Datasets:

pszemraj

/

scientific_lay_summarisation-plos-norm

like 6

Plasmid mediated antimicrobial resistance in the Enterobacteriaceae is a global problem. The rise of CTX-M class extended spectrum beta lactamases (ESBLs) has been well documented in industrialized countries. Vietnam is representative of a typical transitional middle income country where the spectrum of infectious diseases combined with the spread of drug resistance is shifting and bringing new healthcare challenges. We collected hospital admission data from the pediatric population attending the hospital for tropical diseases in Ho Chi Minh City with Shigella infections. Organisms were cultured from all enrolled patients and subjected to antimicrobial susceptibility testing. Those that were ESBL positive were subjected to further investigation. These investigations included PCR amplification for common ESBL genes, plasmid investigation, conjugation, microarray hybridization and DNA sequencing of a blaCTX–M encoding plasmid. We show that two different blaCTX-M genes are circulating in this bacterial population in this location. Sequence of one of the ESBL plasmids shows that rather than the gene being integrated into a preexisting MDR plasmid, the blaCTX-M gene is located on relatively simple conjugative plasmid. The sequenced plasmid (pEG356) carried the blaCTX-M-24 gene on an ISEcp1 element and demonstrated considerable sequence homology with other IncFI plasmids. The rapid dissemination, spread of antimicrobial resistance and changing population of Shigella spp. concurrent with economic growth are pertinent to many other countries undergoing similar development. Third generation cephalosporins are commonly used empiric antibiotics in Ho Chi Minh City. We recommend that these agents should not be considered for therapy of dysentery in this setting. Enterobacteriaceae that have the capability to express CTX-M (so named because of their hydrolytic activity against cefotaxime) family extended spectrum beta lactamases (ESBLs) have emerged as a major health threat worldwide [1], [2]. Most of the research in this area is conducted in industrialized countries, where organisms, such as Escherichia coli and Klebsiella spp. , mostly from urinary tract infections are the commonest source [3], [4], [5]. Relatively little is known about the distribution of such genes in organisms found developing or countries undergoing an economic transition, where the circulating pathogens may differ. Enterobacteriaceae capable of producing ESBLs have been described previously in South East Asia [6], [7]. Ho Chi Minh City in southern Vietnam is typical of many cities where patterns of infectious diseases are changing due to rapid economic growth, better access to health care and improving infrastructure. We recently showed that 42% of healthy people carried ESBL producing bacteria as part of their regular intestinal flora [8]. This previous work suggested that commensal organisms play a role in the dissemination and maintenance of such antimicrobial resistance genes in the population. Furthermore, the uncontrolled use of antimicrobials in the human population and in livestock rearing may lead to further problems with drug resistance and even more limited therapeutic options. Shigellosis is a gastrointestinal infection caused by members by Shigella spp. Due to the faecal oral route of transmission of the Shigellae, children less than five years old and living in developing countries have the highest incidence [9], [10]. In our hospital in Ho Chi Minh City, shigellosis is the leading cause of paediatric diarrhoeal admission with bacterial aetiology. The infection is typically self limiting, although antimicrobial treatment is necessary for the young and those that are severely ill as it ensures fewer complications and curtails the duration of the disease [11]. Fluoroquinolones are the drugs of choice to treat Shigella infections in both adults and children [12]. However, as with many other members of the Enterobacteriaceae, mutations in the genes encoding the target proteins for fluoroquinolones are common in Shigella [13], [14]. Our recent findings show that patients with shigellosis are staying in hospital for longer periods compared with 5 and 10 years ago and the disease severity has concurrently increased [15]. Interestingly, at the same time there has been a significant species shift from S. flexneri to S. sonnei isolated from patients [15]. Patients here are treated with fluoroquinolones, however, those patients that do not respond to the standard therapy are treated with third generation cephalosporins (mainly ceftriaxone). The intravenous third generation cephalosporins are amongst the most commonly used antimicrobials in hospitals in Ho Chi Minh City and the oral second and third generation cephalosporins are also widely available in the community. Antimicrobial resistance in the Shigellae is common; these organisms are closely related to E. coli and are readily transformed by exogenous DNA [16], [17], [18]. The distribution of antimicrobial resistance is, however, often different depending on the species. A multi-centre study across Asia demonstrated that S. flexneri were more likely to be resistant to ampicillin, whilst S. sonnei were more likely to be resistant to co-trimoxazole [19]. Resistance patterns and species dominance are variable depending on the specific location [20], [21], [22]. We have previously reported the rapid emergence of third generation cephalosporin resistant Shigella in Vietnam, where we noted the routine isolation of a number of ESBL producing microorganisms [15]. Here, we present data suggesting that ESBL negative organisms have been replaced with ESBL positive organisms. This study was conducted according to the principles expressed in the Declaration of Helsinki. This study was approved by the scientific and ethical committee of the HTD and Oxford tropical research ethics committee (OXTREC) number 010-06 (2006). All parents of the subject children were required to provide written informed consent for the collection of samples and subsequent analysis. The work was conducted on the paediatric gastrointestinal infections ward at the hospital for tropical diseases (HTD) in Ho Chi Minh City in Vietnam. The HTD is a 500 bed tertiary referral hospital treating patients from the surrounding provinces and from the districts within Ho Chi Minh City. All patients from which Shigella spp. were isolated were enrolled into a randomized controlled trial comparing treatment with ciprofloxacin and gatifloxicin as described previously [15] (trial number ISRCTN55945881). Briefly, all children (aged 0–14 years) with dysentery (defined as passing bloody diarrhoea or mucoid stools with additional abdominal pain or tenesmus) whose parent or guardian gave fully informed written consent were eligible for admission to the study. The primary outcome of the trial was treatment failure, defined as the patient not clearing symptoms after five days of antimicrobial treatment. Stool samples were collected from patients and cultured directly on the day of sampling. Samples were cultured overnight in selenite F broth (Oxoid, Basingstoke, UK) and plated onto MacConkey and XLD agar (Oxoid) at 37°C. Colonies suggestive of Shigella were sub-cultured on to nutrient agar and were identified using a ‘short set’ of sugar fermentation reactions (Kliger iron agar, urea agar, citrate agar, SIM motility-indole media (Oxoid, United Kingdom) ). Serologic identification was performed by slide agglutination with polyvalent somatic (O) antigen grouping sera, followed by testing with available monovalent antisera for specific serotype identification as per the manufacturer' s recommendations (Denka Seiken, Japan). Antimicrobial susceptibility testing of all Shigella isolates against ampicillin (AMP), chloramphenicol (CHL), trimethoprim – sulfamethoxazole (SXT), tetracycline (TET), nalidixic acid (NAL), ofloxacin (OFX;) and ceftriaxone (CRO) was performed by disk diffusion (Oxoid, United Kingdom). The minimum inhibitory concentrations (MICs) were additionally calculated for all isolates by E-test, according to manufacturer' s recommendations (AB Biodisk, Sweden). Those strains that were identified as resistant to ceftriaxone using the disk diffusion susceptibility test were further subjected to the combination disc method to confirm ESBL production [23], [24]. The combination disc method utilizes discs containing only cefotaxime (CTX) (30 µg) and ceftazidime (CAZ) (30 µg) and both antimicrobials combined with clavulanic acid (CLA) (10µg). ESBL producing strains were identified as those with a greater than 5 mm increase in zone with the single antimicrobial compared to the combined antimicrobial, i. e. demonstrating ESBL inhibition [25]. All antimicrobial testing was performed on Mueller-Hinton agar, data was interpreted according to the Clinical and Laboratory Standards Institute guidelines [26]. Genomic DNA was isolated from strains that were subjected to PCR and DNA microarray hybridisation from 1 ml of a 5 ml overnight bacterial culture using the wizard genomic DNA extraction kit (Promega, USA), as per the manufacturer' s recommendations. For characterization of gene content of isolated Shigella strains, genomic DNA was hybridized to an active surveillance of pathogens (ASP) oligonucleotide microarray [27], [28]. The ASP array contains over 6,000 gene markers, including species signature genes, virulence genes and antimicrobial resistance genes from over a hundred bacterial species. Thus the ASP array provides data for assessing horizontally transferred genes, such data is helpful for diagnosis and for guiding antimicrobial therapy. The ASP array used in this study was version 6. 2 and was designed and constructed as described previously [28]. Test samples were labelled and hybridised as described previously [29]. Briefly, 5 µg genomic DNA was labelled with Cy5 and hybridised with a formamide based hybridisation buffer solution in a final volume of 48 µl at 50°C for 16–20 hours. The ASP arrays were washed as described previously but with the initial wash at 50°C [29]. The ASP arrays were scanned using a 418 microarray Scanner (Affymetrix, USA) and intensity fluorescence data acquired using ImaGene 7. 5 (BioDiscovery, USA). Data was analysed as described previously by Stabler et al. [28]. Briefly, a reporter was considered positive if the background corrected mean reporter signal from duplicate spots was both greater than one standard deviation of reporter signal (reporter variation) and the mean reporter signal was greater than the whole background corrected microarray mean plus one standard deviation, as shown for S. sonnei EG1007 in Dataset S1 in supporting information. The raw microarray data for all isolates is presented in Dataset S2 in supporting information. Plasmid DNA was isolated from ESBL positive and ESBL negative Shigella isolates using a modified version of the methodology previously described by Kado and Liu [30]. The resulting plasmid DNA was separated by electrophoresis in 0. 7% agarose gels made with 1× E buffer. Gels were run at 90 V for 3 h, stained with ethidium bromide and photographed. For DNA sequencing plasmid DNA containing an ESBL gene was extracted from an E. coli transconjugant using a NucleoBond® Xtra Midi kit as per the manufacturers recommendations (Clontech, USA) Genomic DNA was subjected to PCR amplification targeting known classes of bla genes using, initially, primers that would recognise sequences encoding SHV, (F; 5′ TCTCCCTGTTAGCCACCCTG, R; 5′; CCACTGCAGCAGCTGC) TEM (F; 5′ TGCGGTATTATCCCGTGTTG, R; 5′ TCGTCGTTTGGTATGGCTTC) and CTX-M (F; 5′ CGATGTGCAGTACCAGTAA, R; 5′ TTAGTGACCAGAATCAGCGG) class ESBLs [31], [32]. Further characterisation of the various sub-group of blaCTX ESBL genes was performed using primers, CTX-M-1; (F 5′ ATGGTTAAAAAATCACTGCG, R 5′ TTACAAACCGTCGGTGAC), CTX-M-2; (F 5′ TGGAAGCCCTGGAGAAAAGT and R 5′ CTTATCGCTCTCGCTCTGT) and CTX-M-9; (F 5′ATGGTGACAAAGAGAGTGCAAC, R 5′ TTACAGCCCTTCGGCGATG) using previously outlined PCR amplification conditions [31], [32]. To identify an association with CTX-M genes and the adjacent ISEcp1 transposase, all ESBL positive strains were subjected to PCR with primers forward primers Tnp24F 5′ CACTCGTCTGCGCATAAAGCGG, Tnp15F 5′ CCGCCGTTTGCGCATA CAGCGG (for blaCTX-M-24 and blaCTX-M-15 respectively) and reverse primer TnpR 5′ AGATATGTAATCATGAAGTTGTCGG. The Tnp24F and Tnp15F were located within the blaCTX-M-24 and blaCTX-M-15 genes respectively and TnpR was located within the ISEcp1 transposase gene. The bla-transposase PCR was performed under the following conditions; 95°C for 1 minute, 30 cycles of 95°C for 30 seconds, 56°C for 30 seconds, 72°C for 1 minute 30 seconds and 72°C for 2 minutes. All PCRs were performed using Taq DNA polymerase and appropriate recommended concentrations of reagents (Bioline, UK). Positive PCR amplicons were cloned into cloning vector pCR 2. 1 (Invitrogen, USA) and sequencing reactions were carried out as recommended by the manufacturer using big dye terminators in forward and reverse orientation on an ABI 3700 sequencing machine (ABI, USA). All sequencing reactions were performed twice to ensure correct sequencing and sequences were verified, aligned and manipulated using Bioedit software (http: //www. mbio. ncsu. edu/BioEdit/bioedit. html). All ESBL gene sequences were compared to other ESBL sequences by BLASTn at NCBI. The DNA sequence of various classes of blaCTX were downloaded and aligned with the produced sequences. Bacterial conjugation experiments were performed by combining equal volumes (3 ml) of overnight Luria-Bertani cultures of donor and recipient strains. The donor strains were Shigella clinical isolates carrying blaCTX genes and the recipient was E. coli J53 (sodium azide resistant). Bacteria were conjugated for 12 hours at 37°C and transconjugants were selected on Luria-Bertani media containing sodium azide (100 µg/ml) and ceftriaxone (6 µg/ml). Potential transconjugants were verified by serotyping and plasmid extraction. Plasmid pEG356 was selected for DNA sequencing and annotation as previously described [33]. The DNA sequence was annotated to identify coding sequences and repeat sequences in Artemis. To identify plasmids with similar sequences, pEG356 was compared by BLASTn at NCBI. pAPEC-01-ColBM (Ac. DQ381420) [34] was downloaded and aligned with pEG356 and viewed in Artemis Comparison Tool (ACT) [35]. Schematic drawing of the sequence of pEG356 was constructed using DNAplotter [36]. Artemis, ACT and DNAplotter are freely available at (http: //www. sanger. ac. uk/Software). The full sequence and annotation of pEG356 was submitted to EMBL with the accession number FN594520. During a 24 month period between April 2007 and March 2009 we isolated 94 Shigella strains from the stools of children admitted with dysentery. Of these 94 strains, 24 were S. flexneri and 70 were S. sonnei, confirming the species substitution previously noted from isolates in this region [15]. The general antibiotic sensitivity patterns in these strains were variable, although resistance to trimethoprim – sulfamethoxazole, tetracycline and latterly nalidixic acid were ubiquitous and there was an overall propensity of sensitivity towards older generation antimicrobials such as chloramphenicol (Table 1). A reversion of sensitivity to older therapies highlights how antimicrobial resistance genes can be maintained (or otherwise) by selective antimicrobial pressure in the population. The first isolation of a ceftriaxone resistant organism during the transitional period occurred in May 2007 and similar strains were isolated in low numbers for the following months (Figure 1). The numbers of Shigellae isolated that were resistant to ceftriaxone fluctuated over the following 18 months. However, there was increase in the proportion of resistant to sensitive isolates 19% to 41% (5 to 11) between the periods from April 2007–September 2007 and April 2008–September 2008, respectively. This trend peaked in March 2009, with six out of seven Shigella strains isolated resistant to ceftriaxone (MIC>256). The overall rate of resistance to ceftriaxone between September 2008 and March 2009 was 75%. We initially cultured a ceftriaxone resistant S. sonnei strain in 2001 (DE 0611) (Table 1), however, this strain was a single, isolated organism and a secondary ceftriaxone resistant Shigella was not isolated again until 2007. Between 2007 and 2009,35 (34%) Shigella isolates cultured were resistant to ceftriaxone (Table 1). Of these strains, 33 were S. sonnei and the other two isolates were S. flexneri. In total, we isolated 36 ceftriaxone resistant organisms between 2001 and 2009. The mechanism of ceftriaxone resistance was examined by the double disc inhibition method to identify ESBL producing organisms. All the S. sonnei and one S. flexneri strain (35 from 36 ceftriaxone resistant Shigella) produced the characteristic ESBL pattern on investigation, whereas the hydrolysing activity of the other S. flexneri organism was not inhibited by clavulanic acid [23], [24] (Table 1). The median age of patients harbouring third generation cephalosporin resistant Shigellae was 32 months (range; 8 to 120 months), the median age of shigellosis patients during the same period was 30 months [15]. Owing to the rapid increase in the rate isolation of such organisms we hypothesised that an individual dominant strain had began circulating in one area of Ho Chi Minh City. However, residence data procured on the time of admission showed that such strains were circulating over a wide area of the city and not purely limited to an isolated outbreak (Table 1). 12 patients were resident in surrounding provinces, some 150 km from the hospital. In conjunction with ceftriaxone, all strains were examined for resistance to an additional five antimicrobials by disc diffusion and MIC (Table 1). As predicted, all strains demonstrated co-resistance to ampicillin. Thirty five of the 36 strains (97%) were resistant to trimethoprim – sulfamethoxazole and tetracycline, whilst 33/36 were resistant to nalidixic acid. Only three isolates; DE0611, EG0419 and EG0471 were co-resistant to chloramphenicol, of which two, EG0419 and EG0471 (6%), were resistant to five of the six antimicrobials tested (Table 1). The most common mechanism of dissemination of ESBL genes in the Enterobacteriaceae is plasmid mediated transfer. Our previous studies have suggested that Vietnam (and other parts of South East Asia) may be hotspot for the origin and further transmission of antimicrobial resistant organisms [8], [13], [37], [38]. Enterobacteriaceae which carry MDR plasmids are common in Vietnam and the isolation of MDR Shigella strains has been repeatedly reported [19], [20], [39]. We hypothesised that the ESBL phenotype was related to the insertion of a transposon carried on an MDR plasmid that had permeated into and was circulating within the Shigella population. To investigate the genetic nature of the ESBL positive isolates compared to the ESBL negative isolates we hybridised genomic DNA to an active surveillance of pathogens (ASP) DNA microarray. In total, 15 isolates (seven ESBL positive and eight ESBL negative) were compared. The ASP array is designed to monitor gene flux, genetic content and the nature of horizontally transferred DNA in a bacterial population. The resulting hybridisation is shown in Figure 2. Concurrently, plasmid DNA was isolated and compared from the same bacterial isolates to assess plasmid content. Figure 2 is a heatmap representation of the 142 ASP microarray reporters which demonstrated positive hybridisation to DNA in two or more of the S. sonnei samples and the 11 reporters representing the S. sonnei Ss046 plasmid pSS_046. The overall hybridisation data and the names and predicted functions of the genes are presented in Dataset S2 (supporting information). The pattern of relative hybridisation across all strains was remarkably homogenous, with only 30% (42/142+11 pSS_046) of the total proportion of the positive coding sequences demonstrating variable hybridisation patterns. The coding sequences demonstrating common hybridisation patterns across all 15 strains included a number of signature E. coli, Shigella spp. regions and sequences corresponding to virulence and antimicrobial resistance (Figure 2 and Supporting information Datasets S1 and S2). The common antimicrobial resistance genes identified between isolates included genes conferring resistance to streptomycin, macrolides, tetracycline, beta lactams and also some unspecific antimicrobial resistance efflux genes. The homogenous nature of hybridisation suggests that variation between isolates is limited and dependent on plasmid content. All the ESBL producing strains demonstrated significant hybridisation to sequences corresponding to bla genes, highlighted in Figure 2, DNA from the ESBL negative strains failed to hybridise to these targets. Plasmid visualisation of plasmid DNA by agarose gel electrophoresis with all hybridised strains revealed that in contrast to the ESBL negative isolates, all the ESBL producing isolates had a large plasmid, we roughly estimated to be greater than 63 Kbp in size (according to the marker plasmid). Despite the ESBL negative isolates lacking a large plasmid; these strains demonstrated similar resistance profiles, with the obvious exception of ceftriaxone (data not shown). These data suggested that the ESBL genes may be located on simple (none MDR) extrachromosomal elements. This hypothesis was supported by evidence of in vivo horizontal plasmid transfer; two strains cultured two days apart from the same patient were identical in serotype, plasmid content and MIC resistance profile, with the exception of the secondary strain carrying a large plasmid and displaying resistance to ceftriaxone (data not shown). Furthermore, sequencing of a conjugative, ESBL encoding plasmid confirmed our suggestion of a simple extrachromosomal element. PCR was performed to detect the blaTEM, blaSHV and blaCTX-M genes. Further PCR amplifications were performed on DNA from all strains that produced amplicons with the blaCTX-M primers. Primers that were specific for the three major CTX-M clusters, blaCTX-M-9, blaCTX-M-1 and blaCTX-M-2 were selected [40]. Three strains (DE0611, EG0187 and EG0356) produced amplicons with the blaCTX-M-9 primers and the remaining 32 isolates produced amplicons with the blaCTX-M-1 primers (Table 2). All 35 PCR amplicon were sequenced. Sequence analysis of the PCR amplicons demonstrated that there were two differing blaCTX-M genes present in the Shigella population, these were, blaCTX-M-24 (n = 3,8%) and blaCTX-M-15 (n = 32,92%) (Table 2). Both genes (blaCTX-M-24 and blaCTX-M-15) share 74% DNA homology with each other; blaCTX-M-15 and blaCTX-M-24 differ by 12 and 6 nucleotides from the precursor genes within their respective parent groups, (blaCTX-M-1 and blaCTX-M-9). Plasmid sizing, by visualisation of the previous agarose gel electrophoresis demonstrated that the estimated plasmid size corresponded with either the blaCTX-M gene (Table 2); blaCTX-M-15 was consistently located on a plasmid larger than that associated with blaCTX-M-24. These observations were confirmed by Southern blotting hybridisation of plasmid DNA extractions (data not shown). The differing plasmid sizes and ESBL genes correlated precisely with two distinct zone clearance areas when strains were susceptibility tested with ceftazidime. The strains expressing CTX-M-24 demonstrated less activity against ceftazidime when compared to CTX-M-15 (median zone size, CTX-M-24; 28mm, CTX-M-15; 20mm) (Table 2). All blaCTX-M harbouring plasmids with the exception of the plasmid in strain EG1020 were transmissible with high conjugation frequencies, ranging from 4. 84×102 to 4. 88×106 (median 1. 55×102) per donor cell (Table 2). The mobilisation of one of these blaCTX harbouring plasmids was further demonstrated by conjugative transfer of the plasmid originally from S. sonnei EG356 from an E. coli transconjugant back into a fully susceptible, naive S. sonnei strain at a similarly high frequency. The ESBL encoding gene blaCTX-M-24 appears to be generally restricted to Enterobacteriaceae in Asia [41], [42], with only sporadic reports of this gene in other locations [43]. Therefore, we selected the plasmid from isolate EG0356, carrying a blaCTX-M-24, as it is applicable to this location, for further characterisation by DNA sequencing. Plasmid pEG356 was found to be a circular replicon consisting of 70,275 nucleotides, similar in size to another blaCTX-M-24 encoding plasmid from Asia; pKP96. pKP96 was isolated from a Klebsiella pneumoniae strain from China in 2002, yet demonstrates limited DNA homology to pEG356, with exception to the ESBL encoding region [44]. pEG356 was comparatively GC neutral (52. 26%) and belonged to incompatibility group FI (on the basis of the DNA sequence homology to the replication region) (Figure 3). pEG356 was predicted to contain 104 coding sequences, of which 14 were considered to be pseudogenes on the basis of apparent premature stop codons, frameshifts or missing start codons. The density of coding sequencing approached 95% and contained four main structural features, a replication region, the ESBL gene encoding region with predicted homology to an ISEcp1 element, an iron ABC transport system and a DNA transfer region (labelled red, pink, dark blue and light blue, respectively in Figure 3). pEG356 encoded the complete tra gene-set encoding a conjugative pilus with high sequence similarity to the transfer region from the F plasmid sequence from E. coli K12 [45] (Ac. AP001918). This is consistent with the in vitro data demonstrating that this plasmid is transmissible into an E. coli recipient. The IncFI replication region was highly similar to other IncF plasmids, including the recently described CTX-M-15 encoding plasmid pEK499 (Ac. EU935739) isolated from an E. coli O25: H4-ST131 epidemic strain circulating in the United Kingdom [46]. Additionally, pEG356 shared another 30 Kbp (position 15,152 to 44,255 in pEG356) of high sequence similarity with pEK499 [46]. This region contains multiple common hypothetical plasmid genes of unknown function, genes involved in conjugative transfer (traM to traC), plasmid partitioning and a predicted single stranded DNA binding protein (ssb). Unlike pEK499 the mok and hok post segregational killing genes are missing from within the plasmid maintenance region [46]. With respect to pEK499 and other ESBL carrying plasmids, pEG356 does not carry multiple antimicrobial resistance genes, transposons, insertion sequences or any additional virulence associated genes [44], [46], [47] (Chen et al. 2007; Shen et al. 2008; Woodford et al. 2009) (Chen et al. 2007; Shen et al. 2008; Woodford et al. 2009). In overall structure, but not size, pEG356 shared the most DNA sequence similarity with the ColBM plasmid pAPEC-O1 (Ac. DQ381420), isolated from an avian pathogenic E. coli strain [34] (Figure 4). pEG356 shared around 80% of the gene content with pAPEC-O1, including the conjugation (tra), replication (rep) and a putative ATP iron transport system (iro). The iro region consisted of four coding sequences, which include, a putative permease, an iron binding protein and an export associated protein. The blaCTX-M-24 was located on an ISEcp1 like element. The overall sequence of the ISEcp1 variant on pEG356 is 4,725 bp and 3,000 bp shares 99% DNA homology with an ESBL gene encoding element from an E. coli strain that was isolated in China; pOZ174 (AF252622) [48]. The blaCTX-M-24 carrying region is also highly similar (99% DNA homology) to the equivalent region in the previously described plasmid, pKP96, including the IS903D downstream of the blaCTX-M-24 gene (Figure 4) [44]. The ISEcp1 element contains two pairs of inverted repeat (Figure 4): the larger inverted repeat (31 bp) flanks the complete element, inclusive of six coding sequences. The 3′end of the ISEcp1 element contained a ISEcp1 transposase and a small hypothetical coding sequence of unknown function which is spanned by two IS1380 elements. The blaCTX-M-24 isadjacent to two pseudogenes, which were understood to have encoded a conserved hypothetical transposon protein and a maltose-inducible porin precursor, it is not clear what significance, if any, these genes are to the overall functionality of the element or the plasmid. All ESBL producing Shigella were subjected to PCR to demonstrate if all bla genes were associated with the ISEcp1 transposase. The location of the PCR primers Tnp24F and TnpR are highlighted in Figure 4 and were designed to produce an amplicon if the bla gene and the adjacent ISEcp1 transposase were in the same location and orientation in strains with a blaCTX-M-24. A secondary forward primer was designed in equivalent location for those strains with a blaCTX-M-15 (Tnp15F). Therefore, if blaCTX-M-24 or the blaCTX-M-15 was consistently adjacent to the ISEcp1 transposase it would produce an amplicon of 414 bp in all strains. All ESBL positive strains (CTX-M-15 and CTX-M-24) generated a PCR amplicon of the predicted size (Table 2). Sequencing of all PCR products demonstrated that all the blaCTX-M-15 and the blaCTX-M-24 gene were associated with an ISEcp1 transposase, The DNA sequence from all PCR products was identical from within the transposase gene up to and including the IS1380. Members of the Enterobacteriaceae that carry CTX-M family ESBLs have been isolated from many parts of the world since the mid 1990s [40]. CTX-M genes have been previously identified from pathogenic Enterobacteriaceae circulating in South East Asia; such as Vietnam, Thailand, Cambodia and Singapore [6], [7], [49], [50]. Additionally, our work has shown that ESBLs are commonly found in organisms which constitute the “normal” gastrointestinal flora in the general population living in Ho Chi Minh City [8]. Such data predicts that intestinal flora may be a considerable reservoir of ESBL encoding genes and the genetic elements they circulate on, permitting potential transmission to their pathogenic counterparts. CTX-M genes in the Shigellae have been previously reported in Argentina, (CTX-M-2) [51], Korea (CTX-M-14) [52] and from a traveler returning from India (CTX-M-15) [53]. More recently, Nagano et al. described a novel CTX-M-64 hybrid from a shigellosis patient infected with S. sonnei after returning to Japan from China [54]. The S. sonnei strains isolated here in Ho Chi Minh City harbored the blaCTX-M-15 and blaCTX-M-24 genes. Current data suggests that blaCTX-M-24 is found mainly in Asia [41], [42], yet may have been transferred to other locations [43]. MDR CTX-M-15 producing E. coli is emerging worldwide as an important pathogen causing hospital-acquired infections [2]. The potential impact of MDR Shigella combined with CTX-M-15/24 carrying plasmids is substantial, with implications for local treatment policy and the transportation of such plasmids into other countries as has been implicated in Canada [43], [55]. The structure of pEG356 as a vector for transferring blaCTX-M-24 implies that such plasmids may be common. The streamlined nature of pEG356, remarkably high conjugation frequency may ensure onward circulation of the genetic cargo as it becomes stable in the bacterial population. The simplistic nature of pEG356, with a lack of additional resistance genes suggests that this is a contemporary element, with the blaCTX-M-24 a recent acquisition. The blaCTX-M-24 gene has been located on a relatively uncomplicated plasmid in Asia, however, pKP96 only demonstrates limited homology to pEG356 [44]. All ESBL gene were located adjacent to a ISEcp1 transposase (as identified by PCR). We are currently unable to substantiate if it is the ISEcp1-like element, the plasmids or the circulation of bacterial clone is responsible for the increasing rate of isolation. However, the geographical spread of these strains suggests that they are widely disseminated throughout southern Vietnam. S. sonnei is a monophyletic bacterial pathogen, and owing to the lack of sensitivity of existing sequence based methods such as multi locus sequence typing [56], we are currently unable to confirm clonality satisfactorily (data not shown). Further epidemiological investigation of CTX-M containing strains combined with a more sensitive sequenced based methodology, such as is used for Salmonella Typhi is required [57]. We are currently assessing the genetic nature of the strain and the plasmids carrying the ESBL genes. Our findings show a transfer from 0% to 75% ceftriaxone resistance in S. sonnei over just two years in the key age group (1 to 3 years) for this disease. By sampling across the Ho Chi Minh City area, covering approximately 150 sq kilometres of Vietnam and a population of approximately 15 million people we have shown that the genetic explanation for this resistance pattern is the dissemination two distinct ESBL genes, of which one is dominant. These are the leading source of ESBLs in clinical Shigella cases and their rapid spread suggests that these organisms are under strong selection pressure. The use of third generation cephalosporins, such as oral cefpodoxime and cefixime in the community is common in Vietnam, and places the even the short term usage of ceftriaxone and other broad-spectrum cephalosporins in jeopardy. Shigella spp. are capable of carrying multiple plasmids with an array of phenotypes including virulence and antimicrobial resistance [16], [18]. The presence of Shigella in the gastrointestinal tract of humans is an ideal environment to acquire horizontally transferred genetic material. Small highly transmissible plasmids that impinge on the fitness of the host may be rapidly disseminated under appropriate conditions. Vietnam is a country that in many respects is representative of many parts of the world. The Vietnamese economy is developing rapidly and the country is undergoing transition with an increasing population, urbanisation and shifting patterns of infectious diseases. In the past decade there has been a transition in species from S. flexneri to S. sonnei in the Southern provinces of Vietnam. With this shift has come the emergence of ESBL S. sonnei. These findings from the Vietnamese population should perhaps serve as a warning for other countries encountering the same economic transition. The progressive evolution of pan-resistant Shigella makes vaccine development an increasingly important objective.

Shigellosis is a disease caused by bacteria belonging to Shigella spp. and is a leading cause of bacterial gastrointestinal infections in infants in unindustrialized countries. The Shigellae are dynamic and capable of rapid change when placed under selective pressure in a human population. Extended spectrum beta lactamases (ESBLs) are enzymes capable of degrading cephalosporins (a group of antimicrobial agents) and the genes that encode them are common in pathogenic E. coli and other related organisms in industrialized countries. In southern Vietnam, we have isolated multiple cephalosporin-resistant Shigella that express ESBLs. Furthermore, over two years these strains have replaced strains isolated from patients with shigellosis that cannot express ESBLs. Our work describes the genes responsible for this characteristic and we investigate one of the elements carrying one of these genes. These finding have implications for treatment of shigellosis and support the growing necessity for vaccine development. Our findings also may be pertinent for other countries undergoing a similar economic transition to Vietnam' s and the corresponding effect on bacterial populations.

Abstract Introduction Materials and Methods Results Discussion

gastroenterology and hepatology/gastrointestinal infections microbiology/medical microbiology infectious diseases/antimicrobials and drug resistance

2010

The Sudden Dominance of blaCTX–M Harbouring Plasmids in Shigella spp. Circulating in Southern Vietnam

Burkholderia pseudomallei is the causative agent of melioidosis, a severe infection prominent in northern Australia and Southeast Asia. The “gold standard” for melioidosis diagnosis is bacterial isolation, which takes several days to complete. The resulting delay in diagnosis leads to delayed treatments, which could result in death. In an attempt to develop better methods for early diagnosis of melioidosis, B. pseudomallei capsular polysaccharide (CPS) was identified as an important diagnostic biomarker. A rapid lateral flow immunoassay utilizing CPS-specific monoclonal antibody was developed and tested in endemic regions worldwide. However, the in vivo fate and clearance of CPS has never been thoroughly investigated. Here, we injected mice with purified CPS intravenously and determined CPS concentrations in serum, urine, and major organs at various intervals. The results indicate that CPS is predominantly eliminated through urine and no CPS accumulation occurs in the major organs. Immunoblot analysis demonstrated that intact CPS was excreted through urine. To understand how a large molecule like CPS was eliminated without degradation, a 3-dimenational structure of CPS was modeled. The predicted CPS structure has a rod-like shape with a small diameter that could allow it to flow through the glomerulus of the kidney. CPS clearance was determined using exponential decay models and the corrected Akaike Information Criterion. The results show that CPS has a relatively short serum half-life of 2. 9 to 4. 4 hours. Therefore, the presence of CPS in the serum and/or urine suggests active melioidosis infection and provides a marker to monitor treatment of melioidosis. Burkholderia pseudomallei is a Gram-negative, soil-dwelling bacillus and the etiologic pathogen of melioidosis, a severe infection endemic in tropical areas with the highest incidence in Southeast Asia and northern Australia [1]. In early 2016, it was predicted that approximately 165,000 individuals worldwide would suffer from melioidosis, while 89,000 of them would die from the infection [2]. B. pseudomallei has also been acknowledged as a potential agent of biological warfare and terrorism because of its ability to cause severe disease via airborne transmission [3,4]. Due to the possibly significant impact on public health and the inherent potential for misuse, the Centers for Disease Control and Prevention (CDC) has classified this organism as a Tier 1 select agent [5]. Currently, there is no licensed vaccine available to prevent the infection. In addition, since B. pseudomallei is resistant to common antibiotics, the success of melioidosis treatment greatly relies on rapid point-of-care diagnosis [6]. At present, bacterial isolation using Ashdown’s selective medium remains the diagnostic gold standard for melioidosis. This technique is only 60% sensitive along with being time consuming, causing treatment delays and increased mortality risk [7]. Rapid diagnostic methods such as latex agglutination, immunofluorescence assay (IFA), ELISA, and PCR have been developed for B. pseudomallei detection [8]. In addition to these techniques, a lateral flow immunoassay (LFI) targeting the capsular polysaccharide (CPS) of B. pseudomallei developed by our group has been shown to be one of the most promising methods for rapid point-of-care detection of melioidosis, especially in resource poor settings [8–10]. The LFI uses a murine monoclonal antibody (mAb) specific to CPS to detect the presence of the bacterium (by detecting CPS) in patient samples. Capsular antigens are outer membrane components expressed by many Gram-negative bacteria, and CPS is known to be one of the most important virulence factors for B. pseudomallei. Structurally, B. pseudomallei CPS is an unbranched homopolymer of 1,3-linked 2-O-acetyl-6-deoxy-β-D-manno-heptopyranose with an approximate molecular weight of 300 kDa [11,12]. Previous animal model studies have found that a CPS-specific antibody provides protection against lethal challenge with B. pseudomallei, suggesting that CPS is a candidate target for melioidosis vaccine development [12–14]. In addition, a recent study from our laboratory revealed that CPS antigen circulates in the bloodstream during infection; this led us to develop the CPS-targeting LFI [15]. Currently, clinical performance of the LFI is being assessed in several endemic areas. However, relatively little is known about the ultimate fate of CPS in vivo. The main focus of this study was to investigate the in vivo distribution and clearance of CPS, information that is essential for improving the clinical use of the LFI. Culture media was inoculated with B. pseudomallei RR2683 (O-polysaccharide mutant; select agent-exempt strain, originating in the Brett laboratory) and incubated overnight at 37°C with vigorous shaking [12]. Cell pellets were obtained by centrifugation and extracted using a modified hot aqueous-phenol procedure [11]. Purified CPS was obtained as previously described [12]. Female, 8-week old CD1 mice (Charles River Laboratories, Inc. , Frederick, MA) were injected with 200 μL of dPBS (Mediatech, Inc. , Manassas, VA) containing 4,20 or 100 μg of purified CPS via the tail vein. The CPS doses were chosen according to previous research investigating the clearance of capsule components of Bacillus anthracis [16]. At 30 min, 2 hours, 4 hours, 8 hours, 12 hours, 1 day, 2 days, 4 days and 8 days post-injection, mice were euthanized using CO2 for sample collection. Urine samples were collected just prior to death. Immediately after euthanasia, blood samples were collected via cardiac puncture and sera were separated. Internal organs including lungs, liver, spleen and kidneys were harvested, weighed and homogenized in 2 mL of dPBS using a PRO250 homogenizer (Pro Scientific, Oxford, CT). Homogenates then were centrifuged and supernatants were collected. All samples were stored at -80°C until quantitative ELISAs were performed. The use of laboratory animals in this study was approved by the University of Nevada, Reno Institutional Animal Care and Use Committee (protocol number 00024). All work with animals at the University of Nevada, Reno is performed in conjunction with the Office of Lab Animal Medicine, which adheres to the National Institutes of Health Office of Laboratory Animal Welfare (OLAW) policies and laws (assurance number A3500-01). An antigen-capture (sandwich) ELISA for CPS quantification was developed using CPS-specific mAb 4C4. Isolation of mAb 4C4 was described previously [17]. Microtiter plates were coated overnight with 100 μL of mAb 4C4 (2. 5 μg/mL in PBS) at room temperature. The plates were washed with PBS-Tween (PBS containing 0. 05% Tween 20) and blocked with a blocking solution (PBS containing 5% skim milk and 0. 5% Tween 20) at 37°C for 1 hour. After blocking, the plates were washed with blocking solution, and then incubated at room temperature for 90 min with 100 μL of a twofold serial dilution of samples (sera, urine or supernatants from tissue homogenates) diluted in blocking solution. A standard CPS sample was prepared by spiking purified CPS into untreated samples diluted in blocking solution. The final concentration of the standard CPS samples was 30 ng/mL. The CPS standard then was added to the plates and serially diluted along with samples to generate the standard curve. After incubation, the plates were washed again with blocking solution, incubated with a mAb 4C4-horseradish peroxidase (HRP) conjugate (0. 5 μg/mL in blocking solution) at room temperature for 1 hour, followed by washing with PBS-Tween. The plates were developed by adding 100 μL of tetramethylbenzidine (TMB) substrate (KPL, Gaithersburg, MD) into each well. The reaction was stopped with 1 M H3PO4, and then the optical density was read at 450 nm (OD450). CPS concentrations in samples were determined by comparison with the standard curve using an OD450 of 0. 5 as the endpoint. The limit of detection of the assay is approximately 0. 25 ng/mL. The amounts of CPS in organ homogenates are reported as micrograms CPS per organ. The CPS amounts reported were corrected by subtraction of the amount of CPS found in the plasma volume in each organ [18], and resulting negative values after subtraction were adjusted to zero. The organ analysis results are presented in comparison with CPS amounts present in serum, which were calculated based on the following information: 1) the blood volume of a mouse is estimated at 5. 77 mL/100g, 2) half of the blood volume is plasma [19], and 3) the average weight of mice used in the study is 25. 9 g. The kinetics of CPS clearance from serum was analyzed using four different exponential decay models: 1) a two-parameter monophasic exponential decay model, y=ae−bx, where a is the Y intercept (concentration of CPS present at time zero) and b is the rate constant of clearance, 2) a three-parameter monophasic exponential decay model, y=ae−bx+y0, where a is the Y intercept (concentration of CPS present at time zero), b is the rate constant of clearance, and y0 is plateau (concentration of CPS persistent in serum), 3) a four-parameter biphasic exponential decay model, y=ae−bx+ce−dx, where a is the proportion of CPS that clears rapidly during the initial clearance step, b is the rate constant of the initial clearance, c is the proportion of CPS that clears more slowly, and d is the rate constant of slower clearance step, and 4) a five-parameter biphasic exponential decay model, y=ae−bx+ce−dx+y0, where a is the proportion of CPS that clears rapidly during the initial clearance step, b is the rate constant of the initial clearance, c is the proportion of CPS that clears more slowly, d is the rate constant of slower clearance step, and y0 is plateau (concentration of CPS persistent in serum). The model fitting was carried out using SigmaPlot 13. 0 (Systat Software Inc. , San Jose, CA). Corrected Akaike' s Information Criterion (AICc), a standard for model selection, was used to evaluate how well each model represents the data. The model that best describes the data (lowest AICc value) among the four equations was selected and used to determine the half-life of CPS in serum. To analyze excreted CPS, urine samples from mice injected with 100 μg CPS were analyzed by immunoblot analysis. Only samples collected at 30 min, 2 hours, and 8 hours contained enough CPS for the analysis. The samples were diluted in SDS-PAGE sample buffer, treated with proteinase K (Fisher Scientific, Waltham, MA) at 60°C for 1 hour, and boiled for 10 min prior to electrophoresis on 7. 5% TGX precast gels (Bio-Rad, Hercules, CA). The volume of each sample loaded onto the gel was adjusted so that an equal amount of CPS (approximately 1 μg) was present in each lane. Western blotting was performed with mini-nitrocellulose transfer packs and a Trans-Blot Turbo transfer system (Bio-Rad). The membranes were blocked with 5% skim milk in TBS-Tween (TBS-T: 50 mM Tris, 150 mM NaCl, 0. 1% Tween 20, pH 7. 6) at 4°C overnight, followed by incubation with 1 μg/mL of mAb 4C4 for 90 min at room temperature. After washing with TBS-T, the membranes were incubated with an anti-mouse IgG-HRP conjugate (Southern Biotech, Birmingham, AL) for 60 min at room temperature to facilitate detection. The final development was carried out using Pierce ECL Western Blotting Substrate (Pierce Biotechnology, Rockford, IL) and a ChemiDoc XRS imaging system (Bio-Rad). The chemical structure of the B. pseudomallei CPS antigen was obtained from previously published work [11]. The structure was drawn in ChemDraw Prime version 15. 0 (PerkinElmer, Waltham, MA) and exported to ChemBio3D Ultra software (PerkinElmer). Due to limitations of the software in processing a large molecule like CPS, only a fragment of CPS that consists of 100 units of 2-O-acetyl-6-deoxy-β-D-manno-heptopyranose was built. Three-dimensional (3D) structure of the CPS was predicted using MM2 energy minimization mode implemented in the ChemBio3D software and exported in protein data bank (PDB) format. The PyMOL program (Schrödinger, LLC, www. pymol. org) was used to visualize and analyze the 3D structure of CPS. Active Melioidosis Detect™ (AMD™) LFI (InBios International, Inc. , Seattle, WA) was used to detect excreted CPS in a urine sample. The urine sample was collected from a mouse injected with 4 μg CPS at 30 min post injection. The sandwich ELISA was used to determine the CPS concentration in this sample. The sample then was serially diluted with mouse control urine to yield the desired concentrations of CPS (0. 04–625 ng/mL). Each dilution of urine (5 μL) was mixed with 20 μL of chase buffer (included in AMD™ kit). The mixture then was applied to the sample pad, followed by an additional 100 μL of chase buffer. The tests were allowed to develop for 15 min. The results were assessed by four examiners in a semi-blinded, randomized manner and photographed. Intensities of the test lines were also quantified using an ESE-Quant lateral flow reader (QIAGEN, Helden, Germany). Mice were intravenously injected with various doses of purified CPS. At the designated time points, serum and urine samples were collected, and CPS concentrations were determined using antigen capture ELISA. Preliminary analysis was performed to ensure that CPS detection was not significantly affected by the presence of serum or urine (S1 Fig). The kinetics data for CPS in serum were fitted to the exponential decay models, and AICc then was used for model selection. According to the AICc results, clearance of CPS from serum is best described by the two-parameter monophasic exponential decay model (y = ae-bx) (S1 Table). Data analysis showed that CPS was cleared rapidly from serum with a half-life (95% confidence interval) of 4 hours (2. 5–6. 6 hours), 4. 4 hours (2. 0–9. 7 hours), or 2. 9 hours (2. 3–3. 9 hours), for the doses of 100 μg, 20 μg, or 4 μg, respectively, suggesting that the half-life values derived from all three doses were comparable (Fig 1). Fig 2 shows the concentrations of CPS in urine at different time intervals. The CPS concentrations in urine were found to be highest at 30 min (the first time point of sample collection), and then decreased rapidly, corresponding to the decrease of CPS in serum. Kidneys, livers, lungs and spleens were collected from the same CPS-injected mice that were used for the CPS clearance study, but only the samples from mice injected with 100 μg CPS were used. The organs were homogenized in PBS for analysis of CPS concentrations by ELISA. The preliminary analysis showed that the presence of tissue homogenates had no effect on assay performance (S1 Fig). CPS amounts in each organ were reported in comparison with amounts of CPS found in serum (Fig 3). The results showed no significant amounts of CPS deposited in the organs. Since we could not detect CPS in organs from mice injected with 100 μg CPS, the internal organs from mice receiving 20 μg or 4 μg CPS were not analyzed in this study. Altogether the results demonstrated that the circulating CPS was eliminated rapidly and predominantly through urine, without accumulation in the major organs. To find out how a large molecule like CPS was excreted, the urine samples from CPS-treated mice were analyzed by Western blot (Fig 4). The Western blots detected full-length CPS in the urine samples, while degraded CPS was not detected. When interpreting these results, however, it is important to note that CPS epitopes might be affected by degradation, so the Western blot analysis might not be able to detect degraded products of CPS. Therefore, it is appropriate to deduce that some (but not necessarily all) of the injected CPS was eliminated without degradation. In some instances, two bands of CPS were observed in our blots. The higher molecular weight band seemed to comprise the largest fraction of CPS in urine samples from these mice; however, how and why this occurred needs to be further investigated (Fig 4). Analysis of the total urine protein was used to assess whether or not exogenous CPS affected renal function, which could result in leakage of high molecular weight compounds into urine. The results showed that there was no difference in urine protein profiles between CPS-treated and untreated mice, suggesting that kidney impairment was unlikely the cause of rapid renal excretion of CPS (S1 Fig). In order to explain how CPS was excreted through the kidneys without apparent degradation, a three-dimensional structure of CPS was predicted (Fig 5). However, due to limitations of the software, only a short fragment (~22 kDa) of CPS was constructed (full-length CPS has a molecular weight of ~300 kDa). The computational 3D model demonstrated that CPS has a rod-like shape with a diameter of approximately 1. 2 nm. The length of a single molecule of CPS, which was calculated from the length of the 22 kDa fragment model, was approximately 490 nm. Previous experiments demonstrated that CPS is cleared rapidly and predominantly through urine, suggesting that urine has the potential to be used as a non-invasive sample for diagnosis of melioidosis. AMD™ LFI is an assay designed to detect B. pseudomallei by targeting CPS molecules in various types of biological samples. To assess whether or not the sensitivity of AMD™ LFI was impacted when it was used to detect excreted CPS in urine samples, serial dilutions of urine from a CPS-treated mouse were tested with the LFI. The results showed that AMD™ LFI could detect excreted CPS as low as 0. 2 ng/mL (Fig 6), comparable with the AMD™ LFI sensitivity reported previously [10]. Like many other pathogenic microorganisms such as Bacillus anthracis, Haemophilus influenzae type b, Streptococcus pneumoniae, and Cryptococcus neoformans, B. pseudomallei expresses a capsular antigen that is shed into the bloodstream during infection [15,20–22]. Based on this finding, immunodiagnostic methods (IFA and LFI, among others) targeting B. pseudomallei CPS were developed. These diagnostic tools showed similar specificity but higher sensitivity when compared to the ‘gold standard’ diagnostic bacterial culture, suggesting that these techniques have the potential to be used clinically [7,10,23]. To use these tools as a routine diagnostic method, however, it is important to understand the fate of CPS in vivo. This could provide insight into the retention and processing time for CPS in a patient during infection, and the potential patient sample types that can be targeted for testing. To understand how CPS is processed in vivo, mice were intravenously injected with purified B. pseudomallei CPS. CPS concentrations in samples (blood, urine, lungs, liver, spleen, and kidneys) at various time points post-injection were determined using an antigen capture ELISA. The doses of CPS used in this study were chosen according to previous research investigating the clearance of B. anthracis capsule antigen [16]. We acknowledge that the concentration of CPS in patient serum reported previously was much lower than the lowest dose of CPS we used in this study [10]. However, the range of CPS concentrations reported was from a limited number of serum samples, and was not associated with the stage of infection. In addition, the concentrations reported previously were too low to allow us to collect accurate and sufficient kinetic data in our study. Thus, the experiments were conducted using 4,20, or 100 μg CPS per mouse. According to our results, B. pseudomallei CPS was rapidly cleared from serum with a half-life of 2. 9–4. 4 h (Fig 1), and not deposited in kidneys, lungs, liver, or spleen (Fig 3). However, relatively high concentrations of CPS were detected in urine shortly after the injection (Fig 2). Notably, at 30 min post-injection, we found that the CPS concentrations in urine were higher than those found in serum for mice receiving 20 or 100 μg CPS (Fig 2). These results indicated that the kidney is the major organ responsible for CPS elimination. Comparison of our findings with the fate of capsular antigens from other organisms suggests that B. pseudomallei CPS has a unique set of characteristics. C. neoformans produces glucuronoxylomannan (GXM) as a major capsule component [24], while B. anthracis produces a capsule composed of a poly-γ-D-glutamic acid (PGA) polypeptide [25]. B. pseudomallei capsule is composed of polysaccharides and is somewhat similar in chemical composition to GXM; however, B. pseudomallei CPS and PGA are apparently similar in geometry, as both of them have a rod-shaped structure [26]. Grinsell et al. demonstrated that GXM has a long serum half-life (~1. 6 days) [19]. However, CPS from B. pseudomallei behaved more like the capsular polypeptide PGA, as both of them showed rapid serum clearance [16]. Previous studies have also found that pneumococcal polysaccharide, GXM and PGA accumulated in many mouse tissues [16,19,27]. In addition, the liver and spleen were found to play important roles in clearance of both GXM and PGA. B. pseudomallei CPS, however, was not deposited in any mouse organs (kidneys, lungs, spleen and liver), and it was cleared predominantly by the kidneys. Altogether, the results suggest that B. pseudomallei CPS exhibits certain characteristics distinct from capsular antigens of other previously reported microbes. Sutherland et al. also showed that PGA was found in urine at high concentrations, which we also found to be true with B. pseudomallei CPS. The study revealed that PGA was excreted in a degraded form [16]. Since B. pseudomallei CPS has a high molecular weight, which is much greater than the molecular cutoff for glomerular filtration [28], we anticipated that we would find degraded CPS in urine, as previously seen in the PGA study. However, our results illustrated that a portion (if not all) of circulating B. pseudomallei CPS was apparently excreted without degradation (Fig 4). The CPS molecule, thus, was further investigated using 3D computer modeling that allowed us calculate a structure of the molecule. The result showed that CPS has a rod-like shape with the dimensions (diameter x length) of 1. 2 nm x 490 nm (Fig 5). We found that the structure of CPS resembles a high molecular weight molecule (~350–500 kDa) of a single-walled carbon nanotube (SWCN, dimension = ~1 nm x ~500 nm) [29]. Ruggiero et al. demonstrated that in vivo SWCN was excreted rapidly, and predominantly by glomerular filtration of the kidneys, even though its molecular weight exceeds the known glomerular cutoff [29]. As explained by Ruggiero et al. , SWCN has a diameter smaller than a glomerular pore (~10 nm), and capillary flow orients the major axis of the rod to align with the glomerular orifice, thereby allowing it to flow through the kidneys. Since B. pseudomallei CPS molecules and SWCN are geometrically similar, it is possible that CPS and SWCN are eliminated via the kidneys by the same mechanism. Our results show that B. pseudomallei CPS has a short serum half-life like PGA from B. anthracis. However, while the half-life of PGA was dose-dependent, we found that the B. pseudomallei CPS half-life was dose-independent (Fig 1). This reflected the possibility that CPS and PGA might be eliminated by different mechanisms. We know that PGA was eliminated following degradation; the degradation capacity, however, can be saturated when a treatment dose of PGA exceeds a certain concentration. As a consequence, PGA accumulated faster than it could be cleared; thus the half-life of PGA became longer when the doses were higher (dose-dependence) [16]. For B. pseudomallei CPS, however, half-life was independent from the treatment doses. We interpret these results to indicate that the CPS elimination capacity was not saturated, at least by the highest concentration used in this study. We found this interpretation fit our proposed mechanism of CPS elimination, in describing that B. pseudomallei CPS is eliminated passively through glomerular filtration, rather than by degradation or carrier proteins. Finally, our results have several implications for the clinical use of immunodiagnostics detecting B. pseudomallei CPS. We know from our previous study that during infection CPS is shed into the blood circulation, indicating that serum can be used as a sample for diagnosis of melioidosis [15]. In this study, we have revealed that CPS can also be detected in urine samples at a high concentration. This finding is consistent with other studies where urine samples from human [10] and non-human primates infected with B. pseudomallei (K. C. Brittingham, A. Leon, P. A. Braschayko, M. S. Anderson, K. A. Knostman and R. E. Barnewall, presented at the ASM Biodefense and Emerging Diseases Research Meeting, Arlington, VA, 8 to 10 February 2016) were analyzed; results showed a large amount of CPS was present in the urine. In this study, we also discovered that CPS has a half-life of approximately 2. 9 to 4. 4 h in serum. We noted that our experiments were performed in non-infected animals that have no antibody to CPS. Antibody is known to play an important role in clearing various exogenous antigens, including capsular antigen [30]. Thus, it is possible that, in infected animals or patients with CPS-specific antibody, the serum half-life of CPS could be even shorter than the half-life reported in this study. Since CPS is cleared rapidly from serum by the kidneys, the presence of CPS in serum or urine may suggest an active source of B. pseudomallei antigen, i. e. acute B. pseudomallei infection. It also suggests that CPS could be a potential biomarker for monitoring efficacy of melioidosis treatment. In addition, we also found that the AMD™ LFI can efficiently detect eliminated CPS in a urine sample (Fig 6). These findings together suggest that perhaps urine, a noninvasive sample that contains a high concentration of CPS, as a sample for melioidosis diagnosis is more appropriate than using serum samples. In summary, the in vivo clearance of B. pseudomallei CPS has a unique set of characteristics, including i) rapid serum clearance, ii) no significant accumulation in internal organs, iii) potentially passive excretion by glomerular filtration, and iv) presence at a high concentration in urine. Rapid serum clearance of CPS suggests that CPS is a significant biomarker for identifying active melioidosis and monitoring treatment progress. In addition, urine, a noninvasive sample, also has a potential to be used as a clinical specimen for melioidosis diagnosis.

An outer membrane component, capsular polysaccharide (CPS), is a virulence factor expressed by many Gram-negative bacteria including Burkholderia pseudomallei, the causative agent of melioidosis. Recently, B. pseudomallei CPS was identified as a useful diagnostic biomarker, leading to the development of a lateral flow immunoassay (LFI) targeting CPS for B. pseudomallei detection. In this current work, we studied the in vivo fate of CPS using a murine model, to better understand the clinical applications and potential limitations of the LFI. Interestingly, we found that B. pseudomallei CPS has a unique set of characteristics (as compared to other bacterial capsule antigens) including rapid kidney clearance from serum, no deposition in major internal organs, and ability to be cleared without degradation. Clinically, these findings suggest that CPS may be a potential biomarker for detecting active melioidosis and monitoring melioidosis treatment outcome. Additionally, urine may be used as a non-invasive sample for detecting melioidosis.

Abstract Introduction Materials and Methods Results Discussion

medicine and health sciences body fluids immune physiology enzyme-linked immunoassays spleen melioidosis urine drug delivery system preparation bacterial diseases routes of administration antigen encapsulation pharmaceutics pharmacology kidneys immunologic techniques research and analysis methods intravenous injections infectious diseases immunoassays biochemistry polysaccharides anatomy physiology biology and life sciences renal system pharmaceutical processing technology glycobiology

2016

In vivo Distribution and Clearance of Purified Capsular Polysaccharide from Burkholderia pseudomallei in a Murine Model

Recent studies of cellular networks have revealed modular organizations of genes and proteins. For example, in interactome networks, a module refers to a group of interacting proteins that form molecular complexes and/or biochemical pathways and together mediate a biological process. However, it is still poorly understood how biological information is transmitted between different modules. We have developed information flow analysis, a new computational approach that identifies proteins central to the transmission of biological information throughout the network. In the information flow analysis, we represent an interactome network as an electrical circuit, where interactions are modeled as resistors and proteins as interconnecting junctions. Construing the propagation of biological signals as flow of electrical current, our method calculates an information flow score for every protein. Unlike previous metrics of network centrality such as degree or betweenness that only consider topological features, our approach incorporates confidence scores of protein–protein interactions and automatically considers all possible paths in a network when evaluating the importance of each protein. We apply our method to the interactome networks of Saccharomyces cerevisiae and Caenorhabditis elegans. We find that the likelihood of observing lethality and pleiotropy when a protein is eliminated is positively correlated with the protein' s information flow score. Even among proteins of low degree or low betweenness, high information scores serve as a strong predictor of loss-of-function lethality or pleiotropy. The correlation between information flow scores and phenotypes supports our hypothesis that the proteins of high information flow reside in central positions in interactome networks. We also show that the ranks of information flow scores are more consistent than that of betweenness when a large amount of noisy data is added to an interactome. Finally, we combine gene expression data with interaction data in C. elegans and construct an interactome network for muscle-specific genes. We find that genes that rank high in terms of information flow in the muscle interactome network but not in the entire network tend to play important roles in muscle function. This framework for studying tissue-specific networks by the information flow model can be applied to other tissues and other organisms as well. In the last decade, several high-throughput experimental techniques have allowed systematic mapping of protein-protein interaction networks, or interactome networks, for model organisms [1]–[4] and human [5], [6]. Interactome networks provide us with a global view of complex biological processes within an organism. However, it has been a challenge to associate network properties with functional relevance. Work on global topology of interactome networks has led to a conclusion that these networks are small-world with power-law degree distributions [7]–[10]. This translates to having a few hub nodes and a majority of nodes with a few partners. This property of interactome networks is very different from random networks where the degree is uniformly distributed. Given that interactomes evolved into this topology, analyzing topological properties of biological networks should provide system-level insights on key players of biological processes. In an interactome network, the ‘central’ proteins, which topologically connect many different neighborhoods of the network, are likely to mediate crucial biological functions. The most straightforward way of quantifying the centrality of a protein in the context of interactome networks is to examine the protein' s degree, e. g. the number of binding partners interacting with the protein of interest. Perturbations of high-degree proteins (hubs) are more likely to result in lethality than mutations in other proteins [7], [11]. However, degree only measures a protein' s local connectivity and does not consider the protein' s position relative to other proteins except for the direct binding partners of the given protein. A metric to estimate global centrality is betweenness. Betweenness determines the centrality of a protein in an interactome network based on the total number of shortest paths going through the given protein [12], [13]. A node partaking in a large fraction of all shortest paths has high betweenness. Such nodes have been termed bottlenecks [14] as they are not necessarily high degree (as are the hub nodes), yet they have a large amount of “information traffic. ” The bottlenecks, like the hubs, are more likely to be essential than randomly sampled proteins in interactomes [11], [15]. Recent evidence shows that high betweenness is correlated with pleiotropy [16], and bottlenecks tend to mediate crosstalks between functional modules [14]. Both degree and betweenness are graph metrics that are not specifically tailored to describe biological networks. Degree measures a protein' s local connectivity and does not consider the protein' s position in the network globally. Betweenness is a better measure for centrality in that it takes into account paths through the whole network, but it still has the disadvantage of only considering the shortest paths and ignoring alternative pathways of protein interactions. More importantly, interactome networks can be error-prone and some interactions in the same network are not as reliable as others. Many studies have been conducted to categorize interaction data into different confidence levels [3], [17], [18]. Neither degree nor betweenness takes the confidence levels of interactions into consideration. To provide a better solution for identifying central proteins, we developed an information flow model of interactome networks. We took the approach of modeling networks as electrical circuits, which had been presented in previous network analyses [19]–[21]. Construing the propagation of biological signals as flow of electrical current, our method identified proteins central to the transmission of information throughout the network. Unlike the previous methods which characterized only the topological features of proteins, our approach incorporated the confidence scores of protein-protein interactions and automatically considers all possible paths in a network when evaluating the importance of proteins. We compared the information flow score to betweenness, and found that the information flow score in the entire interactome network is a stronger predictor of loss-of-function lethality and pleiotropy, and better tolerates the addition of large amounts of error-prone data. For a multi-cellular organism, not all interactions have the same propensity to occur in every tissue. However, the current network metrics usually treat interactome networks as a whole, disregarding the possibility that some interactions may not occur at all in certain types of tissues. To address this, we developed a framework for studying tissue-specific networks using the information flow model. We constructed an interactome network for muscle enriched genes in C. elegans, and showed that genes of high information flow in the muscle interactome network but not in the entire interactome network are likely to play important roles in muscle function. We modeled an interactome network as an electrical circuit, where interactions were represented as resistors and proteins as interconnecting nodes (Figure 1). In the circuit, the value of resistance for each resistor is inversely proportional to the confidence score of the interaction. According to Kirchhoff' s circuit laws, the current entering any node is equal to the current leaving that node. By applying a current source to one node and grounding another, we determined the exact amount of current flowing through each node in the network (see Materials and Methods). We iterated over all pairwise combinations of “source” and “ground” nodes in the network and summed up the absolute values of current through the node of interest from all iterations. We defined the information flow score of a protein as the sum of absolute values of current through the corresponding node. A node that actively participates in the transmission of current for other nodes ends up with a high sum of absolute values of current, and the corresponding protein receives a high information flow score. Unlike degree that only considers direct interactions or betweenness that only scores proteins along the shortest paths interpreted as the dominant paths, the information flow model weighs proteins along all the possible paths. Therefore, the information flow model is able to rank “runner-up” proteins participating in many paths of information transmission, instead of only the seemingly prominent ones. This aspect of the information flow model reflects the property of biological pathways more faithfully: there have been plenty of observations for multiple pathways acting in parallel to achieve a specific biological function [22]–[26], and the active pathways may not always be the shortest ones. We applied the information flow model to two publicly available interactome networks: a S. cerevisiae interactome consisting of 1516 proteins involved in 39,099 interactions [3] and a C. elegans interactome consisting of 4607 proteins involved in 7850 interactions [2], [27], [28] (see Materials and Methods). Every interaction in the yeast interactome is accompanied by a socio-affinity index, which quantifies the tendency for a pair of proteins to identify each other when one of the pair is tagged and to co-purify when a third protein is tagged [3]. A high socio-affinity index indicates a high confidence level for an interaction. We used all the interactions with socio-affinity indices of 2 or higher. The worm interactome does not have numerical scores for the interactions, so we regarded all of the interactions for worms equally. Using these two interactomes, we were able to evaluate the information flow model under situations where interactions are treated equally or interactions have different confidence scores. Similarly to degree and betweenness, information flow scores of proteins in the yeast or worm interactome network did not follow a Gaussian distribution (data not shown), so we converted information flow scores into ranks and percentiles to reflect their relative values in an interactome network. Although the information flow score is a very different network metric from betweenness or degree, there might be relationships between the information flow score and these two topological metrics. We obtained scatter plots for the ranks of information flow scores versus the ranks of betweenness or degree for both the yeast interactome and the worm interactome (Figure 2). Although the information flow score and betweenness are correlated, a given betweenness rank usually corresponds to a wide range of information flow ranks, and vice versa (Figure 2A and 2C). The information flow score and degree are less correlated (Figure 2B and 2D). Low degree does not necessarily imply low information flow score, although very high degree often implies high information flow score. We propose that the information flow model is able to identify proteins central to the transmission of biological information in an interactome network. If this model works, eliminating the proteins of high information flow scores should be deleterious. The perturbation of information flow and the disintegration of functional modules are likely to result in lethality or multiple phenotypes (pleiotropy). To test our hypothesis, we performed a correlation analysis between the percentages of essential proteins or pleiotropic proteins and the percentiles of information flow scores (see Materials and Methods). For each bin containing proteins within a certain range of information flow scores (in percentiles), we calculated the percentage of proteins whose loss-of-function strains exhibit lethality and the percentage of proteins whose loss-of-function strains exhibit two or more phenotypes. We observed a strong increasing trend for the percentage of essential proteins and the percentage of pleiotropic proteins when information flow scores increase (Figure 3). For S. cerevisiae, the Pearson correlation coefficient (PCC) between the percentages of essential proteins and the percentiles of information flow scores is 0. 84, and the PCC between the percentages of pleiotropic proteins and the percentiles of information flow scores is 0. 60. For C. elegans, the PCC between the percentages of essential proteins and the percentiles of information flow scores is 0. 95, and the PCC between the percentages of pleiotropic proteins and the percentiles of information flow scores is 0. 85 as well. In contrast, betweenness is a poorer predictor for both essentiality and pleiotropy. For S. cerevisiae, the PCC between the percentages of essential proteins and the percentiles of betweenness is −0. 02, and the PCC between the percentages of pleiotropic proteins and the percentiles of betweenness is −0. 31. For C. elegans, the PCC between the percentages of essential proteins and the percentiles of betweenness is 0. 67, and the PCC between the percentages of pleiotropic proteins and the percentiles of betweenness is 0. 49. To determine the statistical significance of the correlation, we generated randomized datasets by shuffling genes among the percentile ranges while keeping the number of genes in each range fixed. Next we obtained the percentage of essential or pleiotropic genes for each range and performed correlation analysis for each randomized dataset. We found that the correlation between essentiality or pleiotropy and information flow scores is generally stronger in the actual datasets than in the randomized datasets (P-value = 0. 0059 and P-value = 0. 055 for essentiality and pleiotropy in S. cerevisiae, respectively; P-value = 0. 00054 and P-value = 0. 0047 for essentiality and pleiotropy in C. elegans, respectively), while the correlation between essentiality or pleiotropy and betweenness is not significant (P-value>0. 05). Information flow outperforms degree in terms of correlation with essentiality or pleiotropy in S. cerevisiae (Figure S1). In the C. elegans interactome where the interactions are unweighted, degree is still a strong indicator of essentiality and pleiotropy (Figure S1). Proteins with similar betweenness in an interactome can differ significantly in terms of information flow scores (Figure 2). We investigated whether the information flow score is well correlated with essentiality and pleiotropy among proteins that rank low in terms of betweenness. We identified 449 proteins that rank the lowest 30% in the yeast interactome and 672 proteins that rank the lowest 30% in the worm interactome. We found that the correlation between the information flow score and essentiality or pleiotropy holds for these two groups of proteins (Figure 4). For example, we found ten yeast proteins that are among the highest 30% of all proteins in terms of information flow but are among the lowest 30% of all proteins in terms of betweenness. Out of these 10 proteins, 8 correspond to lethal phenotypes when deleted, and the other 2 correspond to multiple other phenotypes when deleted (Table S1). In contrast, we found three yeast proteins that are among the highest 30% of all proteins in terms of betweenness but are among the lowest 30% of all proteins in terms of information flow, and none of them are essential or pleiotropic. Similarly, we found that the information flow model is predictive of essentiality or pleiotropy among medium- or low-degree proteins as well (Figure S2). What properties make some proteins low in betweenness but high in information flow scores? From the information flow model, we can expect two typical situations: one situation is that a protein lies on alternative paths that are slightly longer than the shortest paths; the other situation is that a protein has a limited number of high-confidence interactions. Betweenness does not take any alternative, longer paths into consideration in the first situation, and betweenness does not give “extra credit” to high-confidence interactions in the second situation. We illustrated the above two situations with example “toy” networks, and analyzed how the information flow model scores nodes that may be important but not recovered by betweenness (Text S1). A closer look at the individual proteins from the interactome networks confirms the existence of both situations in biological networks. Every interaction in the yeast interactome has a socio-affinity index that measures the likelihood of a true interaction [3]. A hub that has many low-confidence interactions may not be rated as high as a protein with a limited number of high-confidence interactions by the information flow model. We defined an average interaction score for a protein as the average of socio-affinity indices for all interactions involving the given protein. For example, SRP68, a core component of the signal recognition particle ribonucleioprotein complex, has a high average interaction score which ranks among the highest 30% in the yeast interactome. SRP68 ranks among the lowest 30% in terms of betweenness but the highest 30% in terms of information flow score. The deletion of this gene results in lethality of the yeast strain. The same situation applies to RPB5, an RNA polymerase subunit. The high average interaction scores are not taken into account in the calculation of betweenness. In the information flow model, we give more credit to the proteins with high-confidence interactions. The C. elegans interactome does not have numerical scores associated with the interactions, so all the interactions are treated equally in our information flow model. Therefore, the discrepancy of information flow scores and betweenness is likely to result from topological features of the network. For example, KLC-1, which has been found to interact with UNC-116/kinesin, KCA-1/kinesin cargo adaptor, and the ARX-2/Arp2/3 complex component by yeast two-hybrid (Y2H) screens [2], is involved in intracellular transport and is required for embryonic viability. KLC-1 is on a topologically central position (Figure 5A) but scores low in terms of betweenness. Another example is TAG-246, an ortholog of mammalian SWI/SNF-related matrix-associated actin-dependent regulator of chromatin subfamily D (SMARCD). TAG-246 is required for LIN-3/EGF signaling in C. elegans vulva development. Just like KLC-1, TAG-246 only has 4 interactions. The loss-of-function of TAG-246 results in lethality as well as several post-embryonic phenotypes, such as protruding vulva and sterile progeny. Figure 5B shows that there are many parallel paths around TAG-246, so TAG-246 does not always lie on the shortest path, thus scoring low in betweenness. Although KLC-1 and TAG-246 are neither high-degree nor high-betweenness, the information flow model ranks them in the top 37% and top 26%, respectively, because it considers all possible paths in the network. Taken together, the information flow model is effective in identifying proteins that are central in interactome networks. Even in cases where betweenness ranks are relatively low, the information score serves as a strong predictor for essential or pleiotropic proteins. As more high-throughput datasets become available, new interactions are added into the networks. High-throughput experiments are error-prone and false positives can be problematic [17]. To address the data-quality issue, there have been many studies attempting to estimate the probability of a true interaction between a pair of proteins instead of weighing all interactions equally [18]. However, previous network metrics such as betweenness do not take the likelihood of interactions into account. By incorporating likelihood of interactions into resistor values, the information flow model is able to more accurately simulate information propagation throughout the network. In order to analyze how well the information flow model tolerates the addition of a large amount of noisy data, we simulated a growing yeast interactome network by adding low-confidence interactions. Higher socio-affinity indices indicate higher confidence of interactions. In total, there are 9,290 interactions with socio-affinity indices of 4. 5 or higher, or 17,159 interactions with socio-affinity indices of 3. 5 or higher, or 39,099 interactions with socio-affinity indices of 2 or higher. We rank both information flow scores and betweenness for all the proteins in each of the three versions of the interactome. We showed that ranks of information flow scores were more consistent than that of betweenness when low-confidence interactions were added to the interactome (Figure 6). The consistency of information flow ranks suggests that the information flow model is not only effective but also robust in the case of noise in the data. In multi-cellular organisms such as C. elegans, a pair of proteins may only interact in certain tissues or cell types. Therefore, the architecture of interactome networks may vary according to tissue or cell types [29]. We hypothesize that proteins of high information flow in a given tissue play crucial roles for the normal function of that tissue. We tested our hypothesis in an interactome network for muscle-enriched genes. From a SAGE (Serial Analysis of Gene Expression) dataset of 12 C. elegans tissues [30], we identified muscle-enriched genes using a semi-supervised learning method [31]. The semi-supervised learning analysis combines the benefits of unsupervised clustering and supervised classification. In other words, both the distribution of data points and prior biological knowledge can be utilized to identify genes enriched in a particular tissue. We manually curated the biomedical literature and found 25 genes known to show enriched expression in muscle cells and 165 genes known not to be expressed in muscle cells (Table S2). These two groups of genes served as positive and negative training data, respectively. For each gene expressed in muscle, the semi-supervised learning procedure gave a probability score (Pi (muscle) ) ranging from 0 to 1 to indicate the gene' s expression enrichment in muscle as compared to other tissues (Table S3). We defined genes scoring 0. 5 or higher (Pi≥0. 5) as muscle-enriched genes and identified 310 such genes (Figure 7). Among the muscle-enriched genes identified by us, promoter: : GFP reporter strains are available for 52 of them, and 31 of them (60%) show clear expression patterns in body wall muscle (Table S4), not including those that might be expressed in other types of muscle. In addition, 260 (84%) of muscle-enriched genes contain cis-regulatory modules that indicate expression in muscle in their promoter sequences [32] (Table S5). From the interactome dataset, we identified direct interacting partners of the muscle-enriched genes. We discarded the interacting genes that, according to the SAGE data, are not expressed in muscle cells. The muscle-enriched genes and their interacting partners which are expressed in muscle form a network of 332 genes and 638 interactions. We defined the weight of an interaction (g12) in the muscle interactome network as the product of the probability scores for the two interacting genes (g12 = P1P2). In other words, the more enriched a given gene' s expression is in muscle, the higher its propensity is to interact with other enriched genes in muscle cells. We applied the information flow model to the muscle interactome network, taking the weights of interactions into account. We ranked all the genes in the muscle interactome network by their information flow scores in the muscle interactome network and by their information flow scores in the entire interactome network, respectively. We found that genes of high information flow in the muscle interactome network and genes of high information flow in the entire network did not completely overlap (Figure 8). In other words, some genes rank high in both the muscle network and the entire network, while others rank high in the muscle network but not in the entire network. We first examined genes ranking high in both networks. We identified the top 35 genes based on the sum of their ranks from both networks and found that 40% of them correspond to loss-of-function lethality, which implies that they are essential for the organism development. We then hypothesized that the genes ranking high in the muscle network but not in the entire network play crucial roles in muscle function, though they may not be essential for the whole organism. We obtained the percentiles of genes in terms of information flow scores in the muscle network and the percentiles of genes in the entire network, calculated the differences between these two percentiles, and ranked the genes by the differences. A C. elegans homolog of human paxillin, tag-327, shows the largest percentile difference (Table 1). This gene is suspected to be part of the worm muscle attachment complex [33]. A homozygous gene knockout of tag-327 resulted in uncoordinated animals arrested at the L1 developmental stage, displaying mild disorganization of the myofilament lattice in their muscle cells [33]. The gene showing the second largest percentile difference is dys-1, which ranks top 15% in terms of information flow scores in the muscle network and 71% in the entire network. dys-1 encodes an orthologue of the human DMD [34], which when mutated leads to Duchenne muscular dystrophy, a severe recessive x-linked form of muscular dystrophy that is characterized by rapid progession of muscle degeneration. The gene showing the third largest percentile difference is lev-11, which ranks in the top 21% in terms of information flow scores in the muscle network and 78% in the entire network. lev-11 encodes an orthologue of the human TROPOMYOSIN 1 [35] (www. wormbook. org), which when mutated leads to familial hypertrophic cardiomyopathy, a genetic disorder caused by the thickening of heart muscle. The gene showing the fourth largest percentile difference is deb-1, which encodes a muscle attachment protein found in dense bodies, and is required for attaching actin thin filaments to the basal sarcolemma [35] (www. wormbook. org). Out of the top 35 genes that show the largest differences, RNAi feeding strains are available for 25 genes from a library [36]. We performed feeding RNAi experiments using the rrf-3 strain, an RNAi-sensitive strain, and found that the perturbation of 6 genes (24%) cause motility defect (Table 1). In contrast, RNAi experiments of only 1 out of 16 genes (6%) that rank the lowest in terms of percentile differences revealed any motility defect (Table 1). As a general reference, in a genome-wide RNAi screen using the rrf-3 strain [37], RNAi experiments of 4. 1% of all tested genes showed paralyzed or uncoordinated phenotypes. Even among the muscle-enriched genes identified by the semi-supervised learning method, only 9% of the genes correspond to a paralyzed or uncoordinated phenotype. The analysis' result supports our hypothesis that genes of high information flow specifically in the muscle network play important roles in normal muscle function. It is plausible that the genes showing higher information flow scores in the muscle network than the entire network can also be distinguished by conventional methods such as betweenness. To clarify this, we obtained the percentiles of genes in terms of betweenness in the muscle network and that of genes in the entire network, and ranked the genes by the differences between the two percentiles (Table S6). The top genes identified by differences in information flow do not necessarily rank high by the differences in betweenness (Table 1 and Table S6). For example, tag-327, dys-1, lev-11, and deb-1, the top four genes identified by differences in information flow, only rank No. 20,23,58, and 59 by differences in betweenness, respectively. This is due to the fact that the information flow model considers the confidence of interactions derived from co-expression while betweenness does not. Similarly, if we rank genes by the probability of expression in muscle, Pi (muscle), as derived from the semi-supervised learning method, tag-327, dys-1, lev-11, and deb-1 rank only at No. 149,269,97, and 124, respectively. The relevance in muscle function of these genes has been reported in the literature [33]–[35], suggesting that the information flow method does identify biologically relevant candidate genes that can be distinguished using neither the gene expression data nor a graph metric such as betweenness. We model interactome networks as large electrical circuits of interconnecting junctions (proteins) and resistors (interactions). Our model identifies candidate proteins that make significant contributions to the transfer of biological information between various modules. Compared to degree and betweenness, our model has two major advantages: first, it incorporates the confidence scores of protein-protein interactions; second, it considers all possible paths of information transfer. When a protein that mediates information exchange between modules is knocked down, the disintegration of multiple modules is very likely to result in lethality. Even if the organism is still viable, pleiotropy may be observed because multiple phenotypes imply the breakdown of multiple modules. In support of our model, we find that the information flow score of a protein is well correlated with the likelihood of observing lethality or pleiotropy when the protein is eliminated. Even among proteins of low or medium betweenness, the information flow model is predictive of a protein' s essentiality or pleiotropy. Compared to betweenness, the information flow model is not only more effective but also more robust in face of a large amount of low-confidence data. Our method is accessible to the public. The MATLAB implementation of the information flow algorithm, along with the information flow scores for proteins in the yeast interactome network and proteins in the worm interactome network, can be downloaded at http: //jura. wi. mit. edu/ge/information_flow_plos/. The information flow model identifies central proteins in interactome networks, and these proteins are likely to connect different functional modules. We developed an algorithm that decomposes interactome networks into subnetworks by removing proteins of high information flow in a recursive manner (Figure 9) (Materials and Methods). Starting from the largest network component, we removed the protein with the highest information flow score. If the proteins remained connected in a single network, we removed the protein with the next highest information flow score one-at-a-time, until the network fell into multiple pieces upon the protein removal. We then counted the number of proteins in each of the subnetworks. If a subnetwork contained between 15 and 50 proteins, we examined whether any Gene Ontology (GO) term was enriched among proteins in the subnetwork [38], [39]. If a subnetwork contained over 50 proteins, we repeated the procedure of removing high information flow proteins from the subnetwork. Overall, we obtained 37 subnetworks, and all but two of them were enriched with proteins from certain GO categories (Table S7). We investigated the effects of varying the minimum and maximum size of subnetworks (Text S2). The selected range of 15 to 50 proteins was based on the number of recovered subnetworks as well as the overall GO enrichment scores. If we increased the minimum subnetwork size to 20 proteins, the number of subnetworks shrank to 24, all of which were functionally enriched. However, in order to recover the additional 11 GO enriched subnetworks for a total of 35, we decided to keep the lower threshold at 15 proteins. The fact that the majority of subnetworks are functionally enriched provides additional evidence that proteins with high information flow score interconnect different modules. It was previously observed in a yeast interactome network that ‘date hubs’, which connect different modules, are more likely to participate in genetic interactions than randomly sampled proteins, because elimination of date hubs may make the organism more sensitive to any further genetic perturbations [40]. We tested whether proteins of high information flow and proteins of high betweenness show the same property in the C. elegans interactome. We found that genes that rank the highest 30% in terms of information flow or betweenness are more likely to participate in genetic interactions than randomly selected genes (P-value = 1. 16×10−10 and P-value = 1. 16×10−10, respectively). This is not particularly surprising because many proteins of high information flow or high betweenness are hubs in the network. Another possible feature of “between-module” proteins is related to the expression dynamics of these proteins and their interacting partners. In general, interacting proteins are likely to share similar expression profiles [41]. Date hubs in yeast interactome networks have been found to be less correlated with their binding partners in terms of expression dynamics than ‘party hubs’ which function within a functional module [40]. Proteins of high betweenness in yeast interactome networks have also been reported to show the lack of expression correlation with their binding partners [14]. On the other hand, it has been argued in another study that the lack of correlation is dependent on the datasets examined [42]. We investigated the correlation of expression profiles [43], [44] for proteins of high information flow or proteins of high betweenness with their interacting partners in the C. elegans interactome. We did not find proteins of high information flow or proteins of high betweenness behaving differently from other proteins in terms of expression correlation with their interacting partners (data not shown). Thus the expression correlation between topologically central proteins and their binding partners may be worth further investigations. The transmission of biological signals is directional while at present interactome networks often reflect the formation of protein complexes [3] and do not contain directionality. We explored whether the information flow model is also applicable to signaling networks with directionality. We generated a signaling network for S. cerevisiae by integrating phosphorylation events [45] and Y2H interactions (see Materials and Methods). In this network, we examined the top 30% versus the bottom 30% of genes ranked by the information flow score. We found a significant increase in the percentage of pleiotropic genes in the former group (17. 0%) as compared to the latter (5. 3%) (Table S8) (P-value = 0. 01), though the percentages of essential genes are similar for the two groups. This analysis suggests that the information flow model is useful for discovering crucial proteins in signaling networks, as well as in networks of protein complexes. The lack of correlation with lethality may reflect the fact that fewer proteins in signaling networks participate in “housekeeping” functions, which are often mediated by multi-protein molecular machines. In the future, with more information integrated into interactome networks, we should be able to improve on the performance of information flow model. In addition, interactome networks can vary at different times or in different spatial locations. After all, we still have very limited understanding of how biological information flows through cellular networks. Most likely, it does not flow exactly as the electrical current flow does. As more knowledge is accumulated, we should be able to modify the information flow model according to the design principles of cellular network and highlight the dynamic nature of cellular networks. All of the data used in our study comes from openly available databases and published high-throughput datasets. We obtained a list of essential genes for S. cerevisiae from the Saccharomyces Genome Database (http: //www. yeastgenome. org/) and a list of essential genes for C. elegans embryos from the WormBase (http: //www. wormbase. org/). We downloaded phenotypic data of S. cerevisiae deletion strains under various conditions [46] and C. elegans post-embryonic phenotypes from genome-wide RNAi screens [37], [47]. We also downloaded interaction datasets for S. cerevisiae [3], [45], [48] and C. elegans [2], [26], [27]. Betweenness is a centrality measure of a node in a network graph. The betweenness of a particular node is determined by how often it appears on the shortest paths between the pairs of remaining nodes [12]. For a graph with N nodes, the betweenness CB () for node is: where σst represents the number of shortest paths from node s to node t, and σst () represents the number of shortest paths from node s to node t that pass through node. To compute shortest path, we used Dijkstra algorithm [49]. Dijkstra algorithm is a greedy search algorithm that solves the single-source shortest path problem for a directed graph with non negative edge weights. We modified it to handle edges without directionality. We model an interactome network as a resistor network, where proteins are represented as nodes and interactions are represented as resistors. The conductance of each resistor is directly proportional to the confidence score of the corresponding interaction. In cases where the confidence levels of interactions are not known, we assume that all resistors have unit conductance. In order to estimate the importance of node k in conducting electrical current in a network of N nodes, we connect node i to a unit current source and node j to the ground, and we compute how much current flows through node k using Kirchhoff' s laws (see Figure S3). We define the information flow score of node k as the sum of current through node k among all pair-wise combinations of source and ground nodes. Since exchanging the source node and the ground node does not lead to different current distributions, we perform the calculation of information flow scores only for cases where i>j. The total number of pair-wise combinations of nodes (i, j), such that i≠k, j≠k and i>j is (N-2) (N-3) /2. The information flow through node k is (1) where I km is the current between the nodes k and m, and Σ m is the sum over all resistors connected to node k. For a given pair of source node and ground node, the standard way of computing resistor currents of a circuit is using nodal analysis and solving the resulting system of (N -1) linear equations for node voltages. For each node m that is not a ground node, we have the following equation: (2) where v l is a voltage at node l, and the sum is over all nodes directly connected to node m. When node m is a source node, the right-hand side of equation (2) is a unit value of current. Node voltages can be computed by solving the following linear system of equations: (3) where G is a symmetric (N-1) × (N-1) conductance matrix, v is a vector of unknown node voltages and J is a vector of currents to every node. The matrix G can be calculated using the following algorithms. Our information flow model identifies central proteins in interactome networks. Very likely the proteins of high information flow scores represent connecting points of functional modules. To test this hypothesis, we designed an algorithm to recursively remove the highest flow proteins and release subnetworks from a large interactome network component. In the algorithm described below, a ‘core module’ refers to a subnetwork composed of 15 to 50 proteins. To evaluate the performance of information flow in signaling networks, we combined a phosphorylation dataset for S. cerevisiae which contained kinases and their target proteins [45] with various sources of Y2H data [48]. Specifically, we searched for Y2H interactions between the target proteins in the phosphorylation dataset. As a result, we obtained a set of 77 kinases involved in 1008 phosphorylation events with 312 target proteins interconnected by 503 Y2H interactions. Each kinase phosphorylates one or more of the 312 proteins in the Y2H network. In order to retain the directionality of phosphorylation in the information flow model, we compute the information flow separately for each kinase. First, we use directed edges to link the kinase to its phosphorylation targets in Y2H network. Next, we set the kinase to be a source and sequentially set the remaining 312 proteins to be sinks as we compute the information flow. Before we move on to the next kinase, we remove the phosphorylation edges specific to the previous kinase. The total information flow score for each of the 312 proteins in the Y2H network is obtained by summing the absolute values of information flow from 77 kinase-specific networks. We performed RNA interference (RNAi) experiments by feeding L4 worms, following protocols from the WormBook [50] (www. wormbook. org). The bacteria strains for feeding RNAi experiments were from an RNAi library [36] that is commercially available.

Protein–protein interactions mediate numerous biological processes. In the last decade, there have been efforts to comprehensively map protein–protein interactions occurring in an organism. The interaction data generated from these high-throughput projects can be represented as interconnected networks. It has been found that knockouts of proteins residing in topologically central positions in the networks more likely result in lethality of the organism than knockouts of peripheral proteins. However, it is difficult to accurately define topologically central proteins because high-throughput data is error-prone and some interactions are not as reliable as others. In addition, the architecture of interaction networks varies in different tissues for multi-cellular organisms. To this end, we present a novel computational approach to identify central proteins while considering the confidence of data and gene expression in tissues. Moreover, our approach takes into account multiple alternative paths in interaction networks. We apply our method to yeast and nematode interaction networks. We find that the likelihood of observing lethality and pleiotropy when a given protein is eliminated correlates better with our centrality score for that protein than with its scores based on traditional centrality metrics. Finally, we set up a framework to identify central proteins in tissue-specific interaction networks.

Abstract Introduction Results Discussion Materials and Methods

genetics and genomics computational biology

2009

Information Flow Analysis of Interactome Networks

The evolution of signal transduction pathways is constrained by the requirements of signal fidelity, yet flexibility is necessary to allow pathway remodeling in response to environmental challenges. A detailed understanding of how flexibility and constraint shape bacterial two component signaling systems is emerging, but how new signal transduction architectures arise remains unclear. Here, we investigate pathway remodeling using the Firmicute sporulation initiation (Spo0) pathway as a model. The present-day Spo0 pathways in Bacilli and Clostridia share common ancestry, but possess different architectures. In Clostridium acetobutylicum, sensor kinases directly phosphorylate Spo0A, the master regulator of sporulation. In Bacillus subtilis, Spo0A is activated via a four-protein phosphorelay. The current view favors an ancestral direct phosphorylation architecture, with the phosphorelay emerging in the Bacillar lineage. Our results reject this hypothesis. Our analysis of 84 broadly distributed Firmicute genomes predicts phosphorelays in numerous Clostridia, contrary to the expectation that the Spo0 phosphorelay is unique to Bacilli. Our experimental verification of a functional Spo0 phosphorelay encoded by Desulfotomaculum acetoxidans (Class Clostridia) further supports functional phosphorelays in Clostridia, which strongly suggests that the ancestral Spo0 pathway was a phosphorelay. Cross complementation assays between Bacillar and Clostridial phosphorelays demonstrate conservation of interaction specificity since their divergence over 2. 7 BYA. Further, the distribution of direct phosphorylation Spo0 pathways is patchy, suggesting multiple, independent instances of remodeling from phosphorelay to direct phosphorylation. We provide evidence that these transitions are likely the result of changes in sporulation kinase specificity or acquisition of a sensor kinase with specificity for Spo0A, which is remarkably conserved in both architectures. We conclude that flexible encoding of interaction specificity, a phenotype that is only intermittently essential, and the recruitment of kinases to recognize novel environmental signals resulted in a consistent and repeated pattern of remodeling of the Spo0 pathway. Responses to changing environmental conditions are mediated by signal transduction pathways that recognize a signal, convey that signal into the cell, and initiate an appropriate cellular response. In bacteria, two-component signaling systems, typically comprised of a histidine kinase (HK) and a cognate response regulator (RR), are a primary mechanism of environmental response (Fig 1A). Signal recognition by the N-terminal sensor region of the HK leads to the autophosphorylation of a conserved histidine residue in the so-called HisKA domain by the catalytic (HK_CA) domain. The signal is then transduced by phosphotransfer from the autophosphorylated HK to a conserved aspartate residue in the N-terminal receiver (REC) domain of the RR [1]. Phosphorylation of the REC domain activates the C-terminal output domain of the RR, initiating a response to the recognized signal. Bacteria typically encode 20 to 30 two-component signaling pathways per genome [2]. A set of non-contiguous, co-evolving residues at the interface of HK and RR proteins, six in the HisKA domain and seven in the REC domain, ensure specific interaction within each cognate pair [3–6]. These specificity residues are partially degenerate: multiple sets of kinase specificity residues permit phosphotransfer to the same receiver (and vice versa [7]), such that each receiver has a spectrum of kinase specificity with which it can interact (Fig 2A). To prevent deleterious crosstalk between non-cognate proteins [8], selection acts to separate the spectra of two-component signaling pathways encoded in the same genome. Acquisition of novel pathways (e. g. , through duplication or horizontal gene transfer) can cause conflicts in interaction space. The degeneracy of these interactions allows for repositioning in interaction space to eliminate crosstalk via mutational trajectories involving compensatory mutations in the cognate pair. However, in the absence of a perturbation, pathways likely remain in the same region of interaction space over the course of evolution [8]. Histidine-aspartate phosphotransfer also admits more complex signal transduction architectures. Examples include multiple-input architectures [9], multiple-output architectures [10], and so-called phosphorelays comprising a sequence of phosphotransfer interactions [11–13]. For example, the sporulation initiation (Spo0) pathway is a multi-input phosphorelay characterized extensively in B. subtilis [13–15] and also observed in closely related species [16–18]. In this architecture, multiple sensor kinases phosphorylate Spo0F, a protein possessed of a REC domain, but lacking an output domain; subsequently, that phosphoryl group is transferred via Spo0B, an intermediate histidine phosphotransferase, to Spo0A, the master regulator of sporulation (Fig 1B). The maintenance of signal fidelity in these more complex pathways entails additional constraints on the genetic determinants of specificity because a single protein must support multiple interactions. The interaction requirements of the Spo0 phosphorelay necessitate precise molecular recognition to allow both Spo0F and Spo0A to interact with Spo0B, but only Spo0F to accept a phosphoryl group from sporulation kinases (Fig 2B). The balance of flexibility and constraint that shapes molecular recognition in these complex architectures is not well understood. To explore this issue, we present here an analysis of the evolution of the Spo0 pathway. The Spo0 pathway controls entrance into a developmental program that produces stress-resistant, dormant endospores. The ability to produce endospores is a common feature of the Firmicutes phylum, observed in numerous species throughout two anciently related classes, the Bacilli and Clostridia, suggesting that this survival mechanism is ancient [19,20]. These two classes are predicted to have diverged 2. 7 billion years ago, coinciding with the atmospheric rise of oxygen during the great oxidation event [21]. The ancestral Firmicute was likely an obligate anaerobe, a trait that has been preserved in the present-day Class Clostridia, whereas the Bacilli are typically facultative aerobes. Many taxonomic families in both classes include both sporogenous and asporogenous species, suggesting that the ability to sporulate is frequently lost [22] through adaptation to a stable niche where sporulation is unnecessary for survival [23]. Strikingly, a comparison of the Spo0 pathways in the type species of the two Firmicutes classes, Bacillus subtilis [13] and Clostridium acetobutylicum [24], reveals that the outputs of these pathways are conserved [25], but the inputs and the signal transduction architectures are not. Spo0A, the terminal component of the pathway in both species, initiates spore development upon phosphorylation [26,27] and is encoded by all known sporulators [22]. Spo0A is a canonical response regulator protein in its domain composition, including a REC domain [28] and a highly conserved, DNA-binding output domain, Spo0A_C [29]. Unlike Spo0A, which is likely orthologous in these distantly related species, the upstream signal transduction architectures are different. In contrast to the B. subtilis multi-input phosphorelay Spo0 architecture, C. acetobutylicum and other closely related species possess a multi-input architecture in which Spo0A is directly phosphorylated by multiple kinases [24,30,31] (Fig 1C). Considering that these two different signal transduction architectures both orchestrate the initiation of sporulation through the phosphorylation of an orthologous regulator, they likely arose from a common ancestral pathway. How, then, did different signaling architectures evolve in present day species? The prevailing view is that the ancestral Spo0 pathway had a two-component direct phosphorylation architecture and the more complex phosphorelay observed in B. subtilis is a derived state [32–34]. This hypothesis was inspired by the apparent lack of Spo0F and Spo0B orthologs in the first Clostridium genome sequenced [32]. The simplicity of the direct phosphorylation architecture and the similarly anaerobic lifestyles of the ancestral Firmicutes and present-day Clostridia, taken together, provided further support for predictions that the original Spo0 pathway also functioned through direct phosphorylation [33]. It was further proposed that the phosphorelay likely arose in the Bacillar lineage, possibly as the result of duplication of a cognate HK-RR pair [35], and that the additional points of control associated with a phosphorelay may have contributed to adaptation to rising oxygen levels in early Bacilli [36]. Regardless of the status of the ancestral pathway, some combination of gains and losses of interaction must have occurred to produce the distinct pathway architectures observed in present day species. We took advantage of the dramatic increase in the number of sequenced Firmicutes genomes available to investigate these remodeling events. Our results challenge the prevailing hypothesis. In silico analyses, combined with in vitro experimental verification of a Clostridial phosphorelay, reveal that phosphorelay architectures are present throughout the Firmicutes. Further, we demonstrate that interaction specificity of representative Bacillar and Clostridial phosphorelays is functionally conserved. In contrast to the prevailing model, our results support a scenario in which the ancestral Spo0 pathway in the Firmicutes ancestor was a phosphorelay. The phylogenetic distribution of Spo0 architectures is patchy, consistent with several independent transitions from phosphorelay to direct phosphorylation architecture. Our results further suggest that these transitions were mediated via changes in sensor kinases, while Spo0A specificity is conserved across the Firmicutes phylum. Our findings provide a framework for reasoning about the forces that act to maintain signaling fidelity in complex signal transduction pathways with multiple interactions. We undertook a survey of Spo0 components in the representative genomes to establish the architecture of modern-day Spo0 pathways. First, to establish which genomes in our representative set likely encode Spo0 pathways, we searched for response regulators encoding a Spo0A_C domain, which uniquely distinguishes Spo0A from other response regulators. We identified 68 genomes that encode an apparent Spo0A ortholog (Fig 3, green dots). Of these, 53 have been observed to form spores (Fig 3, red leaves in Class Clostridia, blue leaves in Class Bacilli; S1 Table). The presence of Spo0A in 15 apparent non-spore-formers could be due to a recent loss of sporulation or an alternate functional role for Spo0A in those species. It is also possible that these species are sporogenous, but spore formation has not been observed under the conditions tested [46]. Next, to determine the prevalence of the Spo0 phosphorelay in present-day Firmicutes, we searched for orthologs of the response regulator Spo0F and the histidine phosphotransferase Spo0B. These proteins are necessary, along with a sporulation kinase, to phosphorylate Spo0A and initiate sporulation in B. subtilis [13] and likely also in other Bacillar Spo0 pathways [16–18]. The only reported functional roles for Spo0F and Spo0B are as signal transduction intermediates in the Spo0 pathway [reviewed in 22,47]. Thus, the presence of homologs of both Spo0F and Spo0B is strong evidence of a Spo0 pathway with a phosphorelay architecture. Prediction of Spo0F and Spo0B homologs via sequence similarity methods has proven challenging due to the specific characteristics of Spo0F and Spo0B. Spo0F contains only a REC domain (PFAM: PF00072), making Spo0F orthologs difficult to distinguish from other response regulators that lack an output domain, such as the chemotaxis protein, CheY. Spo0B sequences lack strong sequence conservation, even within the same genus. Over broader evolutionary distances, sequence comparison cannot distinguish between Spo0B proteins and histidine kinases unambiguously [47]. PFAM domains annotated to known Spo0B homologs are either too general (SPOB_a, PFAM: PF14689) or too specific (SPOB_ab, PFAM: PF14682) to be useful identifying features for Spo0B. Having concluded that sequence similarity and domain content do not provide sufficient information to identify phosphorelay protein orthologs, we devised an alternative method for the identification of orthologs of Spo0F and Spo0B, based on genome context. As a guide, we considered the several dozen proteins from strains of B. subtilis and its closest relatives that are annotated as Spo0F or Spo0B in the RefSeq database [48]. This guide set includes three experimentally verified Spo0F proteins [13,17,49] and two experimentally verified Spo0B proteins [13,47]. Unlike many canonical two-component signaling proteins, the sporulation phosphorelay proteins are encoded in dispersed regions of the genome. The Spo0F guide set revealed that Spo0F homologs are almost always encoded immediately upstream of a fructose bisphosphate aldolase (fbaA) gene. Two other proteins, a CTP synthase and a transaldolase, are also commonly encoded in close proximity. Using these three genes as Spo0F neighborhood markers, putative orthologs of Spo0F were identified (S3 Table, S4 Fig). All but two spore-forming Class Bacilli genomes investigated contain a Spo0F. We also identified candidate Spo0F genes in 17 spore-forming genomes within Class Clostridia (Fig 3, dark blue dots). In contrast, 13 spore-forming Class Clostridia genomes do not encode a Spo0F-like gene in the vicinity of any of the Spo0F neighborhood markers, nor do they encode any two of the neighborhood markers in close proximity to each other. In particular, no Spo0F candidates were found in species in which a direct phosphorylation architecture has been experimentally verified (C. acetobutylicum [24], R. thermocellum [30], and C. difficile [31]). These results suggest that Spo0F homologs can be identified by conserved genome neighborhoods and are found not only in Class Bacilli, but also in many early-branching Class Clostridia taxa. We next investigated whether genome neighborhood conservation could also be used to predict Spo0B-encoding genes. Each gene annotated as spo0B in RefSeq is flanked by two downstream genes encoding ribosomal proteins, L21 and L27, and an upstream gene encoding the GTPase ObgE (see also [47]). In our set of representative Firmicutes, 75 genomes encode this trio in close proximity (Fig 3, orange dots; S4 Table, S5 Fig). All but two spore-formers in Class Bacilli were found to encode a Spo0B-like protein within a five gene window that includes all three marker genes. The genomes of 18 of the 30 spore-formers within the Class Clostridia also had a region containing the three marker genes and a candidate Spo0B ortholog. The remaining Class Clostridia genomes encoded the three Spo0B neighborhood markers in close proximity, but did not encode a protein meeting the criteria of Spo0B in that vicinity. No putative Spo0B was identified in any Class Clostridia species in which direct phosphorylation of Spo0A has been verified experimentally (S4 Table). Thus, Spo0B homologs can also be identified by conservation of genome neighborhood and are found in almost all genomes in which a Spo0F homolog was identified. In summary, we predicted Spo0F and Spo0B orthologs in most Class Bacilli genomes and in genomes broadly distributed within Class Clostridia. As the only known function of Spo0F and Spo0B proteins is phosphotransfer within the Spo0 phosphorelay, these proteins likely also perform this role in Class Clostridia species. This identification of phosphorelay proteins in multiple spore-forming taxa in Class Clostridia conflicts with the standing hypothesis that Spo0 phosphorelays, and therefore phosphorelay proteins, are restricted to Class Bacilli. The presence of putative Spo0F and Spo0B proteins in some spore-forming Class Clostridia species suggests that these organisms may, like those in Class Bacilli, signal the initiation of sporulation through a phosphorelay architecture. To determine whether the Spo0 proteins predicted by our method do, in fact, participate in a phosphorelay, we sought to test the in vitro phosphotransfer properties [50] of the putative phosphorelay proteins from Desulfotomaculum acetoxidans DSM771 (Class Clostridia, starred in Fig 3), a spore-forming species in the Peptococcaceae [51]. The predicted homologs of Spo0F and Spo0B in this genome have conserved genomic neighborhoods. Comparison of the predicted Spo0 proteins in D. acetoxidans with their B. subtilis counterparts indicated a high degree of similarity in their respective specificity residues (Table 1, see Methods for specificity residue prediction). Experimental testing of a possible D. acetoxidans phosphorelay also required prediction of the sporulation kinase (s). Experimentally verified Spo0 kinases possess few shared sequence features that definitively distinguish sporulation kinases from other sensor histidine kinases. Analysis of the regions flanking known sporulation kinases did not reveal any conservation of the genomic neighborhood. The HisKA and HK_CA domains of sporulation kinases are not markedly more similar to each other than to those of other sensor kinases, and the N-terminal sensor regions of bona fide sporulation kinases vary substantially. However, all experimentally verified sporulation kinases (S2 Table) are orphans, i. e. , are not co-located with genes encoding other two-component signaling system proteins. Moreover, N-terminal PAS domains are observed more frequently in sporulation kinases than in the set of all kinases in the same species (S2 Table). Thus, orphan status, combined with the presence of an N-terminal PAS domain, furnishes a signature for predicting candidate Spo0 kinases. D. acetoxidans has seven orphan kinases, six of which encode a PAS domain. Strikingly, all six have putative specificity residues similar to verified Bacillus sporulation kinases (Table 1) suggesting that they may target Spo0F. Of these six kinases, Dtox_1918 was chosen for the phosphotransfer experiments as it has specificity residues differing from B. subtilis KinA at only one position. To test the hypothesis that phosphotransfer to D. acetoxidans Spo0A will only be observed in the presence of a sporulation kinase, Spo0F, and Spo0B, we purified affinity-tagged variants of the four predicted D. acetoxidans Spo0 proteins (Table 2, rows 1–4; see also Methods). For the multidomain proteins (the kinase, Dtox_1918, and Spo0A, Dtox_2041), we used truncated sequences that contain the interaction domains. To assess phosphotransfer connectivity, the purified kinase, Dtox_1918, was first incubated alone with radiolabeled ATP for 15 minutes and then examined by SDS-PAGE and autoradiography (Fig 4A, lane 1). Two bands representing autophosphorylated Dtox_1918 were seen, consistent with different kinase oligomers. Inclusion of Spo0F in the reaction (Fig 4A, lane 2) produced an additional band indicating that Spo0F can be directly phosphorylated by Dtox_1918. Similarly, the addition of Spo0F and Spo0B to autophosphorylated Dtox_1918 produced bands for each of the three proteins (Fig 4A, lane 3) and addition of Spo0F, Spo0B, and Spo0A produced bands corresponding to all four proteins (Fig 4A, lane 4). Importantly, Spo0A is only phosphorylated in the presence of both Spo0F and Spo0B (Fig 4A, lanes 5 and 7). Further, the phosphorylation of Spo0B requires the presence of Spo0F (Fig 4A, lanes 6 and 7, S8 Fig). Finally, we confirmed that Dtox_1918 cannot directly phosphorylate Spo0A under these conditions (Fig 4A, lane 8), or even following a longer incubation (Fig 4B). Collectively, these results indicate that the Spo0F, Spo0B, and Spo0A homologs identified above, in conjunction with Dtox_1918, comprise a bona fide phosphorelay similar in architecture to that first characterized in B. subtilis. The D. acetoxidans phosphorelay is, to our knowledge, the first experimentally verified Spo0 phosphorelay outside of Class Bacilli. The experimental confirmation of our computational predictions in D. acetoxidans attests to the reliability of our prediction signatures for both Spo0F and Spo0B. This is corroborated by the consistency of the predictions across the complete data set: almost all species either encode both Spo0F and Spo0B or encode neither. Given that a similarity in specificity residues correlates with phosphotransfer capability in vitro [4,7, 52], we further compared the specificity residues of the predicted Spo0F and Spo0B orthologs, and their putative interaction partners, with the specificity residues of their counterparts in experimentally verified pathways. The predicted specificity residues (S5 Table, Materials and methods), represented as logos (Fig 5A and 5B), show strong similarity to specificity residues in experimentally verified phosphorelay proteins (Table 1; S2 Table). This further supports our prediction of Spo0F and Spo0B orthologs and suggests that they function as intermediate proteins in a phosphorelay architecture. Based on the combined evidence, we predict that the 33 spore-formers that encode orthologs of Spo0F, Spo0B, and Spo0A possess a phosphorelay. The remaining 15 spore-formers, in which no Spo0F or Spo0B were identified, likely possess a direct phosphorylation architecture. All 48 spore-formers in our data set, with the exception of Erysipelatoclostridium ramosum, possess at least one orphan kinase. Of those, 41 have at least one orphan kinase with an N-terminal PAS domain. The predicted pathway architectures agree with the experimental evidence in all species in which the Spo0 pathway architecture has been investigated [13,16–18,24,30,31]. Examination of the phylogenetic distribution of these predicted pathways reveals abundant phosphorelays, not only in Class Bacilli, but also in Class Clostridia. Further, comparison of the specificity residues in phosphorelays predicted in the two classes reveals striking similarity in each Spo0 component (Fig 5A and 5B, see S2 Text for a quantitative comparison), suggesting that the genetic determinants of specificity are similarly encoded in both classes. To test this hypothesis, we asked whether phosphorelay proteins in D. acetoxidans could recapitulate the function of the corresponding proteins in B. subtilis in vitro (Table 2, lines 5–8). Specifically, we examined phosphotransfer in the B. subtilis phosphorelay, as described for D. acetoxidans above (Fig 4), systematically replacing each B. subtilis protein with its D. acetoxidans counterpart (Fig 6, lanes 1–6). For each step in the pathway, a band corresponding to the replacement D. acetoxidans protein was observed, demonstrating that each D. acetoxidans protein was capable of accepting a phosphoryl group from the upstream B. subtilis Spo0 pathway component (Fig 6, lanes 2,4, and 6). Moreover, bands were observed for downstream components of the B. subtilis phosphorelay, where included, indicating that the D. acetoxidans replacement was also capable of transferring a phosphoryl group to the downstream component in the B. subtilis phosphorelay (Fig 6, lanes 1,3, 5). Additionally, we noted that phosphorylation of D. acetoxidans Spo0A by the B. subtilis phosphorelay required the presence of both B. subtilis Spo0F and Spo0B proteins (Fig 6, lanes 7–9), indicating that D. acetoxidans Spo0A cannot be directly phosphorylated by B. subtilis KinA. Taken together, our results demonstrate that, in all cases, the D. acetoxidans proteins were able to recapitulate the function of their counterparts in B. subtilis. Thus, not only do these pathways independently interact as phosphorelays [13, this work], they also encode sufficiently similar phosphotransfer specificity to render them functionally interchangeable, at least in vitro. Therefore, either these two phosphorelays arose independently with interchangeable genetic determinants of specificity, which we deem highly unlikely, or these pathways are the descendants of a common ancestral pathway. Having established evidence of common ancestry and conserved specificity in the Spo0 phosphorelay, we next considered the evolutionary history of the direct phosphorylation Spo0 pathway. Examination of the phylogenetic distribution of the predicted architectures in Clostridia (Fig 3, summarized in Fig 5D) reveals that neither predicted architecture is monophyletic. To ensure that this patchiness is not a byproduct of taxon sampling or phylogeny reconstruction artifacts, we repeated the computational analysis with two other Firmicutes phylogenies [43,44], one of which includes a much larger set of genomes (see S1 Text, S2 and S3 Figs). Both analyses revealed similar patchiness. The patchy distribution of predicted Spo0 architectures is consistent with multiple changes in pathway architecture over the course of evolution. Each change in architecture requires the gain of one phosphotransfer interaction and the loss of another. This could occur via changes in kinase specificity residues, resulting in phosphotransfer to a different response regulator (Fig 7A), changes in the Spo0A specificity spectrum, allowing phosphorylation by a different phosphodonor (Fig 7B), or a combination of the two. To investigate the changes in specificity that resulted in the present-day distribution of Spo0 architectures, we constructed specificity residue logos for orphan kinases and Spo0A proteins encoded in genomes possessing phosphorelays (Fig 5A and 5B), and compared them with the corresponding direct phosphorylation architecture logos (Fig 5C). This comparison revealed that the similarity across architectures is greater for Spo0A proteins than for orphan kinases. When Clostridial and Bacillar phosphorelays are considered separately, Spo0A specificity residues are more similar within the same taxonomic class, than within the same pathway type (Fig 5A–5C; S2 Text, S5 Table). The opposite is true for candidate sporulation kinases. The specificity residues of candidate phosphorelay kinases from both the Clostridia and the Bacilli differ markedly from those of kinases predicted to phosphorylate Spo0A directly. This difference is even more dramatic when experimentally verified sporulation kinases associated with the two architectures are compared (Fig 5E). These results suggest that architectural remodeling was driven primarily by changes in kinase specificity and not changes in Spo0A specificity. To test this prediction, we probed the heterologous interactions between Spo0 proteins (Table 2) from a Bacillar phosphorelay (B. subtilis), a Clostridial phosphorelay (D. acetoxidans), and a Clostridial direct phosphorylation pathway (C. acetobutylicum). In addition to the phosphorelay kinases, B. subtilis KinA and D. acetoxidans Dtox_1918, we included two direct phosphorylation kinases (CA_C0903, CA_C3319) that were chosen to span the diversity of specificity residues observed in experimentally verified sporulation kinases in C. acetobutylicum [24]. Each of these kinases was incubated with each of the five receiver proteins (two Spo0F and three Spo0A proteins) in separate reactions. To test Spo0B interaction connectivity, each Spo0B protein was incubated with each of the three Spo0A proteins, in separate reactions, in the presence of its autologous kinase and Spo0F. Interactions between a phosphodonor and a heterologous phosphoreceiver (Figs 8 and 9, summarized in Table 2) allow us to infer the relative positions in interaction space of receiver specificity spectra from various species, because the spectra of receivers that can be phosphorylated by the same donor must overlap. Thus, observation that phosphorelay proteins from B. subtilis and D. acetoxidans are functionally interchangeable (Fig 6 and systematically probed in Fig 8) suggests that the specificity spectra of phosphorelay receivers, Spo0F and Spo0A, have changed very little. Moreover, both of the Spo0B proteins tested are capable of phosphorylating Spo0A proteins associated with either phosphorelay or direct phosphorylation architectures (Fig 8A and 8B). Importantly, this shows that the overlapping region of the Spo0F and Spo0A spectra in phosphorelays also overlaps with the specificity spectrum of the directly phosphorylated Spo0A protein in C. acetobutylicum. Heterologous interactions also allow us to test hypotheses for the changes in specificity associated with pathway remodeling by probing differences in receiver specificity spectra in phosphorelay and direct phosphorylation architectures. One possibility is that the change in pathway architecture arose through changes in kinase specificity, with little or no change in the Spo0A specificity spectrum. This hypothesis predicts that direct phosphorylation kinases, which phosphorylate Spo0A in their native environments, will also phosphorylate Spo0A proteins from phosphorelays (Fig 7A). An alternate scenario is that the change in pathway architecture is due to changes in Spo0A specificity. This hypothesis predicts that direct phosphorylation kinases will phosphorylate heterologous Spo0F proteins and phosphorelay kinases will phosphorylate Spo0A proteins associated with direct phosphorylation architectures (Fig 7B). Consistent with the first scenario, both direct phosphorylation kinases tested (CA_C0903, Fig 9A and CA_C3319, Fig 9B) were able to phosphorylate both phosphorelay Spo0A proteins, while only one (CA_C3319, Fig 9C, lanes 3 and 4) was able to phosphorylate Spo0F. In contrast, we only observe two of the four interactions expected in the second scenario: C. acetobutylicum CA_C0903 did not phosphorylate either of the heterologous Spo0F proteins tested (Fig 9C, lane 1 and 2) and no interaction was observed between B. subtilis KinA and C. acetobutylicum Spo0A (Fig 8C, lane 1). Thus, we observe heterologous interactions that are predicted by the first hypothesis, as well as two additional interactions: First, in addition to both Spo0A proteins, CA_C3319 phosphorylated both Spo0F proteins, indicating that the specificity of CA_C3319 is located in the overlapping region of Spo0F and Spo0A spectra. Selection against crosstalk would act to exclude phosphorelay kinases from this overlapping region; however, such selection would not act on CA_C3319 because Spo0F is not encoded in the same genome. Second, the phosphorelay kinase, Dtox_1918, although unable to phosphorylate the Spo0A encoded in its own genome (Fig 8C, lane 7), phosphorylated both Spo0A proteins encoded in other genomes (Fig 8C, lanes 6 and 8), suggesting a minor shift in the Spo0A spectrum in D. acetobutylicum, relative to the other species. The observed heterologous interactions, taken together, support the hypothesis that changes in kinase specificity are the driving force in remodeling of the Spo0 pathway (Table 3, Fig 7C). The requirement that Spo0B must interact with both Spo0F and Spo0A keeps the specificity spectra of those proteins in close proximity in interaction space. Thus, small changes in specificity could easily result in heterologous interactions that would be selectively disadvantageous if they occurred in the native environment, such as those we observe with CA_C3319 and Dtox_1918. The intimate proximity of the Spo0F and Spo0A spectra may contribute to the evolutionary flexibility of the Spo0 phosphorelay. The juxtaposition of two distinct architectures controlling a homologous sporulation program in anciently related species raises intriguing questions about the role of pathway remodeling during the evolution of the Spo0 pathway. The prevailing hypothesis is that the ancestral Spo0 pathway was a conventional two-component pathway with the emergence of the Bacillar phosphorelay (Fig 10A) following the separation of the classes, Bacilli and Clostridia. This model predicts that phosphorelays will be observed only in Class Bacilli and direct phosphorylation architecture pathways only in Class Clostridia. Our results challenge this model. Using evidence from the conservation of genomic neighborhoods, we identified homologs of the phosphorelay-specific proteins, Spo0F and Spo0B, in several independent lineages in Class Clostridia. Homologs of Spo0F and Spo0B were also found in all spore-formers in Class Bacilli, except two Erysipelatoclostridium strains that may be in the process of losing the sporulation phenotype (see S1 Text, Section 3). This patchy distribution calls for a reconsideration of the evolutionary history of the sporulation pathway in the Firmicutes. We hypothesize that the ancestral Spo0 pathway was a phosphorelay (Fig 10B) and is the ancestor by vertical descent of all present day Spo0 phosphorelays. The sole genesis of the phosphorelay occurred prior to the divergence of Class Bacilli and Class Clostridia. In this scenario, the present-day direct phosphorylation architectures arose through multiple, independent episodes of pathway remodeling, resulting in a patchy distribution of pathway architectures. Whereas multiple independent inventions of a phosphorelay would also result in a patchy distribution, the complexity of the pathway, coupled with the dramatic similarities between predicted phosphorelay proteins from the two classes, render multiple independent inventions of Spo0 phosphorelays unlikely. Further, when we replaced B. subtilis phosphorelay proteins with their D. acetoxidans counterparts (Fig 6), every component of the Clostridial phosphorelay was able to recapitulate the interactions of the Bacillar phosphorelay such that connectivity was maintained. It is unlikely that these two pathways encode similar specificity by chance; therefore, we conclude that specificity in the Spo0 phosphorelay has been preserved over 2. 7 billion years of independent evolution. If, as predicted by the standing hypothesis, the phosphorelay first arose in a Bacillar ancestor after the divergence of the Bacilli and the Clostridia (Fig 10C), the present-day distribution of pathway architectures could only occur through horizontal transfer of the phosphorelay to Clostridial taxa. However, acquisition of the phosphorelay through horizontal gene transfer entails an improbable series of events. Multiple independent acquisitions through transfer would be required to produce the present-day distribution, because genomes that harbor a phosphorelay are not monophyletic in the Clostridia. Moreover, each acquisition of the phosphorelay would likely require multiple, independent horizontal transfer events, because the genes encoding Spo0 components are dispersed throughout the genome (S6 Fig). Further, because the genomic neighborhoods of Spo0F and Spo0B are conserved, this scenario requires that every transfer of a gene encoding one of these proteins result in insertion into the same neighborhood. Thus, we conclude that the phosphorelay was most likely present in the ancestor of all Firmicutes and all present-day phosphorelays are derived from it by vertical descent (Fig 10B). According to the ancestral phosphorelay hypothesis, present-day direct phosphorylation pathways are a result of multiple, independent transitions, wherein Spo0F and Spo0B were lost and direct phosphorylation of Spo0A was gained. Our results support a scenario in which these transitions arose through changes in or replacement of the kinases. Similarities in the genetic determinants of Spo0A specificity reflect shared taxonomic relationships, not shared pathway architecture, consistent with conservation of the Spo0A specificity spectrum throughout the phylum (Fig 5, see also S2 Text). Kinase specificity residues, in contrast, are most similar within the same architecture, consistent with the hypothesis that changes to sensor kinase specificity, and not Spo0A specificity, are responsible for the change in pathway architecture. Further, phosphorelay and direct phosphorylation kinases harbor HK_CA: 3 and HK_CA: 2 type catalytic domains, respectively (see S3 Text, S6 Table), suggesting that most phosphorelay orphan kinases are distantly related to those of direct phosphorylation architectures. The heterologous phosphotransfer assays (Figs 6–8) also support a history of remodeling through changes in sporulation kinases, and not Spo0A (shown schematically in Fig 10). Here, we highlight several scenarios in which changes in sporulation kinase specificity could result in acquisition of direct phosphorylation of Spo0A. More complex scenarios, for example, involving the interactions between multiple kinases, can also be envisioned. One possibility is that substitutions in an autologous sporulation kinase resulted in a loss of specificity for Spo0F and gain of specificity for Spo0A (Fig 11A). Given the requirement that Spo0F and Spo0A specificity spectra must overlap (since both receivers interact with Spo0B), only a few substitutions may be required. Alternatively, an autologous hybrid histidine kinase, consisting of fused kinase and REC domains, could encode a HisKA domain with pre-existing specificity for Spo0A (Fig 11B), as the REC domain of a hybrid kinase insulates it from interaction with non-cognate receivers [53]. Loss of the REC domain would result in direct phosphorylation of Spo0A. Acquisition, via horizontal transfer, of a novel kinase already possessing specificity for Spo0A would result in immediate remodeling to a direct phosphorylation architecture. This scenario requires that a kinase encoded in a different species be able to phosphorylate the local Spo0A. This could occur if the donor were a non-sporulating species, in which the specificity spectra associated with Spo0A in spore-formers were occupied by the receiver from an unrelated pathway (Fig 11C). The donor could also be a spore-former if the transferred kinase was insulated from Spo0A in its own cell, but within the Spo0A specificity spectrum of the recipient (Fig 11D), due to minor shifts in the specificity spectra. The interactions observed between D. acetoxidans Dtox_1918 and heterologous Spo0A proteins suggest that such shifts in the specificity spectra of Spo0F and Spo0A do occur. The C. acetobutylicum sporulation kinase, CA_C3319, which exhibited affinity for both Spo0A and Spo0F proteins (Fig 9) may be an example of this last scenario. CA_C3319 harbors a HK_CA: 3 type catalytic domain (Agfam [54], see S3 Text), which is commonly observed in phosphorelay sporulation kinases, but not those that directly phosphorylate Spo0A. Further, it possesses unusual specificity residues (SVGLQL) that do not match the typical signatures of either architecture (Table 1; S2 Table). These distinct characteristics suggest that CA_C3319 could be a recently acquired phosphorelay kinase that was specific for Spo0F in the donor cell. Upon acquisition, it may have interacted weakly with Spo0A initially, as there is no Spo0F present in C. acetobutylicum, and subsequently evolved a stronger preference for Spo0A. Our results suggest an evolutionary history wherein remodeling of an ancient phosphorelay resulted in a simpler, direct phosphorylation signal transduction pathway. This is consistent with recent theories of reductive genome evolution, which posit that present-day species with streamlined genomes evolve from gene-rich ancestors via a process of specialization [55,56]. The observation of repeated, independent episodes of pathway remodeling may indicate that the Spo0 pathway has a particular susceptibility for this type of reorganization. The propensity for pathway remodeling may result from juxtaposition of the particular interaction requirements of the Spo0 phosphorelay and the ecological role of the phenotype that it controls. The specificity spectra of Spo0F and Spo0A must intersect to some extent, since both interact with Spo0B. Given their proximity in interaction space, the mutational trajectories required to lose interaction with Spo0F and gain direct interaction with Spo0A may be short. Further, since sporulation is only essential in survival conditions, selection acting on these mutational trajectories may be relatively permissive. Thus, pathway remodeling via substitutions that change interaction specificity may arise easily. A second mechanism of pathway remodeling, by acquisition of a foreign kinase with specificity for Spo0A, may be a byproduct of adaptation to changing environments, since acquisition of novel sensor kinases is a source of novel signal recognition capabilities. The diversity of environmental conditions that induce sporulation in various taxa [57], as well as the diversity of lineage specific sporulation kinase repertoires [35] (see also S2 Table), are both consistent with a process of ongoing, lineage-specific turnover of sporulation kinases. Pathway remodeling via acquisition of novel kinases could also be linked to the loss and recovery of the spore formation phenotype. Sporulation is a metabolically expensive process and is lost frequently in stable conditions [23]. Loss of Spo0F or Spo0B is one scenario that would result in loss of sporulation. If environmental conditions subsequently became less favorable, acquisition of a kinase with specificity to Spo0A would restore sporulation, albeit with a direct phosphorylation architecture. Indeed, several Clostridium sensu stricto species, which likely encode a direct phosphorylation pathway, nevertheless possess a Spo0B-like protein (Fig 3; S2 and S3 Figs), as might be expected in this scenario. Further, we observe that clades harboring direct phosphorylation architectures tend to encode a mix of spore-formers and non-spore-formers (Fig 3; S2 and S3 Figs), which is consistent with the hypothesis that Spo0 pathway remodeling is linked to loss of sporulation. What we have learned about the Spo0 phosphorelay suggests general design principles for signaling pathways in which a single protein must interact with multiple partners and specificity is enforced by molecular recognition. It also provides a perspective on the properties that distinguish the Spo0 phosphorelay from other phosphorelays. Sporulation is initiated by multi-input pathways in which each step in the cascade is encoded in a separate protein, requiring that interaction specificity be controlled entirely by molecular recognition. In most phosphorelays that have been studied, two or more of the four interaction domains are encoded in the same protein, such that interaction specificity is controlled by spatial tethering [53]. Signal transduction based on fused proteins that enforce specificity via spatial tethering may be more robust, but less easily reconfigured or expanded. The differences between spatial tethering and molecular recognition in a phosphorelay could represent different trade-offs between flexibility and constraint. To generate a species phylogeny representative of the Firmicutes phylum, we constructed a maximum likelihood tree using 50 concatenated ribosomal protein sequences from 84 Firmicutes genomes, broadly sampled from the major taxonomic families of the Firmicutes phylum. An initial set of aligned ribosomal protein sequences was obtained from [44]. Profiles were constructed from these multiple sequence alignments and used as queries in HMMer [58] to find ribosomal protein family members in the full complement of 84 genomes. A multiple sequence alignment for each ribosomal protein family was constructed separately using GUIDANCE2 [59] with MAFFT [60] to construct the underlying multiple sequence alignment. Columns possessing at least 50% gaps or a GUIDANCE alignment score below 92% were trimmed from the alignment. Next, the 50 trimmed multiple sequence alignments were concatenated into a supergene alignment. After concatenation, TIGER was used to eliminate uninformative sites [61]. TIGER analysis was performed to group sites into ten bins that are predicted to be evolving at similar rates. The most rapidly evolving sites (Bin_10) were removed, along with columns that were less informative than a randomized site (ptp test with defaults, Bin_Disagreement). The maximum likelihood species tree (Fig 3) was built from the resulting alignment using RaxML version 8. 2 [62] with the CAT model [63], which accounts for site-specific heterogeneity, and bootstrapped with 100 replicates (bootstrap values greater than 50 shown as branch labels in Fig 3). Candidate homologs of Spo0F and Spo0B were identified based on genome neighborhood conservation. Genome neighborhoods were collected from MistDB. com, version 2. 2 [64], which supports protein searches based on RefSeq annotation or domain content (Pfam version 26 [65], Agfam version 1 [54]). Loci up and downstream can also be obtained through the MistDB protein interface. For the purposes of homology identification we refer to the genes encoded four adjacent ORFs up- and downstream as a protein’s genome neighborhood. First, characteristic neighborhood genes for Spo0F were identified from the genome neighborhood of proteins annotated as Spo0F in Refseq, defined as the Spo0F guideset. RefSeq annotations are incomplete and genes within the neighborhood of many of the genes labelled Spo0F lack a RefSeq gene name annotation. Comparison of genome neighborhoods by sequence similarity is preceded by a matching problem (i. e. which genes should be aligned between neighborhoods). Instead, characteristic domain content, which is algorithmically applied and available as a search term through MistDB, can be used to compare genes within a neighborhood. Thus, to identify potential gene markers we analyzed the domain content of genes within the Spo0F guideset neighborhoods and selected the three domains that were the most frequently observed within the neighborhood and least frequently outside of the neighborhood. Genes encoded close to the Spo0F homolog were also favored as this increases the likelihood that the protein will remain in the neighborhood of Spo0F, even in more distantly related species. By these criteria, we identified the following domains as marker domains for Spo0F neighborhoods: F_bp_aldolase (Fructose Bisphosphate Aldolase, PFAM: PF01116), Transaldolase (recently renamed to TAL_FSA in PFAM version 31, PFAM: PF00923), and CTP_Synth_N (N-terminal CTP synthase, PFAM: PF06418). To identify candidate Spo0F homologs outside of the guideset, MistDB searches were performed to identify loci encoding a marker domain (see S3 Table, S4 Fig). The genome neighborhoods for all marker genes were collected. Each potential Spo0F neighborhood was searched for proteins matching the Spo0F criteria: a protein encoding only a single REC domain, taking up 90% or more of the total protein (as measured by amino acid coverage). In many cases, the marker genes were encoded within the genome neighborhood. If two or more ORFs containing domains of interest were separated by no more than four ORFs (regardless of length of interstitial non-coding regions), they were combined into the same neighborhood and searched for proteins matching Spo0F, as well as for presentation in S3 Table and S4 Fig. S3 Table presents all identified marker genes and lists a locus for Spo0F in the same row if it was identified in the neighborhood of that marker gene neighborhood. No genome encodes more than one Spo0F and no genome neighborhood contained more than one protein matching the characteristics of Spo0F. A similar procedure was used to identify candidate Spo0B orthologs (see S4 Table, S5 Fig). Analysis of the neighborhoods of proteins annotated as Spo0B in RefSeq resulted in the selection of three marker genes for the Spo0B neighborhood: GTP1_OBG (PFAM: PF01018), a GTPase domain found on a protein called ObgE in Bacillus subtilis, and ribosomal proteins L21 (Ribosomal_L21p, PFAM: PF00829) and L27 (Ribosomal_L27, PFAM: PF01016). For retrieval from MistDB version 2. 2 (which uses PFAM version 26), candidate Spo0B proteins fit the criteria of no PFAM domains and included a region alignable to the first 50 amino acids of B. subtilis Spo0B. In most cases, each of the genes that the marker domains were associated with were singletons and encoded in the same neighborhood of four ORFs. Further, all candidate Spo0Bs were identified encoded between the proteins encoding Ribosomal_L27 and GTP1_OBG proteins. If two or more ORFs containing domains of interest were separated by no more than four ORFs (regardless of length of interstitial non-coding regions), they were combined into the same neighborhood for presentation in S4 Table and S5 Fig. The specificity residues of candidate Spo0B and sporulation kinase sequences were predicted by manual alignment to the HisKA domains of three Escherichia coli kinases, EnvZ, RstB, and CpxA, for which the specificity residues have previously been determined [3]. The conserved histidine residue that holds the phosphoryl group was used to anchor the alignment. Note that Spo0B specificity residues are likely equivalent to those of HisKA because the Spo0F-Spo0B interaction was instrumental in uncovering interfacial contact residues [66] (see also PDB: 1F51). Similarly, specificity residues for candidate Spo0F and Spo0A sequences were determined by manual alignment to REC domains of response receivers with known specificity residues (OmpR, RstA, and CpxR from E. coli) [4]. The resulting predicted specificity residues for phosphodonors and receivers are given in S5 Table. The Spo0F sequence in Solibacillus silvestris is truncated and was excluded from this analysis. Plasmids encoding Spo0 proteins were constructed for subsequent protein purification and phosphotransfer analysis. Oligonucleotides encoding Desulfotomaculum acetoxidans DSM 771 (CP001720. 1) and Clostridium acetobutylicum ATCC 824 (NC_003030. 1) Spo0 protein sequences were designed with codon usage and GC content optimized for expression in E. coli and synthesized by Genewiz Inc. and Thermo Fisher Scientific GeneArt, respectively. Graphical Codon Usage Analyzer [67] and GeneWiz or Thermo Fisher software were used for coding sequence optimization. Native nucleic acid sequences from the Bacillus subtilis subsp. subtilis str. 168 genome (NC_000964. 3) were used for constructs encoding B. subtilis Spo0 proteins. To increase protein yield and solubility, truncated sequences possessing intact interaction domains were used in three cases: D. acetoxidans kinase Dtox_1918 (residues 301–535), C. acetobutylicum kinase CA_C0903 (residues 244–683, as used in a previous study [24]), and D. acetoxidans Spo0A (Dtox_2041, residues 1–134). The remaining sequences encode the full-length protein (see S7 Table for GenPept accession numbers). The nucleotide sequences used in all expression constructs are provided in supplementary file S1_Sequence. fasta. Expression plasmids were created using the Gateway (Invitrogen) recombinational cloning system, as previously described by Laub et al. [68]. Briefly, the nucleotide sequences described above were cloned into pENTR/D-TOPO entry vectors using the pENTR Directional TOPO Cloning Kit (Thermo Fisher Scientific). The coding sequences were subsequently transferred to destination vectors using the Gateway LR reaction (Thermo Fisher Scientific), yielding expression plasmids encoding affinity-tagged proteins under control of an IPTG-inducible promoter. Three N-terminal affinity tags were used: a hexahistidine sequence followed by a thrombin protease cleavage site (His6-thrombin); a thioredoxin domain followed by a hexahistidine sequence and a TEV cleavage site (TRX-His6-TEV); and a hexahistidine sequence, a maltose-binding protein, and a TEV cleavage site (His6-MBP-TEV). The thioredoxin and maltose-binding domains aid protein folding and stability, leading to higher yield during protein purification. Table 2 gives the affinity tag used in each construct, as well as the size and molecular weight of the resulting fusion protein. See S4 Text for a detailed description of the N-terminal fusion sequences used. The complete amino acid sequences of the 11 fusion proteins are provided in S2_Sequence. fasta. Proteins were expressed and purified as described in Laub et al. [68]. Briefly, constructs in the destination vectors were transformed in E. coli BL21 cells. These cells were grown in LB medium to an OD600 of approximately 0. 6 at 37 °C. Protein expression was induced by the addition of 300 μM IPTG, after which cells were incubated at 30 °C for 4 hours. Cells were harvested by centrifugation, transferred to lysis buffer (20 mM Tris-HCl, pH 7. 9,0. 5 MNaCl, 10% glycerol, 20 mM imidazole, 0. 1%Triton X-100,1 mM PMSF, 1 mg/ml lysozyme), and sonicated. Cleared lysate was obtained by centrifugation at 30,000g for 60 minutes, and was added to 1mL of equilibrated Ni-NTA agarose slurry (Qiagen). Binding was performed at 4°C for 30 minutes. Next, the Ni-NTA agarose slurry was washed twice in wash buffer (20 mM HEPES-KOH, pH 8. 0,0. 5M NaCl, 10% glycerol, 20 mM imidazole, 0. 1% Triton X-100,1 mM PMSF). Tagged proteins were eluted from the slurry using an Econo-column (Bio-Rad) with elution buffer (20 mM HEPE-KOH, pH 8. 0,0. 5MNaCl, 10% glycerol, 250 mM imidazole). Finally, PD-10 columns were used to exchange the purified protein into HKEG buffer (10 mM HEPES-KOH, pH8. 0,50mM KCl, 10% glycerol, 0. 1 mM EDTA, 2mM DTT) and concentrated, as required. Purified proteins and a Novex Pre-stained Protein Ladder were visualized on a 12% SDS-Page gel (7. 5 μL protein, 2. 5 μL 4x LDS Loading Buffer) by staining with Colloidal Blue (all products by Thermo Fisher Scientific). Table 2 summarizes the expected sizes for the proteins used in Figs 4,6, 8 and 9. Phosphotransfer profiling reactions were assayed following the protocols described in Laub et al. [68]. Autophosphorylation was performed at an estimated final concentration of 5 μM kinase in HKEG buffer supplemented with 5 mM MgCl2,500 μM ATP, and 0. 5 μCi/μL [γ32P]-ATP from a stock at ~6000Ci/mmol (Perkin Elmer). Preliminary autophosphorylation experiments with Dtox_1918 demonstrated that peak autophosphorylation is achieved within 15 minutes and is maintained for at least 30 additional minutes (S7 Fig). We further verified that D. acetoxidans Spo0B phosphorylation is only observed when both a sporulation kinase and Spo0F are present, suggesting that it does not undergo autophosphorylation (Fig 4B, lane 3, and S8 Fig). For phosphotransfer profiling of the candidate D. acetoxidans phosphorelay (Fig 4), Dtox_1918 was incubated with [γ32P]-ATP for 15 minutes to allow autophosphorylation. Next, a solution containing radiolabeled Dtox_1918 was split to accommodate three different series of component addition: 1) addition of Spo0F, Spo0B, and Spo0A at 4 minute intervals; 2) addition of Spo0B and Spo0A at 4 minute intervals; 3) addition of Spo0A. A sample was taken 3 minutes after the addition of the next component at each step in the series (Fig 4A). A second sample was taken 10 minutes after the addition of Spo0A in each series (Fig 4B). In all reactions, the estimated final concentrations of the kinase, Spo0F, and Spo0B were 4–6 μM, while the estimated final Spo0A concentration was 10 μM. The reaction for each sample was stopped by the addition of 4X Novex LDS Loading buffer (Life Technologies) and analyzed by 12% SDS-Page gel and phosphorimaging. For the cross species complementation of the B. subtilis phosphorelay with D. acetoxidans phosphorelay components (Fig 6) and the cross-species phosphotransfer profiling experiments (Figs 8 and 9), each kinase was incubated with [γ32P]-ATP for 10 minutes to allow autophosphorylation. Next, a mixture of all other Spo0 components was added together to a final volume of 10 μL. The estimated concentration of histidine kinase in the final mixture was 5 μM; all other components had estimated 10 μM concentrations. After a 5 minute incubation, the reaction for each sample was stopped by the addition of SDS-PAGE loading buffer (500 mM Tris-HCl pH 6. 8,8% SDS, 40% glycerol, 400 mM mercaptoethanol) [68] and analyzed by 12% SDS-Page gel and phosphorimaging.

Survival in a changing world requires signal transduction circuitry that can evolve to sense and respond to new environmental challenges. The Firmicute sporulation initiation (Spo0) pathway is a compelling example of a pathway with a circuit diagram that has changed over the course of evolution. In Clostridium acetobutylicum, a sensor kinase directly activates the master regulator of sporulation, Spo0A. In Bacillus subtilis, Spo0A is activated indirectly via a four-protein phosphorelay. These early observations suggested that the ancestral Spo0A was directly phosphorylated by a kinase in the earliest spore-former and that the Spo0 phosphorelay arose later in Bacilli via gain of additional proteins and interactions. Our analysis, based on a much larger set of genomes, surprisingly reveals phosphorelays, not only in Bacilli, but in many Clostridia. These findings support a model wherein sporulation was initiated by a Spo0 phosphorelay in the ancestral spore-former and the direct phosphorylation Spo0 pathways, which are observed in distinct sets of Clostridial taxa, are the result of convergent, reductive evolution. Further, our evidence suggests that these remodeling events were mediated by changes in kinase specificity, implicating flexible pathway remodeling, potentially combined with the recruitment of kinases, in Spo0 pathway evolution.

Abstract Introduction Results Discussion Materials and methods

bacteriology phosphorylation medicine and health sciences gut bacteria protein interactions pathology and laboratory medicine chemical compounds clostridium pathogens bacillus microbiology marker genes organic compounds bacterial sporulation prokaryotic models experimental organism systems basic amino acids amino acids molecular biology techniques bacteria bacterial pathogens research and analysis methods microbial physiology proteins medical microbiology microbial pathogens chemistry molecular biology biochemistry bacterial physiology histidine organic chemistry post-translational modification bacillus subtilis protein domains biology and life sciences physical sciences organisms

2018

Flexibility and constraint: Evolutionary remodeling of the sporulation initiation pathway in Firmicutes

Dissemination of HIV in the host involves transit of the virus and virus-infected cells across the lymphatic endothelium. HIV may alter lymphatic endothelial permeability to foster dissemination, but the mechanism is largely unexplored. Using a primary human lymphatic endothelial cell model, we found that HIV-1 envelope protein gp120 induced lymphatic hyperpermeability by disturbing the normal function of Robo4, a novel regulator of endothelial permeability. HIV-1 gp120 induced fibronectin expression and integrin α5β1 phosphorylation, which led to the complexing of these three proteins, and their subsequent interaction with Robo4 through its fibronectin type III repeats. Moreover, pretreatment with an active N-terminus fragment of Slit2, a Robo4 agonist, protected lymphatic endothelial cells from HIV-1 gp120-induced hyperpermeability by inhibiting c-Src kinase activation. Our results indicate that targeting Slit2/Robo4 signaling may protect the integrity of the lymphatic barrier and limit the dissemination of HIV in the host. HIV becomes established at mucosal sites by infecting dendritic cells, CD4+ T lymphocytes and macrophages in the lamina propria after its entry. From there, virus and infected cells disseminate via lymphatic endothelial channels to the draining lymph nodes, and subsequently pass into the bloodstream [1]–[4]. An impaired lymphatic barrier may accelerate HIV dissemination. Generally, endothelial cells do not express CD4, the major receptor of HIV, but express varying levels of CXCR4 [5] and CCR5 [6], the co-receptors of HIV, depending on the tissue of origin [7]. While HIV can infect endothelial cells, its biological importance in the pathogenesis of AIDS is unclear [8]–[10]. The HIV-1 envelope glycoprotein gp120 and the HIV transactivator of transcription (Tat) may contribute to HIV-associated vasculopathy. HIV-1 gp120 induces apoptosis in endothelial cells [11], [12] and Tat stimulates angiogenesis [13], [14], which is often concomitant with hyperpermeability. Current knowledge of the effects of HIV-associated hyperpermeability are limited to disrupting the integrity of vascular structures and/or enhancing inflammatory reactions. However, these phenomena are characteristic of many infectious diseases [15] and do not explain the unique biology of HIV. In addition, while a pivotal role for the lymphatic system in the pathogenesis of HIV/AIDS has been suggested [16], the pathobiology of HIV interaction with lymphatic endothelium has not been extensively characterized. The Slit2/Robo4 (Roundabout 4) signaling pathway is a recently identified regulator of endothelial permeability [17]. The Slit/Robo family members were originally discovered as axon guidance molecules that mediate repulsive signaling mechanisms in the central nervous system [18]–[20]. Recent studies from animal models strongly implicate a central role for Slit/Robo in vascular biology [17], [21]. For example, Robo4 knockdown zebrafish embryos have vascular sprouting defects [22], and Robo4 knockout mice display abnormal vascular hyperpermeability [17]. Moreover, Slit2/Robo4 interactions can maintain the integrity of the vascular network and its barrier function by inhibiting cytokine-mediated vasculogenesis and enhanced permeability [17], [23], and the Slit2-Robo4-paxillin-GIT1 network inhibits neovascularization and vascular leakage [24]. Slit2 belongs to a family of three glycosylated extracellular proteins containing at least four different motifs and sharing cognate Robo receptors (Robo1-4) [25], [26]. Slits are secreted by midline glial cells and other tissues [19], [27], [28], and can be processed by proteolytic cleavage to yield a shorter C-terminus fragment of unknown function and a longer, active N-terminus fragment that agonizes the Robos [29], [30]. Robo4 is predominantly expressed in endothelial cells, including embryonic endothelium and tumor vascular endothelium, and shows significant structural differences from the other Robos [26], [31]. Robo4 has only two immunoglobulin (Ig) domains and two fibronectin type III domains in the extracellular region, whereas the other Robos have five and three, respectively [32], [33]. The Robo4 cytoplasmic domain also differs from the other family members, e. g. while Robo1 has four conserved motifs in this region, Robo4 retains only two [26]. Structure-effect studies have revealed that the Slits bind via their N-terminal leucine-rich repeat domain to the Robos, and that the first Ig domain of the Robos is highly conserved and important for Slit binding [34]–[36]. Slit2/Robo4 signaling activates Rho GTPases in endothelial cells, but the precise mechanism by which they interact with each other remains controversial [37]–[39]. There are two prevailing hypotheses for their interaction. One posits that Slit2 activates Robo4 and initiates a signaling cascade [17], [21]. Alternatively, Slit2 may interact with Robo1, and then transactivate Robo4 [39], [40]. In this study, we explored if and how HIV-1 gp120 modulates the Slit2/Robo4 signaling pathway in primary human lung lymphatic endothelial cells. We found that HIV-1 gp120 elevated fibronectin levels, activated fibronectin and α5β1 integrin, and induced a physical association between α5β1 and Robo4. This complexing of Robo4 resulted in hyperpermeability in a lymphatic cell monolayer; however, pretreatment with Slit2N, an active N-terminal fragment of Slit2, inhibited significantly these HIV-1 gp120-induced effects. We suggest that the Slit2/Robo4 pathway may play a key role in modulating HIV-1 gp120-induced lymphatic hyperpermeability, and its manipulation may be used to inhibit the dissemination of HIV in the host. The effects of HIV-1 gp120 on vascular endothelium have been well characterized [41]–[43], however, very little is known about how HIV-1 gp120 specifically affects the lymphatic barrier. To address this issue, we studied the effects of HIV-1 gp120 from two different HIV-1 strains (M-gp120 which utilizes the CCR5 co-receptor on target cells, and T-gp120 which utilizes the CXCR4 co-receptor) on lung lymphatic endothelial cells (L-LECs) in an in vitro, vascular permeability assay. Permeability was quantified by the translocation of FITC-conjugated Dextran particles through an L-LEC cell monolayer seeded in the top chamber of a transwell plate, into the bottom chamber, after incubation with specified concentrations of M-gp120 or T-gp120. We observed a significant increase in permeability of the lymphatic cell monolayer after treatment with both M-gp120 and T-gp120 (Figure 1A). We then assessed in the L-LECs, the expression of CD4 (the major receptor for HIV-1 gp120 on target cells) and the co-receptors, CCR5 and CXCR4, by immunohistochemistry. While we detected no expression of CD4 or CCR5 in these cells (data not shown), we observed a robust expression of CXCR4 on the cell surface and in the nucleus (Figure 1B). However, inhibiting the effects of CXCR4 with a neutralizing antibody had no effect on the HIV-1 gp120-induced permeability of the monolayer (data not shown). These data suggest that HIV-1 gp120 induces hyperpermeability in an L-LEC monolayer by a mechanism independent of CD4, CCR5 and CXCR4 binding. Fibronectin is important for maintaining vascular integrity [44] and is involved in lymphangiogenesis [45], [46]. Previous studies showed that HIV-1 gp120 can bind to fibronectin through its heparin-binding domains, and facilitate HIV infection [47]–[49]. Therefore, we assessed fibronectin expression by Western blot analysis in L-LECs and their supernatant after incubation with various concentrations of HIV-1 gp120 (M-gp120 was used in all experiments unless specifically stated otherwise). We observed marked increases of fibronectin (predominantly as a dimer) in cell lysates after treatment with HIV-1 gp120, and less pronounced increases of soluble, monomeric fibronectin in the supernatant (Figure 2A). We interpret our data to indicate that HIV-1 gp120 can enhance fibronectin expression in lung lymphatic endothelial cells. Interestingly, we observed that low concentrations of gp120 (10–50 ng/ml) induced a decrease in FN secretion (vs. untreated) as compared with higher gp120 concentrations (100–500 ng/ml). Few experimental studies focus on gp120 at such low levels, however, we hypothesize that the effects of gp120 at these low concentrations may be an in vitro correlate for HIV latent infection in vivo and a low viral load, although this has yet to be confirmed. With the recent discovery that Slit2/Robo4 signaling regulates endothelial permeability [17], [24], and our data that demonstrate HIV-1 gp120-induced hyperpermeability and fibronectin up-regulation in L-LECs, we postulated that fibronectin, Slit2 and Robo4 may be interacting to regulate lymphatic permeability after HIV exposure. By confocal microscopy, we observed the expression and localization of fibronectin and Robo4 in L-LECs with or without treatment with Slit2N or HIV-1 gp120. After stimulation with Slit2N, no co-localization of fibronectin and Robo4 was observed (Figure 2B, middle panel). When the L-LECs were treated with HIV-1 gp120, however, fibronectin and Robo4 displayed strong co-localization (Figure 2B, right panel). These expression patterns and interactions were corroborated by a Robo4 immunoprecipitation assay in which L-LECs, stimulated with HIV-1 gp120, showed fibronectin activation (by serine/threonine phosphorylation, Figure 2C) and a significantly enhanced physical association between fibronectin and Robo4 (Figure 2C). Since Slit2/Robo4 signaling is known to inhibit cytokine-induced vascular permeability [17], [23], we compared Slit2 expression in L-LECs in the presence or absence of HIV-1 gp120, a known inducer of endothelial permeability. Using a semi-quantitative RT-PCR assay, we found that at low concentrations, HIV-1 gp120 enhanced Slit2 expression in L-LECs, while higher concentrations of HIV-1 gp120 inhibited the expression of Slit2 (Figure 3). The inhibition of Slit2 by HIV-1 gp120 at 250 ng/ml and 500 ng/ml is consistent with the characterization of Slit2 as an inhibitor of pathological hyperpermeability [24]. Taken together, these data suggest that fibronectin, Robo4 and Slit2 may cooperate in mediating permeability induced by HIV-1 gp120 in lymphatic endothelium. When fibronectin interacts with vascular endothelium it commonly binds to either α5β1 integrin or αvβ3 integrin on the cell surface. Activation of these integrins dramatically enhances this interaction [50]. Therefore, we examined the expression of these two integrins in L-LECs by Western blot analysis. Since we detected α5β1, but not αvβ3 in L-LECs (data not shown), we investigated only α5β1 in subsequent experiments. We treated L-LECs for various times with either HIV-1 gp120 or Slit2N. By Western blotting we measured the levels of β1 phosphorylation, a reflection of α5β1 activation (Figure 4A). We observed no change in α5β1 activation after Slit2N treatment, however, incubation with HIV-1 gp120 induced significant phosphorylation of β1 (Figure 4A). Furthermore, we observed co-localization of HIV-1 gp120 and activated α5β1 integrin on the L-LEC cell surface by confocal microscopy (Figure 4B, “Merge” panel). Based on these results, we hypothesized that the physical interaction between HIV-1 gp120 and integrin α5β1 may play a role in HIV-1 gp120-induced effects. Therefore, we examined the effect of blocking this interaction on lymphatic hyperpermeability. Using the previously described in vitro transwell permeability assay, cells were pre-treated with either a neutralizing anti-α5β1 antibody or an isotype control before incubation with M-gp120 or T-gp120. We observed increased permeability through the L-LEC monolayer after treatment with either of the HIV-1 gp120 isotypes (Figure 4C); pretreatment with the anti-α5β1 antibody prevented much of the increase in permeability associated with HIV-1 gp120 (Figure 4C). These data indicate that the increased activation of α5β1 integrin by HIV-1 gp120 and their physical association are required for HIV-1 gp120-induced hyperpermeability of L-LECs. Based on the results from our expression and co-localization studies of Slit2, Robo4, gp120, fibronectin and α5β1, we sought to investigate further the physical interactions that contribute to HIV-1 gp120-induced hyperpermeability, and to explore the specific effects of Slit2 on these processes. To these ends, we examined the physical interaction of Robo4 and α5β1 in L-LECs after treatment with Slit2N or HIV-1 gp120 in a Robo4 immunoprecipitation assay. The basal association between Robo4 and α5β1 integrin was not affected by the differential expression of Slit2N (Figure 5A). However, we observed a significant increase in this physical association after treatment with HIV-1 gp120 (Figure 5A). We then pretreated L-LECs with Slit2N or a negative control before incubating the cells with HIV-1 gp120. While the association between Robo4 and α5β1 integrin appeared to peak 15 minutes after HIV-1 gp120 incubation (Figure 5B), pretreatment with Slit2N greatly diminished this interaction (Figure 5B). Based on these data, we theorized that Slit2 may antagonize the effects of HIV-1 gp120 on a lymphatic cell monolayer, and therefore, may protect lymphatic endothelium against HIV-1 gp120-induced hyperpermeability. To test this hypothesis we utilized the L-LEC transwell permeability assay previously described. While incubation with M-gp120 and T-gp120 increased L-LEC monolayer permeability (Figure 5C, “Control” bars), the extent of this HIV-1 gp120-induced hyperpermeability was significantly inhibited by pretreatment with Slit2N (Figure 5C, “Slit2N” bars). We interpret these data to indicate that Slit2N significantly inhibits HIV-1 gp120-induced hyperpermeability in lymphatic endothelium by blocking the physical association between Robo4 and α5β1 integrin. To demonstrate that Slit2N and gp120 can induce similar effects in various types of lymphatic endothelium, we repeated the lymphatic permeability assay using primary human dermal lymphatic endothelial cells (D-LECs). Similar to the results using L-LECs, gp120 increased the permeability of D-LEC monolayers in a dose-dependent manner, and pretreatment with Slit2N significantly decreased the gp120-induced hyperpermeability (Figure 6A). To confirm that the changes in permeability were not due to the origin of the gp120, we repeated this experiment with another M-tropic gp120 protein, gp120CM, and observed similar effects (data not shown). To demonstrate that the effects of the gp120 protein reflect accurately those of intact HIV-1 virions on lymphatic hyperpermeability, we pretreated L-LEC monolayers with Slit2N or a negative control, followed by incubation with HIV-1 virions or gp120. We found that HIV-1 virions significantly increased lymphatic permeability within 5 hours, whereas gp120 induced only a mild increase during the same time period (overnight incubation was needed for full in vitro effect of gp120 on permeability) (Figure 6B). Pretreatment with Slit2N significantly inhibited the permeability induced by both the HIV-1 virions and the gp120 protein (Figure 6B). Taken together, our results indicate that intact HIV-1 virions increase lymphatic monolayer permeability, and preincubation with Slit2N can effectively inhibit this increase. These data indicate that HIV-1 virions can induce lymphatic endothelial monolayer permeability similar to that induced by gp120, suggesting that our in vitro model of gp120-induced endothelial cell monolayer permeability may reflect the actions of HIV-1 in vivo. To elucidate the signaling cascade (s) responsible for HIV-1 gp120-induced hyperpermeability in lymphatic endothelium, we analyzed the effects of HIV-1 gp120, Slit2N and Robo4 by Western blot analysis on Src kinase, a key molecule in the regulation of vascular endothelial permeability [51], [52]. We found that preincubation of L-LECs with Slit2N significantly inhibited HIV-1 gp120-induced phosphorylation of c-Src (Figure 7A), indicating inhibition of the Src signaling pathway. We theorized that the modulation of Src kinase signaling by Slit2N and Robo4 may be the result of a physical complexing between the two proteins. To test this hypothesis, we transiently expressed both Robo4 and Myc-tagged Slit2 in 293 cells, and examined their physical association in a Robo4 immunoprecipitation assay. We observed a physical association between Slit2 (c-Myc) and Robo4 in these cells (Figure 7B). Since pretreatment with Slit2N inhibited c-Src signaling and there appeared to be a physical association between Slit2 and Robo4, we asked if the inhibition of c-Src signaling was a result of Slit2 sequestering Robo4 to deplete its cellular levels and render it unavailable for binding to a competing protein. To approximate this situation, we pretreated L-LECs with a mixture of Robo4 siRNAs or a control siRNA before incubating the cells with HIV-1 gp120. We did not observe the same inhibition of c-Src activation as we had with the Slit2N preincubation (Figure 7A). Instead, the constitutive activation of c-Src increased dramatically in the Robo4 knockdown cells as compared with the control siRNA-transfected cells (Figure 7C). These findings are consistent with the phenotype of Robo4 knockout mice which display heightened vascular permeability [17]. These data suggest that a sufficient endogenous level of Robo4 in lymphatic endothelium is necessary to block c-Src signaling, and that its binding to Slit2 is required to protect against lymphatic hyperpermeability. Additionally, HIV-1 gp120 did not enhance c-Src signaling in the Robo4 knockdown cells as it did in the control siRNA-transfected cells (Figure 7C). We hypothesize that the elevated constitutive level of c-Src kinase signaling in the Robo4 knockdown cells prevented HIV-1 gp120 from enhancing this effect in the L-LECs. To determine if Src signaling is involved in HIV-1 gp120-induced lymphatic permeability, we pretreated L-LECs with a Src kinase inhibitor or a DMSO control before measuring HIV-1 gp120-induced permeability, as described previously. While treatment with HIV-1 gp120 resulted in increased permeability through the L-LEC monolayer preincubated with DMSO, HIV-1 gp120 had no effect on the L-LECs preincubated with a Src kinase inhibitor (Figure 7D). Taken together, we interpret these data to indicate that Src kinase signaling is required for HIV-1 gp120-induced lymphatic hyperpermeability, and that Slit2/Robo4 interactions can inhibit this signaling cascade. To characterize more precisely the role of Robo4 in HIV-1 gp120-induced effects on lymphatic permeability, we transfected L-LECs with control siRNAs or Robo4-specific siRNAs (to reduce Robo4 levels), and confirmed a decrease in Robo4 expression by Western blot analysis 24 hours later (Figure 8A). We compared the permeability of L-LEC monolayers expressing endogenous levels of Robo4 (Figure 8B, “Control siRNA” columns) and reduced levels of Robo4 (Figure 8B, “Robo4 siRNA” columns) in the presence or absence of Slit2N or HIV-1 gp120. In the L-LEC monolayers with endogenous levels of Robo4, incubation with Slit2N had no significant effect on permeability, but HIV-1 gp120 significantly increased the permeability of this monolayer. We observed a significantly higher basal level of permeability in the L-LEC monolayers with reduced Robo4 levels. Slit2N had no significant effect on the permeability of these monolayers, and HIV-1 gp120 failed to cause any significant change in the permeability of the L-LEC monolayers with reduced Robo4 levels. We hypothesize that HIV-1 gp120 did not enhance the permeability of these monolayers, because reducing Robo4 levels had already markedly increased their permeability. These data suggest that sufficient endogenous levels of Robo4 are required to maintain an intact lymphatic barrier. The Robo4 receptor contains two fibronectin (FN) type III domains in its extracellular region [32], [33]. While a study by Kaur et al. , demonstrated that these motifs are important for the interaction of Robo4 with fibronectin [38], no other function for the domains has been documented. Fibronectin regulates the permeability of vascular endothelium [44]. We observed that HIV-1 gp120 elevated FN levels and enhanced lymphatic monolayer permeability in L-LECs. Therefore, we examined the effects of fibronectin on c-Src activation, and its effects after pretreatment with Slit2N. We observed that fibronectin enhanced the activation of c-Src, and that pretreatment with Slit2N significantly inhibited the FN-induced activation of c-Src (Figure 9A). We hypothesize that Slit2N may be interacting with Robo4 to block the FN-induced c-Src activation, and that the FN domains of Robo4 may be involved in the inhibition of FN-induced c-Src activation by Slit2 and L-LEC monolayer hyperpermeability. To explore the potential role of the Robo4 FN domains in HIV-1 gp120-induced effects, we compared the effects of HIV-1 gp120 on the permeability of L-LEC monolayers transfected with wild-type Robo4 (WT), mutant Robo4 (MT), which lacks the FN type III domains, or a vector control (V). We found that HIV-1 gp120 induced significantly less permeability in L-LEC monolayers with elevated levels of wild-type Robo4 as compared to those with endogenous Robo4 levels (Figure 9B). These results indicate that Robo4 inhibits HIV-1 gp120-induced permeability in L-LEC monolayers, and may protect the integrity of the lymphatic barrier after HIV-1 infection by interacting with Slit2. We also observed that HIV-1 gp120-induced permeability was inhibited to a significantly greater extent in the L-LEC monolayers transfected with mutant Robo4 vs. wild-type Robo4. In fact, treatment with HIV-1 gp120 resulted in no change in the permeability of the L-LEC monolayers expressing mutant Robo4 (Figure 9B). We interpret these results to indicate that the complexing of FN and Robo4 (through its FN type III domains) is necessary for HIV-1 gp120-induced hyperpermeability of L-LEC monolayers, and that the FN type III domains of Robo4 may be required for the interaction of HIV-1 gp120, Robo4 and FN. To explore this hypothesis, we transiently transfected L-LECs with plasmids encoding wild-type Robo4 (WT), mutant Robo4 (MT), or a vector control (V). After 48 hours, we analyzed the effects of HIV-1 gp120 on c-Src pathway activation in each of the transfected cell types by Western blot analysis. We observed that the basal level of c-Src activation was lower in L-LECs with elevated Robo4 expression as compared to those with endogenous Robo4 expression (Figure 9C, WT/− and V/−, respectively). HIV-1 gp120 increased c-Src activation in both cell types (Figure 9C), however, overall HIV-1 gp120-induced c-Src activation levels were significantly lower in L-LECs with elevated Robo4 levels as compared to those with endogenous Robo4 levels (Figure 9C, WT/+ and V/+, respectively). In L-LECs expressing elevated levels of mutant Robo4 (MT), both basal c-Src activation and HIV-1 gp120-induced c-Src activation were equivalent to the L-LECs expressing elevated wild-type Robo4 (Figure 9C). These data indicate that elevated levels of Robo4 inhibit basal c-Src activation and HIV-1 gp120-induced c-Src activation. We hypothesize that since Slit2 inhibits c-Src activation, elevated Robo4 levels after transfection may magnify the effects of Slit2, by providing more receptors to which endogenous Slit2 can bind. We also examined the levels of HIV-1 gp120-induced phosphorylation of ERK1/2, key signaling molecules for endothelial cell function, by Western blot analysis, using the same three groups of L-LEC transfectants. HIV-1 gp120 induced a significant increase in ERK1/2 phosphorylation in the L-LECs transfectants with elevated wild-type Robo4 expression as compared to those with endogenous Robo4 expression (Figure 9C). Although HIV-1 gp120 increased the phosphorylation of ERK1/2 in the L-LECs transfected with mutant Robo4, the increase was significantly lower than the wild-type Robo4 transfectants (Figure 9C). These data indicate that while the FN domains of Robo4 are not required for the inhibition of gp120-induced c-Src activation, they are required for gp120-induced phosphorylation of ERK1/2. We hypothesize that the FN domains of Robo4 may participate in the activation of other key signaling molecules like ERK1/2, however, further investigation is needed to fully understand their function. The integrity of the lymphatic barrier requires a dynamic interaction between fibronectin, other extracellular matrix (ECM) proteins, and their receptors, cell-surface integrins [46], [53]. As a result of HIV-1-induced inflammation and increased protease expression, fibronectin fragments are detected in the blood of HIV-infected patients [54]–[56]. These fragments are believed to promote the transendothelial migration of HIV-1-infected and non-infected leukocytes, and to promote viral stability and cell-to-cell transmission [48], [56], [57]. We found that lymphatic endothelial cells produce elevated levels of cell-bound fibronectin after exposure to HIV-1 gp120 (Figure 2A). This elevation appeared to modulate the integrity of the lymphatic barrier. In particular, HIV-1 gp120 induced activation of α5β1 integrin which enhanced the physical complexing of HIV-1 gp120, fibronectin, α5β1 integrin and Robo4, and resulted in lymphatic hyperpermeability (Figures 1A, 2B, 2C, 4A and 4B). While FN/integrin [45] and Slit2/Robo4 [17] interactions are both important for endothelial permeability, little is known about their relationship. α5β1 and αvβ3 are two major integrins expressed on the surface of endothelial cells [45]. We and others have shown that integrin α5β1, but not αvβ3, clustered in focal contacts of endothelial cells during stressful cellular conditions (Figure 4B) or incubation with fibronectin [58]–[60]. In this study, we found that Robo4 formed a complex with fibronectin and integrin α5β1 at low, basal levels in uninfected lymphatic endothelial cells (Figures 2B, 5A and 5B). Slit2N did not alter this association, which is important for maintaining the integrity of the lymphatic barrier (Figures 5A and 5C). However, exposure to HIV-1 gp120 enhanced the association of α5β1 and Robo4 (Figure 5A) and resulted in increased lymphatic permeability (Figures 1A and 5C). Moreover, pre-incubation with Slit2N blocked the HIV-1 gp120-induced enhanced complexing of Robo4 and α5β1, (Figure 5B) and lymphatic hyperpermeability was reduced (Figure 5C). These data suggest that α5β1 integrin may also participate in the effects of HIV-1 gp120 on lymphatic permeability, and Slit2 may help sustain the integrity of the lymphatic barrier after HIV-1 exposure. Activation of the Src kinases modulates cytoskeletal remodeling and affects cell-to-cell and cell-to-ECM adhesion [61], [62]. Our data indicate that HIV-1 gp120 and Slit2 exert opposing effects on c-Src kinase signaling, namely, HIV-1 gp120 activates c-Src signaling, while pretreatment with Slit2N significantly reduces these effects (Figure 7A). Moreover, the enhancement or inhibition of Src kinase signaling and the resulting effect on lymphatic permeability is critically dependent on Robo4 levels (Figures 9A, 9B, and 9C). Robo4 displays unique structure and function, but the relationship between these characteristics is largely unknown [26], [32], [33]. Although the first Ig domains of the Robos are highly conserved and important for Slit binding, direct binding of Slit2 to Robo4 is still debated [37]–[39]. The Robo proteins, including Robo4, contain fibronectin type III domains. Previous studies, which found that the FN domains were required for adhesion to fibronectin, suggest that these domains may play a central role in modulating vascular permeability [38]. Our data strongly support their function in Robo4-mediated lymphatic permeability upon HIV-1 gp120 stimulation, and imply a potential role for Robo4 in fibronectin-associated vasculopathies, such as HIV-associated pulmonary hypertension [63]. Based on our new data and that of others, we propose a hypothetical model for the interactions of HIV-1 gp120, FN, α5β1 integrin, Robo4 and Slit2, and their effect on lymphatic permeability (Figure 10). Robo4 and α5β1 integrin are transmembrane proteins expressed in lymphatic endothelium. We hypothesize that under normal, physiological conditions, soluble FN and Slit2 are expressed at low, basal levels, and they interact with Robo4 via its FN type III domains and Ig domains, respectively. FN also binds to α5β1 integrin, which is expressed on the endothelial cell surface. Under these conditions, the integrity of the lymphatic endothelial barrier is intact, and transmigration through the endothelial barrier is severely restricted. We propose that upon HIV infection, HIV-1 gp120 elevates FN levels significantly and complexes with FN. FN then activates α5β1 integrin, which results in enhanced intracellular signaling through α5β1 integrin, a significantly stronger interaction between FN and Robo4, and enhanced intracellular signaling through Robo4. These changes activate the c-Src signaling pathway and induce hyperpermeability of the lymphatic endothelial barrier. The resulting “leaky” barrier may facilitate the dissemination of HIV-1 and virus-infected cells throughout the body. Furthermore, we propose that elevated levels of Slit2 may protect the lymphatic channels from HIV-induced vasculopathy and HIV spread. We hypothesize that at sufficiently elevated levels, Slit2 will bind strongly to the Ig domains of Robo4 and inhibit c-Src pathway activation and HIV-1 gp120-induced lymphatic hyperpermeability. Slit2 may affect this inhibition by various means. A likely senario is that upon binding, Slit2 alters the protein conformation of Robo4, which may lessen/abolish its ability to interact with FN, alter α5β1 integrin intracellular signaling, and inhibit the activation of c-Src. In addition, the binding of Slit2 may alter also the signaling through Robo4, which may inhibit c-Src pathway activation and lymphatic hyperpermeability. Although our data strongly support this model, further investigation is needed to confirm it, or posit alternative mechanisms for the effects of HIV-1 gp120, FN, α5β1 integrin, Robo4 and Slit2 on lymphatic permeability. Multiple studies indicate that the lymphatic channels play important roles in the establishment of HIV infection, and its dissemination throughout the host [1]–[4]. HIV-induced lymphadenopathy, including lymphoedema, is commonly seen among HIV-infected individuals with Kaposi' s sarcoma, a vascular neoplasm which is derived from lymphatic endothelial cells [64], [65]; however, dysfunction of the lymphatic vasculature and its effects on HIV biology are largely unexplored. We established an in vitro endothelial monolayer model to study the effects of HIV on lymphatic permeability. In this model, HIV-1 gp120 and HIV-1 virions both induced lymphatic hyperpermeability, which was significantly inhibited by Slit2 preincubation (Figure 6B). These results suggest key roles for gp120, FN, and Slit2/Robo4 in HIV-associated lymphatic hyperpermeability, and implicate lymphatic hyperpermeability in HIV infection and spread throughout the body. Future studies to explore the traversion of HIV virions or virus-infected cells through the lymphatic endothelium and its contribution to HIV infection should provide more evidence on HIV-induced lymphatic hyperpermeability and HIV dissemination in a humanized mouse model of HIV infection. In summary, we found that the balance between HIV-1 gp120/FN/α5β1 integrin-induced signaling and Slit2/Robo4-induced signaling in L-LECs modulates lymphatic monolayer permeability. Targeting these pathways may offer novel approaches to inhibit HIV-induced lymphatic injury, and limit the dissemination of HIV in the host. Human embryonic kidney cells (293 cells) (Stratagene, La Jolla, CA, USA) were cultured in Dulbecco' s modified Eagle' s medium with 10% fetal calf serum. Primary human lung lymphatic endothelial cells (L-LECs) and dermal lymphatic endothelial cells (D-LECs) were purchased from Lonza, Inc. (Allendale, NJ, USA) and maintained in EBM-2 medium with EGM-2MV SingleQuots (Lonza, Inc.). Recombinant human Slit2N (the active fragment of Slit2) was provided by Dr. Dean Li, Department of Oncological Sciences at the University of Utah. The following reagents were obtained through the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH: recombinant HIV-1Ba-L gp120 (M-gp120) protein and HIV-1 virus (strain HIV-1 Ba-L, which was from Dr. Suzanne Gartner, Dr. Mikulas Popovic and Dr. Robert Gallo). Per NIH data sheet, HIV-1Ba-L was originally isolated from a primary culture of adherent, human, infant lung tissue cells, and amplified in human monocytes/macrophages. Virus was harvested 10 days post-infection. This virus was used in the transendothelial monolayer permeability assays. HIV-1 gp120 LAV (III B) (T-gp120) was purchased from Protein Sciences Corporation (Meriden, CT, USA). HIV-1 gp120CM was purchased from ProSpec-Tany TechnoGene Ltd. (Ness Ziona, Israel). Src inhibitor-1 was purchased from Sigma-Aldrich, Corp. (St. Louis, MO, USA). Mouse monoclonal antibody to the HIV-1 gp120 was purchased from Advanced Biotechnologies Inc. (Columbia, MD, USA). Anti-integrin β1, anti-phospho-integrin β1 (Tyr783), and neutralizing anti-integrin α5β1 antibodies were purchased from Millipore Corp. (Billerica, MA, USA). Anti-phospho-Src kinase family antibodies were purchased from Cell Signaling Technology, Inc. (Beverly, MA, USA). All other antibodies were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA, USA). Total cell RNA was extracted using the RNeasy Mini Kit from Qiagen, Inc. (Valencia, CA, USA). RT-PCR was performed using a one-step RT-PCR kit from Clontech (Mountain View, CA, USA). Specific primers for CXCR4 and CCR5 were purchased from R & D Systems (Minneapolis, MN, USA). The primers for human Slit2 were synthesized by Invitrogen Corp. (Carlsbad, CA, USA). The sequences are: upstream: 5′-GGTGTCCTCTGTGATGAAGAG -3′; downstream: 5′- GTGTTTAGGACACACACCTCG -3′. Cells cultured in 8-well chamber slides (Thermo Fisher Scientific Inc. , Waltham, MA, USA) were fixed with 4% (v/v) paraformaldehyde solution for 1 hour at room temperature, incubated in fresh permeabilization solution (0. 1% sodium citrate in 1% Triton X-100 in 1× PBS) for 2 minutes on ice, incubated with 3% BSA/1× PBS on ice for 30 minutes, and then with anti-human CXCR4 rabbit polyclonal antibody or normal rabbit IgG (Millipore Corp.) at 4°C for 1 hour. The slides were then washed 3 times in 1× PBS, and incubated with a FITC-conjugated, goat anti-rabbit IgG antibody (Vector Laboratories, Burlington, CA, USA) at 4 °C for 30 minutes. The slides were washed again 3 times in 1× PBS, and then air dried and mounted with mounting medium (Vector Laboratories). Lymphatic endothelial cells were seeded in the top chamber of transwell plates, according to the manufacturer' s instructions (Millipore Corp.), starved for one hour, and then incubated with different reagents or their respective controls as indicated. Subsequently, FITC-Dextran was added to the top chamber and allowed to permeate through the monolayer to the lower chamber for 5 minutes. The extent of permeability was determined by measuring the fluorescence of the solution in the lower chamber by a standard plate reader (BioTek Instruments, Inc. , Vinooski, VT, USA). The gp120 control was prepared by boiling gp120 for 10 minutes to inactivate its protein activity while preserving its inherent endotoxin activity. It was employed here, and in all other experiments that required a gp120 control. Cells were starved for 2 hours in serum-free media, and then stimulated as indicated. Cells were lysed in RIPA buffer (Cell Signaling Technology, Inc.) after stimulation. Immunoprecipitation and Western blotting were performed as described previously [66]. For quantitation, the ratio of protein expression, phosphorylation, or association vs total protein in each lane was obtained by densitometry with a gel imaging system (Cell Biosciences, Inc. , Santa Clara, CA, USA). The pCMV6 Entry expression plasmid encoding Myc-DDK-tagged Slit2 was purchased from OriGene Technologies, Inc. (Rockville, MD, USA). The expression plasmid encoding RFP-tagged Robo4 was constructed as follows. Robo4 cDNA was amplified from the pCMV-SPORT6 containing Robo4 cDNA (Thermo Fisher Scientific Inc.), using primers purchased from Invitrogen Corp. (upstream sequence: 5′-GAGGCGATCGCATGGGCTCTGGAGACAGCCTCCTG-3′; downstream sequence: 5′-GCGACGCGTGGAGTAATCTACAGGAGAAGCACCAGC-3′). The purified PCR product was digested with Sgf I plus Mlu I, and inserted into the pCMV6-AC-RFP plasmid digested with same restriction endonucleases to create the pCMV6-AC-RFP-Robo4. To make the mutant Robo4 expression plasmid, we designed a pair of primers to amplify a section of the pCMV6-AC-RFP-Robo4 plasmid by PCR. The primers are: 5′-CCCCCCCCGCTAGCTCTAGGCTTGGGGCCCTCTGCAGGATC-3′ and 5′-TTTTTTTTGCTAGCCCTGTCTGCCTCCTTTTAGAGCAGGCC-3′. The PCR product was digested with Nhe1, purified, and ligated with T4 DNA ligase at 16°C overnight. The ligation product was used to transform competent DH5 α cells. Positive clones were screened and confirmed by DNA sequencing. Specific Robo4 siRNAs and control siRNAs, purchased from Santa Cruz Biotechnology, Inc. , were used to transfect L-LECs using HiPerFect transfection reagent from Qiagen, Inc. Cells were grown to 60% confluence in tissue culture dishes. Transfections were done using Super Effectene transfection reagent according to the manufacturer' s instructions (Qiagen, Inc.). At 3 hours post-transfection, cells were washed once with 1× PBS, then cultured in full medium for 48 hours. The transfection efficiency was determined by detection of red fluorescent cells under a fluorescent microscope (Nikon Diaphot 300, Tokyo, Japan). Cells cultured in 8-well chamber slides (Thermo Fisher Scientific Inc.) were serum starved for 2 hours, and then treated with HIV-1 gp120 as indicated. Subsequently, cells were fixed with 4% (v/v) paraformaldehyde for at least 1 hour at room temperature and permeabilized for 2 minutes on ice. Cells were then incubated with primary antibodies or normal IgG controls overnight at 4°C, and washed 3 times with 1× PBS. Fluorescence-conjugated secondary antibodies were added for 30 minutes at 4°C and the cells were washed 3 times in 1× PBS. Finally, the chambers were removed and coverslips were affixed with mounting medium containing DAPI (Vector Laboratories). Slides were examined under a Leica TCS-NT laser scanning confocal microscope (Leica Microsystems, Bannockburn, IL, USA). Each experiment was repeated at least 3 times, and representative blots, images, or graphs are shown in the figures. Statistical significance was determined using the ANOVA test (*p<0. 05).

The most common route of HIV transmission is through unprotected sexual contact. By this route, HIV first infects cells in the mucous membranes of the mouth, vagina or rectum. From the mucosa, virus and virus-infected cells move through lymphatic endothelial channels to draining lymph nodes where they infect various cells, including their major target cells, CD4+ T lymphocytes. The virus and infected cells then transmigrate through the lymphatic barrier, enter the blood stream, and spread throughout the body. We found that HIV-1 gp120 compromises the lymphatic endothelial barrier by inducing hyperpermeability. We hypothesize that an impaired barrier may facilitate the dissemination of HIV. Likewise maintaining a “normal” barrier may help slow the dispersal of HIV, thereby protecting the body from HIV spread and progression after initial mucosal exposure. We demonstrated that in lymphatic endothelium the interactions among Robo4, Slit2, fibronectin and α5β1 integrin modulate the effect of HIV-1 gp120 on lymphatic permeability. Moreover, we found that Slit2 inhibits the complexing of Robo4 with fibronectin and protects cells from gp120-induced hyperpermeability. These data suggest that by interacting with Robo4, Slit2 may help maintain the integrity of the lymphatic barrier, thereby interfering with the dissemination of HIV beyond the draining lymph nodes.

Abstract Introduction Results Discussion Materials and Methods

medicine infectious diseases hiv vascular biology viral diseases cardiovascular

2012

Slit2/Robo4 Signaling Modulates HIV-1 gp120-Induced Lymphatic Hyperpermeability

Ubiquitination relies on a subtle balance between selectivity and promiscuity achieved through specific interactions between ubiquitin-conjugating enzymes (E2s) and ubiquitin ligases (E3s). Here, we report how a single aspartic to glutamic acid substitution acts as a dynamic switch to tip the selectivity balance of human E2s for interaction toward E3 RING-finger domains. By combining molecular dynamic simulations, experimental yeast-two-hybrid screen of E2-E3 (RING) interactions and mutagenesis, we reveal how the dynamics of an internal salt-bridge network at the rim of the E2-E3 interaction surface controls the balance between an “open”, binding competent, and a “closed”, binding incompetent state. The molecular dynamic simulations shed light on the fine mechanism of this molecular switch and allowed us to identify its components, namely an aspartate/glutamate pair, a lysine acting as the central switch and a remote aspartate. Perturbations of single residues in this network, both inside and outside the interaction surface, are sufficient to switch the global E2 interaction selectivity as demonstrated experimentally. Taken together, our results indicate a new mechanism to control E2-E3 interaction selectivity at an atomic level, highlighting how minimal changes in amino acid side-chain affecting the dynamics of intramolecular salt-bridges can be crucial for protein-protein interactions. These findings indicate that the widely accepted sequence-structure-function paradigm should be extended to sequence-structure-dynamics-function relationship and open new possibilities for control and fine-tuning of protein interaction selectivity. Biological systems critically rely on selective and specific protein interactions, creating building blocks that cooperatively form the basis of functional complexity [1]. In particular, regulatory and signaling pathways, such as conjugation of ubiquitin (Ub) or similar ubiquitin-like modifiers (UBLs), depend on specific recognition of binding partners whilst discriminating against non-specific interactions [2], [3]. Ub becomes conjugated to substrates through the action of the E1-E2-E3 enzymatic system. This cascade follows a pyramidal hierarchy wherein two human Ub-activating enzymes (E1s) and over 30 human Ub-conjugating (E2) enzymes and hundreds of E3 Ub protein ligases cooperate to catalyze subsequent substrate modification [4], [5]. Whereas the Ub-E1 enzymes have different selectivity [6], [7], it is mainly the selective associations between E2 and E3 enzymes among thousands of possible interactions that are primarily responsible for effective Ub conjugation [2], [3], [8], [9]. The molecular basis for this selectivity is provided by surface residues of E2 enzymes and by “cross-braced”, zinc-binding domains of the Really Interesting New Gene (RING) sub-family of E3 ligases [2]. These RING finger domains generally mediate binding with E2s through their highly conserved UBC-fold, although exceptions have been described [8], [10]. This fold is a strictly conserved structure surrounding the active site cysteine that covalently accommodates the activated Ub [8], [11], [12]. Regarding the E2 enzymes, the major structural determinants for E3 binding have been identified in residues located in helix 1 (H1), loop 1 (L1) and loop 2 (L2) within the classical E2-E3 interaction interface [10]. Efficient ubiquitin-conjugation depends on specific interactions between E2 and E3 enzymes. The high degree of sequence and structure conservation in E2s and, to some extent, among the E3 RING-finger domains and their reported E2-E3 interfaces, remains compatible with a highly selective binding for their cognate E2 or E3 partners [9], [13]. Conserved sub-families of E2 enzymes, such as those belonging to the UbcH5 branch, have been shown to cooperate with similar sub-sets of E3 ligases [13]. At the other side, structurally more divergent E2s display interactions with more specialized E3 enzymes, such as for example the APC/C-UbcH10 pair [14]. Furthermore, several studies have indicated that the ability of E2-E3 pairs to mediate biochemical Ub- or UBL-conjugation is directly dependent on the ability of physical E2-E3 interactions [15], [16]. Based on this, physical E2-E3 interactions can be regarded as primary determinants for efficient conjugation reactions. We demonstrate here that a subtle and minute change – an aspartic to glutamic acid substitution (one methylene group difference) – is sufficient to completely change the selectivity profile of an E2 enzyme toward E3 RING-finger domains. Using molecular modeling and molecular dynamics simulations, a network of intra-molecular salt-bridges was identified that controls the balance between a binding-competent and a binding-incompetent state. Perturbation of network components located both in and outside the classical interaction surface resulted in a switch in the E3 interaction binding profile. These results suggest a new and delicate mechanism of how protein-protein interaction selectivity is achieved within the promiscuous ubiquitination system. The hypothesis of a side-chain equilibrium controlling UbcH8 and UbcH6 E2 interactions with E3 RING-finger domains was tested experimentally by perturbing key residues involved in the intra-molecular salt-bridge network. First, the significance of K117 was assessed in UbcH8, since K117 was predicted to play a central role in alternatively contacting D113 and D145, thereby, preventing D113 of interacting with the RING domain. By substituting UbcH8 K117 with a histidine, the E3 interaction pattern of UbcH8 was reverted to that of UbcH6 (Figure 4A). Histidine was chosen in order to maintain the total charge of the system. Note that, since the intracellular pH of yeast has been found to be around 5. 5 in most growth phases, it can be expected that the solvent-exposed histidine side-chains are protonated [20]. The result of this substitution is that the positively charged K117 side-chain is replaced by the shorter imidazole side-chain of histidine. This conservative mutation, K117H, does no longer allow a favorable salt-bridge formation with D113 due to the more rigid histidine side-chain, leaving the latter free to interact with the RING domain. This indeed yielded an interaction pattern that resembles that of UbcH6. In contrast, UbcH8 K117R preserves the UbcH8 interaction profile because arginine, which resembles lysine in terms of length and charge, is still able to bridge D113, shielding it away from interacting with the E3 (Figure 4A). These experimentally validated observations confirm the central role of K117 of UbcH8 as a regulator of the orientation of D113 and D145 and thereby as a regulator of RING interactions. We next investigated the role of D145, the most remote (as seen from the E3 interaction site) component of the network. Interaction screening of an UbcH8 D145K mutant demonstrates that D145K interacts in a similar manner as UbcH8 WT (Figure 4A). The introduced charge repulsion between D145K and K113 reinforces the bridging between K117 and D113, locking the latter into a “closed” salt-bridged conformation preventing RING-contacts. In contrast, the introduction of D145E in UbcH8 alters the interaction profile of UbcH8 WT to that of UbcH6 WT. Here, the longer glutamine side-chain can more effectively contact K117, forming a stable salt-bridge, and consequently freeing D113 for forming contacts with RING. To ascertain that this indeed involves D113, the double mutant UbcH8 D113A/D145E was also characterized, showing the loss of the previously gained E3 interactions (Figure 4A). These results also indicate that both ASP and GLU at position 113 are binding competent, provided they are not stabilized in their closed conformation by intra-molecular salt-bridge formation, highlighting again the key role of this salt-bridge network. Taken together, these results point toward a delicate network of intra-molecular salt-bridges that dynamically control the positioning of a crucial E3-interacting residue (schematically represented in Figure 4B). Residues located in the classical E2-E3 interface, as well as those that are more distantly located dynamically control the positioning of K117 that acts as a gatekeeper for the negatively charged hot-spot residue capable of mediating E3 interactions. While pH-dependent [21] or phosphorylation [22] regulatory mechanisms have been described previously, the subtle equilibrium identified here does not involve any change in net charge, but originates from a minimal difference of a single methylene group along a side-chain. Although the catalytic UBC-fold of E2 enzymes is characterized by high levels of sequence and structural similarities, it is adapted to selectively recognize RING-finger domains to allow transfer of Ub to substrates [2], [23]. By comparing sequence and structure of two highly similar E2 enzymes with their global RING-finger interaction patterns, in combination with molecular dynamics simulations, we were able to identify an intra-molecular network of salt-bridges that actively control RING interactions. Perturbation of components of this dynamic network in combination with Y2H screening of interactions subsequently confirmed its crucial role in controlling the interaction specificity. Residue K109/K117 was identified to play a crucial role within this network. It binds in an exclusive manner either E105/D113 or D137/D145. This lysine appears to be strictly conserved among E2 enzymes. In UbcH5B, for example, the corresponding lysine residue at this position is the central K63, which has been intimately linked to UbcH5B selectivity within the UbcH5B-CNOT4 RING-interaction [15], [24]. Interestingly, UbcH5B K63E failed to interact with the WT CNOT4 RING-finger domain, but this could be fully restored by incorporating the charge-swapping D48K/E49K substitutions in CNOT4 [15]. Although K63 interacts directly with D48/E49, this is not the case for K109/K117, indicating that even among conserved residues there are different ways of involvement in generating selectivity and E2-E3 pairing. Additionally, the most remote component of the identified salt-bridge network, D137/D145 is also conserved in the E2 superfamily, but it is not directly involved in RING interactions. The demonstration in this work that mutations at this position can switch selectivity emphasizes that E2-E3 interaction specificity can be a consequence of a subtle, dynamic interplay between interface and non-interface residues. In the majority of E2 enzymes, residues involved in establishing E3 interaction specificity are not concentrated in a single hotspot, but dispersed over the N-terminal helix one and two relatively flexible and divergent loop regions (L1 and L2) [2], [11]. Mutagenesis of these residues can abolish or modulate E2-E3 specificity [10], [11], [15]. Apart from sequence information, additional structural characteristics are affecting RING selectivity, like the length of H1, the flexibility of L1 and L2 and the triangular distances between these elements [12], [24], [25]. Comparing the sequences of UbcH6 and UbcH8 reveals additional amino acid differences at several key positions in these regions. One of them, D58/E66, lies within H1 and it is not orientated toward the E2-E3 interface. D58/E66 is in close proximity to T144/152 and S146/154, which are solvent exposed and directly interacting with RING domains. Therefore, D58/E66 might affect E3 interactions either indirectly through T144/152 and S146/154 or via a repositioning of helix H1. Indeed, the linker region that connects H1 with the rest of the E2 enzyme is found to be flexible [26] and D58/E66 might be involved in a hinge-like mechanism that might allow some conformational freedom of H1, which influences the distances between the three triangle points and thereby RING-interactions. Finally, the N-terminal extension of the tested E2 enzymes, not present in the crystal structures, might also affect their E3 binding preference. This could explain why reverse mutations in UbcH6 that mimic UbcH8 did not restore the E3 interaction profile of UbcH8. The presence of another intermolecular salt-bridge in the UbcH6-TOPORS complex generated by HADDOCK, involving the residues LYS43 (UbcH6) and GLU28 (TOPORS) (data not shown), could explain why UbcH6 E105D did not affect the selectivity profile of UbcH6. This salt-bridge cannot be formed with UbcH8, where LYS43 is replaced by ALA51 in the sequence. Comparing the sequences of UbcH6 and UbcH8 indeed reveals amino acid differences at several key positions in this N-terminal extension known to be primarily involved in the interaction with the E1 ubiquitin-activating enzyme, but also to play a role in E3 interaction selectivity. We should note that Y2H screening of physical protein-protein interactions between E2 enzymes and E3 RING-finger domains does not directly address biological functionality, e. g. the ability to transfer the activated Ub. However, we previously demonstrated that E2-RING E3 interactions found by LexA-B42 yeast two-hybrid assays are good predictors for enzymatic functionality [13] and therefore believe that the results presented here are also relevant in a functional context. It is unlikely that minor alterations in side-chain characteristics described in this work would affect protein stability in vivo (and our in silico results are supporting this). In addition, a recent systematic analysis of interaction dynamics across different technologies reported that high-throughput yeast two-hybrid is the only available technology for detecting transient interactions on a large scale [27], which support the recourse to this technique to unravel labile E2-E3 interactions. Despite their high percentage of identity, it has been reported that UbcH6 plays a major role in ubiquitin-conjugation while UbcH8 has only a minor role in ubiquitination but rather is the key conjugating enzyme of the ISGylation pathway [17]. This functional difference is in line with our observations [13] and supports the importance of an E3 selectivity mechanism that must differenciate not only among Ub-E3 ligases but also between Ub and UbL E3 ligases. Finally, this unreported dynamic intra-molecular salt-bridges network constitutes a new fundamental principle to understand the structural and evolutionary determinants of multispecific recognition. In conclusion, a dynamic equilibrium of conserved residues in two highly homologous E2 enzymes was identified, that mediate RING interactions. Amino acids located both within the classical interaction surface as well as residues that are remote from this surface are actively involved in modulating side-chain conformations and thus availability for binding of crucial residues. The subtle and dynamic nature of the identified regulatory switch suggests new ways how protein interactions can be controlled. Furthermore, the observation that minimal sequence differences between two highly similar proteins can control protein interaction networks serves as a cautionary tale and raises new challenges for bioinformatics analysis and modeling of protein interactions. Finally, these findings indicate that the widely accepted sequence-structure-function paradigm should be extended to sequence-structure-dynamics-function relationship. High-copy yeast-two hybrid (Y2H) shuttle plasmids expressing human UbcH6 (UBE2E1) and UbcH8 (UBE2E2) and RING-finger domains of 250 human RING-type E3 ubiquitin protein ligases as fusions with the E. coli LexA binding domain (BD) or with the B42 acidic activator domain (AD), respectively are described in [13]. All amino acid substitutions were introduced using QuickChange II Site-Directed mutagenesis (Stratagene), according to the manufacturers' instructions. Appropriate mutagenesis was validated using DNA sequencing. Both wild-type and mutant UbcH6 and UbcH8 were used as starting structures. Wild-type UbcH6 and UbcH8 structures were taken from the Protein Data Bank (PDB) (PDB-ID 3bzh and 1y6l, respectively). Selective mutations were introduced with CNS (Crystallographic and NMR System) [28], keeping the Ca and the Cb atoms fixed to preserve the side-chain rotamer. Each system was simulated in duplicate for 200 ns using the GROMOS G53a6 force field [29] and GROMACS 4. 5. 5 package [30]. The starting structures were solvated in a dodecahedral box with a minimal 14 Å distance between solute and box, resulting in systems of about 11500 SPC water molecules, 34 Na+ and 36 Cl− ions (±1, depending on the system). Simulations were performed under periodic boundary conditions at 300 K and constant pressure (1 bar), using either Berendsen [31] or v-rescale [32] coupling algorithms for temperature control and Berendsen [31] coupling algorithm for pressure control (with coupling constants of 0. 1 and 1. 0 ps, respectively). Bond lengths were constrained with the Linear Constraint Solver (LINCS) algorithm [33] and the time step for the integration was 2 fs. Electrostatic interactions were calculated using either the reaction-field method [34] with a 14 Å cut-off distance or Particle Mesh Ewald (PME) [35], [36] with a 10 Å cut-off distance. Non-bonded interactions were updated every 10 fs (with a 10 Å cut-off distance for the short-range neighbor list). MD analysis was performed excluding the first 1 ns of the trajectories. All various simulation settings resulted in similar trajectories and only the results obtained with PME and v-rescale temperature control are discussed here. Structural models of UbcH6 and UbcH8 in complex with the RING-finger domain of TOPORS were generated by structural alignment of single components onto the bound structures of existing E2-E3 complexes (PDB: 1fbv, 1ur6,3eb6,2c2v and 2oxq). The resulting models were subjected to a short refinement in explicit water using HADDOCK [18], [19]. Electrostatic potential surfaces were determined using the Adaptive Poisson-Boltzmann Solver package (APBS) [37] within Pymol (version 1. 3) [38]. Manipulation of yeast cells and two-hybrid techniques are described in [13]. Briefly, WT and mutant LexA-E2 fusions were co-transformed together with the pSH18-34 LacZ reporter plasmid in EGY48α cells. To evaluate the effects of E2 mutants on human RING-type interactions, 250 sequence-verified B42-RING constructs were arrayed in EGY48a cells. Mating between α and A cells was performed on non-selective Yeast Peptone Dextrose (YPD) -medium (24 hours at 30°C). Diploids were selected for 48 hours on synthetic complete (SC) medium lacking the amino acids histidine, tryptophan and uracil (HWU− medium) by manual transfer of yeast spots using a 384-well replicator pinning tool (V & P Scientific, San Diego). Putative interactions were assessed in triplicate by growing diploids on SC HWU− medium supplemented with 5-bromo-4-chloro-3-indolyl β-D-galactopyranoside (X-gal) or on SC HWUL− (idem as HWU− but also lacking leucine), including either galactose or glucose as main carbon source. Quantification of interactions was done after 72 hours as reported earlier [13].

During their life, proteins undergo various modifications ranging from structural marking or signaling to degradation. One major biochemical process involves ubiquitin, a small and evolutionary conserved protein. This regulatory protein serves as a tag that, when attached to a protein substrate, alters its function, cellular sub-location or commits the labeled protein to destruction in the proteasome. The high specificity of the ubiquitination pathway is achieved through interactions between two large protein families, E2 and E3, that ensure the efficient covalent conjugation of ubiquitin. By comparing two “almost identical” E2 enzymes, we identified a single minute substitution that, operated by a dynamic network of salt-bridges, functions as a subtle switch that controls interaction selectivity toward E3 proteins. Using a combination of bioinformatics and modeling techniques, complemented by mutagenesis and experimental screening of E2-E3 interactions, we unraveled an equilibrium between an “open”, binding-competent and a “closed”, binding-incompetent state. Subtle modifications in this network are sufficient to switch the selectivity profile. These findings should serves as a cautionary tale and raises new challenges for bioinformatics analysis, modeling and experimental engineering of protein-protein interactions. The dynamic nature of the identified regulatory switch suggests that the widely accepted sequence-structure-function paradigm should be extended to sequence-structure-dynamics-function.

Abstract Introduction Results Discussion Materials and Methods

protein interactions molecular dynamics macromolecular assemblies signaling networks protein structure biophysics simulations sequence analysis biochemistry simulations proteins chemistry biology proteomics biophysics biochemistry biochemical simulations computer science computational chemistry computer modeling biophysic al simulations computational biology macromolecular structure analysis

2012

Dynamic Control of Selectivity in the Ubiquitination Pathway Revealed by an ASP to GLU Substitution in an Intra-Molecular Salt-Bridge Network

Scrub typhus is a serious public health problem in the Asia-Pacific area. It threatens one billion people globally, and causes illness in one million people each year. Caused by Orientia tsutsugamushi, scrub typhus can result in severe multiorgan failure with a case fatality rate up to 70% without appropriate treatment. The antigenic heterogeneity of O. tsutsugamushi precludes generic immunity and allows reinfection. As a neglected disease, there is still a large gap in our knowledge of the disease, as evidenced by the sporadic epidemiologic data and other related public health information regarding scrub typhus in its endemic areas. Our objective is to provide a systematic analysis of current epidemiology, prevention and control of scrub typhus in its long-standing endemic areas and recently recognized foci of infection. The traditional endemic area of scrub typhus is known as the “tsutsugamushi triangle”. It is a region covering more than 8 million km2, from the Russian Far East in the north, to Pakistan in the west, Australia in the south, and the Japan in the east [23,48,49]. There are one billion people at risk of infection; the endemic area is highly populated [1]. The progress of globalization and associated travel contributes to the exportation of the infected persons to non-endemic areas [50]. The antigenic and genetic diversity of O. tsutsugamushi strains, and their unclear correlation with virulence for humans, confound the epidemiological study of scrub typhus [51]. Better understanding of the epidemiology of scrub typhus will help efforts to prevent and control the disease. This part of the article describes studies of the geographic distribution and risk factors of scrub typhus in both the endemic areas and in travelers from the rest of the world. Even with its recent re-emergence in the traditionally endemic areas and worldwide, scrub typhus is still a neglected infectious disease [6,63,64]. The geographic distribution of scrub typhus is determined by the distribution of its vector and reservoir—mites, primarily of the genus Leptotrombidium. Humans are accidental hosts [1,51]. Outdoor workers, especially field workers in rural areas, have a higher risk of acquiring the disease [65]. It is reported that rice fields are an under-appreciated location where the biting of mites and transmission of O. tsutsugamushi occurs in the endemic areas [63]. Tropical weather provides stable and ideal conditions for transmission of the disease. High temperature and high humidity are optimal for mite activity. In more temperate climates, the transmission of scrub typhus is more seasonal due to the temporal activity of chiggers [45,63,66]. Scrub typhus has been a nationally notifiable disease in Bhutan, China, Japan, South Korea, Thailand, and Taiwan [67,68,69,70,71,72]. The reported seroprevalence of O. tsutsugamushi for each country is shown in Table 1 [73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106]. The reports of outbreaks of scrub typhus are summarized in Table 2 [43,94,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137]. The list of other countries with published cases of scrub typhus is found in S2 Table [80,136,138,139,140,141,142,143,144,145,146,147,148,149,150,151,152,153,154]. All countries with reported cases are designated in Fig 1. The literature review confirmed that the majority of scrub typhus cases were reported in the “tsutsugamushi triangle” in the Asia-Pacific region (Table 3) [23,45,105,155,156,157,158]. There were a few cases reported in Central Asia and the Middle East, which are outside the traditional definition of the Asia Pacific region, but neighboring it [49,145,159]. The cases are primarily found in southwest China, and the southeast coastal and eastern regions of China. May is usually the start of the scrub typhus season, and June and July are the peak months. The pattern correlates with the weather and life cycle of mites [160]. Recent studies showed that the geographic distribution of the disease has expanded to northern China. It has existed in southern China for thousands of years [158,161]. Scrub typhus cases can be divided into 6 clusters in China. Cluster 1, the significant primary cluster, is located in southern and southeast China, which includes provinces of Guangdong, southern Fujian, Jiangxi, and Guangxi. The secondary cluster is mainly in southwest China, which includes Yunnan and Sichuan Provinces. Jiangsu, Anhui and Shandong provinces in East China are the third cluster for scrub typhus. Shaanxi province in the Northwest and Beijing Municipality were recognized as the fourth and fifth clusters in the analysis done by Zhang et al. The provinces of Zhejiang and northern Fujian were cluster 6 [158]. In other studies, there were reported cases in Hunan Province and Tibet [160]. Data collected between 2006 and 2012 show that the highest cumulative incidence was in the 60–69 year-old age group (0. 66 per 100,000), and the lowest one was in the 10–19 year-old age group (0. 11 per 100,000). The 50–60 year-old group accounted for the largest portion of all scrub typhus patients in China (21. 36%). There was no difference in incidence between genders [158]. A remarkable resurgence and a prominent outbreak occurred between 1976 and 1984 due to an increase of mite populations carrying O. tsutsugamushi. There was no explanation available for the increased number of mites during that period [162]. The disease has now been found in almost all areas of Japan except in Okinawa and Hokkaido prefectures. A retrospective study in 1998 demonstrated that Kyushu (51% of total cases), Tohoku-Hokuriku (27%) and Kanto (19%) had the largest numbers of cases [45]. In contrast to China, November accounted for the largest proportion of reported cases driven by the large number of cases in Kanto and Kyushu. May had the second highest monthly number of cases due to the cases in Tohoku-Hokuriku [45]. The age distribution differs between Japan and China. In Japan, 62% of cases were 51–75 years old, while in China this age group accounted for less than half of the total patients (~48. 2%) [45,158]. No significant gender differences were observed in Orientia infection in Japan. Not surprisingly, working in farming and forestry is an important risk factor for scrub typhus [45,162]. Scrub typhus was first reported in South Korea during the Korean War, but it was still unfamiliar to Korean civilians until 1986 [163]. The disease has subsequently been recognized as the most common rickettsial disease in South Korea [156,163,164]. Nation-wide seroepidemiologic and microbiologic surveys demonstrated that 27. 7% to 51% of acute febrile illness patients in South Korea were seropositive for O. tsutsugamushi between 1986 and 1993 [163]. The study confirmed that scrub typhus was widely spread in the country, and that it was frequently underdiagnosed [163]. Scrub typhus became a reportable disease in South Korea in 1994. Physicians must report all confirmed or suspected scrub typhus cases to both the local health bureau and the Korean Centers for Disease Control and Prevention (CDC). The gender inequality of scrub patients is unique in this country. Several studies confirmed that more female patients were reported than male patients (~65% vs 35%) [70,156]. One possible explanation is due to the conventional working behavior in farms in South Korea. Female workers typically work in a squatting position in dry fields, while male farmers tend to stand with tools in rice fields during work [70]. Other characteristics of the epidemiology of scrub typhus in South Korea are the increasing incidence in urban areas and expansion to northern regions [70,164,165]. The proportion of cases identified in urban areas increased from 20% (388 cases) in 2002 to 26. 9% (1,345 cases) in 2009, while that in farmers decreased from 43. 3% to 25%. However, further analysis revealed that outdoor activity in urban areas is the most common risk factor [70,164]. Similar to Japan, October and November are the peak months for scrub typhus cases. The age group 60–69 years old is the largest group for scrub typhus cases in South Korea (27. 48%), and 72. 2% patients were 50–79 years-old [70,156]. Scrub typhus was recognized as a typhus-like fever in India in 1917 [1,166]. It was a major cause of fever among military personnel along the Assam-India-Myanmar (formerly Burma) border during World War II, and the 1965 Indo-Pak war [1,43]. The disease resurged at the Pakistan border of India in 1990 [43]. The widespread use of insecticides and empiric treatment of febrile illness as well as changes in lifestyle all contributed to the subsequent decrease in incidence [43,167]. However, scrub typhus is still an under-diagnosed disease in India [43]. Field epidemiology studies indicate that the disease occurs all over India, from South India to Northeast India and Northwest India. There were cases reported from Maharashtra, Tamil Nadu, Karnataka, Kerala, Himachal Pradesh, Jammu and Kashmir, Uttaranchal, Rajasthan, West Bengal, Bihar, Meghalaya, and Nagaland [27,43,110,168,169]. The peak of the disease is between August and October. Leptotrombidium deliense is reported as the primary vector of O. tsutsugamushi [1,155]. Socioeconomic status and occupation are important risk factors for scrub typhus. Most scrub typhus patients in India are uneducated and live in rural areas [27,43]. The tropical climate of Thailand provides an ideal environment for the vectors of O. tsutsugamushi, L. deliense and L. chiangraiensis [1]. Nationwide sero-epidemiological studies revealed a high prevalence of scrub typhus in Thailand [157,170]. The rates of O. tsutsugamushi antibodies varied from 13% to 31% of residents in suburban Bangkok, to 59% to 77% of residents in the northern and northeastern regions [157]. A human case was first reported from the central region of Thailand in 1952 [157,171]. The pathogen was first isolated from rodents in the same part of the country two years later [172]. There was a substantial increase in the number of confirmed cases in Thailand from the 1980s to the 2000s [157]. Growing awareness of the disease and development of new diagnostic tools may at least partially contribute to this trend [157,173,174]. Different from China and Japan, the male to female gender ratio of scrub typhus patients in Thailand is 2: 1 [157]. The age distribution of the disease in Thailand is that the 50–59 year old group is the largest group (22. 3%), but both the 30–39 year old and 40–49 year old groups are similar, ~20%. Outdoor activity, especially occupational exposure, is a critical risk factor [157]. The earliest reports of scrub typhus cases in Vietnam can be dated back to the 1960s during the Vietnam War [175,176,177,178]. Most of the patients in the 1960s and 1970s were military personnel, especially American servicemen in South Vietnam [46,178]. The disease had been neglected in Vietnam since then until the end of the last century and the beginning of 21st century, resulting in a gap in publications during that time [179]. Differing from past studies, current studies and case reports of scrub typhus in Vietnam have been focused on the northern part of the country. There was only one study in the central part of Vietnam, Quang Nam province, which identified that the main genotype in the area was the Karp group [180]. Recent studies of scrub typhus in North Vietnam demonstrated that the cumulative incidence of scrub typhus was about 1. 1% among the general population, and ~3. 5% among patients admitted to hospitals [103,105]. The peak season in North Vietnam is summer though cases occur throughout the year. The transmission pattern of scrub typhus in tropical South Vietnam may be different because a seasonal pattern is more obvious in a temperate climate [66,105]. There was no significant difference between urban and rural areas [103]. In addition to the countries described above, there are quite a few other countries with reported scrub typhus cases in the tsutsugamushi triangle. Scrub typhus has been recognized on the islands of the southwest Pacific including Indonesia and the Philippines, and the continent of Australia for almost a century [149,181,182,183,184]. It was recognized as “coastal fever” in 1913 and scrub typhus after the 1920s in Australia. The endemic areas in Australia are the tropical coastal periphery of northeastern Queensland, the tropical region of the Northern Territory, and the adjacent Kimberly region of Western Australia [1,181,182]. A new strain, Litchfield, different from those from other Asia-Pacific area countries was isolated in Australia in 1998 [185]. The Philippines did not confirm the occurrence of scrub typhus until World War II. The first but failed US scrub typhus vaccine was prepared from the lungs and spleens of white rats infected with Volner strain of O. tsutsugamushi. The Volner strain was originally isolated from the blood of a soldier in the Philippines [149,186]. The history of scrub typhus in Malaysia could be traced back to 1915 [187,188,189]. World War II made the disease known in the Solomon Islands, Republic of Vanuatu, and Papua New Guinea [128,146,190,191,192]. Our literature search found isolated Orientia infection in human patients in Far East Russia and Pakistan, and further studies may be necessary to confirm the distribution and other epidemiological features of scrub typhus in these countries [48,153,154,193,194,195,196]. The distribution of scrub typhus covers a large and diverse area. In the Asia-Pacific region alone, different countries in the endemic area have different climates, environment, and culture, which all contribute to the different characteristics of epidemiology of the disease [1,64]. Our study has the limitation that we only analyzed literature in English within the databases of PubMed and Google Scholar. There were quite a few publications in Chinese, Japanese, Korean, Russian, and other languages not included in the analysis. Our literature review determined that there are a few sporadic scrub typhus cases from countries and regions outside the traditional “tsutsugamushi triangle” in the Asia-Pacific area (Fig 1). UAE is outside the traditional endemic triangle. However, the case reported in 2010 demonstrated a scrub typhus case confirmed to be caused by a new Orientia species, O. chuto [49]. The Australian patient concerned had traveled to Dubai, UAE and the United Kingdom before the onset of her febrile illness. She noticed an eschar on her abdomen after the Dubai visit. An IFA, polymerase chain reaction (PCR), and sequencing were employed to determine the etiologic pathogen. The molecular variance of the 47-kDa gene, 56-kDa gene and other nucleotide sequences, and geographical difference led the researchers to propose the new Orientia species. There was only one known Orientia species, i. e. , O. tsutsugamushi, before this case [23,49]. Before the two scrub typhus reports in Chile, there was no reported autochthonous scrub typhus case in the Western Hemisphere. A patient was bitten by terrestrial leeches on Chiloé Island in southern Chile in 2006 [139]. Both PCR, a molecular biological test, and, IFA, a serological test, showed diagnostic confirmation of O. tsutsugamushi infection. The molecular analysis further indicated that the pathogen is closely related but not identical to other O. tsutsugamushi and O. chuto species. It was significantly closer to Orientia sp. than to other rickettsiae. The results suggested that the pathogen from the Chilean sample was not a simple import from an endemic area. In addition, the case also reminds us that there might be other vectors, such as leeches, for Orientia [139]. The latest study reported three native cases on the same island in 2016 [151]. The researchers used both serological testing and molecular analysis to diagnose the patients. At least one patient received diagnosis by two serological tests from both the local hospital and the Mahidol-Oxford Tropical Medicine Research Unit in Thailand, and molecular analysis at the Lao-Oxford-Mahosot-Hospital-Wellcome Trust Research Unit. They used paired antibody titer comparison for diagnosis, which is more reliable than a single titer diagnosis [197]. The second patient was documented by molecular analysis of the agent but half of the IFA serologic tests were negative. The third patient had two high single-titer IFA results but negative PCR results. No leech bite was observed in these cases [151]. More follow-up studies could fill these gaps. There are case reports from Cameroon, Kenya, Congo, Djibouti and Tanzania in Africa [102,143,144,147]. The only case report from Cameroon was an American who visited Cameroon before developing a febrile illness [143]. The patient’s IFA titer to O. tsutsugamushi increased from 1: 256 to >1: 1,024 two weeks after admission, but PCR analysis of formalin-fixed, paraffin-embedded skin samples was negative. Several clinical features were not typical [143]. The researchers in Kenya screened samples of reactive sera from patients with febrile illness in Kenya. Western blot was performed to confirm the specificity. About 5% of the serum samples contained antibodies reactive with O. tsutsugamushi [102]. The clinical features and serological test, IIP, confirmed the diagnosis of O. tsutsugamushi Kato strain in the only case in Congo. However, the patient resided and was diagnosed in Japan. The patient visited Congo for 23 days, and noticed symptoms 8 days after he left Congo. The researchers contacted local centers for disease control in Japan where the patient lived and worked. They did not find similar reports in Japan so they concluded that the patient contracted the disease in Congo. No other case in Congo has been reported [147]. Therefore, it is reasonable to suspect the possibility of domestic infection in Japan instead of acquisition in Congo. More epidemiologic studies of scrub typhus are necessary to confirm the existence of scrub typhus in Congo. A new study of 49 abattoir workers in Djibouti demonstrated that three workers were seropositive against Orientia, and one worker seroconverted during the study [80]. The titers observed in the study were 100–400 for ELISAs, and 1: 128 for IFA. The cut-off titer and methods used in this study may be controversial, and require further validation [197]. In addition, the study did not provide the participants’ travel history even though the authors played down this confounder due to the subjects’ socioeconomic status [80]. A Dutch traveler to Tanzania contracted O. tsutsugamushi there. Researchers in the Netherlands confirmed the case with clinical features (fever, eschar, etc.) and serological tests, i. e. , IFA [144]. No other case has been reported in Tanzania.

Scrub typhus is a serious public health problem in the Asia-Pacific area. There is an estimated one million new scrub typhus infections each year, and over one billion people around the world are at risk. Without appropriate treatment, the case fatality rate of scrub typhus can reach 30% or even higher. Scrub typhus has long been a neglected infectious disease so many aspects of the disease, including its diagnosis to prevention, are unknown. We here provide a comprehensive review of the epidemiology, prevention and control of scrub typhus.

Abstract Introduction Methods

dermatology invertebrates typhus medicine and health sciences pathology and laboratory medicine japan pathogens geographical locations microbiology animals bacterial diseases signs and symptoms infectious disease control bacterial pathogens orienta tsutsugamushi infectious diseases eschar medical microbiology microbial pathogens mites scrub typhus arthropoda people and places eukaryota diagnostic medicine asia fevers biology and life sciences organisms

2017

A review of the global epidemiology of scrub typhus

Reconstructions of cellular metabolism are publicly available for a variety of different microorganisms and some mammalian genomes. To date, these reconstructions are “genome-scale” and strive to include all reactions implied by the genome annotation, as well as those with direct experimental evidence. Clearly, many of the reactions in a genome-scale reconstruction will not be active under particular conditions or in a particular cell type. Methods to tailor these comprehensive genome-scale reconstructions into context-specific networks will aid predictive in silico modeling for a particular situation. We present a method called Gene Inactivity Moderated by Metabolism and Expression (GIMME) to achieve this goal. The GIMME algorithm uses quantitative gene expression data and one or more presupposed metabolic objectives to produce the context-specific reconstruction that is most consistent with the available data. Furthermore, the algorithm provides a quantitative inconsistency score indicating how consistent a set of gene expression data is with a particular metabolic objective. We show that this algorithm produces results consistent with biological experiments and intuition for adaptive evolution of bacteria, rational design of metabolic engineering strains, and human skeletal muscle cells. This work represents progress towards producing constraint-based models of metabolism that are specific to the conditions where the expression profiling data is available. Commonly, genome-scale metabolic networks are reconstructed to contain all known metabolic genes and reactions in a particular organism [1]. These reconstructions are thus a superset of the metabolic reactions that are functioning in the organism at any one time. The processes that determine which enzymes are active in a cell are often overlooked in constraint-based studies. Of particular interest are the transcriptional regulatory processes in cells which choose a subset of possible enzymes for activity at any given time. Knowledge of transcriptional regulation of metabolism comes from different sources. At a low level, from the bottom up, some of the regulatory proteins that control the transcription of sets of metabolic genes are known [2]. At a higher level, from the top down, gene expression data provides a picture of what genes are being transcribed at a particular time, and hence which enzymes are probably active in the cell [3]. Both of these types of knowledge can be used to refine metabolic networks under given conditions. There are three ways to study how regulation tailors gene-expression under a specific condition. First, if a transcriptional regulatory network network (TRN) is available, then the transcription state of the cells can be computed for a given input [4], [5]. However, genome-scale TRNs are not available. Even for Escherichia coli, it has been estimated from dual perturbation experiments that only about one-fourth or one-third of its TRN is currently known [6]. ChIP-chip data and other approaches may soon enable more comprehensive reconstructions. Second, in the absence of a TRN, optimization procedures based on the assumption that the organism picks out the best set of reactions to meet a physiological objective have been used [7]. However, there are multiple solutions to such problems [8] and no real way to determine which internal reactions are used in the absence of flux data. In addition, the statement of an objective introduces a ‘user-bias’ and such objective may not actually be relevant to the true physiological state. The third approach to study regulation relies on the available of expression profiling data. If such data is available for the conditions being examined we can directly examine the expression of the ORFs accounted for in a genome-scale reconstruction. Metabolic network reconstructions can be combined with gene expression data from different states to identify regulatory principles in organisms [9]. Pathway-based analysis methods can be used to predict the usage of entire pathways of reactions based on the expression state of multiple genes [10]. Gene expression data has previously been applied with yeast to predict which reactions may be inactive on a gene-by-gene basis [11]. More recently, gene expression data has been interpreted in terms of elementary modes [12], moving towards a more functional view of analysis. The results of these methods are dependent on the quality of the expression data that is used as input. Expression data is known to be noisy, and the variety of methods for converting the fluorescence intensity of thousands of spots on a chip to semi-quantitative readings of mRNA molecule counts do not produce equivalent results [13], [14]. Importantly, due to the noise, it is impossible to define a comprehensive set of present mRNA transcripts without a large number of false-positives. Practically speaking, one can say either (1) these few mRNAs are almost certainly present in the cell, or (2) some of these many mRNAs are maybe present in the cell. Pathway-based methods, for example [10], attempt to avoid the noise problem by assuming that all mRNAs assigned to a particular pathway should be present or absent together. This is dependent on biased, human-imposed pathway definitions, and reactions that function within a pathway can also function outside of that pathway, limiting the use of such assumptions. Here we use gene expression data in combination with objective functions to create functional models despite potentially noisy data. We describe the use of genome-scale transcriptomic data to constrain reactions in both bacteria and human cells, enabling context-specific metabolic networks to be reconstructed and compared. We quantitatively define the consistency of gene expression data with assumed functional states of a cell, demonstrating agreement with physiological data. Context-specific metabolic networks will be virtually essential to accurately model human metabolism due to the variety of cell types and their corresponding metabolic processes. The approach to the construction of context-specific metabolic networks is termed Gene Inactivity Moderated by Metabolism and Expression (GIMME) and is illustrated in Figure 1. As inputs, the algorithm requires: (1) a set of gene expression data, (2) the genome-scale reconstruction, and (3) one or more Required Metabolic Functionalities (RMF) that the cell is assumed to achieve. Preliminary tests (not shown) suggest that proteomic data can be substituted for expression profiling data. Given these three inputs the algorithm produces a list of reactions in the network that are predicted to be active and an inconsistency score (IS) that quantitatively classifies the disagreement between the gene expression data and the assumed objective function. This inconsistency score is converted to a normalized consistency score (NCS), allowing for relative comparisons of how well each gene expression data set agrees with a particular metabolic function. Simply speaking, reactions that correspond to mRNA transcript levels below a specified threshold are tentatively declared inactive. If the cell cannot achieve the desired functionality without at least one of these reactions, linear optimization is used to find the most consistent set of reactions to reactivate. Inconsistency scores are calculated based on the product of distance from threshold and necessary flux for each reaction required to be reactivated, as illustrated in Figure 2. A smaller inconsistency score indicates that the data is more consistent with the RMF. The GIMME algorithm produces the network with the minimal inconsistency score through the following two-step procedure: Part A is achieved through flux balance analysis (FBA) [15]. Part B involves the solution of the following linear programming problem: In the formulation, xi is the normalized gene expression data mapped to each reaction. xcutoff is a cutoff value set by the user above which a reaction is definitely present; there is no contribution to the inconsistency score from reactions that are above this threshold. S is the stoichiometric matrix with reactions as columns, metabolites as rows, and stoichiometric coefficients as elements. v is the flux vector, quantitatively describing the flow through each reaction. ai and bi are the lower and upper bounds, respectively, for each reaction and define the minimum and maximum allowable flux. These bounds are set according to the maximal value of the RMF (s) found in step (1), in general by setting the lower bound corresponding to each RMF to some fraction of its maximal value. The great majority of the lower and upper bounds do not correspond to an RMF and are set to the same value as in a standard FBA problem, usually to arbitrarily high and low values, but sometimes to finite values as for input constraints (glucose uptake, for example) and irreversible reactions. The above optimization problem would generally be difficult to solve due to the presence of an absolute value operator, but in this case, a trivial simplification converts the above problem to a standard LP problem. Each reaction defined as possibly reversible (containing a negative lower bound) is converted to two irreversible reactions, thus restricting all fluxes to be positive, and removing the need for the absolute value. In general, some reactions will not have available data. The algorithm takes a conservative approach and designates these reactions as active; hence the term “gene inactivation” is part of the method name. The algorithm treats these reactions as if they had data that surpassed the cutoff; this is a conservative approach to avoid any penalty for absent data. The lack of data does have implications for the interpretation of results. It is entirely possible that given better data, these reactions would be determined to be absent, perhaps necessitating the activation of other reactions. Clearly, with limited data, the results must be considered with caution. In general, this is far more of a concern for human metabolic networks than for E. coli. We have used the GIMME algorithm to produce context-specific metabolic networks for E. coli for several different conditions and to compare inconsistency scores for different strains of the bacterium. We show that the inconsistency scores agree with experimental data in nearly all cases. Gene expression data from different conditions of E. coli growth are the input data, and the independent validation data is phenotypic data describing the relative growth and product secretion. The wide variety of human cell types in the body do not share a simple objective such as cellular growth, but rather have a multiplicity of functions necessary for multi-cellular life. Accordingly, understanding the metabolism of any particular cell type requires a model that contains only the reactions present in that cell type, without potentially thousands of extraneous reactions. Human Recon 1 [22] will contain many reactions that are inactive in particular cell types. Accurate models require their removal, and the GIMME algorithm provides a framework for this process. Herein we describe the first functional genome-scale metabolic models for particular human cells, in this case, skeletal muscle cells in different conditions. The work reported herein details the first available method to both produce a guaranteed functional metabolic model specific to a set of gene expression data and quantify the agreement between gene expression data and one or more metabolic objectives. We have demonstrated the functionality of this GIMME method with gene expression data from E. coli and human skeletal muscle cells. We have shown that (1) the computed consistency between gene expression data for different conditions and RMF agrees with physiological data, (2) the most consistent networks depend on the metabolic objective and media conditions, and (3) the most consistent networks for human skeletal muscle cells contain significantly fewer reactions than the global human model. Initially, we expected that the results for human models would be more interesting than those for any other organism reconstructed to date, principally because we expected that human cells would show the most variability across conditions. However, the lack of available data for a substantial number of human metabolic reactions confuses attempts at comparison. We showed that reducing the number of reactions considered by 5% can change the apparent differences between different datasets. In addition, the lack of replicates in human gene expression data sets and the difficulty in obtaining high quality biological controls complicates matters and reduces the statistical power of comparisons. We have higher confidence in the results presented for E. coli because nearly all of the gene-associated reactions have data available, replicates are available, and controls are present. We also found that a substantial number of reactions in E. coli do vary in activity when different input conditions are provided. In the end, we conclude that a tool originally conceived to plug a key gap in the analysis of human cellular metabolism actually provides more immediate use in the analysis of microbial metabolism. With metabolic reconstructions growing in size and becoming available for more and more organisms, tools to filter global reaction lists into context-specific reaction lists will be highly useful. Meaningful analysis of the human metabolic network will require procedures such as GIMME in order to accurately predict phenotypes. The gene expression data was obtained as CEL files and processed using Bioconductor [25]. The data for E. coli was processed using GCRMA as implemented within Bioconductor [14]. The data for human skeletal muscle was processed using the affy package [26] and the mas5calls function. The p values were subtracted from 1 and the resulting value used as a quantitative measure of likelihood that the gene was available. The default parameters were used. For all datasets, the expression level of each reaction was determined by mapping any available data from genes associated with that reaction. If data was not available for any gene associated with a reaction, it was given a score of −1. If data was available for one or more genes, a single score was computed by evaluating the boolean GPR associations; OR' s would evaluate to the greater of the two values, AND' s to the lesser. The end result was a score for each reaction from each set of data, either −1 or non-negative, with greater numbers implying greater certainty that reaction is present. This is the data that was input into the GIMME algorithm to compute the consistency scores and context-specific metabolic networks. The GIMME algorithm is implemented in Matlab, using functions in the COBRA Toolbox. In general, any robust linear programming solver should work; we used Tomlab (Tomlab Optimization, Pullman, WA). The output from the GIMME algorithm is an inconsistency score, and a higher score means that the gene expression data is less consistent with the model achieving the desired objective. For visualization purposes only, these scores are converted into normalized consistency scores, with a higher score indicating greater consistency between the data and the modal achieving the objective. For a given set of scores, each inconsistency score is subtracted from 1. 02 * (maximum inconsistency score) to produce a set of consistency scores. Each consistency score is divided by the maximum consistency score to produce a set of normalized consistency scores. The 1. 02 factor assures that the smallest consistency score is slightly greater than zero and easy to visualize on a graph. Permutation testing with 10,000 randomizations was used to determine the statistical significance of all results with regard to consistency scores. This testing was implemented in Matlab. Heat-map type representations were produced in Matlab. Other graphs were produced in Excel (Microsoft, Redmond, WA).

Systems biology aims to characterize cells and organisms as systems through the careful curation of all components. Large models that account for all known metabolism in microorganisms have been created by our group and by others around the world. Furthermore, models are available for human cells. These models represent all possible biochemical reactions in a cell, but cells choose which subset of reactions to use to suit their immediate purposes. We have developed a method to combine widely available gene expression data with presupposed cellular functions to predict the subset of reactions that a cell uses under particular conditions. We quantify the consistency of subsets of reactions with existing biological knowledge to demonstrate that the method produces biologically realistic subsets of reactions. This method is useful for determining the activity of metabolic reactions in Escherichia coli and will be essential for understanding human cellular metabolism.

Abstract Introduction Results/Discussion Materials and Methods

computational biology/systems biology genetics and genomics/gene expression

2008

Context-Specific Metabolic Networks Are Consistent with Experiments

Pericardial fat is a localized fat depot associated with coronary artery calcium and myocardial infarction. We hypothesized that genetic loci would be associated with pericardial fat independent of other body fat depots. Pericardial fat was quantified in 5,487 individuals of European ancestry from the Framingham Heart Study (FHS) and the Multi-Ethnic Study of Atherosclerosis (MESA). Genotyping was performed using standard arrays and imputed to ∼2. 5 million Hapmap SNPs. Each study performed a genome-wide association analysis of pericardial fat adjusted for age, sex, weight, and height. A weighted z-score meta-analysis was conducted, and validation was obtained in an additional 3,602 multi-ethnic individuals from the MESA study. We identified a genome-wide significant signal in our primary meta-analysis at rs10198628 near TRIB2 (MAF 0. 49, p = 2. 7×10-08). This SNP was not associated with visceral fat (p = 0. 17) or body mass index (p = 0. 38), although we observed direction-consistent, nominal significance with visceral fat adjusted for BMI (p = 0. 01) in the Framingham Heart Study. Our findings were robust among African ancestry (n = 1,442, p = 0. 001), Hispanic (n = 1,399, p = 0. 004), and Chinese (n = 761, p = 0. 007) participants from the MESA study, with a combined p-value of 5. 4E-14. We observed TRIB2 gene expression in the pericardial fat of mice. rs10198628 near TRIB2 is associated with pericardial fat but not measures of generalized or visceral adiposity, reinforcing the concept that there are unique genetic underpinnings to ectopic fat distribution. Obesity is a heterogeneous condition, and its attendant metabolic sequelae may not be adequately captured by using traditional metrics of generalized adiposity [1]. In part, this is because different fat depots may be associated with differential metabolic risk. For example, visceral abdominal fat is thought to be a unique pathogenic fat depot [2], [3]. Ectopic fat depots, defined as fat depots in non-classical locations [4], may mediate vascular disease due to their local toxic effect on nearby anatomic structures. We and others have shown that pericardial fat, defined as fat surrounding the heart and attendant structures, but not visceral fat, is associated with coronary artery calcification and coronary heart disease [5], [6]. The hypothesized local toxic effect of pericardial fat is supported by experimental research demonstrating perivascular inflammation [7] and smooth muscle cell proliferation [8]. Prior studies have shown that measures of generalized adiposity, including body mass index, are heritable [9]. In addition, more recent work has demonstrated that markers of body fat distribution, including waist-hip-ratio [10], subcutaneous and abdominal visceral fat [3], and fatty liver [11] also have a heritable component. Recent large scale genome-wide association studies (GWAS) have identified genomic loci for indices of body fat distribution that are independent of BMI [11]–[13], further supporting the concept that unique genetic variants exist that are associated with ectopic fat depots. To explore this further, we conducted a GWAS of pericardial fat to determine whether genetic loci are associated with the propensity to store fat around the heart. The study sample characteristics are shown in Table 1 and Table S1. The mean age ranged from 55 years in the Framingham Heart Study to 62 years in MESA. In the MESA cohort, mean pericardial fat differed significantly between race/ethnicity groups. Compared to European Americans, mean pericardial fat was significantly lower in African Americans (P = 1. 1E-38) and Chinese Americans (P = 4. 8E-11), and was higher in Hispanic Americans (p = 0. 02). The heritability (h2) of pericardial fat in the Framingham Heart Study was 50%. Upon additional adjustment for height and weight, the h2 was 52%. We also calculated genetic correlations between pericardial fat, visceral fat (VAT), and BMI in the Framingham Heart Study. Genetic correlations between pericardial fat and VAT were 0. 57; between pericardial fat and BMI 0. 41, and between VAT and BMI 0. 75. In all cases, we confirmed that there are genes that are associated with pair-wise comparisons of all three traits (all p-values<1. 7*10E-15 for overlapping genetic correlations), although our results also suggest that not all genes are shared (all p-values<1. 4*10E-22 for non-overlapping genetic correlations). The quantile-quantile plot (Figure S1) of GWAS of 5487 individuals of European ancestry demonstrated deviation from the null with no evidence of population stratification (lambda 0. 99). The Manhattan plot (Figure S2) shows a genome-wide significant locus on chromosome 2 (p = 2. 7E-08). The lead SNP (MAF 0. 49) is rs10198628 located ∼80 kb upstream from the TRIB2 gene. Per copy of the A allele, pericardial fat volumes were 4. 4 cm3 lower in the Framingham Heart Study and 3. 6 cm3 lower in MESA. All SNPs with p-values<1E-04 are shown in Table S2. We observed no evidence for a sex interaction for rs10198628 (p = 0. 33). The variance of pericardial fat explained by the lead SNP in the Framingham Heart Study was 0. 5% and 0. 3% in MESA. We performed validation in a multi-ethnic sample of African ancestry (n = 1442, β = 3. 31, p = 0. 001), Hispanic (n = 1399, β = 3. 62, p = 0. 004), and Chinese (n = 761, β = 4. 56, p = 0. 007) participants from the MESA study, with a combined Stage 1 and Stage 2 p-value of 5. 4E-14 (Figure 1A and Figure 1B). To assess whether rs10198628 is specific to pericardial fat, we assessed its associations with other fat depots (Table 2). We observed no association with body mass index from the GIANT consortium (p = 0. 38) [14] or with visceral or subcutaneous fat from the Framingham Heart Study (p = 0. 17 and 0. 34, respectively). We observed nominal direction-consistent associations with waist-hip-ratio adjusted for body mass index from the GIANT consortium (p = 0. 01) [12] and with visceral fat adjusted for body mass index in the Framingham Heart Study (p = 0. 01). We tested whether previously published SNPs in association with waist-hip-ratio adjusted for BMI [12] and BMI [14] are associated with pericardial fat in our meta-analysis dataset (Tables S4 and S5). Among the 14 well-validated SNPs for body fat distribution, we observed direction-consistent associations with CPEB4, a gene involved in cell survival [15]. Among SNPs associated with BMI, we observed nominal direction-consistent associations with FTO; no other associations were observed. Because of the proposed locally toxic effect of pericardial fat and cardiovascular disease outcomes, we evaluated the association of rs10198628 with several CVD phenotypes. We observed nominal, direction-consistent associations with carotid intimal medial thickness from the MESA individuals of European ancestry (n = 2505, p = 0. 04), but we observed no association with myocardial infarction from the CARDIoGRAM consortium (p = 0. 37, OR 1. 01 [95% CI 0. 985–1. 04], n = 81929) [16] or with coronary artery calcification from the MESA study (p = 0. 25, n = 2527). We performed a look-up of 25 validated SNPs for coronary heart disease from the CARDIoGRAM consortium (Table S3) [16], and found that rs12190287 at TCF21 was associated with pericardial fat in a direction-consistent fashion (p = 0. 0019). No additional SNPs met the bonferroni corrected p-value threshold of p<0. 002 (0. 05/25). We queried available human gene expression genetics data (see Methods) and identified eQTL associations with TRB2 in omental adipose (rs890069,1. 79e-9) [42] and two independent subcutaneous adipose samples (rs4669887,3. 62e-9; rs12616457,1. 28e-6) [41], [42]. All of these variants are in LD with our lead SNP rs10198628 (rs890069, rs4669887, rs1261657 have r2 0. 38,0. 38 and 0. 51 with rs10198629, respectively). Next, we tested for gene expression in multiple subcutaneous (inguinal, axillary, and gluteal) and visceral (epididymal, retroperitoneal, mesenteric, omental) adipose depots as well as classical brown, pericardial, and perivascular adipose tissue which were dissected from high-fat fed male mice. mRNA expression levels of the adipocyte markers aP2 and PPARγ2 were comparable across all adipose depots. Trib2 was expressed in pericardial adipose tissue, as well as all of the other adipose depots surveyed (Figure S3). Expression in pericardial adipose tissue was comparable to that in other adipose depots. Lipin 1 is another annotated gene near our lead SNP. It was also expressed in pericardial adipose tissue, as well as all other depots surveyed. Furthermore, expression in pericardial adipose tissue was comparable to that in other adipose depots. We have identified a SNP near the TRIB2 locus that is associated with pericardial fat but not with body mass index or visceral abdominal fat. This SNP is also associated with pericardial fat in a multi-ethnic sample consisting of individuals of European, African, Hispanic, and Chinese ancestry. Finally, we identified a nearby eQTL, suggesting the potential for altered gene expression associated with our top SNP, or a correlated variant. An important question of our work is whether rs10198628 is uniquely associated with pericardial fat, or merely represents a manifestation of generalized adiposity. Our results suggest that this SNP is unique in its association with pericardial fat, given its strong association with pericardial fat in our stage 1 and stage 2 analysis. In contrast, this SNP was not associated with visceral abdominal fat, an ectopic fat depot that is correlated with pericardial fat. Further, we observed no association with our lead SNP and body mass index in more than 100,000 individuals from the GIANT consortium. We note that we observed nominal significance with our lead SNP with VAT-adjusted-for-BMI and waist-hip-ratio-adjusted for BMI, traits representing fat distribution. Gene expression analysis in mice showed that Trb2 is expressed in all adipose depots. Expression was not enriched in any depot. Confirmation that Trb2 is expressed in adipose tissue supports a functional role for this gene product in this tissue. Future studies are needed to investigate the specific function of Trb2 in adipose tissue, and in particular in pericardial fat. Finally, although our lead SNP is closest to Trb2, it is also possible that this is part of a regulatory region for another gene. Prior genome-wide association studies have primarily used easily-obtainable anthropometric measurements to estimate body fat distribution [12], [13]. While these studies benefit from concomitant enhanced power, the lack of detailed phenotyping renders the precise meaning of the measurement uncertain. In the current study, we have made use of well-validated measurements of pericardial fat that have been previously associated with coronary artery calcium [5], [17], myocardial infarction [6], [18], measurements of left ventricular structure and function [19], and carotid intimal medial thickness [20]. In this context, we sought to examine our lead SNP with coronary calcium, coronary heart disease, and carotid intimal medial thickness. Given the modest epidemiologic associations that have been observed in concert with the relatively small genetic effect sizes that are typical of GWAS, it is not surprising that we observed only nominal associations with carotid intimal medial thickness. Our findings are also notable for some enrichment of association with a SNP in TCF21, previously identified in association with coronary heart disease in CARDIoGRAM [16]. TCF21 encodes a transcription factor of the basic helix-loop-helix family and is a molecular marker of white adipose tissue [21]. TCF21 is expressed in the epicardium of the developing zebrafish, and is associated with perivascular cells but not cardiomyocytes [22]. This is relevant given the anatomic location of pericardial fat and thus rendering it a form of perivascular fat surrounding the coronary arteries. Our lead SNP is located about 80 kb downstream from TRIB2. TRIB2 is the tribbles homolog 2 gene, part of the Tribbles gene family. TRIB2 expression has been shown to be elevated in lung cancer, and has been found to induce apoptosis through downregulation of the transcription factor CCAAT/enhancer-binding protein alpha (C/EBPα) [23]. Via this mechanism, TRIB2 may also suppress adipocyte differentiation via AKt inhibition and C/EBPα degradation [24]. TRIB2 has also been associated with hematologic abnormalities including acute myelogenous leukemia [25]. TRIB2 has also been shown to be a regulator of the inflammatory activity of monocytes [26], suggesting a possible mechanism by which it may link low density lipoprotein cholesterol to plaque formation. TRB1 and TRB3 have both been linked to obesity and related phenotypes, with TRIB1 gene expression linked to adipose tissue inflammation [27] and TRB3 gene expression associated with insulin resistance [28]. Our lead SNP also lies near the LPIN1 gene, a compelling candidate gene that has previously been identified in association with lipodystrophy syndromes [29]. While it is tempting to implicate this gene in our association analyses, it is notable that only rare, not common variants have been detected in human populations in association with lipodystrophic phenotypes [30]. In addition, our best SNP in the LIPN1 gene has a p-value = 0. 03, rendering it less likely. Finally, our lead SNP is ∼1 MB from LPIN1. Strengths of our study include the well-characterized pericardial fat data present in both the Framingham Heart Study and the MESA Study. An additional strength is the extension of our findings to multi-ethnic populations, underscoring the generalizability of this finding. Typical of GWAS, we have identified an associated locus, but the causal variant and gene remains unidentified. Limitations include the relatively modest sample size of our study, leading to relatively low power to detect small effects. While our finding that rs12190287 at TCF21 was associated with pericardial fat in a direction-consistent fashion, we are unable to perform a formal mediation analysis. These findings support the concept that unique genetic variants exist in association with different ectopic fat depots. These findings are important because they suggest that different ectopic fat depots may each have their own unique genetic signature that is independent of generalized adiposity. Future work should focus on identifying the molecular mechanisms that link these genomic loci to ectopic fat depots, as this could ultimately lead to the identification of novel pathways and new therapeutic targets. A SNP near TRIB2 is associated with pericardial fat but not measures of generalized or visceral adiposity, reinforcing the concept that there are unique genetic underpinnings to ectopic fat distribution. Pericardial fat was measured on CT using protocols determined by the participating studies, as described in the Study-Specific Methods. Sex-specific residuals were created, with adjustment for age, height, and weight, as well as principal components derived from genotypes denoting population stratification where necessary. Heritability of pericardial fat was calculated in the Framingham Heart Study using standard methods. Sex-and-cohort specific residuals were created and then pooled for analysis using variance components analysis (SOLAR) [31]. We used SOLAR [31] to calculate pair-wise genetic correlations between pericardial fat, visceral fat, and BMI in the Framingham Heart Study. We used residuals adjusted for age and sex. We tested two separate hypotheses: RhoG = 0 is the test for overlapping genetic correlations, whereas RhoG = 1 is the test for non-overlapping genetic correlations. Table S1 and the study specific methods describe the genotyping that was conducted. Quality-control filters were used to exclude low-quality samples or SNPs. Each study imputed ∼2. 5 million Phase 2 HapMap SNPs based on CEU samples; allelic dosage was used in the analysis. Each cohort separately conducted the regression analysis, using an additive genetic effect model with accounting for family structure when necessary. Next, we conducted a fixed effects weighted Z-score meta-analysis given possible differences in phenotype scaling between the participating studies using METAL [32]. Statistical significance was considered when SNPs reached a meta-analysis P value≤5×10−8 [33]. Discovery analyses were performed on European ancestry participants. SNPs were filtered at a minor allele frequency<2% and an imputation quality score<0. 3. We conducted Stage 2 validation using non-white ethnic samples from the MESA study. Statistical significance was achieved when a direction-consistent p-value was at least p<0. 05. For our lead SNP, we performed look-up in the publically available GIANT datasets [12], [14]. We also obtained specific look-up results in the CARDIoGRAM (for coronary heart disease) [16] and MESA (for coronary artery calcification and carotid intimal medial thickness). We also tested whether the 25 previously-identified SNPs for coronary heart disease from the CARDIoGRAM consortium [16] were associated with pericardial fat. To determine statistical significance, we used the false discovery rate q-value; SNPs with q-value<0. 05 were considered statistically significant using the QVALUE package in R [34]. We tested for a formal sex interaction of rs10198628. Each study computed the interaction regression coefficient, standard error, and p-value. For the sex interaction, we included, age, height, weight, and any principal components (and study center) that were used in the original discovery analysis. We additionally added rs10198628 and the cross-product rs10198628*sex. Interaction terms were meta-analyzed using the weighted z-score approach. We calculated the variance explained using the following formula: 2*MAF* (1−MAF) * ( ( (beta) ∧2) / ( (SD) ∧2) ). We searched for eQTLs in a region bounded by the LIPN1 and TRIB2 genes using expression SNP (eSNP) datasets availably publically or via collaboration including lymphocytes [35], leukocytes [36], leukocytes from patients with Celiac disease [37], lymphoblastoid cell lines (LCL) from children with asthma [38], HapMap LCL from 3 populations [39], a separate study on HapMap CEU LCL [40], peripheral blood monocytes [41], [42], subcutaneous and omental adipose tissue [43], [44], and blood samples [43], 2 studies on brain cortex [41], [45], three large studies of brain regions including prefrontal cortex, visual cortex and cerebellum (Emilsson, personal communication), liver [44], [46], osteoblasts [47], skin [48], and additional fibroblast, T cell and LCL samples [49]. Statistical significance was considered using the association with gene transcript levels as originally described. Adipose tissue was dissected from male C57Bl/6J mice (Jackson Labs) (n = 4) following 20 weeks of ad libitum feeding with a 60% high-fat diet (Research Diets, New Brunswick NJ). High fat feeding was used because many adipose depots (such as pericardial fat) are either not visible or extremely limited in size in standard chow fed mice. Animals were sacrificed and adipose tissues were dissected and frozen in liquid nitrogen. The following adipose depots were dissected: inguinal, axillary, gluteal, brown adipose, epididymal, retroperitoneal, mesenteric, omental, pericardial, and perivascular. All animal experiments were done according to procedures approved by the Institutional Animal Care and Use Committee of Beth Israel Deaconess Medical Center. Total RNA was isolated using TRIzol (Invitrogen, Carlsbad, CA) combined with RNeasy mini-columns, (Qiagen, Valencia, CA) according to the manufacturer' s instructions. For real-time PCR analysis, cDNA was synthesized from RNA using the high capacity cDNA synthesis kit (ABI, Carlsbad, CA). cDNA was used in quantitative PCR containing SYBR-green dye (ABI). mRNA expression levels for each gene were normalized to TBP. Quantitative PCR was performed using an ABI-7900HT PCR machine. In addition to Lipin 1 and Trb2, the expression of the adipocyte markers aP2 and PPARγ2 were also measured. The primer sequences were as follows: aP2, forward 5′ CAT CAG CGT AAA TGG GGA TT 3′ reverse 5′ CCG CCA TCT AGG GTT ATG AT 3′; Lipin 1, forward 5′ CGT ACG TGC GGC TCT GCG AA 3′ reverse 5′ GCT CGG TCG CGT CAA GCT GA 3′; PPARγ2, forward 5′ GCA TGG TGC CTT CGC TGA 3′ reverse 5′ TGG CAT CTC TGT GTC AAC CAT G 3′; TBP, forward 5′ CCC CTT GTA CCC TTC ACC AAT 3′ reverse 5′ GAA GCT GCG GTA CAA TTC CAG 3′; Trib2, forward 5′ CCC GCC CGA GAC TCC GAA CT 3′ reverse 5′ GCA CAG CGC GGA AAA CGT GG 3′.

Pericardial fat is a localized fat depot associated with coronary artery calcium and myocardial infarction. To test whether genetic loci are associated with pericardial fat independent of other body fat depots, we measured pericardial fat in 5,487 individuals of European ancestry. After performing an unbiased screen using genome-wide association, we identified a genome-wide significant signal in our primary meta-analysis at rs10198628 near TRIB2 (MAF 0. 49, p = 2. 7×10-08). This SNP was not associated with visceral fat (p = 0. 17) or body mass index (p = 0. 38). Our findings were robust among multi-ethnic participants from the MESA study, with a combined p-value of 5. 4E-14. We observed TRIB2 gene expression in the pericardial fat of mice. rs10198628 near TRIB2 is associated with pericardial fat but not measures of generalized or visceral adiposity, reinforcing the concept that there are unique genetic underpinnings to ectopic fat distribution.

Abstract Introduction Results Discussion Methods

medicine myocardial infarction clinical genetics genetics and genomics cardiovascular

2012

Genome-Wide Association of Pericardial Fat Identifies a Unique Locus for Ectopic Fat

Following repeated encounters with adenoviruses most of us develop robust humoral and cellular immune responses that are thought to act together to combat ongoing and subsequent infections. Yet in spite of robust immune responses, adenoviruses establish subclinical persistent infections that can last for decades. While adenovirus persistence pose minimal risk in B-cell compromised individuals, if T-cell immunity is severely compromised reactivation of latent adenoviruses can be life threatening. This dichotomy led us to ask how anti-adenovirus antibodies influence adenovirus T-cell immunity. Using primary human blood cells, transcriptome and secretome profiling, and pharmacological, biochemical, genetic, molecular, and cell biological approaches, we initially found that healthy adults harbor adenovirus-specific regulatory T cells (Tregs). As peripherally induced Tregs are generated by tolerogenic dendritic cells (DCs), we then addressed how tolerogenic DCs could be created. Here, we demonstrate that DCs that take up immunoglobulin-complexed (IC) -adenoviruses create an environment that causes bystander DCs to become tolerogenic. These adenovirus antigen loaded tolerogenic DCs can drive naïve T cells to mature into adenovirus-specific Tregs. Our study reveals a mechanism by which an antiviral humoral responses could, counterintuitively, favor virus persistence. Human adenoviruses (HAdVs), of which there may be 85 types (based on serology and genome analyses), typically cause self-limiting respiratory, ocular, and gastro-intestinal tract infections in immunocompetent individuals. After repeated encounters, most young adults generally harbor cross-reactive, long-lived humoral and T-cell responses [1–3] that are thought to work together to efficiently blunt subsequent HAdV-induced morbidity. However, in spite of the robust anti-HAdV immune responses, HAdVs routinely establish decades-long, subclinical infections that are characterized by low level shedding of progeny virions [4,5]. While potential molecular mechanisms by which HAdVs evade the immune response have been proposed [6], we suspected that complementary mechanisms also exist. Of note, in T-cell compromised patients the loss of cellular control of persistent HAdV infection can lead to fulminant and fatal disease [4,5]. It is noteworthy that serological evidence that the patient has been infected by a given HAdV type before hematopoietic stem cell transplantation is predictive of escape from the same HAdV type during immune suppression [7]. While T-cell therapy has shown a notable potential to prevent HAdV disease in immunocompromised patients [8,9], immunoglobulin therapy has had remarkably little impact [4]. Due to omnipresent anti-HAdV antibodies, it is not surprising that immunoglobulin-complexed HAdVs (IC-HAdVs) are detected in some patients with HAdV disease [10–12]. In a broader view, immunoglobulin-complexed viruses can form during prolonged viremia, secondary infections, primary infections when a cross-reactive humoral response exists, and in the presence of antibody-based antiviral therapy. It is important to note that IC-HAdVs are potent stimulators of human dendritic cell (DC) maturation [13,14]. In immunologically naïve hosts, immunoglobulin-complexed antigens are efficient stimulators of antibody and cytotoxic T-cell responses [15]. However, most studies using immunoglobulin-complexed antigens have used prototype antigens that have little impact on their intracellular processing. This is not the case for IC-HAdVs. The endosomolytic activity of protein VI, an internal capsid protein, prevents the canonical processing of the IC-HAdVs by enabling the escape of HAdV capsid and its genome from endosomes into the cytoplasm [14]. In the cytoplasm, the HAdV genomes are detected by absent in melanoma 2 (AIM2), a cytosolic pattern recognition receptor (PRR) [16]. AIM2 engagement of the 36 kb HAdV-C5 genome induces pyroptosis, a pro-inflammatory cell death in conventional DCs [17]. Pyroptosis entails inflammasome formation, caspase 1 recruitment/auto-cleavage/activation, pro-IL-1β processing, gasdermin D (GSDMD) cleavage, GSDMD-mediated loss of cell membrane integrity, and IL-1β release [18,19]. Just as immune responses need to be initiated, suppression of cellular responses are primordial to avoid excessive tissue damage and feature prominently in acute and chronic infection [20–22]. Control of antigen-specific T cells can be mediated in part by peripherally induced antigen-specific regulatory T cells (Tregs) [23], which can favor the establishment of persistent viral infections. Moreover, tolerogenic DCs are required for antigen-specific Treg formation. The variable phenotype and functionality of tolerogenic DCs are globally characterized by a semi-mature profile encompassing cell surface costimulatory molecules, cytokine expression and secretion, and antigen uptake and processing [24,25]. The goals of our studies were to determine how anti-HAdV humoral immunity impacts the cellular response to HAdVs, and whether this might affect persistence. Initially, we found that healthy adults harbor HAdV-specific Tregs. We then demonstrated that IC-HAdV5-challenged human DCs induce a tolerogenic phenotype in bystander DCs. We show that the bystander DCs are capable of taking up and presenting HAdV antigens, and can drive naïve T cells to mature into HAdV-specific Tregs. Our study reveals a mechanism by which an antiviral humoral responses could, counterintuitively, favor virus persistence. Initially, we asked if healthy adults harbor HAdV-specific Tregs and if so, are they capable of dampening anti-HAdV T-cell proliferation. To address these questions, we pre-screened a cohort of healthy individuals using an IFN-γ ELISpot assay for a memory T-cell response to HAdVs using a pool of overlapping HAdV5 hexon peptides (hexon is the major protein in the HAdV capsid). It is important to note that the anti-HAdV T-cell response is not species—or type-specific as the hexon sequence is highly conserved among all HAdVs. While the majority of 58 donors in this assay had a HAdV-specific T-cell response, PBMCs from 11 individuals with a spot forming unit ratio 5-fold greater than mock-treated cells were selected for further analyses. Because inducible Tregs can produce IL-10 in response to their cognate antigen, the ability of HAdV-specific CD4+ T cells to produce IL-10 as well as IFN-γ, TNF, and IL-2 was assessed by multi-parametric flow cytometry. Consistent with our previous results [13], the cytokine profile of HAdV-specific memory CD4+ T cells was dominated by polyfunctional IFN-γ+/IL-2+/TNF+/IL-10- cells (approximately 25% of total HAdV-specific CD4+ T cells) and IFN-γ+/IL-2-/TNF-/IL-10- cells (approximately 20%) (Fig 1A, a representative donor). We then characterized the combinations of the responses and the percentage of functionally distinct populations in all donors (Fig 1B). Each slice of the pie chart corresponds to HAdV-specific CD4+ T cells with a given number of functions, within the responding T-cell population. Of note, IL-10-producing HAdV-specific CD4+ T cells, which were approximately 5% of total, were predominantly IFN-γ-/IL-2-/TNF-. To determine if the IL-10 producing T cells have a Treg phenotype, the expression of conventional Treg markers, CD45RO, CD25, FoxP3, and CD127 [26], were assessed. We found that approximately 8% of the IL-10 producing T cells were CD25+/FoxP3+/CD127dim. By contrast, most of IFN-γ producing HAdV-specific CD4+ T cells harbored a conventional memory phenotype (CD45RO+/FoxP3-/CD25-/CD127+) (Fig 1C). These data demonstrate the presence of HAdV-specific Tregs in healthy adults. To determine if putative HAdV-specific Tregs have regulatory functions, we used PBMCs from 5 individuals that harbored an anti-HAdV T-cell response. The CFSE-labeled PBMCs, or CFSE-labeled PBMCs depleted in CD25-expressing cells, were incubated with HAdV5 or the hexon peptide pool and T-cell proliferation was quantified. We found that depletion of CD25+ cells caused CD4+ cells to proliferate greater than control peptide-challenged CD4+ cells (Fig 1D), suggesting that the HAdV-specific Tregs in the CD25+ population can restrict the proliferative anti-HAdV T cells. Taken together, these data indicate that a fraction of HAdV-specific CD4+ T cells harbors an inducible Treg phenotype, and that healthy adults likely have CD25+ Tregs that dampen the proliferation of HAdV-specific T cells. A prerequisite for antigen-specific Treg formation is the presence of antigen-presenting tolerogenic DCs [27,28]. Because the cellular profile of HAdV5-challenged DC [29] is inconsistent with that of tolerogenic DCs [29], we asked if IC-HAdV5 could be involved in the generation of HAdV-presenting tolerogenic DCs. When HAdV5 is mixed with neutralizing antibodies from human sera, 200 nm-diameter complexes are formed that induce DCs to undergo pyroptosis, or, if the DC does not die, a hypermature profile [13,14]. As these profiles are also inconsistent with that of tolerogenic DCs, we hypothesized that it was not due to IC-HAdV5-activated DCs, but rather an effect on bystander DCs. To assess the impact of IC-HAdV5-induced pyroptosis and DC maturation on bystander DCs we developed a transwell assay (see S1A Fig for schematic). Briefly, CD14+ monocytes isolated from fresh buffy coats were induced to differentiate into immature DCs for 6 days. Immature DCs seeded in 12-well plates were mock-treated, challenged with bacterial lipopolysaccharides (LPS) as a generic control for DC reactivity, HAdV5, IgGs, or IC-HAdV5 (these cells will be referred to “direct DCs”). At 6 h post-challenge, a transwell insert was added and naive immature DCs (bystander DCs) from the same donor were seeded in the upper chamber (see S1B–S1D Fig for controls concerning transfer of HAdV5 particles between chambers and cell death). Twelve hours after adding the bystander DCs to the upper chamber, the direct and bystander DCs were collected and assayed as described below. Compared to bystander DCs stimulated by direct DCs challenged with IgG or HAdV5, bystander DCs stimulated by IC-HAdV5-challenged DCs increased their cell surface levels of the maturation/activation markers CD80, CD83, CD86 (Fig 2A), CD40, and MHC II (S2A Fig). The level of CD86 on bystander DCs tended to increase as the number of IC-HAdV5 particles increased during the stimulation of the direct DCs (S2B Fig). The cell surface increase of CD86 and CD83 was also accompanied by an increase in total (cell surface + intracellular) CD86 and CD83 levels (Fig 2B). Together, these data demonstrate that IC-HAdV5-challenged DCs enhanced the synthesis and cell surface expression of maturation/activation markers on bystander DCs. To characterize bystander DC functional capabilities we used an 84-plex inflammatory cytokine, chemokine and their receptor mRNA array to quantify transcriptional changes (see S2C Fig for the list of mRNAs that gave unique amplification profiles). Stimulation of bystander DCs with the milieu from HAdV5-challenged DCs (without IgGs) led to notable increases (>50 fold) in mRNA levels of Th1/Th17 cell activation/differentiation markers (e. g. CXCL9, CXC10 & CXC11) (see S2C Fig for all data and Fig 3A left hand columns for selected data). By contrast, the bystander DC response to the IC-HAdV5-challenged DCs was greater with respect to the number of mRNAs altered (>20) and magnitude (up to 10,000-fold increase) (Fig 3A right column, and S2C Fig middle column). Of particular relevance was the lack of TNF mRNA by bystander DC because tolerogenic DCs should not, a priori, secrete TNF. To better understand the transcriptional responses of the different conditions, we applied a principal component analysis (PCA) to find patterns in these data sets. We found that two principal components (see Materials & Methods for genes in the F1, F2, and F3 axes) explained 89% and 39% of the total information, respectively, and each stimulus is distinguishable from the others (Fig 3B). Because a cell infected by one HAdV particle could produce >104 virions ~36 h later, local and global HAdV levels, as well as IC-HAdV formation, are dynamic at early stages of infection. Of note, IC-HAdV5 causes a dose-dependent induction of pyroptosis in direct DCs [14]. We therefore extended the mRNA array analyses by quantifying dose-dependent response of bystander DCs. Using RT-qPCR we analyzed TNF, IFNβ and CXCL10 (Fig 3C) and IL1β, IL12 (p40), CCL3 and IL6 (S2D Fig) mRNA levels. In all cases the transcriptional response of bystander DCs varied depending on the IC-HAdV5 challenge dose. These data suggest that the bystander DC response is linked to the percentage of direct DCs undergoing pyroptosis [14]. To characterize time-dependent transcriptional changes in direct DC and bystander DCs, we compared mRNA levels of TNF, IFNβ (Fig 3D), Mip-1α and IL6 (S3 Fig, which also includes dose-dependent response). Globally, mRNAs that code for pro-inflammatory molecules were 2 to 10-fold greater in direct DCs than in bystander DCs. In addition, only IL1β and Mip1α mRNA levels changed significantly (p < 0. 01) over time. These data demonstrate that bystander DCs have a semi-mature transcriptional profile, which is linked to DC pyroptosis, and lack noteworthy levels of TNF mRNA [30]. To examine the events downstream the transcriptional response, we quantified the secreted cytokine from direct and bystander DCs. Because proteins can readily diffuse across the transwell membranes, bystander DCs were removed from the upper chamber 12 h post-challenge, rinsed, and then placed in a separate well with fresh medium for 9 h before collecting the medium. The direct DC medium was collected at 12 h post-challenge, or after a wash at 12 h and then collected 9 h later (21 h) to compare conditions similar to that used for bystander DCs (see Fig 4A for schematic). Challenging DCs with HAdV5 alone had a modest effect on their secretome with the exception of a 5- to 10-fold increase in TNFSF10, and CXCL9 & 10 levels (Fig 4B, second column from the left). By contrast, IC-HAdV5-challenged DCs responded with increases of >15 fold in approximately half of the cytokines (Fig 4B, middle columns). These data are consistent with previous results showing the robust maturation of IC-HAdV5-challenged DCs [13,14]. HAdV5-challenge DCs that were rinsed 12 h post-stimulation had overall lower cytokine levels than prior to washing, but TNFSF10, CXCL9, CXCL11, and CCL5 levels remained robust (Fig 4C, middle columns). Interestingly, instead of a positive correlation between the cytokine secretion and the IC-HAdV5 dose, we found that as the IC-HAdV5 dose increased, the cytokines secreted by direct DCs tended to decrease (Fig 4C, middle columns). Using HAdV5-challenged DCs (without IgGs) to generate bystander DCs, we found that the latter secreted 3- to 12-fold higher levels of TNFSF10, CCL5, CXCL9, CXCL10 and CXCL11 compared to bystander DCs exposed to the medium from IgG-challenged DCs (Fig 4D, second column). Similarly, when bystander DCs were generated using IC-HAdV5-challenged DCs, the level of the above five cytokines also increased. In addition, three chemokines involved in immune cell recruitment (CCL15, CCL20, and CCL2) increased >3 fold. Consistent with the transcriptome analyses, we did not find a notable dose-dependent effect on bystander DCs when direct DCs were incubated with increasing IC-HAdV5 particles (Fig 4D, middle columns). Together, these data suggest that the release of pathogen-associated molecular patterns (PAMPs), danger-associated molecular patterns (DAMPs), and/or the increased levels of cytokines secreted by a greater number of DCs that do not undergo pyroptosis, are key factors in bystander DC maturation. In addition, the environment created by IC-HAdV5 induces a semi-mature cytokine secretion profile in bystander DCs. To determine how cytokines and pyroptosis impact bystander DCs, we used a combination of drugs and mutant HAdVs to selectively modify the environment created by IC-HAdV5-challenged DCs. To determine the impact of IL-1β, direct DCs were pre-treated with ZVAD, a pan-caspase inhibitor that blocks caspase 1 auto-cleavage and pro-IL-1β processing. Importantly, ZVAD has no effect on TNF and canonical protein secretion (S4A Fig and reference [14]). We found that blocking IL-1β production by direct DCs reduced bystander DC maturation as demonstrated by their lower levels of CD86 and CD83 (Fig 5A and 5B). We then used brefeldin A to block ER to Golgi-mediated cytokine secretion in direct DCs (see S4B Fig for controls). Of note IL-1β release is not significantly affected by brefeldin A, (S4C Fig). In brefeldin A-treated IC-HAdV5-challenged DCs the levels of CD83 and CD86 did not change markedly (Fig 5C), while the bystander DCs responded with lower levels of CD83 and CD86 (Fig 5D). Next, we generated ICs using AdL40Q [31], an HAdV5 capsid containing a mutated protein VI that attenuates endosomolysis. While IC-AdL40Q poorly induces pyroptosis in direct DCs [14], they secrete levels of TNF that are similar to IC-HAdV5-challenged DCs. Furthermore, IFNβ and IL1β mRNA levels are lower [14]. We found notably lower levels of CD86 and CD83 on bystander DCs following stimulation with the response from IC-HAdV5 versus IC-AdL40Q-challenged DCs. In addition, the reduced maturation/activation effects were only modestly altered by increasing the IC-AdL40Q dose (Fig 5E). Together, these data demonstrate a role for pyroptosis-associated factors in the maturation of bystander DCs. We then compared cytokine mRNA levels in bystander DCs stimulated by HAdV5-, IC-AdL40Q-, or IC-HAdV5-challenged DCs (Fig 5F and S5A–S5C Fig). Consistent with the phenotype, the transcriptional responses of bystander DCs to both ICs were globally higher than to HAdV5 alone. The bystander DC transcriptional response to IC-AdL40Q-challenged DC milieu was generally lower than in IC-HAdV5-challenged DCs, and it was qualitatively distinguishable as determined by PCA (Fig 5G). We then assessed the effect of pyroptosis using IC-Ad2ts1. Ad2ts1 has a hyper-stable capsid due to a mutation in protease that results in failure to process the capsid pre-protein [32,33]. We previously showed that IC-Ad2ts1 poorly induces DC pyroptosis, likely because the HAdV genome does not escape from the capsid and therefore does not nucleate AIM2 (see reference [14] and S5D–S5F Fig for Ad2ts1 controls). Of note, TNF levels are comparable in DCs challenged with IC-Ad2ts1 or IC-HAdV5 [14]. Here, we found that IC-Ad2ts1-challenged DC induced an increase of CD86 cell surface levels on bystander DCs (Fig 5H). Together, these data demonstrate that cytokines, DAMP, and PAMPs play a role in the activation and semi-maturation of bystander DCs. To characterize how bystander DCs are activated, we focused on Toll-like receptor 4 (TLR4). TLR4 is a multi-functional cell surface PRR that can directly or indirectly (by forming a complex with MD-2, CD14, or other PRRs) be activated by extracellular viral components (PAMPs) and, under inflammatory conditions, extracellular high-mobility group box 1 and heat shock proteins (DAMPs) [34–36]. Of note, MD-2 acts as a co-receptor for recognition of both exogenous and endogenous ligands [37–40]. While TLR4 does not bind to, or become activated by, HAdV5 alone [41], TLR4 might be activated by PAMPs or DAMPs that interact directly with the HAdV5 capsid. We therefore used TAK-242 to disrupt TLR4 signaling in bystander DCs (see S6 Fig for TAK-242 control). As readouts, we used the upregulation of TNF and IL1β mRNAs, and activation/maturation cell surface markers. When TLR4 signaling was blocked in bystander DCs stimulated by the IC-HAdV5-challenged DC milieu, there was a significant (p < 0. 05) decrease in IL1β mRNA levels and 2-fold decrease of TNF mRNA (Fig 6A). CD83 and, to a lesser extent, CD86 levels were also reduced (Fig 6B). These data suggest that bystander DCs use TLR4 to detect PAMPs and DAMPs released by IC-HAdV5-challenged DCs, leading to changes in bystander DC maturation. Immature DCs survey the extracellular environment by random phagocytosis. Once PRRs are engaged, DC maturation is accompanied by decreased uptake of fluid phase molecules [42]. Of note, a functional hallmark of tolerogenic DCs is their ability to retain some antigen uptake properties. To address the functional maturation of IC-HAdV5-challenged DCs and bystander DCs, we incubated cells with FITC-labeled dextran and quantified uptake by flow cytometry. We found that phagocytosis was modestly decreased in direct DCs stimulated with HAdV5 or LPS (Fig 7A). By contrast, IC-HAdV5-challenged DC phagocytosis was near background levels, consistent with complete maturation (see S7 Fig for controls) [29]. While bystander DCs had reduced phagocytosis when created by IC-HAdV5-challenged DCs, the bystander DCs still took up 17-fold more FITC-dextran than background levels (Fig 7B). These functional data are consistent with semi-mature, tolerogenic DC profile. While tolerogenic DCs can induce, recruit, and maintain Treg homeostasis, tolerogenic DCs can also create a feedback loop to promote their own generation [43]. Because monocytes are recruited to sites of inflammation [44,45], we compared the recruitment capabilities of direct DCs and bystander DCs (see S8 Fig for setup and controls). Unexpectedly, we found that IC-HAdV5-challenged DCs inhibited monocyte recruitment in an IC-HAdV5 dose-dependent manner (Fig 8A & 8B). Of note, the inhibition was abrogated when the IC-HAdV5-challenged DCs were washed, suggesting that inhibitory factors were generated <3 h post-IC-HAdV5 challenge (Fig 8B). To determine if pyroptosis-related factors (i. e. IL-1β, DAMPs and PAMPs) are responsible for the inhibition of monocyte recruitment, we used ZVAD and IC-AdL40Q to reduce pyroptosis. ZVAD, which prevents caspase 1 auto-activation, IL-1β maturation, and GSDMD-associated pore formation, modestly increased monocyte recruitment induced by IC-HAdV5-challenged DC (Fig 8C & 8D). In contrast to the IC-HAdV5-challenged DC response, the IC-AdL40Q-challenged DC response significantly (p < 0. 05) increased monocyte recruitment, in a dose-dependent manner (Fig 8E & 8F). These data suggest that DC pyroptosis inhibits monocyte recruitment. We then examined the ability of bystander DCs to recruit monocytes. In contrast to IC-HAdV5-challenged DCs, bystander DCs promoted monocyte recruitment (Fig 8G). These data are consistent with the bystander DC milieu containing more chemoattractants (Fig 5). There was also a trend towards greater recruitment when higher IC-HAdV5 doses were used to stimulate the direct DCs. Once monocytes migrate into an inflammatory environment they acquire distinct phenotypic and functional profiles [46]. One phenotypic hallmark of monocyte differentiation is CD14, which is high on monocytes and macrophages, but lower on DCs. We therefore characterized migrating and static monocytes for CD14 and CD86 levels at 24 and 72 h (see schematic at the left of each panel in Fig 8H–8J for the times and location of cells, and S8 Fig for controls). At 24 h the level of CD14 on monocytes that had migrated into the bystander DCs environment did not change markedly, while CD86 levels were lower (Fig 8H). At 72 h the recruited monocytes had two distinct populations based on CD14 levels (Fig 8I). The decrease in CD14 levels suggested that they differentiated into DCs, while the CD86 levels suggest the maintenance of an immature phenotype. In addition, monocytes recruited by bystander DCs had increased CD14 levels. By contrast, CD86 levels decreased on monocytes in the upper chamber (bottom chamber containing bystander DCs) (Fig 8J). Together, these data demonstrate that DCs challenged with IC-HAdV5 inhibit monocyte recruitment. Monocytes recruited to the bystander DC environment was abetted by pyroptosis of the direct DCs. Recruited monocytes had reduced CD14 levels, possibly due the engagement and internalization of TLR4/CD14 complexes. Monocyte-DC contact also appeared to favor the increase in cell surface levels of activation/maturation markers. We concluded that the dynamic environment created by bystander DCs is consistent with a feed-forward loop to foster tolerogenic DCs. A functional characteristic of tolerogenic DCs is that they can take up and present antigens. Therefore, we asked if some of the bystander DCs generated in our ex vivo model are capable of inducing proliferation of HAdV5-specific memory T cells. We used IC-HAdV-challenged DC to generate bystander DCs, which were then added to CFSE-labeled PBMCs. Seven days post-incubation we found that CD3+/CFSElow cells harbored memory T cell markers (CD45RO+/CD45RA-) (Fig 9A). These data are consistent with the potential of some of the bystander DCs to maintain fluid phase uptake and subsequent presentation of HAdV5 antigens to memory T cells. In addition to antigen presentation, tolerogenic DCs can induce naïve CD4+ cells to become Tregs. To address this functional characteristic, bystander DCs were generated and incubated with autologous naïve CD4+/CD45RAhigh cells for 3 or 7 days. The T cells were then assayed by multi-parametric flow cytometry for CD4, CD25, CD127 and FoxP3, markers that are indicative of Tregs. While activated T cells transiently express FoxP3 (S9 Fig), the relatively low-level does not result in acquisition of suppressor activity [27]. By contrast, stable and high levels of FoxP3 can be used to identify bona fide Tregs. At day 3, naïve T cells expressed Treg markers in all conditions (except mock-treated direct DCs) (Fig 9B). At day 7, the number cells with Treg phenotype was near background following incubation in the milieu of mock-, IgG-, or HAdV5-challenged direct DC (Fig 9C). By contrast, bystander DC created from IC-HAdV-direct DCs had a significant (p < 0. 05) increase in cells with a Treg profile. These data demonstrate that bystander DCs can induce naïve CD4 into cells with a Treg phenotype, further supporting our conclusion that they are tolerogenic DCs. As shown in Fig 1D, healthy adults harbor CD25+ cells can inhibit HAdV-specific CD4+ cell proliferation. We therefore asked if the tolerogenic bystander DCs generated in our ex vivo assay could induce the production of HAdV-specific Tregs. To address this question we isolated PBMCs, CD14+ monocytes, and naïve CD4+ T cells from 3 donors that harbored anti-HAdV memory T cells (see S10 Fig for flow chart). Briefly, monocytes were used to create direct DCs that were incubated with IC-HAdV-C5. Bystander DCs were generated as previously described. VPD 450-labeled naïve CD4+ T cells were incubated with bystander DCs to generate Tregs. VPD 450low/CD4+/CD25+ cells (600 to 5,000 cells) were isolated by FACS and mixed with CFSE-labeled PBMCs ± hexon peptides. We found that the ex vivo generated Tregs from 3/3 donors reduced the proliferation of anti-HAdV T cells (CFSElow/CD4+) (Fig 9D). These data demonstrate that HAdV-specific Tregs can be generated via bystander DCs. HAdV infections lead to multifaceted, robust, long-lived cellular and humoral responses in most young immunocompetent adults. Nonetheless, several HAdV types somehow circumvent immune surveillance to establish persistent infections. It is well documented that HAdV neutralizing antibodies are type specific, while the anti-HAdV cellular response is cross-reactive [1,3, 8,47–49]. In addition, it is the anti-HAdV cellular response that protects us from reactivation of persistent infections. The dichotomy between the two arms of the adaptive immune response led us to address how anti-HAdV antibodies influence anti-HAdV T-cell responses. In this study, we initially asked if healthy adults harbor HAdV-specific Tregs, which would be indicative of a path towards HAdV persistence. We then explored how tolerogenic DCs and HAdV-specific Tregs could be generated. We previously showed that IC-HAdV5s are internalized by, and aggregate in, DCs [14]. Following protein VI-dependent endosomal escape of the capsid, the viral genome is engaged by AIM2 in the cytoplasm. AIM2 nucleation induces ASC (apoptosis-associated speck protein containing a caspase activation/recruitment domain) aggregation, inflammasome formation, caspase 1 auto-activation, pro-IL-1β and GSDMD cleavage, and GSDMD-mediated loss of cell membrane integrity. Here we demonstrate that the pyroptotic environment induced by IC-HAdV5 plays a significant role the creation of tolerogenic bystander DCs. We further show that some of these bystander DCs can induce HAdV-specific memory T cells to proliferate, and/or drive naïve CD4 cells towards a Treg phenotype. The Tregs generated in this ex vivo assay are capable of inhibiting the proliferation of anti-HAdV T cells. We therefore propose that an antiviral humoral responses can, counterintuitively, abet HAdV persistence. Our assays using a human pathogen, naturally occurring human antibodies and primary blood-derived human cells address one possible immune cell-based mechanisms of adenovirus persistence. Yet, ex vivo results cannot unequivocally show causality. Host-pathogen-based studies have often used mice to address questions underlying disease-immune relationships. However, the impact of HAdVs on human and mouse DCs is notably different. Furthermore, we are unaware of studies directing addressing the impact of murine adenovirus (MAV) on murine DCs. In 1964, D. Ginder showed that a MAV can cause persistent infections for 10 weeks in outbred Swiss mice [50]. K. Spindler and colleagues then showed that MAV-1 infections persist for at least 55 weeks in outbred Swiss mice [51]. In addition, Spindler and colleagues demonstrated that in contrast to humans, mice that lack B cells are highly susceptible to MAV-1 infection, while mice that lack T cells are not susceptible [52]. In light of our results, the question could be raised as to whether anti-MAV-1 antibodies are needed to generate Tregs to reduce the potential impact T-cell induced immunopathology [27]. To address this one could use a single pre-injection of sera from MAV-1-challenged mice into B-cell deficient mice and quantify disease progression. Using nonhuman primates (NHPs) to address the dichotomy between the two arms of the adaptive immune response to adenoviruses is likely a more informative option, but use of NHPs entails unique challenges when it comes to pre-existing exposure to their own set of adenoviruses. Nonetheless, Miller and colleagues showed that rhesus macaques harboring a neutralizing antibody response against a HAdV5 host-range mutant, and then re-challenged with the same virus, respond with a significant increase in circulating Tregs [53]. These in vivo observations, which hinge on pre-existing HAdV5 neutralizing antibodies, are consistent with our proposed mechanism. One also needs to take into account the dynamic, recurrent exposure to multiple HAdV types during childhood and adolescence. Although our study focused on HAdV5, a relatively common species C HAdV, we believe that recurrent exposure provides numerous opportunities for the formation of IC-HAdVs, from multiple HAdV types, and the impetus to form cross-reactive HAdV Tregs. Our data also complement the mechanism for HAdV persistence described by Hearing and colleagues [6]. Using human cell lines, they showed that IFN-α and IFN-γ production block HAdV5 replication via an E2F/Rb transcriptional repression of its E1A immediate early gene [54]. The E1A gene product is essential for activating expression of the other early genes and reprogramming the cell into a state that allows virus propagation. Of note, type 1 IFN secretion is significant from IC-HAdV5-challenged DCs and may allow HAdVs (including those that are covered with non-neutralizing Abs) to be taken up by neighboring cells to establish persistent infections. Mechanisms by which DCs promote tolerance include induction of Tregs, the inhibition of memory T-cell responses, T-cell anergy, and clonal deletion [24–26]. The semi-mature phenotype of tolerogenic DCs provide insufficient stimulatory signals and drive naïve T cells to differentiate into Tregs rather than effector T cells [55]. The global anti-viral response by DCs acts via a combinatorial cytokine code to direct the response of neighboring immune cells. The cytokine profile produced by the IC-HAdV5-challenged DCs and bystander DCs is noteworthy, particularly in the context of the combination and dose that influences activation of other immune cells. Recently, a biochemical and functional chemokine interactome study suggested that several chemokines form heterodimers that have unique functions in certain conditions [56]. Based on these interactome data, we plotted the possible combinations that could influence the direct and bystander DCs in our assays (S11 Fig). What impact these heterodimers could have on HAdV persistence will require future study, in particular because we did not find notable levels of TGFβ secreted by direct or bystander DCs. More than other cytokine families, the IL-1 family may be primordial because it is tightly linked to IC-HAdV-induced DC pyroptosis. Indeed, the intracellular domain of the IL-1R1 shares similar signaling function properties with TLRs. In general, IL-1β release from monocytes is tightly controlled; less than 20% of the total pro-IL-1β precursor is processed and released. IL-1β also increases the expression of intercellular adhesion molecule-1 and vascular cell adhesion molecule−1, which, together with the chemokines, promote the infiltration of cells from the circulation into the extravascular space and then into inflamed tissues [57]. While circulating monocytes do not constitutively express IL1β mRNA, adhesion to surfaces during diapedesis induces the synthesis of large amounts that are assembled into large polyribosomes primed for translation [58]. At least two aspects of the IC-HAdV-induced DC immune response that remain unknown are the impact of neutrophils and the phenotype/function of recruited monocytes. Neutrophils are pertinent because they can secrete/release proteinase 3 (PR3), elastase, cathepsin-G, chymase, chymotrypsin, and meprin α or β, which can process extracellular pro-IL-1β into its active form [59,60]. In addition, IC-HAdVs activate neutrophils (L-selectin shedding) via Fc receptors and complement receptor 1 interactions [61]. Moreover, neutrophils are a major source for anti-microbial peptides (e. g. , defensins and LL-37) and proteins (e. g. lactoferrin) for which a pro- or anti-viral roles in HAdV infection has been proposed [62]. With respect to the phenotype/function of recruited monocytes, Ly6Chi monocytes [63], which suppress T-cell proliferation during HAdV-induced inflammation [64], may also impact the creation of HAdV antigen-presenting tolerogenic DCs and HAdV-specific Tregs. The dynamic equilibrium between recurrent HAdV infections and IC-HAdV formation, DC maturation/pyroptosis, recruitment and generation of bystander DC, and Tregs production/activation, likely starts in childhood and develops nonlinearly over decades. While it is hard to argue that the generation of persistent infections is not beneficial to the pathogen, it is possible that the sustained anti-HAdV cellular and humoral responses partially shield a healthy host from infections by other pathogens (e. g. hepatitis C virus [65]) or the related immune-induced tissue damage [66]. Avoiding chronic tissue damage is particularly important because, as mentioned previously, HAdVs infect the eye, respiratory and gastrointestinal tracts. However, in a T-cell compromised host IC-HAdV-induced pyroptosis of FcγR+ cells (neutrophils, monocytes, macrophages, DCs) may also prime the host for HAdV-disseminated disease. In summary, our findings suggest a mechanism by which humoral immunity to HAdV fosters tolerance. Understanding this complex virus-host interplay may enable us to identify high risk patients undergoing immunosuppression and develop therapies to treat disseminated HAdV-disease [67,68]. Blood samples from anonymous donors (~120 from the Etablissement Français du sang, Montpellier, France, and 58 from Lausanne University Hospital/CHUV) were used during this study. All donors provided written informed consent. DCs were generated from freshly isolated CD14+ monocytes in the presence of 50 ng/ml granulocyte-macrophage colony-stimulating factor (GM-CSF) and 20 ng/ml interleukin-4 (IL-4) (PeproTech, Neuilly sur Seine, France) [3]. DC stimulations were performed 6 days post-isolation of monocytes. THP-1 cells purchased from ATCC (TIB-202) were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS). Similar to DCs, THP-1 cells were differentiated into DCs using 50 ng/ml GM-CSF and 20 ng/ml IL-4 for 6 days. Adβgal is a ΔE1/E3 HAdV5 vector harboring a lacZ expression cassette [69]. AdL40Q is an HAdV5-based vector with a leucine to glutamine mutation of an amino acid in protein VI that decreases its membrane lytic activity [31]. Alexa555- and Alexa488-HAdV5 were generated from Adβgal by using an Alexa555 or Alexa488 Protein Labeling Kit (Life Technologies, Villebon-sur-Yvette, France) as previously described [70]. Ad2ts1 harbors a mutation in protease and results in several unprocessed capsid proteins and a hyper-stable capsid [71]. All HAdV viruses/vectors were produced in 293 or 911 cells and purified by double banding on CsCl density gradients as previously describe [14]. Vector purity typically reaches >99%. HAdV concentrations (physical particles/ml) were determined as previously described [72]. The hexon peptide pool (PepTivator AdV5 hexon, Miltenyi) is overlapping sequences of the HAdV5 hexon protein. Anti-human CD4-PE (cat 300508) was from BioLegend. Anti-human CD83-FITC (cat 556910), anti-human HLA-ABC-PE (cat 555553), anti-human HLA-DR-PE (cat 555812), anti-human CD80-FITC (cat 557226), anti-human CD86-APC (cat 555660), anti-human CD25-PE (cat 555432), anti-human CD127-FITC (cat 561697), anti-human CD4-PE-Cy7 (BD) (cat 348809), anti-TNF-PE-Cy7 (cat 557647), anti-human IL-2-PE (cat 559334), anti-human CD3-APC-H7 (cat 560176), anti-human CD4-CF594 (cat 5562281), anti-human CD4-PB (pacific blue), anti-human CD8-BV605 (cat 564116), anti-human CD8-PerCP-Cy5. 5 (cat 341050), anti-human IFN-γ–AF700 (cat 557995), anti-human CD35-PE-Cy7 (cat 557741), anti-human IL-10 (cat 554707) were from Becton Dickinson, Pharmigen. Anti-human Foxp3-APC (cat 17-4776-41) was from eBioscience. Anti-human CD14-PE (cat A07764) was from Beckman Coulter. Anti-human CD45RO-APC/Cy7 (cat 304227), anti-human CD45RA-PE (cat 304205), anti-human CD3-APC (cat 300411), and anti-human CD40-APC (cat 313008) were from BioLegend). Anti-human IL-10-BV421 (cat 501421), anti-human CD45RA-BV711 (cat 304137), anti-human CD127-BV711 (cat 351327) from Biolegend, anti-human CD45RO-ECD (cat IM2712) (BC), anti-human FoxP3-PE (cat 12-4777-42), from eBiosciences. DCs (4 x 105 in 400 μl of complete medium) were incubated with HAdV5 or IC-HAdV5 (or IC) (2 x 104 physical particles (pp) /cell, unless otherwise indicated) for the indicated times. IC-HAdV5s were generated by mixing the virus (8 x 109 physical particles) with 2. 5 μl of IVIg (human IgG pooled from 1,000 to 50,000 donors/batch) (Baxter SAS, Guyancourt, France) for 15 min at room temperature. IVIg is used in patients with primary or acquired immunodeficiency as well as autoimmune diseases. Z-VAD-FMK 20 μM (ZVAD) was added 2 h before stimulation. Brefeldin A was used at 3 μg/ml after 6 h stimulation or for the same time with stimulation. DCs (1. 5 x 106 in 1. 5 ml of full media) were incubated ± LPS 100 ng/ml, HAdV5, and IgG in the lower compartment of the well (12 mm diameter polyester membranes with 0. 4 μm pores; (Corning, Bagneaux-sur-Loing, France). After 6 h incubation, fresh immature DCs (6 x 105 in 600 μl of media) were added to the upper compartment and are referred to as bystander DCs. TAK-242 was added to DCs 1 h pre-challenge. Expression levels of cytokine and chemokine genes were evaluated using RT-qPCR assays. Total RNA was isolated from cells using the high pure RNA isolation Kit (Roche, Berlin, Germany) with a DNase I treatment during the purification and subsequent elution in 50 μl of RNase-free water (Qiagen, IN, USA). Reverse transcription was performed with the superscript first-strand synthesis system (Invitrogen) using 10 μl of total RNA and random hexamers. The cDNA samples were diluted 1: 20 in water and analyzed in triplicate using a LightCycler 480 (Roche, Meylan, France). SYBR green PCR conditions were as follows: 95°C for 5 min and 45 cycles of 95°C for 15 s, 65°C or 70°C for 15 s, and 72°C for 15 s using GAPDH as a standard. See S2 Table for primers sequencers. Relative gene expression levels of each respective gene were calculated using the threshold cycle (2-ΔΔCT) method and normalized to GAPDH mRNA levels. Expression levels of cytokine and chemokine mRNAs were analyzed using PCR array assays. Total RNA was isolated from cells using the High Pure RNA isolation Kit (Roche, Berlin, Germany) with a DNase I treatment during the purification and elution in 50 μl of RNase-free water (Qiagen). Reverse transcription was performed with the RT2 First strand Kit (Qiagen, Courtaboeuf, France), and the cDNA samples were analyzed in duplicate using a RT2 Profiler TM PCR array (Qiagen). SYBR green PCR conditions were 95°C for 10 min and 40 cycles of 95°C for 15 s, and 60°C for 1 min using 84 human inflammatory and receptor genes. The potential mRNAs were chosen and then confirmed by RT-qPCR. The genes that contributed in each axis in the PCA were as follows: F1 = CCL1,2, 4,5, 7 13,15,17,20,22, CSF1, CX3CL1, CXCL 1 to 3,5, 8 to 11, FASLG, IFNG, IL10RA, IL10RB, IL15, IL1a, IL1b, IL7, NAMPT, TNFSF4,10,11,13,13B, and VEGFA. F2 = AIMP1, C5, CCL1,2, 13,17,23, CRR1,2, 3,4, 5, CSF1, CX3CR1, CXCR2, IL10RA, I10RB, IL15, LTA, LTB, MIF, SPP1, TNF, TNFSF4,10,11,13, and 13B. F3 = CCL17,23, CCR5, CX3CR1, IL10RA, IL5, IL9, MIF, and OSM. Surface levels of CD83, MHCII, CD80, CD40, and CD86 were assessed by flow cytometry. Cell membrane integrity was assessed by collecting cells via centrifugation at 800x g; the cell pellets were then resuspended in PBS containing 10% FBS, propidium iodide (PI) (Sigma-Aldrich, Missouri, USA), or 7-aminoactinomycin D (7AAD) (Becton-Dickinson, New Jersey, USA). The cell suspension was incubated for the indicated times and analyzed using a FacsCalibur flow cytometer (Becton-Dickinson) and FlowJo software. Surface and intracellular levels of CD83 and CD86 (total protein) were stained with a BD Cytofix/Cytoperm Fixation/Permeabilization Kit, and then measured by flow cytometry. To assess cell membrane integrity, the cells were collected and centrifuged at a speed of 800x g; the cell pellets were then resuspended in PBS, 10% FBS, PI (Sigma), or 7AAD and analyzed on a FacsCalibur flow cytometer (Becton-Dickinson) and FlowJo software. Monocyte migration was evaluated using a 5. 0 μm-diameter pore transwell system (Corning, Bagneaux-sur-Loing, France). Monocytes (2 x 105 in 200 μl of full media) were added into inserts and DCs or DCs (7. 5 x 105 in 750 μl of full media) and ± LPS (100 ng/ml), HAdV5, HAd555, or IgG in the lower wells. Monocytes were stained by carboxy-fluorescein diacetate-succinimidyl ester (CFSE) (Molecular Probes, Eugene, OR, USA) (CellTrace CFSE Cell Proliferation Kit). DCs incubated for 30 min with HAd555 or HAdV5 and IgG in the lower chamber were or were not washed in medium before adding the stained CFSE monocytes. After 3,6, and 24 h incubation at 37°C, the cells in the upper and lower compartment were detected quantified using a FacsCalibur flow cytometer (Becton-Dickinson) and FlowJo software. Supernatant from the cells were collected and cytokine secretion was measured by ELISA and Luminex assays. The secretion of TNF and IL-1β was quantified by ELISA using an OptEIA human TNF ELISA Kit (Becton Dickinson) and human IL-1β/IL-1F2 DuoSet ELISA (R&D Systems, Lille, France) following the manufacturer’s instructions. Additionally, 22 other cytokines and chemokines were detected by Luminex using a Bio-plex pro human chemokine, cytokine kit (Bio-Rad, Marnes-La-Coquette, France) following the manufacturer' s instructions. PBMCs were isolated using standard gradient separation techniques. Half were CD25+-depleted, using anti-CD25 in a human CD4+CD25+CD127dim/- Regulatory T Cell Isolation Kit II and MACS separation system. PBMCs were washed and suspended in PBS for labeling with CFSE or Violet Proliferation Dye 450 (VPD 450) (BD Horizon, Le Pont de Claix, France) at a final concentration of 2. 5 μM or 1 μM, respectively, for 3 min at room temperature. Labeling was terminated by the addition of fetal calf serum (FCS) (40% of total volume). PBMCs ± CD25+ were stained with CFSE and cultivated in 96-well U-bottom plates; (cell concentration 1 x 106/ml and a final volume of 200 μl; PBMC CD25+/ PBMC CD25- ratio 1: 10). HAdV5 hexon peptides (PepTivator, Miltenyi, Paris, France) were added at 0. 3 nmol. On days 3 and 5 the cells were split and IL-2 was added (final concentration 100 U/ml). Cells were analyzed on a FACS Canto II using FlowJo software. Naïve CD4+ T cell were isolated using naïve CD4+ T Cell Isolation Kit II and MACS separation system. DCs indirectly activated for 12 h with LPS, HAdV5, IC-HAdV5 and IgG, and then were co-cultured with CD4+ naïve T cells labeled VPD450 (with ratio bystander DCs/ T cells is 3: 1) in RPMI 1640 supplemented with 10% FCS and IL-2 (Proleukin 18 x 106 IU, CA, USA) (100 U/ml) for 3 or 7 days. Recombinant IL-2 was added on day 3 and day 5. CD25, CD127, and FoxP3 levels were quantified by flow cytometry using FACS Canto II. All experiments were performed at least in duplicate a minimum of three independent times, and the results are expressed as mean ± SEM unless otherwise stated. The statistical analyses were performed using the Student’s t-test unless otherwise stated. A p value < 0. 05 is denoted as significant. Statistical analyses of the global cytokine profiles (pie chart) were performed by partial permutation tests using the SPICE software.

While numerous studies have addressed the cellular and humoral response to primary virus encounters, relatively little is known about the interplay between persistent infections, neutralizing antibodies, antigen-presenting cells, and T-cell responses. Our studies suggests that if adenovirus–antibody complexes are taken up by professional antigen-presenting cells (e. g. dendritic cells), the DCs can generate an environment that causes bystander dendritic cells to become tolerogenic. These tolerogenic dendritic cells favors the creation of adenovirus-specific regulatory T cells. While this pathway likely favors pathogen survival, there may be advantages for the host also.

Abstract Introduction Results Discussion Materials and methods

blood cells flow cytometry innate immune system medicine and health sciences immune cells immune physiology cytokines immunology messenger rna vertebrates animals mammals organisms developmental biology molecular development research and analysis methods white blood cells animal cells t cells immune response spectrophotometry immune system cytophotometry biochemistry rna eukaryota cell biology monocytes cats nucleic acids physiology biology and life sciences cellular types regulatory t cells amniotes spectrum analysis techniques

2018

Humoral immune response to adenovirus induce tolerogenic bystander dendritic cells that promote generation of regulatory T cells

Human babesiosis, especially caused by the cattle derived Babesia divergens parasite, is on the increase, resulting in renewed attentiveness to this potentially life threatening emerging zoonotic disease. The molecular mechanisms underlying the pathophysiology and intra-erythrocytic development of these parasites are poorly understood. This impedes concerted efforts aimed at the discovery of novel anti-babesiacidal agents. By applying sensitive cell biological and molecular functional genomics tools, we describe the intra-erythrocytic development cycle of B. divergens parasites from immature, mono-nucleated ring forms to bi-nucleated paired piriforms and ultimately multi-nucleated tetrads that characterizes zoonotic Babesia spp. This is further correlated for the first time to nuclear content increases during intra-erythrocytic development progression, providing insight into the part of the life cycle that occurs during human infection. High-content temporal evaluation elucidated the contribution of the different stages to life cycle progression. Moreover, molecular descriptors indicate that B. divergens parasites employ physiological adaptation to in vitro cultivation. Additionally, differential expression is observed as the parasite equilibrates its developmental stages during its life cycle. Together, this information provides the first temporal evaluation of the functional transcriptome of B. divergens parasites, information that could be useful in identifying biological processes essential to parasite survival for future anti-babesiacidal discoveries. Human babesiosis is a rapidly emerging, zoonotic, infectious disease causing potentially life-threatening malaria-like symptoms in humans. It is caused by intra-erythrocytic protozoan parasites of the genus Babesia [1] and it is transmitted to humans via an ixodid tick vector or through a blood transfusion from asymptomatic carriers [2]. Bovine babesiosis is well regarded as one of the most important diseases of livestock, especially in the tropical and sub-tropical regions of the world [3]. However, human babesiosis disease prevalence has escalated over the past 50 years from a few isolated cases to global endemic areas now being recognized [4,5]. In Europe, cattle associated B. divergens is the most common causative agent of human babesiosis, especially throughout regions with extensive cattle industries, as the distribution geographically correlates with both pathogen infected host species and tick-vector infested regions, allowing for zoonotic transmission potential [6]. Disease burden outside North America and Europe is currently poorly described but considering the worldwide distribution of Babesia parasites, improved surveillance is required. Since the symptoms of human babesiosis resemble that of malaria and diagnosis is predominantly reliant on microscopic evaluation of blood smears, this disease may be misdiagnosed as a malaria infection, especially in areas of co-endemicity. Early disease detection, diagnosis and treatment with effective anti-babesiacidal compounds are therefore vital for both human and animal health [7]. In humans, Babesia parasites can be cleared by anti-malarials including atovaquone (with azithromycin) or quinine (plus clindamycin) but highly immuno-compromised individuals respond poorly to these treatments. As early reports of resistance against these combinations have been noted in the past few years, the need for alternative treatments is evident [8,9]. Against this background of potential zoonotic human babesiosis medical emergencies, it is quite surprising that our understanding of the basic biological processes underlying Babesia pathophysiology is still poorly understood, even with the recent application of genetic manipulation for transfection of Babesia parasites as well as the sequencing of the B. bovis genome [10,11]. Particularly intriguing is the fact that the precise progression and duration of the intra-erythrocytic, asexual developmental cycle (IDC) has not been clarified. During its IDC, Babesia parasites undergo asexual replication by binary fission (budding) of trophozoites to form 2–4 merozoites [12]. Each merozoite is thought to undergo a single cycle of division and then escape via cell lysis to re-infect new erythrocytes [13]. This establishes a perpetual, asynchronous asexual parasitic growth cycle, which is thought to last approximately 8 hours [14] and encompasses several developmental stages all present at the same point in time within the hosts’ bloodstream [15]. However, the description of the IDC and its different stages are fraught with uncertainties: historically different stages were described only based on light microscopy; little attention has been paid to their sequence of development and descriptions of the various Babesia parasitic in vitro stages do not share a consensus in literature and display considerable morphological pleiomorphism. Moreover, fundamental biological questions remain unanswered, particularly concerning the molecular descriptors governing the IDC of Babesia parasites. In this study, the in vitro IDC of the human pathogen B. divergens was comprehensively evaluated by employing various high-content cell biological and molecular strategies as has been previously applied to the more widely studied but related hemoprotozoan malaria parasite, Plasmodium falciparum. The study particularly focused on B. divergens as human pathogen and model organism for Babesia since it is amenable to in vitro cultivation. This is to our knowledge the first quantitative description and temporal evaluation of intra-erythrocytic B. divergens development and enabled clear characterization of the stage-specific development, based on nuclear proliferation in these parasites. The information is novel not just from a biological perspective, but will also be essential in future prioritization of anti-babesiacidal compounds. Based on the findings presented here that correlate the linear increase of DNA content to its associated nuclear content and morphological classification, the progression of intra-erythrocytic B. divergens parasites in its development from one IDC stage to another could be proposed. Temporal evaluation of in vitro proliferation for each developmental form (ring, paired piriforms, tetrads and/or multiple infections) was performed over a 16-hour period. Parasitemia steadily increased to 15% over the 16-hour period monitored in the asynchronous culture (Fig 3). However, between 4–6 hours of development, the ring formation population doubled (from a 2% to a 4% contribution to parasitemia, P<0. 05, n = 3) (Fig 3). This was additionally associated with an increase in paired piriforms (2% increase in contribution to total parasitemia for paired piriforms). Similar observations were made between 8–10 hours of development, where a significant increase in paired piriforms was again observed (8% to 10%, P<0. 05, n = 3) (Fig 3). Overall, the population distribution in the mixed culture remained predominantly paired piriforms (9. 04 ± 1. 03% contribution to total parasitemia that ranged from 10. 3–15. 3%) followed by ring formations (3. 33 ± 0. 68% contribution to total parasitemia) (Fig 3). The tetrad (or multiple infection) population is the most stable throughout the temporal evaluation with no significant variation in population size (average 1. 08 ± 0. 28% contribution to total parasitemia maintained throughout). The increase in ring and piriform populations between 4–6 hours of development additionally contributed to a 3% increase in total parasitemia (from 10 to 13% total parasitemia) and again a 2% increase in parasitemia between 8–10 hours (from 13 to 15% total parasitemia). This is indicative of life cycle progression through re-infection of erythrocytes (contributing to new rings formed) and parasite developmental maturation (contributing to new paired piriforms). If significant increases in the ring populations are taken as an indicator of merozoite invasions of erythrocytes and initiation of new development cycles, it appears therefore that the in vitro development of B. divergens is typified by a 4 hour progression window and that, after two parasitic life cycles and under the culture conditions employed, equilibrium could be established and maintained. As such, a multiplication index of 3-fold was subsequently observed for continued B. divergens in vitro development. There are currently no data describing the molecular events associated with B. divergens intra-erythrocytic development, information that is essential to understanding the nature of stage-specific progression of this parasite’s IDC. Since clear contributions of morphologically distinct life cycle compartments were observed to contribute to the IDC progression of B. divergens parasites, we set out to describe the global transcriptome of these parasites through its IDC as an indicator of the physiological processes involved. Transcriptome analysis (mRNA abundance determination) was subsequently conducted on asynchronous, newly initiated cultures with a custom designed DNA microarray containing 15744 target features covering 97% of the B. bovis genome (3703 independent ORFs) and using a reference pool design strategy (Fig 4). B. divergens parasites seem to adapt to culturing under unlimited growth conditions (4% increase in parasitemia, P<0. 05, n = 3) (Fig 4A). Again, this adaptation was resultant of a significant increase in both ring and paired piriform parasites (from 2. 5 to 5% and 4. 75 to 7. 5%, respectively, P<0. 05, n = 3); after this the parasite population distribution equilibrated. These observations were further evident from correlation data, indicating that the transcriptome of equilibrated parasites showed the best correlation (r = 0. 19) between parasites that have been in culture for 9 and 16 hours (C9 vs. C16, Fig 4B). Comparatively, initiate cultures showed complete disconnect from established cultures (C0 vs. C16 anti-correlated at -0. 03). This clearly indicates a physiological adaptation event underscored by a predominant transcriptional repression in the initiate culture (931 undefined transcripts, 387 transcripts with increased abundance, compared to 2385 repressed transcripts). Differential expression between the initiate culture and the parasites 4 hours after inoculation resulted in 164 transcripts significantly affected (82 decrease and increase in abundance, respectively) (S1 Table). Of these, only 50 could be annotated using the dCAS annotation system [18] based on each transcript’s COG classification (E-value cut-off of less than 1x10-4) (Table 1). The processes mostly affected include transcription (26%), translation (8%), protein turnover (21%), cellular (30%) and metabolic (11%) processes, and stress defense mechanisms (2%). All activated transcripts may be associated with the changes in parasite population distributions (increases in ring and paired piriforms). Clustering of transcripts based on co-expression profiles indicated the expected similarity in expression profiles across the transcriptome (Fig 4C). Moreover, transcripts in some of these co-expressed clusters additionally showed clear chromosomal synteny (e. g. BBOV_III006240 and BBOV_III007280 on chromosome III; BBOV_IV007380 and BBOV_IV001730 on chromosome IV) implying transcriptional level regulation for these transcripts. Further temporal evaluation of the B. divergens transcriptional landscape during its IDC indicated correlation across the transcriptional landscape (Fig 5). However, clear transcriptional activation to a permissive state was observed as the parasites progressed in their life cycle, particularly evident for parasites in culture for at least 9 hours (e. g. in C4,45% of the transcripts showed increased abundance; this increased to 60% at C9) (Fig 5A). However, after 16 hours in cultivation, the transcriptome seems more unbiased and show a more equal distribution of abundances in transcripts with an inclination towards transcriptional repression. This may be correlated to a slight increase in the paired piriform population in the morphological profiles observed between parasites in culture for 9 or 16 hours. Further comparison of the transcriptomes of the parasite populations indicated the presence of differential expression patterns across the 4–16 hour evaluations. To distinguish only processes associated with IDC progression and minimize adaptation responses as a result of culture initiation (time 0–4 h), transcriptional profiles were evaluated between parasites in culture for 4,9 and 16 hours, respectively (Fig 5B). Functional annotation of transcripts associated with the alternative expression patterns identified several transcripts associated with normal biological, cellular and functional pathways throughout the investigated period. This included 11 major functional categories (catalytic activity 16%; energy production and conversion 5%; lipid transport metabolism and synthesis 2%; membrane protein components 6%; mitochondrial components 3%; proteolysis 13%; ribosomal components 17%; transcription and translation 25%; transport 6%, and variant surface antigen expression 9%), which were present across all clusters and time points. However, certain biological processes proved more variable over the temporal evaluation, particularly protein turnover, transcription and translation with increased levels of activity at 9 hours in culture. Comparatively, energy production is, as expected, maintained throughout the temporal evaluation. Our current understanding of the intra-erythrocytic developmental cycle of B. divergens has been clouded by imprecise and conflicting classifications, mostly due to historical analyses relying for the most part only on morphological microscopic observations. For instance, discrepancies in morphological evaluation of different life cycle forms are reported, where trophozoites are sometimes referred to as merozoites [12]. Additionally, one of the major challenges with studying the stage-specific development of B. divergens intra-erythrocytic development is the inability to synchronize these parasites in vitro to a single stage. A well-developed synchronization strategy induced by either sorbitol or mannitol is widely applied in P. falciparum research and is mainly associated with changes in membrane permeability and buoyant density of erythrocytes parasitized by this organism [19]. However, these techniques were evaluated and found to be ineffective against B. divergens in vitro, resulting in parasite death within 24 hours. The data obtained by combining advanced cell biological and molecular strategies allow for the first time objective and clear evaluation of the chronology and stage-specificity of intra-erythrocytic development of B. divergens parasites. These are sensitive and quantitative and allow various developmental stages to be distinguished. Flow cytometry, light and fluorescent microscopy allowed for the accurate detection of intra-erythrocytic B. divergens parasites as well as determination of dynamic proliferation, developmental stage assessment and isolation of asynchronous B. divergens parasites, based on morphology as well as nuclear content. Asynchronous, independent nuclear division occurs during intra-erythrocytic P. falciparum development, where daughter merozoites follow a non-geometric expansion and parasitic multiplication consequently deviates from what is expected from equal numbers of binary divisions [20]. Similar findings were observed with light microscopy and flow cytometry for the asynchronous in vitro B. divergens cultures in this study and enabled the isolation of specific developmental stages over a 16-hour period. The increase in MFI values observed between the three infected populations corresponds to the fluorescent microscopy images, which ultimately indicate an increase in DNA nuclear content from one developmental stage to the next. Based on the DNA measurements, which underlie the morphological findings of the present study, asynchronous in vitro Babesia dynamics was further evaluated. With primary parasitology classifications in mind, intra-cellular and actively metabolizing parasites are usually classified as trophozoites, which divide asexually (merogony) with the resultant formation of daughter merozoites. Erythrocytes infected with a ring formation, contain a single parasitic nucleus (1N) and erythrocytes infected with either paired piriforms, tetrads or multiple infected erythrocytes, contain two or more nuclei (2N or >2N). These findings correlate nuclear content to a particular isolated cell population (based on morphology); previously unclear for Babesia parasites. Here we define intra-erythrocytic B. divergens parasites directly after invasion as mono-nucleated rings (1N nuclear content), which then rapidly progress to metabolically active but still mono-nucleated, haploid trophozoite populations. These forms are only morphologically distinguishable based on anaplasmoidy in rings compared to the more rounded / ovoid trophozoites and not on differences in nuclear content; further specific classification would require metabolic flux data. With the associated nuclear content information provided in this paper, we were able to indicate the subsequent progression of mono-nucleated ring / trophozoites to bi-nucleated paired piriforms (2N nuclear content) during binary fission in which a single parasite undergoes a single nuclear division event resulting in two daughter merozoites. During this nuclear division event, the nucleus becomes typically V-shaped (as observed with fluorescent microscopy). The formation and origin of multi-nucleated tetrads is less clear. These parasites may undergo two nuclear division cycles that may result in the formation of four daughter merozoites. Binary fission dictates the duplication of nuclear content in a cell, followed by DNA segregation and finally cytokinesis. The formation of tetrads would therefore imply either that (a) a duplicate binary fission event occurred simultaneously from a single ring, visible as the characteristic cross morphology prior to cytokinesis or (b) that a minor proportion of paired piriforms would not undergo cytokinesis after DNA replication but rather undergo a second round of DNA replication, resulting in the tetrad formations with a 4N nuclear content prior to cytokinesis and ultimately to the formation of 4 daughter merozoites. High-resolution real-time microscopy evaluation is needed to address these possibilities. Since tetrad forms are infrequently observed in culture compared to the other morphological forms, tetrad formation is either relatively rare (only 10–25% of the total parasite population) or occurs rapidly (with quick kinetics of cytokinesis) such that these forms are rarely observed. With an optimal multiplication index of 2–3, it does not make a major contribution to B. divergens parasite proliferation in vitro. All merozoites of B. divergens parasites are typically piriform and joined by their pointed ends and also do not fill the complete erythrocyte, as expected of ‘small babesiae’. The kinetics of invasion (and re-invasion) after initial contact between the parasite and the erythrocyte, proceeds rapidly, between 45 seconds and 10 minutes [21]. We therefore hypothesize a parasitic propagation diagram based on the morphological observations, DNA measurements, temporal distribution and transcriptome expression dynamics (Fig 6). If an increase in parasitemia and associated increased ring populations is taken as indicators of new infections, then the significant increase in newly infected erythrocytes (ring formations) observed here during the first 4–6 hours in culture provides indications of the kinetics of the B. divergens in vitro IDC merogony. This is markedly quicker than previous reports where the in vitro life cycle of B. divergens parasites was claimed to last around 8 hours under the culture conditions used in that study [14]. However, the 2 hourly evaluation performed in our study enabled a finer analysis of the IDC progression kinetics that was not probed previously. The asexual Babesia parasitic life cycle has only been described within the erythrocytes of their vertebrate hosts and the salivary glands of the tick vector, with limited data currently available concerning the Babesia sexual life cycle [22]. Only two studies have, however, reported on the visualization of intra-erythrocytic gametocytes in B. divergens [12,23]. Our temporal evaluation of B. divergens revealed contrasting results to that achieved with the optimized flow cytometry strategy during the interrogation of mixed infected erythrocyte populations. The overall population distribution (associated with temporal evaluation) remained predominantly paired piriforms followed by rings, tetrads and multiple infections. Comparatively, the overall population distribution observed with flow cytometry during the interrogation of mixed infected erythrocyte populations was predominantly ring formations followed by paired piriforms, tetrads and multiple infections. The inability to accurately visualize and characterize gametocytes with light microscopy within mixed B. divergens cultures may have contributed these contrasting results. However, we analyzed our transcriptional data set for genes characterizing gametocytes (i. e. bdccp 1; bdccp 2 and bdccp 3) [24]), but these were not differentially affected in our data set. This can be interpreted as: 1) we did not observe gametocytes in B. divergens cultures in vitro (supporting our morphological descriptors characterising merozoites) or 2) the parasites are not under undue physiological stress, which have been implicated to result in increased gametocytogenesis. The transcriptional response that we do observe therefore mimics typical parasite physiology. Flow cytometry measurements may therefore include ring forms as well as gametocytes, thereby increasing the DNA content. Morphological findings suggest that a trend in parasitic development from one stage to the next can be observed and that all stages reach a plateau in their developmental process. Transcriptome analyses of the initiate culture revealed an expression pattern for B. divergens parasites. The draft B. divergens genome (recently deciphered) in combination with the information presented here, can facilitate future Babesia related studies and improve our understanding of the parasites biology, host-parasite interaction as well as improve control and treatment strategies [25,26]. Additionally, the information presented here represents a preliminary catalog for B. divergens gene expression during its IDC, measured over time. The differential transcript abundance analysis identified several activated and repressed transcripts associated with parasitic growth and development, which may have contributed to changes in parasite population distributions. The transcriptional analysis was subsequently linked to the morphological findings. The initial stress response (between 0–4 hours) potentially induced by the addition of complete culture media, may have influenced the expression pattern. To minimize the possible stress response effects, an alternative expression pattern was suggested for the hierarchical clustered data set, which ranged between 4–16 hours. Functional annotation of the visualized transcripts and transcriptional response revealed predominantly activated and repressed components associated with transcription, translation, protein turnover, cellular and metabolic response as well as stress defense mechanisms, confirming the observed change in parasite population distributions (increases in ring and paired piriforms). The transcriptome mirrors the Babesia genome with glycolysis and components of the TCA cycle, glycerolipid and glycerophosphospholipid metabolism, pyrimidine and associated nucleotide synthesis, amino acid synthesis and certain components associated to apicoplast metabolism [11]. Moreover, almost a tenth of the differential transcriptome is associated with expression of transcripts in the ves1 family (expressing VESA1). Similar to the malaria parasite, Babesia spp use antigenic variation to escape detection by the host’s immune system [26]. The hierarchical clustering employed here indicated that some of these biological processes do however show differential expression (e. g. protein expression and transport). Overall, the transcriptionally permissive state of the B. divergens IDC resembles that of the IDC transcriptome of P. falciparum parasites (with the exception of silenced virulence genes) [27]. In the latter, phase-ordering indicated a ‘just-in-time’ transcriptional activation, with transcripts finely associated with highly synchronized specific stages in the IDC. Even through the data presented here for B. divergens implies molecular control factors involved in the IDC, fine resolution temporal analysis of this parasite will be confounded as the population is composed of mixed stages, making stage-specific analysis difficult. More comprehensive identifications of novel compounds against veterinary Babesia species have only recently gained attention [28]. However, for screening platforms to identify effective anti-babesiacidal compounds, several lessons may be learnt from highly advanced screening strategies from other thoroughly investigated diseases like malaria. As such, anti-malarial screening strategies have been clearly delineated and require early decision making based on target product and target candidate profiles. Particularly for in vitro hit identification, the ability of novel compounds to target specific (or all) stages of malaria parasite development as well as their speed of action is used to classify these in vitro specific profiles, thereby enabling their further prioritization [29,30]. Similar strategies are required for screening of anti-babesiacidal drugs, as evaluation of the stage-specific nature of compounds would be imperative in understanding it’s mode-of-action and examine new treatment and dosage strategies. Current evaluation of growth-inhibiting effects of potential anti-babesiacidal compounds rely on either microscopic examination of Giemsa-stained smears and / or the evaluation of incorporation of isotopes into in vitro Babesia cultures [31], with neither of these allowing stage-specific or temporal evaluation of compound action. Light microscopy is time-consuming, subjective, labor intensive and operator dependent with poor quantitative robustness. Although isotopic techniques overcome the latter, this is being replaced by cost effective, reliable and non-radioactive fluorescence-based assays. Apart from the use of such dyes in plate-reader formats (enabling high-throughput screening platforms) [28] these assays are usually performed without evaluation of stage-specificity. However, the possibility to combine these dyes with techniques such as flow cytometry will allow for an additional level of characterization that has proved useful to address these caveats. One such example has been the successful application to the stage and temporal evaluation of anti-malarial compounds [17,32]. The expansion of our biological knowledge of B. divergens parasites’ intra-erythrocytic development through the molecular blueprint of its complete transcriptome provided in this paper should enhance future discovery of novel anti-babesiacidal drugs. Approval for the importation and in vitro cultivation of B. divergens was obtained from the South African Department of Agriculture, Forestry and Fisheries. Human blood and sera was collected from volunteer donors and used for cultivation with ethical approval (University of Pretoria Faculty of Natural and Agricultural Sciences Ethics Committee approved the project protocol with identification number EC120821-077). Volunteer donation was based on written informed consent from only adult donors at a registered phlebotomy facility. No minors were allowed to donate blood in this study. Parasite cell culturing was based on CDC criteria for such research. No animals were used in this study. All parasites were grown under in vitro conditions only. The Rouen 1987 strain of B. divergens was kindly provided by Dr. Stephane Delbecq (Laboratoire de Biologie Cellulaire et Moleculaire UFR Pharmacie, Montpellier, France). Asynchronous B. divergens parasite cultures were maintained in human erythrocytes (type O+) suspended in complete culture medium [RPMI-1640 medium supplemented with 25. 2 mM HEPES, 22. 2 mM D-glucose, 50 mg/l hypoxanthine, 21. 4 mM sodium bicarbonate, 48 mg/l gentamycin (Sigma) ] further supplemented with 10% human serum [33]. Cultures were maintained at 37°C in a gaseous environment of 90% N2,5% O2 and 5% CO2 on a rotary platform (60 rpm) at ~5% hematocrit and 10–15% parasitemia, with daily media replacement. Comparatively, intra-erythrocytic P. falciparum (3D7 strain) parasites were obtained from the MR4 (www. mr4. org) and were maintained at a 5% hematocrit, 2–5% parasitemia in complete culture medium [RPMI-1640 medium supplemented with 25. 2 mM HEPES, 22. 2 mM D-glucose, 50 mg/l hypoxanthine, 21. 4 mM sodium bicarbonate, 48 mg/l gentamycin (Sigma) ] further supplemented with 0. 5% Albumax II (Invitrogen) [34] in the same gaseous environment under shaking conditions as above. For both intra-erythrocytic B. divergens and P. falciparum parasites, Giemsa-stained thin smears light microscopy was used for the daily determination of both the parasitemia and morphology. Thin blood smears were prepared, air dried, fixed with methanol and stained for 5 minutes prior to examination. The percentage parasitemia was determined as the number of infected erythrocytes per 100 cells, with a minimum of 1000 erythrocytes counted [35]. Asynchronous intra-erythrocytic B. divergens parasites (5% hematocrit, 10–15% parasitemia) were examined every two hours over a 16-hour time period using Giemsa-stained smears and light microscopy. The asexual developmental life stages were classified as rings, paired piriform and tetrad and/or multiple infection formations. Fluorescent microscopy (Zeiss Axiovert 200 fluorescent microscope) was also used to examine intra-erythrocytic B. divergens parasites using the Axiovision release 4. 8. 2 software for analyses. A custom designed DNA microarray slide was used for gene expression analysis. The base composition probe design strategy and probe selection parameters were selected according to Agilent Technologies eArray 60-mer platform specifications. Available B. bovis sequences were downloaded and retrieved from the National Center for Biotechnology Information (NCBI) (http: //www. ncbi. nlm. nih. gov/) and supplemented with additional published sequence data [5]. Selected probes were randomly distributed across the array using the 8x 15K design format. All arrays were ordered from Agilent Technologies (https: //earray. chem. agilent. com/earray/).

Vector-borne parasitic diseases are still the major cause of morbidity and mortality in both humans as animals. Some of these parasites have been well studied, including the malaria parasite, Plasmodium falciparum, since these cause major fatalities in humans. However, other parasites like Babesia divergens, who closely resemble malaria parasites, are currently becoming a major concern since these are zoonotically transmitted to humans from their natural hosts (e. g. cattle) by ticks. Their rising levels indicate the possibility of human medical emergencies occurring. Unfortunately, basic biological processes in these parasites are poorly understood due to their neglected status. In this manuscript, we describe for the first time high-content analysis of the development of B. divergens parasites in vitro over time. We applied cell biological and advanced functional molecular strategies to also provide the first descriptors of the parasite’s transcriptome during life cycle development. This information is unique and to not only dramatically expand our information base on Babesia biology, but also provide information that can be exploited for future drug discovery endeavors.

Abstract Introduction Results Discussion Materials and Methods

2015

Morphological and Molecular Descriptors of the Developmental Cycle of Babesia divergens Parasites in Human Erythrocytes

We investigated whether Tbx1, the gene for 22q11. 2 deletion syndrome (22q11. 2DS) and Foxi3, both required for segmentation of the pharyngeal apparatus (PA) to individual arches, genetically interact. We found that all Tbx1+/-; Foxi3+/- double heterozygous mouse embryos had thymus and parathyroid gland defects, similar to those in 22q11. 2DS patients. We then examined Tbx1 and Foxi3 heterozygous, null as well as conditional Tbx1Cre and Sox172A-iCre/+ null mutant embryos. While Tbx1Cre/+; Foxi3f/f embryos had absent thymus and parathyroid glands, Foxi3-/- and Sox172A-iCre/+; Foxi3f/f endoderm conditional mutant embryos had in addition, interrupted aortic arch type B and retroesophageal origin of the right subclavian artery, which are all features of 22q11. 2DS. Tbx1Cre/+; Foxi3f/f embryos had failed invagination of the third pharyngeal pouch with greatly reduced Gcm2 and Foxn1 expression, thereby explaining the absence of thymus and parathyroid glands. Immunofluorescence on tissue sections with E-cadherin and ZO-1 antibodies in wildtype mouse embryos at E8. 5-E10. 5, revealed that multilayers of epithelial cells form where cells are invaginating as a normal process. We noted that excessive multilayers formed in Foxi3-/-, Sox172A-iCre/+; Foxi3f/f as well as Tbx1 null mutant embryos where invagination should have occurred. Several genes expressed in the PA epithelia were downregulated in both Tbx1 and Foxi3 null mutant embryos including Notch pathway genes Jag1, Hes1, and Hey1, suggesting that they may, along with other genes, act downstream to explain the observed genetic interaction. We found Alcam and Fibronectin extracellular matrix proteins were reduced in expression in Foxi3 null but not Tbx1 null embryos, suggesting that some, but not all of the downstream mechanisms are shared. The pharyngeal apparatus (PA) is an evolutionarily conserved structure that forms early in vertebrate embryos. The PA develops as a series of bulges, termed arches, found on the lateral surface of the head region of the embryo. During mammalian development, five pairs of pharyngeal arches numbered PA1, PA2, PA3, PA4, and PA6 (the fifth PA is transient) form subsequently, over time, from the rostral to caudal part of the head region of the embryo [1]. The process involved in the formation of each arch is referred to as pharyngeal segmentation. Each arch contributes to different craniofacial muscles, nerves and skeletal structures. PA1 contributes to the skull, incus and malleus of the middle ear, jaw, nerves and muscles of mastication. PA2, contributes to the skull, stapes in the middle ear, facial muscles, jaw, and upper neck skeletal structures. In addition to skeletal structures, muscles and nerves, PA3 is required to form the thymus and parathyroid glands. In mouse embryos, the thymus and parathyroid glands are derived only from PA3, but in humans the inferior parathyroid gland is derived from PA3 whereas PA4 contributes to the superior parathyroid gland. PA4 and PA6 contribute to the aortic arch and arterial branches [2,3]. Each pharyngeal arch is surrounded by endoderm and ectoderm derived epithelial cells forming pharyngeal pouches and clefts, respectively. Mesoderm and neural crest derived mesenchyme cells occupy the center of each arch [1,4]. The epithelia are needed to invaginate and promote segmentation to individual arches. The pharyngeal endoderm (PE), which is the focus of this study, receives and sends signals from the mesenchyme to initiate morphogenesis and invaginate [5]. Once each arch segments, proper patterning is also required to form derivative structures. The PE sends distinct signals in each arch to promote normal patterning [6–9]. Abnormal PA segmentation or patterning during development will cause defects within the later structures derived from the PA and leads to human birth defect disorders. One particular gene important for PA segmentation is Tbx1, encoding a T-box transcription factor implicated in 22q11. 2 deletion syndrome (DiGeorge syndrome [MIM# 188400]; velo-cardio-facial syndrome [MIM# 192430]). Tbx1-/- homozygous null mutant mouse embryos die at birth with hypoplastic and intermittent missing craniofacial muscles [10], cleft palate, absent thymus and parathyroid glands, as well as a persistent truncus arteriosus (PTA) with a ventricular septal defect (VSD) [11–13]. The first arch forms in Tbx1-/- embryos but the distal PA fails to become segmented, thereby explaining, in part, why the PA derived structures are malformed [11–13]. Tbx1 is expressed in the mesoderm of the head region in early mouse embryos and then throughout the endoderm, mesoderm, and distal ectoderm of the PA, while each arch forms, and becomes reduced at mouse embryonic day (E) 10. 5 [12,14,15]. Tissue specific inactivation of Tbx1 has been performed using the Cre-loxP system [4,16–19]. It was found that Tbx1 is required in all three tissues for development of the derivative organs affected in the null mutant embryos [17,19–21]. Since the PE is critically important for segmentation of the distal PA [22,23], it is important to understand the genes and processes that might act downstream. Another gene that has been shown to be important for normal PA segmentation is Foxi3, which encodes a Forkhead box (Fox) transcription factor. Foxi3 is expressed in the ectoderm in the head region early in embryonic development and is then expressed in the epithelia of the PA from around the same stages as when Tbx1 is expressed [24]. Foxi3 is also important for epithelial cell differentiation within the epidermis [25]. Heterozygous mutations were discovered in Foxi3 in several hairless dog breeds with hair follicle and teeth defects [26]. Global Foxi3-/- null mutant mouse embryos fail to form endodermal pouches and this results in failed PA segmentation leading to severe defects in the skull, jaw, and ears [27–29]. It has been shown that Foxi3 may have a cell non-autonomous effect on craniofacial neural crest cell survival because these cells undergo apoptosis in the mutant embryos at E10. 0 [27]. In this study, we tested whether there is a genetic interaction between Tbx1 and Foxi3 during mouse embryonic development. We discovered that these two factors interact at minimum, in the third pharyngeal pouch endoderm, needed to form the thymus and parathyroid glands. Further, we found that inactivation of Foxi3 results in cardiovascular anomalies. We characterized the process of pharyngeal segmentation and found that global inactivation of Tbx1 and Foxi3 both result in failure of the epithelia to properly invaginate along with an expansion of multilayers of PE cells leading to failed segmentation of the distal PA. We identified some shared downstream genes that were reduced in expression in either null mutant and suggest that they may share some similar molecular mechanisms. Since loss of Foxi3 or Tbx1 disrupt the segmentation of the PA, we tested whether Foxi3 might act upstream or downstream of Tbx1. Foxi3 is normally expressed throughout the epithelia of the PA, while Tbx1 is more broadly expressed (Fig 1A and 1C). Whole mount in situ hybridization (WMISH) using a Foxi3 antisense mRNA probe on Tbx1-/- mouse embryos and wildtype (WT) littermate controls at E9. 5, revealed that Foxi3 expression was reduced in PA1 and absent in the unsegmented distal pharyngeal apparatus in Tbx1 null mutant embryos (Fig 1A and 1B). To determine if Tbx1 expression is affected in Foxi3-/- embryos, WMISH followed by tissue sectioning using a Tbx1 probe on Foxi3+/- control littermate and Foxi3-/- mouse embryos was performed and we found that the Tbx1 expression pattern was maintained in the pharyngeal mesoderm and endoderm despite the lack of segmentation of the distal PA (Fig 1C–1H). This indicates that either Foxi3 acts downstream of Tbx1 in the same genetic pathway and/or that the cells expressing Foxi3 were lost in the Tbx1 null mutant embryos. It has been previously shown that Foxi3-/- embryos have absent jaw bones, abnormal mandible, deformed maxilla bones, absent jugal (bony arch of zygoma, cheek bone) and smaller palatines, misshapen Meckel’s cartilage, and absent ears [27,29]. At E9. 5, segmentation of the PA to individual arches did not occur in Foxi3-/- embryos (Fig 1F and 1G), which is consistent with previous findings [27]. There is little known about its role in formation of later embryonic structures from the distal PA derived from PA3-6. We found that at E15. 5, Foxi3-/- embryos had absent thymus and parathyroid glands (100%; n = 11), interrupted aortic arch type B (IAAB, 63%; n = 7), ventricular septal defect (VSD; 100%; n = 11) and retro-esophageal right subclavian artery (RRSA, 55%; n = 6) as listed in Fig 1I and shown in Fig 1L–1S. Some had both a RRSA and IAAB (18%; n = 2), while the remaining had either RRSA or IAAB (81%; n = 9; Fig 1I and 1P–1S). Based upon the similarities in the distal PA derived defects, and that Foxi3 was reduced in expression in Tbx1-/- embryos, we tested whether there could be a genetic interaction between the two genes. We first tested whether single heterozygous Foxi3+/- [27] or Tbx1+/- [11] embryos had defects. At E15. 5, Fox3+/- embryos were normal (n = 8) and Tbx1+/- embryos had a normal thymus gland and had ectopic or absent parathyroid glands in 38% of the embryos (n = 5; Fig 1I–1K; S1A–S1H Fig). Normally, the parathyroid glands should be found adjacent to the thyroid glands, where the two carotid arteries are present nearby in the same section. When ectopic in Tbx1+/- or Tbx1+/-; Foxi3+/- embryos, parathyroid glands were found in a more caudal position in the embryo that is separate from the thyroid glands (Fig 2C; S1C and S1 Fig) versus controls (Fig 2A; S1A Fig). When ectopic in Tbx1+/-; Foxi3+/- embryos, thymus glands were more rostrally located than normal and were present at the same level of the embryos as the carotid arteries (Fig 2D; S1F Fig), as compared to control embryos, where the thymus glands were located at the branchpoint between the innominate and right carotid artery (Fig 2B; S1B and S1D Fig). Hypoplastic thymus glands were smaller in size than normal glands (Fig 2D; S1F Fig). At E15. 5, all double heterozygous Tbx1+/-; Foxi3+/- embryos had either a hypoplastic and/or ectopic thymus and parathyroid glands (n = 14) and this increase is statistically significant (Fig 1J and 1K). More than half of double heterozygous embryos had both a hypoplastic thymus and ectopic parathyroid glands in comparison to WT controls (57%; n = 8/14; Figs 1J, 1K and 2A–2D; S1A–S1H Fig). Expression of Gcm2 (glial cells missing homolog 2) and Foxn1 (Forkhead box protein N1) mark the parathyroid-fated and thymus-fated domains in the pharyngeal endoderm of PA3, respectively [30–32]. We performed RNAscope in situ hybridization on tissue sections with probes for Gcm2 and Foxn1 at E11. 5 when both of these genes are expressed. In Tbx1+/-; Foxi3+/- embryos, expression of both markers was slightly reduced in intensity in PA3 in comparison to WT littermate controls (Fig 2G and 2H). When expression was quantified in comparison to WT embryos, both genes were reduced in expression in the PA3 derivative region in the double heterozygous embryos, but only reduction of Foxn1 was statistically significant (Fig 2J). The presence of some expression of these two genes is consistent with the occurrence of milder thymus and parathyroid gland defects (hypoplastic thymus and/or ectopic parathyroid) in these embryos as compared to null mutant embryos (Fig 1I; S1A–S1H Fig; Fig 2A–2D). Histology sections of Tbx1+/-; Foxi3+/- embryos at E10. 5 were examined to see if there were defects in PA3. We noted a slightly narrowing of the space between PA3 and PA4 in the double heterozygous embryos, but no other malformations were detected (S1I–S1M Fig). Again, the phenotype at E15. 5 is relatively mild in comparison to either null mutant, in which the organs were completely absent. To gain more insights into the basis of the observed defects, we examined where Tbx1 and Foxi3 are co-expressed. The PA develops from E8. 5-E10. 5, in which PA3 forms by E9. 5. In situ hybridization using RNAscope probes was performed on coronal tissue sections from WT embryos to determine if there is overlap between Foxi3 and Tbx1 mRNA expression at E9. 5 (Fig 3A–3D). There was strong expression of Tbx1 in the pharyngeal endoderm and cardiopharyngeal mesoderm (Fig 3B), as has been previously reported [33,34]. Tbx1 was also expressed in the pharyngeal endoderm and ectoderm of the distal PA (Fig 3B) as has been previously reported [33]. Foxi3 expression was localized exclusively to the pharyngeal pouches and clefts as published in the past [35]. Expression of Foxi3 was particularly strong in the junction between the pharyngeal pouch and cleft that lies between PA2 and PA3 (Fig 3C). Expression of Foxi3 was also detected where invagination of the epithelia is taking place to form the separation between PA3 and PA4 (Fig 3C). Co-expression of Tbx1 and Foxi3 in the same cells was detected in the second and third pharyngeal pouch and cleft and where invagination was taking place (Fig 3D). The third pharyngeal pouch is where co-expression of both genes occurred and this is the same region where the thymus and parathyroid glands will form. This provides supporting evidence that there could be a genetic interaction between the two genes. We then decided to inactivate both alleles of Foxi3 using the Tbx1Cre mouse line [36], where we would expect more obvious developmental defects at these early stages than when we inactivate one allele. To establish the role of Foxi3 within the Tbx1 lineage to explain the basis of PA3 derived defects in double heterozygous embryos, we inactivated it using Tbx1Cre/+ knockin mice [36]. For this, Tbx1Cre/+ mice were crossed with a Foxi3 floxed allele (Foxi3f/+) and the double heterozygous mice were crossed with Foxi3f/f mice to inactivate both alleles of Foxi3. We also crossed the Tbx1Cre/+ mice with a Rosa26GFP/GFP allele to detect the Tbx1 lineage using the GFP reporter. GFP fluorescence was observed in the pharyngeal mesoderm at E8. 5 (Fig 3E and 3F) and in both the pharyngeal mesoderm plus epithelia of the PA at E9. 5 (Fig 3G and 3H). The Tbx1Cre/+ is a knock-in allele for Tbx1 and it is heterozygous for Tbx1 [36]. GFP fluorescence was not detected in the epithelia at E8. 5. This was unexpected because Tbx1 mRNA expression occurs in the epithelia at this stage [37]. Thus, there is a difference in timing of Tbx1 expression and detection of GFP fluorescence, which marks recombination of loxP sites and translation of sufficient GFP to be visualized. This timing difference may explain why Tbx1Cre/+; Foxi3f/+ embryos at E15. 5 did not exhibit a more severe phenotype than what occurred in Tbx1+/- embryos, as compared to that in Tbx1+/-; Foxi3+/- embryos (Fig 1I). At E15. 5, Tbx1Cre/+; Foxi3f/f mutant embryos were compared to Tbx1Cre/+; Foxi3f/+ controls (n = 13) to determine if there were any PA derived defects (Fig 1I). Reduction of Foxi3 expression in the pouches, but not the pharyngeal clefts, within the PA in Tbx1Cre/+; Foxi3f/f embryos was observed in S2 Fig. The Tbx1Cre/+; Foxi3f/f embryos had absent thymus and parathyroid glands (100% n = 14; Fig 2E and 2F), similar to what occurred in Foxi3-/- embryos (Fig 1I). To determine if Foxn1 (thymus) and Gcm2 (parathyroid) expression [32] was affected, we performed RNAscope in situ hybridization with probes for these genes and found that the expression of both genes in the PA3 region were significantly reduced in conditional null versus wildtype controls (Fig 2G, 2I and 2J). We noted that there was no separation between PA3 and PA4 at stage E10. 5 (Fig 2L) as compared to the presence of a separation in Tbx1Cre/+; Foxi3f/+ controls (Fig 2K). Reduction in expression of Foxn1 and Gcm2 as well as the presence of morphology defects at E10. 5 might explain why the thymus and parathyroid glands did not form. Despite having absent thymus and parathyroid glands, these mutant embryos did not have cardiovascular or aortic arch defects (Fig 1I and S3A–S3F Fig). India ink injections confirmed presence of aortic arch arteries 3,4 and 6, in WT and Tbx1Cre/+; Foxi3f/f embryos (S3G and S3H Fig). While expression of Foxi3 was significantly reduced as determined by WMISH (S3A–S3D Fig) in conditional mutant embryos, we also tested Tbx1Cre/+; Foxi3f/- embryos to further inactivate Foxi3 and found that the embryos lacked the thymus and parathyroid glands but similar to above, they had no intracardiac or aortic arch defects (n = 8; S3I–S3L Fig). Therefore, there is an interaction between Tbx1 and Foxi3 in the third pharyngeal pouch endoderm. We did not detect PA4 derived defects in the Tbx1Cre/+; Foxi3f/- embryos at E15. 5. There are two different possibilities to explain basis for the lack of aortic arch or branching anomalies derived from PA4. One possibility is that there is no genetic interaction in the fourth pharyngeal pouch. Another possibility could be due to timing of Tbx1 gene expression and delayed timing of Cre activity using the Tbx1Cre allele, within the PE (Fig 3E–3H), although its less likely an issue in Tbx1Cre/+; Foxi3f/- embryos. We then next decided to inactivate Foxi3 in the PE to understand its tissue specific function. Foxi3 is expressed in the epithelia in the PA (Figs 1A, 3C and 3D and [24]). To determine the role of Foxi3 within the PE, we performed tissue specific inactivation of Foxi3 using the Sox172A-iCre/+ allele [38]. We first confirmed that the PE population (and endothelial cells) is marked by GFP expression using Sox172A-iCre/+; ROSA26GFP/+ embryos (Fig 3I–3L). Foxi3 is not expressed in endothelial cells. Recombination occurred prior to E8. 5 as indicated by robust green fluorescence that was present at E8. 5 (Fig 3I–3L). Foxi3 expression was reduced in these conditional mutants as detected by WMISH (S2E–S2H Fig). We found that Sox172A-iCre/+; Foxi3f/f embryos had failed segmentation of the distal PA as compared to Sox172A-iCre/+; Foxi3f/+ controls (S4A and S4C Fig). The PA was not as severely affected as in the Foxi3-/- null mutant embryos (S4B Fig), as PA2 was present, albeit hypoplastic (S4C Fig). At E15. 5, all Sox172A-iCre/+; Foxi3f/f embryos had absent thymus and parathyroid glands (100%; n = 12; Fig 1I). Cardiovascular defects occurred in 66% of embryos that included IAAB (50% n = 6), and/or RRSA (50% n = 6; Fig 1I and S4D–S4F Fig). As compared to Foxi3-/- mutant embryos that always had an aortic arch defect plus a VSD, Sox172A-iCre/+; Foxi3f/f mutant embryos had an IAAB but had normal septation of the ventricles (Fig 1I). Several mutant embryos had both an RRSA and IAAB (33% n = 3), while 33% (n = 3) had just an RRSA and 33% (n = 3) had an IAAB, whereas a few had normal aortic arches and normal branching (33% n = 3; Fig 1I). India ink injections showed that the 4th aortic arch artery was absent in Sox172A-iCre/+; Foxi3f/f embryos at E10. 5, similar to what we observed in Foxi3-/- embryos (S4G–S4L Fig). Absence of the 4th aortic arch arteries would lead to the observed cardiovascular defects that occurred in the embryos. Based upon this data, Foxi3 expression is required in the PE for classic 22q11. 2DS phenotypes. Further, Foxi3 in the ectoderm may have an additional role in the septation of the ventricles. In the Tbx1Cre/+; Foxi3f/f mutant embryos, only PA3 derivative structures were affected, being the thymus and parathyroid glands. As indicated above, the pouch and cleft between PA3 and PA4 did not form and was missing at E10. 5, when segmentation was complete (Fig 2K and 2L). E-cadherin is a cell-cell adhesion protein forming adherens junctions that bind cells tightly to each other and it marks epithelial cells (Reviewed in [39]). ZO-1 (Zona occluden-1) forms permeable barriers in adherens junctions and is a marker for the presence of apical/basal polarity among epithelial cells, with expression specifically on the apical side of the cell facing a lumen [40]. Immunofluorescence using antibodies to E-cadherin and ZO-1 was performed on WT, Tbx1Cre/+; Foxi3f/+ and Tbx1Cre/+; Foxi3f/f embryos at E9. 5, to visualize the structure of the epithelial cell population. At E8. 5 and E9. 0 there was no noticeable difference between the mutant and control embryos (Fig 4A–4H). But, at E9. 5, invagination of the partially stratified pharyngeal endoderm and ectoderm between PA3 and PA4 did not take place (Fig 4L–4N) as compared to controls (Fig 4I–4K). The PE of PA3 maintained its epithelial identity and cell polarity as marked by expression of E-cadherin (Fig 4I, 4J, 4L and 4M) and ZO-1 (Fig 4K and 4N). It is possible that instead of failure to invaginate, there was a delay. However, at E10. 5, the third pouch and cleft were absent in the conditional mutant embryos (Fig 4Q and 4R) in comparison to controls (Fig 4O and 4P), so that there was no distinction between PA3 and PA4 as shown by H&E staining (Fig 2K and 2L). Thus, the genetic interaction between the two genes is at minimum, in PA3. Immunofluorescence using antibodies to E-cadherin and ZO-1 was performed to detect differences between WT, Tbx1 and Foxi3 null mutant embryos at E9. 5. While individual arches were present in the PA in WT embryos, segmentation of the distal PA, PA2-6, did not occur in Tbx1 or Foxi3 null mutant embryos at this stage (Fig 5A–5C). In the Tbx1-/- and Foxi3-/- embryos, there was a multilayered stratified epithelium, possibly in areas where cells would begin to invaginate, which appeared thicker than in WT embryos (Fig 5A–5C). To examine whether multilayers of epithelia are present as a normal part of epithelial cell dynamics, we carefully examined the PA segments in WT embryos at E9. 5 (Fig 5A and 5D–5F). During normal segmentation of the distal PA, the endoderm and ectoderm invaginate toward each other to form a pouch and cleft, which becomes juxtaposed thereby providing a physical separation of the rostral and caudal arch. This process occurs dynamically in a rostral to caudal manner over time, such that the process of segmentation of different arches can be observed at one stage. The junction of the pouch and cleft between PA2 and PA3 was completely formed by E9. 5 and consisted of a tightly organized intercalated dual layer of cells, likely one layer of endoderm and one of ectoderm, in which ZO-1 is expressed at the outer face of each layer, on their apical surface (Fig 5D and 5D’). To determine the process of segmentation, we examined the region where the junction of the pouch and cleft between PA3 and PA4 was forming (Fig 5E and 5E’). We noted that there were multiple layers of a partially stratified epithelium where the pharyngeal epithelia became juxtaposed to each other (Fig 5E and 5E’). The outer cells expressed ZO-1 on their apical surface, but the inner cells did not (Fig 5E and 5F). It was as though a zippering process was beginning in the central region such that ZO-1 negative cells rostrally and caudally were being pushed out. More caudally, epithelial cells on either side of the mesenchyme appeared to be extending processes towards each other at the point where the cells were beginning to invaginate to form the next segment (Fig 5F and 5F’). Thus, the process of segmentation involves cell movement and repositioning as well as communication of the endoderm and ectoderm in order to form mature pouch-cleft junctions. Additional stages of E8. 5, E8. 75, E9. 0, E10. 0 and E10. 5 were examined to further characterize epithelial cell dynamics and ascertain whether there were fundamental differences at different stages in WT embryos (Fig 5G–5R). The first transition of pouch morphogenesis began at E8. 5 when the endoderm and ectoderm initiate the process of invagination to eventually separate PA1 from PA2 (Fig 5G and 5H). We did not observe multilayers of epithelial cells at E8. 5 (Fig 5G and 5H). As invagination was completing between PA1 and PA2, at E8. 75 (Fig 5I), the pharyngeal endoderm and ectoderm, consisting of two to a few layers, became loosely juxtaposed (Fig 5J and 5K). This is similar to the pouch-cleft formation for PA3-4 at E9. 5 (Fig 5E and 5E’). At E9. 0 (Fig 5L), the junction between first pouch and cleft that separated PA1 and PA2 formed a tight intercalated dual cell layer (Fig 5M) that appears similar to that observed in Fig 5D and 5D’. At E9. 0, the second pouch and cleft between PA2-PA3, consisting of a few layers of epithelium, became juxtaposed and intercalated (Fig 5N). At E9. 0 and E9. 5, there were two to three layers of endoderm cells in the caudal PA, as compared to less layers in the rostral PA, where the mature pouch-cleft junction occurred. At E10. 0, the pouches and clefts formed a pouch-cleft junction that was almost mature in between PA3-4 (Fig 5O and 5P). At E10. 5, PA formation was complete with the presence of dual layer mature pouch-cleft junctions between each arch (Fig 5Q and 5R). As indicated in Fig 5B, segmentation of the distal PA in Tbx1-/- embryos did not occur and additional layers of epithelia were present at E9. 5, possibly where the cells would begin to invaginate. However, proliferation assays at E8. 5 and E9. 5 revealed no significant changes in cell proliferation in null mutant embryos as compared to controls (S5A–S5G Fig) as has been previously reported at E9. 5 [18]. ZO-1 expression was normal at E8. 5 (S5H and S5J Fig) and E9. 5 (S5I and S5K Fig), and the apical side of only the outer facing cells expressed ZO-1 (S5H–S5K Fig). The PA in Tbx1-/- embryos was shorter in length and cell number quantification revealed that there were significantly more epithelial cells within the shortened PA at E9. 5, but not at E8. 5 (S5L Fig). This suggests that more cells were packed into a smaller PA at E9. 5. We then performed endoderm specific inactivation to determine whether this process was cell type autonomous. Inactivation of Tbx1 in the PE using the Sox172A-iCre/+ allele results in a normal first and hypoplastic second arch, as compared to an absent second arch in global null mutant embryos. This is similar to the situation with endoderm inactivation of Foxi3 (S4C Fig). Invagination of the PE did not take place and this resulted in failed segmentation of the distal arches [18], as we confirmed (S5M and S5N Fig). Similar to what was observed in the global Tbx1 null mutant embryos, E-cadherin and ZO-1 expression was normal in the conditional mutant embryos as compared to the Sox172A-iCre/+; Tbx1f/+ littermates at E9. 5 and E10. 5 (S5M and S5P Fig). As compared to the global null mutant, we did not observe excessive multilayers in the conditional mutant embryos at E9. 5 (S5M and S5N Fig). We did observe multilayers of PE cells in the distal PA at E10. 5 as compared to controls (S5O and S5P Fig), suggesting some differences between the conditional mutant embryos compared to null mutant embryos. We next examined WT, Foxi3-/- and Sox172A-iCre/+; Foxi3f/f embryos at E8. 5-E10. 5 to understand if the defects observed in Foxi3 null mutant embryos occurred in a tissue specific manner (Fig 6). Invagination defects began at E8. 5 in both Foxi3-/- and Sox172A-iCre/+; Foxi3f/f embryos (Fig 6A, 6G and 6M). At E9. 5, regions with excessive stratified multilayers of endoderm cells, especially at the points where the cells would be turning inwards to invaginate, were found in Foxi3-/- (Fig 6H–6J) and Sox172A-iCre/+; Foxi3f/f embryos (Fig 6N–6P) versus WT controls (Fig 6B–6D). All endoderm cells in WT, Foxi3-/- and Sox172A-iCre/+; Foxi3f/f embryos, expressed E-cadherin and the outermost cells expressed ZO-1 on the apical side, indicating that cells did not lose epithelial identity or polarity (Fig 6A–6R). At E9. 5 and E10. 5, the epithelial cells in Foxi3-/- embryos began to invaginate but never advanced (Fig 6K and 6L) as compared to WT embryos (Fig 6E and 6F). At E10. 5, the epithelia partially invaginated in the Sox172A-iCre/+; Foxi3f/f embryos, that seemed more complete for the ectoderm than endoderm (Fig 6Q and 6R). Overall, this data indicates a tissue autonomous role of Foxi3 in the endoderm during segmentation of the PA. E-cadherin positive epithelial cells within the PA in WT versus Foxi3-/- and WT embryos were quantified and there was a 45% (P-value <0. 03) increase of epithelial cell number within the PA of Foxi3 null mutants as compared to controls (S6 Fig) This may be correlated with the additional cell layers observed in the mutant embryos. There was a 40% increase in the number of endodermal cells (P-value <0. 006), but no increase (P-value <0. 4) was detected of ectoderm cells in the PA of Sox172A-iCre/+; Foxi3f/f mutant embryos versus controls, at E9. 5 (Fig 6T). We then decided to test whether cell proliferation was increased at E8. 5 and E9. 5. For this, a pH3 antibody was used to mark proliferating cells on serial sections on mutant versus control embryos at E8. 5 and E9. 5 (Fig 6U and 6V; S6 Fig). When calculating the ratio between proliferating cells versus total E-cadherin positive cells, there was a significant increase at E8. 5 in the endoderm of Foxi3-/- and Sox172A-iCre/+; Foxi3f/f embryos versus controls (Fig 6U and 6V; S6 Fig). At E9. 5 there was no difference in proliferation between mutant and control embryos (Fig 6U and 6V; S6 Fig). This indicates that increases of cell proliferation at E8. 5 can partially explain why there is an increase of layers of epithelial cells within the PA at E9. 5 but not at E10. 5. Notch signaling is critical for many aspects of embryonic development such as for skeletal development [41,42] and cardiovascular development [21,43,44]. Notch signaling might have a possible role in thymus gland development [21,45]. It has also been shown that Notch pathway genes, Jagged1 (Jag1) and Hes1 act downstream of Tbx1 during embryogenesis [21,46,47]. We therefore wanted to determine if Jag1, Hey1, and Hes1 may be regulated by Tbx1 and Foxi3 during PA formation. To test this, we performed WMISH and RNAscope experiments on WT, Tbx1-/- and Foxi3-/- mutant embryos at E9. 5. By WMISH, Jag1, Hes1 and Hey1 expression in the pharyngeal pouch-cleft regions in WT embryos was reduced in both Foxi3 and Tbx1 null mutant embryos (Fig 7A–7I). Three-color RNAscope assays were performed on tissue sections from WT, Tbx1-/- and Foxi3-/- mutant embryos at E9. 5, to examine expression level changes of Jag1, Hes1 and Hey1 (Fig 7J–7X). As in the WMISH experiments, Jag1 expression was localized to the pouch-cleft junctions, Hes1 was expressed at a low level throughout the PA in WT embryos and Hey1 was expressed more broadly in the PA (Fig 7J–7X). Expression was quantified and we found that levels of all three genes were significantly reduced in null mutant embryos (Fig 7Y). We also examined expression patterns of additional genes that are expressed in the PE. Isl1 (Islet1) is expressed in the second heart field mesoderm and endoderm, among other tissues during early embryogenesis. Based upon WMISH, there was not a dramatic difference in expression in WT, Foxi3-/- and Tbx1-/- embryos at E9. 5 (S7A–S7C Fig). FGF signaling has been shown to act genetically downstream of Tbx1 [48,49] and Foxi3 [27]. We previously found that Fgf3 is reduced in expression when Tbx1 is inactivated [50]. Using WMISH, we also found that Fgf3 expression was reduced in Foxi3-/- embryos at E9. 5 (S7D and S7E Fig). Expression in the otic vesicle was gone because the structure doesn’t form in Foxi3 null mutant embryos. Expression of another PE specific transcription factor gene, Pax8 (Paired box 8) was reduced within the PA in both Tbx1 and Foxi3 null mutants at E9. 5 (S7F–S7H Fig). Pax8 is important for thyroid gland development and acts downstream of Foxi3 [51]. Pax9 is a transcription factor that is normally expressed within the PE and marks the pouches during embryogenesis [27]. We also confirmed that Pax9 mRNA expression is reduced but not absent in the PA in Tbx1 mutant embryos [52] in comparison to WT controls using WMISH and RNAscope (S7I–S7M Fig). Other studies reported that Pax9 expression was mis-regulated in Foxi3-/- embryos [27]. Our data indicated that Pax9 expression was reduced in Foxi3-/- embryos, although this does not rule out that it was also mis-regulated (S7J and S7M Fig). All together this shows that there are genes that act downstream of both Tbx1 and Foxi3. In addition to observing expression of known genes important for pharyngeal segmentation, we also investigated expression of activated leukocyte cell adhesion molecule (Alcam; also called CD166, Neurolin, or DM-GRASP), Ephrinb2, and Fibronectin (Fn1) in Foxi3 and Tbx1 null mutant embryos. This is because these extracellular proteins have roles in epithelial cell function and endodermal pouch formation in zebrafish [53] [54–57]. In Foxi3-/- embryos, Alcam expression was reduced in the pharyngeal endoderm (Fig 8C and 8D) as compared to heterozygous controls (Fig 8A and 8B). In contrast, Alcam protein expression appeared unchanged in Tbx1-/- embryos (Fig 8G and 8H) in comparison to WT controls (Fig 8E and 8F). In zebrafish, ephrinb2 is required to prevent epithelial cells from rearranging once PE segmentation is complete [58]. We did not observe a change of Ephrinb2 expression in epithelial cells in Foxi3+/- and WT controls as well as Foxi3-/- or Tbx1-/- mutant embryos (Fig 8A, 8C, 8E and 8G). Fibronectin protein is an extracellular matrix protein that is present outside the basal surface of the epithelia closest to the adjacent mesenchyme and is also expressed in the mesenchyme [56,57]. Fibronectin expression was absent or expression was spotty in cells adjacent to the endoderm and ectoderm in Foxi3-/- mutant (Fig 8K and 8L) versus Foxi3+/- control embryos (Fig 8I and 8J). Fibronectin expression in Tbx1-/- mutant embryos was increased (Fig 8O and 8P) as compared to control littermates (Fig 8M and 8N), which is consistent with previous findings [59]. This indicates some differences between Tbx1 and Foxi3 functions. We note that some of the changes in patterns of expression could be due to morphological defects in the null mutant embryos. During vertebrate embryonic development, the segmentation of the distal PA is needed to create individual arches that later form derivative structures including the thymus and parathyroid glands [23,60]. We revisited the process of pharyngeal segmentation to better understand the functions of Tbx1 and Foxi3. Our data indicates that there are a few major epithelial transitions required for morphogenesis, as shown in the model in Fig 9. In the first transition, invagination of the endoderm and ectoderm takes place. Next, a few layers of a partially stratified epithelium forms in the region where invagination occurs starting at E8. 75-E9. 0. The internal layers of epithelial cells in the partially stratified epithelium do not express the cell polarity protein, ZO-1. Interestingly, it appears as if endoderm and ectoderm cells extend processes towards each other as illustrated in Fig 9. In the final transition, as invagination is completed and the endoderm and ectoderm meet, the multilayers of loosely organized cells form a tightly organized dual intercalated pouch-cleft junction, in which ZO-1 is expressed on the apical surfaces. It can be hypothesized that a zippering process initiates in the center of the forming junction of the epithelia, in which cells become reorganized. E-cadherin expression remains throughout the process indicating that the cells retain at least some of their epithelial properties. A similar process has been described for branching morphogenesis to form pancreatic ducts from the distal foregut endoderm [61]. In pancreatic organogenesis, a single layer of polarized foregut endoderm will invaginate into the mesenchyme to form branches and ducts composed of differentiated cells. In this process, a single layer of polarized epithelial cells is dynamically transformed to a multilayered epithelium, followed by a second transition back to a monolayer of polarized epithelial cells in the newly formed duct [61,62]. As for the pharyngeal endoderm, the pancreatic endoderm cells express E-cadherin during this process and the row of cells on the apical side expresses ZO-1 [61,62]. A few years ago, dynamic transitions of the pharyngeal endoderm were described in zebrafish [54]. During pharyngeal pouch formation, there is a two-step transition to form a temporary stratified epithelium from two layers of cells, which revert back to two opposing layers when pouch formation is complete [54]. In zebrafish and in mouse embryos, E-cadherin is expressed throughout the segmentation process. Apical/basal polarity is lost as detected by lack of ZO-1 expression in internal epithelial cells in zebrafish [54,63] as we found in mouse embryos. In zebrafish, this process is regulated by signals from the ectoderm and mesoderm to the PE, and is in part, non-autonomous. Specifically, in zebrafish it was found that non-canonical Wnt (Wnt11r) signaling emanating from the mesoderm is required to regulate the process of segmentation [54]. Independently, Wnt4a signals from the ectoderm upstream of the extracellular matrix protein, Alcama (Alcam in mammals) and E-cadherin within cells, are needed to transition from multilayers to two cell layers [54]. Tbx1 is expressed in the mesoderm as well as the epithelia. In zebrafish, mesodermal tbx1 regulates expression of wnt11r (non-canonical Wnt) and fgf8a (Fibroblast growth factor 8a) morphogens that signal to the PE to promote pouch formation [20]. Similarly, in mice, inactivation of Tbx1 in the mesoderm results failed segmentation of the distal PA [34]. Thus, the data in zebrafish and mouse is consistent for non-autonomous roles of mesodermal Tbx1 in pharyngeal segmentation. It was previously found that Tbx1 has autonomous roles in the PE for segmentation of the distal PA using the Sox172A-iCre/+ allele [18]. We also found that invagination failed when Tbx1 was inactivated using this allele. In Tbx1 null mutant embryos, excessive multilayers formed by E9. 5. We did not observe excessive multilayers when Tbx1 was inactivated in the endoderm at E9. 5, although there were some additional layers by E10. 5. We speculate that excessive multilayer formation may be partially suppressed in endodermal conditional mutant embryos because Tbx1 is still expressed in the mesoderm. In mammals and zebrafish, there are three Foxi class genes, Foxi1, Foxi2 and Foxi3. In zebrafish, foxi1 has a similar expression pattern and function to that of Foxi3 [51]. Recently, it was found that inactivation of foxi1 resulted in a failure to transition from a multilayered epithelium to a simple dual layered pharyngeal pouch and excessive multilayers formed [64]. This is similar to our findings for Foxi3 function in mammals, however we found that Foxi3 is also required in the invagination process. Further, in zebrafish, foxi1 appears to have its major role in the ectoderm to signal non-autonomously to the PE [64]. In foxi1 null mutant zebrafish, Alcama expression was normal. In contrast, in the mouse, loss of Foxi3 in the PE resulted in failed segmentation along with reduced Alcam expression. It is possible that ectodermal Foxi3 might have important signaling roles, but these would be independent to PE functions. Nevertheless, it appears that there are some differences in the function of these homologs in different vertebrates. There are also some differences in zebrafish in regards to tbx1 function. In zebrafish, the tbx1 gene does not have an autonomous role in the PE in pouch formation [20]. In contrast, Tbx1 has autonomous roles in the PE in the mouse. Differences in Alcam expression levels in Foxi3 and Tbx1 null mutant embryos, implicate some mechanistic differences in gene function on the extracellular milieu. Eph-ephrin signaling in adjacent cells is important for cell migration. In zebrafish, EphB2 and EphB3 are required to maintain E-cadherin expression during budding morphogenesis of the endoderm from the foregut [62]. During pharyngeal pouch morphogenesis in zebrafish, EphrinB signaling is required to increase E-cadherin expression in the second transition of pharyngeal segmentation [58]. In foxi1 mutant zebrafish, expression of EphrinB2a was not changed [64]. Ephrin2a is expressed in a similar pattern to ZO-1. As in zebrafish, we found that Ephrin b2 was not reduced in Foxi3 mutant mouse embryos, and therefore is not directly implicated downstream of Foxi3. In zebrafish, foxi1 in the pharyngeal ectoderm initiates wnt4a signaling and that this is required for the second transition of the endoderm to form a final mature segment [64]. In endoderm specific Foxi3 conditional mutant mouse embryos, the ectoderm was able to invaginate properly, but segmentation failed. It is possible that endodermal cells might not have been able to properly respond to signals emanating from the ectoderm. Our data also indicates that invagination of the ectoderm is not dependent on invagination of the endoderm. Rather, the ectoderm can initiate invagination independently. This is consistent with studies performed in shark and chick embryos where the endoderm remains separate from the ectoderm throughout epithelial cell invagination [23]. Further work will need to be done to understand the role of extracellular matrix proteins and signaling on epithelial cell dynamics in mouse models. The FGF signaling pathway was previously found to be disrupted in the pharyngeal epithelia of both Tbx1 and Foxi3 null mutant embryos. Fgf8 [49] is expressed in the pharyngeal epithelia as well as the mesoderm and is required for early zebrafish [65] and mouse embryogenesis [66–68]. Inactivation of Fgf8 in the pharyngeal epithelia in mouse embryos resulted in similar phenotypic defects in the distal pharyngeal apparatus as the Tbx1 null mutant embryos [69]. Previous published work shows that Fgf8 expression is reduced in the pharyngeal pouches in Tbx1 null mutant mouse embryos and the two genes, Tbx1 and Fgf8 genetically interact [48]. Both fgf3 and fgf8 are required for segmentation of the distal PA in zebrafish [7]. Relevant to this report, Fgf8 expression was reduced Foxi3 null mutant mouse embryos and addition of exogenous fgf3 partially rescued defects in foxi1 morphants [27]. As expected, in this report, we found that Fgf3 is reduced in expression in Foxi3 mutant embryos. This suggests that Tbx1 and Foxi3 might act in the same genetic pathway as Fgf3 and Fgf8 as well as other FGF ligand genes. We found Pax8 and Pax9 expression was also reduced in the PE in Foxi3 and Tbx1 null mutant embryos, and it is possible that in particular, Pax9 is critical for PA segmentation [70]. In addition, we found genes in the Notch pathway reduced in expression in both Tbx1 and Foxi3 null mutant embryos as well. The Notch pathway has many diverse roles in embryogenesis by regulating Notch effectors of the Hey/Her/Hes class of transcription factors. We previously noted that Jagged1 (Jag1), encoding a cell surface Notch ligand, was reduced in expression the pharyngeal arch epithelia in Tbx1 null mutant embryos [46]. In another report, expression of Hes1, encoding a Notch downstream effector, was reduced in expression in Tbx1 null mutant embryos and further, Hes1 null mutant embryos had similar thymus and aortic arch artery defects as Tbx1 null mutant embryos [21]. More recently, it was reported that Notch pathway genes were altered downstream of failed pharyngeal segmentation in mouse embryos due to inactivation of the transcription co-activator, Eya1 in mouse models [71]. In that study, Jag1, Hes1 and Hey1 expression was altered or reduced in the pharyngeal epithelia [71]. In this report, we found that Jag1, Hes1 and Hey1 were reduced in expression in Tbx1 and Foxi3 null mutant mouse embryos. Tbx1 is still expressed in the abnormal pharyngeal epithelia in Foxi3 null embryos, suggesting that perhaps the loss of expression that is observed might be due to downregulation of expression of Notch pathway genes and other genes described above. It is possible that some of the defects that were observed could be due to reduction of Notch signaling. Although our data support a possible role, it is not known if Notch pathway genes are required for segmentation of the PA. Therefore, more studies are needed to be done in the future to test a possible role for Notch signaling in this process. Patients with 22q11. 2DS have defects within structures derived from the PA including craniofacial dysmorphism, T-cell deficiencies or dysfunction, hypocalcemia, as well as aortic arch and cardiac outflow tract defects [72]. TBX1 is the major candidate gene for these defects, and it is required for PA segmentation [17,18,20]. Based on results presented in this report, we suggest that Tbx1 may act upstream of Foxi3 in this process. One question is whether individuals might be identified that have mutations in FOXI3. There has been one report of a patient with a deletion of one allele of FOXI3 that had severe ear defects, mild craniofacial defects, and missing arteries derived from PA1 and PA2 [73]. These symptoms are due to defects of structures derived from the PA but are different from those typically observed in patients with 22q11. 2DS. The phenotypic expression of 22q11. 2DS varies extensively, implicating the existence of genetic or environmental modifiers. It would be interesting to determine whether DNA sequence variants in FOXI3 or other downstream genes, such as FGF pathway genes, PAX9, or Notch pathway genes, might act as potential modifiers of phenotype in individuals with 22q11. 2DS. Analysis of sequence from a large cohort of individuals with 22q11. 2DS will be required to test this possibility. All experiments using mice were carried out according to regulatory standards defined by the National Institutes of Health and the Institute for Animal Studies, Albert Einstein College of Medicine (https: //www. einstein. yu. edu/administration/animal-studies/), IACUC protocol # 2016–0507. The following mouse mutant alleles used in this study have been previously described: Foxi3f/f (flox = f), Foxi3+/- [27], Tbx1Cre/+ [36] Sox172A-iCre/+ [38], Tbx1+/- [11] and ROSA26GFPf/+ (RCE: loxP) [74]. Foxi3-/- embryos were generated by inter-crossing Foxi3+/- mice. Double Tbx1 and Foxi3 heterozygous embryos were generated by inter-crossing Tbx1+/- and Foxi3+/- mice. Tbx1Cre/+; Foxi3f/f and Sox172A-iCre/+; Foxi3f/f embryos were generated by crossing male Tbx1Cre/+; Foxi3f/+ or Sox172A-iCre/+; Foxi3f/+ mice with Foxi3f/f females. Foxi3+/-, Tbx1+/-, Tbx1Cre/+; Foxi3f/+, Sox172A-iCre/+; Foxi3f/+, Sox172A-iCre/+; Tbx1f/+ and wildtype littermates were used as controls for the experiments, as indicated. The Foxi3+/-, Sox172A-iCre/+, and Tbx1Cre/+ mice were backcrossed 10 generations to a Swiss Webster background from a mixed C57Bl/6, Swiss Webster background. The PCR strategies for mouse genotyping have been described in the original reports and are available upon request. Mouse embryos were isolated in phosphate-buffered saline (PBS) and fixed overnight in 10% neutral buffered formalin (Sigma Corp.). Following fixation, the embryos were dehydrated through a graded ethanol series, embedded in paraffin and sectioned at 10 μm. All histological sections were stained with hematoxylin and eosin (H&E) using standard protocols in the Einstein Histopathology Core Facility (http: //www. einstein. yu. edu/histopathology/page. aspx). A total of 80 embryos, including controls, at E15. 5 were obtained from more than 10 independent crosses and analyzed morphologically using light microscopy. Fisher’s exact test was used to determine if parathyroid and thymus gland defects were significant in Tbx1+/-; Foxi3+/- compared to Tbx1+/- embryos. RNAscope in situ hybridization with non-radioactive mRNA probes was performed as previously described [75]. Tissue was fixed in 4% paraformaldehyde (PFA) for 24 hours at 4°C and then cryopreserved in 30% sucrose in PBS overnight at 4°C. Embryos were embedded in OCT and cryosectioned at 10 μm thickness. RNAscope probes for Tbx1, Foxi3, Pax9, Hey1, Hes1, Jag1, Foxn1, and Gcm2 were generated by Advanced Cell Diagnostics. Quantification was performed using Volocity Software (Perkin Elmer Corporation) where each nuclei and mRNA signal dot were counted. Each probe was calculated separately. The ratios of cell number to number of signal dots was calculated for each embryo (n = 3). P-values were determined using the Student’s t-test. Whole-mount RNA in situ hybridization with non-radioactive probes was performed as previously described [76,77], using PCR-based probes for Foxi3 [24], Tbx1 [78], Jag1 [79], and Hey1 [80]. The probe for Hes1 was generated from a cDNA plasmid clone [21,81]. Following the whole mount RNA in situ hybridization protocol, the embryos were fixed in 4% PFA and then dehydrated through a series of graded ethanol steps, embedded in paraffin, and sectioned at 10 μm thickness. Minimum of 2–4 embryos from 2–3 independent litters were analyzed for each experiment. Embryos were collected at various stages: E8. 5 (7–10 somite pairs), E9. 0 (15–19 somite pairs), E9. 5 (20–23 somite pairs) E10. 0 (24–29 somite pairs), and E10. 5 (30–33 somite pairs). Fixation was carried out in 4% PFA in PBS at 4°C for two hours. After fixation, tissue was washed in PBS and then cryoprotected in 30% sucrose in PBS overnight at 4°C. Embryos were embedded in OCT and cryosectioned at 10 μm. After fixation, frozen sections were obtained as described and then permeabilized in 0. 5% Triton X-100 for 5 min. Blocking was performed with 5% goat serum in PBS/0. 1% Triton X-100 (PBT) for 1 hour. Primary antibody was diluted in blocking solution and incubated for 1 hour. Primary antibodies used included: E-cadherin (BD Transduction laboratories 610181,1: 200 mouse), ZO-1 (Invitrogen 61–7300,1: 200 rabbit), Fibronectin (ab2413,1: 100 rabbit), Alcam (R&D BAM6561,1: 50 mouse), Ephrin b2 (ab150411,1: 200 rabbit) and GFP (ab6673 1: 500 goat). Proliferation of cells was assessed by immunofluorescence using antibody anti-phospho Histone H3 (Ser10), which is a mitosis marker (06–570 Millipore). Sections were washed in PBT and incubated with a secondary antibody for 1 hour. Secondary antibodies used were Alexa Fluor 488 goat a-mouse IgG (Invitrogen A32723) at 1: 500 and Alexa Fluor 568 donkey a-rabbit IgG (Invitrogen A11019) at 1: 500. Slides were mounted in hard-set mounting medium with DAPI (Vector Labs H-1500). Images were then captured using a Zeiss Axio Observer microscope with an apotome. To count epithelial cell number, we obtained 10 μm serial coronal sections of control, Tbx1-/-, Foxi3-/-, and Sox172A-iCre/+; Foxi3f/f embryos, which were collected and stained with an antibody for E-cadherin. To ensure that the cell quantification was accurate, we counted E-cadherin positive cells in the PA in every other section throughout each embryo. We did not count epithelial cells that were not part of the PA. When counting cells, we matched the embryos by stage using somite counts, and we matched the sections by position within the embryo. We also ensured that for each pair of control and mutant embryos, we counted the same number of sections (10–12 per embryo). We counted all phosphoH3 positive epithelial cells in each section and calculated the ratio of proliferating cells within the pharyngeal epithelium. For Tbx1-/- mutant embryos, since the PA is shorter in comparison to WT littermates, we also calculated the size of the PA using ImageJ. We then counted the E-cadherin positive cells of the PA, marking the epithelium, and divided this number by the size of the PA. The mean and standard error of the average cell counts for controls and mutant embryos were determined and they were compared using the t-test. Representations of the complete PA region from at least 3–6 embryos per genotype from at least 3 independent litters were used in each assay.

The mechanisms required for segmentation of the pharyngeal apparatus (PA) to individual arches are not precisely delineated in mammalian species. Using mouse models, we found that two transcription factor genes, Tbx1, the gene for 22q11. 2 deletion syndrome and Foxi3, genetically interact in the third pharyngeal pouch endoderm during thymus and parathyroid gland development. When examining wildtype embryos, we found that each arch is surrounded by epithelial cells derived from the endoderm and ectoderm that undergo dynamic processes during PA segmentation. Invagination co-occurs with formation of multilayers of epithelia that become juxtaposed. The cells are then reorganized to form a dual-layer of tightly intercalated cells at the mature pouch-cleft junction. In Tbx1 and Foxi3 null mutant embryos, these processes are disrupted. Further, the endoderm cells form extensive multilayers in the region where cells should normally invaginate. When examining genes that may act downstream of Tbx1 and Foxi3 we found several including Notch pathway genes Jag1, Hes1, and Hey1 are downregulated in both mutant embryos. Together, we show that Tbx1 and Foxi3 are important for regulating PA segmentation through cellular and genetic mechanisms that may be critical in 22q11. 2 deletion syndrome patients.

Abstract Introduction Results Discussion Materials and methods

medicine and health sciences fish morphogenic segmentation immunology parathyroid vertebrates animals epithelial cells animal models osteichthyes developmental biology model organisms experimental organism systems embryos morphogenesis research and analysis methods ectoderm embryology animal cells animal studies thymus biological tissue immune system zebrafish eukaryota cell biology anatomy endoderm endocrine system epithelium biology and life sciences cellular types organisms

2019

Tbx1 and Foxi3 genetically interact in the pharyngeal pouch endoderm in a mouse model for 22q11.2 deletion syndrome

Zika virus (ZIKV) is a mosquito-borne flavivirus distributed throughout much of Africa and Asia. Infection with the virus may cause acute febrile illness that clinically resembles dengue fever. A recent study indicated the existence of three geographically distinct viral lineages; however this analysis utilized only a single viral gene. Although ZIKV has been known to circulate in both Africa and Asia since at least the 1950s, little is known about the genetic relationships between geographically distinct virus strains. Moreover, the geographic origin of the strains responsible for the epidemic that occurred on Yap Island, Federated States of Micronesia in 2007, and a 2010 pediatric case in Cambodia, has not been determined. To elucidate the genetic relationships of geographically distinct ZIKV strains and the origin of the strains responsible for the 2007 outbreak on Yap Island and a 2010 Cambodian pediatric case of ZIKV infection, the nucleotide sequences of the open reading frame of five isolates from Cambodia, Malaysia, Nigeria, Uganda, and Senegal collected between 1947 and 2010 were determined. Phylogenetic analyses of these and previously published ZIKV sequences revealed the existence of two main virus lineages (African and Asian) and that the strain responsible for the Yap epidemic and the Cambodian case most likely originated in Southeast Asia. Examination of the nucleotide and amino acid sequence alignments revealed the loss of a potential glycosylation site in some of the virus strains, which may correlate with the passage history of the virus. The basal position of the ZIKV strain isolated in Malaysia in 1966 suggests that the recent outbreak in Micronesia was initiated by a strain from Southeast Asia. Because ZIKV infection in humans produces an illness clinically similar to dengue fever and many other tropical infectious diseases, it is likely greatly misdiagnosed and underreported. Zika virus (ZIKV) is a member of the Spondweni serocomplex within the genus Flavivirus, family Flaviviridae [1]. Other mosquito-borne flaviviruses of public health importance include yellow fever, dengue, St. Louis encephalitis, West Nile and Japanese encephalitis viruses. Although research efforts have focused on many of these viruses, other medically important members of the mosquito-borne flaviviruses, such as ZIKV, have received far less attention. Zika virus was first isolated from a sentinel rhesus monkey placed in the Zika Forest near Lake Victoria, Uganda in April 1947; a second isolation from the mosquito Aedes africanus followed at the same site in January 1948 [2]. Since that time, sporadic isolations have been made from humans and a variety of mosquito species in both Africa and Asia, with studies of human and animal seroprevalence confirming this distribution (Table 1). Zika virus is most likely maintained in a sylvatic cycle involving non-human primates and mosquitoes [3], [4], with cyclic epizootics in monkeys reported in Uganda [5], [6], [7], [8]. In the sylvatic transmission cycle, humans likely serve as incidental hosts. However, in areas without non-human primates, humans probably serve as primary amplification hosts and potentially as reservoir hosts if their viremia is sufficient in duration and magnitude [9]. Although it is thought that enzootic ZIKV is maintained primarily in a monkey/mosquito transmission cycle, antibodies have been detected in numerous other animal species including water buffalo, elephants, goats, hippos, impala, kongoni, lions, sheep, rodents, wildebeest, and zebras [5], [10]. Human case reports of clinically diagnosed ZIKV infections include self-limiting acute febrile illnesses with fever, headache, myalgia and rash, similar to that caused by many other arboviruses found throughout the tropics [9], [11], [12], [13], [14], [15]. This clinical picture could easily be mistaken for dengue (DEN) or chikungunya (CHIK) fevers, two common arboviral infections which both produce similar clinical presentations. The latter two infections are much more commonly diagnosed in tropical Africa and Asia than ZIKV. Clinical DENV and CHIKV infections are familiar to local clinicians and most diagnostic laboratories can detect them. In contrast, few physicians are aware of ZIKV and few laboratories test for clinical infection. Consequently, most ZIKV infections are probably missed or incorrectly diagnosed, as suggested by the high prevalence of ZIKV antibodies found in serosurveys of human populations in Africa and Asia (Table 1). A recent epidemic on Yap Island, Federated States of Micronesia, and a pediatric case of ZIKV infection in Cambodia demonstrate that ZIKV is also capable of causing human disease and may be expanding its geographic distribution [9], [15]. Zika virus, has a positive-sense, single-stranded RNA genome approximately 11 kilobases in length [16]. The genome contains 5′ and 3′ untranslated regions flanking a single open reading frame (ORF) that encodes a polyprotein that is cleaved into three structural proteins: the capsid (C), premembrane/membrane (prM), and envelope (E), and seven non-structural proteins (NS1, NS2A, NS2B, NS3, NS4A, 2K, NS4B, and NS5) [16]. A previous genetic study using nucleotide sequences derived from the NS5 gene indicated three ZIKV lineages: East African (one strain examined), West African (three strains examined), and Asian (one strain examined) [17]. Although ZIKV circulates widely in sub-Saharan Africa and Southeast Asia, little is known of the genetic relationships among isolates from these two geographic regions, which may have different vector/host transmission cycles. Furthermore, the geographic origin of the strain responsible for the epidemic on Yap Island epidemic and of the recent Cambodian case of ZIKV infection is unknown. To answer these questions, we determined the nucleotide sequences of the ORF of five ZIKV strains collected between 1947 and 2010 in Cambodia, Malaysia, Nigeria, Uganda, and Senegal and constructed phylogenetic trees to assess their relationships. The five strains sequenced in this study were obtained from the World Reference Center for Emerging Viruses and Arboviruses (WRCEVA) at the University of Texas Medical Branch (Table 2). The viruses were passaged in cell culture and harvested following the observation of diffuse cytopathic effect. Viral RNA was extracted from cell culture supernatants using the QIAamp Viral RNA Kit (Qiagen, Valencia, CA, USA). The ORF of the five viruses were amplified using the Titan One Tube PT-PCR System (ROCHE, Mannheim, Germany) and primers designed against conserved sequences to produce overlapping genome segments using African and Asian ZIKV strains: MR 766 (Prototype, Uganda, 1947, GenBank accession number AY632535) and EC Yap (Yap Island, Micronesia, 2007, GenBank accession number EU545988). Purified DNA was then sequenced using the PCR primers and additional internal sequencing primers. The Applied Biosystems BigDye Terminator version 3. 1 Cycle Sequencing Kit (Foster City, CA, USA) and the Applied Biosystems 3500 genetic analyzer were used to sequence the amplicons. Nucleotide sequences derived from the five ZIKV strains were assembled and aligned with three other sequences of ZIKV and one sequence of Spondweni virus that were retrieved from GenBank using the Vector NTI Suite (Invitrogen, USA) (Table 2). The two GenBank sequences of the MR766 strain exhibited considerable nucleotide and amino acid variation; therefore we resequenced this strain (Table 2). Phylogenies were generated using neighbor-joining (NJ), maximum-likelihood (ML) and maximum-parsimony (MP) methods using the default settings implemented in the PHYLIP package [18]. The Spondweni virus strain SM-6 V-1 was used as the outgroup for all phylogenies, as Spondweni virus is the most closely related flavivirus (antigenically and genetically) to ZIKV [1], [19], [20], [21], [22]. Robustness of the phylogenies was evaluated by resampling with 1,000 bootstrap replicates and horizontal branches were scaled according to the number of nucleotide substitutions per site. All three methods of phylogenetic inference (NJ, ML and MP) identified two major lineages (African and Asian) (Figure 1). The most recent common ancestor of MR 766 (Uganda, 1947) diverged first, followed by the divergence of the most recent common ancestor of the ArD 41519 (Senegal, 1984) and IbH 30656 (Nigeria, 1968) strains, the P6-740 strain (Malaysia, 1966), and lastly the EC Yap (Micronesia, 2007) and FS13025 (Cambodia, 2010) strains. Based on nucleotide and amino acid sequence composition, the African strains were the most divergent from the Asian strains, and strains from the same geographic regions were the least divergent (Africa and Asia) (Table 3). There were several deduced amino acid differences among the strains, which in turn correlated to geographic area of virus collection. The two MR 766 strains that had been previously sequenced exhibited extensive genetic variation (6. 3% nucleotide, 1. 8% amino acid divergence) (Table 2). To investigate this discrepancy (Table 2), we sequenced an additional strain of MR 766 from the WRCEVA collection. The MR 766 sequence with accession number AY632535 [16], was ultimately chosen for use in our analyses due to its low passage history, nucleotide and amino acid similarity to the high passage MR 766 strain that we sequenced (0. 4% nucleotide and 0. 6% amino acid divergence), and its position closest to the root of the MR 766 lineage in a tree including all three sequences (not shown). Deletions in a potential glycosylation site of several strains were observed following their alignment (Figure 2). A 4-codon deletion was observed beginning at amino acid position 153 of the E protein of the MR766 strain (GenBank accession number AY632535), a 6-codon deletion at position 156 of another MR 766 strain (GenBank accession number DQ859059), and a 6-codon deletion at position 156 of the IbH 30656 strain sequenced in this study. The MR 766 strain (Passage 147) sequenced here did not exhibit any deletions in the predicted amino acid sequence and provided evidence of passage-associated changes in potential glycosylation site (s). Prior to this study, only two ZIKV strains had been genetically characterized [16], [17], [22]. Our phylogentic analyses with the inclusion of five newly sequenced strains revealed the existence of two major ZIKV lineages. Analyses showed the basal divergence of the African and Asian lineages. Furthermore, results of the present study show that the FSS13025 (Cambodia, 2010), P6-740 (Malaysia, 1966) and EC Yap (Micronesia, 2007) strains belong to the Asian lineage, suggesting a recent common ancestor. The clinical similarity of ZIKV infection to classical DEN fever and CHIK fever may be one reason why this disease has rarely been reported in Asia. During World War II, for example, DEN fever was a major medical problem among Allied and Japanese troops in Southeast Asia and the South Pacific [23]. At that time, there were no specific laboratory tests that could differentiate between the three diseases. In fact, CHIKV and ZIKV still had not been isolated. Consequently, cases of acute febrile illness were likely diagnosed clinically as DEN or possibly as scrub typhus or malaria, diseases that were known and frequently diagnosed among the many foreign military and civilian personnel present in the region at that time [23]. Carey has reviewed the historical confusion in differentiating DEN fever from CHIK fever [24]; but ZIKV infections have probably been misdiagnosed or not reported for some of the same reasons. Given the low level of nucleotide divergence among the ZIKV isolates here (≤11. 7%), conserved regions could be utilized for the development of diagnostic assays that will not only aid in detecting new ZIKV infections but to also differentiate them from other arbovirus infections. Our results strengthen previous epidemiologic evidence that the EC Yap strain originated in Southeast Asia [9], [17]. This conclusion is further substantiated by the geographic proximity of Yap Island to known areas of ZIKV transmission (Indonesia and Malaysia). It has been reported that wind-blown mosquitoes can travel distances of several hundred kilometers over the open ocean [25], [26]. However, due to the great distances involved, it seems likely that the virus was introduced as a result of travel or trade activities whereby either a viremic person, enzootic host species, and/or an infected and subsequently infective mosquito (adult or immature) was transported to the island as suggested by Duffey et al. [9]. This hypothesis is further supported by the fact that no monkeys were present on Yap Island during the 2007 epidemic [9]. The phylogenetic results indicate that the Cambodian strain diverged from the Malaysian strain in the recent past. Therefore, the most recent common ancestor of the Cambodian strain has been circulating in Southeast Asia since at least the mid-1900' s. These data indicate that Cambodian strain was either recently introduced or that it has been circulating in the region and has remained undetected until 2010. Seroprevalence surveys might help to determine when ZIKV was introduced into Cambodia. Several of the ZIKV strains we analyzed exhibited the deletion of a potential N-linked glycosylation site that has been previously described in some ZIKV and West Nile virus strains [16], [17]. It has been hypothesized that extensive mouse brain or cell culture passage could lead the deletion of the potential glycosylation site [27]. Therefore, it is important to note that several of the strains in our analyses had previously undergone mouse brain passages (MR 766, IbH 30656, and P6-740) (Table 2). Of these strains, two different sequences of the MR 766 strain (s) AY632535 and DQ859059 [16], [22], and the IbH 30656 strain sequenced in this study had a deletion in the potential N-linked glycoslyation site. The high passage MR 766 strain that we sequenced, did not exhibit this deletion. These results provide strong evidence that passage history has influenced glycoslyation sites in the MR766 strain. Since all of the MR766 strains have undergone passage in mouse brains it is impossible to determine if the deletion was present in the original strain, as is also the case for the IbH 30656 strain. Further sequencing of geographically distinct, low passage strains that have not undergone mouse brain passage is needed to ultimately resolve whether this glycoslyation site polymorphism occurs in circulating strains or if it only reflects passage history. We had access to only a small number of ZIKV strains. However, these strains were broadly distributed over time and space, and the phylogenetic analyses were robust. Several of the ZIKV strains had been passaged intracranially in mice, and included two highly passaged strains, IbH 30656 (Nigeria, 1968, passage history: suckling mouse 21, Vero 1) and MR 766 (Uganda, 1947, passage history: suckling mouse 146, C6/36 # 1). It is likely that nucleotide/amino acid changes may have resulted from the passage history of the IbH 30656 strain, which may have slightly influenced the corresponding, terminal branch lengths in our tree but not its overall topology. This investigation indicates that the Yap Island epidemic, which occurred in the Federated States of Micronesia in 2007, most likely resulted from the introduction of a Southeast Asian ZIKV strain (s) pointing to an expansion of the Asian ZIKV lineage. Although ZIKV has one of the earliest and best-documented widespread geographic distributions among arboviruses, many unanswered questions remain concerning its evolution, ecology and epidemiology. In Asia, evidence suggests that the primary mosquito vectors are Ae. aegypti and/or Ae. albopictus [13], [28], though several ecologically or geographically distinct mosquito vectors may be responsible for the transmission and/or maintenance of ZIKV throughout Asia. As such, further studies are needed to determine the primary and secondary mosquito vectors responsible for ZIKV transmission throughout the Asian region. In addition, human seroprevalence studies throughout Asia may provide insight into the expansion of the Asian lineage and clues as to why certain geographical regions maybe more suitable for virus maintenance and transmission than others. Additional work is needed to better understand the clinical presentation, tropism and pathogenesis of ZIKV infection in humans. Finally, continued ZIKV isolations in currently affected regions coupled with active surveillance in presently naïve areas will allow researchers to follow the possible geographic expansion of the virus and predict the potential emergence of ZIKV into uncharted territories. The views expressed in this article are those of the authors and do not reflect the official policy or position of the Kingdom of Cambodia, U. S. Department of Defense, or the Department of the Navy.

Zika virus (ZIKV) is a mosquito-transmitted flavivirus found in both Africa and Asia. Human infection with the virus may result in a febrile illness similar to dengue fever and many other tropical infections found in these regions. Previously, little was known about the genetic relationships between ZIKV strains collected in Africa and those collected in Asia. In addition, the geographic origins of the strains responsible for the recent outbreak of human disease on Yap Island, Federated States of Micronesia, and a human case of ZIKV infection in Cambodia were unknown. Our results indicate that there are two geographically distinct lineages of ZIKV (African and Asian). The virus has circulated in Southeast Asia for at least the past 50 years, whereupon it was introduced to Yap Island resulting in an epidemic of human disease in 2007, and in 2010 was the cause of a pediatric case of ZIKV infection in Cambodia. This study also highlights the danger of ZIKV introduction into new areas and the potential for future epidemics of human disease.

Abstract Introduction Methods Results Discussion

medicine infectious diseases neglected tropical diseases vectors and hosts

2012

Genetic Characterization of Zika Virus Strains: Geographic Expansion of the Asian Lineage

The conserved TOR kinase signaling network links nutrient availability to cell, tissue and body growth in animals. One important growth-regulatory target of TOR signaling is ribosome biogenesis. Studies in yeast and mammalian cell culture have described how TOR controls rRNA synthesis—a limiting step in ribosome biogenesis—via the RNA Polymerase I transcription factor TIF-IA. However, the contribution of TOR-dependent ribosome synthesis to tissue and body growth in animals is less clear. Here we show in Drosophila larvae that ribosome synthesis in muscle is required non-autonomously to maintain normal body growth and development. We find that amino acid starvation and TOR inhibition lead to reduced levels of TIF-IA, and decreased rRNA synthesis in larval muscle. When we mimic this decrease in muscle ribosome synthesis using RNAi-mediated knockdown of TIF-IA, we observe delayed larval development and reduced body growth. This reduction in growth is caused by lowered systemic insulin signaling via two endocrine responses: reduced expression of Drosophila insulin-like peptides (dILPs) from the brain and increased expression of Imp-L2—a secreted factor that binds and inhibits dILP activity—from muscle. We also observed that maintaining TIF-IA levels in muscle could partially reverse the starvation-mediated suppression of systemic insulin signaling. Finally, we show that activation of TOR specifically in muscle can increase overall body size and this effect requires TIF-IA function. These data suggest that muscle ribosome synthesis functions as a nutrient-dependent checkpoint for overall body growth: in nutrient rich conditions, TOR is required to maintain levels of TIF-IA and ribosome synthesis to promote high levels of systemic insulin, but under conditions of starvation stress, reduced muscle ribosome synthesis triggers an endocrine response that limits systemic insulin signaling to restrict growth and maintain homeostasis. Nutrient availability is a critical determinant of cell, tissue and body growth in developing animals. Nearly two decades of research has identified the Target-of-Rapamycin (TOR) kinase signaling pathway as a major nutrient-responsive growth pathway in eukaryotes [1], [2]. TOR functions in two distinct complexes – TORC1 and TORC2 – and it is TOR kinase activity specifically within TORC1 that has been established as a growth driver. A complex intracellular signaling network activates TOR kinase activity within TORC1 in response to availability of extracellular nutrients such as amino acids and glucose. TORC1, in turn, stimulates many cell metabolic processes that drive growth and proliferation [2], [3]. In contrast, when nutrients are limiting, TORC1 activity is inhibited and cells switch their metabolism to promote homeostasis and survival during starvation conditions. One important metabolic target of nutrient/TOR signaling in the control of growth is ribosome biogenesis [4]–[10]. A limiting step of ribosome synthesis is the RNA Polymerase (Pol I) -dependent transcription of ribosomal RNA (rRNA). Studies predominantly in yeast and mammalian cell culture have described mechanisms by which TOR promotes rRNA synthesis [4], [5], [7]–[13]. One target of TOR signaling emerging from these studies is the Pol I-specific transcription factor Transcription Initiation Factor-IA (TIF-IA). TIF-IA associates with Pol I and recruits it to rDNA genes to initiate transcription [5], [6], [9], [14]–[16]. This function of TIF-IA is stimulated by nutrient-dependent activation of TOR, and a handful of reports have proposed mechanisms involving TOR-dependent changes in TIF-IA phosphorylation, levels or localization to rDNA genes [9], [17]. These effects may also involve TOR functioning directly at nucleolar rDNA genes [13]. While these studies provide a molecular basis for understanding how nutrients and TOR control Pol I and rRNA synthesis in cells, the contribution of rRNA and ribosome synthesis to tissue and body growth in developing animals is not as clear. Genetic studies in model organisms, most notably Drosophila, have provided most detail into how nutrient/TOR signaling controls tissue and body growth. During the four-day Drosophila larval period, animals increase in mass almost 200-fold [18]. This dramatic growth is nutrition-dependent and mostly occurs in non-dividing polyploid cells that make up the bulk of the larval organs. TOR signaling is central to this size control and functions by coupling dietary protein to growth [19]–[22]. Loss of TOR function in cells or tissues leads to a reduction in cell size or tissue mass, whereas TOR over-activation leads to increased cell and tissue growth [19]–[22]. TOR activity in specific tissues can also influence overall body size through non-autonomous endocrine or systemic effects [23], [24]. An example is the role of TOR in the larval fat body [25], [26]. When dietary proteins are abundant, amino acid uptake into fat body cells stimulates TOR activity. This triggers release of a fat-to-brain secreted signal that promotes the production and release of Drosophila insulin-like peptides (dILPs) from neurosecretory cells in the brain [25], [26]. These dILPs then circulate throughout the animal and promote growth in all larval tissues via a conserved insulin receptor/PI3K/Akt signaling pathway [27]. In contrast, when larvae are starved, TOR signaling in the fat body is suppressed leading to reduced circulating dILP levels, and decreased insulin signaling and growth. In this way, TOR activity in the fat body links nutrition to larval growth and development. TOR activity in larval muscle has also been reported to exert systemic effects to promote overall body growth and development [28]. This ability of TOR activity in specific tissues to control whole body metabolism and growth is an emerging theme in both mouse and fly genetic studies [23], [24], [29], [30], and emphasizes the importance of non-autonomous mechanisms in the control of body growth. In this paper, we describe our ongoing work exploring the role for rRNA synthesis in controlling tissue and body growth in larvae. We find that the nutrient-dependent TOR pathway is required to maintain TIF-IA mRNA and protein levels in larval tissues, especially the muscle, during development. We also find that TIF-IA-dependent ribosome synthesis is required in muscle to maintain systemic insulin signaling and promote normal body growth and development, and loss of TIF-IA in muscle blocks the body growth-promoting effects of TOR signaling. This work emphasizes the importance of non-autonomous, tissue-specific effects of ribosome synthesis on endocrine signaling and body growth during development. In previous work, we showed that the nutrient/TOR pathway controls rRNA synthesis in developing larvae and that TOR signaling promotes TIF-IA recruitment to rDNA genes [6]. Here, we examined whether TOR signaling may function by controlling TIF-IA levels. Deprivation of dietary protein leads to reduced TOR signaling and decreased rRNA synthesis in larvae. We found that under protein starvation conditions (induced by transferring larvae to a sucrose-only diet), TIF-IA protein levels were reduced compared to fully fed controls (Figure 1A). We also found that TIF-IA protein levels were also reduced in tor null mutant (torΔP) larvae compared to wild-type controls (Figure 1B). TOR can promote growth in part via its downstream effector kinase, ribosomal protein S6 kinase (S6K) [31]. However, we found that that TIF-IA protein levels were unaltered in s6k mutant (s6kL1) larvae compared to wild-type (Figure 1C). These results prompted us to examine TIF-IA mRNA levels. We found that both starved larvae and torΔP mutant larvae had reduced levels of both TIF-IA mRNA (Figure 1D, F) and pre-rRNA (Figure 1E, G) consistent with a reduction in synthesis of rRNA and hence ribosomes. Thus, during larval development nutrient/TOR signaling is required to maintain appropriate levels of TIF-IA mRNA and protein. As well as controlling cell-autonomous growth, TOR activity in specific larval tissues is required for overall body growth in Drosophila. For example, reduced TOR signaling in larval muscle [28] and fat [25], [26] leads to reduced body growth. Given the importance of ribosome synthesis as an effector of TOR in the control of cell-autonomous growth, we examined whether TIF-IA-dependent ribosome synthesis could also exert non-autonomous effects on body growth. We first examined larval muscle. As with whole larvae, we found that protein starvation decreased both TIF-IA protein (Figure 2A) and mRNA (Figure 2B), and also pre-rRNA (Figure 2C) in larval muscle. To explore the consequence of this reduction in TIF-IA levels, we examined the effects of RNAi-mediated knockdown of TIF-IA in muscle, using a UAS-TIF-IA inverted repeat (IR) line. Ubiquitous expression of this TIF-IA IR line in larvae using the daughterless (da) -GAL4 driver (da>TIFIA-IR) phenocopied tif-ia mutants, and led to reduced TIF-IA protein levels (Figure S1B) and larval growth arrest (Figure S1A). Both of these effects were fully reversed by co-expression of a UAS-TIF-IA transgene (Figure S1A), confirming the specificity of the UAS-TIF-IA IR line. We then used the UAS-TIF-IA IR line to knock down TIF-IA specifically in muscle (using the dMef2-GAL4 driver – Figure S2). We found that RNAi-mediated knockdown of TIF-IA muscle mimicked the decrease in both TIF-IA mRNA (Figure 2D) and pre-rRNA (Figure 2E) levels following starvation. When we monitored larval growth and development, we observed that dMef2>TIF-IA IR larvae were smaller than age-matched control larvae (Figure 2G). Moreover, dMef2>TIF-IA IR larvae were significantly delayed in pupal development with respect to control (dMef2>+) larvae (Figure 2F), and only approximately 20% of dMef2>TIF-IA IR larvae formed pupae. These dMef2>TIF-IA IR pupae were malformed compared to control (dMef2>+) pupae (Figure 2H). We examined feeding by transferring dMef2>+ (control) and dMef2>TIF-IA IR larvae onto yeast paste colored with blue food dye. After 4 hours, we observed that both the control and TIF-IA IR larvae contained blue food in their guts (Figure S3), suggesting that knockdown of TIF-IA in larval muscle did not impair feeding. Together, these findings indicated that TIF-IA-dependent ribosome synthesis in muscle is required to maintain normal body growth and development. We also examined the organismal effects of TIF-IA knockdown in other tissues. We used two fat body GAL4 drivers (r4-GAL4 and ppl-GAL4) to express UAS-TIF-IA IR during larval development. We found that r4>TIF-IA IR larvae showed a modest, although statistically significant delay in both developmental timing - time from larval hatching to pupation (Figure 3A) - and growth (Figure 3B), but showed no significant change in body size compared to control (r4>+) animals (Figure 3C). Co-overexpression of UAS-Tsc1 and UAS-Tsc2 - negative regulators of TORC1 – using r4-GAL4 led to marked reduction in body growth, thus confirming the effectiveness of the driver (Figure S4). We also found that ppl>TIF-IA IR larvae showed no significant difference in developmental timing (Figure 3D) or final body size (Figure 3E) compared to controls (ppl>+). We also examined the effects of TIF-IA knockdown in the larval lymph gland and hemocytes using two different drivers, hemolectin (hml) -GAL4 and peroxidasin (pxn) -GAL4. In both cases, we observed no statistically significant decrease in larval development (Figure 3F, G). In fact, larval development was modestly, although significantly, accelerated in hml>TIF-IA IR larvae. TOR activity in muscle is required for normal larval growth and development [28]. We confirmed this finding by inhibiting TOR in the muscle by two different methods, expression of a dominant negative form of TOR in muscle (dMef2>TORTED) [32] and co-overexpression of UAS-Tsc1 and UAS-Tsc2 - negative regulators of TORC1 - in muscle (dMef2>Tsc1, Tsc2). We measured pupal volume, as an indicator of final body size. Our data showed that in both cases, inhibition of TOR in larval muscle reduced pupal volume (Figure 4A, B). Amino acid availability is an important activator of TOR kinase signaling, and we also found that knockdown of the amino acid transporter slimfast (using a UAS-slifAnti antisense [25]), in the larval muscle led to a significant reduction in pupal volume (Figure 4C). Finally, we also examined whether over-activation of TOR in muscle was sufficient to drive systemic growth. We found that overexpression of Ras homolog enriched in brain (Rheb), an upstream activator specifically of TORC1, in muscle (dMef2>Rheb) was sufficient to increase pupal volume compared to control (dMef2>+) pupae (Figure 4D). Together, these findings suggest that TOR activity in muscle is both necessary and sufficient to control overall systemic growth. We next examined whether TIF-IA function was required for these muscle effects of TOR. As described above, overexpression of Rheb in muscle led to increased body size, as indicated by larger larvae (Figure 4E) and increased pupal volume (Figure 4F), while RNAi-mediated knockdown of TIF-IA showed the opposite effects. We found that co-expression of UAS-TIF-IA IR (dMef2>Rheb; TIF-IA IR) phenocopied dMef2>TIF-IA IR animals and abrogated the Rheb induced increase in body size. We quantified the pupal volume and found that reducing TIF-IA in muscle reduced the Rheb induced increase in pupal volume (Figure 4G). Overall, these data indicated that TIF-IA activity in muscle is required for TOR signaling to drive systemic growth. The insulin pathway is the major endocrine regulator of body growth in larvae. Under nutrient-rich conditions, several dILPs are expressed and released into the larval hemolymph [33]. These dILPs then bind to a single insulin receptor in target cells and promote growth [26]. In contrast, starvation leads to reduced systemic insulin signaling and decreased growth. We therefore explored whether the growth inhibitory effects of muscle-specific TIF-IA knockdown were due to reduced systemic insulin signaling. Under nutrient rich conditions, high level of insulin signaling leads to activation of Akt kinase and phosphorylation and nuclear exclusion of the FOXO transcription factor. But when insulin signaling is reduced, FOXO relocalizes to the nucleus and activates target genes such as eIF4E-Binding Protein (4EBP). Therefore, changes in FOXO nuclear localization and transcriptional activity serve as a reliable ‘read-out’ of insulin signaling [34]–[36]. As previously reported, we found that FOXO was excluded from nuclei in fat body cells from fed larvae (Figure 5A), but showed strong nuclear accumulation in fat body cells from starved larvae (Figure 5B). When we knocked-down TIF-IA levels in muscle (dMef2>TIF-IA IR), FOXO showed strong, statistically significant nuclear accumulation in fat body cells (Figure 5C, D). We next measured the levels of 4EBP, a FOXO target gene, and found that dMef2>TIF-IA IR larvae had increased 4EBP mRNA levels with respect to control (dMef2>+) larvae (Figure 5E). Finally we measured the examined levels of phosphorylated Akt – the kinase downstream of insulin signaling that is responsible for phosphorylation and inhibition of FOXO. Using western blotting with an anti-phospho Akt (Ser505) antibody, we found that dMef2>TIF-IA IR had markedly reduced levels of phospho Akt compared to control (dMef2>+) larvae (Figure 5F). Levels of total Akt were also lower, but much less so than the suppression in levels of phosphorylated Akt. Together these data suggest that TIF-IA knockdown in muscle leads to reduction in systemic insulin signaling. An important source of dILPs is a cluster of neurosecretory cells in the larval brain [33], [37]. These cells secrete three dILPs (2,3 and 5), and expression and/or release of these dILPs are suppressed upon protein starvation [26]. Moreover, loss of these neurons leads to slow growing and small larvae [33], [37], [38]. We found that dMef2>TIF-IA IR larvae had reduced dILP3 and dILP5 mRNA levels compared to control (dMef2>+) larvae, while dILP2 mRNA levels were unaltered (Figure 5G). Previous studies showed that nutrient-deprivation leads to reduced dILP2 secretion and hence increased retention in the neurosecretory cells [26]. This retention can be easily visualized by staining with anti dILP2 antibodies. Using, this approach we observed an increase in dILP2 staining in the neurons of dMef2>TIF-IA IR larvae compared to control larvae (Figure 5H, I, J). Together, these data suggest that one mechanism by which reduced TIF-IA activity in muscle suppresses peripheral insulin signaling is by reduced expression and release of brain-derived dILPs. In addition to the dILPs, other secreted factors can influence insulin signaling in Drosophila. One factor is Imaginal morphogenesis protein-L2 (Imp-L2), which is the Drosophila homolog of insulin-like growth factor binding protein 7 (IGFBP7) [39]. Imp-L2 can bind to dILPs and inhibit their ability to signal through the insulin receptor [39]–[41]. Moreover, a recent report showed that mitochondrial perturbation in adult muscle leads to increased Imp-L2 expression and subsequent suppression of systemic insulin signaling [42], [43]. We found that dMef2>TIF-IA IR larvae had upregulated Imp-L2 mRNA levels in their muscle compared to control (dMef2>+) larvae (Figure 5K). These data suggest that upregulation of Imp-L2 may provide another mechanism by which perturbation of TIF-IA in muscle suppresses systemic insulin signaling. Indeed, we found that overexpression of Imp-L2 in the muscle led to delayed larval development and reduced pupal size (Figure S5). Our data suggest that TIF-IA function in muscle is required to maintain systemic insulin signaling in fed animals. We next examined whether TIF-IA-mediated ribosome synthesis in muscle may provide one mechanism to couple nutrient availability to systemic insulin signaling. We overexpressed a UAS-TIF-IA transgene in muscle (dMef2>TIF-IA) and observed a very slight, but statistically significant acceleration in development compared to (dMef2>+) larvae (Figure S6A), although final body size was not affected (Figure S6B). Similar effects were observed with a second, independent UAS-TIF-IA transgene (Figure S6C, S6D). We then examined effects of muscle TIF-IA overexpression in starved animals. When larvae are deprived of dietary protein, insulin signaling is reduced leading to upregulated levels of FOXO target genes such as 4EBP and InR, an effect we observed here following 24 hr starvation. However, when we overexpressed TIF-IA in muscle (dMef2>TIF-IA), the starvation-induced increase in both 4EBP and InR mRNA was partially reversed compared to control (dMef2>+) larvae (Figure 6A, B). This result suggests that TIF-IA function in muscle can, in part, couple nutrient availability to systemic insulin signaling. The findings presented here suggest that TIF-IA function in muscle is required for normal nutrient-dependent systemic insulin signaling and growth. Hence, upon knockdown of TIF-IA in muscle, we saw reduced growth and delayed development. To further implicate a role for reduced insulin signaling in these effects, we tested whether restoring insulin signaling to some degree could have any effect on the phenotypes we observed. To achieve this we examined partial loss of negative regulators of insulin signaling. We first tested the effects of reducing foxo gene dosage. We found that the decrease in larval growth seen in dMef2>TIF-IA IR larvae was partially reversed in larvae that were heterozygous for a loss-of-function mutation in foxo (foxo25) (Figure 7A). We next examined the effects of reducing the levels of Imp-L2, whose expression was increased in dMef2>TIF-IA IR larval muscle. We found that co-expression of a UAS-Imp-L2 inverted repeat (IR) line with the UAS-TIF-IA IR in muscle, also partially reversed the growth defects seen with expression of UAS-TIF-IA IR alone (Figure 7B). Loss of one copy of foxo (foxo25/+) alone or expression of UAS-Imp-L2 IR alone in the muscle had no effects on larval size (Figure S7). When we measured developmental timing, we also saw that both the delayed larval development and reduced numbers of pupating larvae seen in dMef2>TIF-IA IR larvae were partially reversed in larvae that either were heterozygous for foxo25, or which co-expressed UAS-Imp-L2 IR in the muscle (Figure 7C). These experiments provide genetic evidence that muscle TIF-IA function is required for normal larval growth and development at least in part by maintaining systemic insulin signaling. The major finding of our work is that under nutrient-rich conditions TIF-IA-dependent regulation of muscle ribosome synthesis is required to maintain systemic insulin signaling and body growth. Work in yeast, mammalian cell culture and Drosophila indicates that TIF-IA links nutrient availability and TOR signaling to rRNA synthesis [4]–[10]. Here we show that in growing tissues in vivo one mechanism by which nutrient/TOR signaling functions is through maintaining TIF-IA levels. Recent studies in yeast also showed TIF-IA levels were reduced following pharmacological inhibition of TOR [17]. Moreover, in previous work, we showed that maintaining high levels of TIF-IA expression could reverse the decrease in rRNA synthesis caused by amino acid starvation in Drosophila larvae [6]. Hence, control of TIF-IA levels represents one mechanism by which nutrient availability and TOR signaling can control the synthesis of rRNA. TOR has also been reported to indirectly control site-specific phosphorylation of TIF-IA, and this phosphorylation modulates TIF-IA nucleolar localization [9]. Hence, TOR may impact TIF-IA function in several ways. When we mimicked the starvation induced decrease in muscle TIF-IA mRNA levels by RNAi-mediated knockdown, we observed that larvae were slower growing and failed to develop. This phenotype was not simply due to a gross motor defect, since the larvae were able to crawl normally and ingest food. Studies from Demontis and Perrimon [28] describe a similar reduced growth phenotype following inhibition of TOR signaling in larval muscle. Here, we extended this work to show that increased TOR in muscle leads to a larger overall body size, and that this effect required intact TIF-IA function. Our data implicate changes in insulin signaling as underlying the effects of TIF-IA-dependent ribosome synthesis in muscle on overall body growth and development. Our findings also suggest that the ability of dietary nutrients to stimulate and maintain systemic insulin rely, in part, on maintaining TIF-IA levels and function in muscle. Muscle TIF-IA appeared to control insulin signaling by at least two mechanisms. First, we saw that expression of brain-derived dILPs required normal muscle TIF-IA function. The expression and release of dILPs (2,3 and 5) from a cluster of neurosecretory cells [33], [37] in the brain is regulated by signals from other tissues. Hence, the changes in systemic insulin signaling that we saw following inhibition of TIF-IA in muscle could be explained by a role for muscle-derived secreted factors (often termed myokines). In mammals, muscle has been shown to secrete many factors, including a host of cytokines, and secretion of these factors is often controlled by nutrients [44]–[50]. In Drosophila, the full complement of factors secreted from muscle is not clear [49]. Nevertheless, one or more secreted factors could potentially signal to the brain to promote dILP release. Indeed, a recent study showed that suppression of ribosome synthesis by overexpression of Mnt in adult muscle led to release of myoglianin, a myostatin-like myokine, which induced remote effects on fat body function [51]. Also, activin signaling in adult muscle can remotely control dILP release and systemic insulin signaling [52]. Second, we saw that knockdown of TIF-IA in muscle led to an increase in expression of Imp-L2, a secreted protein that functions to suppress insulin signaling [39]. A recent paper showed that perturbation of mitochondrial function in Drosophila muscle can also lead to upregulation of Imp-L2 expression [42]. Together with our data, this finding suggests that upregulation of Imp-L2 may be a common response triggered by metabolic stress in muscle cells. Importantly, we were able to partially rescue both the reduced growth and delayed development seen with muscle knockdown of TIIF-IA by either loss of one copy of foxo or RNAi-mediated knockdown of muscle Imp-L2. In both cases, the rescue was partial probably because neither genetic manipulation would be predicted to completely restore systemic insulin signaling. Nevertheless, the findings provide further support for our model that muscle-specific ribosome synthesis can control systemic insulin signaling and body growth. A previous report described how inhibition of PI3K/TOR signaling in muscle led to both reduced muscle cell size and a non-autonomous reduction in size of other tissues and overall body size [28]. These non-autonomous effects were proposed to be mediated through altered endocrine signaling from the muscle to other tissues, although it is unclear whether this occurred solely as a result of reduced muscle cell size, or whether it reflects a cell size-independent role for TOR in controlling the endocrine function of muscle. Our findings here suggest that altered insulin signaling is one important endocrine response that links changes in muscle ribosome synthesis to altered physiology and growth in other tissues, although as with the effects of TOR we cannot discern whether this occurs only due to reduced muscle cell size. Interestingly we showed that inhibition of TIF-IA in the fat body had only a weak non-autonomous effect on body growth, although TIF-IA knockdown can limit fat cell size and ploidy [6]. Thus the mechanisms that couple TIF-IA and ribosome synthesis in muscle to the endocrine control of systemic insulin may not operate in the fat body. Ultimately, it is likely that the role for TIF-IA and ribosome synthesis in controlling overall body growth depends on a combination of cell-autonomous and non-autonomous influences. For example, inhibition of ribosome synthesis in the prothoracic gland was shown to extend larval development by altering endocrine ecdysone hormone signaling [53]. Muscle is a metabolically active tissue that probably has a high demand for continued ribosome biogenesis and protein synthesis to maintain autonomous growth. Our studies suggest that muscle ribosome synthesis may also act as a checkpoint for overall body growth. If muscle ribosome synthesis is perturbed (e. g. by nutrient deprivation), this may cause muscle cells to trigger a suppression of systemic insulin signaling to limit body growth. In using ribosome synthesis as a checkpoint for controlling systemic insulin, muscle cells may simply sense and respond to general changes in bulk translation. Alternatively, altered translation of a select subset of mRNAs may influence either Imp-L2 expression or the ability of muscle to remotely control brain dILP expression. In either case, our findings suggest that larval muscle is also an important nutrient-sensing tissue, in addition to the fat body, that can control systemic insulin signaling via endocrine signaling. The endocrine mechanisms by which either fat or muscle control systemic insulin signaling are nor clear and may be different in both cases. However, it seems that both tissues rely on protein synthesis, although perhaps through different mechanisms. Our data suggest that control of rRNA synthesis is an important limiting step in muscle, while previous work suggests that regulation of tRNA synthesis and signaling via Myc is important in fat [36], [54], [55]. The ‘checkpoint’ response to perturbation of muscle ribosome synthesis may be important for controlling not just growth, but also other organismal responses. For example, upon starvation or other environmental stressors, a reduction of muscle TIF-IA and ribosome synthesis may function to suppress systemic insulin signaling to alter whole body metabolism in order to maintain animal survival under adverse conditions. Reducing insulin signaling has been well described as mediator of stress resistance and extended lifespan in many animals including Drosophila, C. elegans and mice [56]–[58]. Indeed, a recent report showed that elevated Imp-L2 in Drosophila muscle increased adult lifespan [42], [43]. Also, overexpression of 4E-BP, a translational repressor, in muscle [59] or in whole organism [60] leads to stress resistance, and extended lifespan. Thus control of muscle protein synthesis, possibly by regulating ribosome biogenesis, may be a common mechanism to control stress responses and lifespan by regulating whole-body insulin signaling. All stocks and crosses were raised at 25°C and maintained on a media containing 100 g Drosophila Type II agar, 1200 g cornmeal, 770 g Torula yeast, 450 g sugar, 1240 g D-glucose, 160 ml acid mixture of propionic acid and phosphoric acid per 20 L of water. The following fly stocks were used: w1118; yw; UAS-TIF-IA; UAS-TIF-IA IR (v20334, Vienna Drosophila RNAi Center, VDRC); UAS-Tsc1, UAS-Tsc2; torΔP/CyO; s6kL1/TM6B; UAS-Rheb; UAS-slifAnti; UAS-TORTED; UAS-GFP; UAS-Imp-L2 IR (15009-R3, NIG, Japan); foxo25/TM6B, dMef2-GAL4; da-GAL4; r4-GAL4; ppl-GAL4; hml-GAL4; pxn-GAL4. For all GAL4/UAS experiments, homozygous GAL4 lines were crossed to the relevant UAS line (s) and the larval progeny were analyzed. Control animals were obtained by crossing the relevant homozygous GAL4 line to either w1118; +; + or yw; +; +, depending on the genetic background of the particular experimental UAS transgene line. Adult flies were allowed to lay eggs on grape juice agar plates supplemented with yeast paste for 4 hours (hr) at 25°C. 24 hr after egg laying (AEL) all hatched larvae were transferred to food vials with a thin brush, in groups of 45–50 larvae/vial and allowed to develop. For all experiments, whole larvae were starved by floating on sterile 20% sucrose in 1× Phosphate Buffered Saline (PBS) at 72 hr AEL for indicated times. Subsequently, larvae were collected and processed as per experimental requirements. Fed larvae were collected at 72 hr AEL. Whole larval or larval muscle tissue extracts were prepared by lysing 72 hr AEL larvae in 4× protein sample buffer (240 mM Tris-HCl pH 6. 8,8% SDS, 5%β-mercaptoethanol, 40% glycerol, 0. 04% bromophenol blue) with a motorized pestle, boiling for 4 minutes at 95°C and immediately loading the samples onto a SDS-PAGE gel. Immunoblotting was performed as previously described [54]. Antibodies used were βtubulin (E7, Drosophila Studies Hybridoma Bank), phospho-Drosophila Akt Ser505 (Cell Signaling Technology, 4054) and Akt (Cell Signaling Technology, 9272). Affinity-purified antibodies were generated against TIF-IA was raised by immunizing rabbits using the synthetic peptide CIVDKRPKNFDLSKSQEFDKQ, corresponding to residues 585–604 (Anaspec Inc.). Whole larval or larval muscle tissues were isolated at definite time points AEL (as indicated in the figure legends). Total RNA was extracted using TRIzol according to the manufacturer' s instructions (Invitrogen; 15596-018). RNA samples were DNase treated as per manufacturer' s protocol (Ambion; 2238G). The DNase treated RNA was reverse transcribed by Superscript II to make cDNA. This cDNA was used as a template to perform qRT-PCR reactions (BioRad Laboratories, MyIQ PCR machine using SYBR Green PCR mix) using specific primer pairs (sequences available upon request). Pre-rRNA levels were measured by using primer pairs against the internal transcribed spacer (ITS) region of 45S pre-rRNA transcript. qPCR data were normalized to β tubulin mRNA, whose levels we found were essentially unchanged across all the experimental conditions. The exception was the qPCR analyses of tor mutants, where values were corrected for actin mRNA levels. For each experiment, a minimum of 3 groups of 5–8 larvae was collected. Each experiment was independently repeated a minimum of 3 times. Larvae were collected at 24 hr AEL and placed in food vials in equal numbers per vial (with a maximum density of 50 larvae per vial). The number of pupae in vials was counted every 24 hr. For each genotype, minimum of 3 replicates were used to calculate the mean percentage of pupae per timepoint. Pupal volume was calculated as previously described [55]. Larval and pupal images were obtained using a Zeiss Stereo Discovery V8 microscope using Axiovision software. Microscopy and image capture was performed at room temperature and captured images were processed using Photoshop CS5 (Adobe). For each experiment all larval and pupal images were captured using identical magnifications. Final figures were generated from these by cropping individual larvae and then simply rotating images to orient them in the same direction, without altering size or scale. These images were then assembled on a single black canvas in Photoshop. Larval sizes were assessed by using Photoshop to measure larval body areas from these microscope images. Tissue images and Differential Interference Contrast (DIC) images were captured by taking serial Z-stacks using the same magnification and time of exposure. Larvae were inverted using fine forceps in 1× PBS at particular time points (as indicated in the figure legends). Inverted larvae were fixed in 8% paraformaldehyde for 40 minutes, washed in 1× PBS-0. 1% TritonX (PBST), blocked for 2 hr at room temperature in 1× PBST with 5% fetal bovine serum (FBS). Larvae were incubated overnight with primary antibody at 4°C, washed several times with 1× PBST and incubated with secondary antibody (1∶4000) for 2 hours, at room temperature. After few washes, fat bodies were isolated from these larvae using fine forceps and mounted on glass slides with cover slips using Vectashield (Vector Laboratories Inc. , CA) mounting media. Primary antibodies used were rabbit anti-FOXO (from Marc Tatar) and rabbit anti-dILP2. Alexa Fluor 488 and 568 (Invitrogen) were used as secondary antibodies. Hoechst 33342 (Invitrogen) was used to stain nuclei. dILP2 immunostaining of larval brains was performed as described [54]. For all experiments, error bars represent standard error of mean (SEM). P values were computed by Student' s t-test, using Microsoft Excel or Analysis of Variance (ANOVA) followed by Tukey' s post-hoc test, using GraphPad prism (version 6). For developmental timing experiments, mean time to pupation was computed using Mann-Whitney U test using GraphPad prism (version 6). P<0. 05 was considered to be statistically significant, as indicated by asterisk (*) or as indicated in the figure legend.

All animals need adequate nutrition to grow and develop. Studies in tissue culture and model organisms have identified the TOR kinase signaling pathway as a key nutrient-dependent regulator of growth. Under nutrient rich conditions, TOR kinase is active and stimulates metabolic processes that drive growth. Under nutrient poor conditions, TOR is inhibited and animals alter their metabolism to maintain homeostasis and survival. Here we use Drosophila larvae to identify a role for ribosome synthesis—a key metabolic process—in mediating nutrient and TOR effects on body growth. In particular, we show that ribosome synthesis specifically in larval muscle is necessary to maintain organismal growth. We find that inhibition of muscle ribosome synthesis leads to reduced systemic insulin-like growth factor signaling via two endocrine responses—decreased expression of brain derived Drosophila insulin-like peptides (dILPs) and increased expression of Imp-L2, an inhibitor of insulin signaling. As a result of these effects, body growth is reduced and larval development is delayed. These findings suggest that control of ribosome synthesis, and hence protein synthesis, in specific tissues can exert control on overall body growth.

Abstract Introduction Results Discussion Materials and Methods

developmental biology cell biology genetics biology and life sciences molecular cell biology

2014

TIF-IA-Dependent Regulation of Ribosome Synthesis in Drosophila Muscle Is Required to Maintain Systemic Insulin Signaling and Larval Growth

Candida albicans bloodstream infection is increasingly frequent and can result in disseminated candidiasis associated with high mortality rates. To analyze the innate immune response against C. albicans, fungal cells were added to human whole-blood samples. After inoculation, C. albicans started to filament and predominantly associate with neutrophils, whereas only a minority of fungal cells became attached to monocytes. While many parameters of host-pathogen interaction were accessible to direct experimental quantification in the whole-blood infection assay, others were not. To overcome these limitations, we generated a virtual infection model that allowed detailed and quantitative predictions on the dynamics of host-pathogen interaction. Experimental time-resolved data were simulated using a state-based modeling approach combined with the Monte Carlo method of simulated annealing to obtain quantitative predictions on a priori unknown transition rates and to identify the main axis of antifungal immunity. Results clearly demonstrated a predominant role of neutrophils, mediated by phagocytosis and intracellular killing as well as the release of antifungal effector molecules upon activation, resulting in extracellular fungicidal activity. Both mechanisms together account for almost of C. albicans killing, clearly proving that beside being present in larger numbers than other leukocytes, neutrophils functionally dominate the immune response against C. albicans in human blood. A fraction of C. albicans cells escaped phagocytosis and remained extracellular and viable for up to four hours. This immune escape was independent of filamentation and fungal activity and not linked to exhaustion or inactivation of innate immune cells. The occurrence of C. albicans cells being resistant against phagocytosis may account for the high proportion of dissemination in C. albicans bloodstream infection. Taken together, iterative experiment–model–experiment cycles allowed quantitative analyses of the interplay between host and pathogen in a complex environment like human blood. Sepsis is a systemic inflammatory response triggered by infection and a major cause of death worldwide [1]–[3]. In recent years, fungal pathogens have caused an increasing number of sepsis cases with high mortality rates [4], [5]. The major fungal pathogen Candida albicans is a common human commensal but can become invasive in patients with a compromised immune system and disturbance of epithelial barrier integrity or may enter the bloodstream by disseminating from biofilms on medical devices [6]–[8]. Among the different components of human immunity, neutrophils (polymorphonuclear neutrophilic granulocytes, PMN) are crucial for antifungal immune responses and neutropenia is associated with impaired prognosis in systemic candidiasis [9]. PMN possess several mechanisms that may contribute to clearing of C. albicans like phagocytosis, oxidative burst, degranulation and formation of neutrophil extracellular traps (NETs) and have been shown to respond specifically to the invasive filamentous form of C. albicans [10]. Other peripheral blood immune cells have also been implicated in the response against C. albicans, including monocytes as well as NK-cells [11], [12]. Furthermore, C. albicans has been shown to strongly activate complement while at the same time recruiting complement regulators to its surface that may protect it against antimicrobial effector functions [13]–[17]. So far little is known about the interplay of these effects in vivo. Studies using purified human immune cells or experiments performed at a molecular level provide important insights into mechanisms of immune recognition but fail to address in vivo complexity. Murine models are mainly used to address in vivo settings but peripheral blood components differ substantially from their human counterparts with regard to quantity and functional aspects [18]. To overcome some of these limitations, a human whole-blood infection model can be used to monitor host-pathogen interactions. Such models have successfully been used in identifying microbial virulence factors [19], analyzing early immune responses [20], determining the influence of genetic polymorphisms on immune response [21] and testing potential therapeutic approaches or vaccine efficacy [22]–[26]. Whole-blood assays provide time-resolved data on localization and physiological state of the pathogen and immune activation. Whereas many parameters are accessible to direct experimental quantification, others are not due to experimental limitations. However, biomathematical modeling can provide tools to overcome these experimental limitations. Here, we formulate a mathematical infection model for C. albicans in human blood and apply a state-based modeling approach to perform computer simulations that predict details on the dynamics of the immune response. The state-based model corresponds to a non-spatial agent-based model that enables decision making depending on the occurrence of specific events, such as first-time phagocytosis, and allows modeling interactions between individual cells occurring in small numbers in a stochastic fashion [27]. We demonstrate that a priori unknown transition rates between any two states can be estimated by fitting the simulation results to the experimental data using the Monte Carlo method of simulated annealing. Therefore, the state-based model allows detailed predictions on dynamics of host-pathogen interaction in human blood and, in particular, on the main course of the immune response. To analyze early immune responses to a fungal pathogen, C. albicans was added to lepirudin-anticoagulated whole-blood of healthy volunteers at different concentrations. After inoculation of C. albicans yeasts, activation of PMN, monocytes and NK-cells but no unspecific early activation of T- and B-cells could be detected by quantification of the general activation marker CD69 (Fig. 1A). Furthermore, no cell death or decrease in host cell numbers was observed with this inoculum throughout the course of the experiment. No or only slight changes in CD69 expression levels could be observed in response to lower concentrations of C. albicans (Fig. 1A). Fungal concentrations of and more resulted in significant host cell death at later stages of infection. Therefore an inoculum of C. albicans yeasts was used in subsequent experiments. Innate immune activation by C. albicans resulted in significantly elevated plasma levels of pro-inflammatory cytokines () as well as chemokines () (Fig. 1B). As PMN have been shown to play a central role in the defense against C. albicans, we quantified activation of these cells in more detail. Early after inoculation of C. albicans a strong induction of reactive oxygen intermediates in PMN could be observed (Fig. 1C). Surface levels of receptors involved in immune recognition like CD11b and CD64 increased, whereas CD16 markedly decreased on PMN after fungal inoculation indicating cellular activation (Fig. 1C). Up-regulated surface exposure of the degranulation marker CD66b and increased plasma concentrations of myeloperoxidase, lactoferrin and elastase confirmed massive degranulation (Fig. 1D). Consequently, activation of neutrophils also resulted in the accumulation of potentially fungicidal activity in plasma [28], [29]. To analyze the distribution of the fungal pathogen in different compartments of human blood we used a C. albicans strain constitutively expressing GFP. Within of blood infection of fungal cells associated with PMN and this interaction was further increased at () and (). Whereas low association of C. albicans to monocytes (maximum association to monocytes at p. i.) could be observed, no interactions with lymphocytes were detectable (Fig. 2A). A significant proportion of C. albicans cells (at) remained extracellularly throughout the observation period and therefore escaped the cellular immune response by developing resistance against phagocytosis. The inoculation of human blood with C. albicans yeasts/ml resulted in similar fungal association patterns indicating that distribution of C. albicans in blood is largely independent of the fungus to immune cell ratio. To test, whether this distribution pattern was characteristic for C. albicans or rather strain specific, we used a set of ten clinical isolates from bloodstream infections. All strains showed similar distribution patterns with a strongly predominant association to PMN (at p. i. median association to PMN: [range, median association to monocytes: [range. For none of the strains, association to lymphocytes could be detected. Concomitant to interaction with immune cells, changes in C. albicans morphology could be observed in microscopic analyses (Fig. 2B). Intracellular organisms were predominantly found in PMN throughout the experiment and showed different morphotypes, in line with a growth arrest of filaments in PMN after phagocytosis [10]. In contrast, extracellular fungi showed small germ tubes after inoculation and mainly occurred as pseudohyphae at later time points, indicating continuous filamentation of these cells during the experiment (Fig. 2B). Plating assays demonstrated a substantial killing of C. albicans over time with only of fungal cells remaining viable four hours after inoculation (Fig. 2C). To model host-pathogen interaction in C. albicans blood infection we used a state-based model that comprises all experimentally validated C. albicans states in human blood (Fig. 3, for details see Methods section and a flow-diagram of the algorithm in Fig. S1). Alive C. albicans cells () may be extracellularly killed () and both, and may turn into cells that are resistant against phagocytosis and further killing, denoted by and, respectively. Non-resistant extracellular cells may be phagocytosed by monocytes or PMN and internalized viable fungal cells could be killed intracellularly. A proper bookkeeping of these intracellular processes in monocytes () or granulocytes () was ensured by the two indices, which refer to the numbers of internalized C. albicans cells that are alive () and killed (), respectively. Transitions between states occur with specific transition rates that determine the time-dependent simulation of the infection process and are summarized in Fig. 4. Of note, we distinguished the initial phagocytosis by PMN with rate from subsequent phagocytosis events by activated PMN that may occur with a different rate [30]. Furthermore, taking into account that the release of antimicrobial peptides by PMN induces extracellular killing, we used a time-dependent rate for extracellular killing that increases with the number of initial phagocytosis events by PMN. Initially, all immune cells occupied states and and the number of immune cells were set to average physiological numbers in blood: and. The initial number of C. albicans cells corresponded to the inocula used in the experiments and these cells were either in the -state or in the -state, while no resistant cells existed at the initial time point. A priori unknown transition rates were estimated by the method of simulated annealing based on the Metropolis Monte Carlo Scheme. Starting with a randomly chosen parameter set, the algorithm searched in the parameter space of transition rates for the global optimum from a fit to the time-resolved experimental data of the whole-blood infection assays with C. albicans (see Materials and Methods section for details). The mean values of the transition rates could be estimated with standard deviations below, indicating the high accuracy of the fitting procedure (Table 1) and the comparison of simulated and experimental data clearly showed quantitative agreement for the whole time course of infection (Fig. 5). The simulations were repeated 100 times for the normally distributed transition rates (Table 1) and the thickness of the solid lines in Fig. 5 represents the mean standard deviation due to these variations. The limiting value of the standard deviations was below for each quantity and the solid lines remained well within the experimental error bars, indicating that the simulation results are robust against variations in the transition rates. Due to experimental limitations it is impossible to quantify the contribution of single effector mechanisms to the overall elimination of C. albicans in the whole-blood infection model, whereas the virtual infection model allowed separate analyses for all killing mechanisms. In the state-based model, we assumed that the phagocytosis rates were constant in time. This assumption was experimentally justified by reinoculation of C. albicans cells into an infected blood sample after. Since we observed a similar distribution pattern for the newly added C. albicans after as in the initial experimental set-up (Fig. 6), it could be concluded that the phagocytosis rates remain fairly constant over time. According to the model, phagocytosis of C. albicans by a monocyte is less probable than uptake by PMN (). To confirm the different phagocytic capacity of PMN and monocytes we experimentally increased the total monocyte number by adding autologous isolated monocytes to blood samples. Distribution of C. albicans to the different immune cell populations in these samples was quantified after and compared to non-substituted blood samples. Despite an almost equal number of PMN and monocytes in the substituted blood samples (PMN to monocytes ratio:), the majority of C. albicans cells still associated with PMN (), clearly indicating that PMN are more efficient in taking up C. albicans than monocytes (Fig. 7). In addition, the model predicted that internalization by PMN that phagocytose for the first time is lower compared to internalization by PMN which did phagocytose more than one C. albicans cell (). We examined the robustness of the prediction by performing four restricted parameter estimations with conditions (i), (ii), (iii) and (iv). For all those conditions, the fitting errors were significantly larger than the fitting error of free parameter estimation (see Fig. S2A). This was verified by Wilcoxon rank-sum test and the variations in the corresponding parameter sets are depicted in Fig. S2B. Surprisingly, the model predicted that intracellular killing of PMN occurs with a lower transition rate than intracellular killing by monocytes (). To test the robustness of this prediction we repeated the parameter estimation procedure under the biologically motivated condition. We found that the fitting error of this conditional parameter estimation was not significantly different from the free parameter estimation, but is again significantly smaller than that of parameter estimations under conditions (i) – (iv) (see Fig. S2A). The parameter estimation with condition yielded, which was mainly due to a decrease of by more than. This was compensated by relatively small variations in all other rates (see Fig. S2B), indicating that the parameter estimation for the virtual infection model is generally robust in all the other rates. The original parameter estimation revealed that most C. albicans cells were killed within PMN (), were killed extracellularly and a small amount was killed within monocytes (). Consequently, elimination of C. albicans in human blood is mainly mediated by PMN which – apart from being present in higher numbers – release antimicrobial peptides inducing extracellular killing and are therefore more effective in eliminating C. albicans than monocytes. The virtual model allowed us to distinguish between intracellularly and extracellularly killed C. albicans cells inside monocytes and PMN. Both immune cell types bear more intracellularly killed than extracellularly killed C. albicans throughout the first of infection (PMN versus, monocytes versus, see Fig. 5). To analyze the average contribution of single PMN to elimination of C. albicans we determined the distribution of alive and killed C. albicans over PMN. The model predicted PMN to phagocytose up to five viable C. albicans cells, with most of the PMN containing one fungus (see Fig. S3A). The amount of PMN that contain viable C. albicans started to decrease after, whereas the amount of PMN containing killed C. albicans increased and reached a maximum after (see Fig. S3B). We found that PMN contained at maximum six C. albicans cells, however, the majority of cells carried only one. After, the relative amount of PMN that contained one C. albicans cell was predominantly greater than the fraction of PMN that contained more than one C. albicans cell (versus, see Fig. S3C). Similar results were obtained for the distribution of C. albicans in monocytes (Fig. S4). These predictions were experimentally verified by manually counting C. albicans cells per PMN in blood smears with quantitatively comparable results, confirming that most PMN which had phagocytosed contained a single C. albicans cell throughout the experiment (see Fig. 8). Excellent fits were achieved for and after inoculation whereas a higher degree of variation was observed at after inoculation, consistent with a higher standard deviation of the experimentally quantified concentrations around this time point (see Fig. 8). These data indicate that activation of PMN triggered by phagocytosis of C. albicans enhances extracellular killing and results into a series of secondary phagocytosis events. Therefore, the distributions of C. albicans cells in PMN and monocytes deviate from the distributions expected for simple Poisson statistics. A comparison revealed a decrease in the number of monocytes containing Candida cells, whereas the number of PMN containing two or more Candida cells was increased (see Supporting Information Text S1 and Fig. S5). These deviations are a direct result of the relationship. Experimental results had shown that a fraction of C. albicans cells remained extracellular and some fungi also survived throughout the experiment (Fig. 5B). These findings could not be explained by proliferation of C. albicans as budding could not be observed and filamentation does not lead to an increase of cell numbers. Lytic escape from phagocytes, which has been described for C. albicans [31], could be excluded as no cell death occurred throughout the experiment. In the model, this was integrated by allowing extracellular C. albicans cells to become resistant against phagocytosis and further killing (Fig. S1). This was required for fitting the virtual infection model to the experimental data as the fractions of extracellular and viable C. albicans cells were not negligible. Our model predicted that almost all C. albicans cells that remained alive had developed resistance against phagocytosis and further killing () and only few fungi remained alive in PMN () and monocytes (). Resistant fungal cells also constituted the major fraction () of extracellular C. albicans at post infection. Using a non-filamentous mutant (C. albicans efg1, cph1) we could demonstrate that development of resistance was not linked to filamentation as this mutant showed an identical distribution as the wild-type without developing filamentous forms (distribution of C. albicans, at p. i. associated to PMN, associated to monocytes and free, for all). Moreover, inoculation of killed C. albicans cells into human blood proved that killed fungal cells developed resistance against phagocytosis with identical rates as viable fungi resulting in similar amounts (for viable versus for inactivated C. albicans) of extracellular fungi (Fig. 9). The simulation results predicted that the amount of alive resistant C. albicans cells was larger than the relative number of killed resistant C. albicans cells, i. e. versus, respectively, which was in line with the observation that extracellular C. albicans showed continued filamentous growth throughout the experiment. Development of resistance was not linked to exhaustion of the host cells. In contrast, immune cells in the model infection system clearly retained their phagocytic capacity throughout the experiment. This was shown by reinoculation of an infected blood sample after, which resulted in identical uptake kinetics as primary infection (Fig. 6). To further confirm these data we added freshly drawn blood of the same donor to an infected blood sample to test whether the new immune cells were able to take up all or part of the extracellular resistant C. albicans population. As expected, no additional uptake of C. albicans cells could be observed. Taken together, the simulation results revealed that development of resistance against phagocytosis and further killing is the only way for C. albicans cells to survive immune activation in human blood. We applied a state-based modelling approach to simulate the host-pathogen interaction for C. albicans in human blood. This approach allowed to set up a virtual infection model that captures the stochastic transitions between systems states, e. g. including all possible configurations of alive and killed C. albicans cells in monocytes and PMN as well as in the extracellular space. In contrast to deterministic models based on differential equations, the bookkeeping of discrete transitions in the state-based model enabled us to accurately model (i) the killing by secreted antimicrobial factors due to the primary phagocytosis of C. albicans cells by PMN and (ii) the dynamic distribution of killed and alive C. albicans in immune cells. This is a consequence of the fact that non-spatial agent-based models represent interactions between cells occurring in small numbers as stochastic events and allow for decision making depending on the preceding occurrence of specific events [27]. A priori unknown transition rates between any two states could be estimated by fitting the simulation results to the experimental data using the Monte Carlo method of simulated annealing. This procedure enabled us to quantify transition rates with high accuracy by identifying the set of parameters that globally minimizes the least-square error between the results of the simulation and the experiment. The current model has been fitted to results obtained with blood samples from several independent blood donors. Furthermore, we have shown that overall distribution rates are highly similar for a set of unrelated clinical bloodstream isolates. Despite this, it has to be noted that our data will most likely underestimate the biological variability of both host and pathogen as a small set of selected donors and C. albicans strains does not cover the complete biological variability of both populations. However, our approach offers an unique option to study this diversity, e. g. by using C. albicans strains that have been shown to differ in their interaction with host immunity [32]. In addition, the ability to use the whole-blood infection assay rather than purified primary immune cell populations bears several other advantages: (i) as no isolation procedure is involved all cells in the assay are completely untouched and should show minimal pre-activation [33], (ii) the whole-blood model allows communication between different effector cells and contains a functional complement system [12], [17], [34], (iii) the whole-blood model enables pharmacological intervention by blocking several arms of innate immune activation [35], [36]. Consequently, several future applications of our approach can be envisioned. These include the comparative analysis of different pathogens, investigation of clinically relevant scenarios (neutropenia) as well as studies on the influence of genetic markers on innate immune activation. The virtual infection model clearly predicts a predominant role of neutrophils in the early immune response mounted in human blood against C. albicans. Although neutrophils have mostly been considered as central in the defense against invasive C. albicans infection, their role in the clinical setting is not unambiguous. In patients with chronic granulomatous disease, a congenital disorder of NADPH oxidase which prevents oxidative burst and formation of NETs, candidemia is surprisingly rare, especially when compared to invasive mould infections like aspergillosis or zygomycosis [37], [38]. In line with this, many studies have failed to identify neutropenia as an independent risk factor for candidemia [39]. As these studies have largely been performed in ICU settings, this may however be due to the rarity of neutropenic patients in these cohorts. In cancer patients, neutropenia has been found to contribute to the risk for developing candidemia [40]–[42] and it is generally accepted that the outcome of candidemia is impaired in neutropenic patients and therefore current therapeutic guidelines recommend intensified treatment protocols for candidemia in neutropenic patients [43]. Our results suggest that neutrophils are of central importance in the immediate response against invading C. albicans and contribute to elimination in two ways. First, they effectively take up viable C. albicans cells and kill them intracellularly. This activity of neutrophils has generally been considered a major route of antifungal activity and was studied in detail using purified neutrophils [10], [44]–[46]. Second, neutrophils release antifungal effector molecules upon activation that result in extracellular killing of C. albicans. Our model predicts that both mechanisms together account for as much as of fungal killing. This clearly underlines the outstanding importance of neutrophils in mounting a protective response against invasive C. albicans infection which has been suggested by experimental in vivo studies [47]. Bloodstream infection with C. albicans frequently results in organ dissemination, which can affect many organs and anatomical sites including liver, eye, joints and even brain. In an early study, 9 of 32 patients with candidemia showed chorioretinitis compatible with Candida infection and routine performance of fundoscopy is advised for patients suffering from candidemia within one week of treatment initiation [43], [48]. Other studies also documented high rates of dissemination in candidemia, resulting in a disease entitity termed acute disseminated candidiasis [49], [50]. Interestingly, profound and prolonged neutropenia can result in a different disease entity known as chronic disseminated candidemia which is defined by a hematogenous infection of liver and spleen by Candida spp. [51]. Our virtual infection model suggests that elimination of C. albicans will be severely hampered in neutropenic blood, which could explain increased levels of dissemination in the respective patients. The ability of C. albicans to disseminate is linked to its ability to interact with endothelial cells in a way that allows invasion of tissue [52], [53]. However, to establish disseminated infection in multiple organs, it is a prerequisite that some C. albicans cells remain viable in the blood for a prolonged time period. Here, we provide clear evidence that this is indeed the case. Furthermore, of several hypotheses that could potentially explain long-term survival of C. albicans in human blood, the model clearly predicts the development of resistance against phagocytosis among an extracellular population of fungal cells to be the most favourable explanation. The molecular basis for development of resistance will have to be addressed in future studies. However, experimental testing of model-generated hypotheses has provided some important clues: (i) development of resistance against phagocytosis does not require viability of the fungus. In contrast, thimerosal-killed yeast cells were able to acquire resistance at identical rates as viable fungi. This also clearly proves that (ii) development of resistance is not linked to filamentation of C. albicans. In line with this finding, a non-filamentous, mutant of C. albicans was also able to acquire resistance at the same rate as C. albicans wild-type. (iii) Finally, the resistance phenotype does not seem to be linked to exhaustion of phagocytes at later stages of infection. This could be shown by reinoculation after two hours of initial infection, which again resulted in unimpaired phagocytosis and killing of the newly inoculated yeast cells. A range of host factors has previously been shown to bind to the fungal cell wall and some Candida proteins may even recruit several host factors at a time [16], [54]. Shielding of the fungal cell wall by host factors may be the basis for developing resistance against phagocytosis and/or killing of C. albicans as observed in our model. Although so far no study has addressed the recruitment of host factors from complex and physiological environments, the established whole-blood infection model in combination with flow-cytometry assisted cell sorting offers a unique opportunity to pursue this hypothesis in future experiments. Moreover, interpreting the experimental results in the light of the virtual infection model will enable quantitative analyses of the dynamic immune response and the relative importance of defence mechanims by iterative cycles between experiment and theoretical modeling.

Candida albicans is the most important fungal pathogen in nosocomial bloodstream infections. So far little is known about the interplay of different cellular and non-cellular immune mechanisms mediating the protective response against C. albicans in blood. The in vivo scenario of C. albicans infection can be mimicked by human whole-blood infection assays to analyze the innate immune response against this pathogen. These experiments reveal the time-evolution of certain mechanisms while leaving the values of other quantities in the dark. To shed light on quantities that are not experimentally accessible, we exploited the descriptive and predictive power of mathematical models to estimate these parameters. The combination of experiment and theory enabled us to identify and quantify the main course of the immune response against C. albicans in human blood. We quantified the central role of neutrophils in the defence against this fungal pathogen, both directly by phagocytosis and indirectly by secreting antimicrobial factors inducing extracellular killing. Other findings include the distribution of C. albicans cells in neutrophils and monocytes as well as the immune escape of C. albicans cells in the course of infection.

Abstract Introduction Results Discussion

medicine immune cells immune activation immunology host-pathogen interaction microbiology fungi immune defense fungal diseases infectious diseases mycology biology infectious disease modeling pathogenesis immune response yeast computer science computer modeling immunity innate immunity

2014

A Virtual Infection Model Quantifies Innate Effector Mechanisms and Candida albicans Immune Escape in Human Blood

The Type VI secretion system (T6SS) mediates toxin delivery into both eukaryotic and prokaryotic cells. It is composed of a cytoplasmic structure resembling the tail of contractile bacteriophages anchored to the cell envelope through a membrane complex composed of the TssL and TssM inner membrane proteins and of the TssJ outer membrane lipoprotein. The C-terminal domain of TssM is required for its interaction with TssJ, and for the function of the T6SS. In Citrobacter rodentium, the tssM1 gene does not encode the C-terminal domain. However, the stop codon is preceded by a run of 11 consecutive adenosines. In this study, we demonstrate that this poly-A tract is a transcriptional slippery site that induces the incorporation of additional adenosines, leading to frameshifting, and hence the production of two TssM1 variants, including a full-length canonical protein. We show that both forms of TssM1, and the ratio between these two forms, are required for the function of the T6SS in C. rodentium. Finally, we demonstrate that the tssM gene associated with the Yersinia pseudotuberculosis T6SS-3 gene cluster is also subjected to transcriptional frameshifting. The Type VI secretion system (T6SS) is a macromolecular machine widespread in proteobacteria that delivers protein toxins into either eukaryotic or bacterial cells [1]–[5]. The Vibrio cholerae T6SS has been shown to inject an effector domain carrying actin cross-linking activity into eukaryotic cells, preventing cytoskeleton rearrangements and allowing the bacteria to escape phagocytosis [6]–[8]. More recently, a number of T6SSs including those of Pseudomonas aeruginosa, V. cholerae, Serratia marcescens, enteroaggregative Escherichia coli and Citrobacter rodentium have been shown to play antagonistic roles in interbacterial competition including competition occuring during host colonization [5], [9]–[13]. Bacterial preys are killed through the actions of toxins that bear peptidoglycan hydrolase, phospholipase or DNase activities [13]–[15]. For toxin delivery, the T6SS is thought to use a dynamic mechanism resembling that of contractile tailed bacteriophages [3], [4], [16]–[18]. Recent cryo-electron microscopy experiments demonstrated that the T6SS is composed of a cytoplasmic tubular structure anchored to the cell envelope by a membrane complex [18]. The tubular structure is structurally and mechanistically similar to the tail of bacteriophages: the Hcp protein forms hexameric rings that stack on each other to assemble a tube resembling the internal tube of phages and tipped by a trimer of VgrG, which shares a fold similar to the trimeric bacteriophage gp27-gp5 hub – or cell-puncturing – complex [16], [19]–[21]. This internal tube is wrapped into a structure composed of the TssB and TssC subunits [18], [22]. This structure has been shown to be dynamic, as TssB proteins fused to the super-folder Green Fluorescent Protein (sfGFP) form long filaments that cycle between extended and contracted conformations, a mechanism reminiscent of bacteriophage sheaths [18], [23]–[25]. The current model proposes that the mode of action of the T6SS is comparable to that of a crossbow [2]–[4]: the sheath assembles around the Hcp internal tube into an extended conformation. Upon contraction, the internal tube will be propelled towards the target cell allowing the VgrG protein to puncture the host cell and effector delivery. Indeed, recent studies have shown that contraction of the T6SS sheath-like structure coincides with killing of the target bacterial prey [23]–[25]. This “phage-related” complex is anchored to the cell envelope through interactions with membrane components. This membrane complex is composed of the TssL and TssM inner membrane proteins and the TssJ outer membrane lipoprotein [3], [4], [26], [27]. TssM is constituted of three trans-membrane helices with a large C-terminal domain of ∼750 residues protruding into the periplasm [28], [29]. TssM is a central component as it interacts with both TssL and TssJ [28], [29]. The interaction with TssJ has been characterized and involves contacts between a specific loop of the lipoprotein and the 150 last residues of TssM [29]. This interaction is critical for T6SS function as disruption of the TssM-TssJ interaction abolishes Hcp release in the culture supernatant [29]. Although this C-terminal region of TssM is an essential determinant of T6SS function, the TssM protein encoded within the CTS1 T6SS gene cluster of C. rodentium, TssM1, is lacking this domain. However, this T6SS is functional as shown by its ability to release the Hcp1 protein in the culture supernatant and to mediate interbacterial killing [30]. Sequence analysis of the tssM1 gene showed that the stop codon is preceded by a poly-adenosine sequence constituted of eleven consecutive adenosine residues [30], [31]. Poly-A runs have been previously shown to be slippery sites for the RNA polymerase that cause frameshifting by the incorporation of additional adenosine bases into the mRNA during transcription [32]–[35]. Here, using a combination of Western-blot, GFP fluorescence and mass spectrometry analyses, we demonstrate that transcriptional frameshifting occurs at the tssM1 poly-A run, yielding two TssM1 size variants. We further demonstrate using Hcp secretion and antibacterial competition assays that both forms of TssM1 are required for efficient Type VI secretion in C. rodentium. The frequency of frameshifting is ∼20–25% and therefore yields a molecular ratio of 3–4∶1 between the truncated and the full-length variants. This ratio between the two forms is critical as inverting this ratio leads to a non-functional T6SS apparatus. Finally, we show that a similar frameshiting mechanism occurs in the tssM gene associated with the Yersinia pseudotuberculosis T6SS-3 gene cluster. Analysis of the Citrobacter rodentium tssM1 gene sequence (accession numbers: ROD_27701, Gene ID: 8713035) showed that the full-length gene is disrupted by the existence of a premature amber stop codon at position 2,421 (from the start codon) [31]. DNA sequencing of a cloned fragment encompassing this region showed that this amber codon is not a sequencing error. Premature arrest of tssM1 translation leads to the production of an 807-amino-acid (aa) TssM1 protein (TssM1[1–807]) (Fig. 1A and 1B). Transmembrane helix predictions and sequence alignment of TssM1[1–807] with T6SS TssM proteins of known topology showed that TssM1[1–807] is constituted of the three N-terminal helices but lacks the C-terminal β-domain which has been previously shown to mediate the interaction with TssJ. Interestingly, a sequence corresponding to a β-domain similar to those of canonical TssM proteins is encoded on the +1 reading frame downstream the stop codon of tssM1[1–807]. Hence, a full-length 1129-aa TssM protein (TssM1-FL) will be produced if +1 frameshifting occurs before the stop codon of tssM1[1–807] (Fig. 1A and 1B). To test whether frameshifting occurs, we monitored TssM1 production by immunodetection. The 3829-bp sequence of tssM1-FL was cloned in an expression plasmid downstream the tet promoter. The cloning strategy included the insertion of (i) a FLAG epitope-encoding sequence immediately downstream the start codon and (ii) a 6×His-encoding sequence upstream the stop codon of tssM1-FL. Western-blot analyses of cell extracts of C. rodentium carrying this expression plasmid demonstrated the production of two proteins of ∼85 and ∼130 kDa, immunostained by the anti-FLAG antibody (Fig. 1C). The molecular weights of these two bands are similar to the expected sizes of TssM1[1–807] (88 kDa) and TssM1-FL (125 kDa). This result suggests that frameshifting occurs in the tssM1[1–807] sequence, yielding a full-length TssM1 protein. The production of TssM1-FL was confirmed as the ∼130-kDa protein was detected by anti-5His immunostaining (Fig. 1C). Quantitative analyses showed that the intensity of the high molecular weight protein is ∼1/3 of that of the low molecular weight protein, suggesting that +1 frameshifting occurs with a frequency of ∼25%. Sequence analysis showed that a stretch of 11 adenosines is localized 28 nucleotides upstream the tssM1[1–807] stop codon (Fig. 2A). It has been previously showed that poly-adenosine tracts might induce ribosome or RNA polymerase slippage [32]–[35]. To test whether this poly-A tract might be involved in the production of TssM1-FL, we used site-directed mutagenesis to engineer (i) a tssM1 variant in which the three AAA codons were substituted by three AAG codons (a construct hereof called tssM1-AAG) hence disrupting the poly-A tract without modifying the amino-acid sequence and (ii) a tssM1 variant with a disrupted poly-A and carrying a deletion of the last A to create an artificial +1 frameshift (called tssM1-AAGΔA) (Fig. 2A). Western-blot analyses showed that disruption of the poly-A tract prevents frameshifting as only the low molecular weight protein corresponding to TssM1[1–807] was detected by the anti-FLAG antibody. By contrast, only the TssM1-FL variant was detected with both anti-FLAG and anti-5His antibodies when the artificial frameshift construct was analyzed (Fig. 2B). To verify that frameshifting occurs, we used an alternate method by engineering translational fusion of TssM1 to the GFP. The GFP-encoding sequence was inserted 48-pb downstream the premature stop codon of TssM1[1-807] (TssM1-WT, Fig. 2C). This construct serves as negative control for fluorescence (Fig. 2D), as no GFP fusion can be produced (because of the in-frame stop codon). Additional GFP fusions were engineered. In the TssM1+1 construct, an additional nucleotide was inserted between the premature stop codon and the GFP-encoding sequence (Fig. 2C). This construct is a reporter of +1 frameshifting in the natural situation, as the fusion protein will be produced only if +1 frameshifting occurs. Indeed, a significant level of fluorescence compared to the WT sequence can be observed with this construct, demonstrating that frameshifting occurs (Fig. 2D). This frameshifting is dependent on the poly-A run as poly-A disruption by AAA to AAG substitutions in the TssM1+1 construct (TssM1+1/AAG, Fig. 2C) decreases the fluorescence to TssM1-WT levels. Finally, a deletion of the 11th A nucleotide in the TssM1+1/AAG construct yields TssM1+1/AAGΔA (Fig. 2C). In this fusion, no frameshifting can occur but the frame is restored by the additional nucleotide placed after the initial stop codon. Hence, all the produced TssM1 proteins are fused to the GFP and the fluorescence levels reflect the fluorescence if +1 frameshifting occurred with an efficiency of 100%. Comparison of the TssM1+1 and TssM1+1/AAGΔA fluorescence levels showed that the +1 frameshift frequency is ∼20% (Fig. 2D), a value comparable to the frequency calculated from protein immunodetection (Fig. 1C). Taken together, the results from the Western-blot analyses and of the GFP fusions demonstrate that a frameshifting mechanism involving slippage onto a poly-A tract restores the reading-frame of the tssM1 gene of C. rodentium to produce two TssM1 variants: TssM1[1–807] and TssM1-FL. The molecular mechanisms that yield frameshifting have been well described. Translational frameshift could occur during translation of mRNA by ribosomes at specific adenosine repeat, but this mechanism usually requires additional determinants within the mRNA such as Shine-Dalgarno-like sequences or specific mRNA secondary structures close to the stretch of adenosines [33], [35], [36]. However, none of these signals are found at proximity to the tssM1 frameshifting site. Long stretches of homopolymeric sequences are better known to induce transcriptional slippage, i. e. , realignment of the growing RNA to its DNA template within the RNA polymerase. The incorporation of extra, nontemplated, A nucleotide (s) by the RNA polymerase during elongation results in the synthesis of a heterogeneous population of mRNA with different molecular masses [32], [37]. To test whether tssM1 frameshifting involves ribosome or RNA polymerase slippage, the molecular masses of tssM1 mRNA products were measured by electrospray ionization mass spectrometry (ESI/MS) as previously reported [37]. Total RNAs were collected from WT C. rodentium cells upon activation of the CTS1 T6SS gene cluster using the recombinant strain carrying inducible promoters [30] and tssM1 cDNA were synthesized and used as template for a PCR reaction. ESI/MS analyses of the PCR products show that products with masses corresponding to molecules bearing 10 to 15 As in the poly-A run were detected (Fig. 3A). The additional adenosines were not incorporated during reverse transcription and PCR amplification as only 11-A PCR products were observed by ESI/MS when a synthetic mRNA corresponding to the region of tssM1 mRNA subjected to reverse transcription was used as initial template (Fig. 3B). Hence, the presence of the stretch of adenosines induces reiterative transcription by the RNA polymerase and this transcriptional frameshifting restores the tssM1 reading-frame. At a functional level, the most frequent use of frameshifting is to allow the synthesis of a product additional to that of standard decoding. The products can have distinct functions and the ratio between the different products might be important. In other cases frameshifting serves a regulatory function [35]. To address the physiological relevance of the tssM1 frameshifting for T6SS function, a strain deleted of the tssM1-FL gene was constructed. As we did not find conditions in which the CTS1 T6SS gene cluster is expressed, we used in this study recombinant strains in which the expression of the cluster is under the control of inducible promoters, as previously described [30]. CTS1 T6SS function was tested by monitoring Hcp1 release in the culture medium and CTS1-mediated interbacterial killing [30]. As shown previously, the CTS1 was non functional in absence of the tssM1 gene as shown by the absence of Hcp1 in culture supernatant and by the inability of CTS1 to confer a growth advantage to C. rodentium in co-culture with E. coli on solid medium (Fig. 4A; [30]). These phenotypes were complemented by the trans-expression of tssM1, which produces both TssM1[1–807] and TssM1-FL. However, when TssM1[1–807] or TssM1-FL (from tssM1-AAG or tssM1-AAGΔA respectively) were produced alone, the CTS1 T6SS was not functional: Hcp1 was not released and CTS1-mediated interbacterial killing was abolished (Fig. 4A). These data indicate that both variants of TssM1 are necessary for T6SS function. To further validate these results, we introduced the AAG and AAGΔA substitutions on the chromosome to yield strains producing only one of the TssM1 variants. As shown in Fig. 4B, the CTS1 T6SS was not functional in these two strains; however, when the second variant was expressed in trans, Hcp release and CTS1-mediated interbacterial competition were restored to levels comparable to the WT strain. Taken together, these results demonstrate that C. rodentium CTS1 T6SS function requires both forms of TssM1, TssM1[1–807] and TssM1-FL, the latter being produced by transcriptional frameshifting. As described above, Western-blot and fluorescence quantifications established that the TssM1[1–807]: TssM1-FL ratio was close to 3–4∶1 (Figs. 1C, 2B and 2C). As both TssM1 variants were required for T6SS function, we asked whether the relative ratio between these two variants is critical. Each form was therefore independently cloned on compatible plasmids: pBAD18 (pBR322 origin) and pASK-IBA37+ (pUC origin). We first verified that producing the natural TssM1 variants in these two vectors (i. e. , retaining the natural ratio between the two forms) did not impact the function of the CTS1 T6SS. Production of the two forms from pBAD18 (Fig. 4A) or from pASK-IBA37+ (Fig. 5A) complemented the T6SS-dependent Hcp1 secretion defect of ΔtssM1 cells. As shown previously for the pBAD18 derivatives, Hcp1 was not released when each length variant was independently produced from pASK-IBA37+ (Fig. 5A). Fig. 5B shows that when TssM1[1–807] was produced at higher levels than TssM1-FL (i. e. , the natural situation; Fig. 5B, lanes 7 and 8), Hcp1 was released in the culture supernatant. However, when the ratio was inversed (TssM1-FL produced at higher levels compared to TssM1[1–807]), Hcp1 release was abolished (Fig. 5B, lanes 3 and 4). These data therefore demonstrate that the CTS1 T6SS is functional only when the truncated form of TssM1 is produced in higher amount than the full-length TssM1 variant. We wondered whether transcriptional frameshifting is a common character among tssM genes. T6SS-associated tssM gene nucleotide sequences collected from the National Center for Biotechnology and Information (NCBI) were used to identify (i) tssM genes with abnormal length and/or (ii) tssM genes bearing A or T homopolymeric runs. Interestingly, we found that the tssM gene encoded within the Yersinia pseudotuberculosis T6SS-3 gene cluster, tssM3 (accession number: YpsIP31758_1373; gene ID 5385222), has a stretch of 9 As at position 2,394 relative to the start codon (Fig. 6A). The tssM3 gene encodes a 130-kDa protein. The tssM3 gene was cloned downstream a FLAG epitope-coding sequence. Western-blot analyses of cell extracts of Y. pseudotuberculosis cells producing FLAG-TssM3 revealed a band at ∼80 kDa in addition to the full-length 130-kDa protein (Fig. 6B). This band results from frameshifting as (i) a stop codon is present downstream the poly-A tract in the +1 reading frame (Fig. 6A) and (ii) disruption of the poly-A tract by AAG substitutions of the AAA codons abolished synthesis of the ∼80-kDa protein (Fig. 6B). These data were confirmed by fluorescence levels of GFP fusion proteins: the putative slippery site was active as fusion of the GFP-encoding sequence in the +1 reading frame downstream the poly-A tract (TssM3+1) led to GFP fluorescence (Fig. 6C and 6D). However, although the slippage mechanisms between the C. rodentium tssM1 and Y. pseudotuberculosis tssM3 genes are probably similar and allow the synthesis of two variants of different lengths, recoding in tssM1 leads to the synthesis of the longer variant, whereas recoding in tssM3 leads to the synthesis of the shorter variant. It is also important to note that the quantification of the anti-FLAG western-blots and the comparison between the fluorescence levels of the TssM3-WT and TssM3+1 GFP fusion constructs showed that the frequency of frameshifting is 15–25%, demonstrating that, in contrast to C. rodentium TssM1, the full-length protein is produced at higher levels compared to the truncated form. In this work, we showed that the sequence of one essential gene of the C. rodentium CTS1 T6SS is disrupted by an early stop codon yielding a 88-kDa truncated protein, TssM1[1–807], that lacks a large part of the C-terminal domain required for interaction with other components of the secretion machine; however, we demonstrated that the full-length 125-kDa TssM1 protein is produced during growth. We further demonstrated that transcriptional frameshifting occurs at a slippery site constituted of 11 consecutive adenosine residues, located a few bases upstream the premature stop codon, that induces RNA polymerase infidelity and realignment. This mechanism, although unusual, is not unprecedented. Several examples of RNA editing have been described in viruses, eukaryotes and prokaryotes [33]–[35], [38]–[41]. Frameshifting is particularly frequent in bacteriophages and bacterial insertion sequence (IS) elements [33]–[35]. Well-studied cases are the phage G gene which encodes two tail proteins, gpG and gpGT, gpGT arising from translational frameshifting [42], [43] and the dnaX gene, which encodes the τ and γ subunits of DNA polymerase III [37], [44]. One striking example is the fusion between pgk and tim, two different genes that can be fused by transcriptional frameshifting at the 3′ end of pgk, yielding a bifunctional chimera protein [38]. The overrepresentation of slippery sites in viruses and bacterial endosymbiots of insects, which have the smallest genomes, suggests that this mechanism helps to condense protein coding in compact genomes [35], [40]. Interestingly, examples of transcriptional frameshifting have been identified in other bacterial secretion systems such as the Shigella flexneri Type III secretion system (T3SS), a machinery that mediates entry of the bacterium into epithelial cells. Slippery sites that induce RNA polymerase infidelity have been identified and characterized in the mxiE gene that encodes a transcriptional activator of this system as well as in three genes encoding structural components of the T3SS, mxiA, spa13 and spa33 [45], [46]. The efficiency of the tssM1 frameshifting was shown to be ∼20-25% leading to a molecular ratio of 3–4∶1 of TssM1[1–807] to TssM1-FL. A similar frequency was measured for Y. pseudotuberculosis tssM3. These frequencies are comparable to those measured for transcriptional slippage of the Shigella flexneri mxiA (15%), mxiE (20–30%), and spa33 (15%) genes and lower than those measured in the case of spa13 (55%) [45], [46]. In this study, we have measured the slipppage efficiencies during growth in rich medium (LB). It would be interesting to test whether the slippage frequency is impacted by the growth conditions or by regulatory elements, such as bacteriophage λ N protein, recently shown to influence transcriptional realignment by stabilizing the RNA/DNA hybrid in the RNA polymerase [47]. Using complementation assays, we further showed that both forms of TssM1 are required for T6SS function. Although further experiments are required to better understand what is the specific function of each of these two variants, this situation is reminiscent to that of the phage lambda G gene, in which both gpG and gpGT variants are required for efficient assembly of functional tails [48]. In this later case, it was shown that the ratio between gpG and gpGT is also important for formation of phage tails [48]. Similarly, we observed that the ratio between the two TssM1 variants is critical for maintaining a functional CTS1 T6SS. In the natural situation, the shorter variant (TssM1[1–807]) is 3-4 times more abundant than the full-length variant. Inversion of the ratio between the two forms abolishes the function of the CTS1 T6SS. One additional intriguing result is the observation that a third TssM1 variant of ∼40 kDa, truncated of the N-terminal region, is immunodetected by the C-terminal 6×His epitope (see * in Fig. 2B). This variant therefore corresponds to the C-terminal portion of the TssM1 80-kDa periplasmic domain that is likely retained into the cytoplasm. This variant might be produced from an internal start codon (although sequence analyses did not identify a potential ribosome binding site or an internal start codon) or might result from a proteolytic processing. Experiments are currently carried out to determine how this third variant is produced and to define whether it is necessary for proper assembly or function of the CTS1 T6SS. Bioinformatic analyses of the T6SS-associated tssM genes showed that transcriptional slippery sites are not common as we only identified the tssM3 gene from Yersinia pseudotuberculosis with a poly-A run. Western-blot and fluorescence studies further demonstrated that this site is active as two TssM3 length variants are produced. Slippage occurs with a frequency comparable to that of the C. rodentium tssM1 situation. Although we have not tested whether these two variants are required for function of the apparatus, it is worthy to note that transcriptional frameshifting in Y. pseudotuberculosis tssM3 leads to synthesis of a shorter protein whereas C. rodentium tssM1 slippage leads to synthesis of the full-length protein. As a consequence, and in contrast to C. rodentium tssM1, the ratio is in favor of the full-length variant. This is particularly intriguing as our data showed that the ratio between the two variants in C. rodentium is critical for the function of the apparatus, and it further suggests that the ratio between the two variants is tailored to fit specific needs during assembly and/or function of the T6SS in different bacteria. In the vast majority of T6SS-associated tssM genes, no slippery site can be identified, suggesting that only the full-length protein is produced. However, stable TssM degradation products of ∼85-kDa have been observed by Western-blot analyses of total extracts of WT cells producing TssM in Agrobacterium tumefaciens [28] and in enteroaggregative E. coli. In these cases, two forms of TssM are therefore produced, the shorter being the result of a degradation mechanism. This observation is particularly fascinating and further experiments are required to understand whether this degradation is a controlled process, whether it is conserved in all TssM proteins, and whether the degradation product is important for the function of the T6SS machines in these bacteria. Strains used in this study are listed in S1 Table. Citrobacter rodentium strains used in this study are derivatives of DBS100 [ATCC51459] (kindly provided by Hervé LeMoual (McGill University, Montreal, Canada) ): RLC2 (a spontaneous nalidixic acid resistant variant), RLC55 (a RLC2 derivative in which the promoter as been swapped with a divergent plac-para promoter) and RLC62 (a ΔtssM1 derivative of RLC55) [30]. Yersinia pseudotuberculosis IP31758 [49] has been kindly provided by Anne Derbise and Elisabeth Carniel (Institut Pasteur, Paris). Escherichia coli DH5α, CC118λpir have been used for cloning procedures. E. coli W3110 has been used for growth competition assays. MFDpir [50] has been used for mating assays with C. rodentium. Strains were routinely grown in Luria-Bertani (LB) broth, or on LB agar plates supplemented with kanamycin (50 µg/mL), chloramphenicol (30 µg/mL) or nalidixic acid (20 µg/mL) when required. MFDpir cells were grown in medium supplemented with 0. 3 mM diaminopimelic acid. Induction of the CTS1 T6SS gene cluster expression was performed with IPTG and L-arabinose, as previously described [30]. PCR fragments used for strain and plasmid constructions were amplified with the Phusion high fidelity DNA polymerase (Thermo scientific). Colony PCR amplifications were performed with Taq DNA Polymerase with Standard Taq Buffer (New England BioLabs). Site-directed Quickchange mutagenesis amplifications were done using the Pfu Turbo DNA polymerase (Agilent Technologies). Restriction enzymes were purchased from New England BioLabs. Custom oligonucleotides (listed in S2 Table) were synthesized by Sigma Aldrich. The chromosomal poly-A sequence of tssM1 was modified with the AAG and AAGΔA mutation by allelic exchange using the sacB-counter selectable suicide plasmid pSR47S. A tssM1 fragments of ∼600-bp surrounding the poly-A tract were amplified with oligo pair icmF-crod1-fwd-C/icmF-crod-rev from mutated plasmids (see plasmid construction below) and cloned by TA cloning into pCR2. 1 (Invitrogen). After DNA sequencing, BamHI/XbaI fragments from pCR2. 1 were inserted into BamHI/SpeI-digested pSR47S, yielding plasmids pRL132 and pRL133. pSR47S derivatives were introduced into C. rodentium RLC55 by conjugation using MFDpir as donor and the first recombination event was selected on kanamycin LB plates as previously described [51]. Insertion of pSR47S derivatives was verified by colony PCR. Sucrose counterselection was obtained as described [51] and insertion of the accurate mutation was verified by PCR on purified chromosomal DNA (DNeasy Blood & Tissue kit, Qiagen) and DNA sequencing. Total RNAs were extracted from exponentially-growing C. rodentium cells using the RNeasy mini kit (Qiagen). RNA preparations were treated with TURBO DNase (Ambion) to avoid DNA contamination prior to Reverse Transcription (RT) -PCR. The absence of contaminating DNA in the Total RNA preparation was verified by PCR. tssM1-specific cDNA encompassing the poly-A run was synthesized from 500 ng of total RNA using oligonucleotide EC955 and the SuperScript II Reverse Transcriptase (Invitrogen). 56-nt PCR products were then amplified from 200 ng of cDNA using primers EC1266 and EC1267 and Phusion DNA polymerase, extracted using the ethanol precipitation procedure and analyzed by electrospray ion mass spectrometry as described previously [37]. As controls for reverse transcription and PCR, PCR products were generated (i) from C. rodentium genomic DNA and (ii) from a 56-nt synthetic RNA (GCCGGCUAUUAUGAGGCGUUUAAAAAAAAAAAUGGGUCCGGGGCUGAUGCUGUUAG) (Eurogentec). The Hcp1 release assay has been performed as previously described [30]. The antibacterial growth competition assay has been performed as previously described, using the E. coli K-12 strain W3110 bearing the pUA66-rrnB plasmid (KanR and strong and constitutive GFP fluorescence) as prey [30]. Briefly, Citrobacter and E. coli cells were mixed to a 4∶1 ratio and the mixture was spotted onto prewarmed dry plates and incubated for 16 hours at 30°C. Fluorescent images were taken using a LI-COR Odyssey imager and the relative fluorescence was measured after resuspension of the bacterial cells using a TECAN Infinite M200 microplate reader. C. rodentium cells carrying the pUA66-rrnB plasmid derivatives were grown in LB at 37°C to an OD600 of ∼1 and normalized to an OD600 of 0. 5. Triplicates of 150 µl were transferred into wells of a black 96-well plate (Greiner) and the absorbance at 600 nm and fluorescence (excitation: 485 nm; emission: 530 nm) were measured with a TECAN infinite M200 microplate reader. The experiments were done in triplicate and the relative fluorescence was expressed as the intensity of fluorescence divided by the absorbance at 600 nm, after subtracting the values of a blank sample. Gene expression from pBAD18 and pASK-IBA37 (+) derivatives was induced in exponentially growing cultures (OD600∼0. 5) using arabinose (0. 2%) for 1 hour and AHT (5 or 10 ng/ml) for 30 minutes respectively. For Western-blot analyses, cells were resuspended in Laemmli buffer (2×1011 cells/ml). Proteins were separated by SDS-PAGE analyses and transferred onto nitrocellulose membranes. Immunoblots were probed with anti-5His (Qiagen) or anti-FLAG (Sigma) antibodies, and anti-mouse secondary antibodies coupled to fluorophores. Immunodetection and band density analyses were performed using a LI-COR Odyssey imager.

Nonstandard decoding mechanisms lead to the synthesis of different protein variants from a single DNA sequence. These mechanisms are particularly important when the genome length has to be limited such as viral genomes, limited by the available space in the capsid, or to synthesize two different polypeptides that have distinct functional properties. Here, we report that tssM, a gene encoded within the Citrobacter rodentium Type VI secretion (T6S) gene cluster, is interrupted by a premature stop codon; however, the stop codon is preceded by a slippery site constituted by 11 consecutive adenosines. Reiterative transcription leads to the incorporation of additional nucleotides in the mRNA and therefore restores the original framing. As a consequence, two different TssM variants are created by transcriptional frameshifting, including a full-length 130-kDa protein and an 88-kDa truncated variant. We further show that both forms, and the ratio between these two forms, are required for the function of the transport apparatus. Interestingly, a similar mechanism regulates the synthesis of two TssM variants in Yersinia pseudotuberculosis.

Abstract Introduction Results Discussion Materials and Methods

bacteriology gram negative bacteria organismal evolution medical microbiology microbial evolution gene identification and analysis gene expression genetics gene regulation biology and life sciences molecular genetics microbiology microbial pathogens evolutionary biology microbial control bacterial pathogens dna transcription gene function

2014

Transcriptional Frameshifting Rescues Citrobacter rodentium Type VI Secretion by the Production of Two Length Variants from the Prematurely Interrupted tssM Gene

Neglected zoonotic diseases (NZDs) have a significant impact on the livelihoods of the world’s poorest populations, which often lack access to basic services. Water, sanitation and hygiene (WASH) programmes are included among the key strategies for achieving the World Health Organization’s 2020 Roadmap for Implementation for control of Neglected Tropical Diseases (NTDs). There exists a lack of knowledge regarding the effect of animals on the effectiveness of WASH measures. This review looked to identify how animal presence in the household influences the effectiveness of water, hygiene and sanitation measures for zoonotic disease control in low and middle income countries; to identify gaps of knowledge regarding this topic based on the amount and type of studies looking at this particular interaction. Studies from three databases (Medline, Web of Science and Global Health) were screened through various stages. Selected articles were required to show burden of one or more zoonotic diseases, an animal component and a WASH component. Selected articles were analysed. A narrative synthesis was chosen for the review. Only two studies out of 7588 met the inclusion criteria. The studies exemplified how direct or indirect contact between animals and humans within the household can influence the effectiveness of WASH interventions. The analysis also shows the challenges faced by the scientific community to isolate and depict this particular interaction. The dearth of studies examining animal-WASH interactions is explained by the difficulties associated with studying environmental interventions and the lack of collaboration between the WASH and Veterinary Public Health research communities. Further tailored research under a holistic One Health approach will be required in order to meet the goals set in the NTDs Roadmap and the 2030 Agenda for Sustainable Development. Neglected tropical diseases (NTDs) are a group of communicable diseases estimated to affect over a billion people globally, particularly those with least economic resources, access to health care, good nutrition, clean water and sanitation facilities; the weak political influence of affected groups as well as the complex nature of these diseases has resulted historically in a lack of attention and resources, precipitating the use of the term “neglected”[1]. This has been acknowledged by the World Health Organisation (WHO) and a global Roadmap was released in 2012 to focus on reducing the burden of 17 NTDs. This “Roadmap for Implementation” [2] includes five ‘key strategies to combat NTDs by 2020’ of which one aims to improve veterinary public health at the human–animal interface, and another emphasises the provision of safe and clean sources of water and effective sanitation infrastructure, and ensuring appropriate hygiene practices (WASH) [3]. The Roadmap, together with the 2015 WHO global strategy on WASH and NTDs [4], espouses a holistic approach to disease control and elimination. The new global development framework enshrined in the Global Goals of the United Nations’ 2030 Agenda for Sustainable Development [5] sets out a One-Health approach to poverty, inequalities, health and the environment, in contrast with the siloed structure of the previous Millennium Development Goals (MDGs), whose agenda ended in 2015. Global Goal 3 within this agenda sets ambitious targets for improving health and wellbeing, including NTDs, and acknowledges the importance of addressing social and environmental determinants of health [6]. A One Health approach that addresses the animal-human interface and defines disease control strategies that enhance livelihoods and reduce poverty can contribute to the achievement of the Global Goals, but also represents a departure from current prevailing practices. Further knowledge on effective programming approaches is therefore urgently needed. Several of the NTDs are zoonotic diseases—infections transmitted between animals and humans, and are therefore referred to as Neglected Zoonotic Diseases (NZDs). These include cysticercosis, rabies, echinococcosis, foodborne trematodiases, zoonotic African trypanosomiasis and schistosomiasis. Several of these are related to WASH elements in terms of prevention and/or treatment. Other diseases recognised by WHO in its “Research Priorities for Zoonoses and Marginalized Infections” include toxoplasmosis, cryptosporidiosis and bacterial zoonoses, for which improved sanitation has proven effective in reducing transmission [3]. The global burden of these zoonotic diseases is considerable. Cystic echinococcosis causes, on average, the loss of 2 million annual disability-adjusted life years (DALYs), with associated costs rising up to US$ 3 billion for human treatment and livestock industry losses [7]. Taenia solium, the causal agent of taeniasis and cysticercosis, is responsible for an estimated cost of 2. 8 million DALYs globally [8]. Mortality due to cysticercosis in humans increased by 58% between 1990 and 2010 [9], and the disease is estimated to affect over 50 million people globally, causing up to 30% of all epilepsy cases [10]. Zoonoses are estimated to contribute to up to 10% of the total DALYs lost, and 26% of DALYs lost due to infectious diseases in low income countries [11]. Zoonoses affect human health directly, but by affecting animal health, they can also cause important economic losses and limitations for affected rural communities that depend on animals for working fields, transportation, as a source of protein and as a source of income when sold in local markets [12]. For example, cysticercosis has been reported to cause $12,6 million in annual losses in Cameroon [13], $150 million in India [14] and 18. 6 to 34. 2 million US dollars in East Cape, South Africa [15]. These zoonotic diseases are neglected due to the relatively low mortality associated with them, their tendency to affect predominantly poor and marginalised populations, and the complex, intersectoral measures required to control them, which include community infrastructure and capacity building, health promotion programmes, improved diagnostics and treatment, vaccination and prevention programmes and policy adaptation at local, regional, national and international level [11]. Zoonotic pathogens have complex life cycles that commonly include different phases in human hosts, animal hosts and the environment before completion. Overlooking one or more of these three elements facilitates the perpetuation of the cycle, and with it, reinfection. A One Health approach to controlling zoonotic transmission is needed, considering animals, people and the environment in a comprehensive approach to public health. Since zoonoses are influenced directly and indirectly by multiple factors, focusing solely on transmission routes wrongfully overlooks socio-cultural, economic, anthropological and ecological elements that may affect transmission as well as delivery of control programmes. The need for intersectoral control measures is especially evident in low income countries [16], where the rural population accounts for an average of 69% of the total [17]. Not only do poor, rural communities have fewer resources and less access to healthcare, they also possess less political influence and power than other population groups to demand services and resources from government authorities [18–20]. A One Health approach helps create resilient solutions for disease transmission by setting measures that can be implemented in the long term by community and government action, meeting the objectives for sustainability set by the Sustainable Development Goals [21]. In poor, rural settings, smallholder animal production of indigenous species of pigs, poultry and ruminants is dominant [22], and hence human and animal interaction within the household is more common in these settings, requiring special attention to this interaction in the control of zoonotic diseases [23]. However, given the dependence of rural households on animals as a major source of livelihood and as an alternate source of income in emergencies, certain measures that may support disease control objectives may not be feasible in practice [24]. For example, pig-corralling is recommended as a main method for control of cysticercosis, and hence programmes may be put in place to improve this practice amongst farmers [25]. However, for many households and communities in middle-low income countries, this is not economically feasible [26], since this would require the family to assume the added cost of feeding the pigs, instead of allowing the animals to forage for themselves [27]. Similarly, protecting water sources from animal access prevents contamination of water for human use with animal faeces and secretions. However, the need to provide livestock and humans with sufficient clean water from a protected source poses a challenge for many communities [28]. A One Health approach can help identify such multi-factorial elements and avoid omitting valuable programme components, including human, environmental and animal factors. Human behaviour factors such as conflict, migration and socio-cultural practices, shape disease patterns, due to relocation, high human density and reduced hygiene levels [29]. Similarly, economic and agricultural development will reshape the land and demands of society, changing animal farming and animal product consumption practices, increasing the risk of food-borne disease transmission and zoonotic influenza [30]. An example of an animal factor to consider is how wildlife reservoirs can help perpetuate infective cycles within local livestock. This poses a great challenge for zoonotic disease control in pastoral communities due to the difficulty of limiting direct and indirect interaction between wildlife and livestock species [30,31]. Additionally, ecological factors like climate change and deforestation have a direct impact on the distribution of vector-borne diseases by altering the habitats of the vector and reservoir species, as well as allowing vectors to sustain their life cycle in new areas due to a rise in average temperatures, leading to emergence and re-emergence of these diseases in new parts of the world [30,32]. Another example of One Health approaches helping to tackle ecological problems can be found in the reuse of animal excreta as crop manure, as incorrect use can lead directly to disease transmission through contact and clothes and indirectly through water contamination [33]. Use of animal excreta as crop manure can also alter the chemical properties of the soil, endangering the environmental sustainability of the area, and subsequently increasing the exposure of humans and animals to contaminated sources of infection [33]. Authors like Nguyen-Viet, Zinsstag and Charron propose an integration method as a solution for optimising the use of human and animal excreta as manure, by combining cross-sectoral knowledge and stakeholder engagement under a One Health framework [33,34]. Such a framework enables the implementation of sustainable control strategies for NZDs in countries where economic resources are scarce. Water, sanitation and hygiene (WASH) programmes can plausibly contribute to control of zoonotic disease given the knowledge about pathogen transmission cycles, through provision of sanitation infrastructure that safely removes human and animal faecal waste from the human environment, provision of clean water sources, and improvement of hygiene practices at the community and household level [4]. The WHO WASH and NTDs strategy is a step towards developing collaboration between WASH and NTDs programmes, both of which reference integration of control measures, but do not offer specific guidance or methods of monitoring on collaboration between the sectors [4]. However, the much needed guidance to encourage a One Health approach through engagement of other sectors such as agriculture and veterinary public health is not included in the remit of the WASH and NTDs strategy [5,35]. The positive relationship between WASH programmes and reduction of NTDs incidence has been proven, yet many of these programmes still lack the multifactorial approach needed to cover the impact of other elements that affect disease transmission [36], such as animal presence within the household and human-animal interaction. Because of this, there are limitations to understanding why WASH programmes may not result in the expected disease control outcomes and how they can be optimized. No systematic research has been done to date on the impact of demand-side sanitation programmes on NZDs transmission [3]. Although the evidence base on the interaction of animals with sub-standard sanitation facilities is weak, it is plausible that the presence of free-roaming household animals alongside conditions of open defecation or poor containment of faeces can contribute to intensified disease transmission [37]. As mentioned in the WHO WASH and NTDs Strategy [4], and as several authors argue [36,38–40], it is necessary to gather more information regarding WASH-related interventions and disease burden reduction. This is particularly relevant for zoonotic diseases, as, out of the existing reviews relating to WASH and disease burden, few focus specifically on zoonotic diseases. Those that do, often disregard the presence of animals in the household and its impact on the effect of WASH interventions on zoonotic disease. There is need to identify these linkages and knowledge gaps that require further study. The aim of this work was to conduct a systematic review to identify the existing published data, on how the presence of animals in the household impacts the efficacy of WASH interventions for zoonotic disease control. The objectives of this review were: to identify how animal presence in the household influences the effectiveness of water, hygiene and sanitation measures for zoonotic disease control in low and middle income countries; to identify gaps of knowledge regarding this topic based on the amount and type of studies looking at this particular interaction. A review protocol was designed to inform and direct the review steps before conducting the systematic review. The protocol was designed based on the guidelines given by “CRD’s guidance for undertaking reviews in health care” and the “WHO Handbook for Guideline Development” [41,42], as well as example systematic review protocols found in various academic sources, approved by peer academic experts. The complete protocol can be found in Text S1. Three databases were used: Medline, Web of Science and Global Health. These were chosen based on other systematic reviews conducted in the area of sanitation, hygiene and NTDs [43–45], and on expert academic advice solicited by the authors. The three databases were systematically searched for publications dating 1980 to 30th April 2016. The search terms relative to WASH were chosen based on other WASH literature reviews and scientific articles. Animal terms were selected based on literature and expert advice, including those species most likely to interact with humans within the household, in low- and middle-income countries. The terms were then divided into four pools: The terms amongst pools were combined by the Boolean operator “OR”, while those between pools were combined by the Boolean operator “AND”. Diseases chosen for the terms were based on the list of neglected zoonotic diseases described in the WHO NTDs Roadmap [2]. The results obtained were sorted by “author” in descending order. Studies were selected through a three-stage process, first by title and abstract screening, then by full text analysis, based on the selection criteria for each stage, and finally by a quality control checklist. References were managed with the use of reference management software EndNote X7. For the first stage, title and abstract screening, studies were included if the abstract mentioned a zoonotic disease term together with a WASH term, if a full text version was available and if the article was published in English or Spanish. Studies not meeting these requirements, and review articles, were excluded. The full text versions of studies selected in this first stage were retrieved and analysed for further selection. In this second stage, articles that did not quantify burden of disease in human or animal populations, did not analyse the role of animals in zoonosis transmission in relation to WASH measures, or did not meet the requirements of the quality check described in the protocol, were excluded from the review. The type of study and its design were not deemed to be crucial inclusion/exclusion criteria, due to a low number expectancy of final study retrieval. Studies selected for the last stage of the systematic review were analysed using a quality checklist based on the guidelines for public health studies from the National Institute for Health and Clinical Excellence [49]. Articles included in the full text review were subjected to data extraction based on the protocol, with special attention to the study population regarding burden of disease, the diagnostic method used, the WASH measures in place, description of animal presence within the household, and the statistical analysis approach taken by the study. Due to the consideration of various types of studies in the inclusion criteria and the expected low count of final studies making the last selection, pooling was not deemed possible. Therefore, a narrative approach was chosen for addressing data synthesis. Zoonotic diseases in which WASH measures play a relevant role in control were included in the analysis and synthesis of the results, as long as the selected study included it in its own analysis, even if said diseases were not considered to be neglected by inclusion in the WHO reference list. Seven thousand five hundred and eighty-eight (n = 7588) studies where obtained after introducing the search terms into the three databases (Fig 1). Screening of titles and abstracts retrieved a total of 80 studies (n = 80) meeting the inclusion criteria for the first stage of the review: 46 from Medline, 28 from Global Health, and six from Web of Science. Of these 80,13 were duplicates and three were unable to be retrieve in full-text form and were therefore discarded. The total number of articles selected for the next stage of the review was 64. Full text for the remaining 64 articles was obtained, analysed and considered for review inclusion. After data extraction and analysis, two articles [50,51] were identified that quantified the burden of disease in humans or animals and analysed the role of animals in zoonosis transmission in relation to WASH measures, hence meeting the final inclusion criteria as set out in the protocol. Due to the low count of studies included in the final review, the 64 articles analysed in this phase were summarised in the form of tables that show the research tendencies when addressing WASH and NZDs. The complete list with the main data extracted from each one can be found in Table 1, including location, type of study, number of participants in the study, disease of interest, diagnostic test used to address presence of disease, WASH and animal component studied, the type of statistical method used for the analysis, and a summary of the results of the study. More than half of the studies (29) focused on cysticercosis, while 12 focused on toxoplasmosis (Table 2). Humans appear as the most studied species, with 36 studies looking at human burden of disease, while pigs were second with 26 citations. Fifty one out of 64 were designed as cross-sectional studies, 46 of these establishing a prevalence value through a serological test and combining it with a questionnaire for associated risk factors. Table 3 shows the study count for each of the categories for water, hygiene and sanitation components, and the proportion of studies that included one, two, or the three types is shown in Fig 2. Three studies had at least one factor in each of the categories. The summarised data suggests the existence of a relationship between NZD epidemiology and the contact of humans and animals in the household, generally showing a negative impact of animal presence on WASH measures or an enhanced negative effect of animal presence on the impact of poor WASH conditions. In the case of cysticercosis, studies show contradictory results regarding the impact of WASH measures and animal presence on disease prevalence. Due to the small number of studies that were selected based on the criteria, the outcome of the quality control check was not considered for further exclusion. The study by Holt et al. (2016) was designed as a cross-sectional study examining prevalence of hepatitis E virus (HEV), Japanese encephalitis virus and Trichinella spiralis in both humans and pigs, as well as Taenia spp. solely in humans in two provinces of Lao PDR, with a multiple correspondence analysis and a hierarchical clustering of several components deemed relevant to disease transmission. Three clusters were identified: one referential (cluster 1) with the best sanitation and lowest pig contact; cluster 2, with moderate sanitation levels and slaughtering of pigs as the main source of animal contact; and cluster 3, with lower sanitation levels and a relative higher rate of free-roaming pigs. The risk of human infection, measured through Odds Ratio (OR), for each of the diseases and clusters when compared to cluster 1 are shown in Table 4. HEV had a very similar OR for risk of infection between clusters 2 and 3, despite the superior WASH conditions of cluster 2. For Taenia spp. and Cysticercosis, risk of infection proved higher in cluster 3 than cluster 2, but with a significant increased risk of infection in cluster 2 compared to the control, despite solid practices of hand washing and water boiling amongst the population. Finally, Japanese encephalitis showed an increased risk of infection in cluster 2 over cluster 3, despite better WASH conditions. Data regarding pig seropositivity was not clustered and WASH factors were not found to be significant in T. spiralis and HEV infection. The other study (Bulaya et al. 2015) was a comparative study pre- and post- community-led total sanitation (CLTS) intervention for porcine cysticercosis control, identifying prevalence performing an Ag-ELISA test. There was no randomization in village selection or house selection, and instead selected based on village characteristics and willingness to participate, respectively. The prevalence pre-intervention was 13. 5%, (6. 8–20. 1,95% C. I.), compared to a value of 16. 4% (12–20. 8,95% C. I.) post-intervention, although this increase was deemed non-significant by the author. After the intervention, latrine presence improved from 67. 2% to 83. 1%, with the percentage of free-roaming pigs changing from an 89. 8% to a 30. 3% of them free roaming, 43. 8% partially free roaming and 25. 8% penned. Home slaughter of pigs increased from 49. 15% baseline to 80. 90% post-intervention. Despite the improvement in latrine presence, animal husbandry was not improved enough to avoid direct and indirect contact between animals and humans within the household. This review showed examples of the way animal-human interaction can affect the effectiveness of WASH interventions for zoonosis control. Importantly, it also highlighted the dearth of studies looking specifically at this interaction. After the search retrieved 7588 articles for this review, 64 were selected in the first screening, of which only 2 were selected for the final review after the second screening. This outcome is likely due to the sectoral focus of the studies. Traditionally, research groups investigating the effectiveness of WASH interventions focus on human factors as positive or negative influences. Similarly, the Veterinary Public Health community focuses more on animal-related factors and disease-transmission routes. The interaction between these two aspects is a research and programming ‘blind spot’, as was demonstrated by this review, and needs to be addressed with further intersectoral research studies. As noted by Zinsstag in 2015 [33], a study in Vietnam showed how a One Health approach for WASH programmes integrates all factors into one framework. This helps identify the relationship between the factors, while exposing the missing links and the areas in need for further research, of which the main one stated is “the boundaries of the sanitation problem”. Sanitation and hygiene programmes have proven effective in reducing NTD burden in numerous studies, as backed by various systematic reviews [43–45]. However, effective, full-coverage implementation of control programmes considering both human and animal sanitation aspects can be challenging in practice. As described by Guilman et al. in 2012 [26], some communities may not have sufficient resources to change their animal farming system to one that limits animal-human contact. In other cases, the community may actually benefit economically from this new farming system [114], but as long as the population believes this is not the case, no change will be embraced by the community [115]. This reinforces the importance of accompanying these type of logistic measures with strong education and hygiene promotion campaigns that involve the community and show the importance and benefits of adopting them. The study by Holt et al. [51] compared Odds Ratio of infection in several pig zoonoses between different sanitation and pig contact factors. For HEV, lower levels of sanitation, as described in the results section, proved to be a risk factor for virus presence, without significant differences between these lower levels specifically. However, increased contact with pigs, particularly through handling and slaughtering, proved significant in its influence on the effectiveness of WASH measures in disease control, as the cluster with moderate sanitation and close pig contact had equal risk of infection as the cluster with poorer sanitation. Pig contact has been described as a risk factor for HEV transmission previously [116], but according to this study, pig corralling impede their access to the household would not make a significant difference in disease transmission as long as the animals are still being slaughtered at home, due to direct human contact with pig blood. In the case of Trichinella, socioeconomic status acted as a confounder, since the main risk factor is pork consumption [117,118], which in this study was associated with higher status due to availability and affordability cost, as are good sanitation and hygiene conditions. In the case of JEV, the cluster with higher direct contact with pigs showed a higher risk of infection, despite better sanitation and hygiene conditions, showing an example of how animal contact can severely hinder the effectiveness of WASH measures. This could be due to its vector-borne nature, which correlates to two factors of this particular cluster: unprotected water sources, which facilitates breeding areas for Culex spp. ; hygiene practices, latrine use or corralling measures would not make a significant impact in its transmission unless done optimally, avoiding contamination of water that could facilitate Culex spp. reproduction. Regarding Taenia solium and cysticercosis, the cluster with higher rates of free-roaming pigs and open defecation showed the highest risk of infection, as expected. However, the high risk of infection presented by the cluster with moderate WASH and close contact with pigs shows how the latter can affect the effectiveness of the former. During the selection process of this review, several studies (Table 1) were screened and later revisited, for further insights on the impact of animals on WASH interventions. Some showed presence, usage or condition of latrines and free roaming of pigs to be significant risk factors in disease transmission [84,119,120], but others had non-significant results [107], rather identifying the source of water for consumption and its quality as a risk factor. In contrast, Nkouawa et al. in 2015 [87] identified that despite having a non-potable (unsafe) water source, disease transmission was reduced by improving hygienic practices and corralling pigs. The study by Holt et al. [51] provided robust results on relative impact of animal and WASH factors, meeting the criteria for selection stated in the protocol of the review. However, future studies should ideally be designed in a way that focuses on isolating the influence of animal factors on the effectiveness of WASH measures. This is particularly difficult to achieve given the circumstances of the communities in which these studies need to be conducted: as noted by Schmidt et al. in 2014 [121], designing impact studies on water, sanitation and hygiene and retrieving significant results is a recurrent challenge for the scientific community: Randomised controlled trials are rarely free from bias, while observational studies usually lack a large enough study population or result significance [121]. Additionally, performing randomised controlled trials in the optimal representative geographical areas is logistically and economically challenging. Another factor to take into account is time, since marketing and promotion campaigns can take several years to have a significant effect, deeming any study that withholds investment in WASH services for such an extended period of time unethical [121]. A relevant limiting factor to assess the efficiency of any WASH programme implementation is the correct use, design and upkeep of sanitation facilities. Several studies show that although latrines were present in the community, they were not consistently used for defecation by all household members or kept in a sufficiently hygienic state [84,85]. The incorrect use of latrines is often associated with socio-cultural and psychological factors, as identified by Thys in 2015 [122], such as a sense of reduced privacy, latrines being too close to the village, comfort of use or trust in its efficacy and need of use. Lack of ownership of the need for latrine construction and lack of ongoing support for maintenance and improvement can undermine potential health benefits of basic latrines. The study by Bulaya et al. in 2015 [50], showed that despite the CLTS intervention resulting in increased latrine presence, net increase in latrine usage and improved pig husbandry, prevalence of disease in pigs increased slightly after the intervention. The study did not specify whether the newly built latrines resulted in safe separation of humans and animals from human faeces. Achieving that level of detail in the analysis is an objective for future studies. Although deemed non-significant, the 95% C. I. shows almost no change in prevalence from pre to post intervention. This was attributed by the authors to infected members of the community still practising open defecation due to lack of resources for latrine construction. Not corralling the totality of the pig population, therefore allowing for interaction of animals and humans within the household, could be the explanation as to why the increase in latrine presence had no effect in decreasing porcine cysticercosis. Free roaming of pigs has been identified as a risk factor for porcine cysticercosis by some of the studies screened before review inclusion [69,75] but was found to be non-significantly others [72]. Similarly, the presence of latrines can be significant [72,73] or non-significant [69] for disease prevalence in pigs, depending on the study, reinforcing the findings by Bulaya et al. (2015). As previously mentioned, low latrine usage has been described as a risk factor for disease transmission [59,84,85] but also as a recurrent sociocultural problem, since many members of the community do not use latrines on a consistent basis for a variety of reasons [59,115,122], or do not keep the latrines in a suitable condition for them to effectively reduce disease transmission [84,115,120]. However, poor programme design, lack of follow up or disputes between NGOs and community leaders on logistics, provisions and payments can be a cause for poor latrine construction and maintenance [123]. This reinforces the suggestion made by Bulaya et al. [50] of the importance of continued hygiene promotion programmes and access to sanitation hardware options in order to ensure the complete effectiveness of sanitation or animal husbandry improvement programmes. As an example of a multifactorial approach to disease transmission control, prevalence of Schistosomiasis was significantly reduced in three studies in China [70,102,124] by implementing a complete WASH programme with sanitation facilities and hygiene educational programmes, reducing the indirect contact of animals and humans through water and reducing the population of the host snail species for Schistosoma. However, programmes that alter animal husbandry in drastic ways such as changing free-roaming farming systems into stabling farming systems, also alter the local economy of the community [125]. In the case of cysticercosis, the penning of pigs is not always possible in certain communities given the resulting increased costs of feed and infrastructure [125]. Substantial investment and economic compensation to farmers and households would therefore be required to maintain and sustain these programmes consistently over time [126]. In the case of toxoplasmosis, principal and consistent risk factors for infection identified throughout the literature, include unsafe water source, inadequate hygienic conditions of the household and cat presence in the household or the vicinity, and were common to human [52,66] or animal [55,58] infection. While providing clean water sources and creating appropriate hygienic conditions decreases the burden of disease, avoiding the presence of cats within the household could potentially increase the presence of rodents in many communities that use cats as the sole method of rodent control. A study showed how, when combined, the presence of cats and dogs in an area significantly reduced the local rodent population [127], however, more research should be conducted to clarify the impact of cat population control on rodent-transmitted diseases in rural communities. The review protocol was designed to include animal-focused studies as well as human-focused studies to ensure a One Health approach to zoonotic disease transmission. Particularly for NZDs, interrupting sustained transmission requires a multifactorial approach considering both zoonotic and anthroponotic transmission paths. Reducing animal burden of disease has a direct effect on human prevalence of disease and vice versa [128], and therefore WASH programmes applied equally to human and animal populations are likely to provide better results than a human-centred approach. The review identified the lack of studies looking at the importance of animal influence in WASH programmes, exposing the existent lack of knowledge in the matter. Further research and programme design need to focus further on animal impact and isolating the study of animal components in the efficiency of WASH control programmes. One of the limitations of the review was the non-inclusion of rodent species in the study. Although rodents are acknowledged to be a source of NZD transmission within the household, they were deemed to overreach the scope and feasibility of this review: on one hand because the review focused in farmed animals kept by the household owners; on the other hand because thorough control of rodent activity in the household is difficult and less reliable than that of farmed animals, mainly due to the complex biological and ecological characteristics of each local rodent species [129,130]. The initial literature review was conducted for fulfilment of an MSc with one student. All three co-authors advised on the approach to be taken and made revisions to the literature. Throughout the writing of the literature there was input from all authors who also held regular review meetings. To further optimise the systematic review, a second reviewer would have performed the search and selection and compared results. Also, had a longer period of time been available, more databases could have been screened, although the final count of studies would most likely be low, since the tendency identified in the review is that of a very low percentage of studies looking specifically at animal influence in WASH measures efficacy. The time constraints were due to the timelines of the MSc. However, all authors had additional input to the manuscript. Whilst the initial literature review was conducted by one student, the manuscript has been prepared after revisions by all authors with additional literature added after further reviews. This has been rewritten to reflect the input following the initial MSc project. This systematic review demonstrated the relevance of human-animal interaction within the household for the effectiveness of WASH measures for control of NZDs. It also shows the significant lack of specific studies tending to the effect of animals on WASH programmes’ effectiveness for zoonotic disease control. Several examples exist in the literature describing prevalence of zoonotic disease and associated risk factors, yet, in the majority of cases, their design fails to assess the specific influence of animal presence in WASH interventions. Further research should be undertaken regarding the influence of animals in WASH programmes, ideally isolating the sanitation component and studying different levels of animal interaction and exposure within the household. Attention to animal burden together with human burden of disease would allow for better understanding and optimisation of WASH programme effectiveness on both disease control and broader development objectives. There exists an evident lack of direct coordination between WHO’s WASH and NTDs official programmes. Further developing of a research agenda around the animal-sanitation-disease link can help set out clear actions on which disease control programmes can be based.

Neglected Tropical Diseases (NTDs) affect the health and economies of populations globally. Many of these diseases are zoonotic, occurring as a consequence of the interaction between humans and animals, particularly at the household level in low- and middle-income countries. Based on the WHO Global Strategy to accelerate and sustain progress on NTDs, including zoonoses, through improvement in sanitation, hygiene and water, this review identifies existing published studies examining the interaction between water, sanitation and hygiene elements, animals and zoonosis transmission within the household. Only two out of 7588 studies screened met the criteria. They showed the relevance of animal influence in the effectiveness of WASH measures, as well as the difficulties of designing studies that look at this particular interaction. A synthesis of several studies analysed in the second selection stage of the review shows a significant relationship between animal and WASH factors for disease transmission. It also shows certain contradictions regarding the importance of key risk factors for some diseases across studies. It is therefore crucial to carry out further studies showing the interaction between animals and water, hygiene and sanitation measures within the household to improve these control measures and reduce zoonotic neglected tropical disease transmission.

Abstract Introduction Material and methods Results Discussion Conclusions

medicine and health sciences tropical diseases vertebrates parasitic diseases animals mammals health care sanitation neglected tropical diseases infectious disease control public and occupational health infectious diseases swine zoonoses hygiene helminth infections environmental health eukaryota cats cysticercosis biology and life sciences amniotes organisms

2018

Animal influence on water, sanitation and hygiene measures for zoonosis control at the household level: A systematic literature review

In the continuous mode of cell culture, a constant flow carrying fresh media replaces culture fluid, cells, nutrients and secreted metabolites. Here we present a model for continuous cell culture coupling intra-cellular metabolism to extracellular variables describing the state of the bioreactor, taking into account the growth capacity of the cell and the impact of toxic byproduct accumulation. We provide a method to determine the steady states of this system that is tractable for metabolic networks of arbitrary complexity. We demonstrate our approach in a toy model first, and then in a genome-scale metabolic network of the Chinese hamster ovary cell line, obtaining results that are in qualitative agreement with experimental observations. We derive a number of consequences from the model that are independent of parameter values. The ratio between cell density and dilution rate is an ideal control parameter to fix a steady state with desired metabolic properties. This conclusion is robust even in the presence of multi-stability, which is explained in our model by a negative feedback loop due to toxic byproduct accumulation. A complex landscape of steady states emerges from our simulations, including multiple metabolic switches, which also explain why cell-line and media benchmarks carried out in batch culture cannot be extrapolated to perfusion. On the other hand, we predict invariance laws between continuous cell cultures with different parameters. A practical consequence is that the chemostat is an ideal experimental model for large-scale high-density perfusion cultures, where the complex landscape of metabolic transitions is faithfully reproduced. Biotechnological products are obtained by treating cells as little factories that transform substrates into products of interest. There are three major modes of cell culture: batch, fed-batch and continuous. In batch, cells are grown with a fixed initial pool of nutrients until they starve, while in fed-batch the pool of nutrients is re-supplied at discrete time intervals. Cell cultures in the continuous mode are carried out with a constant flow carrying fresh medium replacing culture fluid, cells, unused nutrients and secreted metabolites, usually maintaining a constant culture volume. While at present most biotechnology industrial facilities adopt batch or fed-batch processes, the advantages of continuous processing have been vigorously defended in the literature [1–5], and currently some predict its widespread adoption in the near future [6]. A classical example of continuous cell culture is the chemostat, invented in 1950 independently by Aaron Novick and Leo Szilard [7] (who also coined the term chemostat) and by Jacques Monod [8]. In this system, microorganisms reside inside a vessel of constant volume, while sterile media, containing nutrients essential for cell growth, is delivered at a constant rate. Culture medium containing cells, remanent substrates and products secreted by the cells are removed at the same rate, maintaining a constant culture volume. The main dynamical variable in this system is the dilution rate (D), which is the rate at which culture fluid is replaced divided by the culture volume. In a well-stirred tank any entity (molecule or cell) has a probability per unit time D of leaving the vessel. In industrial settings, higher cell densities are achieved by attaching a cell retention device to the chemostat, but allowing a bleeding rate to remove cell debris [9]. Effectively only a fraction 0 ≤ ϕ ≤ 1 of cells are carried away by the output flow D. This variation of the continuous mode is known as perfusion culture. By definition, a continuous cell culture ideally reaches a steady state when the macroscopic properties of the tank (cell density and metabolite concentrations) attain stationary values. Industrial applications place demands on the steady state, usually: high-cell density, minimum waste byproduct accumulation, and efficient nutrient use. However, identical external conditions (dilution rate, media formulation) may lead to distinct steady states with different metabolic properties (a phenomenon known in the literature as multi-stability or multiplicity of steady states) [10–14]. Therefore, for the industry, it becomes fundamental to know in advance, given the cell of interest and the substrates to be used, which are the possible steady states of the system and how to reach them. Moreover, to satisfy production demands, it may be advantageous to extend the duration of a desired steady state indefinitely [6], implying that their stability properties are also of great interest. Fortunately, in the last few years it has been possible to exploit an increasingly available amount of information about cellular metabolism at the stoichiometric level to build genome-scale metabolic networks [15,16]. These networks have been modeled by different approaches [17,18] but Flux Balance Analysis (FBA) has been particularly successful predicting cell metabolism in the growth phase [19]. FBA starts assuming a quasi-steady state of intra-cellular metabolite concentrations, which is easily translated into a linear system of balance equations to be satisfied by reaction fluxes. This system of equations is under-determined and a biologically motivated metabolic objective, such as biomass synthesis, is usually optimized to determine the complete distribution of fluxes through the solution of a Linear Programming problem [20]. This approach was first used to characterize the metabolism of bacterial growth [21], but later has been applied also to eukaryotic cells [22,23]. Alternatively, given a set of under-determined linear equations, one can estimate the space of feasible solutions of the system and average values of the reaction fluxes [24–26]. To consider the temporal evolution of a culture, FBA may be applied to successive points in time, coupling cell metabolism to the dynamics of extra-cellular concentrations. This is the approach of Dynamic Flux Balance Analysis (DFBA) [27] and has been applied prominently either to the modeling of batch/fed-batch cultures or to transient responses in continuous cultures, being particularly successful in predicting metabolic transitions in E. Coli and yeast [23,27,28]. However, to the best of our knowledge, the steady states of continuous cell cultures have not been investigated before. First, because DBFA for genome-scale metabolic networks may be a computational demanding task, particularly when the interest is to understand long-time behavior. Second, because it assumes knowledge of kinetic parameters describing metabolic exchanges between the cell and culture medium, that are usually unknown in realistic networks. Moreover, although the importance of toxic byproduct accumulation has been appreciated for decades [29,30], its impact on steady states of continuous cultures has been studied mostly in simple metabolic models involving few substrates [31,32], while it has been completely overlooked in DFBA of large metabolic networks. Lactate and ammonia are the most notable examples in this regard and have been widely studied in experiments in batch and continuous cultures [30,33–36]. Our goal in this work is to introduce a detailed characterization of the steady states of cell cultures in continuous mode, considering the impact of toxic byproduct accumulation on the culture, and employing a minimum number of essential kinetic parameters. To achieve this and inspired by the success of DFBA in other settings we couple macroscopic variables of the bioreactor (metabolite concentrations, cell density) to intracellular metabolism. However, we explain how to proceed directly to the determination of steady states, bypassing the necessity of solving the dynamical equations of the problem. This spares us from long simulation times and provides an informative overview of the dynamic landscape of the system. The approach, presented here for a toy model and for a genome-scale metabolic network of CHO-K1, but easily extensible to other systems, supports the idea that multi-stability, i. e. , the coexistence of multiple steady states under identical external conditions, arises as a consequence of toxic byproduct accumulation in the culture. We find and characterize specific transitions, defined by simultaneous changes in the effective cell growth rate and metabolic states of the cell, and find a wide qualitative agreement with experimental results in the literature. Our analysis implies that batch cultures, typically used as benchmarks of cell-lines and culture media, are unable to characterize the landscape of metabolic transitions exhibited by perfusion systems. On the other hand, our results suggest a general scaling law that translates between the steady states of a chemostat and any perfusion system. Therefore, we predict that the chemostat is an ideal experimental model of high-cell density perfusion cultures, enabling a faithful characterization of the performance of a cell-line and media formulation truly valid in perfusion systems. We study an homogeneous population of cells growing inside a well-mixed bioreactor [37], where fresh medium continuously replaces culture fluid (Fig 1). The fundamental dynamical equations describing this system are: d X d t = (μ − ϕ D) X (1) d s i d t = − u i X − (s i − c i) D (2) where X denotes the density of cells in the bioreactor (units: gDW/L), μ the effective cell growth rate (units: 1/hr), ui the specific uptake of metabolite i (units: mmol/gDW/hr), and si the concentration of metabolite i in the culture (units: mM). The external parameters controlling the culture are the medium concentration of metabolite i, ci, the dilution rate, D (units: 1 / time), and the bleeding coefficient, ϕ (unitless), which in perfusion systems characterizes the fraction of cells that escape from the culture through a cell-retention device [9] or a bleeding rate. For convenience of notation, in what follows an underlined symbol like s _ will denote the vector with components {si}. Eq 1 describes the dynamics of the cell density as a balance between cell growth and dilution, while Eq 2 describes the dynamics of metabolite concentrations in the culture as a balance between cell consumption (or excretion if ui < 0) and dilution. One must notice that at variance with the standard formulation of DFBA, the terms involving the dilution rate in the right-hand side of both equations enable the existence of non-trivial steady states (with non-zero cell density) which are impossible in batch. These are the steady states that are relevant for industrial applications adopting the perfusion model and that we study in what follows. Still, we require a functional connection between variables describing the macroscopic state of the tank (X, s _) and the average behavior of cells (u _, μ). We start assuming that metabolites inside the cell attain quasi-steady state concentrations [38], so that fluxes of intra-cellular metabolic reactions balance at each metabolite. If Nik denotes the stoichiometric coefficient of metabolite i in reaction k (Nik > 0 if metabolite i is produced in the reaction, Nik < 0 if it is consumed), and rk is the flux of reaction k, then the metabolic network produces a net output flux of metabolite i at a rate ∑k Nik rk, where Nik = 0 if metabolite i does not participate in reaction k. This output flux must balance the cellular demands for metabolite i. In particular we consider a constant maintenance demand at rate ei which is independent of growth [39,40], as well as the requirements of each metabolite for the synthesis of biomass components. If yi units of metabolite i are needed per unit of biomass produced [41,42], and biomass is synthesized at a rate z, we obtain the following overall balance equation for each metabolite i: ∑ k N i k r k + u i = e i + y i z, ∀ i (3) It is also well known that a cell has a limited enzymatic budget [17]. The synthesis of new enzymes, needed to catalyze many intracellular reactions, consumes limited resources, including amino acids, energy, cytosolic [22,43,44] or membrane space [45] (for enzymes located on membranes), ribosomes [46], all of which can be modeled as generic enzyme costs [17]. We split reversible reaction fluxes into negative and positive parts, r k = r k + - r k -, with r k ± ≥ 0, and quantify the total cost of a flux distribution in the simplest (approximate) linear form [17]: α = ∑ k (α k + r k + + α k - r k -) ≤ C (4) where α k +, α k - are constant flux costs. The limited budget of the cell to support enzymatic reactions is modeled as a constraint α ≤ C, where C is a constant maximum cost. Thermodynamics places additional reversibility constrains on the flux directions of some intra-cellular reactions [47], which can be written as: lb k ≤ r k ≤ ub k, lb k, ub k ∈ { - ∞, 0, ∞ }. (5) Additionally, some uptakes ui are limited by the availability of nutrients in the culture. We distinguish two regimes. If the cell density is low, nutrients will be in excess and uptakes are only bounded by the intrinsic kinetics of cellular transporters. In this case ui ≤ Vi, where Vi is a constant maximum uptake rate determined by molecular details of the transport process. These will be the only kinetic parameters introduced in the model. When the cell density increases and the concentrations of limiting substrates reach very low levels, a new regime appears where cells compete for resources. In this regime the natural condition si ≥ 0 together with the mass balance equation (Eq 2 in steady state) imply that ui ≤ ci D/X. In summary, - L i ≤ u i ≤ min { V i, c i D / X } (6) where Li = 0 for metabolites that cannot be secreted, and Li = ∞ otherwise. Thus, an important approximation in our model is that low concentrations of limiting nutrients are replaced by an exact zero. The ratio D/X in Eq 6 establishes the desired coupling between internal metabolism and external variables of the bioreactor. The appendix contains an alternative derivation of Eq 6. Next, we reason that, although cellular clones in biotechnology are artificially chosen according to various productivity-related criteria [48], the growth rate is typically under an implicit selective pressure. We will consider then that the flux distribution of metabolic reactions inside the cell maximizes the rate of biomass synthesis, z, subject to all the constrains enumerated above. Note that to carry out this optimization it is enough to solve a linear programming problem, for which efficient algorithms are available [49]. This formulation is closely related to Flux Balance Analysis (FBA) [20,21,50,51], but some of the constrains imposed here might be unfamiliar. In particular, Eq 4 has been used before to explain switches between high-yield and high-rate metabolic modes under the name FBA with molecular crowding (FBAwMC) [43,52,53], while the right-hand side of Eq 6 is a novel constraint introduced in this work to model continuous cell cultures. If multiple metabolic flux distributions are consistent with a maximal biomass synthesis rate [54], the one with minimum cost α (cf. Eq 4) is selected [17]. Summarizing, from the complete solution of the linear program we obtain the optimal z, and the metabolic fluxes u _ feeding the synthesis of biomass. Finally, the net growth rate of cells μ (see Eq 1) is essentially determined by the cellular capacity to synthesize biomass (rate z), but it may also be affected by environmental toxicity. In the examples presented below we considered that: μ = z - σ (s _) or μ = z × K (s _), corresponding to two different mechanisms explored in the literature [36,55]. In the first case σ (s _) is easily interpreted as the death rate of the cell, while 0 ≤ 1 - K (s _) ≤ 1 represents a fraction of biomass that must be expended on non-growth related activities, for example, due to increased maintenance demands on account of environmental toxicity (but see also Refs. [56,57] and in particular B. Ben Yahia et al. [37] for a recent review of the subject). Both σ (s _) and K (s _) depend on the concentrations of toxic metabolites in the culture, such as lactate and ammonia. Numerical simulations were carried out in Julia [77]. Linear programs were solved with Gurobi [78]. The CHO-K1 metabolic network [70] was read and setup with all relevant parameters using a script written in Python with the COBRApy package [79–81]. All scripts (which also include parameter values) are freely available in a public Github repository [82]. In this section we present the general procedure to determine the steady states of Eqs 1 and 2 and discuss some general results of our model that are independent of the specificities of the metabolic networks of interest. The first step is to set the time-dependence in Eqs 1 and 2 to zero, d X d t = (μ − ϕ D) X = 0 (12) d s i d t = − u i X − (s i − c i) D = 0 (13) Note that Eq 13 depends on X and D only through the ratio 1/ξ = D/X (known in the literature as cell-specific perfusion rate, or CSPR [83]), such that ξ is the number of cells sustained in the culture per unit of medium supplied per unit time (the units of ξ are cells × time / volume). If we recall that in our cellular model, ui is constrained by a term that also depends on X and D only through ξ (cf. Eq 6), it immediately follows that the values of the uptakes and metabolite concentrations in steady state must be functions of ξ, which we denote by u i * (ξ) and s i * (ξ) respectively. To compute u i * (ξ), solve the linear program of maximizing the biomass synthesis rate (z) subject to Eqs 3–5, but replacing Eq 6 with: - L i ≤ u i ≤ min { V i, c i / ξ }. (14) The resulting optimal value of z will be denoted by z* (ξ). Moreover, once u i * (ξ) is known, the stationary metabolite concentrations in the culture follow from Eq 13: s i * (ξ) = c i - u i * (ξ) ξ (15) Then, given z* (ξ), the effective growth rate in steady state can also be given as a function of ξ, μ* (ξ), by evaluating K or σ using the concentrations s i * (ξ) from Eq 15. Next, Eq 12 implies that the dilution rate at which a steady state occurs must also be a function of ξ, that we denote by D* (ξ). Combining this with the relation ξ = X/D, we obtain the steady state cell density, X* (ξ), as well: D * (ξ) = μ * (ξ) / ϕ, X * (ξ) = ξ μ * (ξ) / ϕ. (16) Note that while Eqs 12 and 13 are satisfied by any D ≥ 0 when X = 0, the value Dmax = D* (0) given by Eq 16 is actually the washout dilution rate, i. e. , the minimum dilution rate that washes the culture of cells. Clearly all steady states with non-zero cell density are required to satisfy D* (ξ) < Dmax. From Eq 16 it is evident that the function μ* (ξ) encodes all the information needed to get the values of X in steady state at different dilution rates and for any value of the bleeding coefficient ϕ. On the other hand, if multiple values of ξ are consistent with the same dilution rate (i. e. if the function D* (ξ) is not one-to-one), the system is multi-stable (i. e. , multiple steady states coexist under identical external conditions). A necessary condition multi-stability is that μ* (ξ) is not monotonously decreasing. Since the biomass production rate z* (ξ) is a non-increasing function of ξ (proved in the Appendix), a change in the monotonicity of μ* (ξ) must be a consequence of toxic byproduct accumulation, modeled through the terms K and σ. A noteworthy consequence of Eq 16 is that a plot displaying the parametric curve (ϕD* (ξ), ϕX* (ξ) ) as a function of ξ is invariant to changes in ϕ. This means that for a given cell line and medium formulation, this curve can be obtained from measurements in a chemostat (which corresponds to ϕ = 1), and the result will also apply to any perfusion system with an arbitrary value of ϕ. Moreover, since s i * (ξ) and u i * (ξ) are independent of ϕ, cellular metabolism in steady states is equivalent in the chemostat and any perfusion system (with an arbitrary value of ϕ), provided that the values of ξ = X/D in both systems match. Finally, we mention that generally a threshold value ξm exists, such that a steady state with ξ = X/D is feasible only if ξ ≤ ξm. When ξ > ξm, some of the constrains in Eqs 3,5 and 6 cannot be met. In degenerate scenarios we could have ξm = ∞ (e. g. , this could be the case if the maintenance demand in Eq 3 is neglected) or ξm ≤ 0 (e. g. , if the medium is so poor that the maintenance demand cannot be met even with a vanishingly small cell density). The parameter ξm arising in this way in our model, coincides with the definition of medium depth given in the literature [84], and it quantifies for a given medium composition the maximum cell density attainable per unit of medium supplied per unit time. We first consider the small metabolic network depicted in Fig 2. In this example, maximization of growth sets the nutrient uptake (u) and respiratory flux (r) at the maximum rates allowed by their respective upper bounds, in Eqs 6 and 7. Employing Eq 8 to determine the waste secretion rate (v) from u, r, we obtain: u = min { V, c / ξ }, r = min { u, r max }, v = r - u. (17) Thus the toy model admits simple analytical expressions giving the rates of metabolic fluxes in steady states as functions of ξ. A minimum nutrient uptake rate um is required to sustain the maintenance demand e. Since most cell types are able to grow under certain conditions without waste secretion, we make the biologically reasonable assumption that um ≤ rmax (which is satisfied by the parameters chosen in Materials and methods). It then follows that um = e/ (NF + NR). There are three critical thresholds in ξ that correspond to important qualitative changes in the culture: ξ m = c / u m, ξ sec = c / r max, ξ 0 = c / V. (18) These transitions can be interpreted in the following way: 1) if ξ ≥ ξm the growth rate is zero because the maintenance demand cannot be met; 2) if ξsec ≤ ξ ≤ ξm, cells grow without secreting waste; 3) if ξ ≤ ξsec, there is waste secretion; 4) for ξ ≥ ξ0 cells are competing for the substrate and growth is limited by nutrient availability (cf. discussion before Eq 6); 5) finally, if ξ ≤ ξ0 there is nutrient excess and cells are growing at the maximum rate allowed by intrinsic kinetic limitations. We emphasize that the threshold ξsec carries a special metabolic significance, because it controls the switch between two qualitatively distinct metabolic modes: if ξ ≤ ξsec, respiration is saturated and the intermediate P overflows in the form of secreted waste, with a lower energy yield; on the other hand, if ξ ≥ ξsec, the cell relies entirely on respiration to generate energy, with a higher yield (cf. Fig 3). The medium carries a concentration c of primary nutrient and zero waste content. Under these assumptions, Eq 15 has the following analytical solution for the steady state values of the metabolite concentrations, s* (ξ), w* (ξ): s * (ξ) = c − min { V ξ, c } (19) w * (ξ) = max { 0, c − s * (ξ) − r max ξ } (20) Note that s* (ξ) is a decreasing function of ξ, while w* (ξ) has at most a single maximum. Eqs 8–10,17,19 and 20 can be used to define μ* (ξ). Then D* (ξ), X* (ξ) are given by Eq 16. Fig 4 shows plots of μ* (ξ), X* (ξ), s* (ξ) and w* (ξ) for this model. Parameter values are given in Materials and Methods. As ξ ranges from ξ = 0 to ξ = ξm, stable and unstable steady states are drawn in continuous and discontinuous line, respectively. The system is stable in two regimes: ξ ≲ 1 × 106 cells · day/mL, with high toxicity, low biomass yield and low cell density; and ξ ≥ ξsec, with no toxicity, high biomass yield and high cell density that decays as ξ increases. The later states rely solely on respiration for energy generation (Fig 3a), while the former states exhibit overflow metabolism (Fig 3b). Waste concentration initially increases with ξ until a maximum value is reached. Then w decays during the unstable phase, all the way to zero at ξsec, where waste secretion stops and the system becomes stable again. Intuitively, unstable states become stressed due to high levels of toxicity, which also makes the system very sensitive to perturbations. The typical behavior of nutrients and waste products (in particular glucose and lactate, respectively) in continuous cell cultures, as observed in experiments [84], is that as ξ increases, nutrient concentration in the culture decreases while waste initially accumulates [84] but eventually phases out as cells switch towards higher-yield metabolic pathways [10,11]. This behavior is qualitatively reproduced by s and w in our toy model. The function μ* (ξ) is not monotonically decreasing. As explained above, this implies a coexistence of multiple steady states under identical external conditions. This is readily apparent in a bifurcation diagram of the steady values of X versus D, as shown in Fig 5a. In a range of dilution rates (0. 25 ≲ D ≲ 0. 7, units: day−1), the system exhibits three steady states, one of which is unstable (discontinuous line in the figure), while the other two are stable (continuous line in the figure). Thus a stable state of high-cell density coexists with another of low cell density, over a range of dilution rates. Cellular metabolism in the former state is respiratory (Fig 3a), whereas cells in the later state exhibit an overflow metabolism (Fig 3b). The unstable state is an intermediate transition state lying between these two extremes. Multi-stability of continuous cultures has been repeatedly observed in experiments [10–14]. In our model it is a direct consequence of toxicity induced by the accumulation of waste [87]. Small variations in the dilution rate near D ≈ 0. 25 day−1 or D ≈ 0. 7 day−1 result in discontinuous transitions where the cell density rises or drops abruptly, respectively. These jumps can be traced around an hysteresis loop, drawn in orange arrows in Fig 5a. More generally one may also expect that the system jumps from one state to the other due to random fluctuations. In particular, note that the basin of attraction of the high-cell density state decreases with D (since the discontinuous line of unstable states eventually intersects with the high cell density states). Therefore, our toy model exhibits a plausible mechanism through which increasing dilution rates translate into high cell density states with diminishing resilience to perturbations. The role of toxicity becomes evident if we consider ideal cells resistant to waste accumulation (setting τ = 0). The resulting plot of X vs. D in this case (Fig 5b) reveals a single stable steady state for each value of the dilution rate. There is a discontinuous transition at the washout dilution rate (Dmax), where the cell density suddenly drops to zero. Away from this value, the system is resilient to perturbations since there are no multiple steady states between which jumps can occur. Multi-stability implies that system dynamics are non-trivial, in the sense that different trajectories might lead to different steady states. Therefore it is important for industrial applications to understand how the system is driven to one or another state. We numerically solved Eqs 1 and 2 by performing the FBA optimization at each instant of time, in a manner analogous to DFBA [27]. With the parameter values given in Materials and Methods, we simulated the response of the system to three different profiles of the dilution in time. First, in Fig 6a, a constant dilution rate of D = 0. 6 day−1 is used. Two possible stable steady states are consistent with this dilution rate, attaining different cell densities (cf. Fig 5a). Starting from a very low initial cell density, the system responds by settling at the steady state of lowest cell-density. As can be appreciated in the bottom row of Fig 6a, this state is characterized by an accumulation of toxic waste that prevents further cell growth. A smarter manipulation of the dilution rate makes the high-cell density state accessible. This is demonstrated in Fig 6b, where D starts from a lower value (D = 0. 2 day−1), and is gradually increased until the final value (D = 0. 6 day−1) is reached. The final cell density resulting from this smooth increase of the dilution rate is five-fold larger than the one obtained with a constant dilution rate. This state is also characterized by very low levels of waste accumulation (cf. last row of Fig 6b). We stress that external conditions in the final steady state (dilution rate and medium formulation) are the same in both cases. Finally, Fig 6c shows how the dilution rate can be manipulated to switch from one steady state to another. Starting from the final state of the simulation in Fig 6a, the dilution rate is first decreased to a low value (D = 0. 2 day−1), and then it is pushed back up to the starting value (D = 0. 6 day−1). The system responds by switching from the state with low cell-density to the state with high-cell density. These simulations nicely reproduce the qualitative features of the experiment performed by B. Follstad et al. [11], where a continuous cell culture under the same steady external conditions (dilution rate and medium) switches between different steady states by transient manipulations of the dilution rate. The response of the cell density to transient manipulations of the dilution rate best illustrated in the X, D plane (cf. first row of Fig 6). Then it becomes obvious that the dilution rate must be pushed down to ≈ 0. 2 day−1, otherwise the system will not leave the low cell density state. We determined the steady states of a continuous cell culture of the CHO-K1 line. Cellular metabolism was modeled using the reconstruction given by Hefzi et al. [70], the most complete available at the time of writing. In the simulations we used Iscove’s modified Dulbecco’s growth media (IMDM), which is typically employed in mammalian systems. Similar to what we found in the toy model (cf. Fig 3), and in qualitative agreement with experimental observations [10,14], cells exhibited several metabolic transitions between distinct flux modes as ξ was varied. However, in contrast to the the toy model, the CHO-K1 genome-scale metabolic network displays a rich multitude of transitions, as expected from its greater complexity. Because of their importance in the performance of the culture, we focused on metabolic changes that have an impact on macroscopic properties of the bioreactor, i. e. , those that affect metabolite exchanges with the extracellular media (ui). Although many classifications are possible, we organized our discussion by focusing on five qualitatively different modes based on the secretion of lactate and formate. Fig 7 shows cartoon diagrams of these phases in order of increasing ξ. On the top of each diagram we annotate the nutrients that became limiting for growth during a phase. Blue arrows indicate consumption and red arrows secretion. We focused on metabolites that changed their role between phases. In particular, NH4 was secreted in all phases and therefore was omitted from the figure to reduce clutter. A more detailed representation of our results is given in Fig 8, which shows metabolite concentrations (si) and uptakes (ui) in steady states as functions of ξ for a sub-set of selected metabolites. Red lines in these plots indicate the transitions depicted in Fig 7. For small values of ξ we found that glucose and almost all the amino acids available in the media were consumed, but none of them reached limiting concentrations. We call this the nutrient excess phase, where substrate uptake is limited only by intrinsic kinetic properties of cellular transporters. Remarkably lactate was not secreted at this stage, since pyruvate was converted instead to alanine [33] (although a small fraction of pyruvate was secreted as well [88]). The cell also produced succinate [89] and formate, the later being an overflow product of one-carbon metabolism of serine and glycine [53]. As ξ continues to increase, the first metabolite that becomes limiting is serine. This marks the end of the nutrient excess phase, coinciding also with the onset of lactate secretion. At this point pyruvate is no longer secreted into the culture. Remarkably, aspartate switches from being a secreted byproduct in the first phase [89], to consumption. Even more striking is that the specific uptake rate of aspartate and proline quickly increase until both reach limiting concentrations. A third phase is entered when succinate and formate production ceases, coinciding with a limitation of glycine. Histidine consumption rises steeply until it too reaches limiting concentrations. Other nutrients that limit growth include tyrosine, tryptophan, arginine, lysine and phenylalanine. This phase is also characterized by secretion of acetaldehyde. Remarkably, formate secretion is resumed in a later phase, where glucose, glutamine and asparagine also become limiting. Finally, as ξ approaches the maximum value ξm, lactate and alanine secretion cease. This ideal state attains the highest possible biomass yield per unit of medium supplied per unit time. Note that the increase of ξ has brought an overall qualitative switch to a state of metabolic efficiency where the number of secreted byproducts has dropped significantly, compared to the nutrient excess phase. Notably, the cell-specific ammonia secretion was sustained even in the states of highest biomass yield, indicating a nitrogen imbalance. This behavior has been seen qualitatively in some experiments. For example, using a CHO-derived cell line [12], secretion of ammonia was sustained even after a transition to an efficient metabolic phenotype with low lactate secretion and high cellular yields. However this observation seems to be cell-line dependent, and in another experiment with an hybridoma, ammonia accumulation decreased with increasing ξ [84]. All of the secreted metabolites predicted by our model have been verified in experiments in mammalian cell cultures [33,53,88,89], with the exception of acetaldehyde. Although some of these experiments do not match the cell line and media used in our simulations (CHO and IMDM), these byproducts have been observed in mammalian cell cultures in a variety of conditions, suggesting that they are not restricted to a specific cell line or media. For acetaldehyde our search in the literature did not reveal any experimental evidence refuting or validating its presence in mammalian cell cultures. A possible explanation is the high reactivity of this metabolite. Acetaldehyde binds covalently to glutathione and proteins, forming adducts that are subsequently detoxified [96,97]. However, since the metabolic network does not account for these interactions [70], the model predicts excretion of pure acetaldehyde instead. We thank an anonymous reviewer for bringing this fact to our attention. The performance of cell-lines and media are typically evaluated by measurements performed in batch experiments [90]. Measurements performed in the exponential phase of batch only reveal the behavior of continuous cultures at very low ξ, in conditions of nutrient excess. The existence of a rich multitude of qualitatively distinct metabolic behaviors at higher values of ξ is missed by these experiments and therefore the assessment should not be extrapolated to high-density perfusion systems. As our analysis reveals, several nutrients may switch from basal rates of consumption to growth limiting at later values of ξ, while others go from secreted byproducts to consumption [91]. These examples indicate that nutrients could be in excess in a batch experiment but need not be so in the ideal regime of high-cell density perfusion cultures, at high ξ. Our model suggests that a better characterization of a cell-line and media formulation can be obtained in a chemostat, since the full spectrum of values of ξ can be explored in this device and it faithfully reproduces all the metabolic transitions found in perfusion. The effects of toxic byproduct accumulation are explored in Fig 9. We considered the toxic effects of the most commonly studied metabolites in this regard: lactate and ammonia, although the model can easily accommodate the effects of additional toxic compounds if necessary. In Fig 9a we plot the effective growth rate, μ as function of ξ. Stable states are drawn in continuous line, unstable states are dashed and the red dots indicate the metabolic transitions depicted in Fig 7. Note that μ* (ξ) is not monotonous. In particular, metabolic transitions resulting in lactate and ammonia secretion peaks produce a sink in the curve μ* (ξ). On the other hand, metabolic transitions associated to the secretion of other non-toxic byproducts do not imply changes in the monotonicity of μ* (ξ). The non-monotonicity of μ* (ξ) results in multiple stable states coexisting at the same dilution rate, as evident in the bifurcation diagram Fig 9b. This resonates with the results obtained in the simpler model considered above, and is also consistent with many experimental observations of bi-stability in the literature [10–14]. The regime with high-cell density corresponds to a higher value of ξ and exhibits a lower accumulation of toxic byproducts (lactate and ammonia). Metabolism in this regime is also more efficient, with less byproducts secreted (cf. Fig 7). On the other hand, low cell density states are wasteful, with high levels of environmental toxicity preventing further cell growth. Again, bi-stability implies the existence of an hysteresis loop (orange arrows in the figure), where the system may suffer abrupt transitions between high and low cell densities. In this work we have presented a model of cellular metabolism in continuous cell culture. Although similar in spirit to DFBA, our dynamic equations include terms accounting for the continuous medium exchange that enables steady states in this system. We presented a simple method to compute the steady states of the culture as a function of the ratio between cell density and dilution rate (ξ = X/D), scalable to metabolic networks of arbitrary complexity. In the literature 1/ξ is known as the cell-specific perfusion rate (CSPR), introduced by S. Ozturk [83] who already made the empirical observation that control of the CSPR can be used to maintain a constant cell environment independent of cell growth [83]. Our model theoretically supports this idea and leads to a stronger conclusion: that for a given cell line and medium formulation, the steady state values of the macroscopic variables of the bioreactor are all unequivocally determined by ξ. Therefore, ξ is an ideal control parameter to fix a desired steady state in a continuous cell culture. The model is consistent with multi-stability, a phenomenon repeatedly observed in experiments in continuous cell cultures where multiple steady states coexist under identical external conditions. Moreover, our model accounts for metabolic switches between flux modes, experimentally observed in continuous cell culture in response to variations in the dilution rate [92]. These transitions affect the consumption or secretion of metabolites and the set of nutrients limiting growth. As a consequence, the metabolic landscape of steady states in perfusion cell cultures is complex and cannot be reproduced in batch cultures. This has the practical implication that assesments of medium quality and cell line performance carried out in batch [90] should not be extrapolated to perfusion, since they might be missleading in this setting. However, our analysis reveals a simple scaling law between steady states in the chemostat and any perfusion system. The landscape of metabolic transitions in the later system can be faithfully reproduced in the chemostat. Thus, for a fixed cell-line and medium formulation, the diagram displaying the values of ϕX versus ϕD in steady state is invariant across perfusion systems with any bleeding ratio (ϕ), cf. Eq 16, while metabolism is equivalent if the ratio ξ = X/D is the same. The practical consequence is that the chemostat is an ideal experimental model where cell-lines and medium formulations can be benchmarked for their performance in high-cell density industrial continuous cultures. Further, the model predicts that multi-stability is a consequence of negative feedback on cell growth due to accumulation of toxic byproducts in the culture. The qualitative complexity of the ϕX versus ϕD diagram depends only on the behavior of toxic metabolites. Moreover, multi-stability implies that the system is sensitive to initial conditions and transient manipulations of external parameters. In practice, the dilution rate must be manipulated carefully to bring the system to a desired state. Thus, starting from a seed of low cell density, sharp increases of the dilution may land the system on a steady state of high toxicity and low biomass. On the other hand, slowly increasing the dilution rate will surely lead towards high-cell density states. The conclusions stated above rely on the validity of our assumptions. In particular, we have considered a homogeneous cell population in a well-mixed bioreactor. Both assumptions are behind many models published in the field and provide reasonable fits to experimental data [37]. Mechanical stirring of the culture typically achieves a well-mixed solution, but care must be taken to prevent mechanical damage to the cells [93] (but see Ref [94]). Moreover, that the cell population can be treated attending only to its average properties is justified by the large number of cells in a typical culture (∼106 − 108 cells/mL), although in some settings cell-to-cell heterogeneity might become relevant [95]. Next, to develop a specific model of cellular metabolism, we adopted a flux-balance approach [38], where cells are assumed to optimize their metabolism towards growth rate maximization. Although this framework is well supported in the literature [20], it is worth noting that we did not consider the kinetics of intracellular metabolites or additional regulatory mechanisms that may also control metabolic fluxes. Additionally, the quantitative predictions of the model rely on the accuracy of parameters found in the literature and databases. Among these, the flux cost coefficients (αi, Eq 4) are not available for many enzymes. If too many of these parameters are absent, calculations from FBA might be degenerate [17,54]. Another important omission from the present model is that we did not consider explicitly the exchange between the culture and a gaseous phase. In particular, this includes oxygen exchange. Therefore our approach is only valid if this exchange does not become limiting to cellular growth. Despite these limitations, we have shown that the model predictions are in qualitative agreement with experimental data. More importantly, the conclusions stated above are independent of the values of model parameters. Finally, we discuss briefly the relevance of our work in tissues. Constant perfusion flows resulting from the blood stream imply that tissues can reach a steady state in principle similar to the chemostat. Although many complications must be considered in this case, including circadian oscillations, heterogeneous cell populations and spatial structure (Ref. [60] presents an attempt to deal with the later two), some of the conclusions derived in this manuscript could still be relevant in this context. For example, it would be very interesting to explore if the metabolic profile of cells from different tissues can be correlated to the ratio between cell density and average perfusion rate experienced at the tissue site (i. e, the ξ parameter defined in this work). Moreover, the majority of in vitro experiments studying the metabolism of human cells, are conducted in batch due to ease of operation. Our work points out that the metabolic profile observed in this setting might be very different from the state in a perfused tissue.

While at present most biotechnology industrial facilities adopt batch or fed-batch processes, continuous processing has been vigorously defended in the literature and many predict its adoption in the near future. However, identical cultures may lead to distinct steady states and the lack of comprehension of this multiplicity has been a limiting factor for the widespread application of this kind of processes in the industry. In this work we try to remediate this providing a computationally tractable approach to determine the steady-states of genome-scale metabolic networks in continous cell cultures and show the existence of general invariance laws across different cultures. We represent a continuous cell culture as a metabolic model of a cell coupled to a dynamic environment that includes toxic by-products of metabolism and the cell capacity to grow. We show that the ratio between cell density and dilution rate is the control parameter fixing steady states with desired properties, and that this is invariant accross perfusion systems. The typical multi-stability of the steady-states of this kind of culture is explained by the negative feedback loop on cell growth due to toxic byproduct accumulation. Moreover, we present invariance laws connecting continuous cell cultures with different parameters that imply that the chemostat is the ideal experimental model to faithfully reproduce the complex landscape of metabolic transitions of a perfusion system.

Abstract Introduction Materials and methods Results and discussion

cell physiology medicine and health sciences pathology and laboratory medicine chemical compounds biological cultures metabolic networks cell processes cell metabolism toxicology physiological processes toxicity metabolites network analysis cell growth cell cultures research and analysis methods computer and information sciences chemistry biochemistry ammonia cell biology physiology secretion biology and life sciences physical sciences metabolism

2017

Characterizing steady states of genome-scale metabolic networks in continuous cell cultures

In view of the current widespread use of and reliance on a single schistosomicide, praziquantel, there is a pressing need to discover and develop alternative drugs for schistosomiasis. One approach to this is to develop High Throughput in vitro whole organism screens (HTS) to identify hits amongst large compound libraries. We have been carrying out low throughput (24-well plate) in vitro testing based on microscopic evaluation of killing of ex-vivo adult S. mansoni worms using selected compound collections mainly provided through the WHO-TDR Helminth Drug Initiative. To increase throughput, we introduced a similar but higher throughput 96-well primary in vitro assay using the schistosomula stage which can be readily produced in vitro in large quantities. In addition to morphological readout of viability we have investigated using fluorometric determination of the reduction of Alamar blue (AB), a redox indicator of enzyme activity widely used in whole organism screening. A panel of 7 known schistosome active compounds including praziquantel, produced diverse effects on larval morphology within 3 days of culture although only two induced marked larval death within 7 days. The AB assay was very effective in detecting these lethal compounds but proved more inconsistent in detecting compounds which damaged but did not kill. The utility of the AB assay in detecting compounds which cause severe morbidity and/or death of schistosomula was confirmed in testing a panel of compounds previously selected in library screening as having activity against the adult worms. Furthermore, in prospective library screening, the AB assay was able to detect all compounds which induced killing and also the majority of compounds designated as hits based on morphological changes. We conclude that an HTS combining AB readout and image-based analysis would provide an efficient and stringent primary assay for schistosome drug discovery. Schistosomiasis is the most important helminth infection and the second most important parasitic disease next to malaria. It continues to spread in parts of the world due to water management and irrigation projects [1]. The major current strategy for control of schistosomiasis is chemotherapy and significant reductions in prevalence of infection and of severe disease have been achieved in several parts of the world e. g. Central and South America, North Africa and China [2]. A common approach is to treat all school-aged children in areas where the prevalence of schistosomiasis is over 10% and the whole community when the prevalence is over 50%. Such control has now been extended to several African countries through the Schistosomiasis Control Programme funded by the Bill and Melinda Gates Foundation and this has led to significant reductions in the prevalence, intensity and morbidity of infection [3]. However, chemotherapy does not interrupt transmission and so for this morbidity reduction to be maintained repeated periodic treatment will need to be continued for the foreseeable future. Since its introduction in the 1980s praziquantel (Biltricide®) has been the mainstay of control programmes and it is now the only drug being used for mass treatment of schistosomiasis. It is an effective drug with broad spectrum activity against all five human schistosome species, low toxicity, few side-effects, simple administration and currently low cost. In mass treatment campaigns at a dose of 40 mg/kg praziquantel usually results in parasitological cure rates of around 70% and egg count reduction rates of more than 90% [4]. The marked increase in use of, and reliance on, repeated treatments with praziquantel has raised concerns about the possible emergence of drug resistance which, if it were to occur, would leave us without an effective schistosomicide [5]. There have been sporadic reports of low efficacy and of treatment failure in individuals from different parts of the world but as yet no convincing evidence of development and selection of resistance in endemic areas. However, strains of schistosomes which have been isolated from drug treatment failures show lower susceptibility to praziquantel and strains with stable elevated tolerance to the drug can be selected in the laboratory [6]. Worries about reliance on one schistosome drug and the possibility of the emergence of drug resistance led to the establishment of the “Helminth Drug Initiative (HDI) ” (http: //apps. who. int/tdr/svc/research/lead-discovery-drugs/workplans/helminth-drug-initiative) by the WHO Special Programme for Research and Training in Tropical Diseases (TDR) for the discovery of new schistosomicides [7]. As part of the HDI, whole organism screening was established with the London School of Hygiene and Tropical Medicine chosen as one of the screening centers. The screen adopted involved culture of adult worms with test compounds for 5 days with activity being determined by microscopic examination of worm death [8]. The HDI was initially based on highly selected libraries and throughput was not then the major consideration. However, with the provision of larger compound sets we implemented a 96 well microplate primary screen using in vitro derived schistosomula which allows markedly higher throughput than the adult worm-based assay. A similar assay was recently described by Abdulla et al [9]. Such assays are limited by the need for manual microscopic reading and, in order to develop a High Throughput Screen (HTS), the assay needs to be automated [10]. Therefore, alongside testing the use of High Content Screening (HCS) to allow automatic image analysis, we have investigated the use of a plate-based assay using Alamar Blue (AB), a redox indicator of enzyme activity which has been successfully used for colorimetric or fluorometric determination of viability of a number of protozoan parasites in whole organism drug screening e. g. African trypanosomes and Leishmania [11]–[12]. MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyl tetrazolium bromide) which is also an indicator of enzyme activity has previously been shown to be reduced by adult schistosomes in vitro and used as a measure of parasite viability [13]. AB has the advantage over MTT in that it is non-toxic and so can be used to monitor effects over time. We report here our results on the development and validation of the AB assay compared with morphology-based (microscopic) assessment of compound activity. For this purpose we have tested: (i) Several compounds (referred to here as “Standards”) with proven activity against schistosomes in vivo either in humans (oxamniquine [OX] [14]; praziquantel [PZ] [15]; methyl-clonazepam [MCZ] [16]; oltipraz [OPZ] [17]), in sub-human primates (Ro15-5458 [Ro15] [18]) or at least in mice (dihydroartemisinin [DHA] [19] [the analogue, artemether, has also been shown to be active in humans [20]]; clonazepam [CZ] [21]). In addition to their proven efficacy in vivo these Standard compounds were chosen because we had shown them to have a range of effects and speed of action against adult worms and schistosomula in vitro. Thus, PZ, MCZ, CZ and OPZ readily kill adult worms in vitro and have IC50s of 0. 36,0. 37,1. 42, and 1. 25 µg/ml, respectively (unpublished observations). However, OX, Ro15 and DHA have IC50s >10 µg/ml, the maximum concentration we use in our primary adult worm screening for WHO-TDR [8]; (ii) A panel of compounds found to be active or inactive by the adult worm assay during prior routine adult worm screening of compound libraries mostly supplied through the HDI (WHO-TDR); (iii) A subsample of compounds from a WHO-TDR library being screened for the first time. An ethics statement is not required for this work. Cercariae of the Puerto Rican strain of Schistosoma mansoni [22] were shed in clean tap water from infected Biomphalaria glabrata snails exposed to direct illumination for 1 h. The cercariae were concentrated to 20 ml using an 8 µM filter (Sartorius CN, Scientific Laboratory Supplies, Ltd) on a concentration apparatus (Millipore) and cooled on ice for 30 min. The water was carefully removed from the pelleted cercariae and replaced by cold serum free medium 169 [23] (M169) supplemented with 300 U/ml Penicillin (Gibco, UK), 300 µg/ml Streptomycin (Gibco, UK), and 160 µg/mL Gentamicin (Sigma, UK) (Incomplete M169). Under sterile conditions, cercariae were mechanically transformed into schistosomula using the ‘Syringe Method’ [24]. The cercarial head and tail suspension was layered onto a sterile gradient of 50% and 60% Percoll (Sigma, UK) in M169 in 15 ml polystyrene tubes [25]. The tubes were spun for 10 min at 350× g at 4°C. After centrifugation each of the layers was carefully removed and the cercarial heads (schistosomula) recovered from the 60% layer, washed twice in 20–25 ml of Incomplete M169 by centrifuging at 400× g and 4°C for 2 min. The schistosomula were washed in M169 supplemented with 100 U/ml Penicillin, 100 µg/ml Streptomycin and 5% foetal calf serum (Sigma, UK) (Complete M169, cM169). They were then transferred into 6-well plates, incubated overnight at 37°C and 5% CO2 and checked for high viability (≤2% dead or damaged larvae) prior to use in screening. Standard drugs: PZ, CZ and DHA were from Sigma, UK; MCZ and Ro15 were a kind gift from Dr H. Stohler (Hoffman-La Roche, Basle, Switzerland); OX was from Pfizer Ltd. Sandwich, UK; OPZ was provided by WHO-TDR, Geneva, Switzerland. These were dissolved in dimethylsulphoxide (DMSO) (Sigma, UK). Test compounds: these were from a number of different compound collections provided by WHO-TDR for routine screening using the adult worm assay. Since this is carried out at up to 10 µg/ml or (when compounds of known molarity are provided) at 12. 5 µM, stock solutions of 10 mg/ml or 12. 5 mM in DMSO were used. Mid-dilutions were performed as necessary in 100% DMSO and 1 µl added to 100 µl/well of cM169 in 96 well plates (Nunc, UK). Finally, 100 schistosomula were added in 100 µl of cM169 to each well. Negative control wells contained schistosomula cultured in either cM169 medium or 0. 5% DMSO in cM169. The final concentration of 0. 5% DMSO did not affect the larvae within the culture periods used. Experiments with standard drugs were carried out in triplicate wells. The compounds were tested at 10 µg/ml (i. e. 28,32,30,44,32,38,35 µM for OX, PZ, MCZ, OPZ, CZ, Ro15 and DHA, respectively) or 1 µg/ml (i. e. 2. 8–4. 4 µM). Viability of schistosomula was assessed on days 1,3, 5 and 7 and in some experiments on day 14 using an inverted microscope (Leitz Diavert Wetzlar, Germany). Drug effects were determined by recording all schistosomula in each well as unaffected, dead (immotile, often showing a characteristic uniform shape and granular appearance) or morphologically damaged (showing a range of altered shapes, granularity and/or blebbing but still with some motility). 20 µl of Alamar Blue (AbD Serotec, UK) was added to each well and the plates incubated at 37°C and 5% CO2 for 24 h unless otherwise stated. The fluorescence intensity was measured using a Spectramax Gemini plate reader (Molecular Devices) using an excitation wavelength of 530 nm and an emission wavelength of 580 nm. Prism 4 (GraphPad Software, Inc) was used for graph drawing and statistical analysis. Student' t test was used to determine the significance of differences between mean values. An initial experiment showed that AB could readily detect metabolic activity during schistosomula culture. As shown in Figure 1 there was induction of significant (P<0. 01) fluorescence as early as 1 h following incubation with 1000 schistosomula (5000/ml). By 3 hrs significant levels of fluorescence were seen with both 100 and 1000 schistosomula and this was significantly enhanced (P<0. 001) at 24 h (4 and 25 fold higher fluorescence levels, respectively, for 100 and 1000 schistosomula compared with control without parasites). At each time point 1000 schistosomula induced significantly greater fluorescence than 100 (at 1 and 3 h P<0. 01; at 24 h P<0. 0001) but use of such high numbers of larvae for screening would greatly reduce possible throughput. Based on such data, future experiments with AB used 100 schistosomula per well read at 24 h or later. In preliminary experiments we showed that OPZ, which kills schistosomula in vitro, inhibited this AB conversion and so we were encouraged to test the assay further. We next tested the larval assay against a random collection of 33 compounds which had proved to be active in vitro at 10 µg/ml against adult schistosomes in our standard assay [8] (i. e. “Adult +ve”) along with 30 randomly selected adult negative compounds (“Adult −ve”). Microscopy-based assay (morphological observations). Figure 3A shows the results using 10 µg/ml of these compounds. 10/33 (30%) of Adult +ves had killed all or nearly all the larvae by day 1 and 80% were classified as hits (≥70% damaged or dead). By day 3,28/33 (85%) had induced killing of most larvae and all were detected as hits. Larval death increased at days 5 and 7 at which time all of the compounds caused >50% death and all but 4 induced 100% parasite death. Using 1 µg/ml (Figure 3B) only 14/33 (42%) would have been classified as hits on morphological grounds on days 3,5 and 7. None of the Adult −ves induced significant parasite death at either drug concentration and by day 3 only two were classed as hits based on morphological damage when tested at 10 µg/ml, giving a false positive rate of ∼7% (NB “false positive” is used here to define activity relative to the adult assay. The latter is taken as the reference because it is used as the secondary and definitive screen for hit progression. Activity in the larval but not the adult screen may be due to higher sensitivity of the schistosomula or to larval specific action). At 1 µg/ml there were no false positives amongst the Adult −ves. In the following experiment we compared the 3 day microscopy and 7 day AB assays during primary screening of 558 compounds. These were from a compound library provided by WHO-TDR at known molarity in which circumstance primary screens are run at 12. 5 µM. As shown in Figure 4, the visual assay detected 8 compounds lethal to the larvae and 27 designated as hits based on damage i. e. a total of 35 “hits” (∼6% hit rate). The AB assay readily detected all 8 of the drugs causing larval death and 21/27 (78%) of the morphologically damaged hits. In addition there were 27/558 compounds causing ≥25% inhibition in the AB assay which were not detected as hits by morphology. Overall, therefore, 56/558 compounds induced AB inhibition of ≥25% i. e. a hit rate of 10%. The 35 visual hits were tested in the secondary ex-vivo adult assay and 5 were shown to be positive. All of these were in the group of 8 compounds which had resulted in schistosomula death by day 3 and which also caused marked inhibition of AB fluorescence at day 7. Of the 35 visual hits 6 caused crystal formation/drug deposition in or on the parasites and in some cases this seems to pierce the larvae e. g. Figure 5. Such effects are not uncommon and assessment of the morphological condition of the larvae can be difficult in such cultures. None of the 6 were hits in the adult assay and two of them gave the highest AB AU fluorescence values in Figure 4 indicating that the larvae were still highly metabolically active. Abdulla et al [9] recently described a medium throughput drug screen based on microscopically determined morphological changes to schistosomula. It was reported that attempts to quantitate cell death in schistosomula using nuclear dyes did not correlate with observed morphological effects seen under bright field microscopy. Our experience supports the conclusion that morphological changes to schistosomula represent a rapid and sensitive means of detecting anti-schistosome activity. An assay based on inhibition of AB reduction was also very effective in detecting severely damaged and dead schistosomula and with a sufficiently long culture period (7 days) was able to detect most of the known schistosome active compounds tested. The microscopy-based morphological assay described by Abdulla et al [9] involved a primary schistosomula screen at 7 days, a secondary similar screen at 24 h to detect fast-acting compounds and then a tertiary screen using ex vivo adults read at day 1–4. Priority for progression to the adult screen was based on activity against schistosomula at 24 h. Based on initial screening of two commercially obtained compound collections it was concluded that the hit rates of 25 and 15% when screening was at 10 and 5 µM were too high whereas the hit rate of 10% at 1 µM was acceptable. In our experiments with a range of known schistosome active drugs (Standards) we found that the microscopy-based larval screen described here, which is standardly read at day 3, detected all 7 compounds as hits when tested at 10 µg/ml (i. e. 28–44 µM) but at 1 µg/ml (i. e. 2. 8–4. 4 µM) CZ, Ro15 and DHA were not detected even at 7 days. The failure to detect Ro15 and DHA at 1 µg/ml may be seen as a failure of the assay but in practice these two compounds would not have been detected in our secondary adult worm screen since they are inactive against the adult worms in vitro at 10 µg/ml. In our testing of a collection of 33 compounds originally detected as hits by primary screening against adult worms, all were detected at 10 µg/ml, but at 1 µg/ml 58% were not detected even when cultured up to 7 days. We conclude that testing at 1 µg/ml (estimated at ∼3 µM) missed an unacceptable percentage of the Adult +ves. As mentioned above the choice of 1 µM in the earlier report [9] was an empirical one based on achieving an acceptable hit rate (∼10%). In the present studies we found that a selection of Adult −ve compounds from the collections which yielded the Adult +ves gave a low false positive rate of ∼7% at 10 µg/ml. Since these Adult −ves were randomly selected from the libraries this value of 7% represents the hit rate which would be expected from these collections using the larval assay at 10 µg/ml. However, in our prospective screening of new collections of compounds using the schistosomula primary screen at 10 µg/ml hit rates up to 16% have been obtained whilst with compounds tested at 12. 5 µM we have found an average of 6% hits. Given the variation in hit rates seen with different libraries, the rational approach is to determine a concentration that gives an acceptable hit rate for progression to secondary screening [9]. It is interesting that all of the compounds we tested here which had been found to be active in the adult worm screening of WHO-TDR collections proved to be highly active against the schistosomula as judged visually. Thus of 38 such adult actives tested all but 5 had killed >75% of larvae and all had caused ≥70% morphological damage by day 3. By contrast the Standards produced more diverse effects, readily detectable by morphological damage but, with the exception of OPZ, not resulting in high levels of parasite death up to a week in culture. AB fluorescence was found to be a reliable indicator of schistosomula viability and the AB assay was able to detect severely damaging and lethal effects caused by test compounds. However, it was much less effective in detecting the more subtle morphological effects such as those caused by most of the Standards. For example, OPZ which kills around 75% of larvae by day 7 in culture at 10 µg/ml was reliably detected as causing >25% inhibition of AB conversion from day 3 onwards, reaching 100% inhibition by day 7. On the other hand PZ, MCZ, CZ and Ro15 which damage all but kill few schistosomula within 7 days had more modest effects on AB conversion although, if cultured up to 7 days, ≥25% inhibition was generally observed. However, having run these Standards as controls in many screens we have noted that occasionally these four compounds fail to show a ≥25% inhibition and so would not be detected as hits. Altering test compound concentrations, culture duration and interval to AB reading did not lead to a protocol which improved this. We conclude that relative to microscopy the AB assay would occasionally miss detecting PZ, Ro11, CZ and Ro15 and would not detect OX. It is, however, worth noting that neither Ro15 nor OX are active in the adult in vitro visual assay and so would not be taken forward as hits following secondary screening. Comparison of the morphology and Alamar blue assays against collections of compounds previously shown to be positive or negative against adult worms in vitro showed that both assays were 100% sensitive in detecting the Adult +ves at 10 µg/ml. However, whereas the microscopy assay was positive for only ∼7% of Adult −ve compounds from these source libraries, the AB assay produced a wide range of values with 40% of these falling below the cut-off for hits in this assay. When the Adult −ves were tested at 1 µg/ml the AB assay showed only 3% as hits and also showed a somewhat higher sensitivity in detecting Adult +ves (67%) than the morphology assay (42%). Since the Adult −ve compounds were randomly chosen from compound collections sent for testing, the 40% and 3% “hit” rates with 10 and 1 µg/ml, respectively, reflect the hit rates which would be obtained if the AB assay were used for primary screening of these collections. As discussed above the optimal concentration for a particular library would need to be chosen empirically. So, for AB screening of this particular compound collection the appropriate concentration would be somewhere between 10 and 1 µg/ml in order to yield a manageable hit rate for progression to the adult worm secondary screen. Notably, in this retrospective testing of the Adult +ves at 1 or 10 µg/ml, all wells, apart from 2 on Day 1, where there was larval death (range 10–100% cf controls 0. 2–2%), the AB values were >25% lower than the controls and in most instances >50% lower. Similarly, in the prospective testing at 12. 5 µM (Figure 4) all wells in which larval death was recorded (37–100% death) were readily detected by the AB assay. In conclusion the AB assay has proved to be very effective in detecting compounds which have severe effects on the larvae resulting in a proportion of the larvae dying. The assay is also very suitable for HTS automation and is valuable when there is crystal formation on the parasites which obscures accurate observation. However, it showed somewhat lower sensitivity and reliability compared with manual visualization in detecting the more subtle damage caused by some known anti-schistosome compounds including praziquantel. Therefore, we envisage an HTS may usefully incorporate both AB and image-based analysis and are currently investigating if automatic HCS image-based analysis can be developed which has comparable sensitivity to the manual visualization assay.

Only one drug, praziquantel, is widely available for treating schistosomiasis, a disease affecting an estimated 200 million people. Because of increasing usage there is concern about development of praziquantel drug resistance and a perceived need to develop new schistosomicides. Possible sources of these are large collections of compounds held by pharmaceutical companies and academic institutions. Anti-schistosome activity can be detected in vitro by visually assessing damage to cultured adult schistosome worms, but these are large and are recovered from mice which somewhat limits screening throughput. By contrast, schistosomula can be produced in vitro and used for screening in microwell plates, thus allowing medium throughput screening. High throughput screening (HTS) would require automated readout of schistosomulicidal action rather than manual microscopy. Here we report on the use of Alamar blue (AB), a fluorescent indicator of cell viability which can be measured rapidly and automatically. The AB assay was readily able to detect compounds causing death or severe damage to the larvae but was less reliable than microscopy for more subtle morphological changes including those induced by some known schistosome drugs. It is concluded that an automated HTS would benefit from integrated use of both AB and automatic image-based morphology assays.

Abstract Introduction Materials and Methods Results Discussion

infectious diseases/helminth infections infectious diseases/antimicrobials and drug resistance infectious diseases/neglected tropical diseases

2010

Comparison of Microscopy and Alamar Blue Reduction in a Larval Based Assay for Schistosome Drug Screening

Human populations outside of Africa have experienced at least two bouts of introgression from archaic humans, from Neanderthals and Denisovans. In Papuans there is prior evidence of both these introgressions. Here we present a new approach to detect segments of individual genomes of archaic origin without using an archaic reference genome. The approach is based on a hidden Markov model that identifies genomic regions with a high density of single nucleotide variants (SNVs) not seen in unadmixed populations. We show using simulations that this provides a powerful approach to identifying segments of archaic introgression with a low rate of false detection, given data from a suitable outgroup population is available, without the archaic introgression but containing a majority of the variation that arose since initial separation from the archaic lineage. Furthermore our approach is able to infer admixture proportions and the times both of admixture and of initial divergence between the human and archaic populations. We apply the model to detect archaic introgression in 89 Papuans and show how the identified segments can be assigned to likely Neanderthal or Denisovan origin. We report more Denisovan admixture than previous studies and find a shift in size distribution of fragments of Neanderthal and Denisovan origin that is compatible with a difference in admixture time. Furthermore, we identify small amounts of Denisova ancestry in South East Asians and South Asians. An archaic genomic segment introgressed into a population is expected to have a high density of variants not found in populations without the introgression. We use a Hidden Markov Model (HMM) to classify genomic segments into states with varying density of such variants. We focus on a scenario where introgression from a deeply divergent archaic population only happened into an ingroup and not the outgroup, see Fig 1A. By removing variants found in the outgroup we can better distinguish introgressed segments from non-introgressed segments based on the density of remaining variants, see Fig 1A. These remaining variants, which we denote private variants (because they are private to the ingroup with respect to the outgroup) can either have occurred on the branch starting from the split of the ingroup and outgroup, or on the introgressing population’s branch. Because the introgressed segments have had a longer time to accumulate variants, they should have a higher density of private variants. Thus, we define a HMM with two states. The hidden states are Ingroup and Archaic, and the probability for changing state in the Ingroup is p and the probability for changing state in the Archaic is q, see Fig 1B. The probability of changing state can also be expressed in terms of a constant recombination rate between windows r ∙ L, the admixture time Tadmix and admixture proportion a, see Fig 1B. For practical purposes we bin the genome into windows of length L (typically L = 1000 bp). The number of private variants observed in a window is Poisson distributed with a rate λIngroup and λArchaic, respectively where λIngroup = μ ∙ L ∙ λIngroup and λArchaic = μ ∙ L ∙ λArchaic, μ is the mutation rate, TIngroup is the mean coalescence time for the ingroup and the outgroup and Tarchaic is the mean coalescence time for the archaic population and the outgroup, see Fig 1C. We make a correction to the rates to take into account the number of missing bases in a window and the local mutation rate. For window i we have λIngroupi=μi∙Li∙TIngroup and λArchaici=μi∙Li∙TArchaic, where μi is the local mutation rate and Li is the number of called bases in a window. The set of transition parameters p, q and the Poisson parameters λIngroup, λArchaic that maximize the likelihood given the observations are found using the Baum-Welch algorithm for an individual genome. These parameters are informative of the mean coalescence times between the ingroup and outgroup and between the archaic and the outgroup, the admixture time and the admixture proportion if we assume a known mutation rate μ and a known recombination rate between windows rL. Once the set of optimal parameters are found they can be used to decode the genome, using posterior decoding to identify candidate introgressed segments as consecutive regions with posterior probability of coming from the archaic state above some threshold. Until now we have assumed the data is phased haploid genomes. But to avoid problems with phasing we run this model on unphased diploid genomes. Heterozygous archaic segments will still stand out from homozygous non-introgressed segments. Formally this is equivalent to assuming that homozygous introgressed segments are sufficiently rare that they can be ignored for model fitting. In practice any homozygous archaic segments will have higher private variant density than heterozygous segments, so in the absence of a homozygous HMM state they will be classified with the heterozygous state. We show how to convert model parameters to demographic parameters, both when analyzing haploid and diploid genomes in S1 Dataset. We note that this method will likely only work in cases where the coalescence time distribution of the ingroup and archaic segments are sufficiently different. This will work better in cases where the variation in the ingroup is a subset of variation in the outgroup so the majority of variation in the common ancestor can be removed, as the case of Non-Africans and Africans. To investigate the ability of our model to identify archaic (Neanderthal and Denisovan) admixture into Papuans we simulated whole diploid autosomal data using a coalescent simulator, with admixture with an archaic hominin 1,500 generations ago replacing 0–25% of the population– (a script with all demographic parameters is shown in S2 Dataset and a graphical representation of the demography is shown in S1 Fig). We simulated different scenarios to test the effects of running the model on haploid versus diploid data, adding missing data, varying recombination rate and varying a mutation rate. First, we simulated five individuals where every base in the genome is called equally well and there is a constant recombination rate of 1. 2 ∙ 10−8 events per basepair per generation. We call this dataset the ideal data. Second, we simulated five individuals with missing data (using the repeatmask track for the human reference genome hg19 [15]) and variations in local recombination rate (using HapMap phase II [16]) to test the effect of missing data and recombination. Third we add variations in local mutation rate to the second scenario as described in the materials and method section. We binned all genomes into bins of 1000 bp, and removed all variants found in any of 500 simulated Africans, 100 simulated Europeans and 100 simulated Asians. We train the model on both haplotype data and unphased diploid genotype data for the simulated individuals. The latter is similar to situations where phased data is not available. We estimated the transition and emission parameters using the Baum-Welch algorithm and used them to get an estimate for the admixture time Tadmix, the admixture proportion a and the mean coalescent times with the outgroup TIngroup and TArchaic for the ingroup and archaic segments respectively. We also show the sensitivity and precision of the model at different admixture proportions. We only show the estimated parameters and error rates for simulations with missing data and varying recombination rate, because the addition of a varying mutation rate has a very minor effect, see Fig 2B. A table containing all parameters from the model and the corresponding demographic parameters are listed in S3 Dataset. We evaluate the performance of the model in terms of precision (amount of predicted archaic sequence that is archaic/amount of predicted archaic sequence) and sensitivity (amount of predicted archaic sequence that is archaic/amount of true archaic sequence). When admixture proportions are low (less than 2%) the model does not fit the emission or transition parameters well and the precision is around 70%. With an admixture proportion greater than 2% the model fits the parameters well with sensitivity above 80% and precision greater than 80% at a posterior probability cutoff at 0. 5 (mean posterior probability of being archaic for all windows in segment). Raising the posterior probability cutoff from 0. 5 to 0. 8 increases the precision to around 90% while the sensitivity is still above 75% for 5% admixture as can be seen in S3 Fig. The effect of changing the cutoff for different admixture proportions are shown in S3 Dataset. Across all scenarios TIngroup is slightly underestimated for data sets, while TArchaic is overestimated, see Fig 2A. The error in estimating TIngroup and TArchaic is likely due to model misspecification: the model effectively represents the distribution of coalescence times with the mean coalescence time only, and assumes all sites are heterozygous for both states. The effect of treating all sites as heterozygous is seen when comparing the TIngroup for haploid (mean = 2,952 generations ago) and diploid data (mean = 2,646 generations ago). The average simulated TIngroup was 3,109 generations ago. The effect is also seen when comparing TArchaic for haploid (mean = 36,844 generations ago) and diploid data (mean = 38,508 generations ago). The average simulated TArchaic was 36,462 generations ago. Furthermore, the model tends to classify deeply coalescing haplotypes from the common ancestor of ingroup and outgroup as admixed. This effect is greater in simulations with low admixture proportions (Fig 2A). The misspecification of the coalescent times is more problematic in cases where the ancestral population of ingroup and outgroup is very large and/or contains strong population structure. This can be overcome to some extent by sequencing more individuals from the outgroup, but the improvement becomes limited if there have been bottlenecks in the outgroup. The effective population size of the archaic source population after its separation from the common ancestor of ingroup and outgroup does not matter, since we assume very few lineages contribute to the admixture in this population that affects the test sample. Finally, a large population size in the ancestral population will increase the variance in coalescent times in the archaic state, but we would expect this to have less consequence on the models ability to discriminate. We find that when admixture proportion are >5% and the recombination rate across the genome is constant the model recovers the right admixture time (S3 Dataset). However with varying recombination rate we underestimate the admixture time, which might be due to the model’s failure to identify around 80% of the short segments (see S2 Fig). This would increase the average segment length and lead to a more recent admixture time. The admixture proportion is fitted well for all simulations except where the admixture proportion is zero. In this case, the “archaic” state is assigned to a set of segments with longer coalescent times to lineages in the outgroup, but their posterior probability is lower than for real admixed segments. The inconsistency between the estimated admixture proportion and the amount of segments recovered by posterior decoding, could potentially be used to discriminate whether or not there is admixture (Fig 2C). For admixture proportions below 2% (0. 5% for data with varying recombination rate) we observe that Tadmix is estimated to be greater than TIngroup (meaning admixture happened before the split of ingroup and outgroup), which is not possible and also indicates breakdown of the model. We also compared the performance of our method to Sstar 2014[17], Sstar 2016[9] and Sprime 2018[13] under scenarios of varying recombination rate, varying recombination rate and missing data and varying mutation rate, missing data and varying mutation rate. Our method shows improved tradeoff between sensitivity and precision, because we take missing data into account when training and decoding the model, (Fig 2B and S3 Fig). Having validated the model, we applied it to 14 Papuan individuals from the Simons Genome Diversity Project [18], 40 Papuans from [19] and an additional 35 Papuans [9]. We also analyzed individuals from West Eurasia, East Asia and South East Asia from the Simons Genome Diversity Project[18]. We note that variants in these datasets have been found using different bioinformatics pipelines with different filters but that the counts of heterozygous and homozygous variants are similar, see S4 Dataset. We estimate the background mutation rate in windows of 100 kb, using the variant density of all variants in African populations from the 1000 Genomes Project. Our model will not be able to distinguish Neanderthal from Denisova segments in Papuans, because the Denisovans and Neanderthals share a common ancestor before they do with humans and therefore the mean coalescence time with humans will be the same [1]. This means that the Poisson parameters will be the same as they both depend on TArchaic. However, we are able to enrich for Denisova versus Neanderthal segments by using different outgroups in our filtering step, because Neanderthal ancestry is common to all non-African populations whereas Denisovan ancestry relatively more private to Melanesia [7,9]. For each individual we used two different sets of variants as outgroup. First, we used only variants found in Sub-Saharan African populations as an outgroup (A total of 324 individuals, where 292 individuals are from 1000 genomes and 32 individuals are from Simons diversity). This should remove variation in the common ancestor of Sub-Saharan Africans and the Papuans, retaining archaic variants of Neanderthal and Denisova origin as both are present in Papuans, but mainly absent in Africa [7,9]. We also used this filter when analyzing Eurasian populations. Second we remove variants found in all non Papuan populations (A total of 2751 individuals, where 2504 individuals are from 1000 genomes and 247 individuals are from Simons diversity), only retaining variants that are unique to Papuan populations. This should remove Neanderthal variants that are shared with other non-African populations [1] and also to some extent remove variants of Denisovan origin that are found in Asians and Native Americans [20,21]. Thus removing all variants from the 1000 Genomes Project should enrich for Denisovan segments, while the segments that are found when using Sub-Saharan Africans but not using all 1000 Genomes Project samples as outgroups should be enriched for Neanderthal segments. We estimated the optimal set of transition and emission parameters for each Papuan individual and found them to be largely consistent across the different datasets, see S4 Fig. The parameters were converted into estimates of Tadmix, a, TIngroup and TArchaic using an average recombination rate of 1. 2 ∙ 10−8 events per base pair per generation and an average mutation rate of 1. 25 ∙ 10−8 mutations per base pair per generation, see Fig 3A and 3B. A table containing all parameters from the model and the corresponding demographic parameters is included in S4 Dataset. We find that the mean coalescence time between Papuans and non-Papuan individuals is more recent (1,395–1,540 generations ago) than that between Papuans and Sub-Saharan Africans (1,953–2,293 generations ago) reflecting that Papuans are more closely related to other Non-Africans than to Africans. The mean coalescence time between Papuans and other non-Africans also provides an upper limit for Neanderthal introgression because it happened in their ancestral populations. Using only Sub-Saharan individuals as an outgroup we find the mean coalescence time between the archaic and outgroup to be between 29,404 and 33,944 generations ago. When using non-Papuans as an outgroup the estimate is between 25,268 and 30,352 generations ago. The lower estimate is likely due to some variants in the common ancestor of Denisovans and Neanderthals having been removed in the latter case. Using Sub-Saharan Africans as an outgroup we estimate the total admixture proportion of archaic sequence into Papuans to between 4. 1–4. 4% and the admixture proportion private to Papuans to be 1. 5–1. 8%. This means that approximately 2. 6% is shared with non-Papuans, (Fig 3A). From the transition parameters, we estimate that the admixture event with non-Africans happened 953–1,254 generations while the Papuan specific admixture event happened 888–1,191 generation ago. Both are likely underestimates as it was for the simulated data with missing data and varying recombination rate. Neanderthal admixture likely occurred closer to 2,000 generations ago after the out of Africa migration [7,22] with Denisovan admixture occurring after that. We used a threshold of 0. 8 posterior probability as it showed a good trade-off between precision and sensitivity on simulated data see Fig 2C and S3 Fig. By comparing to the Vindija Neanderthal [10] and Denisova [2] genomes we find that this cutoff removes around 65% of the segments that don’t share variants with any archaic reference genome that were found with a cutoff of 0. 5, while only removing 10. 4% of the total length of archaic segments, see S5 Fig. Short segments that do not share variants with any archaic reference genome may be enriched for false positive variant calls, or may be deeply coalescing modern human haplotypes, in addition to the possibility of containing material from an unknown archaic source. We note that all segments with very high confidence share variants with Neanderthal and/or Denisova references. When we use a cutoff of 0. 8 we find that 84% of the segments unique to Papuans (80% of the total sequence) shared more variants with the Denisova genome than with the Vindija Neanderthal, and that 78% the segments that are shared with other non-Africans (83% of the total sequence) shared more variants with the Vindija Neanderthal than the Denisova (Fig 3C). This is consistent with most archaic sequence unique to Papuans coming from Denisovans, and most shared archaic sequence coming from Neanderthals. However, segments that are unique to Papuans are longer on average (94. 2 kb) compared to those shared with other non-African populations (76. 9 kb), see Fig 3D. The differences in length distributions are not seen as clearly when using Sstar, Sprime or CRF, see S6 Fig. Moreover, the length distributions of archaic segments that are not unique to Papuans (putative Neanderthal segments) are more similar to those found in other non-African populations, see S7 Fig, consistent with a single Neanderthal admixture event. We compared our archaic segments to those previously reported using other methods [7,9, 13]. We find that our method recovers 67% of the archaic sequence found using CRF, 84. 9% of the archaic sequence found using Sprime and 74% of the archaic sequence found using Sstar. When comparing the detected segments to the archaic reference genomes our method finds more Denisova segments in Papuans than Neanderthal ones, unlike Sstar and CRF. Our method also detects a smaller amount of additional Denisova segments in East and South East Asians, (Table 1). A dataset with all inferred segments from Papuans and Simons Genome Diversity Project individuals can be found in S5 Dataset. Our method examines the number of private variants in the ingroup compared to the outgroup and implicitly, the distance between the ingroup and the outgroup haplotypes, which interestingly are the features found to carry the highest weights in the new method of Durvasula et al. [14]. Our implementation also allows for missing data, unlike Sprime and Sstar, which may potentially be useful for analysis of ancient DNA samples. Since emission probabilities are very different between the human and archaic states in our model, we expect a low rate of false positive archaic inference, and this is also what we see in simulations. However, since recombination rates are highly variable, we expect many very short archaic segments and these have a high false negative rate. Our inability to identify these causes us to underestimate the admixture time. Nevertheless, the model does recover the correct size distribution for longer segments (> 50 kb), (S2 Fig). The mean coalescence times of modern and archaic humans are reasonably well estimated in simulations. One issue of interest is that if there were additional super-archaic introgression into the sequenced Denisovan as previously proposed [1], this would cause the mean coalescence time in Denisovan introgressed segments to be greater than that for Neanderthal segments. We did not observe this, although we note that there may be confounding from a low level of Denisovan admixture also present in East Asians which form part of our contrast population, reducing the observed mean divergence. We report more Denisova segments than approaches relying on the Denisovan reference. This is possibly because our method does not rely on matching to the Altai Denisova sequence, which is believed to be considerably diverged from the source population for “Denisovan” admixture into Papuan ancestors, probably shortly after the split of Denisovans from Neanderthals [1]. Furthermore, because of this early split, many segments may be equally close to the Vindija Neanderthal and the sequenced Denisova sample, and we expect that a fraction of segments introgressed from the Denisovan are more closely related to Vindija and vice versa due to incomplete lineage sorting. We find no clear evidence for an introgression with a new archaic hominin in Papuans, but we do find segments that do not share variation with any of the sequenced archaic populations. These segments could represent variation in Neanderthals and Denisovans that is not captured by the three high coverage archaic reference genomes, or another source. In the future it will be interesting to compare these segments to other human populations that might also have archaic segments of unknown origin [11,12]. Our model is not restricted to being applied to humans. We have called the admixture source “archaic” so far, which is standard in the human context, but more generally we are modelling a particular form of population structure involving an admixture event from a distantly diverged lineage. We note that other types of population structure, for example involving continual gene flow, could also create signals under our model. Subject to this caveat, the method can be applied where samples have been sequenced from a population that is hypothesized to have received admixture from a perhaps unknown source, and there is comparable data from an outgroup population that did not receive the admixture. The performance of the method depends on the ratio of signal to noise. The signal is stronger the more admixture there is, the more divergent the admixture source is, and the more recently the admixture happened. The noise increases if the outgroup diverged longer ago from the test samples, and if the common ancestor of the ingroup and outgroup had larger population size. The latter problem can be mitigated by sequencing more individuals from the outgroup. Therefore, as an increasing number of individuals are being sequenced in other species, our method could be used to explore introgression in those species, for example chimp and bonobo [23], bears [24], elephants [25] or gibbons [26]. To simulate data we used Msprime [27]. We simulated 5 Papuans and as an outgroup we simulated 500 Africans, 100 Europeans and 100 Asians using demographic parameters from [19]. We simulated data where we varied the recombination rate according to HapMap recombination maps [16] for 5 individuals and removed variants within non-callable regions and variants that were found in the simulated outgroup. We grouped all autosomes into bins of 1000 base pairs and counted the number of variants. For each 1000 bp window we calculated the number of called bases using the repeat masked segments. We simulated 22 autosomes with varying mutation rate in segments with a mean length of 1 Mb across the genome. The mutation rate in a segment could either be 1. 25*10^-8 mutations per base-pair per generation–the average rate, 50% decrease of the average rate or 50% increase of the average rate. We picked the mutation rate in each segment randomly. The choice of segment length and mutation rate is based on [28]. We trained and decoded the segments using our HMM, which is available at: https: //github. com/LauritsSkov/Introgression-detection/ We used 14 Papuans, 71 WestEurasians, 72 East Asians and 39 South Asians individuals from the Simons Genome Diversity Project (SGDP) [18], 40 Papuans from [19] and an additional 35 Papuans [9]. We used two sets of outgroups. One is all Sub-Saharan Africans (populations: YRI, MSL, ESN) from the 1000 Genomes Project [29] and all Sub-Saharan African populations from SGDP [18] except Masai, Somali, Sharawi and Mozabite, which show signs of out-of-Africa admixture. The other outgroup is all individuals from the 1000 Genomes Project [29] plus all non-Papuans from SGDP. For all human data sets, we also removed sites that fell within repeatmasked [15] regions, and sites that were not in the strict callability mask for the 1000 Genomes Project. hgdownload. cse. ucsc. edu/goldenpath/hg19/bigZips/chromFaMasked. tar. gz Strict callability mask for 1000 genomes: ftp: //ftp. 1000genomes. ebi. ac. uk/vol1/ftp/release/20130502/supporting/accessible_genome_masks/StrictMask/ The background mutation rate was calculated using the density of all variants from populations YRI, LWK, GWD, MSL and ESN in windows of 100 Kb divided by the mean variant density of the whole genome. We called Neanderthal and Denisova segments in the 14 Papuans and compared them to the segments called with CRF with more than 50 posterior probability [7] available at: https: //sriramlab. cass. idre. ucla. edu/public/sankararaman. curbio. 2016/. The path to the haplotypes is: summaries/2/denisova/oceania/summaries/haplotypes/CRHOM. thresh-50. length-0. 00. haplotypes. We called Neanderthal and Denisova segments in the 35 Papuans and compared them to the segments called with Sstar with more than 99 posterior probability [9] available at: https: //drive. google. com/drive/folders/0B9Pc7_zItMCVWUp6bWtXc2xJVkk The path to the haplotypes is: introgressed_haplotypes/LL. callsetPNG. mr_0. 99. den_calls_by_hap. bed. merged. by_chr. bed We called Neanderthal and Denisova segments in the 14 Papuans and compared them to the segments called with Sprime with a score greater than 150,000 [13]. The path to the data is: https: //data. mendeley. com/datasets/y7hyt83vxr/1

The genetic history of present-day individuals includes episodes of mating between divergent groups, which have led to' introgressed' genetic material persisting in modern genome sequences. Perhaps the most notable examples of such events in humans are the introgressions from Neanderthals into non-Africans 50,000 or so years ago, and from a related archaic group known as Denisovans into the ancestors of indigenous people from Papua-New Guinea and Australia. Methods to identify introgressions and the genomic regions that derive from them generally involve the use of reference genome sequences for the source populations. However, there are advantages in having methods independent of reference sequences, both to reduce bias and to detect possible introgression from groups for which we currently lack a reference genome. In this paper we describe such an approach, in a statistical framework which exploits the fact that introgressed regions will contain a high density of genetic variants that are private to the group receiving the divergent material. We apply this method to 89 Papuan genome sequences, estimating times of introgression and initial divergence between archaic and modern humans, and compare it to other related methods.

Abstract Introduction Results Discussion Materials and methods

markov models computational biology social sciences anthropology genetic mapping simulation and modeling neanderthals mathematics paleontology population biology paleoanthropology research and analysis methods population density hidden markov models hominids introgression comparative genomics probability theory hominins population metrics haplotypes physical anthropology heredity earth sciences genetics biology and life sciences physical sciences genomics evolutionary biology evolutionary processes

2018

Detecting archaic introgression using an unadmixed outgroup

The widespread emergence of resistance to insecticides used to control adult Aedes mosquitoes has made traditional control strategies inadequate for the reduction of various vector populations. Therefore, complementary vector control methods, such as the Sterile Insect Technique, are needed to enhance existing efforts. The technique relies on the rearing and release of large numbers of sterile males, and the development of efficient and standardized mass-rearing procedures and tools is essential for its application against medically important mosquitoes. In the effort to reduce the cost of the rearing process, a prototype low-cost plexiglass mass-rearing cage has been developed and tested for egg production and egg hatch rate in comparison to the current Food and Agriculture Organization/International Atomic Energy Agency (FAO/IAEA) stainless-steel cage. Additionally, an adult-index was validated and used as a proxy to estimate the mosquito survival rates by counting the number of male and female mosquitoes that were resting within each of the 6 squares at a given point of time each day in the cage. The study has shown that the prototype mass-rearing cage is cheap and is as efficient as the FAO/IAEA stainless-steel cage in terms of egg production, with even better overall egg hatch rate. The mean numbers of eggs per cage, after seven cycles of blood feeding and egg collection, were 969,789 ± 138,101 and 779,970 ± 123,042, corresponding to 81 ± 11 and 65 ± 10 eggs per female over her lifespan, in the prototype and the stainless-steel-mass-rearing cages, respectively. The longevity of adult male and female mosquitoes was not affected by cage type and, the adult-index could be considered as an appropriate proxy for survival. Moreover, the mass-rearing cage prototype is easy to handle and transport and improves economic and logistic efficiency. The low-cost mass-rearing prototype cage can be recommended to produce Ae. aegypti in the context of rear and release techniques. The proposed adult-index can be used as a quick proxy of mosquito survival rates in mass-rearing settings. Aedes aegypti (Linnaeus) is a highly invasive, medically important mosquito species which has received a considerable increase in attention after being linked to the Zika outbreak in Brazil in 2015 [1]. Together with Aedes albopictus (Skuse), the species also transmits several arboviral diseases including dengue, chikungunya and yellow fever [2]. Dengue viruses alone are estimated to infect 390 million people per year, including 96 million cases with clinical manifestations [3]. The heavy reliance on insecticides to control adult Aedes mosquitoes, especially during disease outbreaks, has led to the emergence of widespread resistance to the limited chemical classes that are currently available, and complementary vector control methods to enhance existing efforts are needed [4]. Amongst those being advocated is the sterile insect technique (SIT), a species-specific and environmentally-friendly pest population control method. According to the International Standards for Phytosanitary Measures No. 5 Glossary of phytosanitary terms, the SIT is a “Method of pest control using area-wide inundative releases of sterile insects to reduce reproduction in a field population of the same species”. The potential of the SIT for mosquito suppression has been demonstrated in a feasibility study in Italy [5], but successful implementation of the technique will rely on maintaining continuous production and repeated release of over-flooding numbers of sterile males [6] that can outcompete their wild counterparts within the target area [7]. To meet these requirements, novel methods and materials for the mass-rearing of mosquitoes are needed [8]. Optimization of mass-rearing conditions for both immature stages (for pupal production) and adult egg production requires a balance between accommodating the biological needs of the species, and achieving high production rates and economic efficiency [9]. For instance, the quality of larval diet [10–12] impacts pupal size and thus female adult size, which in turn determines bloodmeal intake and egg production [13,14]. The fine-tuning of the rearing cycle also requires adult cages to support as close-to-natural as possible behaviour of the mosquitoes. The ideal cage would also be practical in terms of handling, maintenance and space requirements, whilst keeping costs low. Cage design and construction requires careful adjustment of its dimensions in relation to the number of mosquitoes it should hold, providing suitable conditions for mating, feeding, and good survivorship [8,15]. This factor, which we call the Density-Resting Surface per mosquito (adult/cm2), or DRS, is obtained by dividing the number of adult mosquitoes by the vertical resting surface area on the four sides of a cage [15]. To date, only limited mosquito-rearing methods are available [9,15–19]. Adult mass-rearing cages made of stainless-steel, resistant to corrosion and allowing high temperature cleaning, have been developed for Ae. albopictus [9,19] but these have not been fully validated for Ae. aegypti. Preliminary results for egg production in Ae. aegypti under routine mass-rearing conditions have shown the capacity of the Food and Agriculture Organization/International Atomic Energy Agency (FAO/IAEA) stainless-steel cage [9] in supporting egg production. However, the stainless-steel cage faces a high production cost (~€2,300/unit). Another serious drawback of the stainless-steel cage is the significant loss of viable material resulting from eggs hatching prematurely within the cage after the first egg collection. In response to these major issues, a mass-rearing cage prototype has been developed to both reduce the purchase cost and to improve egg yield and adult survival. The final goal is to develop equipment for a mosquito mass-rearing facility, such as those found in established fruit-fly production facilities [7]. This study presents a novel mass-rearing cage and aims to test the cage in terms of egg production and hatch rate in comparison with the FAO/IAEA stainless-steel cage. Additionally, an adult index was validated and used as a quick and reliable proxy to estimate survival rates. Studies on mosquito species do not require a specific permit according to the document 2010/63/EU of the European Parliament and the Council on the protection of animals used for scientific purposes. All mosquito strains used in the present study were maintained in the biosecure insectaries of the Joint FAO/IAEA Insect Pest Control Laboratory (IPCL), Seibersdorf, Austria. All the experiments were performed based on standard operating procedures in the IPCL (FAO/IAEA, 2018). The blood used for routine blood-feeding of mosquitoes was collected in Vienna, Austria during routine slaughtering of pigs or cows in a nationally authorized abattoir (Rupert Seethaler, Himberg bei Wien) at the highest possible standards strictly following EU laws and regulations. The Ae. aegypti strain used in this study originated from field collections in Juazeiro (Bahia), Brazil and were transferred to the ICPL from the insectary of Biofabrica Moscamed, Juazeiro, Brazil in 2016. The strain is maintained following the “Guidelines for Routine Colony Maintenance of Aedes mosquitoes” [20]. Immature stages were reared under controlled temperature, humidity and lighting conditions (T = 28± 2°C, 80± 10 RH%, and 12: 12 h light: dark, including 1h dawn and 1h dusk) whereas adults were maintained in a separate room at 26± 2°C, 60± 10 RH%, 12: 12h light: dark, including 1h dawn and 1h dusk. All pupae were produced with the FAO/IAEA free-standing larval mass-rearing trays (L × W × H = 100 × 60 × 3 cm, Glimberger Kunststoffe Ges. m. b. H. , Austria) [21]. Enough eggs to produce 18,000 larvae (L1) were added to 5L of reverse osmosis purified water per tray (5,000 cm2 inner surface of the tray) to give a rearing density of 3. 6 larvae/cm2. To estimate the number of eggs required, 2-week-old eggs were brushed off egg collection papers and 3 samples of 100–150 eggs were hatched overnight using a 50mL centrifuge tube (VWR, UK) filled with 40mL of boiled and cooled reverse osmosis purified water with 2mL of 4% larval FAO/IAEA diet [22]. On the following day (about 20 hours later), the egg hatch rate of 100 eggs from each sample was verified under a stereomicroscope, and used to determine how many eggs would be needed to obtain 18,000L1/tray. The required egg numbers were estimated using an equation (Weight (mg) = (0. 0088 × Number of counted eggs) - 0. 3324) described by Zheng et al. [23]. The egg batches to produce 18,000 L1 for each tray were hatched separately. To hatch eggs, jam jars (IKEA of Sweden AB SE-343 81 Almhult, Germany) filled with 700mL of boiled and subsequently cooled (deoxygenated) reverse osmosis purified water stored at laboratory temperature were opened to add the eggs, before the jars were quickly closed again to avoid re-oxygenation. A volume of 10 mL (0. 022 mg/larva) of 4% larval FAO/IAEA diet [22] was added to jars to synchronize egg hatching and improve larval development. All jars were kept for 20 hours before their contents were sieved and transferred to mass-rearing trays previously filled with 5L of osmosis water and covered (1 day before being seeded with larvae). A volume of 50 mL of larval food was added to each tray on day 1,100 mL on day 2,200 mL on days 3 and 4,150 mL on day 5 and 50 mL on each additional day until day 9, corresponding to 0. 11,0. 22,0. 44,0. 33 and 0. 11 mg of ingredients per larva per day respectively [22]. Trays were individually tilted on day 6 after larval seeding, and larvae, male, and female pupae were separated using mechanical sexing tools [24] between 9am and 1pm. Larvae were returned to the rearing trays refilled with the retained rearing water. Batches of 500 pupae of each sex were counted manually and their volume estimated using a small cylindrical plastic tube (15 mL centrifuge tube, VWR, UK) covered with mesh on one side to hold pupae. The level of 500 pupae was daily marked on the tube and thus allowed estimation of the number of pupae. An adult-index was developed by drawing three 5. 7 × 5. 7 cm squares in the middle of the netted sides of a 30 × 30 × 30 cm Bugdorm cage (BugDorm-1 Insect Rearing Cage, Taiwan) using a fine marker. Each cage was loaded with approximately 3,000 Ae. aegypti (Brazil strain) mosquitoes (female: male ratio of 3: 1 with a DRS of 0. 8 mosquito/cm2). Mosquitoes were offered a 10% sugar solution using a 100mL urine cup holding a cellulose sponge cloth (SKILCRAFT, MR 580, USA). No blood feeding was performed during this experiment. A count of the number of male and female mosquitoes that were resting within each of the squares at a given point of time each day was used to estimate the density of adults in the cage, a measure named the ‘adult-index’. Dead mosquitoes were removed from the cages using a mouth aspirator and were counted and recorded according to sex for each of the 5 cages for a period of 4 weeks (except during the weekends). The adult mosquitoes that survived the 4 week-period were also removed, counted and sex-separated allowing the initial number of mosquitoes per cage at the beginning of the experiment to be calculated. Three and 4 replicates were performed for the stainless-steel cage and the mass-rearing cage prototype, respectively. Each cage was loaded with around 13,333 female and 4,444 male pupae (female: male ratio of 3: 1 with a DRS of 0. 8 mosquito/cm2) equally distributed over 4 consecutive days (Table 2) allowing the total final stocking of the cage with 12,000 and 4,000 adult females and males, respectively. This figure is reached using a pre-determined estimate 90% of adult emergence and survival rate until the first blood feeding. For the mass-rearing cage prototype, pupae were loaded into the black containers before these were fixed onto the bottom of the cage. Subsequent loading was made possible by sliding the containers backwards and adding pupae. Pupae were split equally into the two containers to avoid crowding and excess mortality. Pupae were introduced into the stainless-steel cage through an inlet valve to load the bottom part of the cage as previously described by Balestrino et al. [9]. Female mosquitoes were offered bloodmeals using two sausage casings (Grade Specification: 3) 26 NC, EDICAS co ltd) each filled with 150 mL of fresh porcine blood [9]. The blood sausages were heated in a water bath (42°C) for 10 min before inserting them into the 17 cm long mesh socks hanging from the top of the cage. The females were fed for 30 min (blood temperature ranged between 25 and 37° C during feeding) before the same process was repeated to allow females to feed on warm blood for a total of 2 hrs in each cage type. Blood feeding was performed 7 times for each cage to give an active period of 32 days (Table 2). Eggs were collected from each cage using seed germination papers (Grade 6, Size: 580 × 580mm, Weight: 145 g/m2, Sartorius Stedium Biotech). For the mass-rearing cage prototype, two egg papers (L × H = 270 × 60 mm) were inserted inside each container (4 egg papers in total) filled with 1L reverse osmosis purified water. To allow access to the same egging surface, 4 egg papers of L × H = 270 × 110 mm size each were inserted at the bottom of the FAO/IAEA stainless-steel cage, which was filled with 2L osmosis water, through four 400 mm-long and 3 mm-wide slots, two on each of the long sides as described by Balestrino et al. [9]. The first egg papers were added on the third day following blood meals and were removed three days later (Table 2). Further tasks were performed following the schedule as described in Table 2. When egg papers were collected from each cage, the remaining water was drained through stacked 300 and 50μm sieves (Retsch, Haan, Germany) to collect floating eggs, which were then dried on coffee filter papers (Melitta 1×4 Original FSC C095206) [25]. The eggs were allowed to mature and dry for 14 days [20] before the eggs were brushed off the papers and sieved through a 300 μm sieve (Retsch, Haan, Germany) to remove any debris. Seven egg batches were collected during the entire active period of each cage type. The mean number of eggs per female was estimated from the initial female pupae count and using the equation (Weight (mg) = (0. 0088 × Number of counted eggs) - 0. 3324) to estimate the number of eggs [23]. Egg hatch rate according to cage type and egg origin (either collected on papers or floating in water) was assessed following the protocol described above (Larval mass-rearing and pupae production for experiments). To estimate adult mosquito daily survival rates in the mass-rearing cage prototype in comparison to the FAO/IAEA stainless-steel cage, the adult-index data were used. Six 10 × 10 cm squares (three on each cage side) were drawn onto the netted sides of the cages using a fine marker (Fig 2). A daily cage density (termed adult-index) check was performed as described above at the same time every day for 4 weeks, except weekends. Counting adults in a given square took 10 to 45 seconds depending on the number of individuals, (longer at the beginning of the cage cycle). All statistical analyses were performed using R version 3. 5. 2 [26]. The correlation between adult-index and survival rates was estimated. The number of daily live mosquitoes in the Bugdorm cages was analyzed using a Gaussian linear mixed-effect model (lme4 package) with the number of live mosquitoes defined as the dependent variable the adult-index (number of mosquitoes counted by square), square location (as a qualitative factor) and their interaction as fixed effects and replicate as a random effect. The best model was selected and the correlation between the number of adult mosquitoes that survived daily in the cage and the adult-index was estimated. The number of adult mosquitoes of a given day equals to the initial number of caged adults minus the number of dead adult mosquitoes until that day. A Pearson' s product-moment correlation-test was used to analyze the correlation between the two values (number of live mosquitoes and the fitted model) for both female and male adult mosquitoes. To correlate the daily mortality rates estimated from counting dead mosquitoes daily and the daily mortality rates from the counts in the squares, a mixed effect Gaussian linear model was used, with the mortality estimated in a given square, the square location and the sex as fixed effects and replicates as a random effect. The egg production was analyzed using a Gaussian linear mixed-effect model (lme4 package) [27] with egg numbers per female defined as the dependent variable, cage type and week of egg collection (as a qualitative factor) and their interaction as fixed effects and replicates as a random effect. A generalized binomial linear mixed-effects model fit by maximum likelihood (Laplace Approximation) with logit link was performed with the proportion of floating eggs defined as the dependent variable and fixed and random effects same as above. The egg hatch rate was analyzed using a generalized binomial linear mixed-effects model fit by maximum likelihood (Laplace approximation) with logit link with the hatch rate defined as dependent variable, cage type, egg origin and their interactions as fixed effects and replicates as a random effect. A comparison of daily mortality rates between the FAO/IAEA stainless-steel cage and the mass-rearing cage prototype was performed using adult-index data. A constant mortality rate was estimated from plotting the logarithm of counts against time [28]. A Gaussian linear mixed-effects model fit by maximum likelihood was then used to analyze mortality rates, with cage type and square location as fixed effects and replicates as a random effect. The best model in all analyses was selected based on the lowest corrected Akaike information criterion (AICc), and the significance of fixed effects was tested using the likelihood ratio test [29,30]. All significant differences are based on p< 0. 05. In addition to the type of material used to manufacture the mass-rearing cage prototype, the main difference to the stainless-steel cage was the cleaning system. While in the prototype, containers can be removed and replaced for cleaning, the stainless-steel cage has inlet and outlet valves on the front side. The inlet valve allows the bottom of the cage to be filled with water for egg laying and the outlet valve allows dead pupae and adults to be removed (after adult emergence) and eggs floating in water (remaining in the cage after egg collection) to be collected. The mass-rearing cage prototype weighs about 5 kg compared to 16 kg for the stainless-steel cage and thus the cost of transportation is significantly reduced. Compared to the prototype (~230€ of total cost), the manufacture of the stainless-steel cage only counts for more than 90% of its total cost (2,300€). The mass-rearing cage prototype produced a comparable number of eggs per initial female to the FAO/IAEA stainless-steel cage (Table 3, Fig 3, p = 0. 79). More eggs were collected in both cage types on the second week of egg collection compared to the first and the third weeks (p = 0. 02). The mean numbers (± S. E.) of eggs per cage after seven blood feedings/egg batches were 969,789 ± 138,101 and 779,970 ± 123,042, corresponding to 81 ± 11 and 65 ± 10 eggs per female throughout each cage’s active period (32 days), for the prototype and the stainless-steel cage, respectively. However, the proportion of floating eggs in the stainless-steel cage was significantly higher (average 46% of the total egg production) than that of the mass-rearing cage prototype (average 41%) (Table 4, p<0. 001, Fig 4). The proportion of floating eggs collected from the stainless-steel cage was significantly lower during the second week (Table 4, p<0. 001) but greater during the third week (Table 4, p<0. 001) than that of the mass-rearing cage prototype during the first week. The egg hatch rate was significantly affected by cage type and egg origin. The mass-rearing cage prototype’s egg hatch rate was higher than that of the stainless-steel cage (Table 5, p = 0. 04, Fig 5). The eggs collected on seed germination papers had a higher hatch rate as compared to eggs floating in water (Table 5, p<0. 001, Fig 5). Egg origin had a greater impact on egg hatch rate compared to cage type. The adult-index was validated and deemed to be a good proxy of mortality. There was a positive correlation between adult-index and daily live mosquitoes in the 30 × 30 × 30 cm cage both for females (Fig 6A, t = 27. 53, df = 298, cor = 0. 84, p<0. 0001) and males (Fig 6B, t = 14. 67, df = 298, cor = 0. 64, p<0. 0001). Moreover, the best model to predict the mortality from the counts was the one using the mortality estimated from the index only (the effects of the square number and the sex did not improve the model). The predicted mortality was correlated to the observed value (cor = 0. 59, t = 3. 89, df = 28, p<0. 0001). Using this index, no significant difference was observed between the stainless-steel and the mass-rearing prototype cages (Table 6, p = 0. 53). However, females lived longer than males in both cages (Table 6, p = 0. 0029). Although the square location impacted the measured mortality (the best model included this factor and its interaction with the cage), no significant effect of the location of the squares was observed (Table 6, p˃0. 05). The ability to colonize and mass rear an insect species in adequate numbers and with relative efficiency are prerequisites for considering an SIT program for that species [31]. We present here a low-cost mass-rearing cage prototype, developed with consideration of the biology of Ae. aegypti, and to the results of a comparative test with the FAO/IAEA stainless-steel cage, measuring egg production and egg hatch rate. In addition, a new adult-index was validated as a proxy for adult density in a cage, and used to estimate mosquito survival rates. We have shown that the mass-rearing cage prototype is much cheaper (ten times) and as efficient as the FAO/IAEA stainless-steel cage in terms of adult mosquito survival and egg production, and gives a better egg hatch rate. In addition, the mass-rearing cage prototype is easy to handle, clean, and to transport. In general, increasing the size of adult cages reduces the total time needed to handle a given number of adults and therefore improves cost-time efficiency. For large scale production a large number of cages would be needed, and so the initial investment in a facility would be high, and developing a low-cost cage allows considerable savings. Furthermore, the proposed mass-rearing cage prototype is lighter (~5 kg compared to the 16Kg for the stainless-steel cage) and enables easy handling, installation and cleaning, reducing the workload and number of staff needed. Based on the DRS of 0. 8 considered in our study, the mass-rearing cage prototype would hold as many adults as six 30 × 30 × 30 cm standard rearing cages (BugDorm-1 Insect Rearing Cage, Taiwan) which cost €60. 54/unit (https: //shop. bugdorm. com/insect-rearing-bugdorm-1-insect-cages-c-5_25. html) and would mean routine operations such as blood feeding and egg collection would need to be performed on one cage instead of six. Moreover, no internal handling is required once the cage is fully stocked with pupae, reducing the number of escapees and therefore protecting staff from bites. Although we did not monitor the exact number of escapees, they seemed to be similarly low for both cages. The middle section of the top of the mass-rearing cage prototype also has an opening (180 mm of diameter covered with netting) allowing the use of the Hemotek device (Discovery Workshops, Accrington, UK; [12]), as an alternative to the sausage casings. A study carried out in Brazil has shown that 28 cages of 30 × 30 × 30 cm were needed to produce 4 million eggs per week, [16]. For the same production with two blood feedings per week, only 7 cages of the prototype would be needed assuming that same number of eggs per female would be produced. For a medium scale mass-rearing facility with a production level of 10 million sterile males per week, about 200 cages would be needed. Considering the difference in cost, the mass-rearing cage prototype manufactured locally can reduce initial investments for equipment by more than €350,000 compared to the stainless-steel cage in a facility of this scale. Other materials such as recycled plastic, fiber glass or aluminium could be used as alternatives to further reduce the costs compared to stainless-steel. Our results did not show any difference in egg production between the cages. However, more eggs were collected during the second week of egg collection. A possible explanation of the delay in egg laying could be female mosquito age, ranging between 3 and 6 days when the first blood feeding occurs. This may lead to differential female blood feeding rates or blood intake volumes, and consequently egg yield, although early studies have reported that both Ae. aegypti and Ae. albopictus females take their first bloodmeal from hosts on the second or third day after emergence [32]. In any case, it was found that more females produced eggs in the second week. We recommend reducing pupal loading to one event on the same day to homogenize adult age, blood feeding rates and increase overall blood volume intake, thereby increasing the total egg production in the first two weeks. This would corroborate the results of Zhang et al. [19] who have recommended a two week-cycle for Ae. albopictus rearing cages in a medium scale rearing facility. Cage structure (height and width) has been shown to play a great role in Ae. albopictus egg production according to their study. Another advantage of the mass-rearing cage prototype described in this study is that the height can be adjusted according to species, if needed. We plan to test this cage for mass-rearing Ae. albopictus in the near future and to adapt its structure to accommodate the mass-rearing of Anopheles arabiensis, particularly by adapting the oviposition containers. Usually Aedes females deposit their eggs above the water level in mainly dark-coloured, man-made containers in and around cities [33,34], and so in rearing facilities eggs are collected on a moist substrate. In our study, the proportion of eggs collected from the water rather than on the wet paper provided found to be above 40% for both prototype and stainless-steel cages, though a greater proportion of floating eggs was found in the stainless-steel cage. Similar results have been observed in Ae. aegypti in the laboratory and in a semi-field setting where a high percentage of floating eggs was observed in the larval sites [35]. Another study described that in field traps, Ae. aegypti females oviposited 22. 7% of their eggs on the water surface [36]. The presence of floating eggs may be due to the phenomenon called skip oviposition behaviour [37], and that the limited number of oviposition papers per cage might play a role in the females’ choice of laying eggs on the water or the paper substrate. It has been shown that gravid female Ae. aegypti favour larval sites that already contain eggs or larvae [38]. Given that the stainless-steel cage has a continuous presence of larvae in the water, the gravid females may lay a larger proportion of their eggs on the water surface instead of the provided paper substrate compared to females in the mass-rearing cage prototype, where no larvae are present in the water. The mean egg hatch rate for both cages was above 85%, with eggs collected from the mass-rearing prototype cages showing a better overall hatch rate after egg storage for two weeks after oviposition, contradicting an observation made by Zhang et al. [19] who observed that the cage structure had no impact on egg hatch rate in their Ae. albopictus strain. However, in our study, egg hatch rate was compared between the eggs found on oviposition papers and floating eggs. The latter showed a decrease in hatch rate when they were collected from the stainless-steel cage. This difference might be due to the pre-existing eggs and larvae remaining from the first egg collection in the stainless-steel cage. Huang et al. [39] observed that older (third and fourth instar) An. gambiae larvae readily eat An. gambiae eggs and first instar larvae. The quality of the water, and egg collection and storage methods affect the egg maturation process and therefore hatch rate. Another major disadvantage of the stainless-steel cage is that the floating eggs are difficult to recover. Although more than 4 L of water was used to rinse the bottom of the cage, a significant number of eggs remained in the cage and hatched prematurely. This causes a loss in yield that is uneconomical in a mass-production factory. Gomes et al. [40] conducted a laboratory experiment and indicated that eggs deposited on the water surface showed faster hatching which could favour faster establishment in wild habitats. Dead mosquitoes (adults, pupae that failed to emerge, and larvae resulting from premature hatching) in the stainless-steel cage could lead to bacterial growth resulting in a rapid decrease in oxygen levels in the water, stimulating egg hatch [41]. The mass-rearing cage prototype has the advantage that egg containers are removed and replaced so that no unwanted egg hatch occurs. Monitoring production and quality of mass-reared insects is important in any rearing facility. Although attention is given to the larval rearing stage to ensure the quality of resulting adults, it is also important to have procedures to monitor and follow the size of the egg producing colony. Miller and Weidhaas [42] suggested that the adult survival rate is the most important factor determining the stability of the population and total egg production. The adult-index method described and validated here could be used as a simple tool to monitor the survival of a cage population as an indicator for the overall status of the mass-rearing colony. A daily adult-index check of one mass-rearing cage would take less than 5 minutes depending on when in the rearing cycle it was. Provided that the index is calibrated for each given environment and strain, any unusual increase in mortality may be spotted based on a decrease in the daily recorded adult index. Other methods based on geometric morphometric analysis (wing size and shape) have been used as a criterion to the quality of mass-rearing procedures [43]. Although the prototype cage structure is different to that of the FAO/IAEA stainless-steel cage, male and female mosquitoes survived equally well in either cage. This could be due to the density of caged mosquitoes being equal and optimized for longevity [8,15]. A DRS value of 0. 9 has been used for the mass-production of Ae. aegypti in Brazil [16] whereas a DRS value of 0. 8 that was measured in our study. Male mosquitoes showed higher mortality rates compared to females, which is likely to be, among other factors, due to mating-related stress. It is well known that mating is costly in terms of energy [44,45] and a recent study pointed out that an increase in female survival could be due to male-derived seminal fluid molecules transferred to the female [46]. We have shown in this study that the proposed mass-rearing cage prototype has several advantages in comparison to the previous stainless-steel cage developed by the FAO/IAEA and that the adult-index is a quick and reliable proxy of mosquito survival rates which could be used in mass-rearing settings. Although an improved mass-rearing cage has been developed, further studies on blood feeding optimization are needed. In addition, egg collection can be optimized by increasing the number of egg papers to reduce the number of floating eggs. Moving egg collection containers to the side instead of along the bottom of the cage will also improve the design and further minimize escapees as well as work load. To better fit the cage mesh to the cage structure, another prototype with Hook-and-loop fasteners (VELCRO) is under investigation and will soon be tested. Alternative materials to lower the production cost of the new mass-rearing cage prototype could also be investigated for developing countries. The mass-rearing cage prototype design and drawings are available for open access on the IAEA website and may be used to locally produce the cages for use in SIT programmes or indeed large-scale rearing for any purpose.

Dengue, among other arboviral infections, is a neglected disease and a major health issue that is re-emerging in tropical countries due to the poor efficacy of conventional vector control methods. Therefore, there is a growing need for more sustainable techniques to control Aedes mosquito species, while reducing the dependence on insecticides. The sterile insect technique, which relies on the mass-production of sterile males, can be used as part of area- wide integrated pest management (AW-IPM) programmes to reduce the vector population below the disease transmission threshold. Therefore, innovations in mosquito mass-rearing techniques including the development of low-cost adult holding cages are essential in the quest to promote an economical and logistically efficient mass -rearing system for the vectors of dengue, chikungunya, yellow fever and Zika diseases.

Abstract Introduction Methods Results Discussion Conclusions

death rates invertebrates medicine and health sciences body fluids engineering and technology animals reproductive physiology developmental biology pupae population biology insect vectors prototypes infectious diseases aedes aegypti technology development life cycles disease vectors insects arthropoda population metrics mosquitoes eukaryota blood anatomy physiology oviposition biology and life sciences species interactions larvae organisms

2019

Reducing the cost and assessing the performance of a novel adult mass-rearing cage for the dengue, chikungunya, yellow fever and Zika vector, Aedes aegypti (Linnaeus)

Testing one SNP at a time does not fully realise the potential of genome-wide association studies to identify multiple causal variants, which is a plausible scenario for many complex diseases. We show that simultaneous analysis of the entire set of SNPs from a genome-wide study to identify the subset that best predicts disease outcome is now feasible, thanks to developments in stochastic search methods. We used a Bayesian-inspired penalised maximum likelihood approach in which every SNP can be considered for additive, dominant, and recessive contributions to disease risk. Posterior mode estimates were obtained for regression coefficients that were each assigned a prior with a sharp mode at zero. A non-zero coefficient estimate was interpreted as corresponding to a significant SNP. We investigated two prior distributions and show that the normal-exponential-gamma prior leads to improved SNP selection in comparison with single-SNP tests. We also derived an explicit approximation for type-I error that avoids the need to use permutation procedures. As well as genome-wide analyses, our method is well-suited to fine mapping with very dense SNP sets obtained from re-sequencing and/or imputation. It can accommodate quantitative as well as case-control phenotypes, covariate adjustment, and can be extended to search for interactions. Here, we demonstrate the power and empirical type-I error of our approach using simulated case-control data sets of up to 500 K SNPs, a real genome-wide data set of 300 K SNPs, and a sequence-based dataset, each of which can be analysed in a few hours on a desktop workstation. The ideal analysis of a genome-wide association (GWA) study for a complex disease would involve analysing all the SNP genotypes simultaneously to find a set of SNPs most associated with disease risk. Such an analysis can improve performance over single-SNP tests, since a weak effect may be more apparent when other causal effects are already accounted for, but also because a false signal may be weakened by inclusion in the model of a stronger signal from a true causal association. To date, analysing all SNPs simultaneously has seemed infeasible, since current GWA platforms can type over one million SNPs, and even larger variable sets may not be far away as genome-wide re-sequencing advances. We exploit recent advances in stochastic search algorithms [1], [2] to develop a computationally efficient tool to simultaneously analyse k SNPs typed in n individuals for association with case-control status, where k » n. We formulate the problem as variable selection in a logistic regression analysis that includes a covariate for each SNP. Our aim is to find a subset of SNPs (a “model”) that best explains the case-control status subject to a specified error rate. The number of possible models is 2k and since k is typically of the order of 106, classical methods such as forward-backward variable selection are computationally expensive and are liable to find sub-optimal modes [3]. Bayesian stochastic search methods have been used to tackle variable selection problems, typically using the “slab and spike” prior formulation [4]. Inference can be made from these models using Markov chain Monte Carlo (MCMC) [5], [6] and this methodology has been extended to the case of more variables than observations [7], [8]. Similar methods have been proposed for the analysis of SNP data [9], [10], which again utilised MCMC. However, despite design of the MCMC algorithms to minimise computational time, these methods are too slow to deal with the size of problem presented by modern SNP chips. Furthermore, these methods have dealt with the computationally easier problem of a continuous outcome. We assign continuous prior distributions with a sharp mode at zero, often referred to as “shrinkag” priors, to the regression coefficients. Our approach is Bayesian-inspired rather than fully Bayesian, since we seek only the posterior mode (s) rather than the full posterior distribution of the regression coefficients. If the signal of association at a SNP is weak or non-existent, the posterior mode for the corresponding covariate will remain at zero. By using continuous shrinkage priors the resulting posterior density is continuous and can thus be maximised using standard algorithms. Our stochastic search maximisation algorithm seeks the (small) subset of SNPs for which the posterior mode is non-zero, corresponding to a signal of association that is strong enough to overcome the prior preference for zero effect. The algorithm can be set to include only additive effects, or it can also consider dominant and recessive terms: only one of these terms is permitted to be non-zero. We consider two shrinkage prior distributions, the Laplace, or double exponential distribution (DE) and a generalisation of it, the normal exponential gamma distribution (NEG), which has a sharper peak at zero and heavier tails, Figure 1. The sharp peak of the NEG at zero favours sparse solutions which is preferable for variable selection when we believe that there are few true causal variables. Further, the heavy tails result in variables being minimally shrunk once included in the model. The NEG is characterised by a shape and a scale parameter. The smaller the shape parameter the heavier the tails of the distribution and the more peaked at zero. Conversely as the shape parameter increases the NEG approaches the DE. For both prior distributions we obtain an explicit expression for the approximate type-I error of our method, so that it can be calibrated without recourse to permutation techniques. With both the NEG and DE prior for n<k, the posterior density can be multi-modal [11] and the mode identified by each run of our algorithm depends on the initial values of the parameters and the order in which they are updated. We implement multiple runs of the algorithm, always starting with all regression coefficients equal to zero but permuting the order in which they are updated, and report the highest of the modes identified. Our checks using more extensive searches in test datasets indicate that typically the largest mode identified in our search corresponds to a model that is very similar to the global optimum model, differing for example in which of two highly-correlated SNPs is included. We demonstrate this in our analysis of a real GWA study and show how the multiple modes found can be utilised to infer a group of SNPs that identify the same signal. As a consequence of modelling all SNPs simultaneously, a SNP will only be included in the model if it significantly improves prediction of case-control status beyond that obtained from the SNPs already included. Thus a SNP with strong marginal effect can be overlooked by our analysis if other SNPs better explain most of its effect. We typically find that our analysis returns only the best SNP characterising the effect of a single detectable causal variant, and when multiple SNPs in close proximity are selected this is an indication of multiple distinct causal variants. Thus, the number of SNPs in the best fitting model gives an estimate of the number of causal variants. This feature of the method also makes it suitable for fine mapping using dense SNP sets, such as those that can arise from imputation methods or re-sequencing, in contrast with single-SNP analyses in which many tightly-linked SNPs may show signs of association, leaving open the problem of locus refinement. Haplotype and interaction effects could be readily implemented using our approach, but these would substantially increase the size of the model space to be explored for genome-wide datasets and we have not pursued these possibilities here. Our software deals with quantitative phenotypes, but here we focus on main effect terms for case-control phenotypes with up to half a million SNPs, and demonstrate that our method improves on single-SNP analyses in terms of false-positive rate, power and interpretability. Our main simulation study used the FREGENE software [12] to simulate 20 Mb of sequence data in a population of 10 K individuals with mutation, cross-over and gene-conversion rates similar to those in humans [13]. From this population we sampled 500 case-control data sets each with six causal variants and 1,000 cases and 1,000 controls. For each simulation we added a further nineteen 20 Mb chromosomes devoid of causal variants. Thus, in effect we analysed 400 Mb genomes consisting of twenty equal-length chromosomes, with all the causal variants concentrated on one chromosome. Marker SNPs were sampled to give an approximately uniform minor allele frequency (MAF) distribution with SNPs spaced on average every 5 Kb, giving 80 K SNPs per data set. The selection ignored causal status, so that the marker SNPs usually included few if any of the causal SNPs. The above data sets were analysed using (i) our algorithm with an NEG shrinkage prior, (ii) our algorithm with a DE shrinkage prior, and (iii) the Armitage trend test (ATT). When using the NEG and DE we standardised the genotype data to have mean zero and variance one. The ATT is the natural univariate comparison for our multivariate method being a score test for a regression coefficient in a logistic regression model [14] and we show in Text S1 that our search procedure, when applied univariately to standardised data, is asymptotically equivalent to the ATT. Detailed analyses on a subset of the data identified 0. 05 as the most suitable value of the NEG shape parameter λ for the selection of truly causal variants; smaller values gave rise to computational problems. With λ = 0. 05 the heavy tails of the prior density (Figure 1), reflect little prior knowledge of effect sizes. However, standardising the genotypes has the effect of incorporating a prior belief that effect sizes may be larger at alleles with smaller MAF [15]. The per-SNP type-I error rate was set at α = 10−5 for all three analyses; see Methods for setting the DE and NEG parameters to achieve this. The results for the NEG and DE were based on the highest posterior mode found from 100 permutations of the search order using the optimisation algorithm described in the Methods. The definition of true and false positives can be problematic when, as here, the causal variant is typically not included among the SNPs analysed. In practice, when a significant result is obtained all SNPs in strong linkage disequilibrium (LD) with it are considered as potential sources of the positive signal. For unlinked SNPs with uniform MAF distribution and typed in 2 K individuals, the upper 99. 9% percentile of the null distribution of r2 is about 0. 005, and so most SNPs having r2>0. 005 with a causal variant are in some sense true associations. However, due to the variable pattern of LD in the human genome, SNPs showing 0. 005<r2<0. 05 with a causal variant can sometimes be hundreds of Kb distant from it, and are likely to be difficult to replicate. Therefore we chose to classify a positive signal as “true” if it has r2>0. 05 with any causal SNP. Furthermore, a cluster of tightly linked false positives might be considered as essentially just one false positive, and in Table 1 false positives were only counted if they were further than a specified distance (between 0 Kb and 100 K) away from any other false positive that had already been recorded. A feature of both the NEG and DE results is the reduction in the number of false positives relative to the ATT, despite the fact that the type-I error rate was set to be the same for all analyses (Table 1). This reflects one of the principal effects of analysing SNPs simultaneously: the signal at a SNP that shows spurious association when analysed singly is often weakened by inclusion in the model of true positives, which may or may not be tightly linked with it. If several SNPs are mutually in high LD, typically at most one of them will be included in the model. Thus the NEG analysis picked 2,097 SNPs over the 500 data sets, many fewer than the 6,810 SNPs selected using the ATT (Table 1). If a causal SNP is detected by the NEG method then typically (87% of the time) it will be tagged by just one selected SNP (Figure 2). In contrast ATT often picks multiple SNPs for each causal variant and picks one SNP just 31% of the time. The higher number of false positives for ATT can be attributed in part to the way the ATT picks many more SNPs per causal variant than the NEG or DE; some of these are remote from, and in low LD with, the nearest causal variant, and fail to reach our threshold of r2>0. 05 for useful tagging. However, the fact that ATT selects many more SNPs than DE or NEG can spuriously inflate its true power, because some of these additional significant SNPs will by chance be in LD with one or more causal variants. In the case of several causal variants, the ATT may produce what appears to be a single signal. Of the 3,000 causal SNPs in the simulations, 1,402 are detected by both the NEG and ATT analyses, 54 are only detected by NEG and 32 are only detected by ATT. Although this difference in empirical power is small (0. 7%), a p-value for the null hypothesis that the NEG and ATT are equally powerful is 0. 011 (binomial probability of ≤32 successes, given 86 trials and success probability 0. 5). Moreover, the NEG empirical power equals or exceeds that of ATT for all six combinations of MAF and allelic risk ratio (Table 2). Thus we have evidence of improved power of NEG over ATT, in addition to its lower false positive rate. DE shows fewer false positives than NEG, but it tags fewer causal SNPs even though it selects 2,622 SNPs, many more than NEG (Table 1). The lighter tails of the DE distribution in comparison with the NEG result in informative variables being shrunk closer to zero. This can result in other correlated SNPs being brought into the model to explain the full effect of the over-shrunk SNP coefficients. Our preliminary analyses varying λ showed that as λ increased the false positive rate decreased but the power also decreased and in doing so approached the results obtained with the DE prior. With λ = 10 the results were very similar to the DE results. To validate our type-I error rate approximation, and to assess the effect on type-I error of allowing for dominant and recessive effects, we permuted the case-control status in one of the 80 K data sets to generate 1,000 data sets representing samples from the null of no genetic effects. The resulting per-SNP type-I error rates of the NEG, DE and ATT methods are shown in Table 3. False positives were only counted if they were further than 20 Kb away from any other false positive that had already been recorded, however the results were the same when the minimum separation was 100 Kb. The type-I error rate was highest for ATT although the differences are not significant. All three analyses result in noticeably fewer false positives than the nominal rate of α = 10−5. Because of LD between SNPs it is not easy to decide if this difference is significant, but it suggests that our type-I error approximation is as conservative as that of the ATT, which is based on the approximation. When dominant and recessive effects are considered in addition to additive terms, the false positive rate approximately doubles. Thus parameter settings that control the type-I error at 2. 5% for additive effects will approximately control the type-I error at 5% when dominant and recessive effects are also included. Recall that our simulations generated an approximately uniform MAF distribution of the marker SNPs, and this result may vary with different MAF distributions. We also generated a data set corresponding to a genome-wide association study consisting of 480 K SNPs, derived from 120 independent 20 Mb chromosomes using the same SNP ascertainment strategy as used in the previous simulation. We chose one 20 Mb chromosome to have ten causal variants each with MAF of 15% and allelic risk ratio of 2. This disease model is unrealistic but was chosen to permit detection of the majority of the causal variants and thus make general comments on the relative merits of the NEG and ATT in one simulation. We analysed this data set as before using the NEG and ATT but with a significance threshold of α = 5×10−7 [16]. As before, we chose the highest posterior mode from 100 permutations of the search order. Figure 3 (A) shows the locations along the 20 Mb chromosome of the ten causal variants, as well as all the SNPs selected by the NEG and ATT analyses. Both methods have detected all ten causal variants, however the NEG analysis selected just 14 SNPs, whereas the ATT identified 35 significant SNPs. We see in Figure 3B that the NEG analysis has improved localisation, it ignores SNPs remote from causal variants that were significant under the ATT, in particular at 8. 3 Mb, and has selected SNPs as close to the causal variant as the ATT has. In Figure 3C the improvement in localisation is less clear, the ATT has selected SNPs closer to causal variant, but at the expense of selecting many more SNPs. In particular the ATT has selected SNPs at 11. 5 Mb, about 400 Kb from the causal variant. Fine mapping of causal variants currently uses very high density markers, obtained either directly from resequencing or from imputation following limited resequencing or from high-density SNPs in public databases [17], [16], [18]. To illustrate the utility of our method for the analysis of imputed and/or sequence data we took the simulated 20 Mb sequences of 10 K individuals used in the previous analyses, which had 192 K polymorphic sites, and sampled 10 case-control datasets using the same sample sizes and disease model as used in the main simulation. All polymorphic sites were included in our analyses. The data sets were analysed with the NEG and ATT with a per-SNP false positive rate of α = 10−5. The ATT and NEG analyses showed similar power over the ten sequence-level datasets. Both methods detected 54 of the 60 causal variants with r2>0. 3, five causal SNPs were missed entirely by both methods and one causal SNP was tagged by both methods at r2≈0. 01. However, NEG showed markedly better localisation than ATT. Figure 4 shows the distribution of the highest r2 value for each selected SNP with a causal variant using the two methods. The NEG selected just 64 SNPs in comparison with 599 selected by the ATT, and a greater proportion of the selected SNPs were in high LD with a causal variant. Of the 60 causal variants, only nine were tagged twice by NEG, in contrast, it is evident from Figure 4 that the ATT often multiply tags causal SNPs. In no simulation did a SNP selected by NEG tag two causal SNPs. The NEG analysis identified two false positive SNPs at the less stringent r2 = 0. 01 threshold for tagging a causal SNP. The ATT analysis generated 14 false positives at this threshold; 11 of these SNPs spanned a 230 Kb region including one of the NEG false positives, while two spanned a 103 Kb region including the other NEG false positive. The 14th ATT false positive had r2 = 0. 009 with a causal variant. From a genome-wide scan on 694 type 2 diabetes cases and 654 controls [19], we reanalyse here genotype data from the Human Hap300 BeadArray, but not the Human Hap100 BeadArray. After removing SNPs with Hardy-Weinberg equilibrium p-value <10−3 or with call rate <0. 95, there were 300,535 SNPs analysed. In the original analysis [19], SNPs were tested for additive, dominant and recessive effects and 42 were significant (permutation p-value <5×10−5), tagging 32 distinct loci (defined here to denote a 1 Mb flanking region). These SNPs, together with 15 SNPs identified using the Human Hap100 BeadArray, were carried forward to a replication analysis using 2,617 cases and 2,894 controls [19] that confirmed eight SNPs tagging five loci. Our NEG reanalysis used λ = 0. 05, while γ was set such that α = 2. 5×10−5 if additive effects only were considered, thus approximately controlling the type-I error rate at 5×10−5 for our actual analysis which also considered dominant and recessive terms. The resulting best-fitting model included 26 SNPs, tagging 25 distinct loci including the five previously-replicated loci (Table 4). Four of our SNPs matched those previously reported while the fifth locus had been tagged by Human Hap100 BeadArray SNPs not included in our reanalysis, but instead we identified rs729287 only 20 Kb distant. We also looked at the seven other best-fitting modes with posterior density within a factor of 10 of the maximum. These modes included 29 unique SNPs. The three extra SNPs in this combined list not included in the best fitting mode were all within 50 Kb of SNPs included in the best fitting mode. In modes which included one of these extra SNPs, the SNP close by was not included. Thus, examining sup-optimal modes can identify alternate SNPs tagging the same causal variants, which can be useful to include in follow-up genotyping when some redundancy is beneficial, or to consider alternative possibilities for the SNP in strongest LD with the causal variant. The SNPs tagging the five previously replicated loci were included in all seven modes suggesting that these are the best tagging SNPs for the causal loci. These results are consistent with the conclusions from our simulation study: we captured the same significant loci as the single-SNP analysis but at the cost of many fewer false positives. In addition, our NEG analysis has picked one SNP from each of the replicated loci, suggesting that there is just one distinct causal variant in each locus tagged by the genotyped SNPs. Our NEG shrinkage-based algorithm provides a computationally-efficient tool for the simultaneous analysis of either genome-wide SNPs or resequencing or hyper-dense SNPs from large regions. The NEG analysis improves on the single-SNP ATT analysis, most notably in terms of false positives, and also in terms of power. It is also superior to the DE analysis in terms of of power, at the expense of a higher false positive rate. The NEG method typically selects one SNP for each causal variant and thus gives a measure of the number of underlying causal SNPs genotyped in the data set under study, as well as improving localisation in comparison with the ATT. The advantages of the NEG analysis are even greater for sequence or very-high density genotype data, such as can be obtained via imputation: it identifies a much smaller subset of SNPs without a reduction of power, and tags the causal variants with higher r2 on average. This reflects the natural advantage of a regression-based approach when causal variants are included in the analysis rather than merely tagged by markers. Significant SNPs from a GWA are usually genotyped in another sample. With cheaper genotyping experimenters may be able to afford to replicate more SNPs than the minimal set suggested by the NEG or DE methods. Candidates for redundant/alternative SNPs can be obtained by considering local modes found by our algorithm. A full Bayesian analysis such as that suggested in [20] would also be possible if limited to a subset of SNPs, and would explore the posterior distribution more completely for that SNP-set. Since we take a regression approach it would be straightforward to include other individual level covariates such as age and sex, as well as covariates to control for population stratification such as eigenvectors from a principal component analysis [21]. Our software can analyse quantitative traits and could be extended to search for haplotype or interaction effects. In the latter case the size of the model space would need to be reduced, perhaps by a strategy of seeking interactions of all SNPs with those SNPs showing marginally significant association. There is growing interest in predicting phenotypes from genotypes in both human genetics [22] and livestock genetics, in which there is interest in predicting breeding values [9]. Since our method is regression based and considers all SNPs simultaneously and will thus account for the LD between SNPs [23], it can also address this application; a weaker significance threshold is often considered appropriate for prediction rather than SNP selection. Software can be downloaded from http: //www. ebi. ac. uk/projects/BARGEN/. For each regression coefficient we assign independent shrinkage priors with a density that is sharply peaked at zero. The DE is a one-parameter distribution that is widely used as a shrinkage prior [1]. It can be represented as a scale mixture of a normal distribution: (1) where N (a, b) is the normal density with mean a and variance b and Ga (c, d) is the gamma density with shape parameter a and scale parameter b. The normal exponential gamma distribution [2] is a generalisation of the DE that has the following scale mixture representation: (2) where λ and γ can be interpreted as shape and scale parameters respectively, κ is the integrating constant and D is the parabolic cylinder function [24]. We can see from (1) and (2) that the NEG can be generated by sampling from a DE distribution with parameter drawn from a gamma distribution. There is a fast algorithm for computing D and its derivatives [25], and Fortran code is available from http: //jin. ece. uiuc. edu/routines/routines. html. As λ and γ both increase such that remains constant, the NEG converges to the DE distribution with parameter ξ. Figure 1 shows the log densities of the DE and three NEG distributions, all with the same density at zero. From the plot we see that as λ decreases the NEG density is steeper near zero and flatter elsewhere, thus shrinking non-zero coefficients less than the DE. For further details of the NEG see Text S1. We seek to maximise the posterior density p (β | x, y) over β = (β1, …, βk), where x = (x11, …, xnk) is the normalised genotype data and y = (y1, …, yn) denotes the case-control status coded as 1 for cases and −1 for controls. Taking logarithms in Bayes Theorem we can write (3) where L denotes the log-likelihood for the logistic regression model and f is minus the log-prior density. The negative sign is introduced to allow f to be interpreted as a penalty function, and so our estimation procedure can be thought of as maximising a penalised log-likelihood. With the DE prior, the maximisation of (3) is equivalent to the Lasso procedure [26]. The EM algorithm has been used for the analogous optimisation problem for linear regression [2] but we found it to converge slowly for binary regression. Instead we use the CLG algorithm [27] which optimises each variable in turn, making multiple passes over the variables until a convergence criterion is met. This algorithm has been implemented for the logistic regression model [1], but not previously with the NEG prior. There is no closed form solution for the univariate optimisation problem in logistic regression, but Newton' s method can be applied using the formula (4) where each ′ denotes a derivative with respect to βj, and (5) where. See Text S1 for justifications. We avoid taking large steps by replacing the L″ with an upper bound [1]. We also disallow any update proposed according to (4) that would change the sign of β: instead, is set to zero. When βj = 0 the algorithm attempts an update in both directions, taking limits as βj approaches zero from above and below, and accepting the move if is, respectively, positive or negative. Since the denominator in (4) is always negative at βj = 0, and f is symmetric about zero so that f′ (0+) = −f′ (0−), where 0+ and 0− denote the limits from above and below, it follows that a move of βj away from the origin occurs whenever (6) The calculation of L′ involves a sum over all individuals and is computationally expensive. Moreover, recall that the ′ denotes derivative with respect to βj and so L′ is required for each j. However, computationally-fast upper and lower bounds for L′ can be derived (see Text S1), which in conjunction with (6), determine whether a move from βj = 0 is possible. Checking this bound avoids the necessity to compute L′ for all but a small proportion of values of j. From (6) we can derive an explicit approximation for the type-I error rate of our procedure. We reject the null if the posterior mode is not at β = 0. By standardising the genotype data to have mean zero and variance one, the type-I error probability is the same for each SNP, regardless of MAF. By writing |L′ (β) | in terms of, the maximum likelihood estimate of β, and assuming asymptotic normality of, the per-SNP type-I error rate will be α if (7) where n0 and n1 are the numbers of cases and controls and Φ−1 is the inverse normal distribution function; see Text S1 for the derivation. To maintain the same type-I error the prior must be chosen such that the penalty f′ increases as the sample size increases. It can be shown that this criterion for controlling the type-I error, when applied multivariately to equal numbers of cases and controls, gives rise to a smaller type-I error rate once one or more β' s are non-zero; see Text S1. For the DE prior f′ (β) = ξ for all β>0, thus to control the type-I error at α we assign ξ to equal the right hand side of (7). For the NEG prior, the value of f′ (β) is given in (5). The NEG has two parameters, whereas (7) imposes only one constraint. We considered a range of values for the shape parameter λ from 0. 01 to 10 and then assign γ by substituting (7) into (5) and rearranging. Both the NEG and DE behave similarly when they have been set to have the same type-I error, when their derivative at the origin is the same, and when β = 0. Solutions diverge however once SNPs are included in the model, since included SNPs are penalised less by the NEG than by the DE. This results in larger parameter estimates using the NEG, and affects how likely a variable is to be pushed out of the model once it has been included. So far, the genotype variable xij is the allele count, standardised to have mean zero and variance one. This corresponds to a model that is additive on the logistic scale. To implement a search for dominant or recessive effects, we simply recode this variable accordingly. For example, to seek a recessive effect, we assign xij = −u if individual i is heterozygote or major-allele homozygote at SNPj, and xij = v otherwise, where u and v are chosen to standardise xij. When dominant and recessive effects were included in the model they were considered in the following order: (1) additive, (2) dominant, (3) recessive; terms (2) and (3) are only considered if no preceding term is already included in the model at that SNP. The allelic risk ratios were multiplicative within and across loci and the disease prevalence was 12%; the multiplicative disease model is similar to, but not the same as, the logistic regression model on which our analyses are based. Two causal SNPs were chosen with each of the following approximate MAF values: 2%, 5%, and 15%. The two allelic risk ratios for each MAF were chosen so that the power to detect an association was around 25% and 75% using the ATT at a significance threshold of 10−5, see Table 2 for the effect sizes. With this disease model, the background disease risk is typically ≈6% for individuals carrying no causal alleles, and this risk can be attributed either to polygenic or environmental effects. Thus, although we explicitly simulate six causal alleles, this does not exclude multiple weaker causal alleles that are unlikely to be detected. Marker SNPs were sampled randomly from disjoint 5 Kb regions on each chromosome with probability proportional to MAF (1–MAF), resulting in an approximately uniform MAF distribution.

Tests of association with disease status are normally conducted one SNP at a time, ignoring the effects of all other genotyped SNPs. We developed a computationally efficient method to simultaneously analyse all SNPs, either in a genome-wide association (GWA) study, or a fine-mapping study based on re-sequencing and/or imputation. The method selects a subset of SNPs that best predicts disease status, while controlling the type-I error of the selected SNPs. This brings many advantages over standard single-SNP approaches, because the signal from a particular SNP can be more clearly assessed when other SNPs associated with disease status are already included in the model. Thus, in comparison with single-SNP analyses, power is increased and the false positive rate is reduced because of reduced residual variation. Localisation is also greatly improved. We demonstrate these advantages over the widely used single-SNP Armitage Trend Test using GWA simulation studies, a real GWA dataset, and a sequence-based fine-mapping simulation study.

Abstract Introduction Results Discussion Methods

genetics and genomics/genetics of disease mathematics/statistics genetics and genomics/population genetics

2008

Simultaneous Analysis of All SNPs in Genome-Wide and Re-Sequencing Association Studies

High blood pressure (BP) is the most common cardiovascular risk factor worldwide and a major contributor to heart disease and stroke. We previously discovered a BP-associated missense SNP (single nucleotide polymorphism) –rs2272996–in the gene encoding vanin-1, a glycosylphosphatidylinositol (GPI) -anchored membrane pantetheinase. In the present study, we first replicated the association of rs2272996 and BP traits with a total sample size of nearly 30,000 individuals from the Continental Origins and Genetic Epidemiology Network (COGENT) of African Americans (P = 0. 01). This association was further validated using patient plasma samples; we observed that the N131S mutation is associated with significantly lower plasma vanin-1 protein levels. We observed that the N131S vanin-1 is subjected to rapid endoplasmic reticulum-associated degradation (ERAD) as the underlying mechanism for its reduction. Using HEK293 cells stably expressing vanin-1 variants, we showed that N131S vanin-1 was degraded significantly faster than wild type (WT) vanin-1. Consequently, there were only minimal quantities of variant vanin-1 present on the plasma membrane and greatly reduced pantetheinase activity. Application of MG-132, a proteasome inhibitor, resulted in accumulation of ubiquitinated variant protein. A further experiment demonstrated that atenolol and diltiazem, two current drugs for treating hypertension, reduce the vanin-1 protein level. Our study provides strong biological evidence for the association of the identified SNP with BP and suggests that vanin-1 misfolding and degradation are the underlying molecular mechanism. Hypertension (HTN) or high blood pressure (BP) is common in populations worldwide and a major risk factor for cardiovascular disease (CVD) and all-cause mortality [1]. Although it is observed across ethnically diverse populations, the prevalence of HTN in the US varies from 27% in persons of European ancestry to 40% among those of African ancestry [2]. BP is a moderately heritable trait and affected by the combined effects of genetic and environmental factors, with heritable factors cumulatively accounting for 30–55% of the variance [3]. After age 20, African Americans have higher BP than other US race/ethnicities [4]–[6] and the progression from pre-HTN to HTN occurs one year eariler on average [7]. Increased rates of HTN among African Americans are the main factor contributing to their greater risk of CVD and end-stage renal disease compared to US whites [8], [9]. Given the widespread occurrence of this condition, and our as yet limited ability to reduce the disease burden, identifying the genetic variants of BP phenotypes could elucidate the underlying biology of high BP and reduce the CVD prevalence. Identification of genetic variants of consequence for HTN remains a significant challenge, owing in large part to the complex and polygenic nature of the disorder and the imprecision with which the phenotype is measured [10]. Using admixture mapping analysis of data from the Family Blood Pressure Program, we recently identified a genomic region on chromosome 6 harboring HTN-associated variants [11]. The same region on chromosome 6 was replicated in an admixture mapping analysis based on the African Americans enrolled in the Dallas Heart Study [12]. By further genotyping the functional variants in the region of interest on chromosome 6, the VNN1 gene, in particular SNP rs2272996 (N131S) was found to account for the association with BP in both African Americans and Mexican Americans, but this association was not observed in European Americans [12]. Fava et al. [13] recently argued that rs2294757 (T26I), rather than N131S, was a more likely functional variant accounting for the effect on BP because it is located in a splicing regulation site in VNN1, but these investigators only found a weak association between T26I and both DBP and HTN in one of the two studies that they carried out. The results of this study are consistent with the lack of evidence for association observed in European Americans in the Dallas Heart Study [12]. VNN1 encodes the protein vanin-1, a glycosylphosphatidylinositol (GPI) -anchored membrane protein [14], [15]. Vanin-1 is widely expressed in a variety of tissues, with higher expression in liver, kidney and blood [16]. Vanin-1 is a pantetheinase, a member of the biotinidase branch of the nitrilase superfamily [17]. Vanin-1 hydrolyzes pantetheine to pantothenic acid (vitamin B5) and cysteamine, a potent regulator of oxidative stress. In vanin-1 null mice free cysteamine is undetectable, indicating vanin-1' s indispensable role in generating cysteamine under physiological conditions [18]. Therefore, vanin-1 plays an essential role in regulating oxidative stress via cysteamine generation. A linkage between oxidative stress and HTN has been hypothesized for many years [19]–[21]. Furthermore, vanin-1 was reported to be involved in cardiovascular diseases [22], [23]. Overexpression of vanin-1 was associated with progression to chronic pediatric immune thrombocytopenia (ITP) [24], and was shown to lead to hyperglycemia [25]. Vanin-1−/− mice showed protective effects against a variety of phenotypes, such as oxidative stress [26], intestinal inflammation [27], and colon cancer [28], mostly due to higher glutathione storage to maintain a more reducing environment. As a consequence, vanin-1' s pantetheinase activity may offer a physiologic rationale for BP regulation with loss of vanin-1 function. In this study, we first investigated the association evidence of the missense variant rs2272996 (N131S) in VNN1 and BP phenotypes by performing a meta-analysis of nearly 30,000 African ancestry subjects from 19 independent cohorts from the Continental Origins and Genetic Epidemiology Network (COGENT). We next examined whether there were other variants in VNN1 associated with BP traits. Lastly, we conducted molecular experiments to establish a functional connection between N131S vanin-1 and HTN. The study samples were the African-ancestry subjects from the COGENT, which includes 19 discovery cohorts. The details are described elsewhere by Franceschini et al [29]. Briefly, the phenotype-genotype association analysis was performed in each cohort separately. Systolic BP (SBP) and diastolic BP (DBP) were treated as continuous variables. For individuals reporting the use of antihypertensive medications, BP was adjusted by adding 10 and 5 mmHg to SBP and DBP respectively [30]. SBP and DBP were adjusted for age, age2, body mass index (BMI) and gender in linear regression models. The results of association between SNP rs2272996 and SBP or DBP for the 18 cohorts are presented in Figure 1. This SNP was not available in the GeneSTAR cohort. The corresponding allele frequencies in the different studies are listed in Supplementary Table S1. Among the 18 cohorts, 12 and 10 have positive effect sizes for SBP (P = 0. 048) and DBP (P = 0. 24), respectively, comparing to 9 expected under null hypothesis of no association between this SNP and BP. We next performed meta-analysis by applying both fixed-effect [31], [32] and random-effect [33] models to estimate the overall effect. SNP rs2272996 was significantly associated with SBP in both fixed-effect (P = 0. 01) and random-effect (P = 0. 04) models (Table 1). However, we did not observe evidence of genotype-phenotype association for DBP. Among the individual cohort analyses, the Maywood cohort had a sample size of 743 and was the only cohort that showed significant association with rs2272996 for both SBP (P = 0. 016) and DBP (P = 0. 0003), however the direction of the association was opposite to what was found in the test for overall effects (Figure 1). The distributions of SBP, DBP, age and BMI did not suggest that the Maywood was an outlier in epidemiologic characteristics (Supplementary Figure S1), with the exception that the sampling strategy for this cohort was based on exclusion of persons on antihypertensive medications (antihypertensive medication rate was 0. 7%). The Nigeria cohort also included a low antihypertensive medication rate but this was a result of inaccessibility to medications (Supplementary Figure S2). When analyses were repeated after exclusion of the Maywood cohort, the association of BP and rs2272996 was substantially improved (P = 0. 003 for SBP, Table 1). The different association evidence between Maywood and other cohorts may suggest genetic heterogeneity or possible interaction between gene and environment factors, although further studies are needed to address this possibility. Since additional genetic variants in VNN1 might be associated with BP, we examined available known variants in VNN1 including 10 kb up- and down-stream of the gene. A total of 105 other SNPs were available in the 19 cohorts. SNP rs7739368 had the smallest p values for association with SBP using either fixed-effect model (P = 0. 004) or random-effect model (P = 0. 004, Supplementary Table S2), but this was not significant after correcting for multiple comparisons. This SNP is ∼7 k bp' s upstream of VNN1 adjacent to the PU. 1 transcription factor binding region (306 bp' s upstream). To understand the function of N131S vanin-1 in relation to HTN, plasma samples from Nigeria HTN patients and normotensives with WT (TT) or homozygous N131S (CC) vanin-1 were collected (6 samples per group, 4 groups). The same amount of total plasma protein from each sample was subjected to Western blot analysis: a clean vanin-1 protein band appeared at 70 kD (Figure 2A), consistent with previous reports [14], [34]. The plasma vanin-1 protein in homozygous N131S vanin-1 was significantly lower than that in WT vanin-1 in both hypertensive and normotensive groups (P = 1. 64×10−5 and 0. 014, Figure 2A, see Figure 2B for quantification), indicating that the N131S mutation is a functional variant that is associated with substantially less steady-state vanin-1 protein. Furthermore, the plasma vanin-1 protein in normotensive groups with WT vanin-1 (samples 13–18) was significantly lower than that in HTN patients with WT vanin-1 (samples 1–6) (P = 0. 042) (Figure 2A, see Figure 2B for quantification). These results demonstrated that vanin-1 expression is associated with both the genotypic N131S mutation and phenotypic HTN, with the former exerting stronger effect. Lastly, the plasma vanin-1 protein in normotensive groups with homozygous N131S vanin-1 (samples 19–24) is also lower than that in HTN patients with homozygous N131S vanin-1 (samples 7–12) although it was not statistically significant (P = 0. 13), probably due to the already exceedingly low vanin-1 quantity. These results suggest that the WT vanin-1 is associated with increased plasma vanin-1 protein expression, and increased HTN risk. We tested two variants, N131S and T26I, as regards how they influence the total vanin-1 protein levels because other investigators have suggested that T26I may be a candidate variant for BP variation as well [13]. We utilized the human embryonic kidney 293 (HEK293) cells stably expressing these vanin-1 variants because HEK293 cells have high transfection efficiency and physiologically-relevant cell environment for vanin-1 protein expression [23]. Significantly lower total vanin-1 proteins were detected in the cells expressing N131S vanin-1, whereas similar vanin-1 protein levels were detected in cells expressing T26I vanin-1 compared to cells expressing WT vanin-1 (Figure 3A, quantification shown below). Because vanin-1 is a GPI-anchored membrane protein, it needs to traffic efficiently to the plasma membrane for its pantetheinase activity. We hypothesized that N131S substantially reduces the trafficking of vanin-1 protein to the plasma membrane, whereas T26I does not. Using a surface biotinylation assay [35], we observed that the N131S mutation led to significantly lower plasma membrane expression, whereas in cells expressing the T26I mutation, vanin-1 surface expression was similar to that observed in WT cells (Figure 3B, quantification shown below). We further confirmed that the variation in vanin-1 protein expression resulted in corresponding functional consequences for N131S and T26I mutations. A cell-based fluorescence assay was carried out to record the kinetics of pantetheinase activity by vanin-1 variants [36]. The cells expressing T26I vanin-1 had similar pantetheinase activity compared to cells expressing WT vanin-1; however, cells expressing N131S vanin-1 retained approximately 9% of the pantetheinase activity, by quantifying the fluorescence signals at the kinetic steady state at 57 minutes (Figure 3C). These data taken together provide evidence of less protein, less membrane trafficking, and lower enzymatic activity of the N131S protein as compared to both the wild type and the T26I variant. To determine the mechanism of loss of surface N131S vanin-1, we sought to confirm that N131S vanin-1 is rapidly degraded. A cycloheximide (CHX) chase assay was used to quantify the half-life of vanin-1 variants in HEK293 cells: WT vanin-1 had a half-life of 240 min; T26I vanin-1,232 min; N131S vanin-1,76 min, respectively (Figure 4A, quantification in Figure 4B). Thus, N131S vanin-1 has a much faster degradation rate than WT vanin-1, whereas T26I vanin-1 is degraded at a rate similar to that of WT vanin-1. To confirm that misfolded N131S vanin-1 is subjected to ERAD, we applied MG-132 to the cells, which is a potent proteasome inhibitor. MG-132 treatment resulted in the accumulation of ubiquitinated proteins and substantially more total vanin-1 proteins (Figure 4C, cf. lane 2 to lane 1), indicating that efficient proteasome inhibition prevents the degradation of N131S vanin-1. Furthermore, using immunoprecipitation against vanin-1, we confirmed that MG-132 treatment resulted in ubiquitination of N131S vanin-1 (Figure 4C, cf. lane 5 to lane 4). These data indicate that N131S vanin-1 is subjected to rapid ERAD, resulting in loss of functional vanin-1 on the plasma membrane. We hypothesize that rapid degradation of N131S vanin-1 resulted from its misfolding in the endoplasmic reticulum (ER). The endoglycosidase H (endo H) enzyme selectively cleaves vanin-1 after asparaginyl-N-acetyl-D-glucosamine (GlcNAc) in the N-linked glycans incorporated in the ER. After the high-mannose form is enzymatically remodeled in the Golgi, endo H is unable to remove the oligosaccharide chain. Therefore, endo H-resistant vanin-1 bands (with higher molecular weight) represent properly folded, post-ER vanin-1 glycoforms, which traffic at least to the Golgi compartment. The N131S mutation resulted in much less intense endo H-resistant bands than WT vanin-1 (Figure 4D, cf. lane 6 to lane 2), whereas T26I did not (Figure 4D, cf. lane 4 to lane 2). The ratio of endo H-resistant to total vanin-1 serves as a measure of vanin-1 trafficking efficiency. The trafficking efficiency of N131S vanin-1 was less than WT vanin-1, indicating that N131S vanin-1 does not fold properly in the ER. These data support the conclusion that N131S vanin-1 is misfolded in the ER and subsequently degraded by the ERAD pathway. To determine whether vanin-1 is a target of current anti-hypertensive drugs, we tested the effect of two commonly prescribed HTN drugs with different known drug mechanisms on endogenous vanin-1 protein level. Human monocyte THP-1 cells were used because they were derived from human blood and have high endogenous WT vanin-1 protein expression levels. Two HTN drugs used are diltiazem [37], an L-type calcium channel blocker, and atenolol, a selective β1 adrenergic receptor blocker [38]. Treatment of THP-1 cells with diltiazem (10 µM) or atenolol (10 µM) for 1d or 3d decreased the endogenous total vanin-1 protein significantly in a time-dependent manner (Figure 5A, quantification shown below). Furthermore, application of diltiazem for 3d decreased the endogenous total vanin-1 protein significantly in a dose-dependent manner (Figure 5B, quantification shown below). This indicates that vanin-1 is a molecular target of current HTN drugs, which was previously unknown and confirms the relevance of vanin-1 to the regulation of blood pressure. Therefore, exploring other compounds that decrease vanin-1 level may lead to discovery of novel antihypertensive drugs, especially those with previously unknown function in HTN. A major methodological issue that has greatly increased the challenges faced in the genetic epidemiology of BP is the high noise-to-signal ratio in the phenotype. This problem has numerous causes, including variation in measurement protocols of SBP and DBP across studies, the dynamic nature of BP levels, and concurrent use of antihypertensive medications. In addition, as with all polygenic disorders, the effect size for any single gene variant is very small and a large number of genes/variants are involved [10]. Recent large-scale BP genome-wide association studies (GWAS) of European, Asian and African ancestry populations demonstrated that the identified genetic variants together explain only 1–2% of BP variation [29], [39], [40]. It is thus not surprising that a large sample size is often necessary to detect genome-wide significant effects. An analysis method complementary to GWAS is admixture mapping, which has been successfully applied to detect BP loci [11], [12], [41]. Our group reported that the missense variant rs2272996 (N131S) in VNN1 was associated with BP through admixture mapping, and we conducted a follow-up association analysis in African and Mexican American samples [11], [12]. The association evidence in European-ancestry population is however less convincing [12], [13] In the current study, we performed meta-analysis using the COGENT consortium consisting of 19 studies with a total sample size of nearly 30,000 African ancestry subjects and confirmed the association evidence between rs2272996 and SBP (P = 0. 01, Table 1). However, statistical evidence alone cannot explain the role of a given variant on disease risk and drug response. Therefore, in our study we decided to analyze the functional effects of the N131S variant. Vanin-1 is a pantetheinase generating cysteamine, which regulates the glutathione-dependent oxidative stress response. We showed that the HTN-associated N131S mutation in vanin-1 significantly reduces vanin-1 total and cell surface expression. Consequently, the N131S vanin-1 only has fractional pantetheinase activity on the plasma membrane, which is associated with decreased HTN risk. Our result is consistent with the recognized link between impaired reduction-oxidation status and the development of HTN [19]–[21], and the observed protective effects in vanin-1−/− mice in a variety of diseases, including oxidative stress [26], intestinal inflammation [27], and colon cancer [28], mostly due to higher glutathione storage to maintain a more reducing environment. We further tested the drug effects of atenolol and diltiazem in human monocyte THP-1 cells, which have high endogenous WT vanin-1 protein expression level. Atenolol is a selective β1 adrenergic receptor blocker and developed as a replacement for propranolol in treating hypertension; diltiazem is a nondihydropyridine member of calcium channel blockers used in treatment of hypertension. We found that both drugs reduce the vanin-1 protein level in the THP-1 cells. The anti-hypertensive drugs may have different and complex mechanisms leading to reduced BP, but whether vanin-1 is targeted was previously unknown. Our experiments filled this gap and showed vanin-1 is involved in the BP regulation pathway. Therefore, other potent vanin-1 inhibitors may prove to have BP reducing effects, which is especially useful given that these inhibitors have not been studied in HTN and thus may provide new therapeutics for HTN. Our study presented the first functional studies of vanin-1 in HTN association, and provides compelling evidence for the essential role of its N131S mutation. Nonetheless, it has been demonstrated that multiple variants in a gene may contribute to a phenotypic variation [42], [43], and it is possible that other closely linked variants may have similar or analogous effects, or act in combination with N131S to regulate the vanin-1 protein expression and function. Current GWAS of HTN related traits mainly focus on testing common variants (MAF: minor allele frequency ≥5%) through pre-built chips and imputations based on HapMap [44] data; to date those findings in general have modest effect sizes [39]. Other functional and rare variants may be identified by deep sequencing, in combination with publicly available databases, such as the 1000 Genome Projects [45] and the Encyclopedia of DNA Elements (ENCODE) [46]. Identification of additional functional SNPs in VNN1 and their association with BP should provide further evidence for vanin-1 function in the regulation of BP. Cell lines were used to determine that ERAD is the underlying mechanism for vanin-1' s loss of function due to the N131S mutation. Cell lines are commonly used for the study of molecular mechanisms because they typically provide efficient transfection and a physiologically-relevant cell environment for the target protein. However, BP has a complex etiology with the involvement of a variety of organs, such as heart, brain and kidney, which cannot be recapitulated solely in cell lines. Although knowledge gained from our cell system provides essential cellular mechanistic insights into the regulation of vanin-1 and its function, the study of BP regulation by vanin-1 calls for studies in animal models. A hypertensive mouse or rat model, vanin-1 knockout mouse or rat model, and N131S vanin-1 knockin mouse or rat model would be of great interest to study the effects of vanin-1 and its mutation in the complex physiological and metabolic systems. Vanin-1 provides a potential candidate to be manipulated to ameliorate HTN. Vanin-1 is a pantetheinase that contains the conserved catalytic triad residue–glutamate, lysine and cysteine–within the nitrilase family [47]. Based on the sequence alignment of vanin-1 with other nitrilase family members, the conserved catalytic triad of vanin-1 is composed of glutamate 79, lysine 178 and cysteine 211 [23]. A three-dimensional atomic model of vanin-1 was built using the I-TASSER server (Figure S3) [48]. Neither T26 nor N131 is in the vicinity of the catalytic sites of vanin-1. Therefore, the T26I and N131S mutations per se are not expected to change the vanin-1 enzyme activity significantly. Indeed, we showed that the T26I mutation did not influence vanin-1 maturation or enzymatic activity. The N131S mutation has much weaker pantetheinase activity, presumably due to exceedingly low concentration of N131S vanin-1 on the plasma membrane; however, the activity is still evident, implying that the catalytic triad is not disrupted by this mutation. Loss of function of vanin-1 is caused by misfolding and rapid degradation of vanin-1 due to a single missense mutation from Asn to Ser at position 131. As a GPI-anchored protein, to function properly, vanin-1 needs to be trafficked efficiently to the plasma membrane, where it acts as a pantetheinase. In accordance with the maturation of general GPI-anchored proteins [15], vanin-1 is co-translationally translocated into the ER for folding. Because human vanin-1 has six potential N-linked glycosylation sites, its maturation is presumably dictated by glycoprotein processing machinery in the ER [49], [50]. Properly folded vanin-1 is trafficked out of the ER, through the Golgi and to the plasma membrane in a fully functional state. Misfolded vanin-1 is recognized by the ER quality control machinery and subjected to ERAD, being retrotranslocated to the cytosol, ubiquitinated and degraded by the proteasome [51]–[54]. Cells need to maintain a delicate balance between protein synthesis, folding, trafficking, aggregation and degradation for individual proteins that make up the proteome in normal physiology. This balance is dictated by the cellular protein homeostasis (proteostasis) network, composed of a variety of sub-networks, including the chaperone, degradation and trafficking networks, and cellular signaling pathways that regulate proteostasis as the core layers [55]–[57]. Therefore, further elucidation of the proteostasis network for vanin-1 should provide a valuable fine-tuning control of vanin-1 expression, function and BP. 19 cohort studies contributed to the meta-analysis of BP and genetic variants in VNN1 in African-Americans as detailed in Franceschini et al [29], including Biological Bank of Vanderbilt University (BioVU); Atherosclerosis Risk In Communities (ARIC); Coronary Artery Risk Development in Young Adults (CARDIA); Cleveland Family Study (CFS); Jackson Heart Study (JHS); Multi-Ethnic Study of Atherosclerosis (MESA); Cardiovascular Health Study (CHS); Genetic Study of Atherosclerosis Risk (GeneSTAR); Genetic Epidemiology Network of Arteriopathy (GENOA); The Healthy Aging in Neighborhoods of Diversity Across the Life Span Study (HANDLS); Health, Aging, and Body Composition (Health ABC) Study; The Hypertension Genetic Epidemiology Network (HyperGEN); Mount Sinai, New York City, USA Study (Mt Sinai Study); Women' s Health Initiative SNP Health Association Resource (WHI); Howard University Family Study (HUFS); Bogalusa Heart Study (Bogalusa); Sea Islands Genetic Network (SIGNET); Loyola Maywood Study (Maywood); and Loyola Nigeria Study (Nigeria). Each study received IRB approval of its consent procedures, examination and surveillance components, data security measures, and DNA collection and its use for genetic research. We selected 24 plasma samples from the International Collaborative Study on Hypertension in Blacks (ICSHIB), in which the study participants were recruited from Igbo-Ora and Ibadan in southwest Nigeria as part of a long-term study on the environmental and genetic factors underlying hypertension [58]. The ICSHIB included 1,188 subjects who were genotyped using Affymetrix platform 6. 0 chip [59]. We selected 6 subjects per group from the high and lower SBP traits in each of TT and CC genotype groups of SNP rs2272996. For each of these 24 subjects, western blot analysis was performed by controlling the same amount of total plasma protein. The detailed statistical analysis of each cohort can be found in Franceschini et al [29]. In brief, each study cohort received a uniform statistical analysis protocol and analyses were conducted accordingly. BP was measured in mmHg. For individuals reporting use of antihypertensive medications, BP was imputed by adding 10 and 5 mmHg for SBP and DBP, respectively. For unrelated individuals, SNP associations for SBP or DBP were assessed by linear regression assuming an additive model, adjusting for age, age2, body mass index (BMI) and gender. Population stratification was controlled by adjusting for the first 10 principal components obtained from selected ancestry informative markers [60], [61]. For family data, association was tested using a linear mixed effect model, where random effects account for family structure [62]. Meta-analysis across the 19 cohorts was performed by applying both fixed-effect [31], [32] and random-effect [33] models to estimate the overall effect. The fixed-effect model assumes that the effect size is the same for all the included studies; the only source of error is the random error within studies, which depends primarily on the sample size for each study. Because the inverse variance is roughly proportional to sample size, the fixed-effect model provides a weighted average of the effect sizes, with the weights being the estimated inverse of the variance of the estimate in each study. The random-effect model assumes that the effect sizes from studies are similar but not identical, dependent on each study protocol; the source of error includes within-study and among-study error [33]. It is more conservative and thus provides relatively wider 95% confidence intervals when heterogeneity across studies exists. All experimental data are presented as mean ± SEM, and any statistical significance was calculated using two-tailed Student' s t-test. MG-132, diltiazem, and atenolol were obtained from Sigma-Aldrich. The pCMV6 plasmids containing human vanin-1 and pCMV6 Entry Vector plasmid (pCMV6-EV) were obtained from Origene. The human vanin-1 missense mutations, N131S and T26I, were constructed using QuickChange II site-directed mutagenesis Kit (Agilent Genomics), and the cDNA sequences were confirmed by DNA sequencing, showing the single-site mutation of these variants. The rabbit polyclonal anti-vanin-1 antibody came from Pierce antibodies, the mouse monoclonal anti-transferrin antibody from Santa Cruz Biotechnology, the mouse monoclonal anti-β-actin antibody from Sigma, and the rabbit polyclonal anti-ubiquitin antibody from Cell Signaling. Human embryonic kidney 293 (HEK293) cells and human monocytic THP-1 cells came from ATCC. THP-1 cells were maintained in RPMI-1640 medium (Hyclone) with 10% heat-inactivated fetal bovine serum (Sigma-Aldrich) and 1% Pen-Strep (Hyclone) at 37°C in 5% CO2. HEK293 cells were maintained in Dulbecco' s Modified Eagle Medium (DMEM) (Hyclone) with 10% heat-inactivated fetal bovine serum (Sigma-Aldrich) and 1% Pen-Strep (Hyclone) at 37°C in 5% CO2. Monolayers were passaged upon reaching confluency with TrypLE Express (Life Technologies). HEK293 cells were grown in 6-well plates or 10-cm dishes and allowed to reach ∼70% confluency before transient transfection using Lipofectamine 2000 (Life Technologies) according to the manufacturer' s instruction. Stable cell lines expressing vanin-1 variants (WT, N131S or T26I) were generated using the G-418 selection method. Briefly, transfected cells were maintained in DMEM supplemented with 0. 8 mg/mL G418 (Enzo Life Sciences) for 15 days. G-418 resistant cells were selected for follow-up experiments. Cells were harvested and then lysed with lysis buffer (50 mM Tris, pH 7. 5,150 mM NaCl, and 1% Triton X-100) supplemented with Roche complete protease inhibitor cocktail. Lysates were cleared by centrifugation (15,000× g, 10 min, 4°C). Protein concentration was determined by MicroBCA assay (Pierce). Endoglycosidase H (endo H) or Peptide-N-Glycosidase F (PNGase F) (New England Biolabs) enzyme digestion was performed according to published procedure [35]. Aliquots of cell lysates or human plasma samples were separated in an 8% SDS-PAGE gel, and Western blot analysis was performed using the appropriate antibodies. Band intensity was quantified using Image J software from the NIH. HEK293 cells stably expressing vanin-1 variants were plated in 10-cm dishes for surface biotinylation experiments according to published procedure [35]. Intact cells were washed twice with ice-cold PBS and incubated with the membrane-impermeable biotinylation reagent Sulfo-NHS SS-Biotin (0. 5 mg/mL; Pierce) in PBS containing 0. 1 mM CaCl2 and 1 mM MgCl2 (PBS+CM) for 30 min at 4°C to label surface membrane proteins. To quench the reaction, cells were incubated with 10 mM glycine in ice-cold PBS+CM twice for 5 min at 4°C. Sulfhydryl groups were blocked by incubating the cells with 5 nM N-ethylmaleimide (NEM) in PBS for 15 min at room temperature. Cells were solubilized for 1 h at 4°C in lysis buffer (Triton X-100,1%; Tris–HCl, 50 mM; NaCl, 150 mM; and EDTA, 5 mM; pH 7. 5) supplemented with Roche complete protease inhibitor cocktail and 5 mM NEM. The lysates were cleared by centrifugation (16,000× g, 10 min at 4°C) to pellet cellular debris. The supernatant contained the biotinylated surface proteins. The concentration of the supernatant was measured using microBCA assay (Pierce). Biotinylated surface proteins were affinity-purified from the above supernatant by incubating for 1 h at 4°C with 100 µL of immobilized neutravidin-conjugated agarose bead slurry (Pierce). The samples were then subjected to centrifugation (16,000×g, 10 min, at 4°C). The beads were washed six times with buffer (Triton X-100,0. 5%; Tris–HCl, 50 mM; NaCl, 150 mM; and EDTA, 5 mM; pH 7. 5). Surface proteins were eluted from beads by boiling for 5 min with 60 µL of LSB/Urea buffer (2× Laemmli sample buffer (LSB) with 100 mM DTT and 6 M urea; pH 6. 8) for SDS-PAGE and Western blotting analysis. The cell-based fluorescence assay to evaluate vanin-1' s pantetheinase activity was performed according to published procedure with modifications [36]. The substrate, pantothenate-7-amino-4-methylcoumarin (pantothenate-AMC) was chemically synthesized according to published method [36]. As a pantetheinase, vanin-1 catalyzed the release of AMC, giving a fluorescence signal at excitation 350 nm and emission 460 nm. HEK293 cells expressing vanin-1 variants were lysed with lysis buffer (50 mM Tris, pH 7. 5,150 mM NaCl, and 1% Triton X-100) supplemented with Roche complete protease inhibitor cocktail. Enzyme activity was performed using 10 µg of total proteins containing the substrate pantothenate-AMC (5 µM), 0. 5 mM DTT, 5% DMSO in a 100 µL final volume in PBS, pH 7. 5. Fluorescence signals at excitation 350 nm and emission 460 nm measuring the released AMC were recorded every 3 min at 37°C in 96-well plates (Greiner Bio-One) using a fluorescence plate reader. A 60-min kinetic assay in four replicates and three biological replicates was carried out. Buffer only and HEK293 cells transfected with empty vector (EV) were used as negative controls for non-specific pantetheinase activity. HEK293 cells stably expressing vanin-1 variants were seeded at 2. 5×105 cells per well in 6-well plates and incubated at 37°C overnight. To stop protein translation, cells were treated with 100 µg/mL cycloheximide (Ameresco) and chased for the indicated time. Cells were then lysed for SDS-PAGE and Western blot analysis. Cell lysates (500 µg) were pre-cleared with 30 µL of protein A/G plus-agarose beads (Santa Cruz) and 1. 0 µg of normal rabbit IgG for 1 hour at 4°C to remove nonspecific binding proteins [63]. The pre-cleared cell lysates were incubated with 2. 0 µg of rabbit anti-vanin-1 antibody (Pierce) for 1 hour at 4°C, and then with 30 µL of protein A/G plus agarose beads overnight at 4°C. The beads were collected by centrifugation at 8000×g for 30 s, and washed four times with lysis buffer. The vanin-1 protein complex was eluted by incubation with 30 µL of SDS loading buffer in the presence of 100 mM DTT. The immunopurified eluents were separated in 8% SDS-PAGE gel, and Western blot analysis was performed.

Hypertension (HTN) or high blood pressure (BP) is common worldwide and a major risk factor for cardiovascular disease and all-cause mortality. Identification of genetic variants of consequence for HTN serves as the molecular basis for its treatment. Using admixture mapping analysis of the Family Blood Pressure Program data, we recently identified that the VNN1 gene (encoding the protein vanin-1), in particular SNP rs2272996 (N131S), was associated with BP in both African Americans and Mexican Americans. Vanin-1 was reported to act as an oxidative stress sensor using its pantetheinase enzyme activity. Because a linkage between oxidative stress and HTN has been hypothesized for many years, vanin-1' s pantetheinase activity offers a physiologic rationale for BP regulation. Here, we first replicated the association of rs2272996 with BP in the Continental Origins and Genetic Epidemiology Network (COGENT), which included nearly 30,000 African Americans. We further demonstrated that the N131S mutation in vanin-1 leads to its rapid degradation in cells, resulting in loss of function on the plasma membrane. The loss of function of vanin-1 is associated with reduced BP. Therefore, our results indicate that vanin-1 is a new candidate to be manipulated to ameliorate HTN.

Abstract Introduction Results Discussion Materials and Methods

biology and life sciences medicine and health sciences research and analysis methods

2014

The Association of the Vanin-1 N131S Variant with Blood Pressure Is Mediated by Endoplasmic Reticulum-Associated Degradation and Loss of Function

Antimicrobial peptides (AMPs) are thought to kill bacterial cells by permeabilizing their membranes. However, some antimicrobial peptides inhibit E. coli growth more efficiently in aerobic than in anaerobic conditions. In the attack of the human cathelicidin LL-37 on E. coli, real-time, single-cell fluorescence imaging reveals the timing of membrane permeabilization and the onset of oxidative stress. For cells growing aerobically, a CellROX Green assay indicates that LL-37 induces rapid formation of oxidative species after entry into the periplasm, but before permeabilization of the cytoplasmic membrane (CM). A cytoplasmic Amplex Red assay signals a subsequent burst of oxidative species, most likely hydrogen peroxide, shortly after permeabilization of the CM. These signals are much stronger in the presence of oxygen, a functional electron transport chain, and a large proton motive force (PMF). They are much weaker in cells growing anaerobically, by either fermentation or anaerobic respiration. In aerobic growth, the oxidative signals are attenuated in a cytochrome oxidase–bd deletion mutant, but not in a –bo3 deletion mutant, suggesting a specific effect of LL-37 on the electron transport chain. The AMPs melittin and LL-37 induce strong oxidative signals and exhibit O2-sensitive MICs, while the AMPs indolicidin and cecropin A do not. These results suggest that AMP activity in different tissues may be tuned according to the local oxygen level. This may be significant for control of opportunistic pathogens while enabling growth of commensal bacteria. Antimicrobial peptides (AMPs, also called host-defense peptides) play a number of important roles in the innate immune response of plants and animals [1]. Important human AMPs include the cathelidicin LL-37 and the defensins [2]. In humans, AMPs are constitutively expressed in phagocytes, including macrophages, neutrophils, and dendritic cells [3,4]. When a pathogen attacks the host, phagocytes initially envelope the invading microbes in internal phagosomes [5]. The phagosome fuses with lysosomes to form the phagolysosome. Presence of the pathogen stimulates a “respiratory burst” in the phagocyte, leading to synthesis of harmful reactive oxygen species (ROS) within the phagolysosome [6,7]. In parallel, AMPs stored in granules are released into the phagolysosome, where their high concentration likely contributes to direct killing of the invading pathogen. AMPs are also released from the phagocyte into surrounding tissue and the bloodstream. In addition to directly attacking pathogens, these external AMPs serve a variety of immunoregulatory functions [8,9]. Most AMPs are cationic and amphipathic. They are attracted to the anionic outer surfaces of bacterial cells and, at sufficient concentration, permeabilize bacterial membranes. In early studies, the halting of growth of bacterial pathogens by AMPs was typically attributed to permeabilization of the cytoplasmic membrane (CM), with concomitant loss of the proton motive force (pmf), loss of critical small molecules, and halting of ATP production. However, over the past 15 years, many studies have shown that AMPs cause a variety of deleterious biophysical and biochemical effects in bacterial cells, including interference with cell wall biosynthesis, DNA replication, transcription, and translation [10–13]. Induction of reactive oxygen species (ROS) has received little attention as a potentially important aspect of AMP action against bacterial cells [14]. We recently used time-resolved, single-cell fluorescence microscopy [15–19] to show that the hybrid synthetic peptide CM15 (15 aa long, net +6 charge) induces oxidative stress within seconds of contact with E. coli growing in aerobic conditions [20]. The minimum inhibitory concentration (MIC) was 20-fold higher in anaerobic (fermentation) conditions than in aerobic growth, suggesting that induction of oxidative stress may be a significant growth-halting mechanism. Additional evidence from the oxidation sensitive dye CellROX Green and an intracellular Amplex Red assay suggested that CM15 may interfere with the electron transport chain, possibly leading to formation of superoxide (•O2–) and hydroxyl radical (•OH) as well as hydrogen peroxide (H2O2) [21]. We observed analogous effects of oxygen for MM63: CHx37, a potent example of a highly cationic, random β-peptide copolymer (mean chain length 35 units, 63% cationic sidechains) [22]. The MIC of this copolymer against E. coli is at least 8-fold lower in aerobic than in anaerobic conditions. Both CM15 and MM63: CHx37 are synthetic peptides. To get a better sense of the generality of these phenomena, here we extend our studies of oxidative effects to four natural AMPs (Table 1): LL-37 (human cathelicidin, α-helical, 37 aa long, net +6), cecropin A (moth, α-helical, 37 aa long, net +7) ), melittin (bee, α-helical, 26 aa long, net +6), and indolicidin (bovine, extended structure, 13 aa long, net +4). LL-37 and melittin exhibit significantly lower MICs against E. coli in aerobic vs anaerobic (fermentation) conditions, and they both induce strong fluorescence signals indicative of oxidative stress. In contrast, for cecropin A and indolicidin the MIC is the same in aerobic and anaerobic conditions. Correspondingly, they induce much weaker fluorescence signals. In addition, we provide a detailed, single-cell comparison of LL-37 attacking E. coli growing under conditions of aerobic respiration, anaerobic fermentation, and anaerobic respiration. In aerobic growth, a burst of oxidative species is induced already on access of LL-37 to the periplasm, i. e. , well before the cytoplasmic membrane is permeabilized to the dye Sytox Orange. The mechanism may involve interference with proteins of the electron transport chain (ETC), leading to improper release of superoxide (•O2–) into the periplasmic space. Mutation studies suggest that LL-37 targets the cytochrome oxidase-bd complex, but not the cytochrome oxidase-bo3 complex. A subsequent burst of oxidative species, detected by an intracellular Amplex Red assay sensitive to H2O2, rose sharply at the moment of CM permeabilization. For cells growing by anaerobic fermentation or by anaerobic respiration using NO3– as terminal electron acceptor, no such signals of abrupt oxidative events were observed. However, the CellROX Green and Amplex Red assays are insensitive to oxidative nitrogen-containing radicals, so that oxidative damage might still be occurring. These new results suggest the possibility that the host may use the degree of tissue aeration for selective control of the potency of AMPs. The present work indicates that LL-37 is most potent against E. coli in oxygen-rich conditions. Earlier work found the same effect for the human beta defensin hBD-3, but the opposite effects for hBD-1 [23]. Reduction of the Cys-Cys linkages in hBD-1, which converts the globular oxidized structure to a linear structure, greatly enhanced its antimicrobial activity against anaerobic Gram positive species. Tuning of cationic AMP activity according to local redox conditions may prove to be important in controlling opportunistic pathogens while enabling growth of commensal bacteria. We have measured MICs in aerobic and anaerobic fermentation conditions using a series of two-fold dilutions for the four natural AMPs in the same rich, chemically defined EZRDM medium at 30°C (Table 1). The MIC in anaerobic fermentation conditions is 4-fold higher for LL-37 and 8-fold higher for melittin. Multiple experimental runs produce the same MIC value within the resolution of the two-fold dilution steps, indicating that 4-fold and 8-fold differences are significant. For cecropin A and for indolicidin, the MIC is the same in aerobic and fermentation conditions. Evidently the activity against E. coli of some, but not all, natural AMPs is enhanced by the presence of oxygen. For LL-37, the primary focus of this study, we also measured the MIC in conditions enabling anaerobic respiration (no oxygen, but supplemented with 10 mM KNO3; Table 2). The MIC is 12 μM, three times higher than in aerobic respiration. The doubling times for MG1655 growing in EZRDM at 30°C under aerobic respiration, anaerobic respiration with KNO3, and fermentation conditions are similar, all in the range 45–53 min (Table 2). Earlier work in M9 medium supplemented with hydrolyzed casein [24] found that E. coli maintained a substantial proton motive force (pmf) in all three growth conditions (–160 mV for aerobic respiration, –144 mV for anaerobic respiration using NO3–, and –117 mV for fermentation). These pmf values may not be transferable to our strain and growth conditions. We chose the human cathelicidin LL-37 for a detailed time-dependent, single-cell study of antimicrobial action. First we investigated the extent to which LL-37 in aerobic conditions at the 6-hour MIC of 4 μM causes cell death (irreversible halting of growth) on the timescale of our microscopy measurements, typically 30–60 min. We monitored cell killing activity using a conventional cell survival assay. A mid-log phase culture of MG1655 E. coli was incubated with LL-37 at 4 μM, sampled after 30 min, 1 hr, and 2 hr of incubation, and serially diluted over the range 5 x 106 to 5 cells/mL. One μL of the diluted sample was spot-plated onto a 3% LB-agar plate for overnight growth at 30°C. In control experiments, no LL-37 was added prior to plating and incubation. At 4 μM LL-37, we observe a significant decrease in colony formation after 30 min of incubation when compared to the control (S1 Fig). After 1-hr incubation at 4 μΜ, essentially no colonies formed, even for the least diluted sample. These results were reproducible over three trials. According to classic clinical microbiological definitions, this indicates that the MIC and the minimum bactericidal concentration (MBC) for LL-37 are essentially the same. At 4 μM of LL-37, both growth inhibition and cell death occur within a 1-hr period. This indicates that the single-cell signs of LL-37 attack as observed under the microscope on the 30–60 minute timescale are likely relevant to cell killing activity. For individual E. coli cells, time-lapse microscopy can determine the timing of the slowing or halting of cell growth, of outer membrane permeabilization, and of cytoplasmic membrane permeabilization following the onset of flow of LL-37 (Methods). Warm (30°C), aeriated EZRDM growth medium flows continuously across the plated cells. Over an observation period of 30–60 min, we alternate phase contrast images with fluorescence images (either one or two colors) and make time-dependent, quantitative measurements of cell length and total fluorescence intensity. In the first set of measurements, the E. coli cells express GFP that is exported to the periplasm by the twin-arginine transport (Tat) system [25]. This produces a characteristic halo image [26]. The medium contains 5 nM of the DNA stain Sytox Green, which becomes fluorescent on crossing both membranes and binding to the chromosomal DNA within the cytoplasm. We directly observe cell length vs time (from phase contrast), the onset of permeabilization of the OM to periplasmic GFP (observed as loss of the green halo surrounding the cytoplasm), and the onset of permeabilization of the CM to Sytox Green (from green staining of the nucleoids). At t = 0, we initiate flow of 4 μM LL-37 (the 6-hr MIC) in aerated medium through the microfluidics observation chamber. For the example cell in Fig 1, the growth rate (slope of the plot of cell length vs time) begins to decrease immediately after injection of LL-37. For at least 90% of the cells in a typical field of 50 cells, we observe gradual slowing or abrupt halting of growth within 10 min of injection. Over the first 30 min after injection of LL-37,60% of the cells lose the “halo” of periplasmic GFP, indicating OM permeabilization to GFP (Fig 1C). During the same 30 min, the other 40% of the cells exhibit an attenuated growth rate, yet continue to elongate without loss of periplasmic GFP. Loss of GFP begins over a wide range of times (2–50 min). Once it begins, complete loss of GFP occurs fairly quickly, over the subsequent 2–3 min. The onset of GFP intensity loss almost always coincides with moderate shrinkage of cell length. As observed before [19], obviously septating cells tend to undergo OM permeabilization earlier than apparently non-septating cells. For the 60% of cells that undergo OM permeabilization before t = 30 min, growth halts. A new signal from Sytox Green begins to rise within 5 min of the OM permeabilization event (example in Fig 1 and S1 Movie). Green fluorescence evolves in the cytoplasm in a spatial pattern reminiscent of the distribution of the E. coli nucleoids, indicating CM permeabilization to Sytox Green. Permeabilization of the CM correlates in time with additional shrinkage of cell length, presumably due to loss of osmolytes from the cytoplasm. The 40% of cells that continued to elongate slowly for the first 30 min did not display a Sytox Green signal during that period, indicating that both the CM and the OM remained intact. However, eventually almost all cells exhibit both OM and CM permeabilization within a 1-hr period after LL-37 addition, as shown by the cumulative distribution function of the lag times to CM permeabilization (Fig 1C). The timescale of the halting of growth observed in these single-cell permeabilization experiments at the MIC is consistent with the results of the bulk, time-lapse bactericidal assay. The MIC data suggest that the halting of growth at 4 μM of LL-37 is mediated by oxygen. Our working hypothesis is that LL-37 induces formation of harmful reactive oxygen species (ROS). In an earlier study of CM15 [20], we developed two single-cell, real-time fluorescence measurements that monitor oxidative stress using the dyes CellROX Green and Amplex Red. CellROX Green (Life Technologies) is a proprietary, permeable, non-fluorescent, oxidation-sensitive dye. Oxidation produces a species we call CellROX*, which fluoresces in the green, but only when bound to ds-DNA. In vitro, CellROX Green is sensitive to superoxide (•O2–) and to hydroxyl radical (•OH), but not to hydrogen peroxide (H2O2) or to a variety of other oxidants including peroxynitrite (ONOO–), NO, and hypochlorite (OCl–). In the cellular environment, other species such as high-valence Fe centers could also oxidize CellROX Green [27,28]. Amplex Red is a permeable dye whose reaction with H2O2 is catalyzed by the non-native peroxidase APEX2, expressed in the cytoplasm from a plasmid [29,30]. The product is resorufin, which fluoresces in the red. The specificity of the enzymatic reaction strengthens the assumption that resorufin fluorescence signals H2O2 formation. We monitor oxidative stress by measuring single-cell fluorescence of CellROX* or resorufin (in cells expressing APEX2) as a function of time after LL-37 addition. The duration of each complete imaging cycle is 12 s for one-color imaging and 6 s for two-color imaging. First we carried out the CellROX* assay in aerobic conditions using 4 μM of LL-37 (the aerobic MIC). At t = 0 the flow was switched to medium including LL-37 and 2. 5 μM CellROX Green. Laser intensities and imaging conditions were held constant, enabling quantitative intensity comparisons across different experiments. More than 90% of 171 cells from three repeats of the experiment exhibit attenuation of growth rate or abrupt shrinkage within 10 min of injection of LL-37 (example in Fig 2A and 2B and S2 Movie), as was observed in the periplasmic GFP experiments. For the particular cell in Fig 2, halting of growth and mild shrinkage occurs shortly after t = 0. At t = 2 min, CellROX* fluorescence intensity begins to increase gradually; the intensity continues to rise for about ten minutes, when it turns sharply downward and decays to a non-zero plateau. At the same moment, cell length begins a second period of gradual shrinkage. For all cells whose length begins to decrease at t < 30 min, we eventually observe a sudden decrease in CellROX* signal to a non-zero plateau. We show below that this decrease typically correlates in time with the moment of CM permeabilization to the DNA stain Sytox Orange. Those cells that continue to grow slowly over the first 30 min (no CM permeabilization), exhibit a much weaker CellROX* signal that increases slowly throughout the observation period. Control experiments indicate that the peak CellROX* signal induced by LL-37 from 58 analyzed cells is on average ten times larger than the magnitude of the slowly rising green fluorescence signal observed at t = 30 min in the absence of LL-37 (Fig 3A). These data suggest that the strong CellROX* fluorescence begins to rise when LL-37 penetrated the OM and gains access to the periplasm. Experiments using two-color imaging of CellROX* and Rhodamine B-LL-37 will corroborate this inference. To directly confirm that CellROX* begins to rise before CM permeabilization, we carried out two-color fluorescence experiments in aerobic conditions, injecting both 2. 5 μM CellROX Green and 5 nM of the DNA stain Sytox Orange into the growth medium at t = 0. Observations were then carried out for 30 min. A typical example of the two fluorescence traces from a single cell that undergoes CM permeabilization during the observation period is shown in Fig 4A. See also S3 Movie. A strong CellROX* fluorescence begins to rise some 10 min before the abrupt onset of Sytox Orange fluorescence, which in turn marks the moment when the CM is permeabilized. The CellROX* signal then typically decreases abruptly by about 60% and stabilizes at a lower value. As before, the abrupt decrease in CellROX* fluorescence occurred in all cells exhibiting CM permeabilization over 30 min. Most often the CellROX* signal decrease begins within 1 min of CM permeabilization, but in about 1/3 of 39 cells the decrease begins 2–15 min after CM permeabilization. A histogram in shown in S6 Fig. Additional evidence that the onset of strong CellROX* fluorescence coincides with entry of LL-37 into the periplasm comes from two-color imaging experiments using CellROX Green and a red fluorescent variant of the peptide, Rh-LL-37. Earlier we showed that Rh-LL-37 and unlabeled LL-37 have the same MIC vs MG1655 E. coli, although membrane permeabilization events occurred somewhat more slowly for Rh-LL-37 [19]. Under aerobic growth conditions, at t = 0 we flowed 25 μM CellROX Green plus 8 μM Rh-LL-37 across plated E. coli cells. A representative result is shown in Fig 4B. As before [19], weak red fluorescence from Rh-LL-37 initially coats all cells uniformly (plateau of red fluorescence at t = 2–17 min). We attribute this to binding of Rh-LL-37 oligomers to the lipopolysaccharide (LPS) layer. The weak plateau of green fluorescence is a background signal that also coats all cells uniformly. This background is not CellROX* fluorescence, because it occurs on addition of Rh-LL-37 in the absence of CellROX Green. Like LL-37, Rh-LL-37 preferentially attacks septating cells. As observed earlier [19], septating cells gradually exhibit a brighter band of red fluorescence that begins at the septal region and slowly spreads to the entire periplasm over 5–10 min (Fig 4B). In salty solution Rh-LL-37 fluorescence is self-quenched due to bundling of multiple helices. We believe that this self-quenching persists during binding to the LPS layer of E. coli, rendering the initial wave of red fluorescence weak. Entry into the periplasm at the septal region unbundles the helices, causing dequenching of fluorescence and gradual development of the brighter red band. Evidently Rh-LL-37 enters at the septal region and binds strongly to some immobile element of the periplasm, possibly the anionic cross-links within the peptidoglycan layer. As local binding sites become occupied, unbundled Rh-LL-37 slowly migrates towards the tips of the periplasm, observed as a gradual outward spreading of the brighter red band. In earlier work, we showed that Rh-LL-37 binds to purified peptidoglycan [19]. As shown in the example of Fig 4B, the green fluorescence from CellROX* and the brighter band of red fluorescence from Rh-LL-37 in the periplasm rise on the same time scale. This occurred within 30 min for 59 cells from three repeat experiments. We infer that oxidative species are formed gradually, as more and more monomeric Rh-LL-37 copies gain access to the periplasm. In an important control experiment, in earlier work [20] we found that permeabilization of both the OM and the CM using Triton-X (without addition of AMP) did not enhance CellROX* fluorescence. This shows that CM permeabilization alone is not sufficient to trigger the signals of oxidative stress observed after LL-37 treatment. To test for the importance of LL-37 stereochemistry on the magnitude of oxidative effects, we repeated the CellROX* assay in aerobic conditions using 4 μM of the all-D stereoisomer of LL-37. The average peak CellROX* signal level was the same within experimental error (Fig 3B). In addition, our earlier work found the same MIC for D- and L-LL-37 [19]. Our working hypothesis is that in aerobic conditions, LL-37 causes formation of ROS (most likely •O2–) in the periplasm by disrupting the electron transport chain. The disruption begins when LL-37 gains access to the periplasm, which also affords access to the outer leaflet of the CM. The electron transport chain employs a series of membrane proteins embedded in the CM [31]. For exponential growth in aerobic conditions, the primary pathway runs through the two NADH dehydrogenases NDH-I and NDH-II (with complex II dominant), passes through ubiquinone (UB), and terminates at the cytochrome oxidase-bo3 complex [32]. Depending on the level of oxygenation, some fraction of the electron flux terminates at the alternative cytochrome oxidase-bd complex. The–bd complex has much higher affinity for O2 than does–bo3; its expression level increases with decreasing O2 concentration in the growth medium. The terminal oxidase converts O2 to H2O, transferring protons to the periplasm and helping to maintain the proton motive force. At sufficient concentration, CN−binds to the key heme iron in both the -bo3 and the–bd complex, blocking O2 binding, halting oxidative respiration and cell growth, and greatly diminishing the proton-motive force [21]. To test whether aerobic respiration is a prerequisite for LL-37-induced generation of the oxidative stress signals, we pre-incubated WT MG1655 E. coli cells with 1 mM KCN for 5 min prior to injection of 4 μM LL-37 in aerated growth medium. According to previous studies of reconstituted respiration in vitro, at this concentration the–bo3 complex is strongly inhibited, but the–bd complex is more weakly inhibited [33]. The inhibiting concentration in vivo is not known. After pre-treatment for 5 min with cyanide, cells do not grow over a 50-min observation time, suggesting that respiration has been blocked. In two-color imaging experiments, we initiated flow of 4 μM LL-37 with 2. 5 μM CellROX Green and 5 nM Sytox Orange at t = 0. The flow also contained 1 mM KCN to block respiration continuously. For the typical cell shown in S2 Fig, no significant rise of CellROX* or Sytox Orange fluorescence was observed on a 50-min timescale. Evidently LL-37 induced neither ROS formation nor CM permeabilization. We measured the maximum CellROX* intensity from 62 cells in three separate experiments with and without pre-incubation with KCN (Fig 3A). The KCN pre-treatment attenuates the mean CellROX* fluorescence per cell by at least a factor of 5. The KCN treatment also greatly reduces the fraction of cells that exhibit significant Sytox Orange fluorescence over 60 min of LL-37 treatment, from ~100% for normally growing cells to ~30% for KCN-treated cells (S2 Fig). This suggests the possibility that formation of oxidative species, possibly ROS, within the periplasm may enhance the ability of LL-37 to permeabilize the CM. Alternatively, reduction of the transmembrane potential may inhibit the ability of LL-37 to permeabilize the CM, as discussed below. The same pre-treatment of cells with KCN also reduces the cell-killing effects of LL-37. We repeated the LL-37 bactericidal assay after KCN treatment. As shown in S1 Fig, more pre-treated cells survive after 4 μM LL-37 incubation for 30 min, and also for 1 hr. Evidently KCN pre-treatment provides some protection against the deleterious effects of LL-37. However, when 8 μM LL-37 was applied to cells pre-treated with KCN, no growth was observed even after 1 hr. In a similar vein, pre-treatment of the cells with the protonophore CCCP at 200 μM for 10 min completely halted growth. Subsequent injection of 4 μM of LL-37 and CellROX Green caused only a small, slowly rising CellROX* signal, again five times smaller than the peak signal from cells growing aerobically (Fig 3A). We tested for H2O2 induction in aerobic conditions by repeating the flow experiment using the MG1655 strain expressing the non-native peroxidase APEX2 from a plasmid [30]. At t = 0, we flowed 4 μM of LL-37 plus 10 μM Amplex Red and alternated phase contrast and red fluorescence images. Inside some 20% of 121 cells from three repeat experiments, a substantial burst of intracellular resorufin fluorescence is observed (example in Fig 5A and S4 Movie). However, for most cells we observe only a weak intracellular resorufin signal that is difficult to measure above a steadily increasing background of red fluorescence outside the cells (S3 Fig). This strongly suggests that for most cells, resorufin formed in the cytoplasm efficiently escapes the cell envelope. As a result, the intracellular resorufin signal is usually very small. To determine the exact timing of the onset of the strong, intracellular resorufin signals relative to CM permeabilization, we carried out two-color imaging experiments using Amplex Red and the green fluorescent DNA stain Sytox Green. In the minority of cells that exhibit appreciable intracellular resorufin fluorescence, the abrupt onset of red and green fluorescence is essentially simultaneous (Fig 5B and S5 Movie). Evidently the burst of resorufin is produced promptly after the CM is permeabilized. However, once again the majority of cells show little or no intracellular resorufin signal. And once again a background of red fluorescence in the surround rises gradually over time, suggesting that while most or perhaps all cells are producing H2O2 in the cytoplasm, the resulting resorufin usually escapes the cell envelope efficiently (S3 Fig). This makes sense; the OM is typically permeabilized to GFP and smaller species long before the CM is permeabilized to Sytox and other small molecules. To summarize, the typical behavior in time of the CellROX* and resorufin fluorescence signal in aerobic conditions is very different. In most cells, CellROX* fluorescence rises gradually as LL-37 slowly gains entry to the periplasm (before CM permeabilization), and then decreases abruptly by a factor of two or more at the moment of CM permeabilization (Fig 2). In contrast, intracellular resorufin fluorescence rises abruptly at the moment of CM permeabilization (Fig 5). This is observed only in a minority of cells, but red background fluorescence rises gradually over the 30-min experiment. To test the possibility that LL-37 is causing release of oxidants by perturbing proper function of cytochrome oxidase-bo3 or of cytochrome oxidase-bd, we carried out a limited number of microscopy experiments on the deletion mutant strains ΔcyoABCDE (–bo3 deletion mutant) and ΔcydAB (–bd deletion mutant). The strains exhibit aerobic doubling times of 53 min and 45 min, respectively, similar to WT cells (Table 2). The 6-hr MICs are both 4 μΜ (Table 2), the same as the WT strain. The expression levels of the–bo3 and–bd oxidases vary with the availability of O2 in the medium. The–bo3 oxidase, which binds O2 more weakly, is more abundant when the level of dissolved O2 is high [31]. The–bd oxidase, which binds O2 more strongly, is more abundant for low O2 concentrations. As shown for the typical cell in S4 Fig, in aerobic growth the ΔcyoABCDE mutant exhibited strong signals in the CellROX* after injection of LL-37 at 4 μM. The mean CellROX* signal level is slightly larger than for the WT strain (Fig 3B). The same mutant also showed strong resorufin signals in the Amplex Red assay, comparable to the signal shown in Fig 5A (S7 Fig). However, the CellROX* signal level for the ΔcydAB mutant strain is only 27% that of the WT strain (Fig 3B). This result implicates cytochrome oxidase-bd in the mechanism by which LL-37 induces oxidative stress on accessing the periplasmic space. No known terminal electron acceptors are present in the standard EZRDM medium. Cells growing in EZRDM with glucose as carbon source but without oxygen (Methods) and without added nitrate carry out fermentation, synthesizing ATP by glycolysis. The doubling time is essentially the same as in aerobic conditions (Table 2), and the pmf is likely reduced by about 25% [32]. At t = 0, we initiated flow of 2. 5 μM CellROX Green and 4 μM unlabeled LL-37 (the aerobic MIC) over wild-type E. coli growing in fermentation conditions. The rate of cell growth, as judged by cell length in phase contrast images, decreased early on, much as it did in aerobic conditions. However, on the 30-min timescale, under anaerobic conditions 80% of 134 cells from three separate experiments continued to grow, albeit more slowly (example in Fig 2C and 2D). In comparison, on the same 30-min timescale under aerobic conditions only 40% of the cells continued to grow. In these fermentation experiments with LL-37, a small green fluorescence signal, possibly due to CellROX*, was typically observed to rise slowly over 30 min as most cells continued to grow (Fig 2C and 2D). Averaged over 20 cells, the maximum green fluorescence intensity achieved during the 30-min observation period was three times smaller in fermentation than in aerobic growth conditions (Fig 3A). There was no abrupt increase in green fluorescence. In contrast, the green signal in aerobic conditions rises more rapidly (over 5–10 min vs 30 min, Fig 2B vs 2D). We repeated the one-color imaging experiments in fermentation conditions using 4 μM LL-37 along with either Sytox Green or CellROX Green. Over the first 30 min 20% of the 92 cells from three separate experiments exhibited cytoplasmic membrane permeabilization to Sytox Green (Fig 1C) and a halting of growth. Importantly, in the CellROX Green experiments no abrupt rise of CellROX* fluorescence was observed for any of the cells. We also repeated the Amplex Red/APEX2 experiments in fermentation conditions using 4 μM LL-37. For all 117 cells studied, we observed no significant resorufin fluorescence signal, either within the cells or in the extracellular background (example in Fig 5A). We also carried out experiments at 16 μM LL-37 in fermentation conditions, a concentration equal to the 6-hr MIC. Under those conditions, all cells shrink within 30 min. A weak signal from CellROX* again rose gradually over 30 min, but there was no abrupt increase and the average maximum signal after 30 min was comparable to that at 4 μM LL-37 in fermentation conditions (Fig 3A). To summarize, in fermentation conditions, the CM (and presumably the OM) of a subset of cells is permeabilized at 4 μM LL-37 and the CM of all cells is permeabilized at 16 μM LL-37. There is no evidence of the same type of rapidly rising CellROX* and resorufin signals of oxidative stress that were observed in aerobic growth. Cells growing in EZRDM in the absence of oxygen but in the presence of glucose and added NO3– carry out anaerobic respiration using the terminal reductase NarGHI, which reduces NO3– to NO2– [31]. To test whether a functional electron transport chain is sufficient to enable LL-37 to induce the CellROX* and resorufin signals, we carried out analogous fluorescence microscopy experiments on cells growing anaerobically in EZRDM supplemented with 10 mM KNO3. In anaerobic respiration conditions, 4 μM LL-37 induced CM permeabilization to Sytox Green in a significantly smaller fractions of cells than in aerobic growth conditions, on both the 30-min and 60-min timescales (Fig 1C). This is congruent with the increase in MIC (Table 2). On average, the maximum CellROX* signal generated over 30 min was fivefold smaller than in the peak signal in aerobic growth (Fig 3A). No resorufin signal, either intracellular or extracellular, was observed from the Amplex Red assay in any of 49 cells studied. At 16 μM LL-37 (higher than the MIC of 12 μM under anaerobic respiration conditions), all cells exhibited cytoplasmic membrane permeabilization within 15 min. We carried out the Amplex Red assay at this higher LL-37 concentration and again observed no signal whatsoever in any of the 45 cells studied. For E. coli in aerobic conditions, on addition of 10 μM melittin (twice the aerobic MIC) we observed a strong, rapidly rising CellROX* fluorescence signal (S5 Fig). Much weaker, more slowly rising CellROX* fluorescence was observed (S5 Fig) on addition of 0. 9 μM cecropin A (1X the aerobic MIC) and of 32 μM indolicidin (1X the aerobic MIC). The average maximum intensity results are shown in the bar graph of Fig 3B. These data are congruent with a strong increase in MIC from aerobic to anaerobic (fermentation) conditions for melittin, but not for cecropin A nor for indolicidin. This work extends our earlier study of the attack of LL-37 on E. coli [19] to include direct observation of the timing of fluorescence signals that monitor oxidative stress. As shown before, the initial step in the attack is binding of the cationic LL-37 to the anionic lipopolysaccharide (LPS) layer, followed by permeabilization of the outer membrane (OM). For LL-37, we observed significant increases in MIC for cells growing under fermentation and anaerobic respiration conditions compared with aerobic respiration (Tables 1 and 2). In and of themselves, differences in MIC under different growth conditions should be interpreted cautiously. In addition to modulating oxidative stress effects, different growth conditions may also modulate the bacterial membrane composition. Such changes could alter the binding propensity of an AMP for the outer membrane and also the surface concentration of AMP required for membrane permeabilization. The present work shows that for cells growing aerobically, an early green CellROX* signal gradually rises as the LL-37 concentration builds up in the periplasm, but before LL-37 has permeabilized the cytoplasmic membrane (CM) to the small dye Sytox Orange (Fig 4). Compared with the WT strain, the maximum CellROX* signal decreases almost 4-fold in the ΔcydAB deletion mutant but increases slightly (nominal 1. 3-fold) in the ΔcyoABCDE deletion mutant. The signal is attenuated almost 6-fold after pre-treatment with KCN, which is known to inhibit aerobic respiration. The signal level was unchanged using the D stereoisomer form of LL-37. Entry of LL-37 into the periplasm gives the AMP access to the outer leaflet of the CM and to external surfaces of cytoplasmic membrane proteins. These observations are consistent with a proposed mechanism in which LL-37 interferes with the terminal cytochrome oxidase-bd, causing inappropriate release of superoxide (•O2–) into the periplasmic space, where it oxidizes CellROX to CellROX* (Fig 6). Similarly, in vitro studies showed that the cyclic, cationic antimicrobial agent gramicidin S interfered with the activity of cytochrome oxidase-bd, but not with cytochrome oxidase-bo3 [34]. The interference might arise from direct interaction of LL-37 with the oxidase. Observation of the same CellROX* signal level using the L- or D- enantiomer of LL-37 argues against existence of a specific binding pocket within the cytochrome oxidase-bd structure; it does not rule out a non-specific interaction due to electrostatic binding, for example. Alternatively, LL-37 may disrupt -bd function indirectly by perturbation of the membrane environment, perhaps by strong interaction of the polycationic peptide with anionic lipids such as cardiolipin (CL) or phosphatidylglycerol (PG) [35]. Importantly, we know from inadvertent experiments that use of old CellROX samples induces cytoplasmic fluorescence in E. coli even without addition of LL-37. Presumably CellROX had already been oxidized to the fluorescent form CellROX*. This makes it plausible that CellROX* created by LL-37 action in the periplasm is able to permeate the intact cytoplasmic membrane, bind to DNA, and fluoresce, prior to permeabilization of the CM to Sytox Orange or presumably to LL-37 itself. The subsequent CM permeabilization event enables Sytox Orange and presumably LL-37 itself to enter the cytoplasm. CM permeabilization correlates in time with abrupt, partial quenching of the CellROX* fluorescence (by an unknown mechanism, Fig 4A) and the abrupt onset of resorufin fluorescence (Fig 5B), presumed to be formed by reaction of Amplex Red with APEX2 and H2O2. The resorufin signal is detected primarily outside the cells, but some 20% of cells retain substantial resorufin inside the cytoplasm. The detailed mechanism of H2O2 production is unclear. The most plausible source is dismutation of •O2– by the superoxide dismutases (SODs) that reside in the cytoplasm. Once LL-37 has permeabilized the CM to Sytox Orange and to LL-37 itself, the •O2– formed in the periplasm may be able to cross the CM and reach the cytoplasm, where it finds SODs. Alternatively, if the CM becomes permeable to globular proteins, both the SODs and also APEX2 may pass from the cytoplasm to the periplasm, where they find •O2– and produce H2O2 (which itself is permeable). In addition, once LL-37 enters the cytoplasm it may induce formation of additional •O2– by some alternative mechanism. Based on the properties of the dyes in vitro, we tentatively attribute the CellROX* signal that rises before CM permeabilization to production of superoxide (•O2–) and the resorufin signal that rises after CM permeabilization to production of hydrogen peroxide (H2O2). However, enhancement of intracellular oxidants other than •O2– (or •OH) might cause conversion of CellROX Green to CellROX* [27]. The specificity of the APEX2 enzymatic reaction with H2O2 and Amplex Red to form resorufin lends support to the assumption that resorufin fluorescence is signaling an increase in hydrogen peroxide flux. We found considerable evidence that aerobic respiration or a robust transmembrane potential or both are prerequisites for LL-37 to induce the early, gradually rising CellROX* signal and the delayed, abruptly rising resorufin signals (Figs 2–5). In Fig 3 we compare the maximum CellROX* intensity observed over 30 min for various conditions. Both oxidative stress signals are greatly attenuated by pre-treatment of cells with cyanide (which blocks O2 binding to the cytochrome oxidases, halts growth, and diminishes the pmf); by pre-treatment of cells with the protonophore CCCP (which abrogates the pmf); and by exclusion of oxygen from the medium, both in fermentation conditions (no operative electron transport chain) and in anaerobic respiration using NO3– in the medium (enabling a different active electron transport chain). It could be that the oxidants induced on entry of LL-37 to the periplasm enhance the ability of LL-37 to permeabilize the CM. Alternatively, the transmembrane electric field associated with the strong aerobic pmf points in the direction that would enhance the ability of a cationic peptide such as LL-37 to penetrate the low dielectric core of the bilayer and its substantial, repulsive dipole potential [36]. Such an effect was observed previously for cationic antibiotics such as gentamicin [37]. In anaerobic fermentation (Figs 1C, 2C, 2D and 5A) and anaerobic respiration conditions (Fig 1C), the pmf is likely lower than in aerobic growth [24]. In those lower-pmf conditions, 4 μM LL-37 permeabilizes the CM and halts growth in a smaller fraction of cells. Even those permeabilized cells do not exhibit the type of CellROX* or resorufin signals characteristic of the rapid oxidative stress induced in aerobic conditions. It is difficult to pinpoint exactly what aspects of the attack of LL-37 on E. coli cause the halting of growth and the eventual killing of cells. The smaller MIC in aerobic vs anaerobic growth conditions (Table 1) suggests that the observed oxidative stress events contribute, perhaps indirectly. Permeabilization of the CM to small species and concomitant loss of the pmf is undoubtedly an important factor as well. In a recent study of E. coli attacked by an analogue of the AMP PMAP-23, at killing concentrations the copy number of AMP per cell was estimated to be 106–107, and the assay was most sensitive to membrane-bound copies [38]. For the synthetic cationic peptide “ARVA”, the analogous copy number was estimated to be >108 per cell [39]. Even at 106 per cell, the AMP concentration would be 1 mM. If LL-37 also binds to E. coli at such high concentrations, it is easy to imagine multiple harmful processes occurring in parallel. The generality of oxidative stress induction by AMP action on bacteria deserves further exploration. As judged by the MIC in aerobic growth vs fermentation conditions (Table 1), the efficacy against E. coli of LL-37 and melittin depends on oxygen levels, while that of cecropin A and indolicidin does not. In previous work, we observed oxygen-sensitive MICs and time-dependent intra-cellular fluorescence signals indicative of oxidative stress during the attack on MG1655 E. coli by the synthetic AMP CM15 [20] and by the synthetic, highly cationic random β-peptide copolymer MM63: CHx37 [22]. In the present work, the oxidative signals induced by LL-37 decreased significantly on deletion of cytochrome oxidase-bd, but not on deletion of the–bo3 oxidase. This suggests a remarkably specific target of LL-37 activity. It seems likely that different AMPs will prove to induce oxidative stress by different mechanisms. Finally, the enhancement of growth-halting effects of LL-37 in aerobic vs anaerobic conditions suggests the possibility that the degree of oxygenation in specific tissues may help to regulate AMP activity. For example, in the human gut the fraction of strict anaerobes increases from proximal to distal; in the colon, the oxygen partial pressure is only 25% of that in the atmosphere [23]. An earlier study of human β–defensin-1 found the AMP to be much more potent in its reduced, unfolded form against the pathogenic fungus Candida albicans and against anaerobic, Gram positive commensals of the Bifidobacterium and Lactobacillus species [23]. The effect was specific to certain microbial species. In contrast, human β-defensin-3, which is extremely potent in its oxidized, folded form, was less potent under reducing conditions. Like LL-37, hBD-3 is more potent in aerobic conditions. More work is needed, but it is already evident that the degree of oxygenation affects different human AMPs in different ways. Future work will test the generality of the induction of oxidative stress by other natural AMPs. The strains are listed in Table 2. The background (“WT”) strain is MG1655 (K12) in all cases. Experiments on periplasmic GFP used strain JCW10, in which TorA-GFP is expressed from plasmid pJW1 as previously described [26]. TorA-GFP is transported to the periplasm by the twin-arginine transport system and the TorA signal peptide is cleaved, leaving free GFP in the periplasm. ZY01 is the strain that expresses the peroxidase APEX2 from a plasmid introduced into the background strain, as described previously [20]. To construct the deletion mutant strain called ΔcyoABCDE, lacking the gene for cytochrome-bo3 oxidase, we performed λ-Red reconstruction, replacing the cyoABCDE gene with a kanamycin resistance gene. The deletion of the cyoABCDE gene and the replacement by a kanamycin resistance gene were confirmed by PCR and DNA sequencing. The deletion mutant strain ΔcydAB was constructed and confirmed analogously. To visualize resorufin generation in the ΔcyoABCDE deletion mutant, we transformed a pASK-IBA3plus vector containing the APEX2 gene into ΔcyoABCDE, yielding strain ZY02. Unlabeled LL-37 lacking a C-terminal amide was purchased from Anaspec (61302). Rhodamine B-LL-37 (no C-terminal amide) was purchased from Bachem (4049885). The oxidation sensitive dye CellROX Green (C10444) and Amplex Red (A22188) were purchased from Invitrogen. The DNA stains Sytox Green (S7020) and Sytox Orange (S11368) were purchased from Thermo-Fisher Scientific. Bulk cultures were grown in EZ rich, defined medium (EZRDM) [40], which is a MOPS-buffered solution at pH = 7. 4 supplemented with metal ions (M2130; Teknova), glucose (2 mg/mL), amino acids and vitamins (M2104; Teknova), nitrogenous bases (M2103; Teknova), 1. 32 mM K2HPO4, and 76 mM NaCl. Cultures were grown from glycerol frozen stock to stationary phase overnight at 30°C. Subcultures were grown to exponential phase (OD = 0. 2–0. 6 at 600 nm) at 30°C before sampling for the microscopy experiments. The aerobic MIC values for the various AMPs (Table 1) were determined using the broth microdilution method as previously described [19]. Two-fold serial dilutions of LL-37 in 1× EZRDM were performed in separate rows of a polystyrene 96-well plate, with each plate containing an inoculum of E. coli MG1655. The inoculum was a 1: 20 dilution from a bulk culture at midlog phase (OD600 = 0. 5) grown at 30°C. The plate was incubated at 30°C and shaken at 200 rpm in a Lab-Line Orbital Environ Shaker (model 3527) for 6 hr. The MIC value was taken as the lowest concentration for which no growth was discernible (<0. 05 OD) after 6 hr. Anaerobic MIC values (Table 1) were measured on a 96-well plate that was sealed with plastic wrap. Cells were incubated in EZRDM containing protocatechuic acid (PCA) at 10 mM and protocatechuate 3,4-dioxygenase (PCD) at 100 nM to scavenge oxygen [41]. The plate was incubated at 30°C for 6 hr, followed by OD measurements. In the earlier study of CM15 [20], we tested that PCA by itself does not interfere with CellROX* fluorescence. The time-lapse recovery assays utilized an MG1655 culture in bulk. Overnight culture of wild-type MG1655 was inoculated at 1: 100 dilution in 2 mL EZRDM at 30°C. When the culture is at midlog phase (OD600 = 0. 5), the culture was diluted with warmed EZRDM to 1: 10 and incubated with different concentrations (0,4, 8, and 16 μM) of LL-37. 100 uL of each culture was sampled at different time point (30 min, 1-hr. and 2-hr incubation). Then, each culture was 10-fold serial diluted with warmed EZRDM into a 96-well plate. Each dilution was plated into fresh LB agar plates and the plates were incubated at 30°C for 24 hr. The plates were then visually inspected for growth of colonies. The control procedure was the same except that the LL-37 was omitted. See S1 Fig for results. As previously described [20], imaging of individual cells was carried out at 30°C in a microfluidics chamber consisting of a single rectilinear channel of uniform height of 50 μm and width of 6 mm, with a channel length of 11 mm. The total chamber volume is ~10 μL. After bonding of the PDMS chamber to the glass coverslip, 0. 01% poly-L-lysine (molecular weight >150,000 Da) was injected through the chamber for 30 min and rinsed thoroughly with Millipore water. E. coli cells are immobilized on the coverslip but grow normally. During imaging experiments, the chamber was maintained at 30°C with an automatic temperature controller. For aerobic imaging experiments, the medium is exposed to air over three hr while held at 30°C in a shaker bath; this ensures full oxygenation of the medium. In addition, the PDMS ceiling of the microfluidics device is permeable to the ambient gases N2 and O2. For anaerobic imaging experiments, O2 must be prevented from entering the chamber through its ceiling. A small anaerobic chamber surrounding the microfluidics device was constructed of aluminum with a nitrogen gas inlet and outlet. Details are provided elsewhere [20]. Before injection of cells, nitrogen gas flowed through the chamber continuously for 1. 5 hr. E. coli were grown in aerobic conditions until injected into the chamber. Fresh deoxygenated EZRDM was made by treating EZRDM with 50 nM protocatechuate 3,4-dioxygenase (PCD) and 2. 5 mM protocatechuic acid (PCA). This was used to wash the cells at 30°C before plating. Deoxygenated EZRDM (with or without addition of 10 mM KNO3) then flowed across the plated cells for 30 min before injection of antimicrobial peptides and CellROX. The subsequent microscopy imaging experiment was carried out as before. Single-cell imaging was performed on two different microscopes: a Nikon TE300 inverted microscope with a 100×, 1. 3 N. A. phase contrast objective and a Nikon Eclipse Ti inverted microscope with a 100×, 1. 45 N. A. phase contrast objective. For the TE300, images were further magnified 1. 45× in a home-built magnification box. GFP, Sytox Green, and CellROX* were imaged using 488 nm excitation (Coherent Sapphire laser), expanded to illuminate the field of view uniformly. The emission filter was HQ525/50 (Chroma Technology). Resorufin and Sytox Orange were imaged using 561 nm excitation (Coherent Sapphire laser). The emission filter was HQ617/73 (Chroma Technology). Laser intensities at the sample were typically ~5 W/cm2 at 488 nm and ~2. 5 W/cm2 at 561 nm. Fluorescence images were obtained with an EMCCD camera, either Andor iXon 897 or Andor iXon 887. In both cases, the pixel size corresponds to 110 ± 10 nm at the sample. For single color experiments, time-lapse movies of 60-min total duration were obtained as 600 frames of 50-ms exposure time each, with fluorescence and phase contrast images interleaved at 6-s intervals (12 s per complete cycle). For dual color experiments, μManager was used to obtain the data and switch filters between frames using a LB10-NW filter wheel (Sutter). The time-lapse movies of 35-min total duration were obtained as 1050 frames of 50-ms exposure time each, with green fluorescence (488 nm excitation), red fluorescence (561 nm excitation), and phase contrast images interleaved (6 s per complete cycle). To minimize spectral bleed-through in the two-color experiments, we utilized the narrower filters HQ510/20 for the green channel and HQ600/50M for the red channel. CellROX Green (Life Technologies) is a proprietary oxidation-sensitive dye whose fluorescence quantum yield at 500–550 nm after excitation at 488 nm increases dramatically on oxidation in the presence of ds-DNA. It readily permeates both E. coli membranes. The manufacturer tested its sensitivity to different reactive oxygen species in the presence of ds-DNA in vitro including hydroxyl radical (•OH), superoxide (O2–), hydrogen peroxide (H2O2), peroxynitrite (ONOO–), nitric oxide (NO), and hypochlorite (ClO–). The only two oxidizing agents that significantly enhanced CellROX* fluorescence were hydroxyl radical and superoxide. Importantly, hydrogen peroxide has no effect. In the CellROX* imaging experiments, MG1655 cells were injected into the microfluidics chamber. After allowing 5 min for plating of cells, the bulk solution was washed away with fresh, pre-warmed, aerated EZRDM. After the wash, cells were grown for 5 min prior to the injection of 4 μM LL-37 plus 2. 5 μM CellROX Green. To maintain good aeration and steady bulk concentrations, the medium with LL-37 and CellROX Green flowed continuously at 0. 3 mL/hr. As previously described [20], the assay for single-cell, time-resolved measurement of H2O2 production following LL-37 treatment is based on the well-established Amplex Red method [30]. Some peroxidases (but not the catalases naturally occurring in E. coli) catalyze reaction of the dye Amplex Red with H2O2 to form the red fluorescent species resorufin (λem = 585 nm). Recently Collins and coworkers [29] adapted the method to carry out the Amplex Red + H2O2 reaction inside the cytoplasm by inserting a plasmid that expresses the peroxidase APEX2 (mutated ascorbate peroxidase). Their method detects H2O2 produced inside the cell using plate-based bulk fluorescence measurements with time resolution of ~60 min. Here we use intracellular APEX2 combined with single-cell, time-resolved detection by fluorescence microscopy. This enables sensitive detection of intracellular H2O2 production with 12-s time resolution and correlation of LL-37-induced H2O2 production with other events in real time.

Antimicrobial peptides play a significant role in the innate immune response of plants and animals, including humans. While it is well known that AMPs can permeabilize bacterial cell membranes, a growing body of evidence indicates that they cause a variety of additional deleterious effects. Here we use single-cell imaging methods to study the induction of oxidative stress in live E. coli by several natural cationic AMPs, including the human cathelicidin LL-37. Strong fluorescence signals indicative of oxidative stress correlate with smaller minimum inhibitory concentrations (MICs) in aerobic vs anaerobic growth conditions. A detailed mechanistic study suggests that LL-37 disrupts the proper flow of electrons through the electron transport chain, releasing oxidative species into the periplasm. Based on these results, we suggest that the degree of aeration in different tissue types may be used by the host to modulate AMP efficacy.

Abstract Introduction Results Discussion Materials and methods

periplasm fluorescence imaging oxidative stress metabolic processes microbiology fermentation deletion mutagenesis molecular biology techniques mutagenesis and gene deletion techniques cellular structures and organelles research and analysis methods stress signaling cascade imaging techniques metabolism molecular biology cytoplasm biochemistry signal transduction cell biology biology and life sciences cell signaling signaling cascades

2017

Oxidative stress induced in E. coli by the human antimicrobial peptide LL-37

Hepatic circadian gene transcription is tightly coupled to feeding behavior, which has a profound impact on metabolic disorders associated with diet-induced obesity. Here, we describe a genomics approach to uncover mechanisms controlling hepatic postprandial gene expression. Combined transcriptomic and cistromic analysis identified hundreds of circadian-regulated genes and enhancers controlled by feeding. Postprandial suppression of enhancer activity was associated with reduced glucocorticoid receptor (GR) and Forkhead box O1 (FOXO1) occupancy of chromatin correlating with reduced serum corticosterone levels and increased serum insulin levels. Despite substantial co-occupancy of feeding-regulated enhancers by GR and FOXO1, selective disruption of corticosteroid and/or insulin signaling resulted in dysregulation of specific postprandial regulated gene programs. In combination, these signaling pathways operate a major part of the genes suppressed by feeding. Importantly, the feeding response was disrupted in diet-induced obese animals, which was associated with dysregulation of several corticosteroid- and insulin-regulated genes, providing mechanistic insights to dysregulated circadian gene transcription associated with obesity. Precise temporal expression of hepatic enzymes is crucial for metabolic homeostasis, and a major part of hepatic circadian protein synthesis is regulated by precisely timed gene transcription and mRNA translation [1]. This is controlled by complex interactions between rhythmic endocrine signaling, fluctuations of the body temperature, oscillating metabolites, and intrinsic circadian networks [2]. It is well established that genetic disruption of the intrinsic cellular clock, controlled by transcription factors such as brain and muscle Arnt-like protein (BMAL), circadian locomotor output cycles kaput (CLOCK), RAR-related orphan receptor (ROR), and reverse c-erbA (REVERB), has a profound impact on circadian gene transcription in the liver [3–6]. However, rhythmic gene transcription can be restored by controlled feeding regimens [7,8], emphasizing the significance of a food-entrainable oscillator for circadian hepatic gene transcription [2]. Importantly, obesity-associated conditions such as hepatosteatosis, insulin resistance, and diabetes are linked to disruption of circadian transcriptional networks in the liver [9]. For example, rodents with ad libitum access to a high-fat diet (HFD) tend to eat in the physical active phase (nighttime) as well as in the resting phase (daytime), in contrast to ad libitum chow-fed mice, eating primarily during the night/active phase [10–12]. Ad libitum HFD feeding leads to a striking disruption of rhythmic gene expression in the liver [13,14] that can be prevented by simple night-restricted feeding (NRF) of HFD [15]. This is associated with diminished HFD-induced obesity, insulin resistance, and diabetes [15], emphasizing the importance of continuous circadian feeding–fasting cycles to prevent obesity-associated disorders. Thus, it is crucial to understand the mechanisms controlling diurnal gene expression and to determine the direct impact of feeding. Housing rodents in a 12-hour light/dark cycle and restricting feeding to the active phase (nighttime) is associated with rhythmic hepatic chromatin remodeling [16], transcription [1,17], translation [1], post-translational modifications [18–20], protein translocation [21], and hepatocyte morphology [22]. Most of these studies used omics-based technology to characterize rhythmicity and genetic disruption of intrinsic circadian regulators to gain comprehensive mechanistic insights to the rhythmic regulation. However, as oscillating molecular processes such as transcription are controlled by a number of additional factors, including feeding–fasting cycles, we lack considerable genome-wide knowledge of regulatory mechanisms controlling rhythmic transcription by factors operating together with the intrinsic molecular clock. To specifically investigate the mechanism by which daily feeding–fasting cycles affect circadian gene expression, we focused on a pre- and postprandial time point at the junction between the light and dark phase of a circadian rhythm. By tampering with food availability, we could isolate a set of circadian-regulated genes and enhancers that were primarily regulated by feeding. Analysis of chromatin accessibility and histone acetylation suggested that the majority of feeding-repressed genes is regulated by changed activity of signaling pathways regulating the glucocorticoid receptor (GR) and Forkhead box O1 (FOXO1). These factors occupy chromatin in the preprandial state, and occupancy is reduced postprandially, leading to reduced enhancer activity. Preprandial injection with dexamethasone (dex) and/or insulin receptor antagonist (S961) demonstrated that the pancreas and the hypothalamic–pituitary–adrenal (HPA) axis regulate specific and overlapping transcriptional programs in the liver, which collectively control a significant part of hepatic circadian gene transcription. To specifically evaluate the transcriptional effects of feeding in the liver during a night-restricted experimental setup, we trained mice to NRF. The subsequent experiments were designed to specifically focus on the transition between the resting (lights on, zeitgeber time [ZT]0–ZT12) and the physical active phase (lights off, ZT12–ZT24) with and without access to food (Fig 1A). Livers from night-restricted–fed mice were isolated at ZT10 and ZT14 to monitor the transcriptional effects of feeding and the transition from light to dark. In parallel, isolation of livers from mice (at ZT14) that did not receive food at ZT12 allowed us to analyze the effect of feeding. We initially evaluated mRNA levels of a few genes known to be either controlled by the intrinsic circadian clock or by food intake. All of these genes were regulated by the transition from ZT10 to ZT14-fed (Fig 1B and Fig 1C). When food was omitted, the mRNA levels of the core circadian clock genes Bmal1, Cry1, Reverba, and Dbp were not significantly different from levels observed in fed mice. In contrast, the level of mRNAs coding for key metabolic enzymes (glucokinase [GCK], fatty acid synthase [FASN], and phosphoenolpyruvate carboxykinase 1 [PCK1]) and a protein involved in lipoprotein and triglyceride metabolism angiopoietin-like 4 (ANGPTL4) was dependent on food intake (Fig 1C). To extend this analysis, we performed RNA-seq experiments on livers isolated at ZT10, ZT14-unfed, and ZT14-fed and focused on genes that were induced (n = 472) and repressed (n = 432) at a false discovery rate (FDR) < 0. 01 by the transition from ZT10 to ZT14-fed (Fig 1D and Fig 1E, respectively). The genes in both groups were ranked by their fold change between ZT10 and ZT14-unfed and grouped into five bins of equal size (Q1–Q5). Q1 and Q2 represent genes that were primarily differentially regulated by the transition from ZT10 to ZT14, regardless of food intake, whereas Q4 and Q5 contain genes primarily regulated by food intake. In agreement with the reverse transcription quantitative PCR (RT-qPCR) data (Fig 1B), Q1 and Q2 contain a number of known core circadian clock genes such as Bmal, Per3, Cry1, Dbp, Clock, and Reverba. Moreover, Q1 and Q2 contain genes involved in heat shock response (e. g. , Hspb1), fatty acid metabolism (e. g. , Acss2, Acsl1, and Abcd2), and cholesterol uptake (e. g. , Abcg8). Q4 and Q5 contain genes such as Gck, Srebp1, Fasn, and Agpat2 (up-regulated by feeding) involved in glucose uptake and fatty acid synthesis and storage and genes such as Pck1 and G6pc involved in gluconeogenesis (down-regulated by feeding). Gene ontology (GO) analysis showed that genes regulated independent of feeding are involved in protein folding, response to heat, and circadian regulation of gene expression (Fig 1F). In addition, GO analysis also identified enriched biological pathways involved in processes such as glucose, glycogen, fatty acid, and cholesterol metabolism. Some of these metabolic pathways were also enriched in the group of genes regulated by feeding. However, some pathways such as triglyceride and ketone body metabolism were specifically enriched in the genes regulated by feeding (Fig 1F). To determine whether the different groups of genes were expressed dynamically throughout a circadian rhythm, we compared our data with previously published circadian RNA-seq data from mice subjected to NRF [1]. More than 80% of the genes, identified by us to be differentially expressed from ZT10 to ZT14-fed, were also expressed in a circadian manner (Fig 1G), and most induced and repressed genes had circadian expression zenith at ZT16–21 and ZT5–10, respectively (Fig 1H). Moreover, average expression analysis of the mRNAs in the different gene groups, defined in Fig 1D and 1E (Q1–Q5), suggested that the circadian expression pattern is very similar between genes regulated by feeding (Q4 and Q5) and genes regulated independent of feeding (Q1 and Q2) (Fig 1I). Thus, a large number of circadian-regulated genes are controlled directly by feeding, indicating that the diurnal feeding response is an important driver of circadian gene expression in the liver. DNase-accessible regions of chromatin harbor putative binding sites for transcription factors regulating gene expression [23]. To identify mechanisms regulating gene expression controlled by feeding, we profiled DNase-accessible regions genome-wide in livers from fed (n = 2) and unfed (n = 2) mice at ZT14. To account for possible differences in the digestion efficiency of DNase, we sequenced libraries from nuclei digested with two different concentrations of DNase I (60 U and 80 U) and subsequently combined all the sequencing data to identify DNase hypersensitive sites (DHSs) irrespective of treatment. (Correlation between biological replicates is shown in S1 Fig) Using this approach, we identified a total of 83,592 DHSs, in agreement with other DNase-seq experiments in mouse liver tissue [24,25]. To specifically probe activity of these putative regulatory regions, we performed histone 3 lysine 27 acetylation (H3K27Ac) chromatin immunoprecipitation-sequencing (ChIP-seq) on livers isolated from three unfed (ZT14-unfed) and three fed (ZT14-fed) animals. (Correlation between biological replicates is shown in S2 Fig) For example, most of the DHSs identified near the genes Fasn, Srebf1, Igfbp1, and Pck1 did change histone acetylation level in response to feeding (Fig 2A and 2B). Genome-wide quantification of H3K27Ac identified a little less than 1,500 DHSs associated with increased H3K27Ac and around 1,900 DHSs associated with decreased H3K27Ac at FDR < 0. 1, n = 3 (Fig 2C). DNase accessibility at these differentially acetylated DHSs was also changed in response to feeding (Fig 2D), and we observed a genome-wide correlation between changed H3K27Ac and DNase accessibility of the DHSs associated with feeding-regulated H3K27Ac (rde = 0. 82, Fig 2E). The correlation between DNase accessibility and H3K27Ac was weaker for all identified DHSs (rall = 0. 29). In addition, we found enrichment of DHSs associated with feeding–up-regulated H3K27Ac near feeding-induced genes compared to DHSs associated with unchanged H3K27Ac (Fig 2F, p = 2. 95 × 10−06, Kolmogorov–Smirnov test). Likewise, DHSs associated with feeding–down-regulated H3K27Ac was enriched near feeding-repressed genes compared to DHSs associated with unchanged H3K27Ac (Fig 2G, p = 2. 81 × 10−09, Kolmogorov–Smirnov test). Collectively, this indicates that the identified DHSs are involved in regulating nearby gene expression, and to uncover transcription factors involved in the feeding response, we mined enriched DNA sequences at DHSs associated with feeding-regulated H3K27Ac. To analyze DHSs primarily associated with feeding-regulated H3K27Ac and DHSs mostly associated with H3K27Ac regulated independently of feeding, we first identified DHSs associated with differential H3K27Ac in livers from mice euthanized at ZT10 and ZT14-fed (FDR < 0. 1, n = 3). H3K27Ac ChIP-seq tags were subsequently quantified at these DHSs from livers isolated at ZT10 (n = 3), ZT14-unfed (n = 3), and ZT14-fed (n = 3) and by hierarchical clustering grouped into two major clusters according to the circadian rhythm and four subclusters according to the feeding status (Fig 3A). Clusters 1 and 3 represent DHSs H3K27 acetylated by processes other than feeding, whereas clusters 2 and 4 were feeding-regulated (Fig 3A). To analyze circadian H3K27Ac at the DHSs differentially H3K27 acetylated between ZT10 and ZT14-fed, we used previously published circadian H3K27Ac ChIP-seq data from livers of mice subjected to NRF [16]. Quantified H3K27Ac at DHSs were normalized and averaged for the four different clusters (Fig 3B). The average H3K27Ac circadian profile of the different clusters were similar (Fig 3B), demonstrating that the feeding-regulated H3K27Ac followed a similar circadian profile as H3K27Ac uncoupled from the feeding response, in agreement with the transcriptomic data (Fig 1I). To identify putative transcription factors associated with differential H3K27Ac, we searched for enriched DNA motifs from known transcription factor–binding sequences curated by hypergeometric optimization of motif enrichment (HOMER) [27]. The most enriched motifs in clusters 1 and 3 were DNA sequences known to interact with BMAL and CLOCK (binding to E-BOX motifs), RAR-related orphan receptor C (RORC), heat shock factors (heat shock response element [HRE]), and REVERB (Fig 3C). Quantification of the motif score between the different clusters confirmed specific enrichment of the RORC and REVERB motifs in cluster 3 (Fig 3D). The HRE was enriched in cluster 3 compared to clusters 1 and 2; however, no significant difference could be observed between clusters 3 and 4. Also, the E-BOX and interferon regulatory factor (IRF) motifs were enriched in cluster 1 compared to clusters 3 and 4; however, no significant enrichment was observed between cluster 1 and cluster 2 (Fig 3D). To support these findings, we evaluated the contribution of a particular DNA motif to H3K27Ac (referred to as motif activity) of DHSs using IMAGE [26]. This analysis suggested that the E-BOX, HRE, and RORC motifs contributed to differential H3K27Ac in livers from mice at ZT10 compared ZT14, irrespective of feeding (Fig 3E). Thus, DNA motifs shown to interact with transcription factors of the core circadian clock network were specifically enriched in the DHSs associated with differential H3K27Ac uncoupled from feeding. The DHSs associated primarily with feeding-regulated H3K27Ac (clusters 2 and 4) were enriched for motifs known to interact with the GR, cAMP Responsive Element Binding Protein (CREB), Forkhead box transcription factors (FOX), GATA, and direct repeats (DR-1) known to interact with transcription factors such as Hepatocyte Nuclear Factor 4 (HNF4), Retinoid X receptor (RXR), and peroxisome proliferator-activated receptor (PPAR) [28]. Quantification of the motif score showed that the GR response elements (GRE), CREB response elements (CRE), and FOX motifs were specifically enriched in cluster 2 (Fig 3D), and IMAGE analyses indicated that the GRE and FOX motifs contributed significantly to reduced H3K27Ac in response to feeding (Fig 3E). Moreover, de novo motif analysis of DHSs associated with feeding-regulated H3K27Ac shown in Fig 2C confirmed enrichment of GRE, CRE, and FOX motifs at DHSs associated with feeding-repressed H3K27Ac (S3 Fig). Collectively, the DNA motif analysis indicated that circadian H3K27Ac is regulated by a combination of core clock transcription factors (e. g. , BMAL, CLOCK, and REVERB), heat shock responsive transcription factors, and transcription factors regulated by endocrine signaling (e. g. , GR, CREB, and FOXO1). To correlate motif enrichment analysis to transcription factor occupancy, we chose to map transcription factors potentially interacting with DHSs associated with feeding-repressed H3K27Ac (Fig 3 and S3 Fig). The identified FOX motifs interact with a range of FOX transcription factors, including FOXO1, known to be negatively regulated by insulin signaling by mechanisms including nuclear exclusion and interaction with transcriptional coregulators [29]. Thus, we mapped FOXO1 together with GR, known to interact with the enriched GRE (Fig 3 and S3 Fig). The CRE motif interacts with CREB, which is post-translationally regulated by glucagon signaling. CREB occupancy of chromatin has been reported previously in 24-hour fasted livers [30,31], and we used these data for analysis of CREB occupancy of chromatin. ChIP-seq identified more than 10,000 GR peaks and more than 8,000 FOXO1 peaks in livers from unfed mice at ZT14. De novo motif analysis of the individual transcription factor ChIP-seq peaks identified their cognate DNA-binding motif in addition to motifs for the linage-determining transcription factors HNF4 and C/EBP (S4A Fig), suggesting that the ChIPs enrich for GR and FOXO1-binding sites. Moreover, the specificity of the FOXO1 antibody used for ChIP was validated by western blotting (S4B Fig). Feeding resulted in significant genome-wide reduction of GR and FOXO1 occupancy (Fig 4A and 4B). Interestingly, in unfed livers, GR co-occupied more than 60% of FOXO1 binding sites genome wide, whereas little overlap was observed between CREB and GR–FOXO1 (Fig 4C, left). This pattern of co-occupancy was observed irrespective of the CREB data set used for the analysis (S4C Fig). Strikingly, FOXO1 and GR co-occupancy was even more pronounced at DHSs associated with reduced H3K27Ac in response to feeding, as defined in Fig 2C (Fig 4C, right). Furthermore, CREB occupancy of these regions is mostly associated with co-occupancy of GR and/or FOXO1. Receiver operating characteristic (ROC) analysis suggested that the level of GR and FOXO1 occupancy in unfed conditions are better predictors of feeding-reduced H3K27Ac compared to CREB (Fig 4D). This suggests that loss of H3K27Ac in response to feeding is primarily caused by reduced GR and FOXO1 occupancy. More than 60% of the DHSs associated with feeding-reduced H3K27Ac were occupied by GR, FOXO1, and/or CREB. If we decreased the FDR for calling differentially regulated H3K27Ac, we observed an increased frequency of GR, FOXO1, or CREB occupancy (Fig 4E), suggesting that DHSs associated with the most robustly feeding-regulated H3K27Ac were more likely occupied by at least one of those three transcription factors. GR and FOXO1 occupancy were most pronounced at distal DHSs repressed by feeding (Fig 4F), whereas CREB preferentially occupied DHSs in the proximal promoter regions (transcription start site [TSS] +/− 2 kb). Correlation with feeding-regulated genes demonstrated that more than 80% of the feeding-repressed genes were associated with at least one GR and/or FOXO1-binding site within 50 kb of the TSS (Fig 4G), corresponding to more than 2-fold enrichment compared to random genes in the genome. Interestingly, a relative high frequency of feeding-induced genes was also associated with at least one GR and/or FOXO1-binding site within 50 kb of the TSS (S4D Fig); however, the enrichment of GR and/or FOXO1 near feeding-induced genes relative to random genes was less pronounced compared to feeding-repressed genes. Moreover, the number of GR and FOXO1 peaks (S4E Fig) and the tag density (S4F Fig) of the peaks near feeding-induced genes were significantly lower compared to GR and FOXO1 peaks identified near feeding-repressed genes. We also observed higher enrichment of feeding-repressed genes with at least one CREB-binding site within 50 kb of the TSS compared to feeding-induced genes (Fig 4F). However, we could not observe any difference in the tag density of CREB peaks (S4F Fig) and the amount of CREB peaks (S4E Fig) near feeding-repressed genes compared to feeding-induced genes. Fig 4H illustrates examples of GR, FOXO1, and CREB occupancy near genes repressed by feeding (Pck1, Angptl4, Tat, and Insig2). Notice that most regions are associated with considerable co-occupancy of GR and FOXO1. In summary, these data indicate that a large fraction of DHSs associated with feeding-repressed H3K27Ac are occupied by GR and FOXO1 in the preprandial state. Feeding leads to reduced occupancy of GR and FOXO1, possibly resulting in postprandial repression H3K27Ac and attenuated transcription of nearby genes. Recruitment of GR to chromatin and subsequent chromatin remodeling is strictly dependent on the glucocorticoid level in the surrounding environment [32,33]. In ad libitum–fed rodents, corticosterone levels surge toward the end of the resting phase (zenith at ZT10-ZT12) when the animals are fasting and declines when animals enter the physical active/dark phase [34]. Thus, postprandial reduction of GR occupancy of the genome is possibly mediated by reduced levels of circulating corticosterone as a result of feeding. Indeed, in our experimental setting, feeding resulted in reduced corticosterone levels in serum (Fig 5A). To determine the significance of the reduced levels of glucocorticoids, we injected dex or vehicle immediately before feeding to short circuit the feeding response and collected livers at ZT14 from fed or unfed animals (Fig 5B). GR occupancy and H3K27Ac were subsequently probed by ChIP at ZT14 in unfed and fed conditions. Importantly, control injections (vehicle) did not perturb feeding-repressed corticosterone levels and feeding-induced insulin levels (Fig 5C and 5D), suggesting that the injection protocol did not interfere with the feeding response. Injection of dex augmented postprandial GR occupancy of several GR-binding sites in the genome (Fig 5E) tested by ChIP-qPCR, and genome-wide analysis indicated that the vast majority of GR binding to chromatin was increased in the fed state after dex injection (Fig 5F and 5G). For example, suppression of GR occupancy near Pck1 and Insig2 in response to feeding was reversed by dex injection (Fig 5H). To identify the functional significance of altered glucocorticoid levels, we quantified H3K27Ac at DHSs (with and without GR occupancy) after dex injection. We observed a general trend of increased H3K27Ac at the DHSs occupied by GR, in contrast to DHSs not occupied by GR (Fig 5I). This suggests that the variation of glucocorticoid concentrations leads to dynamic regulation of H3K27Ac at DHSs occupied by GR. Interestingly, however, not all GR occupied DHSs were acetylated at H3K27 in response to dex injection. For example, a GR-binding site upstream of Pck1 was H3K27 acetylated in response to dex, whereas H3K27 acetylation of the GR binding site in the proximal promoter was largely unaffected (Fig 5H). Similarly, feeding-repressed H3K27 acetylation of GR-binding sites near Insig2 persisted after dex treatment (Fig 5H), indicating that a subset of GR-occupied enhancers was unresponsive to dynamic glucocorticoid signaling. To systematically compare the constituents of DHSs associated with differential H3K27Ac in response to glucocorticoid treatment, we divided DHSs occupied by GR into three groups depending on the H3K27Ac responsiveness to dex (Fig 5J). The DHSs associated with most pronounced increase in H3K27Ac in response to dex were also associated with a significant reduction of H3K27Ac and GR occupancy in response by feeding (Fig 5K). This analysis was confirmed by a linear correlation analysis showing a negative correlation between increased H3K27Ac in response to dex and reduction of H3K27Ac and GR occupancy in response to feeding (S5A and S5B Fig). Moreover, dex-induced H3K27Ac correlated positively with the level of GR occupancy (Fig 5L and S5C Fig) and the GR motif strength (Fig 5M and S5D Fig). In agreement, the level of GR occupancy and the GRE motif strength are better predictors of dex-induced H3K27Ac compared to FOXO1 and CREB occupancy and presence of their respective motifs (S5E Fig). Interestingly, GR occupied DHSs associated with low H3K27Ac in response to dex showed prominent FOXO1 and CREB occupancy (S6A and S6B Fig), indicating that these transcription factors may contribute to feeding-regulated H3K27Ac of the weak GR-binding sites. Feeding-repressed genes are enriched for nearby DHSs associated with reduced H3K27Ac, which is functionally linked to GR occupancy and dynamic glucocorticoid levels (Fig 4D and Fig 5K), suggesting that GR is a putative regulator of feeding-controlled gene transcription. However, since GR binding per se is not indicative of feeding-regulated H3K27Ac, it is likely that many genes, harboring DHSs occupied by GR, are nonresponsive to circulating corticosteroids. To test this, we evaluated expression of genes harboring nearby occupancy of GR. Feeding-mediated repression of Tat, Fkbp5, and Pck1 expression was restored or partially restored upon dex treatment (Fig 5H and S6D and S6E Fig), whereas dex did not reestablish expression of Angptl4, Insig2, and G6pc (Fig 5H and S6F and S6G Fig), despite considerable nearby occupancy of GR. Genome-wide analysis by RNA-seq showed that dex rescued expression of 22% of the feeding-repressed genes at FDR < 0. 05 (Fig 5N), demonstrating that dynamic glucocorticoid levels controlled a subset of genes regulated by feeding. To link dynamic gene expression with H3K27Ac at DHSs, we quantified the number of feeding-regulated genes harboring nearby DHSs associated with dex-regulated H3K27Ac (high, medium, and low) and compared this to analysis of random selected genes (i. e. , relative gene enrichment). Relative enrichment of genes with nearby occupancy of dex-induced H3K27Ac was particularly evident for genes regulated by dex (Fig 5O). Genes not regulated by dex were less enriched for nearby dex-induced H3K27Ac at DHSs. Collectively, this demonstrates that reduced levels of circulating corticosterone controls genome-wide GR occupancy of chromatin. Subsequent regulation of H3K27Ac depends on the strength of the underlying GRE and level of GR occupancy and ultimately determines the regulation of nearby target genes. Expression of genes associated with glucocorticoid-regulated H3K27Ac at DHSs was more likely controlled by circulating glucocorticoids. It is evident from the above analysis that feeding-repressed gene expression is regulated partly by dynamic GR signaling, and additional signaling pathways regulating transcription factors such as FOXO1 and CREB activity are involved. To address the importance of insulin signaling, we injected insulin receptor antagonist S961 immediately before feeding and isolated livers two hours after feeding at ZT14 (S7A Fig). Treatment with S961 resulted in hyperglycemia and hyperinsulinemia but did not affect corticosterone levels (Fig 6A), indicating that circulating insulin and corticosterone levels are regulated independently of each other. Moreover, acute disruption of insulin signaling resulted in elevated levels of postprandial nonesterified fatty acids (NEFAs), likely as a result of perturbed insulin-mediated repression of adipocyte lipolysis. Interestingly, expression of the dex-unresponsive genes, Angptl4, Insig2, and G6pc (S6C Fig), was restored in the liver of fed animals injected with S961, whereas expression of the dex-responsive genes Tat, Pck1, and Fkbp5 was unaffected by S961 (S6C Fig). These results indicate that insulin signaling regulates expression of a subset of feeding-regulated genes, whereas glucocorticoids preferentially control expression of other feeding-regulated genes. To characterize these transcriptional programs genome wide, we performed RNA-seq on livers from unfed and fed mice injected preprandial with S961 and S961 together with dex (n = 4 for all treatment groups). Coinjection resulted in hyperglycemia and hyperinsulinemia and suppressed corticosterone levels as a result of negative feedback inhibition of the HPA axis (Fig 6A). Initially, we evaluated the acute pharmacological effect on RNA expression of the clustered circadian-regulated genes (Q1–Q5, down-regulated from ZT10 to ZT14-fed) defined in Fig 1E. In agreement with the initial separation of circadian-regulated genes into feeding and nonfeeding-regulated, we observed that genes primarily down-regulated by the intrinsic circadian program were not affected by feeding (Q1 and Q2) (Fig 6B). In contrast, the mRNA levels of genes down-regulated by the feeding response (Q4 and Q5) were reduced in response to feeding. Treatment with S961 or dex resulted in increased expression of feeding down-regulated genes (Q4 and/or Q5) in fed condition, whereas intrinsic circadian-regulated genes were unaffected (Q1 and Q2). This was augmented when animals were treated with dex in combination with S961. To illustrate the cooperative effect of dex and S961 on feeding-repressed genes, we identified all feeding-repressed genes in mice injected with vehicle (at FDR < 0. 05,369 genes were suppressed by feeding) and, in parallel, analyzed differential mRNA expression in response to dex, S961, and dex + S961. From this analysis, we extracted the feeding-repressed genes and illustrated the response to dex and/or S961 as M (log ratio) and A (mean average) (MA) plots (Fig 6C). This demonstrated that coinjection of S961 and dex resulted in a more pronounced induction of mRNA expression compared to individual treatments. In agreement, principle component analysis showed that the combined action of dex and S961 resembled the mRNA expression pattern of unfed animals (Fig 6D), indicating that the cooperate action of the two separate signaling pathways controls postprandial suppression of gene transcription. To identify clusters of genes preferentially regulated by dex and/or S961, we identified all feeding-repressed genes and separated genes regulated by S961, dex, and/or dex + S961 at FDR < 0. 05. Based on this analysis, we binned the genes into five major clusters. Cluster 1 did not show any statistically significant response to treatment (Fig 6E). Two clusters of genes responded either to dex (cluster 2) or S961 (cluster 3), demonstrating that the pancreatic and HPA-signaling pathways to some degree operate specific transcriptional programs in the liver. For example, Pck1, Tat, and Klf15 were activated primarily by dex, whereas Angptl4, G6pc, and Insig2 were regulated preferentially by S961, in agreement with the RT-qPCR data (Fig 6F and 6G, respectively). Moreover, we did observe clusters of genes that were activated by both dex and S961 independently (cluster 4) and when dex and S961 were given in combination (cluster 5), suggesting that insulin and corticosteroid coregulate a subset of genes, either additively or synergistically. This includes genes such as Igfbp1 and Ulk1 (Fig 6H). To verify these findings and to determine liver-specific effects of insulin receptor and GR signaling, we performed similar sets of NRF experiments in mice harboring liver-specific disruption of GR (L-GRKO) (S7B–S7D Fig) and previously characterized mice with liver-specific disruption of Insulin Receptor Substrate 1 (IRS1) and IRS2 (L-IRSdKO) [35]. Quantification of RNA expression by RNA-seq of feeding-repressed genes regulated by dex and/or S961 confirmed that the expression of genes regulated by dex was primarily disrupted in livers from L-GRKO mice (S7E Fig), and the expression of genes regulated by S961 was perturbed in livers from L-IRSdKO mice (S7F Fig). To determine whether any of the clusters described in Fig 6E were predominantly enriched for GR or IRS1/2 regulated genes, we identified differentially regulated genes in the livers of L-GRKO mice and L-IRSdKO mice relative to wild-type (WT) controls at FDR < 0. 05, n = 4. This demonstrated that the differentially regulated genes in the L-GRKO mice were enriched in cluster 2 and cluster 4 (Fig 6I), corresponding to dex-regulated genes. In contrast, the differentially regulated genes in the L-IRSdKO mice were enriched in cluster 3 and cluster 4, corresponding to S961-regulated genes (Fig 6J). Interestingly, genes synergistically regulated by dex and S961 (cluster 5) were less likely regulated in the L-GRKO model and the L-IRSdKO models (Fig 6I and 6J and S7E and S7F Fig), suggesting that insulin- and corticosterone-mediated regulation of these genes requires both signaling pathways to be intact to maintain regulation. Alternatively, this group of genes may be primarily controlled by signaling through extrahepatic tissues. For example, S961 disrupts feeding-regulated NEFA levels, likely by perturbed insulin-mediated repression of lipolysis in adipose tissue, impacting lipid metabolism and gluconeogenesis in the liver [36], and may also affect hepatic gene expression. In summary, these data indicate that postprandial repression of hepatic gene expression is controlled by cooperative action of insulin and corticosterone signaling. Circadian transcription in liver from diet-induced obese animals is severely disrupted by a number of metabolic pathways interacting with the intrinsic circadian clock machinery [9,37]. It is known that mice on an HFD have disordered feeding patterns and abnormal transcriptional rhythmicity [12–14] and that NRF of HFD reduces obesity and minimizes the risk of developing insulin resistance and diabetes [15]. Thus, the disrupted circadian hepatic gene expression observed in HFD-induced obese animals is likely strongly associated with a disrupted circadian feeding response. To investigate the circadian feeding response in diet-induced obese animals, we fed mice an HFD for 10 weeks. HFD feeding resulted in significant weight gain (S8A Fig) and resulted in reduced glucose tolerance (Fig 7A). Obese and lean control mice were trained to NRF, and livers were subsequently isolated from unfed and fed lean and obese mice at ZT14 (S8B Fig). In agreement with the reduced glucose tolerance, liver triglycerides (Fig 7B) and insulin levels (Fig 7C) were significantly increased in unfed obese animals at ZT14 compared to lean unfed controls. Moreover, preprandial corticosterone levels were severely reduced in unfed obese animals (Fig 7D) compared to lean controls. This effect was independent of the duration of HFD feeding (S8C Fig). To examine the effect of HFD on feeding-regulated gene expression in liver, we performed RNA-seq on livers isolated from unfed and fed lean and obese animals at ZT14. Although RNA expression analysis demonstrated disrupted feeding-regulated expression of a number of genes (Fig 7E–7G), the cause of impaired feeding response was gene specific. For example, repression of Ppara, Insig2, Acot1, and Angptl4 expression by feeding in lean animals was decreased in obese animals as a result of elevated mRNA expression in the fed obese animals compared to lean controls (Fig 7E). And expression of Pck1, G6pc, Igfbp1, and Tat was significantly decreased in unfed obese mice compared to lean controls, leading to a reduced feeding response (Fig 7F). A similar tendency could be observed for feeding-induced gene expression. For example, feeding-induced expression of Arrdc3 was blunted as a result of increased expression in the liver of unfed obese animals, whereas feeding-induced Fasn expression was impaired as a result of reduced Fasn expression in fed obese animals (Fig 7G). To probe genome-wide effects on gene expression, we identified all differentially regulated genes in response to feeding in chow-fed animals at FDR < 0. 01 (n = 3) and subsequently performed hierarchical clustering of all biological replicates (n = 3 for each condition) from fed and unfed chow and diet-induced obese animals. This approach is independent of a specific FDR cutoff, which allowed us to group genes despite some variability between biological replicates. Hierarchical clustering revealed eight distinct clusters of feeding–up-regulated and feeding–down-regulated genes (Fig 7H), and differential gene expression analysis was performed to confirm disrupted feeding response of specific clusters (S8D Fig). Five of these gene clusters (3,4, 5,6, and 7) showed clear impaired feeding response in obese animals (Fig 7H and S8D Fig). Two clusters of feeding-induced genes demonstrated an impaired feeding response with seemingly different mechanisms. Expression of genes in cluster 3, including the transcription factor Srebp1c and its target genes Fasn, Aacs, Sqle, and Ldlr, was reduced in the liver of fed obese animals compared to lean controls. However, whereas Srebp1c expression was significantly increased in obese unfed animals compared to lean controls, the Srebp1c target genes including Fasn, Aacs, Sqle, and Ldlr were not (Fig 7G and S8E Fig). Expression of genes in cluster 7 was increased in obese animals in unfed condition to a level similar to fed lean animals. This demonstrates mechanistic differences of impaired feeding-mediated induction of gene expression in the liver of obese animals. Likewise, three clusters (4,5, and 6) of impaired feeding-repressed genes could be identified (Fig 7H). Expression of genes in cluster 4 was reduced in unfed obese animals (Fig 7I), resulting in a pronounced impaired feeding response compared to genes in cluster 1 (Fig 7I and S8D Fig). Cluster 4 contains genes such as Pck1, Tat, and Klf15, which we characterized to be regulated by the HPA axis during the circadian feeding response (Fig 6F). In agreement, genes in cluster 4 were significantly enriched for genes regulated by dex and not S961 (Fig 7J) in contrast to genes in cluster 1,5, and 6. In agreement with reduced corticosterone levels in obese animals, we observed reduced GR occupancy to several GR-binding sites (S9A Fig), likely resulting in the attenuated expression of GR target genes in obese animals. In contrast, expression of genes such as Insig2 and Angptl4 in clusters 5 and 6 were increased in fed obese animals compared to fed lean controls, and these clusters of genes were primarily enriched for genes regulated by S961 (Fig 7I and 7J). This indicates that expression of these genes is increased as a result of increased activity of insulin-repressed transcription factors such as FOXO1. However, FOXO1 occupancy of chromatin at several FOXO1-binding sites was suppressed in unfed obese animals compared to lean controls (S9B Fig). This suggests that the increased preprandial insulin level observed in the short-term night-restricted fed obese mice was able to impair FOXO1 binding to chromatin. Thus, under these experimental conditions, FOXO1 is likely not involved in the observed dysregulated postprandial expression of genes such as Ppara, Insig2, and Angptl4, and this also explains why expression of FOXO1 target genes such as G6pc and Igfbp1 was suppressed in unfed obese animals compared to lean controls. Hepatic circadian gene transcription is coupled with environmental cues such as physical activity, light sensed by the suprachiasmatic nucleus (SCN), and feeding behavior. Here, we demonstrate that a large fraction of the circadian transcriptome in the liver is regulated by feeding, and we identified thousands of DNase accessible regions differentially acetylated at H3K27 in response to feeding at the light/dark transition. Importantly, feeding-regulated transcription correlated with differential H3K27Ac DNase-accessible regions. Moreover, we could divide regulatory regions into two major groups: one group regulated by feeding and one group controlled independent of feeding. Both groups had a similar overall circadian H3K27Ac profile. Extensive DNA motif analysis of feeding-regulated enhancers as well as diurnal regulated enhancers uncoupled from feeding uncovered putative transcription factor networks operating at the transition between the unfed and fed state of a circadian rhythm. In this study, we functionally addressed postprandial repression of gene expression and H3K27 acetylation. We found enrichment of GR, FOXO1, and CREB DNA–binding motifs specifically within DHSs associated with feeding-repressed H3K27Ac and observed that GR and FOXO1 occupancy of chromatin was significantly suppressed by feeding. In addition, the GR and FOXO1 cistromic analysis suggested considerable coregulation of enhancer activity and gene expression by upstream signaling pathways regulating GR and FOXO1 activity. The surge of circulating insulin observed after feeding leads to protein kinase B (PKB/AKT) -dependent FOXO1 phosphorylation and subsequent nuclear exclusion [38] in agreement with the observed reduced postprandial FOXO1 occupancy of chromatin. In parallel, we observed reduced corticosterone levels after feeding in an insulin-independent manner. GR occupancy of chromatin is highly dependent on the level of circulating corticosterone [39]. Accordingly, preprandial injection of dex resulted in recruitment of GR to chromatin in the fed state, suggesting that the level of circulating glucocorticoids is a major determinant for postprandial suppression of GR occupancy of chromatin. Interestingly, despite a genome-wide reappearance of GR occupancy upon preprandial dex injection, only a subset of GR-bound feeding-regulated enhancers was reacetylated at H3K27. Accordingly, only a subset of the feeding-repressed genes was induced by dex, suggesting that GR controls a group of postprandial repressed genes by selective enhancer regulation. Enhancers responsive to dynamic glucocorticoid regulations were highly enriched for canonical GREs and relatively high levels of GR occupancy, suggesting that direct interaction of GR with its canonical binding site drives histone acetylation of specific enhancers. In contrast, other enhancers occupied by GR via degenerated GREs are less likely to be regulated by dynamic glucocorticoid signaling. Thus, apart from being a synchronizer of the intrinsic circadian clock in the liver [40,41], GR also directly regulates expression of a number of circadian-expressed genes controlled by daily fed/fasting rhythms. Injection of an insulin receptor antagonist identified postprandial repressed genes controlled predominantly by insulin receptor signaling. Interestingly, several of these insulin-regulated genes were not affected by dynamic glucocorticoid signaling, suggesting that postprandial suppression of circulating corticosterone and increased insulin levels independently regulate the mRNA levels of a subset of feeding-regulated genes in the liver. This includes genes such as Pck1, Tat, Insig2, and G6pc, regulated primarily by glucocorticoids (Pck1 and Tat) or insulin receptor signaling (G6pc and Insig2). Yet studies using primary hepatocytes and/or conditional KO models have shown that Pck1 expression can be regulated by insulin receptor signaling, and G6pc and Insig2 expression can be regulated by GR [42–44]. Thus, we cannot rule out that under certain experimental conditions, genes regulated primarily by glucocorticoids are also under the control of insulin signaling and vice versa. Additionally, increased levels of NEFA in response to insulin resistance may activate transcription factors such as PPARα in the liver [45], which could also contribute to the regulation of genes such as G6pc, Angplt4, and Insig2 [46]. When postprandial insulin and glucocorticoid signaling were independently disrupted, we observed partial disruption of postprandial repression of gene expression; however, combined disruption of these signaling pathways resulted in a postprandial expression profile approximating preprandial gene expression. This demonstrates cooperative actions of insulin and corticosterone signaling to repress postprandial gene expression in the liver and emphasizes that signals operating independent of insulin receptor signaling are involved in postprandial repression of transcription in the liver. These signaling pathways may be involved in hepatic metabolic homeostasis of animals lacking both FOXO1 and upstream components suppressing FOXO1 activity such as AKT, IRS, or IR [35,42,47,48]. For example, FGF15/19 has been shown to reduce circulating corticosterone levels [49] and may act in parallel with insulin receptor signaling to regulate postprandial hepatic gene expression [50]. FGF15/19 is secreted from the small intestine in response to feeding, and signaling via the FGFR4/betaKlotho receptor is important for increased postprandial hepatic glycogen storage, protein synthesis, and decreased gluconeogenesis [51,52]. Transcriptional repression of gluconeogenic genes by FGF15/19 involves suppression of CREB, FOXO1, and PPARG Coactivator 1 (PGC-1) activity [51,53]; however, it has yet to be determined whether other transcription factors, such as GR, are involved. Leptin is another possible postprandial suppressor of circulating corticosterone [54]; however, leptin secretion from adipose tissue is dependent on insulin [55], suggesting that leptin is not involved in the glucocorticoid-specific regulation of postprandial gene expression in the liver, independent of insulin signaling. Circadian gene expression in the liver is altered in diet-induced obese animals, partly linked to perturbed feeding patterns [12–14]. To study the feeding response in diet-induced obese animals, we replicated the experimental feeding setup in lean and corresponding diet-induced obese mice. Using this approach, we observed a clear perturbed feeding response in obese animals. For example, we observed suppressed postprandial lipogenic gene expression (e. g. , Fasn and Aacs) and disrupted preprandial gluconeogenic gene expression (e. g. , G6pc and Pck1. Interestingly, however, the disrupted feeding response in this experimental setup contradicts the general notion that diet-induced hyperinsulimia and insulin resistance leads to increased expression of lipogenenic and gluconeogenic genes by unresolved mechanisms [56]. The discrepancy between this study and other studies may be linked to the short-term night-restricted experimental setup used, which could increase insulin sensitivity. Hence, the preprandial hyperinsulinemia could lead to activation of insulin signaling, abrogating FOXO1 activity. Alternatively, as diet-induced obesity has been shown to dramatically change the circadian gene expression profile [14], the inconsistency may be linked to isolation of livers at different time points during the day. Consistent with other studies, we observed increased preprandial Srebp1c expression in diet-induced obese animals in agreement with hyperinsulimia, yet expression of SREBP1c targets were not induced in the preprandial state nor in response to feeding. This discrepancy may be a result of high pre- and postprandial expression of Insig2, a suppressor of SREBP1c activity [57]. Because FOXO1 occupy a number of regions near the Insig2 gene, elevated Insig2 expression may be a result of hepatic insulin resistance and increased postprandial FOXO1 activity. However, we observed that FOXO1 occupancy of chromatin is disrupted in diet-induced obese animals in a night-restricted feeding setup, suggesting that Insig2 expression is not regulated by FOXO1 under these conditions. Instead, Insig2 may be controlled by increased expression of Ppara [46], found to be up-regulated in the fed diet-induced obese animals compared to controls. The observed suppressed FOXO1 occupancy of chromatin in diet-induced obese animals contradicts a number of reports showing increased FOXO1 activity in hyperinsulimic and insulin resistant diet-induced obese animals evaluated by, for example, nuclear abundance [58] and state of phosphorylation [59]. However, as mentioned above, we cannot rule out that the short-term night-restricted feeding regime used in this study increased insulin sensitivity of the liver in diet-induced obese animals, enabling Akt dependent FOXO1 phosphorylation by elevated insulin levels in the preprandial state. In line with this, we observed that postprandial repression of well-known FOXO1 target genes (e. g. , G6pc and Igfbp1) was blunted as a result of reduced preprandial expression. Unbiased clustering uncovered a group of feeding-regulated genes suppressed in the preprandial state of diet-induced obese animals. This group of genes was particularly enriched for glucocorticoid-sensitive genes. This correlated with reduced preprandial corticosterone levels in obese animals. Dampened circadian corticosterone levels have been reported in diet-induced obese mice [12] and obese humans [60], although other reports have indicated a positive correlation between obesity and corticosterone/cortisol levels [61]. Interestingly, obese, hyperglycemic leptin deficient (ob/ob) mice have elevated circadian corticosterone levels [62], demonstrating a clear difference in the activity of the HPA axis between ob/ob mice and diet-induced obese C57BL/6 mice. Yet ob/ob mice have reduced 11beta-hydroxysteroid dehydrogenase (11β-HSD) activity, likely leading to reduced production of active corticosterone in the liver [63], suggesting that both of these models of obesity may have impaired preprandial GR activity. A body of evidence clearly shows that treatment with glucocorticoids or hypersecretion of cortisol lead to obesity, insulin resistance, and type II diabetes [64], demonstrating that elevated glucocorticoids induce metabolic dysregulation. But clearly metabolic dysregulation caused by diet-induced obesity alone does not necessarily lead to hypercortisolemia. Thus, suppressed preprandial corticosterone levels observed in diet-induced obese animals may be a compensatory mechanism to, for example, impede hyperglycemia when the total energy balance of the animal is in surplus. Collectively, this study identified hepatic circadian–regulated genes and enhancers controlled by feeding. Specifically, we show that two specific signaling pathways control postprandial repression of hepatic gene expression. On one hand, corticosterone levels decrease in response to food intake, leading to reduced expression of specific genes. This is governed by selective regulation of H3K27Ac by GR, mediated by GR recruitment to canonical GREs. In parallel, food intake leads to increased insulin levels, resulting in reduced expression of a set of glucocorticoid independent genes. Both pathways are disrupted during a feeding response in diet-induced obese animals, and impacting postprandial gene regulation in the liver. Thus, a major part of the hepatic circadian–regulated gene program, controlled by feeding, is operated by combined dynamic glucocorticoid and insulin signaling. All mouse work was approved by the Danish Animal Inspectorate (case number 2014−15−0201−00437 and 2017-15-0201-01232), the Danish Environmental Protection Agency, and the Children’s Hospital Boston Institutional Animal Care and Use Committee. Approximately 300 μl of blood was collected postmortem and stored on ice for further analysis. The blood was centrifuged at 10,000 rcf for 10 minutes at 4°C and prepared for insulin, corticosterone, and NEFA measurements, as indicated by UltraSensitive Mouse Insulin ELISA kit (Crystal Chem), the Corticosterone ELISA kit (Enzo Life Sciences), and the NEFA quantification kit using the ACS-ACOD method (Wako #434–91795, #436–91995, #270–77000) according to manufacturer instructions. Liver triglyceride levels were measured as described previously [66]. Approximately 5 mg of liver tissue was homogenized using Ultra-Turrax, and RNA was purified using TRIzol-RNA lysis reagent (Thermo Fisher) in EconoSpin columns (Epoc Life) according to manufacturer instructions. The amount of 1 μg of total RNA was exposed to 10 U DNAse I (Thermo Fisher) for 15 minutes at 37°C, followed by cDNA synthesis using random primers and Moloney murine leukemia virus reverse transcriptase (Thermo Fisher). Expression levels were measured by qPCR using SYBR Green reagent mix (Roche 06924204001) and primers listed in S2 Table. The qPCR was performed on a LightCycler480 (Roche). The expression was normalized to TFIIB expression. Approximately 20 mg of liver was shortly homogenized in 700 μl lysis buffer (50 mM Tris-HCl [pH 6. 8], 10% glycerol, 2. 5% SDS, 10 mM beta-glycerolphosphate, 10 mM NaF, 0. 1 mM NaOrthovanadate, 1 mM PMSF, and 1x protease inhibitor cocktail) and treated with benzonase (Sigma E1014). Protein concentration was determined using the Pierce BCA protein assay kit (Thermo Fisher 23225). A total of 80 μg protein was mixed with laemmli buffer and separated by SDS-PAGE and blotted onto a PVDF membrane (10600023, Amersham Hybond P0. 45 PVDF, lot#G9889898). The membrane was probed with primary antibody against GR (sc-1004, Santa Cruz), FOXO1 (sc-11350, Santa Cruz), TFIIB (sc-225, Santa Cruz) or β−Tubulin (05–661 Merck Millipore), and secondary HRP conjugated antibody (DAKO). In short, ChIP was performed as described previously [66] with minor modifications. Frozen livers (approximately 50 mg per IP) were homogenized in 1% Formaldehyde PBS solution using Ultra-Turrax homogenizer on level 5 for 10 seconds, with following cross-linking at room temperature for 10 minutes. The suspension was quenched by adding 0. 125 M glycine and incubating additional 10 minutes at room temperature. Cross-linked cells were washed in PBS, resuspended in lysis buffer (0. 1% SDS, 1% Triton X-100,0. 15 M NaCl, 1 mM EDTA, 20 mM TrisHCl [pH 8. 0], and BSA 1 mg/ml), and sonicated using Bioruptor (Diagenode) or ME220 Focused-ultrasonicator (Covaris). Chromatin was immunoprecipitated over night at 4 °C using antibodies and Protein A/G agarose beads (Santa Cruz, sc-2003). H3K27Ac ChIP was done with 0. 2 μl/IP of H3K27Ac antibody Ab4729 (Abcam). GR ChIP antibody cocktail consisted of 1 μg/IP of MA1-510 (Thermo Fisher), 1 μg/IP of PA1-511a (Thermo Fisher), and 1μg/IP of sc-1004 (Santa Cruz). FOXO1 ChIP was done with 3 μg/IP of sc-11350 (Santa Cruz). Immunocomplexes were washed, and chromatin was eluted (1% SDS with 0. 1 M NaHCO3) and decross-linked over night at 65 °C. DNA was phenol/chloroform purified and ethanol precipitated. Recovery was analyzed by qPCR using primers listed in S2 Table and/or sequenced. Livers were isolated from 3–4 mice in each treatment group (ZT14-fed or ZT14-unfed), and nuclei were immediately purified from a pool of 200–300 mg liver tissue using a previously described protocol [24]. This was repeated to form a second biological replicate. Purified nuclei were resuspended in buffer (15 mM Tris-HCl [pH 8. 0], 15 mM NaCl, 60 mM KCl, 1 mM EDTA, 0. 5 mM EGTA, 0. 5 mM Spermidine, and protease inhibitors) in a final concentration of 10 million nuclei per milliliter. DNase digestions were performed by adding 100 μl 10X digestion buffer (60 mM CaCl2 and 750 mM NaCl) containing 40 U, 60 U, or 80 U of DNase I (Sigma). Digestions were incubated for 3 minutes at 37 °C, and reactions were terminated by addition of 1 volume of stop buffer (50 mM Tris-HCl, 100 mM NaCl, 0. 1% SDS, 100 mM EDTA and 50 μg/ml Proteinase K [Ambion]). Digested chromatin was incubated at 55 °C for 2 hours and stored at 4 °C until further use. DNase I digestion efficiency was evaluated by qPCR, and samples with optimal digestion efficiency were incubated with 90 μg/ml RNase A (Sigma) for 30 minutes at 37 °C before 50- to 500-bp DNA fragments were purified using ultracentrifugation. DNA was subsequently phenol/chloroform purified and ethanol precipitated. Samples treated with 60 U and 80 U of DNase I were sequenced. De novo motif analysis, log odds motif score, and enriched motifs were identified using HOMER [27]. A set of 2,000 randomly picked DHSs from the total amount of DHSs identified in liver was used as background for the motif enrichment analysis. The GC content of the randomly selected DHSs was 48%, while the mean GC content of DHSs in clusters 1–4 (Fig 3A) ranged from 46% to 49%. Motif analysis using IMAGE was performed using quantified H3K27Ac ChIP-seq reads at DHSs and RNA-seq reads at genes as described previously [26]. Three biological H3K27Ac ChIP-seq replicates were used from livers of mice euthanized at ZT10, ZT14-fed, and ZT14-unfed. Gene expression input for IMAGE was based on replicate RNA-seq data from ZT10, ZT14-fed, and ZT14-unfed. Statistical analysis was performed using Student t test, Mann–Whitney–Wilcoxon test, and Kolmogorov–Smirnov test, as indicated in figure legends. FDR for NGS-read counts at genes, DHSs, and ChIP-seq peaks between different conditions were calculated using DESeq2 [68]. Enrichment of DNA motifs was calculated by HOMER [27] and IMAGE [26]. GO enrichment was performed by GOseq [69]. All sequencing data can be found at GEO: GSE119713. Previously published sequencing data used for analysis include the following: Circadian RNA-seq: GSE73554 [1]; Circadian H3K27Ac ChIP-seq: GSE60430 [16]; and CREB ChIP-seq: GSE45674 [31] and GSE72084 [30].

The liver is an essential organ regulating metabolic homeostasis in response to fluctuations of metabolites induced by daily rhythms of food intake. Homeostasis is maintained by precise dynamic regulation of signaling pathways controlling a wealth of enzymatic reactions involving lipid, bile acid, amino acid and glucose synthesis, storage, and redistribution in hepatocytes. Precise temporal expression of hepatic enzymes is crucial for metabolic homeostasis; a major part of circadian hepatic protein expression is regulated by precisely timed gene transcription. Here, we use a genomics approach to identify genes and regulatory regions of the genome involved in feeding-regulated gene expression. We find that transcription factors acting downstream of glucocorticoid and insulin signaling are enriched at regulatory regions repressed by feeding. Importantly, insulin and glucocorticoid signaling operate to cooperatively control the majority of feeding-mediated gene repression, and these signaling pathways are dysregulated in diet-induced obesity impacting dynamic hepatic gene expression.

Abstract Introduction Results Discussion Materials and methods

body weight medicine and health sciences forkhead box diabetic endocrinology gene regulation regulatory proteins dna-binding proteins hormones endocrine physiology circadian oscillators physiological parameters obesity transcription factors chronobiology insulin proteins endocrinology gene expression biochemistry physiology genetics protein domains biology and life sciences insulin signaling

2018

Insulin signaling and reduced glucocorticoid receptor activity attenuate postprandial gene expression in liver

Bacterial pathogens are frequently distinguished by the presence of acquired genes associated with iron acquisition. The presence of specific siderophore receptor genes, however, does not reliably predict activity of the complex protein assemblies involved in synthesis and transport of these secondary metabolites. Here, we have developed a novel quantitative metabolomic approach based on stable isotope dilution to compare the complement of siderophores produced by Escherichia coli strains associated with intestinal colonization or urinary tract disease. Because uropathogenic E. coli are believed to reside in the gut microbiome prior to infection, we compared siderophore production between urinary and rectal isolates within individual patients with recurrent UTI. While all strains produced enterobactin, strong preferential expression of the siderophores yersiniabactin and salmochelin was observed among urinary strains. Conventional PCR genotyping of siderophore receptors was often insensitive to these differences. A linearized enterobactin siderophore was also identified as a product of strains with an active salmochelin gene cluster. These findings argue that qualitative and quantitative epi-genetic optimization occurs in the E. coli secondary metabolome among human uropathogens. Because the virulence-associated biosynthetic pathways are distinct from those associated with rectal colonization, these results suggest strategies for virulence-targeted therapies. Urinary tract infection (UTI) is a highly prevalent infectious disease with a well-known female predilection and a high incidence of recurrence [1]. E. coli is responsible for up to 85% of community-acquired UTI, and previous studies suggest that the same E. coli strain can cause recurrent UTI' s despite initial antibiotic treatment [2], [3], [4]. UTI has classically been considered to follow inoculation of the bladder through urethral ascension of rectal flora [5]. Urethral ascension to the bladder is considered to be more common in women due to their shorter urethral length and facilitated by mechanical effects on the urethra during intercourse, which is a major risk factor for UTI. The events preceding clinical UTI where colonization progresses to symptomatic bacteriuria are poorly understood and difficult to study. Whether selection of UTI-associated strains from gut E. coli populations is stochastic or the result of intrinsic strain properties has been the subject of multiple investigations. Genes involved with iron acquisition routinely emerge as correlates of urinary pathogenesis in these studies. In one such study, a genome-wide search in the model uropathogen UTI89 revealed extensive selection of 29 genes including those involved in synthesis of the siderophore enterobactin [6]. These siderophore genes were also subject to increased transcription during experimental urinary tract infection [7]. Finally, PCR-based studies of candidate virulence factor genes have identified high frequencies of siderophore receptor genes among urinary isolates although expression of the corresponding siderophores was not determined [8], [9]. Siderophores are a chemically diverse family of small molecules that are produced by a wide variety of pathogenic and non-pathogenic bacteria to scavenge ferric iron (Fe3+) [10]. During iron scarcity, these bacteria synthesize and secrete siderophores, which avidly bind ferric iron and become internalized. Siderophores effectively compete with mammalian iron storage proteins and may be of particular importance in acquiring this critical nutrient during infection [11]. Additional horizontally-acquired genes facilitating siderophore biosynthesis may confer new or enhanced properties that may render a bacterium more virulent. To date, genes for various combinations of four genetically distinct siderophore systems have been described in clinical E. coli isolates with enterobactin being the only system conserved in all isolates (Table 1). Among the non-conserved siderophores, the synthesis of salmochelin is encoded in the iroA gene cluster, involving the IroB-mediated glucosylation and IroE-mediated linearization of enterobactin. The Yersinia high pathogenicity island (HPI) encodes the genes necessary for the synthesis and uptake of yersiniabactin. Aerobactin biogenesis is encoded in the iucABCD cluster of genes. In this study we have used a quantitative metabolomics approach together with microbiologic, genomic, and clinical strategies to uncover a preferential metabolic signature among E. coli isolates from the urines of women with recurrent UTI (urinary E. coli). Comparisons of coincident urinary and rectal strains from patients with recurrent UTI revealed that urinary strains exhibited significantly higher production of yersiniabactin and salmochelin, even amongst genotype-positive strains, but not enterobactin and aerobactin. Also, the siderophore receptor genotype did not always correspond to production of the associated siderophore, in contrast to previous assumptions [8], [12], [13]. Thus, a quantitative metabolomic approach revealed important differences in siderophore production not detectable by genotyping alone. Our analysis of the metabolomic network necessary for siderophore biosynthesis revealed that in addition to its role in salmochelin biogenesis, IroE also converts a conserved siderophore (enterobactin) into a more virulent one (linearized enterobactin) better suited to the infectious microenvironment. These studies demonstrate that E. coli strains associated with recurrent urinary tract infection have a preferred metabolomic profile involving a complex metabolic network. Siderophore production in 18 previously characterized UPEC strains [14] isolated from the urine of women with UTI was examined. To determine what siderophores are expressed by these E. coli isolates, we compared culture supernatants from strains grown for 18 hours in iron-poor and iron-rich minimal media (Fig. 1, Table 2). Comparison of full scan LC-MS profiles from each growth condition revealed a more abundant metabolite signature in iron-poor cultures, consistent with induction of siderophore expression during iron scarcity. rUTI2 was chosen as a model strain to develop a quantitative metabolomic approach because it produced all four known E. coli siderophores. Thus, in iron-poor culture supernatants of strain rUTI2 we identified chromatography peaks corresponding to the [M+H]+ ions of aerobactin (14), salmochelin (15), and enterobactin (16), and the [M−2H+Fe (III) ]+ ion of ferric yersiniabactin (17). These siderophore peaks elute from a reversed phase column in the order reported previously (19). Confirmatory structural information was available by comparing the m/z difference between the [M+H]+ of salmochelin and its precursor, enterobactin. The salmochelin [M+H]+ is 342 m/z units greater (Fig. 2), consistent with enterobactin di-C-glucosylation and trilactone hydrolysis catalyzed by IroB and IroE, respectively (20,21). To further confirm the identity of presumptive siderophore ions, rUTI2 was grown in defined minimal media in which 13C3-glycerol or 15N-ammonium sulfate were substituted for the unlabeled compounds. This heavy isotope labeling strategy resulted in mass shifts for each ion peak based on the number of carbons or nitrogens in their empiric formulae (Fig. 2). Labeling efficiency was high and no unlabeled siderophores or M+1 or M+2 carbon isotope distributions were observed for the most abundant 13C-labeled enterobactin, salmochelin, and aerobactin ions. The prominent M+1 and M+2 ions uniquely present in the 13C-labeled ferric yersiniabactin spectrum are consistent with the presence of iron and sulfur in this species. In this sample, the presence of a monoisotopic 13C ion clearly differentiates the M+1 ion from 57Fe (base peak contains 56Fe) and the M+2 ion from 34S (base peak contains 32S). After mixing labeled and unlabeled supernatants, the labeled and unlabeled siderophore ions all co-eluted, consistent with their expected identical structures. This isotope labeling technique provides both structural confirmation and a source of stable isotope labeled internal standards for MS-based quantification. MS/MS fragmentations were also studied for further structural confirmation using strain rUTI2 (Table S1). The enterobactin [M+H]+ at m/z 670 fragmented predominantly at the ester bonds to yield dihydroxybenzoyl serine monomer (m/z 224) and dimer (m/z 446) as previously reported [15]. In contrast, MS/MS spectra of salmochelin gave the distinct lower molecular weight species expected from fragmentation within the glucose moieties, a finding supported by unchanged neutral losses from 15N-labeled products. Consistent with the hallmark C-glucosylation in salmochelin, we observed no loss of glucose (neutral loss of m/z 162) as is typically seen with O- or N-linked sugars. The aerobactin [M+H]+ at m/z 565 fragmented to give the neutral losses of water (m/z 547) and HCOOH (m/z 519) of a multiply hydroxylated and carboxylated compound. MS/MS of ferric yersiniabactin [M−2H+Fe (III) ]+ gave a complex spectrum, as expected from a heterocyclic compound, that included the prominent m/z 489 peak observed in previous MALDI spectra [16]. Together, these ion fragmentation patterns were consistent with the known structures of the corresponding siderophores. These MS/MS fragmentations were used to quantify siderophores in a multiplexed LC-MS/MS assay. The sequenced model uropathogen UTI89 was observed to produce enterobactin, salmochelin, and yersiniabactin. To validate the stable isotope dilution LC-MS/MS metabolomic assay, we analyzed UTI89 strains with deletion mutations in selected siderophore biosynthetic genes (Fig. 3A). Ions corresponding to the catecholate siderophores enterobactin and salmochelin were absent in an entB [17] mutant while yersiniabactin production was preserved. Conversely, the ybtS [18] mutant produced enterobactin and salmochelin but not yersiniabactin. Because ybtS encodes a salicylate synthase, yersiniabactin expression could be restored in UTI89ΔybtS by growth in the presence of exogenous 0. 3 mM sodium salicylate (data not shown). Selective loss of salmochelin was observed with deletion of iroB, which forms C-glucose bonds with enterobactin [19]. These findings show that metabolomic profiling is sensitive to alterations in siderophore biosynthetic pathways. Single deletion mutants in the siderophore biosynthetic pathways described above (entB, ybtS, iroB) remained positive for siderophore production by the chrome azurol S plate assay based on blue-to-yellow transformation surrounding streaked colonies [20] (Fig. 3B). To determine whether enterobactin, salmochelin, and yersiniabactin together correspond to the total siderophore activity expressed by UTI89, we constructed a UTI89ΔentBΔybtS double mutant which was predicted to selectively abolish synthesis of all known siderophores in UTI89. Metabolomic profiling of UTI89ΔentBΔybtS confirmed the absence of all three siderophores in this mutant and the chrome azurol S assay revealed unchanged colony growth without siderophore production. Thus, for UTI89, total siderophore activity is accountable using this metabolomic analysis as confirmed using a combined biochemical, genetic, and chemical approach. Unlike the K12 E. coli strains described previously [21], [22], a single mutation within the enterobactin gene cluster is insufficient to abolish siderophore production from UTI89. To compare siderophore expression phenotype to bacterial genotype, we examined siderophore production by three fully sequenced E. coli strains MG1655 [23], UTI89 [6], and CFT073 [24] and in our panel of 18 previously genotyped clinical E. coli urinary isolates [14] described above (Table 2). The yersiniabactin receptor (fyuA) and aerobactin receptor (iutA) genotypes were known for all 21 strains. Of the 16 fyuA-positive strains, three pyelonephritic strains, CFT073, pyelo1, and pyelo3 produced no detectable yersiniabactin despite producing other siderophores. Of the 8 iutA-positive strains, two, rUTI4 and pyelo3, produced no detectable aerobactin while still producing other siderophores. The inability of CFT073 to synthesize yersiniabactin is presumably due to mutations that have been identified within essential yersiniabactin biosynthetic genes in this strain [25]. Thus, receptor genotype does not consistently predict the cohort of siderophores that are presented by an organism in response to low-iron conditions. To determine whether urinary E. coli strains exhibited preferential siderophore expression when compared to distinct, coincident rectal strains, we collected a new set of E. coli strains from 18 recurrent UTI patients and we PFGE-typed the isolates (this is a distinct set of isolates from those described in Table 2). From this collection, we identified 13 patients in whom distinct, coincident urinary and rectal PFGE types associated with a UTI were recovered. Clinical characteristics of these 13 patients are described in Table 3. Coincident rectal strains were defined as those isolated from rectal swabs obtained up to one month prior to isolation of a distinct urinary PFGE type. Thus, for this study 14 urinary and 16 rectal PFGE types were chosen from this set of 13 patients. Similarity analysis based on PFGE typing revealed marked diversity (Fig. S1). Two patients yielded isolates with the same PFGE type (Patients 34 and 50, Table 4), one from urine and the other from the rectum. All 30 urinary and rectal strains from these 13 patients exhibited similar growth patterns under iron-limited, minimal media conditions [mean difference 0. 007; 2-tailed p = 0. 8273], indicating that observed metabolite differences were not related to differences in growth density. To determine if quantitative and qualitative differences in the siderophore metabolome distinguish urinary from rectal isolates in this set of 13 patients, we used quantitative metabolomic profiling and then compared these results to a genotypic analysis. In the metabolomic analysis, we measured differences between coincident urinary and rectal strains within individual patients. For each patient, the quantity of each siderophore produced by rectal strains was subtracted from the quantity produced by the coincident urinary strains to yield a difference (Fig. 4). In the four patients from whom multiple coincident urinary and rectal strains were recovered, the mean difference in siderophore production is reported. This analysis revealed significantly (p<0. 05) higher salmochelin and yersiniabactin production among urinary strains. It was also notable that we found no instances in which the rectal-only strain produced yersiniabactin or salmochelin while the urine strain did not. Furthermore, urinary strains always produced more salmochelin, even when non-urinary strains also expressed salmochelin (n = 3) suggesting that salmochelin biosynthesis was more active in urinary strains. Among all urinary strains in this study, prevalence was in the order enterobactin (100%), yersiniabactin (71%), salmochelin (50%) and aerobactin (14%). These data show that, while all strains made enterobactin, a biosynthetically active Yersinia HPI and iroA cassette were common among urinary isolates in this population and that production of these siderophores may have a clinically evident impact on UTI recurrence. The siderophore expression analysis of the rectal and urinary strains from the 13 patients described above was compared to PCR genotyping using established PCR primers for siderophore receptor genes (fyuA for yersiniabactin, iroN for salmochelin, iutA for aerobactin) [8], [26] (Table 4). Previous genotypic data suggesting that enterobactin genes are conserved among E. coli was supported by product analysis, which revealed enterobactin production by all patient isolates. For yersiniabactin and salmochelin, the siderophore receptor genotyping analysis had the limitation of being unable to discern strain differences when both a rectal and urinary strain pair were genotype positive. This was a frequent occurrence. In the ten patients from whom at least one fyuA+ strain was recovered, product analysis revealed rectal-urinary differences in yersiniabactin production in all ten, while genotyping predicted differences in four. Similarly, in the eight patients from whom at least one iroN+ strain was recovered, product analysis revealed differential yersiniabactin production in seven, while genotype predicted differential expression in four. Among fyuA− or iroN-positive strain pairs, mean and median yersiniabactin or salmochelin production remained higher among the urinary strains. For aerobactin, differences between iutA PCR and product analysis were even more striking. Aerobactin production was only detected in four of nine iutA+ strains. As a result, differences in siderophore production predicted by genotype were changed by product analysis in four patients. Agreement between the two methods occurred in one patient with an iutA+ strain and in the seven patients with iutA− strains that produced no detectable aerobactin. These data comparing PCR genotyping to product analysis demonstrate that differences in biosynthetic activity are not solely a reflection of the presence or absence of siderophore gene loci. Yersiniabactin was the most prevelant non-enterobactin siderophore and strikingly, was co-expressed in 90% of the strains that expressed salmochelin. This was the most frequent co-association of any of the non-enterobactin siderophores (p = 0. 006) and was mirrored by a significant association between fyuA and iroN positivity (p = 0. 008). Salmochelin and yersiniabactin co-expression was seen more often among patient urinary (6/13 (46%) ) than rectal (3/13 (23%) ) strains, although this trend did not reach statistical significance (p = 0. 18). Thus, salmochelin expression tended to occur in addition to yersiniabactin expression and co-expression was a common feature among urinary strains in this population. These data raise the possibility that these two siderophore types exhibit complementary activities. Clinical isolates were noted to produce varying amounts of linearized enterobactin, which is distinguished by a distinct retention time and an ion at m/z 688, consistent with hydrolysis (+18 amu) of a single ester bond (Fig. 5A). While somewhat slower at scavenging iron than enterobactin, it has been proposed that linear enterobactin is better suited to avoid sequestration by hydrophobic surfaces [27]. Enterobactin linearization was quantified by the precursor/product relation: Salmochelin expressors exhibited over threefold higher enterobactin linearization than non-expressor strains (Fig. 5B). To explore genetic correlates of this relationship, we examined enterobactin linearization among mutants in the iroA gene cluster in UTI89 (Fig. 5C). Mutants containing a deletion of the iroA gene cluster were unable to express salmochelin and exhibited a decrease in enterobactin linearization compared to the wild type control. Similar low levels of linear enterobactin were observed in clinical isolates that did not express salmochelin. To determine whether this linearizing activity was attributable to the esterase IroE [28], we examined enterobactin linearization in an iroE mutant. Linearization by the iroE mutant was decreased to the same levels as in the K12 strain MG1655, which lacks all of the iroA genes. The remaining baseline level of linear enterobactin in the absence of the iroA genes has been observed previously and may derive from premature release during biosynthesis [29], cleavage of ferric enterobactin by the enterobactin esterase Fes, or spontaneous ester bond hydrolysis. Together, these data show that, in addition to synthesizing salmochelin, iroA also directs enterobactin linearization through the action of the esterase IroE to produce linear enterobactin, a siderophore that may be better suited to iron-scavenging during infection. We have used a combined chemical, genetic, and patient-oriented approach to examine clinical correlates of siderophore production among human E. coli isolates associated with recurrent urinary tract infection. Development of a quantitative metabolomic approach allowed assessment of the complex multiprotein biosynthetic pathways for siderophores rather than inferring these activities from genotype or transcription analysis alone. PCR genotyping of a single siderophore system gene was not an efficient predictor of siderophore production during iron-restricted growth. Thus, the quantitative metabolomic approach was used to determine whether successful uropathogens exhibit systematic differences from coexisting rectal and urinary strains in individual patients. Rather than comparing strains across a population, we examined strain differences within individual patients in order to compare each urinary strain to a more valid reference population. This study compares siderophore production between coexisting bacterial strains associated with urinary disease and gut colonization. Since uropathogens are thought to arise from the gut flora, comparison of these populations should represent the most valid study design. The dichotomy between commensalism and pathogenicity is a common theme among bacteria and is particularly compelling in the case of E. coli urinary tract infections. Siderophore expression has been linked to virulence and here we show that yersiniabactin and salmochelin were the most common non-enterobactin siderophores associated with UTI recurrence in this typical young female population. Notably, yersiniabactin synthesis was observed in almost all strains that expressed salmochelin. In this study population of young women with recurrent UTI, urinary strains produced greater amounts of two siderophores, yersiniabactin and salmochelin and co-expression of both of these siderophores was common. In this study, quantitative product analysis provided information beyond conventional siderophore receptor genotyping in two circumstances: 1) when a genotype-positive strain was unable to produce detectable levels of the corresponding siderophore and 2) when siderophore production differed significantly between genotype-positive strain pairs. Ten of the thirteen patients in this study yielded at least one strain pair in which either or both of these circumstances was operative. Deficient or enhanced siderophore biosynthesis may arise in multiple environmental contexts. Pathogenic bacteria may benefit from increased production of siderophores that are better adapted to the infection microenvironment, as may be the case with salmochelin and yersiniabactin. Alternatively, bacterial strains in polymicrobial communities may benefit from inactivated siderophore production if they retain the ability to “steal” ferric siderophores produced by a neighbor, thereby avoiding the metabolic cost of siderophore biosynthesis [30], [31]. Thus, extrapolation of siderophore receptor genotype to biosynthetic phenotype is inefficient, often inaccurate and suggests that optimization of siderophore biosynthesis may occur in pathogenic strains. Enterobactin or aerobactin production was not preferentially associated with urinary strains in this population. The lack of increased enterobactin production among urinary strains suggests that qualitative shifts in siderophore type may be more conducive to uropathogenesis than a quantitative shift in enterobactin production. Although iutA positivity among pathogenic strains is often used to conclude that aerobactin is an important virulence factor, we did not observe preferential expression of this siderophore when urinary and rectal isolates were compared. The sample size may not have allowed us to discern preferential aerobactin production in this population. Alternatively, iutA-positive clinical isolates might exhibit urinary virulence properties that are unrelated to aerobactin production. These results suggest that yersiniabactin and salmochelin expression may facilitate infection of the human urinary tract. This effect is not absolute, as there are urinary strains in this study that express neither siderophore. Although rectal isolates in this and other studies [32] have been observed to produce yersiniabactin and/or salmochelin, the impact of these siderophores upon fitness for gut colonization is unclear. The relatively high prevalence of yersiniabactin and salmochelin expression among urinary pathogens (Table 5), makes these nonessential metabolic pathways intriguing prospects for virulence-targeted therapies. Interestingly, the yersiniabactin and salmochelin biosynthetic pathways converge at chorismic acid, where each pathway uses related enzymes to synthesize the aromatic precursors 2-hydroxybenzoic (salicylic) acid and 2,3-dihydroxybenzoic acid (Fig. 6). Targeting either or both of these initial points in siderophore biosynthesis may represent a promising target for anti-virulence drug discovery or design. This study was conducted according to the principles expressed in the Declaration of Helsinki. The study was approved by the Institutional Review Board of the University of Washington. All patients provided written informed consent for the collection of samples and subsequent analysis. To examine siderophore production in liquid culture, previously published conditions were used [33]. 3 hour cultures of E. coli grown in LB broth were diluted 1∶100 into M63 medium containing 0. 2% glycerol and 10 mg/mL niacin and incubated for 18 h at 37 C in a rotary shaker. Deletion mutations were made using the red recombinase method, as previously described, using pKD4 or pKD13 as a template and the primers as listed in Table S2. [34], [35] PCR was performed with flanking primers to confirm the appropriate deletions. Antibiotic insertions were removed by transforming the mutant strains with pCP20 [36] expressing the FLP recombinase. The resultant strains subsequently had no additional antibiotic resistance compared with the parental wt strain. The mass spectrometer used for the studies was a Thermo-Finnigan LCQ Deca (San Jose, CA) coupled to a Waters CapLC (Waters MA) equipped with a Vydac C18 MS column (0. 3×150 mm). The flow rate was 6 ul/min with a gradient as follows: Solvent A (0. 1% formic acid) was held constant at 95% and solvent B (80% acetonitrile in 0. 1% formic acid) was held constant at 5% for 5 minutes, increased to 44% B in the next 60 minutes, and then to 95% B in the next 20 minutes. All data was collected in a positive mode. The spray voltage on the mass spectrometer was held constant at 4. 5 K and the capillary temperature was 200°C. For CID experiments helium was used as the collision gas with the collision energy set to 32% of the maximum (∼5 eV). The isolation width was 3 amu. Quantitation was carried out in the SRM mode using 13C labeled standards as described above. Data was collected in the positive centroid mode. Ions were monitored with a window of +/−0. 5 amu. 0. 1 M ferric chloride was added to cell supernatants to a final concentration of 3. 75 mM. After a 15 minute room temperature incubation the precipitate was removed by centrifugation. The supernatant was applied to a column packed with 200 uL of DEAE slurry [33]. The loaded column was washed with 0. 5 mL of water and siderophores were eluted with 3 mL of 7. 5 M ammonium formate adjusted to pH 3. 6 with 7. 5 M formic acid. The DEAE eluate was further purified and desalted by application to a Chrom P solid phase extraction column (250 mg, Supelco). The loaded column was washed with 2 mL of 0. 1% formic acid in 10% acetonitrile. Siderophores were eluted following application of 2 mL of 0. 1% formic acid in 80% acetonitrile. The eluate was then concentrated to 100–200 uL final volume in a centrifugal evaporator for MS analysis. Internal standards were produced by rUTI2, a clinical urine isolate found to express all four known E. coli siderophore types, or UTI89. Strains were each grown for 3 hours in LB broth, which was subsequently inoculated 1∶100 into M63 medium containing 0. 2% 13C3-glycerol (99+ atom %, Isotec), and 10 mg/mL niacin and incubated for 18 h at 37 C in a rotary shaker. Cells were removed by centrifugation and a frozen stock of supernatant was kept for use as internal standard. Isotopic labeling was confirmed by LC-MS. Strains to be compared, along with the reference strain rUTI2, were prepared together on the same day using the same media, reagents, and internal standard. Siderophore quantities are expressed as reference strain equivalents determined through the stable isotope dilution method. 13C-labeled internal standard was added 1∶1 to each clarified culture supernatant and mixed prior to siderophore extraction. Siderophore extracts subject to comparison were then prepared and analyzed by LC-MS/MS using the parent and daughter ions described above and listed in Table S1. Each siderophore type was first quantified by determining the ratio of the analyte peak to the co-eluting 13C-labeled internal standard peak. These peak area ratios were then converted to molar ratios by comparison to standard curves generated by mixing known ratios of unlabeled and labeled rUTI2 supernatants. Siderophore quantities were expressed as rUTI2 equivalents by normalizing each molar ratio to that observed for strain rUTI2 under identical culture conditions. Isolates were grown to log phase on 5 ml LB medium. Primer combinations FyuA f' –FyuA r (880 bp product) /AerJ f–AerJ r (300 bp) /IRONEC-F–IRONEC-R (665 bp) were used for amplification of fyuA, iutA, and iroN genes, respectively [12], [37] Amplification reactions were carried out individually in a Bio-Rad MyCycler instrument using 5 µl of heat-inactivated culture and 35 cycles of 95×30″/57×30″/72×60″. Patients presenting with UTI were enrolled and monitored prospectively as described previously [14]. Rectal specimens were collected during clinic visits using a sterile, rayon-tipped swab and transported to the laboratory in Amies Medium (BBL™ CultureSwab™ Plus, Becton, Dickinson). To avoid inclusion of rectal strains that may have been introduced after the UTI event or that may have been only transiently present [38], [39] prior to the UTI event, rectal strains were excluded if they were recovered after or >30 days before the urinary isolate. Statistics and graphs were generated using GraphPad Prism 4. For groupwise comparisons of siderophore production, the Mann-Whitney U Test was performed. Analyses of paired strain differences in siderophore production were performed using the Wilcoxon signed rank test for significance. For analysis of stationary phase density, the data passed the F test for equal variances and the t test was used to compare urinary versus rectal strain growth as well as growth differences between paired strains. Categorical data was analyzed using Fisher' s exact test.

Urinary tract infections (UTIs) are among the most common bacterial infections treated by physicians worldwide. Although symptoms of acute infection are often resolved with a course of antibiotics, the same bacterial strain often causes subsequent bouts of symptomatic infection. Escherichia coli are the most common bacteria causing UTI and the infecting strains are widely believed to originate from the gastrointestinal tract where multiple E. coli strains reside. Here, we use a novel mass spectrometric technique in a population of patients with recurrent UTI to identify how strains that cause UTI differ from other strains that were present in the gastrointestinal tract at the same time. We found that urinary E. coli strains preferentially expressed two small molecules called yersiniabactin and salmochelin. These molecules are called siderophores, meaning they are able to scavenge iron to support bacterial survival and growth. Synthesis and transport of these small molecules requires a coordinated network of proteins encoded by a collection of different genes. These findings suggest that new antibiotics directed against yersiniabactin or salmochelin-producing E. coli strains may be an improved, and more targeted, strategy to prevent recurrent UTIs.

Abstract Introduction Results Discussion Materials and Methods

infectious diseases/bacterial infections microbiology/microbial physiology and metabolism biochemistry/small molecule chemistry urology/urological infections

2009

Quantitative Metabolomics Reveals an Epigenetic Blueprint for Iron Acquisition in Uropathogenic Escherichia coli

X-linked Glucose-6-phosphate dehydrogenase (G6PD) A- deficiency is prevalent in sub-Saharan Africa populations, and has been associated with protection from severe malaria. Whether females and/or males are protected by G6PD deficiency is uncertain, due in part to G6PD and malaria phenotypic complexity and misclassification. Almost all large association studies have genotyped a limited number of G6PD SNPs (e. g. G6PD202 / G6PD376), and this approach has been too blunt to capture the complete epidemiological picture. Here we have identified 68 G6PD polymorphisms and analysed 29 of these (i. e. those with a minor allele frequency greater than 1%) in 983 severe malaria cases and controls in Tanzania. We establish, across a number of SNPs including G6PD376, that only female heterozygotes are protected from severe malaria. Haplotype analysis reveals the G6PD locus to be under balancing selection, suggesting a mechanism of protection relying on alleles at modest frequency and avoiding fixation, where protection provided by G6PD deficiency against severe malaria is offset by increased risk of life-threatening complications. Our study also demonstrates that the much-needed large-scale studies of severe malaria and G6PD enzymatic function across African populations require the identification and analysis of the full repertoire of G6PD genetic markers. Amongst the approximately 190 genetic variants causing clinical deficiency of Glucose-6-phosphate dehydrogenase (G6PD) that have been characterised [1], the A- deficiency is the most common in sub-Saharan Africa populations, and is associated with protection from severe malaria [2,3]. An understanding of how this protection works may assist with the design of anti-malarial vaccines and drugs. Establishing whether malaria patients are G6PD deficient is also important because of the potential use of 8-aminoquinoline drugs (e g, primaquine and its derivatives) for malaria elimination in sub-Saharan Africa [4]. Primaquine is active against all liver stages of Plasmodium, and also offers activity against P. falciparum gametocytes, thereby blocking transmission to mosquitoes [4]. However, primaquine is haemotoxic, and can cause haemolytic anaemia in G6PD-deficient individuals. G6PD status can be quantified using enzymatic activity assays and is required for unambiguous identification of G6PD-deficiency, especially in mosaic female heterozygotes due to the X-linkage of the trait [5]. Cytochemical methods have been suggested as an alternative [5], but are not efficient for large studies, and genotyping has been used as a high throughput approach. Whilst genotyping approaches have been advocated, there is evidence of extensive diversity at the G6PD locus (X chromosome, 16. 2kb), with more than 150 single nucleotide polymorphisms (SNPs) reported [1]. Many of these known genetic variants result in amino acid changes and have been detected through sequencing the G6PD gene locus in enzyme deficient individuals. The G6PD and the Inhibitor of kappa light polypeptide gene (IKBKG, involved in immunity, inflammation and cell survival pathways [6], and with mutations linked to Incontinentia Pigmenti [7]) loci overlap each other, including a shared conserved promoter region that has bidirectional housekeeping activity [7]. The region containing the G6PD gene and the 5-prime end of the IKBKG gene contains Alu elements [7]. The genetic variability in G6PD and IKBKG is complex [7], and new alleles are still being discovered, making a simple G6PD genetic approach unreliable [8,9]. Despite these limitations, genotyping of the 202A/376G G6PD A-allele (with ∼12% of normal enzymatic activity [10]) has been used extensively in epidemiological studies to investigate protection against severe malaria [8,10–19]. It has been shown that coexistence of the two mutations is responsible for enzyme deficiency in G6PD A- because they act synergistically in causing instability of the enzyme [20]. They also lead to structural changes in the enzyme protein. However, even in large well-powered studies, associations between 202A/376G G6PD and protection from severe disease have been inconsistent, revealing protective effects in female heterozygotes [8,11,17,18,19], in male hemizygotes [12,13], in both [14], or no protection [15]. These phenotype-genotype inconsistencies may be explained in part by variation in study design, G6PD and malaria phenotypic complexity and misclassification and incomplete experimental data [8]. However, it has been recognised that allelic heterogeneity, specifically other unknown polymorphisms, has a role [3,5, 8], with evidence from studies in West Africa [5,8] for A- deficiency and in Southeast Asia and Oceania for other deficiency types [3]. In particular, in the West African setting, the frequency of the 202A allele is often substantially lower than rates of enzyme deficiency indicating a role for other alleles; inclusion of other G6PD polymorphisms (Santamaria 542T/376G—∼2% residual enzymatic activity, Betica-Selma 968C/376G—∼11% activity) [10,16] was required to capture an association between G6PD deficiency and severe malaria in The Gambia [8]. Further understanding is required of the true extent of genetic diversity within the G6PD locus, how this relates to enzyme function, and how it varies between regions and ethnic groups, if genetic epidemiological studies are to provide robust and reproducible findings. A recent study in Mali using 58 SNPs across the G6PD gene found differences in core haplotypes and their frequencies between Dogon and Fulani ethnic groups [9]. The latter group is known to have substantially reduced susceptibility to malaria when compared to sympatric populations [9]. Whilst some ethnicity specific SNP associations were observed with mild malaria, the prevalence of severe malaria was too low for any robust associations to be detected. Here we investigate associations between 68 SNPs within the G6PD and surrounding loci (IKBKG and CTAG1A/B), including the 202,376,542,680 and 968 A- deficiency polymorphisms (referred to here as G6PD202, G6PD376, and so forth), and severe malaria. The work is set within a case-control study (n = 983; 506 cases and 477 controls) conducted in an area of intense malaria transmission in the Tanga region in northeastern Tanzania [17]. To complement the case-control collection, we genotyped samples from 60 healthy parental and child trios (120 parents, 60 children), collected in the same geographical region. We find very strong associations between multiple SNPs across the G6PD gene and protection from severe malaria in female heterozygotes but not in hemizygous males. Very high linkage disequilibrium across this locus allowed us to distil this SNP diversity into just 4 G6PD alleles, ranging in frequency from ∼6% to >60%, and 8 common genotypes (>1%), 2 of which are associated with protection from severe malaria. In summary, this study identifies specific G6PD alleles that confer resistance to severe malaria in this population and reveals a potentially important role of female heterozygotes in maintaining the high frequency of G6PD polymorphisms in malaria endemic populations. Of the severe malaria cases (n = 506), many had severe malarial anaemia (48. 6%) or acidosis (57. 5%) phenotypes (Table 1). Compared to controls (n = 477), malaria cases tended to be younger and male, and with more individuals outside the 7 main ethnic groups (P<0. 05). Malaria cases were less likely to be of blood group O (O vs. A/B/AB, OR 0. 726,95% CI 0. 534,0. 986; P = 0. 04), with alpha thalassaemia of α-/α- (α-/α- vs. αα/αα or αα/α-, Odds Ratio (OR) 0. 639,95% CI 0. 401–1. 018, P = 0. 06) or present with the sickle cell protective AS genotype (AS vs. other, OR 0. 053,95% CI 0. 021–0. 132). The sickle cell AS genotype frequency in parents (6. 3%) and children (5. 4%) in the trio validation study lay between the estimates for the cases (1. 0%) and controls (16. 5%). As expected, the G6PD542,680 and 968 polymorphisms found in West African populations [8,9] were all monomorphic in both cases and controls, as well as in the 60 parental-child trios, and were therefore excluded from further analysis. The G6PD202A and G6PD376G A- alleles were among the 29 SNPs retained with minor allele frequency (MAF) in excess of 1% (S2 Table). Both G6PD202A (case 16. 3% vs. control 20. 0%) and G6PD376G (37. 4% vs. 38. 5%) allele frequencies were lower in malaria cases than in controls (P<0. 02) (Table 1), and broadly similar to the trio study parents (202A 16. 8%, 376G 31. 3%) and children (202A 15. 0%, 376G 24. 1%) (S3 Table). A SNP-by-SNP association analysis revealed 11 multiple loci where female heterozygotes appeared to be protected from severe malaria in all its clinical phenotypes (Table 2, Fig. 1, S1 Fig.) except for cerebral malaria where although there was evidence of heterozygous advantage effects (OR ∼ 0. 5), they were non-significant due to the small number of cases (99) (P>0. 018). The G6PD376 and rs762515 polymorphisms (both flanking G6PD202) were the only SNPs associated with all non-cerebral malaria clinical phenotypes. The association hits across clinical phenotypes included a “core” region consisting of 7 SNPs (rs5986990, rs2515905, rs2515904, G6PD376, G6PD202, rs762515, rs762516) in perfect linkage disequilibrium (D’ = 1), where female heterozygotes were 48. 2% and 72. 4% less likely to be a severe malaria case (any definition) than female homozygote genotypes (P<0. 006, Table 2). By comparison, there were no significant associations between G6PD genotype and severe malaria in hemizygous males (P>0. 310). The correlation between the 29 SNPs was high (linkage disequilibrium D’ median (IQR): all subjects 0. 987 (0. 811–0. 997); female controls 0. 988 (0. 731–0. 998) ). Similarly, LD was high across this region in the trio parents (all: 0. 998 (0. 995–0. 999); female only: 0. 998 (0. 995–0. 999) ) and children (0. 998 (0. 996–0. 999) ) (S2 Fig.). This high LD allowed us to define a small number of haplotypes/G6PD alleles (4) that accounted for 99. 6% of all alleles typed for the “core” region (haplotype 1 = GGGAGTC, 2 = AACGGCT (6 mutations), 3 = AACGACT (7 mutations), 4 = AGGGGCC (3 mutations) ). Female controls had a higher frequency of the three haplotypes (2–4) containing mutations. Whilst protective effects were observed in females (and not males) for these three haplotypes (OR 0. 683–0. 783) compared to the common type (haplotype 1, frequency ∼60%), they were not statistically significant (P>0. 186), due to the heterozygous nature of the protection in females (S4 Table). Further analysis accounting for the genotypic combinations of G6PD alleles confirmed that a combination of haplotypes 1 and either 2 or 3 were protective (OR<0. 38, P<0. 006) compared to a double haplotype 1 (wild-type) genotype (Table 3). This result shows that haplotypes with the 376G mutation have similar protective effect in heterozygotes irrespective of the presence or absence of the 202A mutation, indicating that the 376G mutation is causal. The genotypic combination of haplotypes 1 and 4 also had a potentially protective effect (OR = 0. 599), but it failed to reach statistical significance (P = 0. 11). It is possible that the greater protective effects of haplotypes 2 and 3, could be due to the presence of more mutations (≥6), leading to a possible compound heterozygous advantage effect. The number of heterozygous genotype calls in female controls was greater than in cases (case vs. control median / mean: All SNPs 10 / 9. 1 vs. 7 / 7. 6, P<0. 001; 7 core SNPs, 3 / 3. 2 vs. 0 / 2. 1, P<0. 0001). The Tajima’s D metric was applied to assess if the excess number of heterozygous alleles led to evidence of balancing selection in the G6PD gene. There was very strong evidence of balancing selection across all groups (Tajima’s D > 2. 6, female controls 2. 9). The magnitude of effect is at the extreme positive tail of an observed negatively centred African population distribution [21], where predominantly negative values demonstrate either slow growth from a small population size, or a bottleneck that is much older than that of non-Africans [21]. This result implies that the (high) allele frequency of the SNPs in the G6PD gene is maintained mainly, and perhaps entirely, by the protection against severe malaria of heterozygous females through a balancing selection mechanism. This selection mechanism is also predicted by population genetic theory [22], and consistent with empirical data from other studies [8,18]. Such mechanisms exist at other malaria candidate loci in the autosomal regions, for example at the HbAS sickle trait [23]. There was no evidence of epistatic effects between HbS and G6PD on severe malaria in females (P = 0. 34), nor males (P = 0. 98). Similarly, no evidence of epistasis between alpha thalassaemia and G6PD (female P = 0. 44; male P = 0. 21). Although G6PD A- deficiency is known to protect against severe malaria in African populations, the underlying genetic mechanisms are not well understood. P. falciparum development is hindered in G6PD deficient red cells [24], slowing the rate of parasite replication and reducing the likelihood of severe disease. Suggested mechanisms include more efficient clearance of the infected erythrocytes [25], lower abundance of P. falciparum 6-phosphogluconolactonase mRNA in parasites from G6PD-deficient children [26], and impaired parasite replication [27]. By using the largest set of G6PD (and surrounding loci) SNPs (n = 68) in a genetic association study, within a Tanzanian case-control setting, we have established a set of new G6PD alleles associated with protection. These SNPs need to be further investigated to assess their effect on enzyme function in light of potential use of primaquine for malaria elimination. After validation, these SNPs may be used to identify G6PD-deficient individuals in studies of primaquine efficacy. Further, we have shown that the protective effect of G6PD deficiency is limited to female heterozygotes. This is entirely consistent with heterozygote advantage and balancing selection, relying on alleles at modest frequency and avoiding fixation, where protection provided by this G6PD deficiency against severe malaria is offset by increased risk of life-threatening complications, such as neonatal jaundice and haemolytic crises. In female heterozygotes, random inactivation of one of the two X chromosomes results in some cells with normal enzyme and others with mutant enzyme [11,28,29], reducing the risk of both anaemia and severe malaria. We expect that the fitness of normal male hemizygotes is the same as that of normal female homozygotes (since all red cells will contain fully functional enzyme), and population genetic theory also suggests that the fitness of G6PD-deficient male hemizygotes is the same as that of G6PD-deficient female homozygotes. Under these conditions, it is expected that the female heterozygote must be the genotype with the highest fitness [22]. Two independent studies [8,18] in two different populations, nearly 40 years apart, are consistent in this regard, with G6PD deficiency A− being a balanced polymorphism with heterozygote advantage. Similarly, as the G6PD deficiency A− has been estimated to be at least 5000 years old [3], balancing selection would account for it not having gone to fixation [22]. Further, balancing selection has been observed in autosomal malaria candidate regions like FREM3, the major histocompatibility complex, and the sickle cell trait loci [23]. Hitherto, there has been much uncertainty about the relationship between G6PD status and susceptibility to malaria, due in part to G6PD and malaria phenotypic complexity and misclassification, and potentially also from the genetic complexity of the G6PD locus with the presence of multiple functional SNPs, each of which may separately modify an individual’s enzyme status and susceptibility to malaria. Until very recently, almost all-large association studies genotyped a limited number of G6PD SNPs (e. g. G6PD202 / G6PD376 for A- deficiency), and this approach has been too blunt to capture the full picture. However, analysis of 58 G6PD SNPs has demonstrated major G6PD haplotypic differences between sympatric ethnic groups in Mali [9] and genotyping of the G6PD968 polymorphism in addition to 202/376 revealed a female protective in a Gambian population [8]. With hindsight, it is clear that genotyping of G6PD968 in another study in the same population [14] would have prevented misclassification of two-thirds of the G6PD-deficient samples and the erroneous reporting of a male hemizygous protective effect. Other studies reporting male hemizygous protective effects may also be confounded by allelic heterogeneity, which could be avoided by more comprehensive genotyping and by phenotypic testing for G6PD enzyme activity. A comprehensive study would include a full genetic survey of the G6PD and surrounding regions, with multiple populations and ethnic groups, leading to a more complete map of G6PD that would guide future evolutionary and association studies. A surprising association result is that the G6PD376 mutation is potentially more influential than G6PD202 and haplotypes that contain the 376G with or without the 202A mutation appear to be similar in terms of protective effect on heterozygotes. The 202A mutation is thought to have a more severe effect on enzyme function than the 376G mutation (∼12% and ∼83% of normal function, respectively [10,30]) and coexistence of 202A/376G is responsible for G6PD A- enzyme deficiency [20], but it is possible that more subtle changes in enzyme structure or function also affect the outcome of malaria infection. Fully understanding the role of G6PD requires further correlation of enzymatic activity with full sequences of G6PD and surrounding loci, set within large severe-malaria case and control studies. There have been no such studies to date. A recent study of four G6PD deficiency polymorphisms (202,376,968, Ilesha) and associated enzymatic activities for 110 sequenced genes in African Americans [31] but included only 54 heterozygous females. Enzymatic activity for G6PD376G (A+, n = 28), 376G/202A (A- deficiency, n = 23), 376G/968C (A-, n = 1), 376G/202A/968C (A-, n = 1) and Ilesha (E156K, Nigeria, non A-, n = 1) alleles was estimated to be ∼83%, ∼53% ∼58%, ∼11% and ∼75% of normal, respectively. These results are consistent with deficiency increasing with additional A- related polymorphism, and by implication will change levels of protection or susceptibility to malaria. Another recent study [32] in 1,828 Kenyan children suggested that G6PD202 was responsible for the majority of G6PD enzyme deficiency but that 376G increases the risk of deficiency in 202AG heterozygotes. Neither study considered malaria outcomes. In summary, through a much better understanding of the true extent of genetic diversity within and around the G6PD locus, we have identified alleles associated with protection from severe malaria in Tanzania, driven by a balancing heterozygous advantage mechanism. Further work should extend the mapping of diversity at this genomic region, and identify how the resulting mutations relate to enzyme function, and how they vary between region and ethnic group. In doing so, genetic epidemiological studies are likely to provide robust and repeatable data, which may be used to develop interventions, and improve malaria disease control. The study was conducted in the Teule district hospital and surrounding villages in Muheza district, Tanga region. In this region, mortality in children under 5 years of age is 165 per 1000 (Tanzanian census 2002) and transmission of P. falciparum malaria is intense (50–700 infected bites/person/year) and perennial, with two seasonal peaks [17]. The community prevalence of P. falciparum parasites in children aged 2–5 years in the study area was recorded as 88. 2% in 2002 [17]. Severe malaria cases (n = 506), aged six months to ten years, were recruited during a one-year period between June 2006 and May 2007, with patent parasitaemia, and fulfilling any one of the following eligibility criteria; history of 2 or more convulsions in last 24 hours, prostration (unable to sit unsupported if <9 months of age or drink at any age), reduced consciousness (Blantyre Coma scale<5), respiratory distress, jaundice, severe anaemia (Hemocue Hb < 5g/dL), acidosis (Blood lactate ≥ 5 mmol/L), hypoglycaemia (blood glucose < 2. 5mmol/L). Cases were defined as having had cerebral malaria if their Blantyre coma score was less than or equal to 3 on presentation or early during admission. Participants with co-existing severe or chronic medical conditions (e. g. bacterial pneumonia, kwashiorkor) unrelated to a severe malarial infection were excluded. All cases were confirmed as having P. falciparum malaria parasites. Parasite infection was initially assessed by rapid diagnostic test (HRP-2—Parascreen Pan/Pf) and confirmed by double read Geimsa-stained thick blood films. Residence and ethnic group of both parents was recorded from information provided by the caregiver for each child [17]. Controls (n = 477) were recruited matched on ward of residence, ethnicity and age using household lists during a four-week period in August 2008. Study participants resided in 33 geographical wards (including Mtindiro 9. 6%, Kwafungo 8. 5%, Mkata 6. 3%, Kwedizinga 6. 0%, others each < 5. 0%) surrounding Muheza town in the Tanga region. The participants had a median age of ∼2. 6 years, and were predominantly from seven ethnic groups (see Table 1). Because of limited sample size, we did not perform a detailed analysis based on different ethnic groups or wards of residence. To complement the case-control collection, we collected samples anonymously from 60 healthy parental and child trios (120 parents, 60 children) during 2007 and 2008 from lowland villages near the West Usambara mountains in the Tanga region of Tanzania, which ranges from high to medium levels of malaria transmission. No malaria phenotypic data is available on these individuals, but their genotypic profiles were used to provide validation data of the genetic aspects of the case-control study. Approximately 3ml of venous blood was collected from participants into EDTA vacutainers. A blood film was prepared and haemoglobin levels measured by hemocue. Children in the control group with haemoglobin levels of <11g/DL were referred to the nearest health facility; those with a positive blood film were treated in line with Tanzanian national treatment guidelines and excluded from the genetic analysis. Samples were spun at 5000rpm for 5 minutes and the plasma removed and stored for future analysis. DNA was extracted and purified from the blood cell pellet using a nucleon kit (see [17] for details). Genomic DNA samples were genotyped on a Sequenom MassArray genotyping platform [17,33]. The iPlex genotyping assays included 68 G6PD single nucleotide polymorphism (SNP) positions (identified through resequencing and the 1000 genomes project, described in [9,32]), HbS (rs334), HbC (rs33930165), HbE (rs33950507), and two SNPs that allow an estimate of the ABO blood group rs8176719, rs8176746). In particular, the rs8176719 derived allele results in a non-functional enzyme, and group O individuals are DD, while non-O Individuals are either II or ID. In addition, rs8176746 is involved in the enzyme' s substrate selection and therefore defines either the A or B blood groups. A full list of SNPs can be found in S1 Table. The α3. 7-thalssaemia deletion was typed separately by PCR [17]. All analyses involving SNPs were stratified by gender. Genotypic deviations from Hardy-Weinberg equilibrium (HWE) in females were assessed using a Chi-square statistical test. SNPs were excluded from analysis if they had at least 10% of genotype calls missing, more than 2% of males genotype calls were (falsely) called heterozygous, or if there was a distortion from HWE in female controls (HWE Chi-square P<0. 00001) [9]. On this basis, 6 SNPs were excluded (rs766419, rs743545, rs743548, b36_153424319, rs2472393, b36_153426354). A further 33 SNPs with minor allele frequency less than 1% were also removed, leaving 29 high quality SNPs for association analysis (listed in S2 Table). The 29 SNPs are located in a genomic region with known regulatory capacity (transcription factor binding and DNase peaks (regulomedb. org), and both promoter and enhancer histone marks, with a number of different binding proteins and regulatory motif changes (compbio. mit. edu/HaploReg) ). Case-control association analysis using SNP alleles or genotypes was undertaken within a logistic regression framework, and included age and ethnic group as covariates. In this approach we modelled the SNP of interest assuming several related genotypic mechanisms (additive, dominant, recessive, heterozygous advantage and general models) and reported the minimum p-value from these correlated tests. Epistatic effects between polymorphisms were considered by inclusion of statistical interactions in these models. The haplotypes of females were inferred from genotypes using an expectation-maximization algorithm [34]. Haplotype association testing was performed using the regression models [34]. Linkage disequilibrium was estimated using the pairwise D-prime (and R-square) metrics [35]. Performing multiple statistical tests leads to inflation in the occurrence of false positives. A Bonferroni correction would be too conservative because all SNPs are from the same gene. A permutation approach that accounted for the correlation between tests estimated that a p-value cut-off of 0. 006 would ensure a global significance level of 5%. All analyses were performed using the R statistical software. The R haplo. stat library was used to implement haplotype analysis. The Tajima’s D metric was used to quantify evidence of balancing selection based on the allele frequency spectrum [36]. A negative Tajima' s D indicates purifying selection and/or population size expansion, while positive values may indicate balancing selection. Values greater than +2 or less than -2 are likely to be significant [36]. All DNA samples were collected and genotyped following signed and informed written consent from a parent or guardian. Ethics approval for all procedures was obtained from both LSHTM (#2087) and the Tanzanian National Institute of Medical Research (NIMR/HQ/R. 8a/Vol. IX/392).

Glucose-6-phosphate dehydrogenase (G6PD) is an essential enzyme that protects red blood cells from oxidative damage. Numerous genetic variants of G6PD, residing in the X chromosome, are found among African populations: mutations causing A- deficiency can lead to serious clinical outcomes (including hemolytic anemia) but also confer protection against severe malaria. Epidemiological studies have used some of the genetic markers that cause A- deficiency to establish who is protected from severe malaria, with differing results. Whether females, with one or two copies of mutant genes, males with one copy, or both genders are protected is uncertain. This uncertainty is due to G6PD and malaria phenotypic complexity and misclassification, and to genetic differences between populations and the limited numbers of genetic markers (usually 2) considered. In this study we analysed more than 30 G6PD genetic markers in 506 Tanzanian children with severe malaria and 477 without malaria. We found that only females with one normal and one mutant copy of the gene (heterozygotes) were protected from severe malaria. Further, we established that the G6PD gene is under evolutionary pressure with the likely mechanism being selection by malaria. Our work demonstrates that studies of severe malaria and G6PD enzymatic function across African populations require, in addition to complete and accurate G6PD phenotypic classification, the identification and analysis of the full repertoire of G6PD genetic markers.

Abstract Introduction Results Discussion Materials and Methods

2015

African Glucose-6-Phosphate Dehydrogenase Alleles Associated with Protection from Severe Malaria in Heterozygous Females in Tanzania

Inference of interaction rules of animals moving in groups usually relies on an analysis of large scale system behaviour. Models are tuned through repeated simulation until they match the observed behaviour. More recent work has used the fine scale motions of animals to validate and fit the rules of interaction of animals in groups. Here, we use a Bayesian methodology to compare a variety of models to the collective motion of glass prawns (Paratya australiensis). We show that these exhibit a stereotypical ‘phase transition’, whereby an increase in density leads to the onset of collective motion in one direction. We fit models to this data, which range from: a mean-field model where all prawns interact globally; to a spatial Markovian model where prawns are self-propelled particles influenced only by the current positions and directions of their neighbours; up to non-Markovian models where prawns have ‘memory’ of previous interactions, integrating their experiences over time when deciding to change behaviour. We show that the mean-field model fits the large scale behaviour of the system, but does not capture the observed locality of interactions. Traditional self-propelled particle models fail to capture the fine scale dynamics of the system. The most sophisticated model, the non-Markovian model, provides a good match to the data at both the fine scale and in terms of reproducing global dynamics, while maintaining a biologically plausible perceptual range. We conclude that prawns’ movements are influenced by not just the current direction of nearby conspecifics, but also those encountered in the recent past. Given the simplicity of prawns as a study system our research suggests that self-propelled particle models of collective motion should, if they are to be realistic at multiple biological scales, include memory of previous interactions and other non-Markovian effects. The most striking features of the collective motion of animal groups are the large-scale patterns produced by flocks, schools and other groups. These patterns can extend over scales that exceed the interaction ranges of the individuals within the group [1]–[4]. For most flocking animals, the rules dictating the interactions between individuals, which ultimately generate the behaviour of the whole group, are still not known in any detail. Many ‘self-propelled’ particle models have been proposed for collective motion, each based on a relatively simple set of interaction rules between individuals moving in one, two or three dimensions [2], [5]–[8]. Typically these models implement a simple form of behavioural convergence, such as aligning the focal individual' s velocity in the average direction of its neighbours or attraction towards the position of those neighbours. Generally such rules are explicitly kept as simple as possible while remaining realistic, with the aim of explaining as much as possible of collective motion from the simplest constituent parts. Each of the models in the literature is capable of reproducing key aspects of the large-scale behaviour of one or more biological systems of interest. Together these models help explain what aspects of inter-individual interactions are most important for creating emergent patterns of coherent group motion. With this proliferation of putative interaction rules has come the recognition that some patterns of group behaviour are common to many models, and that different models can have large areas of overlapping behaviour depending on the choice of parameters [4]. Common patterns of collective behaviour are also observed empirically across a diverse range of animal and biological systems. For example, a form of phase transition from disorder to order has been described in species as diverse as fish [9], ants [10], locusts [11], down to cells [12] and bacteria [13]. In all these systems, as density of these species is increased there is a sudden transition from random disordered motion to ordered motion with the group collectively moving in the same direction. These studies indicate that a great deal can be understood about collective behaviour without reduction to the precise rules of interaction. In many contexts however the rules of interaction are of more interest than the group behaviour they lead to. For example, when comparing the evolution of social behavior across different species, it is important to know if the same rules evolved independently in multiple instances, or whether each species evolved a different solution to the problem of behaving coherently as a group [1]. Recently researchers in the field have become interested in using tracking data from real systems on the fine scale to infer what precise rules of motion each individual uses and how they interact with the other individuals in the group [14]–[19]. This is an important trend in the field of collective motion as we move from a theoretical basis, centred around simulation studies, to a more data-driven approach. The most frequent approach to inferring these rules has been to find correlations between important measurable aspects of the behaviour of a focal individual and its neighbours. For example, Ballerini et al. [14] looked at how a focal individual' s neighbours were distributed in space relative to the position of the focal individual itself in a group of starlings. Significant anisotropy in the position of the nearest neighbour, averaged over all individuals, was regarded as evidence for an interaction between each bird and that neighbour. More recently Katz et al. [18] and Herbert-Read et al. [19] investigated how the change in velocity of each individual in groups of fish was correlated to the positions and velocities of the neighbouring fish surrounding the focal individual. This provides evidence not only for the existence of an interaction between neighbours but also estimates the rules that determine that interaction. In these studies the rules of interaction are presented non-parametrically and cannot be immediately translated into a specific self-propelled particle model. Nor are these models validated in terms of the global schooling patterns produced by the fish. An alternative model-based approach that does fit self-propelled particle and similar models to data is proposed by Eriksson et al. [16] and Mann [17]. Under this approach, the recorded fine-scale movements of individuals are used to fit the parameters of, and select between, these models in terms of relative likelihood or quality-of-fit. This approach has the advantage of providing a parametric ‘best-fit’ model and can provide a quantitative estimate the relative probability of alternative hypotheses regarding interactions. What all previous empirical studies have lacked is a simultaneous verification of a model at both the individual and collective level. Either fine scale individual-level behaviour is observed without explicit fitting of a model [18], [19] or global properties, such as direction switches [11], [20], speed distributions [21], [22] or group decision outcome [23] have been compared between model and data. Verification at multiple scales is the necessary next step now that inference based on fine-scale data is becoming the norm. Just as simulations of large-scale phenomena can appear consistent with observations of group behaviour without closely matching the local rules of interaction, so can fine-scale inferred rules be inconsistent with large-scale phenomena if these rules of inferred from too limited a set of possible models or from correlations between the wrong behavioural measurements. The closest that any study so far has come to finding consistency between scales has been Lukeman et al. [15]. In their study the local spatial distribution of neighbouring individuals in a group of scoter ducks was used to propose parametric rules of interaction, with some parameters measured from the fine-scale observables, but with others left free to be fitted using large-scale data. We suggest that if group behaviour emerges from individual interactions, then the form of these interactions should be inferable solely from fine-scale data without additional fitting at the large-scale. An inability to replicate the group behaviour using a selected model demonstrates that the model space has been insufficiently explored. When faced with alternative hypothesised interaction rules, model-based parametric inference provides the best means of quantitatively selecting between them. In this paper we study the collective motion of small groups of the glass prawn, Paratya australiensis. Paratya australiensis is an atyid prawn which is widepsread throughout Australia [24]. Although typically found in large feeding aggregations, it does not appear to form social aggregations and has not been reported to exhibit collective behaviour patterns in the wild. We conduct a standard ‘phase transition’ experiment [9], [11], [12], studying how density affects collective alignment of the prawns. We complement this approach by using Bayesian inference to perform model selection based on empirical data at a detailed individual level. We select between models by calculating the probability of the fine scale motions using a Bayesian framework specifically to allow fair comparison between competing models of varying complexity. Comparison of the marginal likelihood, the probability of the data conditioned on the model, integrating over the uncertain parameter values, is a well developed and robust means of model selection that forms the core of the Bayesian methodology [25]–[28] and which has been applied to compare models in the biological sciences, particularly neuroscience [29]. Bayesian methods are also well established in animal behaviour through consideration of optimal decision making in the presence of conflicting information, both environmental [30] and social [31], [32]. In adopting this approach, we reject the dichotomy of model inference based on either fine scale behaviour of the individuals or the motion of the group. Instead we use reproduction of the large scale dynamics through simulation as a necessary but not sufficient condition of the correct model. Next we investigated a series of interaction models as to their ability to reproduce the fine scale interactions of the prawns. We predict the probability, , that a focal prawn will change its orientation, given one of a number of potential models. The direction changes are determined by the data from the six-prawn treatment. This treatment provides the best balance between the number of data points, density of direction changes, clear large scale behaviour and tracking accuracy. Each model specifies the probability that a focal prawn will change its direction in the next time step conditioned on the relative positions and directions of the other individuals in the arena. We use a logistic mapping to ensure probabilities remain between zero and one, so each model uses the relevant variables to determine a latent ‘turning-intensity’, , such that, (2) where is a function of the relative positions and directions of the other prawns, both now and potentially in the recent past, and the model parameters. The models are, in increasing degree of complexity, as follows. Firstly to consider models that do not include zones-of-interaction – non-spatial models. We establish a baseline with a Null model. This simply posits that direction changes occur at random, at the rate established from the single prawn data, and the prawns do not interact in any way that changes this direction-changing probability. Therefore is given simply by a baseline constant, , which is determined by the rate of direction changing in single prawns. (3) We also consider two models where the interaction is independent of absolute spatial separation. The Mean Field (MF) model includes interactions between all prawns regardless of position, such that their relative directions alter the probability of changing direction. Since the number of prawns in the experiment is fixed, the probability for a direction change is influenced by the number of individuals moving in the opposite direction (negative prawns), . Each negative prawn increases the turning intensity by an amount, (4) A Topological (T) model restricts these interactions to a limited number of nearest-neighbours, , the individuals closest to the focal prawn. The turning intensity is now influenced by the number of negative prawns, within the set of nearest-neighbours. (5) Secondly we consider a class of Spatial models (S1–S4). These models closely resemble the classic one-dimensional self-propelled particle models from the literature [5]. The focal prawn interacts with neighbours within a spatial zone-of-interaction, . The number and directions of individuals within this interaction zone determine the probability of changing direction. A number of further variations are possible; interactions can be limited to prawns ahead of the focal prawn and/or to prawns travelling in the opposite direction to the focal prawn. We consider four variations, indicated in Table 1. The general form for this model is given by, (6) where and are the number of negative and positive (travelling in the same direction) prawns within the interaction zone, and and parameterise the influence of each individual on the turning intensity. . Interactions can occur with negative prawns only, , or with both negative and positive oriented prawns, . The spatial interaction zone is either a symmetrical area centred on the focal prawn, of width radians around the ring (spatial symmetric models in Table 1), or is only directed radians ahead of the focal prawn (spatial forward models). Visual inspection of the movements of the prawns suggests that interactions often follow a particular pattern. Two prawns, travelling in the opposite directions, collide. After the prawns have passed each other one of the prawns may subsequently decide to change direction. Self-propelled particle and other models of collective motion do not capture this type interaction. Such interactions are non-Markovian, i. e. the change in direction is not just the result of the environment now, but of the past environment as well. We proposed a third class of models (D1–D4), simple non-Markovian extensions of the basic spatial models, where each prawn would ‘remember’ the other individuals it encountered, with those memories fading at an unknown rate after the interaction was complete. As such the prawn would integrate those interactions over time, building up experiences which would alter its chance of changing direction. Mathematically this means that the turning intensity is now auto-regressive, depending on its own value at the previous time step as well as the current positions and directions of the neighbouring individuals. We introduce a decay parameter, , which determines how quickly the turning intensity returns to normal after an interaction with a neighbour has occurred. The same variations of interaction are allowed as for the spatial models, giving a general form for the non-Markovian turning intensity as, (7) where now indicates the turning intensity at time, which depends on the value of the turning intensity at the previous time step, . The number of prawns still in the interaction zone from time is indicated by, while the number of new arrivals in the interaction zone is given by. Hence raised (or lowered) turning intensities persist over time, with a duration controlled by the value of. After the focal prawn changes direction the turning intensity is reset to the baseline, , at the next time step. Table 1 specifies the interaction zone structure for each of eleven alternative models, grouped according to the description given above. For each model we calculate the marginal likelihood of the data, conditioned on the interaction model (see Materials and Methods). The marginal likelihood is the appropriate measure for performing model selection, especially between models of varying complexity. More complex models, by which we mean models with a larger number of free parameters, are penalised relative to simpler models when integrating over the parameter space, since less probability can be assigned to any particular parameter value a priori. The marginal likelihood indicates how likely a particular model is, rather than a model and an chosen optimal parameter value (see, for example, Mackay [33] Chapter 28 and other standard texts for discussions on this topic). The marginal likelihoods of each model are shown in Figure 3A. We also measure the consistency between the large scale results of our experiments and the results predicted by simulation of each model, using the parameter values in Table 1. We set a consistency condition that any model that accurately approximates the true interactions must fulfil. We measure the large scale quality-of-fit between the model simulations and the experiments using the Kullback-Leibler divergence [34] between the distribution of simulated and experimental outcomes and performing a G-test for quality-of-fit [35] (see Materials and Methods). The p-value associated with this quality-of-fit for each model is shown in Figure 3B, showing which models are deemed to be consistent with experiments (those with). Large scale results from the simulation of each model are shown individually in Figures S2, S3, S4, S5, S6, S7, S8, S9, S10, S11, S12 in the supplementary materials. The Null model, in which prawns do not interact, performs significantly worse than the mean-field model. Figure 4 shows that the mean-field reproduces both the global alignment of the prawn groups, with an increase in polarisation with time and group size. These results show that the prawns interactions involve matching their directions to that of others, producing alignment. Are local spatial interactions important in reproducing observed direction changes? We note first that a topological interaction zone, where the focal prawn interacts with its nearest neighbours, has a marginal likelihood slightly lower than the mean field model. The topological model is ‘punished’ for having more parameters than the mean-field model, since the most probable value of the topological interaction range encompasses all neighbours. However, interactions between prawns are local. Figure 5 shows how the probability of changing direction depends on the position of the nearest opposite facing neighbour. An opposite facing neighbour within approximately radians of a focal prawn strongly increases the chance that the focal prawn will change direction. This observation suggests that a local interaction spatial model should outperform the mean field model, and we can use the approximate observed range of interaction (radians) to inform our prior probability on the interaction zone for models that include one. However, Figure 3A shows that with this limit on the interaction zone, the spatial models (S1–S4) all have a marginal likelihood lower than the mean field model. Simulating these models with most-probable parameters inferred from our analysis of the data (see Table 1) shows that these fit poorly on the large scale too, having a relatively large divergence between the simulated outcomes and the observed large scale alignment patterns and are therefore showing significant differences in the quality-of-fit test (Figure 3B). Both Figure 5 and our biological reasoning insist that locality must be maintained in interactions between individual animals. Therefore the poor performance of these spatial models indicates that they are an incomplete description of the true behaviour of the prawns. The models incorporating a non-Markovian delayed response together with a spatial interaction zone (models D1–D4) all outperformed the most probable Markovian spatial model on both the fine and large scales (Figure 3). Model D3 is the best performing model on both scales, and is the only model with a greater marginal likelihood than the Mean Field model. This then is the best model we can infer from our selection of possibilities. Figure 6 shows that simulations of model D3 produce collective alignment of the prawns and consistently stronger and faster alignment for larger group sizes, fulfilling our large-scale consistency requirement for a realistic model. The inferred value of the memory parameter associated with this model (see Table 1) puts the half-life of these memories at approximately one second. Combined with the average angular speed of the prawns (radians/s) this means that prawns can be separated by a full half of the arena while still exerting a considerable influence on each other' s behaviour. This potentially explains the strong performance of the mean field model in explaining the fine scale interactions between individuals. A number of physical [36]–[38], technological [39] and biological systems, including animals [9]–[11], [40], tissue cells [12], microorganisms [13], [41] are known to increase their collective order with density. Glass prawns are one additional example of such a system, which is particularly interesting since they are not known as gregarious or social species. By confining the prawns to a ring we facilitated their interactions and in doing so generated collective motion. This adds further support to the idea that collective motion is a universal phenomenon independent of the underlying interaction rules [4], [11], [42]. While we do not expect that prawns often find themselves confined in rings in a natural setting, they and other non-social animals do aggregate in response to environmental features such as food and shelter. Such environmental aggregations can, above a certain density, result in an apparently ‘social’ collective motion. The true value of this study, however, is found not in the addition of one more species to this growing list, but in demonstrating a rigorous methodology for selecting an optimal and multi-scale consistent model for the interactions between individuals in a group. We have used a combination of techniques to identify the optimal model for our experiments: Bayesian model selection, validation against global properties and consistency with biological reasoning. We applied Bayesian model selection to identify the model that best predicts the fine-scale interactions between prawns. This approach allows us to perform model selection in the presence of many competing hypotheses of varying complexity, while avoiding over fitting [17]. This indicated the selection of a non-Markovian model with a persistent ‘memory’ effect. We find that interactions are governed by a perceptual range which is symmetric about the focal individual which is somewhat greater than the average body length of the prawns (approximately radians). Reproduction of the large-scale dynamics is frequently used to validate mathematical models of biological systems, but presents only a necessary and not a sufficient condition for model validation. Indeed, all of the models we have assessed in this work can, with the appropriate parameters, generate aligned motion consistent with experiment. The fact that our mean-field model reproduces global dynamics, but fails at a fine-scale level is not particularly surprising. Mean-field models are not designed to reproduce spatially local dynamics [1]. More illuminating, however, is the failure of Markovian spatial models to reproduce the fine-scale dynamics when the locality of interactions between individuals is imposed. Models S1, S2, S3, S4 are variants of the standard one dimensional Vicsek self-propelled particle model [43], which has previously been validated against the global alignment patterns of marching locusts [11]. For the prawns these models perform poorly on both capturing the fine scale dynamics of interactions and in reproducing the large scale alignment patterns seen in the data. This inconsistency allowed us to reject standard self-propelled particle models as a good model of the data. To identify a better model we first visually inspected the interactions between the prawns. These observations suggested a ‘memory effect’, whereby a prawn would remain influenced by individuals beyond the moment of interaction. The resulting models are able reproduce the fine scale and large scale dynamics of the prawns, while also maintaining the biologically-intuitive locality of interactions between individuals. More generally, we would expect other examples of animal motion to be non-Markovian, with individuals taking time to react to others, to complete their own actions and also potentially reacting through memory of past situations. In this context, it is important to consider the limitations of recent studies identifying rules of interaction of fish [18], [19]. These studies concentrated on quantifying local interactions, but do not try to reproduce global properties. It may be that non-Markovian and other effects are needed to produce these properties. In what circumstances can we expect non-Markovian effects to play an important role in collective behaviour? Inference based on a Markovian model must account for behavioural changes of a focal individual in terms of their current environment. As such the crucial factor is how much the local environment changes between when the animal receives information and when it responds. Large changes in the local environment can be caused by long response times or by rapid movements of other animals relative to the focal individual. Where behavioural changes are strongly discontinuous, such as the binary one-dimensional movement in this study, non-Markovian effects may become especially important. This is because the focal individual may have to execute a number of small changes (such as stopping and turning through a several small angles) in order to register as having changed its direction of motion. Over the course of making many adjustments the environment can change dramatically from the moment that the change was initiated. We have compared the models on the large scale by evaluating the quality-of-fit between the distribution of large scale outcomes predicted by model simulations with that seen in experiments. The model we select from the fine scale analysis is also evaluated as the best on this large scale analysis, and produces simulation results that are qualitatively consistent with experiment (see Figure 6). Because the same model is selected from both analyses we have not been forced to weight the relative importance of each. In future it may be necessary to decide on an appropriate weighting of these different criteria where they disagree on the optimal model. The research presented here provides a first step towards the use of multi-scale inference in the study of collective animal behaviour and in other multi-level complex systems. The frame-by-frame movements of the prawns are imperfect representations of the true orientation, since a prawn will often stop or even drift slightly backwards without physically turning around. A Hidden Markov Model (HMM) allows the underlying orientation of the prawns to be discovered from the noisy frame-by-frame movements by demanding a higher degree of ‘evidence’ for a direction change, in essence only identifying direction changes when the prawn makes a sustained movement in the new direction. This gives a better estimate of the true orientation than given by the instantaneous velocity alone. We constructed a two-state HMM [44] for the observed changes in position of the prawn, as shown in Figure 7. The two states represent clockwise (CW) or anti-clockwise (anti-CW) orientation. In a CW oriented state it is assumed that the prawn will normally move in CW direction over the course of one frame, but because the prawns movements are noisy it may move in the reverse direction over short time periods while remaining oriented CW. We model the distribution of these movements as a Gaussian distribution. We further assume a symmetrical model, such that the distribution of movements in the CW state is anti-symmetric to the distribution of movements in the anti-CW state. Thus a movement of zero is equally probable in either state. We use the Baum-Welch algorithm [44], [45] to learn the transition probability and the mean and standard deviation of the Gaussian observation probability distribution, using data from single-prawn experiments. We then apply this learnt model to identify the most probable state sequence for each of the prawns in the three-, six- and twelve-prawn experiments, using the Viterbi algorithm [44], [46]. We further reduce the number of artifactual detected direction changes by removing any instances where a prawn changes direction twice within one second, since inspection suggests these events are caused by tracking errors. A given model, describes the probability of a change of direction for the focal prawn at time, conditioned on the current, and potentially past, positions of the other prawns, and and the parameters of the model. The likelihood for a given parameter set of the model is the probability of the data, , conditioned on the parameters and the model and is the product over both time steps and focal prawns of the probability for the observed outcome - either a change of direction or no change. Let equal one when prawn in experiment changes direction at time, and is zero otherwise, then, (8) where and indicate the number of experiments and the number of prawns in each experiment respectively. The marginal likelihood of the model is given by integration over the space, , of unknown parameters, (9) The prior distribution of the parameters, is chosen to represent the available knowledge about the parameters and is split into independent parts. We use the empirical observations in Figure 5 to inform the prior distribution on the interaction range and possible interaction strengths. The prior distribution over the number of interacting neighbours in the topological model is set to the entire possible range for the analysed six-prawn experiments, and the prior distribution for the memory factor is naturally between 0 (no memory) and 1 (permanent memory). The prior for the same parameter over different models is the same to allow fair comparison. (10) where indicates a continuous uniform distribution, indicates a discrete uniform distribution and is the Dirac delta function. Numerical integration over the appropriate parameters was performed using annealed importance sampling [47], with 1000 parameter samples. We select the most probable parameter values, for each model as those which maximise the posterior probability distribution, (11) where the posterior probability distribution is given in terms of the likelihood, prior distribution and model evidence defined above (12) In practice we evaluate the posterior probability for each parameter sample generated within the annealed importance sampling algorithm [47] and select the most probable for each model. Given the most probable parameter values (maximum a posteri) for a given model inferred from the fine scale data via equation 12, simulations of that model can be performed to assess the likely large scale results of the interactions the model encodes. To perform these simulations we treat individual prawns as particles moving on a circular ring. Each particle is initially set to have either CW or CCW motion at random. At each time step each particle, taken in a random order, moves around the ring in its direction of motion, moving a distance sampled from a distribution matched to the mean and variance of the experimentally observed motions (radians/s). After this motion, the distance between all the particles is calculated, and for each particle a decision is made whether to change the direction of motion, based on the rules encoded by the model being simulated. The time step used is s, which is matched to the time spacing in the analysed data. It is observed in model simulations that the rate at which the group aligns is highly dependent on the speed of individuals, which we have not attempted to model accurately. However, the final state after 360 seconds of simulation (the length of the experiments) is not sensitive to this factor. Therefore we evaluate the quality-of-fit between the model and experimental data by examine the distribution of final states in the experiments and simulations – that is, how many individuals are travelling clockwise when the experiment or simulation ends. We average this over the final 10 seconds of the experiment or simulation to increase the accuracy of this judgement. The quality-of-fit for the model is given by the Kullback-Leibler (KL) divergence [34], from the experimental distribution of outcomes, to the simulated distribution, . This is a canonical measure of how well one distribution (the simulated outcomes) approximates another (the experimental outcomes). If is the proportion of experiments where prawns are travelling clockwise, and similarly the proportion of simulations where particles are travelling clockwise, then the divergence is given by (13) where is the total number of prawns in the experiment or simulation. We calculate this divergence between experiment and simulation for scenarios with 3,6 and 12 prawns to check for consistency over varying group size. The statistical significance of these divergences can be calculated using the G-statistic, , where is the number of experiments, and the KL divergence is evaluated using the natural logarithm. The null hypothesis that the experimental results come from the simulated distribution implies a -distribution for the G-statistic [35]. This article is a revised version of a paper of the same title [48] that was previously published in PLOS Computational Biology and was subsequently retracted when a computational error was discovered.

The collective movement of animals in a group is an impressive phenomenon whereby large scale spatio-temporal patterns emerge from simple interactions between individuals. Theoretically, much of our understanding of animal group motion comes from models inspired by statistical physics. In these models, animals are treated as moving (self-propelled) particles that interact with each other according to simple rules. Recently, researchers have shown greater interest in using experimental data to verify which rules are actually implemented by a particular animal species. In our study, we present a rigorous selection between alternative models inspired by the literature for a system of glass prawns. We find that the classic theoretical models do not accurately predict either the fine scale or large scale behaviour of the system. Instead, individual animals appear to be interacting even when completely separated from each other. To resolve this we introduce a new class of models wherein prawns ‘remember‚ their previous interactions, integrating their experiences over time when deciding to change behaviour. These show that the fine scale and large scale behaviour of the prawns is consistent with interactions only between individuals who are close together.

Abstract Introduction Results Discussion Materials and Methods

statistical mechanics applied mathematics bayes theorem marine biology animal behavior mathematics theoretical ecology stochastic processes zoology freshwater ecology complex systems theoretical biology probability density biology probability theory behavioral ecology physics ecosystem modeling ecology computational biology markov model

2013

Multi-scale Inference of Interaction Rules in Animal Groups Using Bayesian Model Selection

Biomolecular recognition entails attractive forces for the functional native states and discrimination against potential nonnative interactions that favor alternate stable configurations. The challenge posed by the competition of nonnative stabilization against native-centric forces is conceptualized as frustration. Experiment indicates that frustration is often minimal in evolved biological systems although nonnative possibilities are intuitively abundant. Much of the physical basis of minimal frustration in protein folding thus remains to be elucidated. Here we make progress by studying the colicin immunity protein Im9. To assess the energetic favorability of nonnative versus native interactions, we compute free energies of association of various combinations of the four helices in Im9 (referred to as H1, H2, H3, and H4) by extensive explicit-water molecular dynamics simulations (total simulated time > 300 μs), focusing primarily on the pairs with the largest native contact surfaces, H1-H2 and H1-H4. Frustration is detected in H1-H2 packing in that a nonnative packing orientation is significantly stabilized relative to native, whereas such a prominent nonnative effect is not observed for H1-H4 packing. However, in contrast to the favored nonnative H1-H2 packing in isolation, the native H1-H2 packing orientation is stabilized by H3 and loop residues surrounding H4. Taken together, these results showcase the contextual nature of molecular recognition, and suggest further that nonnative effects in H1-H2 packing may be largely avoided by the experimentally inferred Im9 folding transition state with native packing most developed at the H1-H4 rather than the H1-H2 interface. Molecular recognition is the basis of biological function. For different parts of the same molecule or different molecules to recognize one another, a target set of interactions need to be favored while other potential interactions are disfavored. Biomolecules accomplish these simultaneous tasks via the heterogeneous interactions encoded by their sequences. For proteins, such energetic heterogeneity is enabled but also constrained by a finite alphabet of twenty amino acids. Thus the degree to which non-target interactions can be avoided through evolutionary optimization is limited [1,2]. Conflicting favorable interactions, referred to as frustration, are often present in biological systems. From a physical standpoint, it is almost certain that some of the frustration is a manifestation of the fundamental molecular constraint on adaptation, although under certain circumstances frustration can be exploited to serve biological function [3,4]. Protein folding entails intra-molecular recognition. Early simulations suggested that nonnative contacts can be common during folding [5]. This predicted behavior applies particularly to models embodying a simple notion of hydrophobicity as the main driving force [6,7]. Experimentally, however, protein folding is thermodynamically cooperative [7,8]. Folding of many single-domain proteins does not encounter much frustration from nonnative interactions in the form of kinetic traps [9]. Celebrated by the consistency principle [10] and the principle of minimal frustration [11], these empirical trends have inspired Gō-like modeling, wherein native-centric interactions are used in lieu of a physics-based transferable potential [12–14]. Extensions of this approach allow nonnative interactions to be treated as perturbations in a largely native-centric framework [15–17]. The success of these models poses a fundamental challenge to our physical understanding as to why, rather non-intuitively, natural proteins are so apt at avoiding nonnative interactions. Solvation effects must be an important part of the answer [18], as has been evident from the fact that coarse-grained protein models incorporating rudimentary desolvation barriers exhibit less frustration and higher folding cooperativity than models lacking desolvation barriers [7,19,20]. More recently, and most notably, folding of several small proteins has been achieved in molecular dynamics studies with explicit water [21,22]. Nonnative contacts are not significantly populated within sections of the simulated trajectories identified as folding transition paths [23] though they do impede conformational diffusion [24]. These advances suggest that certain important aspects of protein physics are captured by current atomic force fields, although they still need to be improved to reproduce the high degrees of folding cooperativity observed experimentally [22,25–28]. In this context, it is instructive to ascertain how atomic force fields, as they stand, disfavor nonnative interactions, so as to help decipher molecular recognition mechanisms in real proteins. We take a step toward this goal by comparing the stabilities of native and nonnative configurations of fully formed helices from a natural protein. By construction, this approach covers only a fraction of all possible nonnative configurations and therefore only provides, albeit not unimportantly, a lower bound on the full extent of frustration. Nonetheless, because of its focus on tractable systems, we obtain a wealth of reliable simulation data from which physical insights are gleaned. We do so by applying explicit-water molecular dynamics simulations to compute potentials of mean force (PMFs) between various helices [29] of the E. coli colicin immunity protein Im9 [30]. Im9 is a small single-domain protein that undergoes two-state-like folding [31,32] to a native structure with four helices packed around a hydrophobic core [33]. Its folding mechanism and that of its homolog Im7 have been extensively characterized experimentally [30–40] and theoretically [41–46]. Of particular relevance to our study are experimental Φ-value analyses suggesting that the Im9 folding transition state has a partially formed hydrophobic core stabilized by interactions between helix 1 (H1) and helix 4 (H4), whereas helix 3 (H3) adopts its native conformation only after the rate-limiting step of folding [32]. These experimental inferences have since been rationalized by simulations showing that H1 and H4 are formed whereas about one half of helix 2 (H2) remains unstructured in the Im9 transition state [41], and that, unlike Im7, there is no significant kinetic trap along the Im9 folding pathway [45,46]. Building on these advances, our systematic PMF analysis provides a hitherto unknown perspective on these hallmarks of Im9 folding. Notably, we found significant packing frustration between H1 and H2, viz. , a nonnative packing orientation can achieve a lower free energy than that afforded by the native packing of these two helices in isolation. Superficially, this simulation result seems at odds with experiments indicating little frustration in Im9 folding. On closer examination, however, our discovery provides an unexpected rationalization for experiments indicating that folding is initiated by the more stabilizing H1-H4 interactions rather than by H1-H2 packing. Because the H1-H2 packing frustration can be circumvented by following such a kinetic order, our finding suggests that the Im9 folding pathway might have evolved to avoid a potential H1-H2 kinetic trap. This example underscores that the inner workings of molecular recognition can be rather subtle and deserves further exploration, as will be elaborated below. We begin by investigating the free energy landscape for the association of H1 with H2, a packing interaction that accounts for the largest two-helix interface in the native state of Im9, burying 5. 3 nm2 or 17% of the total surface area of H1 and H2. Throughout this study, surface areas of helical bundles are computed as the solvent-accessible surface areas of the given bundles in isolation, irrespective of the solvent exposure of the configurations in the complete Im9 folded structure. Using an enhanced sampling technique known as umbrella sampling with virtual replica exchange (US-VREX, see Methods) for restrained helical configurations at systematically varied target packing angles, we compute PMFs for H1-H2 association in the absence of their intervening loop (the H1→H2 system in Fig 1B and Table 1). The PMFs are determined for the native orientation as well as for nonnative orientations and nonnative crossing angles entailed by the imposed rotational preferences (Methods and S1 Text). Our technique allows these simulations to converge rapidly (S1 Fig). Each PMF is then integrated over a free-energy basin to provide a binding free energy, ΔGbind, for a specific inter-helix geometry. Unexpectedly, H1-H2 association is favored by a 20–30° positive rotation of H1 against H2. Binding in this nonnative orientation is 10–12 kJ/mol more stable than that in the native orientation (black circles in Fig 2A and Table 2), a free energy difference equivalent to a ~50-fold increase in bound population (S1 Text). In contrast, the binding free energy profiles for rotating H2 against H1 (Fig 2A, red squares) or changing the H1-H2 crossing angle (Fig 2A, blue triangles) indicate that the state corresponding to native packing (0° angle in Fig 2A) is situated well within the basin of lowest free energy with respect to these degrees of freedom, although a ≤50° positive change in H1-H2 crossing or a ≤20° negative rotation of H2 against H1 would leave the system approximately iso-energetic with the native packing (Fig 2A). As mentioned, these binding energies are computed from PMFs such as those in Fig 2B and S2 Fig. A broader view of the orientation-dependent H1-H2 packing free energy landscape can be seen in Fig 2C. Instead of fixing either H1 or H2 in its native orientation (as in Fig 2A), Fig 2C provides the relative favorability of packing orientations resulting from simultaneous rotations of H1 and H2. This two-dimensional PMF is generated by combining sampling data for H1 and H2 rotations under harmonic biasing potentials (S1 Text). It is clear from this two-dimensional landscape that native packing [ (H1, H2) rotations equal (0°, 0°) ] is less favored than the free energy minimum at (+19°, +4°). Indeed, this minimum is situated in a rather broad basin encompassing many nonnative orientations with simultaneous H1 rotation from approximately +5° to +25° and H2 rotation from approximately ‒3° to +15° that are energetically more favorable than the native H1-H2 orientation (0°, 0°). Fig 2C reveals further that there exists another basin of favorable nonnative H1-H2 packing for which both helices rotate by approximately ‒20°. In short, our systematic analysis in Fig 2 demonstrates unequivocally that packing frustration exists in Im9, in that when H1 and H2 are considered in isolation, nonnative packing is favored over native packing. To assess the prospect that intervening loop residues may provide additional guidance for native packing of H1 against H2, we also simulate this helix-loop-helix as a single chain (H1LH2 system; Table 1). Because the covalent connection of H1 to H2 is incompatible with the large helical separations used in our importance sampling, we study the H1LH2 system without inter-helical distance bias in simulations initiated in either the native state or one of 20 different nonnative orientations in which H1 or H2 is rotated by ±10–50°. [Because the actual rotations sampled during simulations are close to those targeted by the restraining potentials (S4 Fig), we do not distinguish between target and actual rotations hereafter]. Although these simulations do not converge to a single conformational distribution, they show broad sampling of H1 rotation with a stable or metastable state near +20° rotation of H1, even when simulation is initiated at the native packing angle (S6 Fig). To explore how the H1-H2 packing frustration might be overcome in Im9 folding, we next investigate the impact of the rest of the protein on the packing between H1 and H2 by computing binding free energies for the association of H1 and H2 not in isolation but in the presence of additional protein fragments involving the other two helices H3 and H4 as well as loop and terminal residues. The conformations of the loop and terminal residues in our simulations are restrained to those in the Im9 PDB structure. We first consider the association of H1 with a bundle comprising helices 2,3, and 4 connected by their intervening loops and extending to the protein' s C-terminus (H1→H2LH3LH4C; Table 1). Interestingly, for this system, native packing is found to be 13 ± 3 kJ/mol more favorable than the nonnative packing resulting from a +30° rotation of H1 (Table 2). The very fact that a nonnative rotation of H1 is substantially favored in H1→H2 (Fig 2A and Table 2) but disfavored in H1→H2LH3LH4C (Table 2) demonstrates clearly that some components of the H2LH3LH4C bundle besides H2 are crucial for overcoming the H1-H2 packing frustration and guiding H1 to pack natively. Furthermore, because native packing is favored in H1→H2LH3LH4C despite the residues N-terminal to H1 (including a short 3–10 helix) being excluded in this model system, these N-terminal residues are likely not necessary for ensuring native packing of H1 against the rest of the Im9 protein. We now dissect the H2LH3LH4C bundle to ascertain the contributions from different parts of this bundle to native H1 packing. To this end, binding free energies for the association of H1 with a variety of subsets of H2LH3LH4C are computed. We first consider a bundle comprising helices 2 and 4 (H1→H2/H4; Table 1). Somewhat surprisingly, native packing in the H1→H2/H4 system is disfavored by as much as 22 ± 1 kJ/mol when compared against nonnative packing with H1 rotated by +30°, even more than the corresponding nonnative preference of 10 ± 1 kJ/mol for H1→H2 (Table 2). This observation implies that H4 by itself is not promoting H1-H2 native packing and therefore H3, loops, and/or the C-terminus must be responsible for driving native packing of H1 with H2LH3LH4C. Indeed, when compared against H2/H4, the presence of these other elements in H2LH3LH4C results in a 26 ± 1 kJ/mol preference for native H1 packing and a 9 ± 3 kJ/mol discrimination against nonnative H1 packing with a +30° rotation (Table 3). To better pinpoint the role of H3 in this intra-molecular recognition process, we compute binding free energies for the association of H1 and a bundle comprising helices 2,3 and 4 but without the intervening loops and the C-terminus (H1→H2/H3/H4; Fig 1D and Table 1). For this model system, native packing is less favorable than +30° rotation of H1 by 11 ± 3 kJ/mol (Table 2). Nonetheless, in comparison to H1→H2/H4, the inclusion of H3 favors native packing more than it favors nonnative packing with a +30° rotation of H1 (Table 2). This observation indicates that H3 is capable of correcting part of the nonnative tendencies of H1 imparted by its interactions with a bundle comprising only of H2 and H4; but H3 is insufficient to ensure native packing in the absence of the connecting loops and/or the C-terminus. To explore whether inclusion of residues neighboring H4 may alter its effect on H1-H2 packing, we consider three residues immediately N-terminal to H4 (Asp62, Ser63, and Pro64). These residues are chosen because they are known to associate directly with H1 in the NMR structure [47] and thus they may contribute positively to native intra-molecular recognition. Consistent with this expectation, once these three residues are included, the H1-binding free energies in the resulting H1→H2/NH4 system (Table 1) for native packing and nonnative +30° rotation of H1 become essentially energetically equivalent (ΔΔGbind = 2 ± 6 kJ/mol; Table 2). Inasmuch as promoting native H1-binding is concerned, this represents a significant improvement over H1→H2/H4 that favors the +30°-rotated nonnative packing by 22 ± 1 kJ/mol (Table 2). Indeed, in the context of H1→H2/H4, addition of these N-terminal flanking residues assists native packing by 31 ± 5 kJ/mol, much more than the 7 ± 3 kJ/mol increase in stability they also impart on the nonnative packing of H1 with a +30° rotation (Table 3). These numbers underscore the important role of Asp62, Ser63, and Pro64 in discriminating against nonnative packing of H1. Another set of helix-flanking residues that may assist native packing in Im9 is its C-terminus. Such an effect is expected because a +30° rotation of H1 would likely place its constituent residue Phe15 into a steric clash with the C-terminal residue Phe83 (S7 Fig) and thus existence of the C-terminus should discriminate against such a rotation of H1. To evaluate this hypothesis, we compute H1-binding free energies with a bundle comprising H2 and H4 as well as the protein' s C-terminus (H1→H2/H4C; Table 1). Similar to the addition of Asp62, Ser63, and Pro64 N-terminal to H4 in H2/NH4 bundle, inclusion of the C-terminus in H2/H4C eliminates the strong nonnative bias in H1→H2/H4, resulting in essentially no discrimination between the native orientation and a +30° rotation of H1 (ΔΔGbind = 1 ± 3 kJ/mol; Table 2). Relative to H1→H2/H4, addition of the C-terminus not only favors native packing by 6 ± 2 kJ/mol but also directly disfavors +30° rotation of H1 by 17 ± 2 kJ/mol (Table 3). The latter penalization of nonnative packing (which does not occur in H1→H2/NH4) is consistent with the aforementioned steric consideration (S7 Fig). Interestingly, the native-promoting effects of N- and C-terminal extensions to H4 are essentially additive. When both extensions are added to H4, the H2/NH4C system (Table 1) is sufficient to favor native packing of H1 by 14 ± 6 kJ/mol over the nonnative packing with +30° rotation of H1 (Table 2). After analyzing systems involving H2, we now turn to the intra-molecular recognition between H1 and H4 without involving H2. Native H1-H4 packing constitutes the second largest two-helix interface in the Im9 folded structure, burying 3. 7 nm2 which amounts to 13% of the sum of individual surface areas of H1 and H4. PMFs for helices 1 and 4 in isolation (H1→H4; Fig 1C and Table 1) are computed in the native orientation as well as nonnative orientations resulting from rotations of H1 or H4. When H1 is rotated while H4 is fixed, native packing is favored (Fig 3A, black circles); however, when H4 is rotated with H1 fixed, a +30° nonnative rotation of H4 leads to 5 ± 1 kJ/mol stabilization (decrease in ΔGbind) relative to native (red squares in Fig 3A and Table 2). Distance-dependent PMFs for the native orientation and ±30° rotations of H4 are shown in Fig 3B, indicating that the favored nonnative packing at +30° is attained at an H1-H4 separation slightly larger than native by about 0. 1 nm. The two-dimensional PMF (Fig 3C) as a function of H1 and H4 rotation angles shows further that native H1-H4 packing (0°, 0°) is situated at the periphery of a broad basin of favored orientations centered roughly around (+10°, +10°). The same two-dimensional landscape suggests that H1 rotations of ≥ +50° or ≤ ‒50° can also be favored with little or no H4 rotation. We noted earlier that a 3-residue N-terminal extension to H4 directly contacts H1 in the native state and that the inclusion of these residues assisted the native packing of H1 against a bundle comprising helices H2 and H4. Consistent with that observation, these three residues—Asp62, Ser63, and Pro64—likewise assist the native packing of H1 against H4, viz. , their inclusion in the H1→NH4 system (Table 1) makes native packing (ΔGbind = ‒44 ± 1 kJ/mol) significantly more favorable than the nonnative packing with a +30° rotation of H4 (ΔGbind = ‒21 ± 2 kJ/mol) while still favoring native orientation of H1 (Table 2). We conclude from these results that helices H1 and H4 are nearly capable of associating in native-like conformations by themselves in isolation; and that they can certainly achieve native packing with the assistance from the 3-residue N-terminal extension to H4. These results suggest that Im9 residues 12–23 and 62–78 may serve as major components of a native-like folding nucleus. To better understand the driving force for nonnative H1-H2 packing, the potential energies between specific pairs of amino acid residues on the H1-H2 interface in the native orientation are compared against those in the nonnative orientation with a +30° H1 rotation. We make this comparison for helix-helix center of mass distance di0 = 1. 10 nm in both the native and non-native configurations, wherefore each pair of helices in question is in close spatial contact (Fig 4). The analysis indicates a prominent role by the more favorable Lennard-Jones interactions between interfacial residue pairs Glu14-Met43, Leu18-Phe40, and Ile22-Phe40 in favoring the nonnative packing, whereas electrostatic interactions between these residue pairs are of similar strengths for the native and nonnative packing orientations. In contrast, the interaction between Ile22 and Leu33 favors native packing, but its effect is more than compensated by the aforementioned multiple residue-residue interactions that drive nonnative packing such that a +30° rotation of H1 is favored over the native orientation for H1-H2 packing in isolation. It is noteworthy, however, that while these residue-residue energetic effects can be significant individually (Fig 4) and collectively (Table 2), they are not accompanied by obvious, drastic structural changes at the level of residue-residue contacts. When contacts between residues on different helices at a helix-packing interface are identified by a commonly used proximity threshold, contact probabilities between the helices are seen to remain essentially unchanged upon a +30° H1 native-to-nonnative rotation in both the H1→H2 and H1→ H2LH3LH4C systems (S8 Fig). Seeking physical reasons for favoring native packing in H1→ H2LH3LH4C but not in H1→H2, we compare the potential energies of these systems in the native and the +30° H1-rotated nonnative configurations (Fig 5). When potential energies are analyzed by the molecular species involved in the interactions, for H1→H2, solvent-protein (solvent-helix) interactions are more unfavorable with nonnative rotation of H1 by +30°, but this effect is overwhelmed by larger, favorable changes in solvent-solvent and intra- and inter-helix interactions (Fig 5A). More specifically, this nonnative H1 rotation favors inter-helix Lennard-Jones interactions (as exemplified by the three residue pairs circled in red in Fig 4A) as well as intra-helix and solvent-solvent electrostatic interactions (Fig 5A), netting an overall favorable (more negative) potential energy for the nonnative orientation (Fig 5A, “sum”). In contrast, the corresponding analysis for H1→H2LH3LH4C yields a set of average potential energies that favors the native state overall (Fig 5B, native “sum” more negative than nonnative). This potential energy (enthalpic) trend is consistent with the above PMF/binding free energy prediction that the native orientation is favored for H1→H2LH3LH4C (Table 2), though entropic effects may make additional contribution to the stability of native packing of H1 against H2LH3LH4C (see below). Because nonnative +30° H1 rotation has opposite effects on intra-H2 (Fig 5A) versus intra-H2LH3LH4C (Fig 5B) Coulomb energies, one of the reasons for disfavoring nonnative +30° H1 rotation in H1→H2LH3LH4C is that this rotation of H1 induces energetic strain within H2LH3LH4C, resulting in a destabilizing increase in intra-H2LH3LH4C Coulomb energy collectively, whereas the same +30° H1 rotation leads to an overall stabilizing decrease in intra-H2 Coulomb energy. The atomic basis of this difference remains to be analyzed. To gain further insight into the differential effects of H2 and H2LH3LH4C on the favorability of the native orientation upon H1 binding, we resolve the distance-dependent H1→H2 and H1→H2LH3LH4C PMFs (Fig 6A and 6B, respectively) into their enthalpic (Fig 6C and 6D) and entropic (Fig 6E and 6F) components. Since the backbones of the helical elements in our simulation systems are restrained to be essentially rigid, the entropic contributions computed here originate almost exclusively from the water solvent and sidechain degrees of freedom, whereas contributions from mainchain conformational entropy are negligible in comparison. Despite sampling uncertainties, several likely trends can be quite clearly discerned: For H1→H2, the lower PMF (ΔG) minimum for the nonnative orientation (Fig 6A) is driven by enthalpy (lower ΔH minimum for +30° H1 rotation than for native in Fig 6C). This effect is partially, but not completely, compensated by the entropic component of the free energy, ‒TΔG. The latter is seen favoring native packing in Fig 6E (red curve below blue curve at distance marked by vertical blue dashed line), although the differences are largely within error bars. Entropy has a similar effect on H1→H2LH3LH4C in stabilizing native packing (Fig 6F). In this case however, unlike H1→H2, enthalpy is also favorable (though only slightly) to the native state (Fig 6D, see also Fig 5B), thus the entropic and enthalpic effects reinforce each other, yielding a ΔG favorable to native packing for H1→H2LH3LH4C (Fig 6B). It should be noted that the trends of entropic stabilization seen here in Fig 6 are similar to those exhibited by a pair of poly-alanine or poly-leucine helices [29]. In both cases, the entropic trends are likely manifestations of the well-recognized solvent-entropic origin of hydrophobic interactions at ambient temperatures. Every helix-helix association in Fig 6 entails an enthalpic barrier at separation ≈1. 5 nm (Fig 6C and 6D). As implied by the absence of PMF barriers at these positions (Fig 6A and 6B), the enthalpic barriers here are compensated by a larger decrease in entropic free energy at the same positions (Fig 6E and 6F). Further examples of enthalpic barriers and entropic compensations are provided in S2 Fig. These results are consistent with burial of hydrophobic surfaces being concomitant with increase in solvent (water) entropy at room temperature and the idea that enthalpic barriers to protein folding [20,29,49,50] may arise largely from steric dewetting [29]. Because steric dewetting creates voids (between the approaching helices in the present cases; S9 Fig), it leads to volume barriers [29] such as those seen in Fig 6G and 6H. As has been discussed, such volume barriers probably amount to part of the activation volume of protein folding [51,52]. For the systems studied in Fig 6, it is not surprising that the enthalpic and volume barriers are higher for H1→H2LH3LH4C than for H1→H2 because the former binding process buries a significantly larger protein surface area. Therefore, we expect a larger transient void volume between the approaching helices before close packing is achieved for H1→H2LH3LH4C than for H1→H2. It is interesting to note that, perhaps because void volumes are largely a consequence of geometry and less of energetics, the volume barrier heights in Fig 6G and 6H are essentially insensitive to the difference between native and nonnative packing. To recapitulate, we have conducted a systematic analysis of the relative stability of native versus nonnative packing of helices in the Im9 protein as a means to address the physical basis of biomolecular recognition. These results are summarized schematically in Fig 7: Relative to native packing, three nonnative configurations (H1→H2, H1→H2/H4, and H1→H2/H3/H4, each with H1 rotated) are significantly stabilized whereas one other nonnative packing orientation (H1→H4 with H4 rotated) is mildly stabilized. Other Im9 systems that we have simulated either favor the native configuration or essentially do not discriminate between native and nonnative packing. As emphasized at the outset, our method is designed to characterize packing frustration of constrained, locally native protein substructures by varying the orientation between interacting substructures that are rigid by construction, viz. , the secondary structure (main-chain conformation) of each of the helices is essentially fixed. It follows that while our substantial computational effort has succeeded in gaining structurally and energetically high-resolution information about frustration that is novel and complementary to that obtained from our previous coarse-grained chain model study of Im7 and Im9 [45], the present investigation—unlike our coarse-grained modeling [45]—cannot by itself address certain general questions regarding folding pathways such as the viability of nucleation-condensation mechanisms [53] because backbone conformational freedom is not treated. For the same reason, the present method does not tackle frustration involving disordered, flexible main-chain segments that may adopt locally nonnative conformations. A notable example in this regard is the second helix of Im7. Among the four respective helices in Im7 and Im9, the amino acid sequence of the second helix varies the most between the homologs [30]. The second Im7 helix has been identified as a part of the protein which is disordered and participates in nonnative interactions that stabilize a kinetically trapped folding intermediate during the process of non-two-state folding of Im7 [45]. However, revealingly, the significant role of a disordered H2 in frustrating Im7 folding is not reflected by its behavior as an ordered helix: Unlike the H1→H2 system of Im9, the H1→H2 system of Im7 exhibits no favorable nonnative packing (S10 Fig). This finding underscores the importance of disordered conformations to frustration in globular protein folding, an effect that the present analysis has not addressed. From a broader perspective, such effects have to be even more critical for molecular recognition among intrinsically disordered proteins [54,55]. Notwithstanding aforementioned limitations of the present approach, several important lessons can already be learned from our extensive computational investigation. First, a majority of the helical systems that we consider favor native packing, indicating that the Im9 amino acid sequence encodes a sufficiently strong native bias such that the native structure can be recognized by the folding protein. Second, frustration exists, manifested most notably by—but not necessarily limited to—the significantly stabilized nonnative H1-H2 packing. Although the conformational space accessible to an 86-residue polypeptide is vast compared to what is accessible via contemporary simulation and thus our ability to identify all possible sources of frustration is limited, the systematic approach taken in the present study does pinpoint one class of frustrated configurations. Third, the native fold is favored overall despite frustration, at least within the class of configurations we tested, because nonnative H1-H2 packing is destabilized when other parts of the protein, especially H4 and its flanking residues, are involved in the interaction. A logical inference from our results is that favorable nonnative interactions can be largely suppressed during Im9 folding by favoring trajectories that assemble H1 and H2 not in isolation but only in the presence of H4 plus flanking residues. Such preference would help avoid kinetic traps to facilitate known two-state folding behaviors of Im9 [32,36]. This expectation is consistent with the Im9 folding mechanism deduced from experimental phi-values (ΦF) by Radford and coworkers, who determined that residues in H2 have the lowest ΦF-values among H1, H2, and H4; but ΦF-values are higher for the hydrophobic residues in H1 and H4. This and other findings led them to conclude that the H1-H4 interface “is the most structured region in the transition state ensemble”, and that the native configuration of H1, H2, and H4 is partially formed in the transition state whereas H3 is formed after the rate-limiting step [32]. Since our simulation results also suggest that H1-H2 interactions should be weaker than those between H1-H4 to minimize kinetic trapping, our data offer a physical rationale as to why the Im9 folding pathways might have evolved. A general theoretical formalism due to Wolynes and coworkers provides quantitative estimates of local frustration [3,42,56]. Of relevance here is their configurational frustration index, which quantifies the likelihood of a pair of residues that are in contact in a protein’s native structure to be engaged in favorable nonnative interactions in alternate conformations. Their web-based “Protein Frustratometer” algorithm [56] predicts a high configurational frustration region in Im9 encompassing residues 25–38, which overlaps substantially with H2 (residues 30–44, Table 1). In contrast, H1 and H4 are predicted to be situated in lower configurational frustration regions on average (S11 Fig). These predictions are consistent with, and therefore lend further support to the aforementioned perspective emerging from our simulation results. It is noteworthy, however, that the Frustratometer-computed configurational frustration Fc of Im7 is not noticeably higher on average than that of Im9 (S11 Fig), notwithstanding the fact that folding is significantly more frustrated for Im7 than for Im9 experimentally [30–40]. In particular, while the predicted frustration of H4 is higher for Im7 than for Im9 (which is consistent with H4’s involvement in nonnative interactions with H2 in Im7 folding), the predicted configurational frustration of H2 of Im7 is similar to, or even slightly lower than that of Im9. It would be instructive to investigate whether this apparent inability of the algorithm to clearly delineate the key experimental difference in Im7 and Im9 folding kinetics is because the decoy inter-residue contact distances used to compute configurational frustration Fc [56] are insufficient to fully capture the conformational possibilities of a disordered H2 that make strong nonnative interactions in Im7 possible [45]. Intuitively, this limitation might be similar or even related to the impossibility of discerning Im7 frustration from the packing of fully formed H1 and H2 alone (S10 Fig) despite the fact that many of the favorable nonnative interactions in Im7 folding are between residues in H1 and H2. This question deserves further attention. Owing to the high computational cost of the present approach, applications have been confined to the commonly used OPLS-AA/L force field. While useful insights are gained as reported above, it should be noted that current molecular dynamics force fields can be limited in their ability to accurately model disordered protein states (reviewed in [57,58]) and to capture subtle effects such as conformational switches [59]. It is important, and would be instructive, to assess how discrimination against nonnative interactions is affected by ongoing efforts to improve current force fields [57,58]. Much work remains to be done before the physical basis of biomolecular recognition can be fully deciphered. Umbrella sampling (US) [60] simulations are employed to quantify the extent to which the residues in two pre-folded regions of the protein are sufficient to drive native-like association. Specifically, we compute orientation-specific free energies for the binding of H1 to a systematic selection of helices from other parts of the protein with and without connecting loops. To enhance computational tractability, the latter helical bundles are prevented from unfolding or changing their relative orientations by imposing harmonic restraints on the positions of all Cα atoms with force constants of 1000 kJ/mol/nm2. Unfolding of H1 is disallowed by Cα position restraints that are enforced only in the Cartesian y and z dimensions, using the same force constant. The US order parameter is the magnitude of the Cartesian x component of the vector connecting the centers of mass of Cα atoms in the two bundles. This linear displacement, d, is harmonically restrained at a specified target value, di0, in each umbrella i, with a force constant of 2000 kJ/mol/nm2. For each system, 39 umbrellas span 0. 7 nm ≤ di0 ≤ 2. 6 nm in 0. 05 nm increments. To further enhance the rate of convergence in these US simulations, we allow equilibrium exchange of umbrellas using the virtual replica exchange (VREX) approach [61,62]. Further details of the US-VREX approach are provided in S1 Text. US-VREX simulations are conducted for the H1→H2, H1→H4, H1→NH4, H1→H2/H4, H1→H2/NH4, H1→H2/H4C, H1→H2/NH4C, H1→H2/H3/H4, and H1→H2LH3LH4C systems of Im9 (Table 1), where the arrow separates the two interacting fragments (bundles) under consideration. The H1→H2 system of Im7 is also simulated using the same method. The two bundles in any given system are on equal footing because their association with each other is mutual. The arrow in our notation serves merely to indicate their spatial association without regard to the arrow’s direction. Each system is simulated for 100 ns/umbrella, except for H1→H2 and H1→H2LH3LH4C in the native orientation and with nonnative H1 rotation by +30°, which are simulated for 500 ns/umbrella. In total, these US-VREX simulations comprise >300 μs of simulated time. Despite the application of position restraints to prevent the helices from unfolding or changing their relative orientations during PMF computations, the rotation angle of helices varies within ±10° of the target packing angle. We identify the simulated systems by the angles to which they are targeted. The H1LH2 system comprising H1, H2, and their connecting loop is simulated from the native [47] and twenty different nonnative initial conformations generated by removing inter-helical loop residues 24–29, rotating H1 or H2 about its long axis by ±10°, ±20°, ±30°, ±40° and ±50°, and then modeling loop residues using the prediction program Loopy [63]. Secondary structure is maintained while allowing changes in helical rotation and separation by applying intra-helical distance restraints on all backbone atom pairs with force constants of 1000 kJ/mol/nm2. Each simulation covers 1 μs, with the first 125 ns discarded in subsequent analysis. MD simulations are conducted with version 4. 5. 5 of the GROMACS simulation package [64]. The water model is TIP3P [65]. Protein is modeled by the OPLS-AA/L parameters [66,67]. Simulation systems are neutralized and excess NaCl is added at 0. 4 M, mimicking experimental conditions [31,68]. Water molecules are rigidified with SETTLE [69] and protein bond-lengths are constrained with P-LINCS [70]. Lennard-Jones interactions are evaluated using a group-based cutoff and truncated at 1 nm without a smoothing function. Coulomb interactions are calculated using the smooth particle-mesh Ewald method [71,72] with a Fourier grid spacing of 0. 12 nm. Simulations are in NPT ensembles by isotropic coupling to a Berendsen barostat [73] at 1 bar with a coupling constant of 4 ps and temperature-coupling the simulation system using velocity Langevin dynamics [74] at 300 K with a coupling constant of 1 ps. The integration time step is 2 fs. The nonbonded pair-list is updated every 20 fs. Further details are provided in S1 Text and S1 Table.

Biomolecules need to recognize one another with high specificity: promoting “native” functional intermolecular binding events while avoiding detrimental “nonnative” bound configurations; i. e. , “frustration”—the tendency for nonnative interactions—has to be minimized. Folding of globular proteins entails a similar discrimination. To gain physical insight, we computed the binding affinities of helical structures of the protein Im9 in various native or nonnative configurations by atomic simulations, discovering that partial packing of the Im9 core is frustrated. This frustration is overcome when the entire core of the protein is assembled, consistent with experiment indicating no significant kinetic trapping in Im9 folding. Our systematic analysis thus reveals a subtle, contextual aspect of biomolecular recognition and provides a general approach to characterize folding frustration.

Abstract Introduction Results Discussion Methods

classical mechanics protein interactions molecular dynamics potential energy protein folding protein structure thermodynamics reaction dynamics physical chemistry proteins chemistry molecular biology free energy physics biochemistry biochemical simulations transition state biology and life sciences physical sciences computational chemistry computational biology macromolecular structure analysis

2017

Molecular recognition and packing frustration in a helical protein

The human DARC (Duffy antigen receptor for chemokines) gene encodes a membrane-bound chemokine receptor crucial for the infection of red blood cells by Plasmodium vivax, a major causative agent of malaria. Of the three major allelic classes segregating in human populations, the FY*O allele has been shown to protect against P. vivax infection and is at near fixation in sub-Saharan Africa, while FY*B and FY*A are common in Europe and Asia, respectively. Due to the combination of strong geographic differentiation and association with malaria resistance, DARC is considered a canonical example of positive selection in humans. Despite this, details of the timing and mode of selection at DARC remain poorly understood. Here, we use sequencing data from over 1,000 individuals in twenty-one human populations, as well as ancient human genomes, to perform a fine-scale investigation of the evolutionary history of DARC. We estimate the time to most recent common ancestor (TMRCA) of the most common FY*O haplotype to be 42 kya (95% CI: 34–49 kya). We infer the FY*O null mutation swept to fixation in Africa from standing variation with very low initial frequency (0. 1%) and a selection coefficient of 0. 043 (95% CI: 0. 011–0. 18), which is among the strongest estimated in the human genome. We estimate the TMRCA of the FY*A mutation in non-Africans to be 57 kya (95% CI: 48–65 kya) and infer that, prior to the sweep of FY*O, all three alleles were segregating in Africa, as highly diverged populations from Asia and ≠Khomani San hunter-gatherers share the same FY*A haplotypes. We test multiple models of admixture that may account for this observation and reject recent Asian or European admixture as the cause. Infectious diseases have played a crucial part in shaping current and past human demography and genetics. Among all infectious diseases affecting humans, malaria has long been recognized as one of the strongest selective pressures in recent human history [1,2]. The Duffy antigen, also known as DARC (Duffy antigen receptor for chemokines) and more recently as ACKR1 (atypical chemokine receptor 1), is a transmembrane receptor used by Plasmodium vivax, a malaria-causing protozoan, to infect red blood cells. P. vivax causes a chronic form of malaria and is the most widespread type of malaria outside of Africa [3,4]. The DARC gene has three major allelic types that are the product of two common polymorphisms, forming the basis of the Duffy blood group system [5,6]. The two variant forms, FY*B and FY*A, are the allelic types commonly observed in non-African populations. FY*B is the ancestral form of the receptor, and is widespread in Europe and parts of Asia. FY*A is defined by a derived non-synonymous mutation (D42G, rs12075) in the P. vivax binding region of the DARC protein. It is the most prevalent of the three alleles in modern human populations, with highest frequency in Asia (predicted frequency >80%) and at 30–50% frequency in Europe [4]. FY*A is also present in southern Africa, despite absence from western and central Africa [4,7–9]. FY*O (also known as Duffy null) is defined by a mutation (T-42C, rs2814778) in the GATA-1 transcription factor binding site in the DARC gene promoter region, and occurs mostly on a FY*B background. The derived FY*O mutation exhibits extreme geographic differentiation, being near fixation in equatorial Africa, but nearly absent from Asia and Europe [4]. Of the three allelic types, FY*A and FY*B are functional proteins, while FY*O does not express the protein on erythrocyte surfaces due to a mutation in the promoter region, which causes erythroid-specific suppression of gene expression [6,10]. The lack of expression of DARC in erythrocytes has been shown to halt P. vivax infection [6,10]. Moreover, recent evidence shows that heterozygous individuals have reduced DARC gene expression and evidence of partial protection against P. vivax [11,12]. It has been proposed that due to the near-fixation of FY*O, P. vivax infection in humans is largely absent from equatorial Africa. An important recent discovery suggests low levels of P. vivax infection in FY*O homozygotes [13–17], which indicates that P. vivax might be evolving escape variants able to overcome the protective effect of FY*O. Phenotypic effects of the FY*A mutation are less clear than FY*O; however, there is evidence of natural selection and reduced P. vivax infection in individuals with this genotype ([18,19], with conflicting reports in the Brazilian Amazon however [12,20,21]). There is long running interest in characterizing the evolutionary forces that have shaped the Duffy locus. The combination of strong geographic differentiation and a plausible phenotypic association (resistance to malaria) has led to the Duffy antigen being cited as a canonical example of positive selection in the human genome (eg. [22–26]); however, details of its genetic structure remain understudied. Though touted as under positive selection, the few early population genetic studies of this locus found complex signatures of natural selection [27,28] and it is rarely identified in whole genome selection scans [29–37]. Some genomic loci display signatures of selection readily captured by standard methods, yet other well-known loci, like FY*O, are overlooked potentially due to intricacies not captured by simple models of hard selective sweeps. Detailed analyses of the haplotype structure can help us better understand complicated scenarios shaping genetic variation in loci under selection. What makes the evolution of FY*O such a complex and uncommon scenario? Plasmodium species and mammals have coexisted for millions of years, with frequent cases of host-shifts and host range expansions along their evolution [38,39]. Great apes are commonly infected with malaria-related parasites [40,41] and recent evidence suggests that human P. vivax originated in African great apes [40], contrasting with previous results that supported an Asian origin for P. vivax [42,43]. In addition to the complex evolutionary relationship among Plasmodium species and mammals, the specific mechanisms of invasion of erythrocytes employed by different species are highly diverse and present commonalities among species. Plasmodium falciparum, the parasite with the highest prevalence currently in Sub-Saharan Africa presents a highly redundant set of targets that enable erythrocyte invasion, but does not include DARC [44]. On the other hand, DARC erythroid expression influences infection in a variety of other species of Plasmodium. For example, it is required for infection by Plasmodium knowlesi, a malaria parasite that infects macaques, and SNPs upstream of the DARC gene homologue in baboons influence DARC expression and correlate with infection rates of a malaria-like parasite [45,46]. Despite the general understanding of the relevance of DARC in the evolution of the interaction between Plasmodium and primates, a thorough analysis of the complex evolutionary history of this locus using recently available large-scale genomic datasets of diverse human populations is still lacking. Here, we analyze the fine scale population structure of DARC using next-generation sequencing data from twenty-one human populations (eleven African populations), as well as ancient human genomes. We estimate the time to most recent common ancestor of the FY*A and FY*O mutations and estimate the strength of selection on FY*O. We propose a model for the spread of FY*O through Africa that builds on previous findings and provides a more complete picture for the evolution of FY*O. We further explore the relationship between the common FY*A haplotype in Asia and the FY*A haplotype found in southern Africa. We were interested in estimating the age of the FY*A and FY*O alleles and the start time of selection for FY*O, based on the average number of pairwise differences between haplotypes. For FY*O, we initially estimate the time to most recent common ancestor (TMRCA) over all FY*O haplotypes to be 230,779 years (95% CI: 169,790–291,039 years), which would imply very old selection under the assumption of a de novo sweep model. We note however that both the presence of two deeply diverged haplotypes with low levels of within-group diversity, as well as observed recombination between FY*O and the FY*A/FY*B alleles may artificially increase the estimated TMRCA. We therefore devised a strategy to remove the effects of recombination and estimate the TMRCA separately on each of the two major haplotypes, in order to obtain an approximate estimate for the start time of selection under the standing variation model. We estimate the major FY*O haplotype class to be 42,183 years old (95% CI: 34,100–49,030) and the minor haplotype class to be 56,052 years old (95% CI: 38,927–75,073) (S8–S10 Tables). For the FY*A allele, the allele age was estimated as 57,184 years old (95% CI: 47,785–64,732). Variation between population-specific TMRCA estimates was low. Additionally, we find that Ust’-Ishim, a 45,000 years old individual from Siberia [51] is heterozygous for FY*A. Under the assumption of no recurrent mutations, this would set a minimum age of 45,000 years for the FY*A mutation. These results corroborate our hypothesis that FY*O was an ancient sweep, likely tens of thousands of years older than most other mutations associated with malaria resistance [57]. These TMRCA estimates were used as a guide to seed our simulations for the following analysis; however, selection scenarios were not limited to these times. FY*O’s two divergent haplotypes indicate it may have reached fixation in Africa via selection on standing variation. To investigate this, we utilized an Approximate Bayesian Computation (ABC) approach to estimate the magnitude of FY*O’s allele frequency at selection onset, followed by the selection coefficient (s) of FY*O. To infer the magnitude of FY*O’s allele frequency at selection onset, we compared the posterior probability of five models of initial frequency at selection onset (de novo mutation (1/2N), 0. 1%, 1%, 10%, 25%), utilizing a Bayesian model selection approach in ABC, based on Peter et al. (2012) [58–61]. It is important to remark that we use additional summary statistics in our ABC implementation, including commonly used scans of selection. We realized that the original method proposed by Peter et al. (2012) [58] did not include any of these statistics but, as we show in this work, those prove to be highly informative of the process. Briefly, for each model we ran 100,000 simulations based on the African demographic model inferred in [62] and centered on an allele with selection coefficient drawn from the distribution 10U (−3, −0. 5). We assumed an additive selective model, as empirical studies predict heterozygotes have intermediate protection against P. vivax infection [11,12] and a selection start time similar to the estimated TMRCA of FY*O’s major haplotype (40 kya). We investigate our power to distinguish between the different models utilizing cross validation. We show that we have high power to distinguish between de novo and higher initial frequencies, though there is some overlap between adjacent models (S1 Appendix). Utilizing a multinomial logistic regression method, we observed strong support for the 0. 1% initial frequency model and low support all other models (posterior probabilities: de novo 0. 0002; 0. 1% 0. 9167; 1% 0. 0827; 10% 0. 0000; 25% 0. 0004) (S1 Appendix). We found these results to be robust to a range of recombination rates, selection start times and demographic models (Table 4 in S1 Appendix). We conclude selection on FY*O occurred on standing variation with a very low (0. 1%) allele frequency at selection onset. We next sought to infer the strength of the selective pressure for FY*O. We estimated FY*O’s selection coefficient via ABC and local linear regression, assuming an allele frequency at selection onset of 0. 1%. We find we have reasonable power to accurately infer s from these simulations; estimated and true selection coefficients have an r2 value of 0. 85 with a slight bias of regression to the mean (Fig 3 in S1 Appendix). We estimate the selection coefficient to be 0. 043 (95% CI: 0. 011–0. 18) (Fig 3). To validate our model choice, we sampled selection coefficients from this posterior distribution and ran simulations with the initial frequency drawn from either 10U (−5, −0. 5) or U (0,1). With the log-based prior distribution, we re-estimate the initial frequency at 0. 15% (95% CI: 0. 018–0. 77%; Fig 4 in S1 Appendix), closely fitting our inference. With the uniform prior distribution, we have much lower power to estimate initial allele frequency and we re-estimate the initial frequency at 6. 86% (95% CI: -20. 3–51. 6%) (Fig 5 in S1 Appendix). This is not surprising as it has previously been shown that it is very difficult to estimate initial frequency with a uniform prior [58]. We also sought to understand the history of these alleles in southern Africa as, unlike equatorial Africa, malaria is not currently endemic in southwestern Africa and past climate was potentially unsuitable for malaria. Thus, we expect there was a lower or no selection pressure for FY*O or FY*A in this region. We analyzed sequences from the Bantu-speaking Zulu and indigenous ≠Khomani San. We find all three allelic classes are present in both populations (Zulu: FY*A 6%, FY*B 16%, FY*O 79%; ≠Khomani San: FY*A: 35%, FY*B 44%, FY*O 21%). The KhoeSan peoples are a highly diverse set of southern African populations that diverged from all other populations approximately 100 kya [63], and the ≠Khomani San represent one of the populations in this group. The Zulu population is a Bantu-speaking group from South Africa; southern Bantu-speakers derive 4–30% KhoeSan ancestry [64] from recent gene flow during the past 1,000 years. We first ask if the FY*O allele in the KhoeSan group represents recent gene flow from Bantu-speakers or whether FY*O has been segregating in southern Africa for thousands of years. We investigated global and local ancestry differences between FY*O carriers and non-carriers. We find a significant difference in genome-wide western African ancestry in ≠Khomani San FY*O carriers vs. non-carriers (17% average in FY*O carriers vs. 5. 4% average in non-FY*O carriers, p = 0. 014 based on a Wilcoxon Rank-Sum test). We also find a significant enrichment of local Bantu-derived ancestry around the FY*O mutation in the ≠Khomani San FY*O carriers (p = 2. 78*10−12 based on Fisher’s exact test; S4 and S5 Figs). Each of these factors indicate that FY*O was recently derived from gene flow into the ≠Khomani San population from either Bantu-speaking or eastern African groups. We then explored the relationship of FY*O in KhoeSan and Zulu samples to Bantu-speaking populations from equatorial Africa. A haplotype network of the ≠Khomani San FY*O carriers indicated that each 20kb haplotype was identical to a haplotype from populations further north (S6 Fig). We tested the Zulu FY*O samples as well, and found that they have identical, though more diverse, haplotypes than other Bantu-speaking populations (Fig 2). However, the increase in diversity may be due to calling biases and recombination between different allelic classes in the Zulus (see discussion). We then sought to understand the prehistory of FY*A in southern Africa. The FY*A allele is common in San populations, despite its absence from equatorial Africa (Fig 1). We compared the FY*A haplotypes found in the ≠Khomani San and Zulu populations with FY*A haplotypes present in Asia and Europe to distinguish between three hypotheses. The FY*A mutation in southern Africa either was 1) segregating in the ancestral human population, 2) due to recent admixture from migrations ‘back to Africa’, or 3) arose convergently in a separate mutation event distinct from the European / Asian mutation. We find that Zulu FY*A haplotypes are highly diverse; some are identical to non-African FY*A haplotypes, while others are unique or ancestral (Fig 2). Global ancestry results show no statistically significant difference between Bantu or KhoeSan ancestry in FY*A ≠Khomani San carriers and non-carriers based on a Wilcoxon Rank-Sum test (San: p = 0. 85, Bantu: p = 0. 101). Our local ancestry results indicate that FY*A carriers are significantly enriched for San ancestry around FY*A compared with non-carriers based on Fisher’s exact test (p = 0. 011). Our results support hypothesis (1), i. e. high ≠Khomani San FY*A haplotype diversity indicates FY*A has an ancient presence in southern Africa. Furthermore, as Bantu-speaking populations from equatorial Africa currently are exclusively FY*O, it is unlikely they transferred FY*A to KhoeSan after the Bantu expansion. Rather, the FY*A haplotypes in the Zulu are largely derived from admixture with the indigenous KhoeSan populations, or potentially very recent gene flow from European/Asian immigrants to South Africa. We estimate the TMRCA of all FY*O haplotypes to be 230,779 years (95% CI: 169,790–291,039 years) and the TMRCA of the most common haplotype class to be 42,183 years (95% CI: 34,100–49,030 years). We note that two of the assumptions of our estimation method (no recombination and star-like phylogeny) are partially violated in our data. However, we developed a strategy to mitigate the effect of recombination (see Methods) and only the TMRCA estimation with all FY*O haplotypes differs greatly from a star-like phylogeny (due to the deep divergence in the two main haplotypes). Estimates of TMRCA are also prone to large confidence intervals due to the stochasticity of the allele frequency trajectory, but all estimates indicate the FY*O mutation is older than most known malaria resistance alleles [57]. Previous estimates of the time of fixation of the FY*O mutation, based on lower density data, range from 9–63 kya (adjusted to our mutation rate and generation times) [28,73]. Other TMRCA estimates ranging from 9 to 14 kya were calculated on microsatellites linked to FY*O [73], which seem to have underestimated the age of the mutation. Perhaps the most comprehensive work on this problem until now was by Hamblin and DiRienzo [28], who estimated the time to fixation of FY*O to be 63 kya (95% CI: 13,745–205,541 years; converted to our mutation rate). This is older than our estimates, but has overlapping confidence intervals. More recently, Hodgson et al. [74] estimated the time necessary for FY*O’s frequency to increase from 0. 01–0. 99 to be 41,150 years, based on an inferred selection coefficient in Madagascar. We inferred FY*A to be an ancient mutation, likely segregating throughout Africa before FY*O swept to fixation. We estimate FY*A to be 57,187 years old (95% CI: 47,785–64,732 years), 15,000 years older than the most common FY*O haplotype and overlapping estimates of the out-of-Africa expansion time [62,75–77]. We note that the San FY*A haplotypes were not used in this TMRCA calculation as there were few homozygous sequenced FY*A San samples and we confined our estimates to homozygotes to reduce issues due to phasing errors. As we are only looking at the out-of-Africa diversity of FY*A, it is likely this TMRCA is more indicative of FY*A’s expansion during and after the out-of-Africa event. Ancient DNA from a Paleolithic hunter-gatherer provides evidence that FY*A was already present in Eurasia by at least 45,000 years ago, thereby setting a lower bound for the age of the mutation. Its intermediate frequency in ≠Khomani San and Zulu populations, and similar haplotypic structure is consistent with FY*A existence in Africa at an appreciable frequency before the out-of-Africa expansion had occurred. The deep divergence of the ≠Khomani San from all other tested populations carrying FY*A strongly supports this ancient origin. Our results are scaled with the mutation rate of 1. 2 * 10−8 mutations / basepair / generation and a 25 year generation time. This mutation rate is supported by many previous whole-genome studies ([51,78–81]; range: 1 − 1. 2 * 10−8 mutations / basepair / generation), but we are aware of recent studies suggesting a higher mutation rate that are either based on exome data ([82–84]; range: 1. 3 − 2. 2 * 10−8 mutations / basepair / generation) or whole-genome data ([85,86]; range: 1. 61 − 1. 66 * 10−8 mutations / basepair / generation). To take into account this uncertainty, we performed additional analyses using a mutation rate of 1. 6 * 10−8 mutations / basepair / generation. With this higher rate, we estimate more recent coalescent times of the FY*O and FY*A mutations; specifically we would estimate the FY*O TMRCA to be 32 kya (vs. 42 kya) and the FY*A TMRCA to be about 43 kya (vs. 57 kya). It is important to consider that most quantities in population genetics are scaled by the mutation rate and effective population size. Therefore, any changes in the mutation rate result in changes not only in our TMRCA estimates, but also in the timescale of the split between African and non-African populations. For example, a recent study of the divergence between African and non-Africans, estimates a median of divergence between 52–69 kya and a final split around 43 kya, using a mutation rate of 1. 2 * 10−8 mutations / basepair / generation [76]. If we use a higher mutation rate of 1. 6 * 10−8 mutations / basepair / generation the median divergence would be 39–52 kya with a final split around 33 kya. Thus, regardless of the mutation rate (and the corresponding demographic scenario), we estimate the FY*O mutation to have occurred soon after the estimated final split. FY*O’s two divergent, common haplotypes in Africa indicate it may have reached fixation due to selection on standing variation. We infer that the FY*O mutation underwent a selective sweep on standing variation with a selection coefficient comparable to some of the most strongly selected loci in the human genome [57]. Utilizing a Bayesian model selection approach implemented in an ABC framework, we find that FY*O likely rose to fixation via selection on standing variation; though the frequency of FY*O at selection onset was very low (0. 1%). We estimate FY*O’s selection coefficient to be 0. 043 (95% CI: 0. 011–0. 18), consistent with previous estimates (>0. 002 in the Hausa [28], 0. 066 in Madagascar [74]). The similarity of these results indicates FY*O may have a similar selective effect in diverse environments. This selection coefficient is similar to other loci inferred to have undergone strong selection in the human genome, including other malaria resistance alleles [57]. The selection coefficients of these other malaria resistance alleles were inferred via a variety of different methods, mostly utilizing simulations and a rejection framework. Our understanding of human demographic history has improved over the past few years with the increase of genomic data. Previous estimates did not consider realistic demographic models, while we utilized the African demographic model inferred in Gravel et al. [62]. Assuming the standard neutral model when the true demography is more complex may result in overestimating the selection coefficient for some of the regions mentioned in Hedrick et al. 2011 [57] (due to recent population expansions). At first glance it would be reasonable to consider such a low initial frequency equivalent to a scenario of selection on a de novo mutation. In order to distinguish between the two possibilities we use the diffusion approximation by Kimura [87,88] to estimate the probability of fixation (equation 8 in [88]) and demonstrate that it is much more likely to reach fixation with an initial frequency of 0. 1% than a scenario of a new mutation arising in the population. We find that an allele with selection coefficient 0. 043 and initial frequency 0. 001 has a 99. 4% probability of fixing, while a de novo mutation with the same s has only an 8. 2% probability of fixing. It is important to note that in our calculation the initial frequency (p) in the equation for the de novo mutation scenario is calculated using the effective population size, as opposed to the census population size. However, if we reasonably assume N ≥ Ne, p is likely at least 0. 1% in the population. This translates in our estimates for the probability of fixation of a de novo mutation being far more optimistic than expected if the ancestral African census population size was much larger than the effective size. This low initial frequency until 40 kya is consistent with FY*O’s absence from non-African present and ancient genomes. It is important to note that selection on standing variation and a soft sweep are not necessarily synonymous. Selection on standing variation (the model we are testing) asks about the frequency of the allele at selection onset. However, it is agnostic to the number of haplotypes that are actually picked up at selection onset. A soft sweep states that multiple haplotypes are picked up at selection onset. Via our msms simulations, we are unable to say, for each individual simulation, whether just one haplotype was picked up or if multiple haplotypes were picked up (either due to additional mutations or recombination). However, we note that of our accepted simulations, summary statistics of the 0. 1% model are much more diverse and much more similar to our data, than the de novo model. We speculate that this may be due to multiple haplotypes that are being picked up in the 0. 1% model. FY*O and FY*A are thought to be under positive selection due to P. vivax, a malaria-causing protozoan that infects red blood cells through the Duffy receptor. Individuals with the FY*O allele do not express the Duffy receptor in red blood cells resulting in immunity to P. vivax [6,10] and individuals with the FY*A allele may have lower infectivity rates [11,12,18–21]. Unlike P. falciparum, the most common and deadly malaria protozoan in Africa that uses multiple entry receptors, P. vivax’s one mode of entry allows the possibility of resistance with only one SNP. Was P. vivax the selective pressure for either the FY*O or FY*A mutations? P. vivax is currently prevalent in equatorial regions outside of Africa; however it is unknown if P. vivax has ever been endemic to Africa. There is an ongoing debate as to if P. vivax originated in Asia or Africa. Previously, it was thought P. vivax originated in Asia, as Asian and Melanesian P. vivax has the highest genetic diversity [42,89] and the most closely related parasite to P. vivax is P. cynomolgi, a macaque parasite [40,42]. However, recent evidence shows global human-specific P. vivax forms a monophyletic cluster from P. vivax-like parasites infecting African great apes, suggesting an African origin [40]. Human-specific P. vivax sequences form a star-like phylogeny likely due to a relatively recent demographic expansion. Our TMRCA estimates of human-specific P. vivax sequences are 70–250 kya (S12 Table), consistent with previous estimates (50–500 kya, [42,43,89]). As the TMRCA of human-specific P. vivax is estimated to be before or overlapping the TMRCA of FY*O, this is consistent with the hypothesis of P. vivax being the selective agent responsible for the rise of FY*O in Africa. However, there are two possible scenarios that could explain the TMRCA estimates for P. vivax. A first scenario is that the estimated TMRCA of human P. vivax indicates the start of the association between host and parasite, thus marking the start of selective pressure on the host. A second scenario is that these estimates overlap the human out-of-Africa expansion times. It is possible that human-specific P. vivax expanded out of Africa with humans, resulting in the estimated TMRCA for P. vivax. The human P. vivax currently in Africa could be from recent migration [43,89]. Additionally, it is yet unclear if such a high selection coefficient is consistent with the fact that the general severity of P. vivax is currently much lower than that observed for P. falciparum, causing more morbidity than mortality. The combination of these observations lead us to suggest that further work is necessary to better understand the evolutionary history of P. vivax to reconcile the demographic scenarios that could have given rise to such a complex pattern. All together, our results suggest that the evolutionary history of the FY*O mutation, a single SNP under strong selection in human populations, has been a complex one. Multiple haplotypes present in highly divergent African populations are consistent with selection on standing variation, shaping the evolution of this locus that was present in very low frequency in ancestral populations. Although more work needs to be done to understand how P. vivax may have shaped the evolution of this locus, our results provide a framework to evaluate the evolution of the parasite and formulate specific hypotheses for its evolutionary history in association with its human host.

Infectious diseases have undoubtedly played an important role in ancient and modern human history. Yet, there are relatively few regions of the genome involved in resistance to pathogens that show a strong selection signal in current genome-wide searches for this kind of signal. We revisit the evolutionary history of a gene associated with resistance to the most common malaria-causing parasite, Plasmodium vivax, and show that it is one of regions of the human genome that has been under strongest selective pressure in our evolutionary history (selection coefficient: 4. 3%). Our results are consistent with a complex evolutionary history of the locus involving selection on a mutation that was at a very low frequency in the ancestral African population (standing variation) and subsequent differentiation between European, Asian and African populations.

Abstract Introduction Results Discussion

parasite groups medicine and health sciences plasmodium parasite evolution geographical locations tropical diseases human genomics parasitic diseases parasitology genetic mapping apicomplexa paleontology paleogenetics africa people and places haplotypes heredity earth sciences genetics biology and life sciences malaria genomics genomics statistics computational biology

2017

Population genetic analysis of the DARC locus (Duffy) reveals adaptation from standing variation associated with malaria resistance in humans

Australia is the only high income country with persisting endemic trachoma. A national control program involving mass drug administration with oral azithromycin, in place since 2006, has some characteristics which differ from programs in low income settings, particularly in regard to the use of a wider range of treatment strategies, and more regular assessments of community prevalence. We aimed to examine the association between treatment strategies and trachoma prevalence. Through the national surveillance program, annual data from 2007–2013 were collected on trachoma prevalence and treatment with oral azithromycin in children aged 5–9 years from three Australian regions with endemic trachoma. Communities were classified for each year according to one of four trachoma treatment strategies implemented (no treatment, active cases only, household and community-wide). We estimated the change in trachoma prevalence between sequential pairs of years and across multiple years according to treatment strategy using random-effects meta-analyses. Over the study period, 182 unique remote Aboriginal communities had 881 annual records of both trachoma prevalence and treatment. From the analysis of pairs of years, the greatest annual fall in trachoma prevalence was in communities implementing community-wide strategies, with yearly absolute reductions ranging from -8% (95%CI -17% to 1%) to -31% (-26% to -37%); these communities also had the highest baseline trachoma prevalence (15. 4%-43. 9%). Restricting analyses to communities with moderate trachoma prevalence (5–19%) at initial measurement, and comparing community trachoma prevalence from the first to the last year of available data for the community, both community-wide and more targeted treatment strategies were associated with similar absolute reductions (-11% [-8% to -13%] and -7% [-5% to -10%] respectively). Results were similar stratified by region. Consistent with previous research, community-wide administration of azithromycin reduces trachoma prevalence. Our observation that less intensive treatment with a ‘household’ strategy in moderate prevalence communities (5-<20%) is associated with similar reductions in prevalence over time, will require confirmation in other settings if it is to be used as a basis for changes in control strategies. Trachoma, caused by serotypes of Chlamydia trachomatis is a major cause of blindness globally. [1] In 1997 The Alliance for the Global Elimination of Blinding Trachoma by 2020 (GET 2020) initiative was launched. Supported by the World Health Organization (WHO), the alliance promotes its goal of elimination through the SAFE strategy, with its key components of surgery to correct trichiasis (S), antibiotic treatment (A), facial cleanliness (F) and environmental improvements (E). [1] Randomised controlled trials have shown that antibiotics, either topical or oral, are effective for treatment. [2] There is a more limited body of trial evidence that has been used to support the strategy of mass drug administration (MDA), or whole community treatment, which is one of the main components of the SAFE strategy in many countries. There have been few comparisons of alternative community treatment strategies, and relatively limited follow up studies of long term trends following implementation of prevention programs. [3,4] Evidence of effective treatment strategies across a range of prevalence settings will become increasingly important as more countries approach the goal of trachoma elimination. Australia is the only high-income country with endemic trachoma. [5] The disease occurs primarily in remote Aboriginal communities in three Australian jurisdictions, the Northern Territory (NT), South Australia (SA), and Western Australia (WA), although it has also been identified in Queensland and New South Wales. [6,7] In 2013 overall prevalence among children aged 5–9 years in endemic areas was estimated to be 4% with substantial variation between communities; an estimated 50% of communities had no clinically detectable trachoma and 8% had hyperendemic levels (>20%). [8] Since 2006 the Australian Government has funded control programs based on regular mapping of trachoma prevalence in endemic areas. Trachoma management has been based on guidelines first endorsed in 2006 [9] and revised in 2014. [10] Unlike the WHO guidelines, the 2006 Australian guidelines recommended screening every community considered at risk annually, regardless of trachoma prevalence, as well as a tiered approach to antibiotic treatment depending on trachoma prevalence (see Table 1). Australia therefore has an opportunity to examine the impact of different treatment strategies, in more detail than has been possible in other trachoma endemic settings, where only MDA has been used, and prevalence is generally monitored at much longer intervals. We report here an analysis based on routinely collected trachoma prevalence data over seven years in Australia’s endemic areas. These data have the potential to inform trachoma control programs both in Australia and internationally. De-identified community-based data were obtained for each year from 2007, when comprehensive data collection began, through to 2013. As the majority of trachoma screening in communities was undertaken through primary school programs targeting 5–9 year olds, we restricted analyses of prevalence to this age group. The unit for analysis was a single episode of screening in 5–9 year olds within a single community. Community trachoma prevalence was estimated by dividing the number of 5–9 year olds with active trachoma during a screening round by the number screened. The treatment strategy adopted for a community in a given year was classified into one of four categories according to what was reported in the national database: no treatment; “active”cases only treated; “household” treatment under which active cases and their households members were treated; and “community-wide” treatment which covered both whole-of-community treatment (also known as “mass drug administration”) and a strategy under which active cases, household members and all children aged <15 years in the community were treated. For all strategies, the treatment administered for those over 6 months of age was a single weight-based dose (20mg/kg) of oral azithromycin [9]. Descriptive analyses by calendar year, examining all communities with eligible screening episodes were conducted. From 2011 onwards treatment coverage in communities for which household or community-wide treatment strategies were recorded was calculated by summing the population aged 0–14 years recorded as treated with azithromycin, and dividing by the total estimated population aged 0–14 years according to both census[13] and local health worker community population estimates. To compare treatment strategies, we undertook two analyses. First we identified all communities for which data on trachoma prevalence in 5–9 year olds were available for pairs of consecutive calendar years. For each such pair of years, we estimated the change between the years in community prevalence, by simple difference. We then grouped communities by the first year of the consecutive pair and by the treatment strategy recorded in that year, and calculated a combined estimate of change for each treatment strategy using a random effects meta-analysis. [14] Second, for each community with at least two years of screening data, regardless of whether they were consecutive, we compared the change in prevalence from the first and final year of available data according to broad categories of treatment strategy (never treated, any non-community-wide treatment, treated at least once with community-wide), using the same meta-analytic method. As the treatment strategy used was strongly influenced by trachoma prevalence, [9] we conducted sensitivity analyses restricting communities to those with moderate prevalence (≥5% to <20%) at the start of the interval. We also stratified results by the two jurisdictions contributing the majority of data (NT and WA), and by community size (<250 versus ≥250 people) based on 2011 Australian census estimates. [13] Finally we conducted post-hoc analyses only including data collected for years 2007 to 2010 with the goal of differentiating secular trends in trachoma prevalence from effects of treatment. We used RevMan 5. 5 software to estimate absolute differences in trachoma prevalence and SAS (version 9. 3) for estimates of trachoma prevalence (function was unavailable in RevMan). The DerSimonian and Laird random effects model was used to obtain pooled estimates of risk difference, using the Mantel-Haenszel method to estimate the variation between studies. We estimated the combined prevalence using an exact likelihood approach. [15] Administrative approvals for the data collection and analyses reported here were provided by the health departments of the three jurisdictions involved. Ethical approval was by the University of New South Wales Human Research Ethics Committee (ref 9-14-042). We identified 914 screening episodes from 215 unique remote Aboriginal communities with children aged 5–9 years screened at least once between 2007 and 2013. Of the 215 communities, the majority were in the NT (n = 90; 42%) and WA (n = 99; 46%). There were 33 communities screened only once, 46 had 2–3 episodes, 59 had 4–5 episodes, and 77 had 6–7. The communities screened less frequently were more likely to have been screened for the first time more recently, with the median year of screening for communities with 3 or fewer years of screening data being 2012 compared to 2010 for those with four or more years of data. Table 2 shows the number of communities screened each year, the proportion of communities screened from each of the three jurisdictions, the median number of children screened, the trachoma prevalence in 5–9 year olds and treatment strategies used. Biannual treatment (a second dose of antibiotics administered in the same year) was recorded following 1% of screening episodes. As biannual treatment was not unique to a particular treatment strategy, and numbers were small, we did not include this as a separate treatment classification. In general, the number of communities screened increased until 2013 when the NT adopted the revised guideline for screening[10] which recommends that screening in communities with high trachoma prevalence takes place every 3 years rather than annually. The median number of 5–9 year old children screened per community remained relatively stable over the seven years numbering about 20 (IQR 10–38). From 2008 to 2013, the proportion of communities with no trachoma detected increased (from 22. 8% to 60. 3%) while the proportion of communities with trachoma prevalence above 5% decreased (from 67. 5% to 27. 6%). In communities with trachoma detected, the median prevalence also decreased, from 23. 1% to 8. 9%. There was an increase in the proportion of communities not treated from 25. 7% in 2008 to 56. 0% in 2013, while from 2009 there was a fall in the number of communities treating ‘active’ cases only. For 2011 and 2012 (Table 3), using local estimates of the population size, treatment coverage among 120 communities that reported having used a “household” treatment strategy was 11. 9% (95%CI 11. 4–12. 5%) while for the 33 communities using a “community-wide” treatment strategy, treatment coverage was 75. 0% (95%CI 73. 5–76. 4%). The estimates were similar when Census population estimates[13] were used. After excluding the 33 communities with only a single year of screening data available, there remained 881 records from 182 unique communities; 77 (42. 3%) were from the NT, 21 (11. 5%) from SA and 84 (46. 2%) from WA. Of 121 communities that applied a treatment strategy in more than one of the years observed, 89 were recorded as having changed strategies over the time period, 30 communities used only household treatments, two used only community-wide treatments, and none applied the “active” case only strategy more than once. Fig 1 shows the estimated change in trachoma prevalence between pairs of successive years, according to the treatment strategy used in the initial year of the pair. Communities recorded as receiving no treatment are separated according to whether they had trachoma detected or not in the initial year of the pair. In the earlier years of the program (2007–2010) for communities without trachoma detected and not treated, there was a significant increase in prevalence between pairs of years (e. g. absolute risk increase of 10% [95%CI 3% to 16%] from 2007 to 2008) but after 2010 there was no substantial change. The number of communities that were not recorded as having been treated despite trachoma being detected decreased over time. In these communities, trachoma prevalence between years did not change significantly between pairs. For all categories of treated communities (active case only, household, or community-wide) there was a reduction in estimated trachoma prevalence between the pairs of years; in most years this was not statistically significant. The largest absolute reductions in trachoma prevalence were in communities that were recorded as having received community-wide treatment, with point estimates ranging from -8% to -31%; the reductions were only statistically significant for the years 2007–2011. These communities receiving community-wide treatment also had the highest prevalence in the earlier of the paired years (range for combined estimates 15. 4%-43. 9%). The majority of communities (n = 176) had annual records of trachoma screening that included at least two years of the eligible period (2007 to 2013), including 68 with data for 7 consecutive years, 47 with 6 years, 25 with 5 years, 12 with 4 and 24 with less than 4. Based on the treatment strategies used between the first and final year of data recorded, communities were grouped into three categories (Table 4): those never treated; those treated but never with community-wide strategies (i. e. only active case or household treatment); and those treated at least once using a community-wide strategy. Fig 2 shows the estimated change in trachoma prevalence between the first and final years of data, by the three classifications of communities in Table 4. For communities never recorded as receiving azithromycin for trachoma, the estimate of prevalence in the first year of screening was 0. 1% and the estimated absolute reduction over time 0% (-3% to 2%). For communities that received only active case and household strategies, the prevalence in the first year of screening was 5. 8% and the reduction over time -4% (-2% to -6%). Among communities treated at least once with community-wide strategies, initial prevalence was 23. 9%, and the reduction -21% (-16% to -26%). These patterns were similar when communities from the NT or WA were considered separately. When we restricted analyses to communities with moderate trachoma prevalence (5-<20%) in their first year of screening (Fig 2) we found that there was only a small difference between those that had received community-wide treatment and those that had not, with reductions of -11% (-8% to -13%) and -7% (-5% to -10%) respectively, from a similar initial prevalence (11. 5% and 10. 1% respectively). Restricting analyses to communities with at least four years of screening data did not change the findings, and we found no differences in treatment effects when we compared smaller (N<250 people) to larger sized communities (N≥250 people). Analyses of data restricted to 2007–2010, the period during which there were substantial increases in trachoma prevalence in previously trachoma-free communities that were untreated (see Fig 1), are shown in Table 5. In the communities that were never treated, and in those treated but not with a community wide strategy, there was no significant fall in trachoma prevalence, while communities with at least one community-wide treatment had a 14% (95%CI 9 to 20%) decline in trachoma prevalence. When we further restricted analyses to communities with moderate trachoma prevalence (5-<20%) we found that those with at least one community-wide treatment had a significant reduction in prevalence but those treated with more targeted strategies did not. In this investigation of the relationship between different community treatment strategies for trachoma control and long-term changes in trachoma prevalence, we found that in high prevalence communities, community-wide administration of azithromycin, or MDA, was associated with a substantially reduced trachoma prevalence after one year or more. In settings with moderate trachoma prevalence (5-<20%), more limited strategies were equally effective in the longer term. As discussed in more detail below, this finding may have particular relevance for countries moving towards elimination, but with localised areas of moderate prevalence remaining. Observational studies have shown that a single MDA of azithromycin in communities with endemic trachoma results in substantial reductions at one year in trachoma prevalence in both hyperendemic (>20% prevalence) [16,17] and moderately endemic (5-<20% prevalence) [3,18] communities. Recent trials have compared annual versus biannual mass azithromycin administration in high prevalence communities, but the trials have not consistently found that larger or more sustained reductions can be achieved with more frequent treatment. [19,20] There are few reports comparing different treatment strategies in moderate prevalence settings. One study found that targeted (household) treatment may be as effective as mass treatment of all children but only had follow-up for 6 months. [21] Another suggested a single mass drug administration may be effective in sustaining a reduction in trachoma prevalence over many years[3] and another suggested that treatment that was not community-wide led to re-infections. [22] Our findings regarding ‘community-wide’ treatment (Fig 1) over one year concur with the observational studies of mass drug administration showing that this approach is effective in substantially reducing trachoma prevalence in high prevalence settings. Our main analyses also suggest that in more moderate prevalence settings, targeted treatment strategies (mostly ‘household’ strategies, whereby active cases and all members of their household were treated with azithromycin), were also associated with reduced trachoma over a year and for longer periods (Figs 1 and 2). In the paired-year analyses (Fig 1), among communities that had no trachoma detected at the start of the observation period, and were consequently not treated, we found that there were annual increases in prevalence between 2007 and 2010. However from 2011 onwards, prevalence remained at zero, i. e. no change. Given the mobility of people between Aboriginal communities, [23] this observation may be evidence that antibiotic treatment programs in communities with trachoma detected can have a “herd” effect, in that transmission to trachoma-free communities is prevented. It is also possible that this resulted from other components that are delivered as part of the SAFE strategy, such as promotion of facial cleanliness and environmental improvements. As the paired-year analysis indicated no overall increase in trachoma prevalence in trachoma-free communities from 2010 onwards, we undertook analyses involving multiple years with the goal of distinguishing effects of treatment from temporal changes in trachoma. In these sensitivity analyses, only communities receiving community-wide treatment were found to have a reduction in trachoma prevalence (see Table 5). It is therefore possible that the trachoma reduction in moderate prevalence settings observed in our primary analyses may in fact have been a result of overall declines in trachoma burden rather than a result of targeted treatment. This was an observational study using routinely collected surveillance data. [5] While diagnosis was undertaken by specialised teams of health care workers following standard international guidelines, there may have been diagnostic error, to an extent that cannot be measured. Our analysis of impact was also limited by the absence of detailed data for all years on the level of treatment coverage achieved in each community. However for 2011 and 2012, data were available for the majority of communities and this indicated that coverage was substantially different between those communities reporting “household” compared to those reporting “community-wide” treatment. There may also be factors that differed between communities or changes over time that were not measured but were associated with treatment strategy and therefore could have affected the summary estimate of difference in trachoma prevalence observed. For example, we did not include in our analyses other factors that may contribute to changes in trachoma prevalence. [1] Facial cleanliness, and facial cleanliness promotion (‘F’ in the SAFE strategy) was reported in the communities screened, but not considered to be sufficiently standardised or validated to use in the analyses presented here. [24] Data on environmental factors (‘E’ in the SAFE strategy) such as improved housing conditions, or the availability of swimming pools, were limited and inconsistent. [25] Despite the absence of information on facial cleanliness and environmental factors, we do not have evidence to suggest that they linked to treatment status and thus had any potential to bias our results. The strengths of our study are the use of annual trachoma screening data from all communities in the three jurisdictions with known endemic trachoma leading to a more comprehensive picture of not only the effects of different treatment strategies on single communities, but also programmatic effects on all communities in a real-world setting. We also had observations for the majority of communities over a significant period of time (at least four years) and were able to observe the effects of a targeted treatment strategy in a moderate prevalence setting. In summary, our study supports current evidence that recommends mass azithromycin administration to reduce trachoma prevalence in high prevalence settings. We also found that less intensive treatment with a “household” strategy in moderate prevalence communities (5-<20%) may be associated with reductions in prevalence similar to mass drug administration. This finding may have implications for countries that are moving to lower levels of endemic trachoma and wish to reduce the amount of azithromycin being used. The strategy does however have the requirement that individual examination must take place, to determine which households have affected members. If a targeted approach is to be considered, trials and health economic analyses are required to determine which option may be more cost-effective in particular programmatic and community contexts. [26] Finally our results also suggest that trachoma program implementation can reduce trachoma prevalence in communities not specifically targeted (“herd effects”) and thereby contribute to reducing trachoma transmission.

Australia is the only high income country with persisting endemic trachoma and a national control program has been in place since 2006. The program involves annual screening of children for trachoma in communities designated to be at high risk of disease and treatment of those affected with the antibiotic azithromycin. Depending on the level of trachoma detected in children, antibiotic treatment is also given to households and other community members. We used data collected annually from 2007 to 2013 to examine what effect the extent of azithromycin treatment had on subsequent levels of trachoma in children aged 5–9 years. We found that in communities with high levels of trachoma, when all community members received azithromycin (community-wide treatment), the greatest reduction in trachoma level was achieved. However in communities with moderate levels of trachoma, using either community-wide treatment or more targeted (household) treatment resulted in equivalent reductions in trachoma. This observation needs to be confirmed in other studies before changes to current recommendations regarding trachoma control strategies are considered.

Abstract Introduction Methods Results Discussion

antimicrobials medicine and health sciences drugs tropical diseases geographical locations australia microbiology census health care bacterial diseases research design pharmaceutics antibiotics eye diseases screening guidelines drug administration neglected tropical diseases pharmacology research and analysis methods infectious diseases treatment guidelines trachoma people and places survey research oceania ophthalmology microbial control biology and life sciences drug therapy health care policy

2016

Relationship between Community Drug Administration Strategy and Changes in Trachoma Prevalence, 2007 to 2013

Dengue, a vector-borne viral disease of increasing global importance, is classically associated with tropical and sub-tropical regions around the world. Urbanisation, globalisation and climate trends, however, are facilitating the geographic spread of its mosquito vectors, thereby increasing the risk of the virus establishing itself in previously unaffected areas and causing large-scale epidemics. On 3 October 2012, two autochthonous dengue infections were reported within the Autonomous Region of Madeira, Portugal. During the following seven months, this first ‘European’ dengue outbreak caused more than 2000 local cases and 81 exported cases to mainland Europe. Here, using an ento-epidemiological mathematical framework, we estimate that the introduction of dengue to Madeira occurred around a month before the first official cases, during the period of maximum influx of airline travel, and that the naturally declining temperatures of autumn were the determining factor for the outbreak' s demise in early December 2012. Using key estimates, together with local climate data, we further propose that there is little support for dengue endemicity on this island, but a high potential for future epidemic outbreaks when seeded between May and August—a period when detection of imported cases is crucial for Madeira' s public health planning. The ongoing spread of dengue, the most important mosquito-borne flavivirus affecting humans, from predominantly tropical and sub-tropical regions into higher latitudes, such as the United States of America, Australia and Europe, is a major public health concern [1]. Globalisation and climate change are some of the possible factors that have facilitated the geographic expansion of its two vector-species, Aedes aegypti and Aedes albopictus [2], [3]. The size of the dengue-naive population together with frequent travels to endemic countries impose a significant risk of large epidemic outbreaks in these regions as well as the possibility of dengue becoming (re-) established as an endemic disease [4]. Understanding and quantifying the potential of dengue outbreaks in previously dengue-free environments is therefore paramount for public health planning. Aedes aegypti, dengue' s main vector, has been considered extinct from continental Europe since the mid-twentieth century but was recently introduced to the The Portuguese Autonomous Region of Madeira [5]. This Atlantic archipelago consists of several islands, two of which are inhabited. From these, the island of Madeira is the largest with a population size of. It has an approximate area of 750 square kilometres and is located around 1000 kilometres from the European continent, sharing roughly the same latitude as central Morocco. The interior of Madeira is particularly mountainous, which has resulted in its population being distributed mainly along the coast, specially in the south, where the capital city of Funchal, harbouring nearly half of the island' s inhabitants, is located. The mixture of densely populated areas with rich and abundant sub-tropical vegetation will have promoted the mosquito' s introduction into Funchal, from where it spread longitudinally along the coast and later to the rest of the island [5]. In contrast to many dengue-endemic cities in tropical regions, mosquito breeding in Funchal can not be linked to poor sanitation, waste disposal or water storage practices [6], [7]. Instead, the well established habit of potting small plants and flowers provides a vast number of potential breeding sites, both indoors and surrounding domestic premises [5]. Although Madeira' s climate is classified as Mediterranean, its heterogenous landscape imposes significant differences in sun exposure, humidity and mean daily temperatures. These local variations, together with influences from the Gulf Stream and the Canary Current, develop into a range of contrasting local microclimates. The island presents monthly average temperatures above 20° Celsius during spring, summer and autumn, peaking around 26° Celsius in August (Figure 1A). Even during the winter months, temperatures often remain above 15° Celsius. The mild climate together with the blend of seaside, mountainous and urban landscapes, and short flight distances to continental Europe, make the island of Madeira an attractive tourist destination. In the past two decades, successive governments have successfully invested in the expansion of the tourism industry, transforming it into the main driving force of the small, local economy. Consequently, the Archipelago has witnessed a major increase in the number of international airline travellers (Figure 1B), mainly from Europe but also from South America (Figures 1C and D). On 3 October 2012, two dengue infections were reported by the Direcção Geral de Saúde (Portuguese health ministry) on the island of Madeira [8], [9]. The patients had no recent overseas travel history, raising an alert for possible autochthonous transmission. In the following weeks the island witnessed its first dengue epidemic with a total of 2187 reported cases, of which approximately were confirmed [9]. The outbreak was characterized by a sharp increase in weekly reported cases throughout October, peaking in November and decreasing rapidly thereafter (Figure 2). It was declared extinct on March 2013, after which one case was imported from Brazil and two others from Angola (until the end of summer 2013) [9], [10]. During the epidemic period, 81 cases were exported to continental Europe, with 11 reported cases in Portugal and 70 in other European countries [9]. Analysis of blood samples from Madeira' s patients identified the circulating virus as belonging to dengue serotype 1 (DENV1) with strong sequence similarity to genotypes circulating in Venezuela, Brazil and Columbia at the time [9], [11], [12]. The reporting of short transmission chains of dengue autochthonous cases in European countries is a recent and increasingly common phenomenon [13]–[15]. This first ever dengue outbreak was therefore a sudden event with wide-ranging public health and economic implications, both locally and at the European level. To date, however, neither the conditions that have facilitated this short epidemic and its extinction nor the associated potential for future outbreaks have been studied in detail. Here, we develop an ento-epidemiological mathematical framework to explore the ecological conditions and human-mosquito transmission dynamics underlying this outbreak. Our results indicate that the declining temperatures of autumn were the determining factor for the outbreak' s sudden decline. We further estimate that the probable time of introduction was around the end of August, weeks before the first clinical cases were officially reported. Importantly, while this matched with the period when airline traffic (to and from the island) was at its yearly maximum, introductions at an earlier timepoint could have resulted in significantly bigger and longer-lasting epidemics, with obvious consequences for local public health and disease spread to other European countries. We devised an ordinary differential equation (ODE) model to capture the transmission dynamics of dengue between human and mosquito hosts. The human population is assumed to have constant size () and to be fully susceptible to the virus. Upon challenge with infectious mosquito bites (), individuals enter the incubation phase () with mean duration of days, later becoming infectious () for days and finally recovering () with life-long immunity. The dynamics of the human population are defined by the following set of ODEs: (1) (2) (3) (4) (5) For the dynamics of the vector population we consider the model previously formulated by Yang and colleagues [16], in which individuals are divided into two pertinent life-stages: aquatic (eggs, larvae and pupae,) and adult females (). We further extend the adult class by subdividing into the epidemiologically relevant stages for dengue transmission: susceptible (), incubating () for days and infectious (). For ease of reading, the temperature-dependent entomological factors are herein distinguished by a (dot) notation (further details in the following sections). The system of equations describing the vector population is: (6) (7) (8) (9) (10) Here, the coefficients and are the fraction of eggs hatching to larvae and the fraction of female mosquitoes hatched from all eggs, respectively. For simplicity and lack of quantifications for the local mosquito population, we assume these to be 1 (see the original publication for a discussion [16]). Moreover, denotes the rate of transition from aquatic to adults, and are the mortality rates, and is the intrinsic oviposition rate. The logistic term can be understood as the physical/ecological available capacity to receive eggs, scaled by the carrying capacity term, used in the fitting approach to indirectly estimate the adult mosquito population size (see below). From the above system, the basic offspring number (), that is, the mean number of viable female offspring produced by one female adult during its entire time of survival (and in the absence of any density-dependent regulation), can be derived as: (11) All parameters defining are temperature-dependent (see below). For a fixed temperature it is possible to derive expressions for the expected population sizes of each mosquito life-stage modelled. These are used to initialize the system, given the temperature present at the initial timepoint: (12) (13) The vector-to-human () and human-to-vector () incidence rates are assumed to be density-dependent and frequency-dependent (respectively), in respect to the type of infected host being considered: (14) (15) Here, is the biting rate and and are the vector-to-human and human-to-vector transmission probabilities per bite. This approach follows the recent framework from Althouse et al. which conforms to the expectations arising from the constant nature of the number of bites per mosquito [17]: conceptually, (i) an increase in the density of infectious vectors should directly raise the risk of infection to a single human; while (ii) an increase in the frequency of infected humans raises the risk of infection to a mosquito biting at a fixed rate. With the two hosts, the expression for dengue' s basic reproductive number is defined equally to previous modelling approaches [18], [19] but without human mortality: In this section we summarize the methodologies used for each of the seven entomological parameters dependent on temperature (Table 1). Here, is temperature in Celsius, is temperature in Kelvin and is the universal gas constant in. In the study by Yang et al. , from where we base the developmental part of our vector dynamical system (see above), temperature-controlled experiments were performed on populations of Aedes aegypti to derive closed-form expressions (based on polynomials) for the model' s rates (see Figures 2,3, 4 and 5 of the original publication [16]). We integrate such solutions into our framework: (16) (17) (18) (19) The relationship between the extrinsic incubation period and temperature has been formulated by Focks et al. [20] using an enzyme kinetics model previously proposed by other authors [21] and used in other dengue modelling approaches [22]. The model assumes that the rate of development is determined by a single rate-controlling enzyme. The expression used is: (20) The probabilities of transmission per mosquito bite and are modelled as previously estimated by Lambrechts and colleagues [23]. The data used in their study was both sampled from several other studies and obtained from de novo experiments that measured the variations in proportion of infected and transmitting vectors according to changes in temperature. The analysis was done for a variety of arboviruses from the flavivirus family, including the Dengue virus, the West Nile virus, Murray Valley Encephalitis virus and St. Louis Encephalitis virus. The expressions used are: (21) (22) The framework described above has only three fixed parameters that are neither temperature-dependent nor estimated in the MCMC approach. These can be found in Table 2. The outbreak time series was compiled from the official weekly reports from the Direcção Geral de Saúde (Portuguese health ministry) [10] issued throughout 2012 and 2013 and the special report by the European Centre for Disease Prevention and Control (ECDC) [9]. Temperature data for the island of Madeira was assembled from Weather Underground, a Weather Channel' s repository [24]. For this we chose a weather station located in the centre of Funchal, Madeira' s capital city, where most cases took place. We resorted to the website of Aeroportos da Madeira (Madeira Airports) for the statistics on airline traffic [25]. Finally, the figures for yearly investment in tourism were obtained from the official local source, the Instituto de Desenvolvimento Regional (Institute for Regional Development) [26]. For the fitting process a Markov chain Monte Carlo approach [27] is used to find combinations of parameters that can describe qualitative properties of Madeira' s outbreak. We define the jumping distribution as being symmetric (Gaussian), effectively defining a random walk Metropolis-Hastings algorithm: (23) (24) (25) (26) (27) Here, the Markov chain state is generally denoted by M, the proposal of new parameters by Y and the ODE system (described above) output by O. In step 1, is the Markov chain state of parameter at step, the pre-defined variance for each jump of parameter and the resulting proposal for time. In step 2, is the probability of acceptance. For this, we calculate the least squares distance between the data series and the ODE output for both the proposal of parameters and the previously accepted parameters. The probability is assumed to decrease exponentially with increases in least squares distances to the data. With this simple approach we explored all possible combinations of values from four open parameters (Table 3) that are able to closely describe the outbreak time series. Amongst these is the carrying capacity K, which we explore in order to indirectly estimate the number of adult mosquitoes per human, and, the timepoint of the first case. We also consider two linear coefficients, and, that scale the mortality rate and incubation period of adult mosquitoes - we argue that these entomological factors, as defined by Yang et al. in laboratory experiments [16], should be adjusted to possible biological/ecological local effects. For example, it has been previously demonstrated that mosquito and virus genotype can have an effect on both susceptibility and incubation [28], [29], while human and predator behaviour, as well as the local geospatial topology can affect adult mortality [30], [31]. By considering these linear effects, we do not change the relative effect of temperature variation on mortality and incubation per se, but rather allow the baselines to be different from the ones obtained from the laboratory, ideal conditions of Yang et al. study. For a discussion on how much field and laboratory entomological factors can differ, see the recent work by Brady and colleagues [32]. We address MCMC convergence by visual inspection and also quantify it using, the Gelman-Rubin statistic, which compares the variance between and within M independent MCMC chains [33]. Consider that each chain has length N steps and that is the parameter value in chain Then, when defining B as the between-chain and W as the within-chain variances, can be obtained using: (28) (29) (30) (31) (32) (33) (34) is expected to approximate 1 when the M chains have converged to the same stationary distribution. Values significantly larger than 1, for instance, indicate that the between-chain variance is greater that the within-chain variance, highlighting that the MCMC may need more time to converge or tuning of jump parameters is required [33]. In our approach we calculate and present for each estimated parameter (Table 3) using 30 independent chains started with random initial conditions. Our jump parameters are chosen to assure that all MCMC chains presented in this study have acceptance rates and chains are run for at least 1 million steps. A stochastic version of the ento-epidemiological framework was developed by introducing demographic stochasticity in the transitions of the dynamical system. We used multinomial distributions to sample the effective number of individuals transitioning between classes per time step. Multinomial distributions are generalized binomials, here defined as, where equals the number of individuals in each class and equals the probability of the transition event (equal to the deterministic transition rate). This approach has been demonstrated elsewhere, see e. g. [34]. The rainfall data displays three distinct peaks during this period, a small one coinciding with the start of the outbreak, one during its peak and another two months after the reported cases dropped close to zero (Figure 2). We first note that the actual amount of rainfall over the whole 6-months period under consideration was very small, and it is reasonable to question how much of an impact this could have made on the local mosquito population during the epidemic, especially given the year-round availability of breeding sites [5]. There are also contrasting observations in relationship to rainfall timing and case-response: the first peak was short and small but followed by a drastic increase in case numbers; in contrast, the second much longer and heavier rain episode coincided with the epidemic peak but was followed by a sudden decrease in case numbers; finally, the third peak in rainfall took place outside the time-range of interest. We therefore argued that rainfall was unlikely to have been a main player in the overall progression of this particular short-lived epidemic, although a small contribution cannot be ruled out. Compared to rainfall, which mostly affects Aedes' s habitat-quality and availability, temperature directly affects both the mosquito life-cycle and viral replication rates within the mosquito [16], [20], [23], [32], [38]. Accordingly, we note that the drop in minimum temperature towards the end of the year correlated with the decrease in case numbers (Figure 2), possibly delayed by the integrated length of the intrinsic cycles of human-vector-human transmission and aquatic-to-adult mosquito development. We initially tested a variety of SIR-based model frameworks with constant parameters in time, but it became clear that these models were unable to fit the sudden decline in case numbers after the epidemic peak in November. That is, models that would match the steep exponential rise in incidence, for example, and which would therefore predict very high reproductive numbers (), would inevitably generate epidemics of significantly higher magnitude and longer duration than the 2012 outbreak. As human intervention can be ruled out [9], and given the apparent strong correlation between declining temperatures and fading case numbers, we instead focused on a temperature-driven human-mosquito transmission model (see Methods) to gain more insight into the dynamic progression of this outbreak. In order to fit our ento-epidemiological model to the case data under temperature variations, or more specifically to derive particular parameter combinations that are able to reproduce the timing, shape and size of the epidemic, we employed a Markov chain Monte Carlo (MCMC) approach (see Methods). Figures 3A and B show the fitted model together with the weekly and cumulative incidence data, respectively. There is an overall close fit between model output and the data, although we notice two apparent deviations: one at the onset of the epidemic and one just after its peak. We argue that these are more likely artefacts of the data, however, rather than real discrepancies. That is, the sudden drop in incidence after the epidemic peak is likely due to the introduction of a new surveillance system in the first week of November [9], whereas the slight overestimation of cases during the initial phase of the outbreak could be due to the deterministic nature of our model and possible under-reporting at the onset of the epidemic. We also note that the new surveillance system did not change the clinical or laboratory-confirmation definitions per se (see annexe of [9] for case definition) but aimed at efficiently integrating data from all health care centres involved, including the private and public sectors. It is therefore not expected that this change in the system affected the sensitivity of case detection but instead increased the time and space resolution of the epidemic data from November onwards. Using this method we estimated that the possible timepoint of introduction of dengue to the island occurred towards the end of August, with the first autochthonous human infection between the and of September, two to three weeks before the first reported clinical case. Convergence to this date (range) was confirmed by independent MCMC runs as demonstrated in Figure 3C. That is, given random initial conditions for the four free parameters (Table 3), the system robustly converged towards equally distributed parameter estimates (by design, this approach produces parameter distributions rather than point estimates); see Methods and Supplementary Figures S1 and S2 for the resulting distributions of other parameters and quantification of convergence. At the time of the first local human infection, temperatures were already on a declining trajectory (Figure 2), which caused a significant reduction in the virus' s reproductive potential due to a combination of shorter mosquito life-expectancy, smaller population size and an increase in the extrinsic incubation period. These effects are demonstrated in Figure 4A, which also illustrates the colder than usual winter and slightly warmer summer during 2012 (compared to the average temperatures over the past 10 years). Dengue' s estimated reproductive potential, , is here given as a time-dependent quantity to highlight the temperature-driven dependencies of entomological factors. It is important to note that this differs from the often used time-varying effective reproduction number, , that takes into account the varying susceptibility levels in the population (see Methods for mathematical expressions). Of particular interest is the increase in the length of the extrinsic incubation period beyond the mosquitoes' average life-expectancies (Figure 4B), substantially contributing to the sharp drop in from at its peak at the end of the summer, to during late autumn and winter. We thus believe that this temperature-driven phenomenon might explain the rapid decrease in dengue incidence and essentially the end of the outbreak, with the expected delay due to the total length of the transmission and developmental cycles. Using these parameter insights we next investigated potential outcomes if the pathogen would have been introduced at different timepoints during 2012 and further considered introductory events during a ‘typical’ year using average temperatures of the past 10 years (2001–2011). In order to take into consideration the probabilistic nature of viral introduction and epidemic outcome, we expanded our framework to include demographic stochasticity and viral extinction (see Methods section). As demonstrated in Figure 4B, there was a time window of several months during 2012 when adult mosquito counts where sufficiently high and, critically, the virus' s temperature-dependent extrinsic incubation period was shorter than the mosquito' s average life-expectancy, thus allowing for efficient vector-human transmission. When simulating introduction events, we found significant differences in the epidemic windows between 2012 and a typical year due to deviations in temperature trends throughout the studied periods. Notably, the winter in 2012 was unusually cold, which resulted in a shorter window during which outbreaks could be measured (Figure 5A). At the same time, slightly warmer temperatures during the summer months of 2012 increased the transmission potential, (Figures 5B and 5C), and resulted in bigger outbreaks when compared to a typical year (Figure 5A). At first sight, differences of 2–4° Celsius in the summer months may seem insufficient to explain the differences in and consequently in outbreak sizes. However, according to experimental evidence, increases in temperature just above the critical point of 20° Celsius will strongly add to the overall vectorial capacity of Aedes mosquitoes [16], [20], [23], [32], [38]. This is a consequence of slight changes in the rates describing mortality, incubation and life-stage progression, which in concert have a cumulative effect and may be involved in positive feedback relationships. Hence, differences of a few degrees, especially when maintained over wide periods of time, can have significant and long lasting effects on the vector population and therefore on dengue' s reproductive number. We further investigated dengue' s success of invasion into the island by quantifying the frequency of stochastic simulations that developed into outbreaks above certain sizes, differentiating between 2012 (Figure 5B) and ‘typical’ years (Figure 5C). Comparing the occurrence of any-size () or major-size () outbreaks, we found the risk for the latter to be strongly linked to introductions during the summer months. In fact, as demonstrated in Figures 5B and 5C, there is a substantial risk for major epidemic outbreaks for introductory events taking place weeks or even months before reaches its full potential. This is because any introduction during that period can enjoy the climate-driven ‘deterministic’ growth in until late summer. We can thus identify a key epidemic window, dictated by temperatures above Celsius, in which efforts to detect and control imported cases are crucial for public health planning in Madeira. In agreement with the estimated differences in transmission potential between 2012 and average years, we also found the invasion success to be generally higher during 2012, which could potentially explain the success of the virus in that particular year. In fact, our results suggest that during a typical year a substantial proportion of introductions () are expected to go extinct before reaching epidemic potential, even during the peak in transmission potential (Figure 5C). Given the homogeneous assumptions of our modelling approach, we argue that these rates should be seen as ‘best’ case scenarios for the successful invasion of dengue in Madeira. In a more realistic scenario, in which heterogeneities in contacts and host and vector densities are present, we expect these rates to be potentially much lower, which could offer an explanation as to why dengue had failed to achieve sustained transmission on the island in the past. The 2012 dengue epidemic in Madeira was the first European outbreak showing significant and prolonged autochthonous transmission. With Aedes aegypti firmly established on the island and travel patterns in place connecting Madeira with other African and South American countries where dengue is now endemic or epidemic, it can be argued that introduction and sustained transmission was only a matter of time. Here, we used a mathematical modelling approach to investigate the underlying drivers of this important epidemic and to highlight the risks of potential future outbreaks. Of particular importance was the date when the virus had been introduced to the island together with the prevailing ecological conditions at this point and the months that followed. Whereas the first official clinical cases were reported on 3 October 2012, our method dates the timepoint of introduction just over a month earlier, at the end of August. There are various reasons for this discrepancy. Firstly, there is evidence that these initial two cases were the result of autochthonous transmission, i. e. they were not the individuals who introduced the virus to the island but rather subsequent cases [9]. Secondly, dengue infections are frequently asymptomatic [2], which means that several people could have been infected before some individuals developed symptoms sufficiently severe and/or specific for health care officials to suspect for dengue fever. Together, this could have led to a significant under-reporting, a common feature of dengue-endemic regions [2], [39], [40], especially during the onset of the epidemic. These initial cases were followed by a rapid rise in dengue incidence over the following month, with the epidemic peaking around early November, indicative of a high transmission potential at this point of the year. Our estimates of, however, showed that its maximum had been reached in August, a few months before the outbreak took place. To further investigate the causes and dynamics of this epidemic we addressed the conditions on the island during the relevant period. We looked for the possible role of local temperature variation and rainfall. We argued that the timing and strength of the three observed rain episodes was insufficient to have played a critical role in the outbreak, especially as the actual amount of precipitation was very small in each of these episodes. Furthermore, it is plausible that the year-round availability of breeding sites in flower and plant pots, as previously described in entomological studies of Madeira [5], [9], [41], may reduce the impact of short and sporadic rain episodes by allowing the mosquito population to persist throughout the year. Temperature, on the other hand, due to its aforementioned influence on the extrinsic incubation period, adult mortality and aquatic developmental rates, appeared to be the predominant driver and essentially limiting factor of the 2012 outbreak. According to our model, the temperatures in autumn not only caused a reduction in the number of adult mosquitoes but, crucially, dropped bellow the critical threshold where the incubation period is shorter than the average mosquito life-span and onward transmission to humans becomes probable. This effectively reduced vectorial capacity and stopped viral propagation, causing a significant decrease in dengue cases. Given the natural annual variation in temperature on the island we found significant differences in dengue' s transmission potential between summer and winter months. This is a consequence not only of varying mosquito population sizes but also of other temperature-dependent entomological and viral factors. During the warmer months, could reach for a few weeks, which stands just above the often reported range of 2–12 for dengue [42]. However, the estimates present in the literature are often based on methods that necessarily average the transmission potential over long periods of time, such as months, transmission seasons or, more commonly, years. In contrast, our estimates of are point estimates that follow temperature variations in real time. Crucially, when averaged over 2012, we obtained values of, in line with estimates from age-stratified sero-prevalence studies [43]. This, on the other hand, highlights some of the dangers in determining dengue' s region-specific based on averages over long periods of time, as the true values might vary significantly within just a couple of weeks due to temperature oscillations (as demonstrated here and in [16], [20], [23]) but also due to the heterogeneous and volatile nature of mosquito populations [6], [30], [44], [45]. Using the parameter estimates from fitting our model to the 2012 outbreak data, we simulated other scenarios where the virus was introduced at different timepoints during the year. For this, we separately considered temperature data for 2012 and the average for the past 10 years (2001–2011). The latter was used to make predictions on the epidemic and endemic potential of dengue during an average year in Madeira. Due to the slightly warmer summer in 2012 we found both the epidemic potential and probability of invasion to be higher when compared to a typical year on the island, potentially explaining the success of the virus in that particular year. However, our results also indicated a reasonable invasion potential between late April and October based on average temperatures and thus identified a key epidemic window, during which efforts by the local authorities should take place to prevent importation, to control the mosquito population and to raise awareness of residents, specially in Funchal. Importantly, there is little support for dengue to become endemic in Madeira, since temperatures regularly drop bellow Celsius outside this window, which severely affects several entomological and viral factors and effectively reduces vectorial capacity to unsustainable levels. Our results can also be used to discuss the potential implications for spreading dengue from the island to other countries. As mentioned in the Introduction and shown in Figures 1B–D, Madeira has a high influx of visitors, mostly from other European countries as well as South America. These are concentrated around two distinct holiday peaks, one in Easter and one during the main summer holiday season in August / September (Supplementary Figure S3). While our estimated timepoint of introduction of dengue coincided with the height in tourism around the end of August, which might explain the dynamics of the events that followed, it is important to note that the outbreak reached its peak when average tourist numbers had already dropped to their annual minimum (Figures 2 and S3), thus limiting the potential for disease exportation. Even so, a total of 81 reported cases were exported to European cities, a number that is possibly underestimated due to asymptomatic dengue infections [2]. This clearly demonstrates the future potential for spreading dengue from the island to continental European areas, with a particularly high risk for those regions with warm climates and where Aedes albopictus is well established, such as Italy or Southern France (Supplementary Figure S4). Some caution must be urged about the interpretation of some of our predictions. Our modelling approach was designed to address the qualitative relationships between viral, human and entomological factors that may have dictated the success and demise of Madeira' s dengue outbreak. However, this dynamic framework includes key assumptions that may affect estimations such as epidemic sizes and invasion success. For instance, recent modelling work suggests that spatial segregation between dengue' s hosts greatly reduces the propensity for large-scale outbreaks by restricting the pathogen' s access to the susceptible pool [19]. It is also known that demographic stochasticity plays an important, if not crucial role for the transmission of human pathogens [46], including dengue virus [19], [47]. We have made an effort to account for this by investigating the epidemic and endemic potential of Madeira using a stochastic version of our model. However, in order to keep the MCMC fitting methodology simple and robust, our parameter estimations were still dependent on deterministic assumptions, and the quantifications of invasion success and epidemic potential should thus be understood as average, if not worst-case scenarios. On the other hand, using this MCMC fitting approach allowed us to capture some of the expected underlying uncertainty, for example with regards to the possible timepoint of introduction (Figure 3C), despite using an underlying deterministic framework. In summary, we have shown that the 2012 dengue outbreak in Madeira was predominantly self-limited, driven to extinction by falling temperatures rather than human intervention. Our results demonstrate that there is little that supports the possibility of dengue to become endemic on the island; there is, however, a major risk for future epidemic outbreaks, with their likelihood significantly enhanced during periods of increased travel from dengue-endemic countries. These outbreaks are only expected within a limited window of time between late spring and summer. Control and social awareness efforts should therefore be placed within this time window to reduce economic and public health consequences, not only for Madeira but also for other European countries with strong tourism links to this island.

In 2012, Europe saw its first dengue epidemic taking place on the Atlantic island of Madeira. Due to strong tourism links, 81 cases were introduced into continental Europe in a short period of three months. Although Aedes aegypti, the mosquito-vector responsible for this particular outbreak, is extinct in mainland Europe, climatic and globalization trends have eased the recent establishment of Aedes albopictus, dengue' s secondary vector, in France, Germany, Italy and Spain. Before this epidemic, dengue had only sporadically achieved short chains of transmission. The presence of fully susceptible populations, however, makes the possible introduction into Europe a major public health concern. Here, using a mathematical approach, we analysed Madeira' s dengue outbreak, focusing on the necessary conditions for introduction, spread and persistence. We find that natural temperature cycles were the determining factor for the 2012' s outbreak demise, and are generally expected to severely disrupt dengue transmission between November and April, suggesting weak potential for endemicity. On the other hand, Madeira demonstrates a high potential for sporadic and potentially large epidemics in the remaining summer months, especially if the virus is introduced early during the warm season.

Abstract Introduction Materials and Methods Results Discussion

infectious diseases plant science medicine and health sciences theoretical biology population modeling infectious disease epidemiology epidemiology disease dynamics population dynamics plant pathology biology and life sciences population biology viral diseases computational biology dengue fever tropical diseases neglected tropical diseases

2014

The 2012 Madeira Dengue Outbreak: Epidemiological Determinants and Future Epidemic Potential

Although great progress in genome-wide association studies (GWAS) has been made, the significant SNP associations identified by GWAS account for only a few percent of the genetic variance, leading many to question where and how we can find the missing heritability. There is increasing interest in genome-wide interaction analysis as a possible source of finding heritability unexplained by current GWAS. However, the existing statistics for testing interaction have low power for genome-wide interaction analysis. To meet challenges raised by genome-wide interactional analysis, we have developed a novel statistic for testing interaction between two loci (either linked or unlinked). The null distribution and the type I error rates of the new statistic for testing interaction are validated using simulations. Extensive power studies show that the developed statistic has much higher power to detect interaction than classical logistic regression. The results identified 44 and 211 pairs of SNPs showing significant evidence of interactions with FDR<0. 001 and 0. 001<FDR<0. 003, respectively, which were seen in two independent studies of psoriasis. These included five interacting pairs of SNPs in genes LST1/NCR3, CXCR5/BCL9L, and GLS2, some of which were located in the target sites of miR-324-3p, miR-433, and miR-382, as well as 15 pairs of interacting SNPs that had nonsynonymous substitutions. Our results demonstrated that genome-wide interaction analysis is a valuable tool for finding remaining missing heritability unexplained by the current GWAS, and the developed novel statistic is able to search significant interaction between SNPs across the genome. Real data analysis showed that the results of genome-wide interaction analysis can be replicated in two independent studies. In the past three years, about 400 genome-wide association studies (GWAS) that focused largely on individually testing the associations of single SNP with diseases have been conducted [1]. These studies have identified more than 531 SNPs associated with different traits or diseases [2] and have provided substantial information for understanding disease mechanisms. Despite the progress that has been made, the significant SNP associations identified by GWAS account for only a few percent of the genetic variance which begs the question where and how the missing heritability can be identified [3], [4]. Possible explanations include [1], [4]: Another way to discover the missing heritability of complex diseases is to investigate gene-gene and gene-environment interaction. Disease development is a dynamic process of gene-gene and gene-environment interactions within a complex biological system which is organized into interacting networks [5]. Modern complexity theory assumes that the complexity is attributed to the interactions among the components of the system, therefore, interaction has been considered as a sensible measure of complexity of the biological systems. The more interactions between the components there are, the more complex the system is. The disease may be caused by joint action of multiple loci. Motivation for studying statistical interaction is to provide increased power for detecting joint acting effects of interacting loci than testing for only marginal association of each of the loci individually. Screening for only main effects might miss the vast majority of the genetic variants that interact with each other and with environment to cause diseases [6]. We argue that the interactions hold a key for dissecting the genetic structure of complex diseases and elucidating the biological and biochemical pathway underlying the diseases [7], [8]. Ignoring gene-gene and gene-environment interactions will likely obscure the detection of genetic effects and may lead to inconsistent association results across studies [9], [10]. GWAS in which several hundred thousands or even a millions of SNPs are typed in thousands of individuals provide unprecedented opportunities for systematic exploration of the universe of variants and interactions in the entire genome and also raise several serious challenges for genome-wide interaction analysis. The first challenge comes from the problems imposed by multiple testing. Even for investigating pair-wise interaction, the total number of tests for interaction between all possible SNPs across the genome will be extremely large. Bonferroni-corrected P-values for ensuring genome-wide significance level of 0. 05 will be too small to reach. The second challenge is the need for computationally simple statistics for testing interactions. The simplest way to search for interactions between two loci is to test all possible two-locus interactions. This exhaustive search demands large computations. Therefore, the computational time of each two-locus interaction test should be short. The third challenge is the power of the statistics for testing interaction. To ensure the genome-wide significance, the statistics should have high power to detect interaction. Developing simple and efficient analytic methods for evaluation of the gene-gene interactions is critical to the success of genome-wide gene-gene interaction analysis. Finally, the fourth challenge is replication of the finding of such interactions in independent studies. This report will attempt to meet these challenges, at least in part. To achieve this, we first should define a good measure of gene-gene interaction. Despite current enthusiasm for investigation of gene-gene interactions, published results that document these interactions in humans are limited and the essential issue of how to define and detect gene-gene interactions remains unresolved. Over the last three decades, epidemiologists have debated intensely about how to define and measure interaction in epidemiologic studies [7], [8], [11]–[15]; The concept of gene-gene interactions is often used, but rarely specified with precision [16]. In general, statistical gene-gene interaction is defined as departure from additive or multiplicative joint effects of the genetic risk factors [17]. It is increasingly recognized that statistical interactions are scale dependent [18]. In other words, how to define the effects of a risk factor and how to measure departure from the independence of effects will greatly affect assessment of gene-gene interaction. The most popular scale upon which risk factors are measured in case-control studies is odds-ratio. The traditional odds-ratio is defined in terms of genotypes at two loci. Similar to two-locus association analysis where only genotype information at two loci is used, odds-ratio defined by genotypes for testing interaction will not employ allelic association information. However, it is known that interaction between two loci will generate allelic associations in some circumstances [19]. Since they do not use allelic association information between two loci, the statistical methods based on the odds-ratio that is defined in terms of genotypes will have less power to detect interaction. To overcome this limitation, we will define odds-ratio in terms of a pseudohaplotype (which is defined as two alleles located on the same paternal or maternal chromosomes) for measuring interaction, and then we will investigate its properties and develop a statistic based on pseudohaplotype defined odds-ratio for testing interaction between two loci (either linked or unlinked). To demonstrate that the pseudohaplotype odds-ratio interaction measure-based statistic for detection of interaction between two loci will not cause false positive problems, we then investigate type I error rates. To reveal the merit and limitation of the pseudohaplotype odds-ratio interaction measure-based statistic for detection of interaction, we will compare its power for detecting interaction with the traditional logistic regression and “fast-epistasis” in PLINK [20]. Although nearly 400 GWAS have been documented, few genome-wide interaction analyses have been performed and few findings of significant interaction reported [8], [21], [22]. Emily et al [23] tested about 3,107,904–3,850,339 pairs of SNPs located in genes with potential protein-protein interaction and reported four significant cases of interactions, one in each of Crohn' s Disease, bipolar disorder, hypertension and rheumatoid arthritis in the WTCCC dataset, but these have not been replicated. To further evaluate the performance of our new statistic and test the feasibility of genome-wide interaction analysis, the presented statistic was applied to interaction analysis of two independent GWAS datasets of psoriasis where 1,266,378,301 pairs of SNPs from 50,327 SNPs in the first dataset and 1,243,782,750 pairs of SNPs from 49,876 SNPs in the second dataset were tested for interactions. These SNPs in the datasets were selected from 501 pathways assembled from KEGG [24] and Biocarta (http: //www. biocarta. com) pathway databases. A program for using the developed statistic to test interaction which was implemented by C++ can be downloaded from our website http: //www. sph. uth. tmc. edu/hgc/faculty/xiong/index. htm. Consider two loci: G and H. Assume that the codes and denote whether an individual is a carrier (non-carrier) of the susceptible genotypes at the loci G and H, respectively. Let D denote disease status where indicates an affected (unaffected) individual. Consider the following logistic model: (1) The odds-ratio associated with G for nonsusceptible genotype at the locus H is defined asSimilarly, the odds-ratio associated with H for nonsusceptible genotype at the locus G is defined asThe odds-ratio associated with susceptibility at G and H compared to the baseline category and is then computed asThe odds for baseline category and are determined asFrom equation (1), we clearly haveDefine a multiplicative interaction measure between two loci G and H as (2 – A) It is clear that (2 – B) If, i. e. , there is no interaction between loci G and H, then. This shows that the logistic regression coefficient for interaction term is equivalent to the interaction measure defined as log odds-ratio. The interaction measure can also be written as The values of odds-ratio defined in terms of genotypes depends on how to code indicator variables G and H. Suppose that alleles and are alleles that increase disease risk. For a recessive model, G is coded as 1 if the genotype is, otherwise, G is coded as 0. For a dominant model, G is coded as 1 if the genotypes are either or, otherwise G is coded as 0. The indicator variable H can be similarly coded. However, in real data analysis, the disease models are unknown. Especially, the types of two-locus disease models are large [25]. We may have a large number of possible coding, and many of them may have larger numbers of degrees of freedom than the allelic model. Similar to the odds ratio for genotypes, we can define odds-ratio in terms of alleles. Let be the probability that an individual becomes affected given they have genotype at locus G and at locus H, where is either or (i. e. is a member of the set {}) and is either or (i. e. is a member of the set {}). We can similarly define. We then can determine the odds-ratio associated with the allele at the G locus and allele at the H locus compared to the baseline asSimilarly, we measure the odds-ratio associated with the alleles and, respectively asSimilar to genotype, we can define a multiplicative interaction measure in terms of log odds-ratio for allele aswhich is equivalent toThe “fast-epistasis” test statistic in PLINK (http: //pngu. mgh. harvard. edu/~purcell/plink/index. shtml) is defined aswhere SE (R) and SE (S) denote the standard deviation of R and S, respectively. Absence of interaction is implied if and only ifThis is the basis of the “fast-epistasis” test in PLINK. Suppose that the locus G has two alleles and and the locus H has two alleles and. Let and be the frequencies of the alleles in the cases and controls, respectively. For the discussion of convenience, we introduce a terminology of “pseudohaplotype”. When two loci are linked, a pseudohaplotype is defined as the regular haplotype. When two loci are unlinked, a pseudohaplotype is defined as a set of alleles that are located in the same paternal or maternal chromosomes. The frequencies of a pseudohaplotype can be estimated by the classical methods for estimation of haplotype frequencies such as Expectation Maximization (EM) Algorithms. For simplicity, hereafter we will not make distinction between the haplotype and pseudohaplotype. When two loci are unlinked, a haplotype is understood as a pseudohaplotype. Let, and, denote the frequencies of haplotypes and in the cases and controls, respectively. We define a penetrance of the haplotype as the probability that an individual becomes affected given they have phased genotype. Let be the penetrance of an individual with the genotype, and be the penetrance of the haplotypes and, respectively. The penetrance of the haplotype can be mathematically defined aswhere and are the population frequencies of the haplotypes and, respectively. and represent a genotype coding scheme. Their represented genotypes depend on the specific genotype coding scheme. It should be noted that the haplotype and and have different meanings. By the same idea in defining genotype-based odds ratio in terms of penetrance of combinations of genotypes, we can determine the odds-ratio associated with the haplotypes compared to the baseline haplotype in terms of penetrance of the haplotypes asSimilarly, we calculate the odds-ratio associated with the haplotypes and, respectively, asIt is noted that replacing and in the definition of odds-ratio in terms of genotypes by leads to the definition of odds-ratio based on the haplotypes. However, the values and biological meanings of these two types of odds-ratios are different. Similar to genotypes, we can compute a multiplicative interaction measure in terms of log odds-ratio for haplotypes as (3) In the absence of interaction, we haveThe multiplicative odds-ratio interaction measure in equation (3) is defined by the penetrance of the haplotypes. From case-control data it is difficult to calculate the penetrance of the haplotypes. However, we can show that the multiplicative odds-ratio interaction measure in equation (3) can be reduced to (Text S1, Appendix A) (4) There are many algorithms and software to infer the haplotype frequencies in cases and controls. Therefore, we can easily calculate the multiplicative odds-ratio interaction measure by equation (4). It can be seen from equation (4) that the absence of interaction between two loci occurs if and only if the ratio of haplotypes frequencies in the cases and the ratio of haplotypes frequencies in the controls are equal. To gain understanding the multiplicative odds-ratio interaction measure, we study several special cases. In the previous section we defined the haplotype multiplicative odds-ratio interaction measure, which can be estimated by haplotype frequencies in cases and controls. By the delta method, we can obtain the variance of the estimator of the haplotype odds-ratio interaction measure [26]: where and are the number of sampled individuals in cases and controls. By the standard asymptotic theory we can define the haplotype odds-ratio interaction measure-based statistic for testing interaction between two loci: (6) where and are the estimators of the corresponding haplotype frequencies in cases and controls, respectively. When sample sizes are large enough to ensure application of large sample theory, is asymptotically distributed as a central distribution under the null hypothesis of no interaction between two loci. Under an alternative hypothesis of of interaction between two loci being present, the statistic is asymptotically distributed as a noncentral distribution with noncentrality parameter proportional to the haplotype multiplicative odds-ratio interaction measure. This statistic can be applied to both linked and unlinked loci. As we explained in Text S1, Appendix B, the proposed statistic is different from the “fast-epistasis” test in PLINK. For the unlinked loci, we can use case only design [27], [28] to study interaction between two loci in which equation is reduced to (7) In the previous sections, we have shown that when the sample size is large enough to apply large sample theory, the distribution of the statistic for testing the interaction between two loci under the null hypothesis of no interaction between them is asymptotically a central distribution. To examine the validity of this statement, we performed a series of simulation studies. MATLAB was used to generate two-locus genotype data of the sample individuals. A total of 100,000 individuals from a general population with an allele frequency, , haplotype frequency and disequilibrium coefficient were generated. A total of 10,000 simulations were repeated. Type I error rates were calculated by random sampling 500–1,000 individuals as cases and controls from the general population. Table 1 and Table 2 show that the estimated type I error rates of the statistic for testing interaction between two loci, assuming and, were not appreciably different from the nominal levels, and. To further examine the validity of the test statistic, we constructed Quantile-quantile (Q-Q) plots of the test statistic in datasets 1 and 2 shown in Figures 1A and 1B, where the P-values of the tests were plotted (as −log10 values) as a function of p values from the expected null distribution. Since the total number of all possible pair-wise tests for interaction between SNPs is too large to store all the results in computer we only stored P-values <. Consequently, Q-Q plots started with 4. Figures 1A and 1B showed good agreement with the null distribution. To evaluate the performance of the statistic for detection of interaction between two loci, we compared the power of the statistic to that of the logistic model and the “fast epistasis” test in PLINK. Power was calculated by simulation. A total of 1,000,000 individuals from a general population with allele frequencies, and and disequilibrium coefficient were generated. Two-locus disease models were used to generate cases and controls, and summarized in Table 3 where odds-ratio was defined in terms of genotypes. We considered three types of genotype coding. For a recessive model, homozygous wild type, heterozygous, and homozygous risk increasing genotypes were coded as 0,0, 1, respectively. For a dominant model, homozygous wild type, heterozygous, and homozygous risk increasing genotypes were coded as 0,1, and 1, respectively. For an additive model, they were coded as 0,1, and 2, respectively. The genotype coding for the logistic regression matched the simulation model. The statistic in equation (6) for the case-control version was used to evaluate the power. In the power simulations, we also assumed that and. An individual who is randomly sampled from the general population was assigned to case or control status depending on the two-locus disease models in Table 3. The process was repeated until a sample of 1,000 cases and 1,000 controls for the dominant and additive models, or a sample of 2,000 cases and 2,000 controls for the recessive model was obtained. A total of 10,000 simulations were repeated. In Figures 2A–2C, power comparisons among the logistic regression model, the “fast-epistasis” in PLINK and the statistic under two-locus recessiverecessive disease model for significance levels, and, respectively are presented. In Figures 3A–3C, power comparisons among the logistic regression model, the “fast-epistasis” in PLINK and the statistic under two-locus dominantdominant disease model for significance levels, and, respectively are shown. In Figures 4A–4C, power comparisons between the logistic regression model and the statistic under two-locus additiveadditive disease model for significance levels, and, respectively are demonstrated. Several remarkable features emerge from these Figures. First, these power Figures indeed demonstrate that the power increases as the measure of the interaction between two loci increases. The power curves were plotted as a function of the traditional genotype odds ratio. We observed that the power curves were a monotonic increasing function of the genotype odds ratio. Therefore, the test statistic can detect the strength of the interaction between two loci. Second, the test statistic had much higher power to detect interaction between two loci than the logistic regression and the “fast-epistasis” test in PLINK. Third, the more complex the disease models were, the larger the differences in power between the test statistic, the “fast-epistasis” test in PLINK and logistic regression that were observed. When two loci are unlinked where we do not observe the allelic association between two loci in the population as a whole, our results also hold. We assumed the following allele and haplotype frequencies in the population: , and. Other parameters were defined as before. A total of 10,000 simulations were repeated to simulate the power of three statistics under three disease models with the significance level. Figures 5A, 5B and 5C showed the power of three statistics for testing interaction between two unlinked loci under two-locus recessiverecessive, dominantdominant, and additiveadditive disease models, respectively. These Figures again demonstrated that the power of the test statistic was still much higher than that of the logistic regression and the “fast-epistasis” test in PLINK. The conclusions still hold for the significance levels and (Data were not shown). To evaluate its performance for detection of interaction between two loci, the proposed test statistic was applied to interaction analysis of two independent GWAS datasets of psoriasis which were downloaded from dbGaP. Psoriasis is a common chronic inflammatory skin disease affecting 2%–3% of the world population. Originally, the first study included 955 individuals with psoriasis and 693 controls, which is considered as dataset 1. The second replication study included 466 individuals with psoriasis and 732 controls, which is designated dataset 2. All cases and controls are of European origin [29]–[31]. After using PLINK [20] to check for contamination, cryptic family relationship and non-Caucasian ancestry, 123 samples were excluded. Subsequently we retained for analysis 915 cases and 675 controls from the first study and 431 cases and 702 controls from the second study. All 2,723 samples had been genotyped with the Perlegen 500K array. In the initial dataset, 451,724 SNPs passed quality control (call rate>95%). To further ensure the quality of the typed SNPs, we used PLINK software to remove the SNPs with >5% missing genotypes, Hardy-Weinberg disequilibrium (P-values <0. 0001), MAF<0. 01 and duplicated markers. In this application, we only considered common SNPs with MAF>0. 01. After quality control filtering, a total of 451,724 SNPs were pruned to 443,018 and 439,201 SNPs with the average genotyping rate 99. 3% in the first and second studies, respectively. Since testing for all possible two-locus interactions across the genome in genome-wide interaction analysis requires extremely large computation, we conducted pathway-based genome-wide interaction analysis. We assembled 501 pathways from KEGG [24] and Biocarta (http: //www. biocarta. com). The assignment of SNPs to a gene was obtained from NCBI human9606 database (version b129). We used the statistic to test interactions of all possible pairs of SNPs located in genes within the assembled 501 pathways. The total number of SNPs in dataset 1 and dataset 2 being tested was 50,327 and 49,876, respectively. The serious problem in genome-wide interaction analysis is multiple testing. We used two strategies to tackle this problem. One is to use false discovery rate (FDR) [32] to declare significance of interaction. Another is replication of the findings in two independent studies, which enhances confidence in interaction tests [22]. We looked for consistent results across the two independent studies. In total, 44 pairs of SNPs showed significant evidence of interactions with FDR<0. 001, which roughly corresponds to the P-value <, in two independent studies (Table S1). These 44 pairs of SNPs were derived from 71 distinct SNPs located in 60 genes, including HLA-C, HLA-DRA, HLA-DPA1, LST1, MICB and NOTCH4. Of 44 pairs of SNPs, only one pair of interacting SNPs: rs2395471 and rs2853950 showed significant marginal association in two independent studies. An additional 211 pairs of SNPs with FDR less than 0. 003 in the two studies is listed in Table S2. These interacting SNPs were mainly located in 19 pathways, including a number of signaling pathways, and immune-related antigen processing and presentation as well as natural killer cell mediated cytotoxicity pathways (Figure 6). Several remarkable features emerge from these results. First, although we can observe a few interactions between SNPs within a gene, the majority of interactions occurred between genes that are often in different pathways. Since the number of SNPs typed within each gene was limited, it is unknown whether this is a general rule or just a special case. Second, a SNP in one gene might interact with multiple SNPs in multiple genes. For example, SNP rs3131636 in the gene MICB interacting with the SNPs rs915895, rs443198, rs3134929 in the gene NOTCH4, the SNP rs1052248 in the gene LAST1/Natural cytotoxicity triggering receptor 3 (NCR3) and the SNP rs1799964 in the gene LTA/TNF. SNP rs1799964 in the gene LTA/TNF interacting with SNPs rs3131636, rs3132468 in the gene MICB, SNPs rs9268658 and rs3135392 in the gene HLA-DRA, SNP rs2227956 in the gene HSPA1L. However, this does not imply that multiple causal SNPs within a gene will interact with multiple causal SNPs within another gene. It is quite likely that this is due to LD between the SNPs within a gene. Third, although interacting SNPs did not form large connected networks, the interacting SNPs connected pathways into a large complicated network. This may imply that many genes and pathways are involved in the development of psoriasis. Fourth, upstream of many pathways included genes with interacting SNPs. For example, genes MICB, CHRM3, HLA-DRA and CIITA, EPHB1 and EPHB2, LAMA1 and LANA5, ITGA1, LTBP1, TNF, and FGF20 that contain interacting SNPs are in the upstream of natural killer cell mediated cytotoxicity, calcium signaling pathway, antigen processing and presentation, axon guidance, ECM-receptor interaction pathway, focal adhesion, TGFB pathway, MAPK pathway and regulation of acting cytoskeleton, respectively. Fifth, most interacting SNPs are in introns and accounted for 77% of total interacting SNPs. Table 4 listed 15 pairs of interacting SNPs that have non-synonymous substitutions. It is unknown how these nonsynonymous mutations are involved in the pathogenesis of psoriasis. From the literature we know that Plexin C1 receptor is a tumor suppressor gene for melanoma [33], NOTCH4 is involved in schizophrenia [34], Phosphodiesterase 4D (PDE4D) is associated with ischemic stroke [35], HLA-DRA is one of the HLA class II alpha chain genes that plays a central role in antigen processing, and neuregulin 1 (NRG1) has been implicated in diseases such as cancer, schizophrenia and bipolar disorder [36]. Table 5 includes five interacting pairs of SNPs, one of which falls in the microRNA (MiRNA) binding region. miRNAs, which are 22 nucleotide small RNAs and regulate gene expressions by pairing the miRNA seed region with the target sites, have been implicated in many biological processes including the immune response, biogenesis and tumorigenesis [37]. Mutations in the target sites will affect miRNA activity. A number of studies have identified polymorphisms in the miRNA target sites associated with the diseases [37]. Interestingly, we identified four SNPs in the miRNA (miR-324-3p, miR-433, and miR-382) target sites which interact with five SNPs to contribute to psoriasis. In previous studies, miR-382 has been associated with dermatomyositis, Duchenne muscular dystrophy and Miyoshi myopathy [38], miR-433 and miR-324 with lupus nephritis [39] and miR-433 with Parkinson' s disease [40]. Some researchers suggest that in genome-wide interaction analysis only SNPs with large or mild marginal genetic effects should be tested for interaction. To examine whether this strategy will miss detection of interacting SNPs, we showed in Table 6 the 20 top pairs of interacting SNPs and in Table S3 all pairs of interacting SNPs with FDR less than 0. 003. Surprisingly, 75% of SNPs with P-values (in dataset 1) larger than 0. 2 and 44% of SNPs with P-values larger than 0. 5 in two studies were observed in Table S3. Table 6 and Table S3 strongly demonstrated that while both SNPs did not demonstrate significant evidence of marginal association, they did show significant evidence of interaction. To further evaluate the performance of the proposed statistic, in Table 7 and Table S4 we list P-values for testing interaction calculated by the statistic, the “fast-epistasis” in PLINK and logistic regression using genotype coding. In Table 7 the 20 top pairs of interacting SNPs and in Table S4 the results of 233 pairs of interacting SNPs are presented. The P-values for interaction calculated by the statistic are much smaller than those from the “fast-epistasis” in PLINK and the logistic regression using genotype coding (Table 7 and Table S4). Moreover, the “fast-epistasis” in PLINK and the logistic regression coded by genotype detect very few interactions that can be replicated in two independent studies (Table 7 and Table S4). In fact, our results for all tested SNPs in 501 pathways showed that the “fast-epistasis” in PLINK and logistic regression coded by genotypes detected very few interactions that can be replicated in two studies (data not shown). Eighteen significantly interacting SNPs identified by Bonferroni correction were listed in Table 8. In dataset1, the total number of SNPs for testing interaction was 50,327. The P-values for declaring interaction between SNPs after Bonferroni correction was. We found that there were 2,210 significant interactions with P-values less than in the dataset 1. Then, interaction for all these 2,210 pairs of SNPs in the dataset 2 was examined. The P-values for declaring interaction between SNPs after Bonferroni correction in dataset 2 was. We identified eight significant interactions that were replicated in dataset 2. Similarly, if we started with dataset 2, the total number of SNPs for testing interaction was 49,876. The P-values for declaring interaction between SNPs after Bonferroni correction was. Significant interactions with the P-values less than in dataset 2 were seen between 1,913 pairs of SNPs. Then, we tested for interaction for all these 1,913 pairs of SNPs in the dataset 1. The P-values for declaring interaction between SNPs after Bonferroni correction in dataset 1 was, and 10 significant interactions were detected that were replicated in the dataset 1. A total of 9 interactions were common in Table 8 and Table S1 and Table S2. The development of most diseases is a dynamic process of gene-gene and gene-environment interactions within a complex biological system. We expect that genome-wide interaction analysis will provide a possible source of finding missing heritability unexplained by current GWAS that test association individually. But, in practice, very few genome-wide interaction analyses have been conducted and few significant interaction results have been reported. Our aim is to develop statistical methods and computational algorithms for genome-wide interaction analysis which can be implemented in practice and provide evidence of gene-gene interaction. The purpose of this report is to address several issues to achieve this goal. The first issue is how to define and measure interaction. Odds-ratio is a widely used measure of interaction for case-control design. The odds-ratio based measure of interaction between two loci is often defined as a departure from additive or multiplicative odds-ratios of both loci defined by genotypes. The genotype-based odds-ratio does not explore allelic association information between two loci generated by interaction between them in the cases. Any statistics that are based on genotype defined odds-ratio will often have low power to detect interaction. To overcome this limitation, we extended genotype definition of odds-ratio to haplotypes and revealed relationships between haplotype-defined odds-ratio and haplotype formulation of logistic regression. To further examine the validity of this concept, we studied the distribution of the test statistic under the null hypothesis of no interaction between two either linked or unlinked loci. Through extensive simulation (assuming allelic association in the controls), we show that the distribution of the haplotype odds-ratio-based statistic is close to a central distribution even for small sample size and that type I error rates were close to the nominal significance levels. The second issue is the power of the test statistic for genome-wide interaction analysis. The genome-wide interaction analysis requires testing billions of pairs of SNPs for interactions. The P-values for ensuring genome-wide significance level should be very small. Therefore, developing statistics with high power to detect interaction is an essential issue for the success of genome-wide interaction analysis. As an alternative to the logistic regression and the “fast-epistasis” in PLINK, we presented a haplotype odds-ratio-based statistic for detection of interaction between two loci and illustrated its power by extensive simulations. The power of the haplotype odds-ratio-based statistic ended up being a function of the measure of interaction and had much higher power to detect interaction than the “fast-epistasis” in PLINK and logistic regression. The third issue is whether the interactions exist with no marginal association and how often they might occur in practice. Our data demonstrated that the majority of the significantly interacting SNPs showed no marginal association. Surprisingly, 75% of interacting SNPs with P-values (for testing marginal association) larger than 0. 2 and 44% of interacting SNPs with P-values (for testing marginal association) larger than 0. 5 in two studies were observed in our analysis. This strongly suggested that testing interaction for only SNPs with strong or mild marginal association will miss the majority of interactions. The fourth issue is that of replication of the results. Genome-wide interaction analysis involves testing billions of pairs of SNPs. Even if after correction of multiple tests, the false positive results might be still high. To increase confidence in interaction test results, replication of interaction findings in independent studies is often sought. To date, very few results of genome-wide interaction analysis have been replicated. This begs the question whether the significant interaction can be replicated in independent studies. In this report, we show that interaction findings can be replicated in two independent studies. The fifth issue is correction for multiple testing. Genome-wide interaction analysis often involves billions of tests, which would require an extremely small Bonferroni-corrected P-value to ensure a genome-wide significance level of 0. 05. Replication of finding at such small P-values in independent studies is often extremely difficult. However, Bonferroni correction assumes that the tests are independent, yet many interaction tests are highly correlated. Correlations in the interaction tests come from two levels [23]. First, two pairs of SNPs may share a common SNP. Second, SNPs in the interaction tests may be dependent due to allelic association. The Bonferroni correction assuming independent tests will be overly conservative due to high correlations among the interaction tests. In this report, two strategies were used to tackle the multiple testing issues. The first is to use FDR to control type I error. The second is to replicate interaction finding. Replication allows us to detect the interactions that are frequent and consistent [22]. This approach still has the limitation that we still make independent assumption of the tests in calculation of FDR. Recently, Emily et al. (2009) [23] proposed to develop a Bonferroni-like correction for multiple tests based on the concept of the effective number of SNP pairs. The concept of the effective number of tests takes correlation among the tests into account and can be applied to both P-value and FDR correction [41]. This may be a promising approach to multiple test corrections in the genome-wide interaction analysis. Although our data show that interactions can partially find the heritability of complex diseases missed by the current GWAS, they are still preliminary. Due to extremely intensive computations demanded by genome-wide interaction analysis we only tested interactions of a small set of SNPs which were located in the genes of 501 assembled pathways in a PC computer. The truly whole genome interaction analysis in which we will test for interactions between all possible pairs of SNPs across the genome has not been conducted. Gene-gene interaction is an important, though complex concept. The statistical interactions are scale dependent. There are a number of ways to define gene-gene interaction. How to define gene-gene interaction and develop efficient statistical methods and computational algorithms for genome-wide interaction analysis are still great challenges facing us. The main purpose of this report is to stimulate discussion about what are the optimal strategies for genome-wide interaction analysis. We expect that in coming years, genome-wide interaction analysis will be one of major tasks in searching for remaining heritability unexplained by the current GWAS approach.

It is expected that genome-wide interaction analysis can be a possible source of finding heritability unexplained by current GWAS. However, the existing statistics for testing interaction have low power for genome-wide interaction analysis. To meet challenges raised by genome-wide interactional analysis, we develop a novel statistic for testing interaction between two loci (either linked or unlinked) and validate the null distribution and the type I error rates of the new statistic through simulations. By extensive power studies we show that the developed novel statistic has much higher power to detect interaction than the classical logistic regression. To provide evidence of gene–gene interactions as a possible source of the missing heritability unexplained by the current GWAS, we performed the genome-wide interaction analysis of psoriasis in two independent studies. The preliminary results identified 44 and 211 pairs of SNPs showing significant evidence of interactions with FDR<0. 001 and 0. 001<FDR<0. 003, respectively, which were common in two independent studies. These included five interacting pairs of SNPs, some of which were located in the target sites: LST1/NCR3, CXCR5/BCL9L and GLS2 of miR-324-3p, miR-433, and miR-382, and 15 pairs of interacting SNPs that had nonsynonymous substitutions.

Abstract Introduction Methods Results Discussion

genetics and genomics/complex traits

2010

A Novel Statistic for Genome-Wide Interaction Analysis

Very little is known about how environmental changes such as increasing temperature affect disease dynamics in the ocean, especially at large spatial scales. We asked whether the frequency of warm temperature anomalies is positively related to the frequency of coral disease across 1,500 km of Australia' s Great Barrier Reef. We used a new high-resolution satellite dataset of ocean temperature and 6 y of coral disease and coral cover data from annual surveys of 48 reefs to answer this question. We found a highly significant relationship between the frequencies of warm temperature anomalies and of white syndrome, an emergent disease, or potentially, a group of diseases, of Pacific reef-building corals. The effect of temperature was highly dependent on coral cover because white syndrome outbreaks followed warm years, but only on high (>50%) cover reefs, suggesting an important role of host density as a threshold for outbreaks. Our results indicate that the frequency of temperature anomalies, which is predicted to increase in most tropical oceans, can increase the susceptibility of corals to disease, leading to outbreaks where corals are abundant. Climatic and oceanographic conditions can modify a wide variety of ecological processes. For example, ocean temperature can control species ranges, the strength of species interactions, the dispersal and survival of marine larvae, and the rates of metabolism and speciation [1–6]. Additionally, anomalously high temperature and other environmental stresses can influence the severity and dynamics of infectious diseases by increasing host susceptibility and pathogen virulence [7,8]. For example, the severity of human epidemics including cholera [9–11] and tick-borne encephalitis [12] are both related to temperature and, possibly, to recent climate change [13]. Temperature and climate change have also been implicated in plant and animal disease outbreaks in both terrestrial and aquatic habitats [7,14–17], and could influence coral disease severity [18–20], potentially accelerating the global loss of coral reefs. Corals are the foundation species of tropical coral reef ecosystems. They directly facilitate thousands of associated species by generating the physically complex reef structure [21,22]. Reductions in coral abundance can cause rapid loss of reef biodiversity [23]. The hypothesized link between anomalously high temperatures and coral disease outbreaks is supported by small-scale field studies indicating that prevalence and the rate of within-colony spread of several coral diseases are higher during the summer [24–30]. Such seasonal changes in disease severity could be driven in part by higher summertime temperature, but could also be caused by a variety of other abiotic factors that vary seasonally within sites. Additionally, such investigations do not directly address the role of temperature anomalies in driving the conspicuous variability of coral disease severity among years and locations [30–32] that has long intrigued coral reef ecologists. Missing are large-scale, longitudinal investigations that combine long-term monitoring of multiple populations with accurate, fine-grained measurements of local temperature anomalies. Longitudinal studies (i. e. , the repeated sampling of individuals or populations) help control for potential confounding factors and inherent temporal variability [33]. Such powerful epidemiological approaches are rarely applied to marine epidemics (but see [34,35]), which has limited our understanding of potential links between temperature and disease outbreaks in the ocean, especially at large spatial scales. Here we describe a regional-scale test of the hypothesis that ocean temperature can influence disease frequency. We analyzed the relationship between the frequency of white syndrome in scleractinian corals and of warm temperature anomalies across the Great Barrier Reef (GBR). Forty-eight reefs were monitored for 6 y (1998–2004), and reef-specific weekly sea surface temperature anomalies (WSSTAs; the frequency of deviations ≥ 1 °C) were derived from a satellite sea surface temperature (SST) database. White syndrome is an emerging disease of Pacific reef-building corals, reported in 17 species from families including Acroporidae, Pocilloporidae, and Faviidae, which comprise the majority of dominant species on the GBR [30]. Severe white syndrome outbreaks can affect coral composition and cover [30]. Little is known about the etiology of white syndrome, although it is presumably infectious and the characteristics are similar to Caribbean white diseases such as white band and white plague [36]. Like the Caribbean white diseases, white syndrome could comprise a group of distinct diseases with similar signs [30]. White syndrome can cause either partial or whole colony mortality and is characterized by a white band of tissue or recently exposed skeleton that moves across the colony as the disease progresses [30,37]. White syndrome has been present on the GBR since at least the beginning of systematic disease monitoring in 1998, but its frequency increased 20-fold in 2002 [30]. This rise came after a year in which the region experienced its second warmest summer in the 20-y satellite record, with 58% of reefs having weekly anomalies of 1 °C or higher. However, even during the peak of the outbreak, there was considerable variation in disease frequency among reefs (0 to 343 cases per 1,500 m2) (Figure 1B). WSSTA also varied substantially among reefs, especially during the warm summers of 1998/1999 and 2001/2002 when some reefs were anomalously warm for 30 wk of the year, but the weekly temperatures on many others never deviated from the long-term local averages (i. e. , WSSTA = 0). Reefs with relatively high coral cover and WSSTA had the greatest white syndrome frequency (Table 1). From the negative binomial regression model, the parameter estimates for the three covariates (WSSTA, coral cover, and the interaction between the two) were positive (i. e. , they predicted an increase in frequency) and highly significant (all p < 0. 000; Table 2). The interaction term (WSSTA × coral cover) explained a statistically significant amount of the increase in frequency of disease among all the covariates in the model (χ2 = 17. 49, df = 1, p < 0. 0000). The deviance statistic for the negative binomial model was 1. 0201, suggesting a very good fit to the data. Disease frequency predicted for nine WSSTA–coral cover combinations (based on the tenth, 50th, and 90th observed quintiles of these covariates) is presented in Table 3. The observed and predicted values indicate that disease frequency only increases substantially with the combination of extreme levels of both covariates. The model is a fairly conservative predictor of this relationship because the observed number of cases with high WSSTA and high coral cover (Table 1) was actually higher than predicted by the model. The frequency of warm temperature anomalies was positively related to white syndrome frequency across the GBR. The disease surveys documented considerable variation in white syndrome frequency (0 to 343 cases per 1,500 m2) among years and reefs. Our results suggest that this variance was caused in part by the number of warm temperature anomalies during the year preceding the disease surveys. A positive effect of high temperature on the severity of coral disease outbreaks might be caused by physiological stress impairing host immunity [8,38,39]. WSSTA, our metric of thermal stress, is based on the frequency of warm anomalies of 1 °C or higher because short-term temperature increases of this degree can cause measurable physiological stress in a coral host [40–46]. WSSTA summarized temperature anomalies throughout the year, including winter anomalies that might also affect the susceptibility of corals to disease [46]. Increased densities of symbiotic dinoflagellate algae (zooxanthellae) at the beginning of winter and the subsequent accumulation of coral-tissue biomass throughout cooler months are thought to influence coral responses to future stresses [47]. These processes are compromised by longer warm periods during the summer or warmer than usual winter temperatures [45]. In fact, winter warming could have the dual effect of predisposing hosts to disease and facilitating more rapid pathogen growth [7]. Summertime anomalies could also increase pathogen virulence by initiating virulence factors [48] or increasing the growth of pathogens [39] for which the normal summertime temperature is below the thermal optima. Our results also indicate that thermal stress is necessary, but not sufficient, for white syndrome outbreaks to occur. Coral cover must also be high; generally 50% or higher (Table 1). White syndrome was uncommon during the 12 mo after the summer of 1998/1999 when WSSTAs were more frequent and occurred at more sites than during 2002/2003. But in 1998/1999, total coral cover was less than 50% at the 20 reefs with the highest WSSTA (Figure 2A), and there was a weak negative relationship between WSSTA and cover (p = 0. 09, linear regression analysis; Figure 2A). In contrast, in 2002/2003, there was a positive association between WSSTA and coral cover (p = 0. 05, Figure 2B). This was possible because there was no reef-specific correlation of WSSTA between 1998/1999 and 2002/2003 (p = 0. 90, Figure 2C). Total coral cover is a reasonable estimation of host abundance in this system because the susceptible species are the competitively dominant space holders [30,49]. A positive relationship between host density and disease prevalence has been clearly demonstrated in many host-pathogen systems [35,50–52], and is considered a hallmark of the infectious process [53]. High host density can have several effects on disease dynamics. For example, it is most often associated with greater rates of horizontal transmission [54–56], leading to localized increases in prevalence. High coral cover reduces the distance between neighboring coral colonies [57] and thus between infected and healthy hosts, increasing the potential for horizontal disease transmission between corals in close proximity. In addition, host density can be positively related to the density of disease vectors [58,59], although no specific vector (s) have been identified for white syndrome. Independent of host density, total coral cover itself, including the abundance of nonsusceptible individuals and species, might also be causally linked with increased white syndrome frequency. A wide variety of biological properties of coral reefs are related to coral cover. For example, the abundance and composition of fishes and invertebrates that could act as disease vectors are tightly linked with total coral cover and reef heterogeneity [21,23]. Competitive interactions among corals increase nearly exponentially with total coral cover and, on the GBR, are relatively rare when cover is below 50% [57]. Corals compete directly by damaging the tissue of neighboring colonies with tentacles and digestive filaments [60]. These encounters usually cause lesions and local tissue necrosis [60] that could facilitate pathogen transmission and colony infection. Additionally, uninfected hosts likely experience physiological stress and a reduction in fitness on high-cover reefs from such direct competitive encounters [61] as well as from indirect competition such as shading [60], which could also reduce disease resistance. Regardless of the relative importance of these and other potential mechanisms for increased host susceptibility or disease transmission where coral cover is high, there is a cover threshold of approximately 50% (Table 1) for white syndrome outbreaks and, frequently, even for the occurrence of this disease on a reef. No white syndrome cases were recorded on 45% of the reefs with cover less than 50% (n = 235). In contrast, 88% of reefs with cover greater than 50% had at least one infected colony (n = 47). Such thresholds for pathogen colonization or persistence based on host density or other factors are theoretically predicted and typical of the dynamics of many wildlife diseases [8,35]. The technique used to measure the intensity of community-wide white syndrome outbreaks (i. e. , counting the number of infected colonies) could lead to a spurious relationship between coral cover and disease frequency, since more colonies could be sampled at higher-cover reefs. This potential artifact was accounted for by including coral cover as a covariate in the statistical model. Additionally, our results indicate that this potential sampling effect did not occur, or at least was undetectable. For example, disease frequency is very low and essentially constant across reefs with coral cover ranging from 0% to 50% (Table 1). Furthermore, the significant WSSTA × coral cover interaction term indicates that the coral cover effect was nonadditive. Finally, on the GBR and other Indo-Pacific reefs, coral cover and colony density generally are not positively related [62]. During the early stages of reef recovery after a major disturbance when nearly all colonies are small and coral cover is very low [62], colony density and cover can be positively related [57,63]. However, when coral cover is high, reefs are usually dominated by large colonies [62], and density and cover are typically negatively related [57]. On the GBR, this frequently observed parabolic relationship between coral cover and colony density is caused by the domination of high-cover reefs by large tabular colonies that exclude smaller non-tabulate species [49]. This was the case in our study on the highest cover reefs in the Cooktown/Lizard Island and Capricorn Bunkers sectors (Figure 1A) where most of the white syndrome outbreaks occurred (C. Page, personal communication). Therefore, our sampling design could in fact underestimate disease severity on very high-cover reefs, diminishing the measured importance of coral cover. Diseases can cause dramatic changes in host populations and can have lasting effects on the structure and functioning of marine ecosystems by reducing the abundance of keystone consumers and habitat-forming foundation species [19,53,64,65]. For example, a pandemic wasting disease of eelgrass populations in the 1930s caused widespread losses along the Atlantic coasts of Canada, the United States, and Europe [66]. In some affected areas, the disease was estimated to have reduced stands to less than 1% of their normal abundance [67]. Oyster diseases in the Chesapeake Bay, where the pathogen Perkinsus marinus has caused annual mortality ranging from 24% to 57%, contributed to the commercial collapse of the regional oyster industry and to the regional loss of oyster reef habitat [68]. Similarly, an unidentified disease decimated populations of the keystone herbivore Diadema antillarum in the 1980s throughout the Caribbean [69,70]. During the same time period, white band disease dramatically reduced the abundance of the two most abundant Caribbean corals, Acropora palmata and A. cervicornis, causing changes in reef structure unprecedented in the last 3,000 y [71,72]. The impacts of marine epidemics could increase if warm temperature anomalies become more frequent or extreme [13,18,19] as predicted by several climate change models [41]. Additionally, temperature could have locally additive or even synergistic impacts if the prevalence of disease or multiple diseases and non-infectious bleaching is increased by warm temperature anomalies [29,38]. For example, bleached corals could be more susceptible to infection [38]. The peak of the white syndrome outbreak occurred after the very warm austral summer of 2001/2002, concomitant with the most severe bleaching episode—in terms of number of reefs affected and intensity of bleaching—ever recorded on the GBR [73]. On some reefs, bleaching and outbreaks of atramentous necrosis, another GBR coral disease, occurred nearly simultaneously [29]. But surprisingly, across the GBR, there was little spatial concordance between bleaching and white syndrome severity in 2001/2002. The most intensive bleaching during 2001/2002 was concentrated in the central latitudes [73] where white syndrome frequency was generally very low (Figure 1). In contrast, there was little or no bleaching on reefs in the southern GBR, including the Capricorn Bunkers sector, where white syndrome outbreaks were most severe (Figure 1). The causes of this negative correlation are unclear, but could include variable host susceptibility, local species composition, thermal history, and prior disturbances. Regardless of potential causes, the segregation of these two impacts of anomalously high temperatures might limit local coral loss, but could lead to additive net declines in coral cover across the region. Alternatively, rising ocean temperature or an increase in summertime anomalies could inhibit marine epidemics. Environmental stress is often assumed to increase disease severity, but stresses that directly reduce host density can have the opposite effect [8] (Figure 3). The role of coral cover in mediating the influence of temperature on disease frequency suggests that temperature could have an important inhibitory effect on white syndrome via bleaching-induced coral mortality. High temperatures only 1–2 °C above the normal summer maximum can cause bleaching and mass coral mortality [41,42], leading to a reduction in host density and total coral cover. Therefore, anomalously high water temperature could, in contrast to our results, reduce the prevalence of coral diseases with host density or coral cover thresholds. However, host density is not always related to the spread of disease, such as when the disease is not infectious, if local secondary transmission is rare, and when pathogens originate outside the local host population or in other host species [8]. In such cases, the relationship between stress and disease severity is generally predicted to be positive [8] (Figure 3). Environmental stress can also reduce the intensity or probability of outbreaks by negatively affecting pathogen fitness or virulence [8]. It is possible that coral pathogens are negatively affected by anomalously high water temperature. In fact, laboratory studies have found that beyond thermal optima, coral pathogens can have reduced photosynthetic [74] and growth rates [39]. Direct, negative effects of environmental stresses on either hosts or pathogens could cause a parabolic relationship between the magnitude or frequency of environmental stresses and disease incidence, with outbreaks occurring mainly at intermediate stress levels (Figure 3). Thus, future increases in thermal anomalies or other forms of environmental stress could decrease the probability and severity of marine epidemics. Paradoxically, management activities that increase host abundance could facilitate epidemics. Indeed, most of the major coral reef epidemics over the last 20 y have been of high-density hosts. Caribbean examples include acroporid white band disease outbreaks [71], the D. antillarum epidemic [69,70], and sea fan aspergillosis [75]. Once the density of these hosts was sufficiently diminished, prevalence often decreased [34], and host populations began to recover [76,77]. Warm temperature anomalies and coral cover are clearly important drivers of white syndrome on the GBR. No previous study has demonstrated a link between ocean temperature and coral disease dynamics, especially at regional spatial scales. Our results are supported by basic epidemiological principles, and could apply to other coral disease systems and to disease ecology in general. However, coral disease dynamics are likely to be affected by a variety of biotic and abiotic factors, the relative importance of which will vary among regions, scales, and species [32]. In some locations, coral disease outbreaks are apparently decoupled from temperature, and several other factors are also known or suspected to influence the dynamics of coral and other marine diseases [19]. For example, the severity of at least three coral diseases is linked with nutrient concentrations [25,32,75], whereas the frequency of others, like white syndrome, is greatest on remote reefs in highly oligotrophic waters [30]. Coral reefs around the world have been dramatically transformed over the last several decades as coral cover decreased and reefs became dominated by macroalgae [71,78–80]. These changes affect entire coral reef ecosystems, resulting in declines in biodiversity, fisheries yield, and other ecosystem services [81]. Our results indicate warm temperature anomalies can drive outbreaks of coral disease under conditions of high coral cover. The general increase in coral disease prevalence and the emergence of several new coral diseases over the last two decades [20,82,83] could also have been caused in part by thermal anomalies. Deciphering these and other effects of increasing temperature on disease dynamics in the ocean presents an urgent challenge to marine scientists. Surveys of white syndrome frequency and total coral cover (i. e. , the percentage of the bottom covered by living scleractinian corals) were performed by the Australian Institute of Marine Science Long-term Monitoring Program. The 48 surveyed reefs are grouped within six latitudinal sectors that span nearly 1,500 km of the GBR from 14° S to 24° S (Figure 1A). Surveys were performed annually between 1998 and 2004 using SCUBA along a depth contour of 6 to 9 m on the northeast flank of each reef. The frequency of white syndrome cases on each reef (number/1,500 m2) was measured by counting the number of infected colonies within 15 permanent 50-m × 2-m belt transects [30,84]. The percentage of the substrate covered with living, hard (scleractinian) coral tissue was quantified on 15 permanent 50-m transects, within a 25-cm-wide belt along the transect using a video camera [85]. A point sampling technique was then used to estimate live coral cover from the videos in the laboratory [85]. We derived weekly sea surface temperature values for each reef from a newly developed 4-km Advanced Very High Resolution Radiometer Pathfinder temperature anomaly dataset (Version 5. 0) developed by the National Oceanic and Atmospheric Association and the University of Miami' s Rosenstiel School of Marine and Atmospheric Science. This dataset covers the longest time period (1985–2004) at the highest resolution of any consistently processed, global satellite temperature dataset. We used nighttime, weekly-averaged values with a quality level of four or better [86]. Some plausible values were given low-quality levels by the Pathfinder algorithm, which eliminates any observation with an SST more than 2 °C different than a relatively coarse resolution SST field based on the Version 2 Reynolds Optimum Interpolation Sea Surface Temperature (OISST) value, a long-term, in situ–based dataset [86,87]. Therefore, we included observations if the SST was greater than the OISST, but less than the OISST + 5 °C. Gaps in the record caused by persistent cloudiness were filled using simple temporal interpolation to provide a complete weekly time series at each reef spanning 1985–2004. We generated a 19-y, weekly SST climatology (i. e. , a long-term record) for the 4 × 4-km grid cell that encompassed each reef. A 5-wk running mean was then used to smooth each gap-free climatology to minimize any unusual fluctuations caused by periods of limited data availability. Although thermal stress metrics have been created to predict bleaching events from satellite SST data [73,88,89], little is known about the thresholds relevant to coral disease. In general, increases of 0. 5–1. 5 °C for several weeks can induce coral bleaching [42]. We assumed that temperatures that may lead to bleaching and physiological stress in corals [42,45] could also potentially increase susceptibility to disease [7]. In a pilot study, we created 16 different metrics of thermal stress. After initial screening, Akaike Information Criteria (AIC) identified WSSTA as the metric that best explained the relationship between temperature and disease (three of the other metrics and the selection procedure are described in [37]). WSSTA quantified the frequency of high-temperature anomalies experienced by coral hosts and by the potential white syndrome pathogen (s), during the 52 wk prior to the annual disease surveys. WSSTAs represent the number of annual deviations of 1 °C or higher from a mean climatology calculated from records between 1985 and 2004 for that calendar week at that reef. Thus, the metric is both week specific and location specific, and considers deviations from local climatological averages, i. e. , typical SST throughout the year, including wintertime high-temperature anomalies that could also influence coral fitness and susceptibility to infection [46]. Because recent field and laboratory studies indicate that corals on the GBR are significantly adapted to local thermal conditions [43,90], we based WSSTA on the local SST climatology created independently for each of the 48 reefs. Furthermore, our long-term, fine-grained measurements of SST and SST anomalies match the scale of the biological surveys, eliminating the usual mismatch between climate and health data that has plagued similar studies of human and wildlife disease dynamics [91]. We used negative binomial regression to model the relationship between thermal stress and coral cover and the frequency of white syndrome cases (i. e. , the number/1,500 m2). Negative binomial regression was ideal for this analysis because the dependent variable was continuous and overdispersed (i. e. , the variance exceeds the mean). The covariates in the model included WSSTA, coral cover, and the interaction term, which represents the multiplicative relationship between coral cover and temperature. Because there is a biologically plausible mechanism by which an interaction between coral cover and temperature could affect the overall outcome (i. e. , the influence of thermal stress could be coral cover dependent), it was important to include this interaction as a covariate. A host density threshold is a common signature of infectious disease outbreaks of humans and other marine taxa such as viral diseases of seals [50,52,53]. Total coral cover or the abundance of susceptible species could both influence disease frequency and the effect of temperature on frequency. Coral cover was also included in the model to account for the potential positive relationship between cover and disease frequency based solely on the fact that the number of surveyed colonies may have increased with coral cover. Because the individual sampling units (reefs) were nested within larger groupings (sectors), this factor was included as a stratification variable to control for the main effect of variance within and between sectors. We used the general estimating equations (GEE) (i. e. , population averaged) to estimate parameters of the negative binomial model, which accounted for the repeated measurement of the individual sampling unit (reefs, each sampled once a year for six consecutive years). An autocorrelative structure was initially included; however, the parameter was sufficiently close to zero (0. 01 ± 0. 05 standard error [SE]) to consider the autocorrelative effects negligible, and thus was not included in the final model. We also calculated a deviance statistic (i. e. , deviance/degrees of freedom) to assess the goodness of fit of the model. If the model and the designated distribution are correct, this value should be approximately 1. 0. Many longitudinal datasets with continuous dependent variables are modeled using Poisson regression [33]. However, the variance structure of the related regression model, the negative binomial, includes a random dispersion term and is thus more flexible and appropriate in assessing the relationship between the covariates and an overdispersed dependent variable [92,93]. We did run a Poisson regression model, and the deviance statistic was 24. 4203, indicating a poor fit to the data. Zero inflation, that is, the possibility of the existence of a population of hosts for which the outcome cannot happen (e. g. , reefs with no susceptible individuals), was also of potential concern. To address this issue, we fit zero-inflated negative binomial and Poisson regression models. There was no difference in parameter estimates from the standard models; thus, the simplest negative binomial model was used in the final analysis. All regression analyses were conducted using Intercooled Stata 9. 1 (Stata Corporation, http: //www. stata. com).

Coral reefs have been decimated over the last several decades. The global decline of reef-building corals is of particular concern. Infectious diseases are thought to be key to this mass coral mortality, and many reef ecologists suspect that anomalously high ocean temperatures contribute to the increased incidence and severity of disease outbreaks. This hypothesis is supported by local observations—for example, that some coral diseases become more prevalent in the summertime—but it has never been tested at large spatial scales or over relatively long periods. We tested the temperature–disease hypothesis by combining 6 years of survey data from reefs across 1,500 kilometers of Australia' s Great Barrier Reef with a new ocean temperature database derived from satellite measurements. Our results indicate that major outbreaks of the coral disease white syndrome only occurred on reefs with high coral cover after especially warm years. The disease was usually absent on cooler, low-cover reefs. Our results suggest that climate change could be increasing the severity of disease in the ocean, leading to a decline in the health of marine ecosystems and the loss of the resources and services humans derive from them.

Abstract Introduction Results Discussion Materials and Methods

corals infectious diseases public health and epidemiology cnidaria (jellyfish ecology etc.) hydra

2007

Thermal Stress and Coral Cover as Drivers of Coral Disease Outbreaks

The current rabies control strategy in Zambia is based on dog vaccination, dog population control and dog movement restrictions. In Nyimba district of Zambia, dog vaccination coverage is low but the incidence of dog bites is high which places the community at risk of rabies infection. The renewed global interest eliminating rabies in developing countries has spurred interest to identify determinants and barriers of dog vaccination in an effort to reduce the overall disease burden. A mixed methods cross sectional design was used in the study. This consisted of three parts: Evaluation of medical records regarding dog bite injuries, implementation and analysis of a household survey and in-depth review of key informant interviews. Data was collected into a Microsoft Excel database and subsequently transferred to STATA for descriptive, inferential and thematic analysis. Dog vaccination coverage overall was 8. 7% (57/655), with 3. 4% (22/655) in urban areas, 1. 8% (12/655) in peri-urban and 3. 5 (23/655) in the rural regions. Financially stable households were more likely to have their dogs vaccinated. Only 10. 3% (31/300) of the respondents had vaccinated their dogs and these had a reliable source of income as 6% (18/300) were peasant farmers, 2% (6/300) were dependants whose guardians were financially stable and 2. 3% (7/300) were in steady employment. Important barriers to dog vaccination included cost, limited awareness of vaccination program and access. Current rabies control strategies in Nyimba district, Zambia, appear quite limited. Improvements in the regional dog vaccination program may provide benefits. Enhancement of educational efforts targeting behavioural factors may also prove useful. Finally, the cost of dog vaccination can be reduced with scaled up production of a local vaccine. Rabies has been a public health concern and has plagued both humans and animals since around 2000 BC [1]. The disease is endemic on all continents except Antarctica [2] and it is believed to cause approximately 59,000 human deaths annually [3]. Rabies is more prevalent in developing countries where management and control measures are poor; consequently, continents such as Asia and Africa have the highest incidence of rabies, accounting for over 95% of the global rabies cases [4]. Rabies can be transmitted between all warm-blooded species including man and studies have shown that several domestic and wild animals such as dogs, cats, cattle, wolves, foxes, jackals, bats and others can get infected with the rabies virus and transmit the disease to humans via bites or scratches. Human rabies is mostly due to dog-transmitted rabies virus (RABV) [5] which is an RNA virus of the Rhabdoviridae family genus Lyssavirus [6]. Following invasion of the central nervous system, rabies infection progresses rapidly [7] and death due to respiratory failure or cardiac arrest ensues [8]. While a very small number of patients with rabies have survived, the disease is untreatable and fatal once signs of encephalitis appear [9]. Fortunately, rabies post exposure prophylaxis (PEP) prevents rabies in humans exposed to the rabies virus [10]. Thus PEP is the cornerstone for rabies prevention in humans, and it is against this background that it is recommended for all persons that have been or suspected to have been exposed to the rabies virus. In developing countries, the domestic dog has been found to be responsible for the transmission of most of the human rabies cases [3] with over 90% of the rabies cases being transmitted via dog bites [11]. It is estimated that owned dogs account for the majority of the hundreds of millions of people that are bitten by dogs in the world each year [2]. Cases of rabies transmission between humans through transplant surgery have been reported but these are very rare [12]. The control of rabies since 1973 as recommended by the World Health Organisation (WHO) includes; mass dog vaccination campaigns and strict dog population control via restricted breeding, restricted movements and culling of unwanted dogs especially stray dogs [13]. Although these control measures have been in place for about 50 years, studies have shown that only a few developed countries are currently rabies free [14]. The vaccination of dogs against rabies is now regarded as the most effective rabies control strategy combined with secondary roles of population control, movement regulations and promotion of responsible dog ownership [11] and [15]. In developing countries were the prevalence of rabies is still significantly high; dog vaccination is a challenge. A number of successful dog vaccination campaigns have been carried out and research has demonstrated that rabies in these countries can be controlled [16]. Studies have further shown that the common assumption that dog vaccination in developing countries is hindered by operational constraints such as lack of dog population knowledge, low public rabies knowledge and inadequate implementation resources, may be erroneous [17]. Dog vaccination may not be a priority in some developing countries because of the limited resources available [18]. In Zambia, rabies is regarded as one of the endemic scheduled or notifiable diseases and the law under the Zambian Animal Health (Control and Prevention of Animal diseases) Order of 2014 stipulates that" animal owners vaccinate their animals against all scheduled or notifiable diseases" [19]. The WHO recommends that vaccinating at least 70% of the dog population against rabies over consecutive years may interrupt rabies transmission chains amongst dogs [20]. It has been found that vaccination coverage lower than 30% of the dog population is a waste of resources [11]. In Zambia, the actual dog population is not well known but it is widely assumed that only a small percentage of the Zambian dog population is vaccinated against rabies. In Nyimba district for instance, the dog population estimate is based on the 2006 Livestock Census. According to the Zambian National Livestock Epidemiologic Information Centre (NALEIC) and District Veterinary records, only 5. 4% (138/2,556) and 5. 6% (157/2,804) of the estimated dog population were vaccinated against rabies in Nyimba district in 2013 and 2014 respectively. Despite the estimated low number of vaccinated dogs in the country, the number of notified dog bite cases has continued to rise. According to the Zambian report on rabies presented at the Southern and East African Rabies Group (SEARG) meeting of 2013; the number of notified dog bite cases in Zambia rose from 620 in 2010 to 732 in 2011. Veterinary records in Nyimba district also show that there has been a steady increase in the number of notified dog bite cases from 84 recorded in 2013 to 134 cases recorded in 2014. The rise in dog bite cases has led to an increase in the demand for PEP in the district as most of the victims bitten by unvaccinated dogs require prophylaxis. The WHO estimates that the global annual cost of PEP to be around $1. 6 billion [2]. This makes the use of PEP more expensive than simply vaccinating dogs against rabies. Although a number of mass dog vaccination campaigns have been carried out in the district, the coverage has been very poor and the majority of the dog population remains unvaccinated. Thus both the dog and human population are at risk of rabies infection. An analysis of suspected rabies cases recorded in Zambia between 1985 to 2004 found 1,088 rabies positive samples from various species, 747 of which were from dogs and 98 were from humans (Table 1) [21]. Another analysis of brain samples collected from suspected rabid dogs between January 2005 and December 2013 found 153 rabies positive cases [22]. Thus the dog mediated human rabies and dog rabies burden in the district and the country at large is still a challenge. The aim of the study was to identify the socio demographic determinants which influence dog vaccination and the local barriers to dog vaccination against rabies. The study tried to explore the perceptions and responses of the dog-owning households in relation to rabies control in Nyimba district. It was hoped that gaining an understanding of the various social norms prevailing in the communities in relation to rabies control would help in the tailoring of vaccination campaigns which would result in a wider coverage. The study was conducted in Nyimba district which is one of the nine districts of Eastern Province in Zambia (please follow this link: https: //en. wikipedia. org/wiki/Eastern_Province, _Zambia#/media/File: Zambia_Eastern_Province_Districts. svg) [23]. The district lies on the southern part of the province and covers a total area of 10,509Km2 which is divided into valley and plateau areas. Nyimba district is situated approximately 340 Km from Lusaka the capital city of Zambia and 230 Km from Chipata the provincial centre of Eastern province. According to the 2010 national population and housing census, the district has a human population of 85,025 with 60% of the population living in the plateau and 40% in the valley. A number of ethnic groups found in the district including the Nsenga, Chewa, Ngoni and Tumbuka with the Nsenga being the indigenous group. Agriculture is the main source of livelihood in terms of crop and livestock farming for about 90% of the population while the rest depend on fishing, casual labour and the civil service. The research was conducted using a mixed methods (qualitative and quantitative) cross sectional design which was divided into three parts (Fig 1). The first part consisted of evaluating dog bite case records at the veterinary office and Nyimba district Hospital to determine the frequency of dog bites and the vaccination and ownership statuses of the dogs involved in the bites. The proportion of reported dog bite cases which received rabies post exposure prophylaxis (PEP) from 2010 to 2013 was 12. 3% (239/1,947). The postulated proportion of reported dog bite cases which received rabies PEP between 2013 and 2015 was hypothesized to be 5%. Calculations using statistical software at 0. 05 significance level and 80% power showed that the required sample size was 130 reported dog bite cases. There were 215 dog bite cases recorded from 2013 to 2015 and they were all included in the study. The second part was a survey of 300 households which responded to a household questionnaire which collected data on socio demographics, community knowledge with regards to rabies and data on dog population and vaccination coverage. The household survey sample was calculated using statistical software at 0. 05 significance and 80% power. According to literature from the Central Veterinary Research Institute (CVRI) in Zambia, the relative prevalence of rabies in the country was hypothesized to be 39. 7% and postulated to be 48%. The resulting sample size was 277 households which were rounded off to 300 households. The households were selected using a cluster randomized sampling method and the questionnaire was administered by the research assistants. The third part included in depth interviews with local rabies experts including the rabies control officers from the Council and Veterinary Department of Nyimba district and the Central Veterinary Research Institute in Lusaka. The in depth interviews were conducted by the researcher. For the in depth interviews with local rabies experts and the dog bite case record evaluation, verbal informed consent was obtained from the informers at both the hospital and the veterinary office. Written informed consent was obtained from respondents of the household questionnaire. Ethical clearance was obtained from Excellence in Research Ethics and Science (ERES) converge (Ref. No. 2015-June-018). All data collection was conducted in English except for the household questionnaire which was done in Nsenga in some cases and then translated into English. Questionnaire and dog bite case evaluation data were entered into Microsoft Office Excel 2007 and coded then transferred to STATA version 12 for analysis. Data from the in depth interviews with local rabies experts was entered into Microsoft Office Word 2007 and thematic analysis was done manually. The evaluation of the dog bite case records found that there were 215 reported dog bite cases recorded between January, 2013 and January, 2015. Thus out of these 46% (99/215) were female and 53. 95% (116/215) were male. Information on age was missing for 45% (98/215) of the cases but in 117 cases, the victims were aged between 1 and 68 years with a mean age of 16. 8 years (SD 14. 7). Approximately 62. 4% (73/117) of the dog bite victims were aged between 1–15 years, 23. 1% (27/117) were aged between 16–30 years and 14. 5% (17/117) were over 30 years of age. The household survey was conducted in order to determine a more accurate estimate of the dog population and vaccination coverage and also to determine community knowledge levels and responses to rabies and its control strategies. A total of 300 households consisting of 1,970 people completed the household questionnaire. The average number of people per household was 6. 57 and occupants ranged from 1 to 15 people in the households. Males made up 74% (223/300) of the respondents and 26% (77/300) were female. Two of the respondents were aged 14 years and they answered the questionnaire in homes where adults were not present at the time of the interview, the other 298 respondents were above 14 years of age (range 14–83 years, mean age = 38. 6 years). The spatial distribution of the 1,970 people covered by the survey was such that 78 (3. 9%) were from the urban areas, 106 (5. 4%) were from the peri-urban areas and 1,786 (90. 6%) were from the rural areas of the district. As a means of livelihood it was found that peasant farming was practiced by 262 (87. 3%) of the respondents while 15 (5%) were formerly employed and 23 (7. 7%) were dependents. The results showed that of the 300 respondents, 49 (16. 3%) did not have any formal education, 169 (56. 3%) had Primary education, 78 (26%) had Secondary education and only 4 (1. 3%) had Tertiary education. The control of rabies in Nyimba district remains challenging. Our study showed that compliance with dog control and dog vaccination against rabies was low. Certainly, there were a number of significant social and economic barriers identified. We believe that interventions to reduce the impact of rabies need to include the local populations in order to maximise benefit. Dog vaccination rates could be improved with sustained efforts at improving dog behaviours, reducing reproduction rates, and increasing access to low cost vaccine. Educational efforts still have room for improvement as well. This study helped to elucidate some of the areas important for targeting efforts to reduce the morbidity of dog bites and the danger of rabies in this region.

Dog vaccination against rabies is the main method of rabies control in Zambia and it is mainly conducted by the Department of Veterinary Services under the Ministry of Fisheries and Livestock. This study explores the factors influencing dog vaccination against rabies and the local barriers to rabies control in Nyimba district of Zambia. The study was done in three parts; a review of dog bite cases (n = 215 cases), a household survey and questionnaire (n = 300 households) and in depth interviews with local rabies experts (n = 5). The study found that the community had adequate knowledge about rabies prevention in dogs and humans. The area of residence, the age and financial capacity of the dog owner determined whether or not the household dog was vaccinated or not. The survey of the 300 households revealed that only 8. 7% of the dogs were vaccinated against rabies and the dog bite case record review showed that the majority of the dog bite cases were caused by unvaccinated dogs most of which were owned. The study discusses the challenges in controlling rabies and possible reasons for non compliance to dog vaccination by the dog owners in Nyimba district.

Abstract Introduction Methods Results Discussion

animal types medicine and health sciences pathology and laboratory medicine pathogens immunology tropical diseases geographical locations vertebrates microbiology pets and companion animals dogs animals mammals viruses preventive medicine rabies rna viruses rural areas zambia neglected tropical diseases vaccination and immunization zoology africa veterinary science rabies virus public and occupational health infectious diseases geography veterinary diseases zoonoses medical microbiology microbial pathogens people and places lyssavirus eukaryota viral pathogens earth sciences geographic areas biology and life sciences viral diseases amniotes organisms

2017

Insights and efforts to control rabies in Zambia: Evaluation of determinants and barriers to dog vaccination in Nyimba district

Root-knot nematodes secrete effectors that manipulate their host plant cells so that the nematode can successfully establish feeding sites and complete its lifecycle. The root-knot nematode feeding structures, their “giant cells, ” undergo extensive cytoskeletal remodeling. Previous cytological studies have shown the cytoplasmic actin within the feeding sites looks diffuse. In an effort to study root-knot nematode effectors that are involved in giant cell organogenesis, we have identified a nematode effector called MiPFN3 (Meloidogyne incognita Profilin 3). MiPFN3 is transcriptionally up-regulated in the juvenile stage of the nematode. In situ hybridization experiments showed that MiPFN3 transcribed in the nematode subventral glands, where it can be secreted by the nematode stylet into the plant. Moreover, Arabidopsis plants that heterologously expressed MiPFN3 were more susceptible to root-knot nematodes, indicating that MiPFN3 promotes nematode parasitism. Since profilin proteins can bind and sequester actin monomers, we investigated the function of MiPFN3 in relation to actin. Our results show that MiPFN3 suppressed the aberrant plant growth phenotype caused by the misexpression of reproductive actin (AtACT1) in transgenic plants. In addition, it disrupted actin polymerization in an in vitro assay, and it reduced the filamentous actin network when expressed in Arabidopsis protoplasts. Over a decade ago, cytological studies showed that the cytoplasmic actin within nematode giant cells looked fragmented. Here we provide the first evidence that the nematode is secreting an effector that has significant, direct effects on the plant’s actin cytoskeleton. Root-knot nematodes (Meloidogyne spp) are small endoparasites with a host range that includes most flowering plants [1]. During the compatible interaction, the motile second stage juveniles (J2) enter the host roots and migrate intercellularly to the plant vasculature [2]. The primary nematode secretory organs, the esophageal glands, produce secretions. The gland secretions are exuded through the nematode stylet into the plant. Secretions that promote nematode parasitism are called effectors [3,4]. Effectors help nematodes with successful parasitism by altering host defenses and/or modifying the plant cells to form the nematode feeding sites [3–5]. The feeding sites are comprised of 2–10 host cells (typically 6) that are reprogrammed to form large, multinucleate “giant cells” [6]. Root-knot nematodes are completely reliant on their giant cells as their sole source for food, and thus, the giant cells must be maintained throughout the nematode’s life for its survival. As part of nematode effector research, several groups have worked to identify secreted root-knot nematode proteins and the plant processes that they affect [7–20]. Work from Bellafiore et al (2008) directly identified secreted proteins from Meloidogyne incognita by exposing the nematodes to root exudates before treating them with resorcinol to induce esophageal gland secretion [8]. Using sensitive mass spectrometry methods, 486 proteins were identified in the M. incognita “secretome” [8]. Although the proteins were categorized based on bioinformatic analyses, the roles of these potentially secreted proteins in plant-nematode interactions have been largely unexplored. Recently, Lin et al. (2016) used the secretome to identify a transthyretin-like protein 5 (TTL5). They showed that M. javanica TTL5 homolog plays an important role in promoting parasitism by suppressing the host’s basal defense responses [21]. This work highlights the utility of studying the nematode secretome for identifying important nematode effectors. Previous reports have used chemical and genetic experiments to show that actin cytoskeletal rearrangements are necessary for giant cell development [22,23]. Therefore, we investigated it the nematodes were secreting effector (s) into the plant cells that were directly targeting actin filaments. In an effort to identify these effectors, we searched the published secretome for proteins that could interact with actin. Of the 486 peptides identified in the M. incognita secretome, 33 fell into the category “cell shape, ” and predicted to interact with actin or microtubules [8]. Of these 33 proteins, two were annotated as profilins (PFNs): Proteins #131 and #240 [8]. Profilins are small actin binding proteins found in all eukaryotes whose main function is to bind globular (G) actin [24,25]. Profilin can also bind to barbed ends of actin filaments, albeit with lower affinity than to G actin [26,27]. In addition to binding to actin, profilin can also bind to polyphosphoinositide molecules, Arp2/3 complex, annexin, and proline-rich ligands [28–32]. Plant profilins do not share high amino acid similarity with profilins from other organisms, but they can complement profilin mutants in yeast and Dictyostelium, suggesting profilins have conserved functional roles across kingdoms [33,34]. By focusing on nematode profilins found in the previously published secretome, we discovered that a nematode profilin, called MiPFN3, is an effector that facilitates parasitism. Our work shows that MiPFN3 expression in plant cells causes a disruption to plant actin filaments. We were interested in specifically studying the two peptides found in the M. incognita secretome that were homologous C. elegans profilins (pfam00235): M. incognita #131 (Mi131) and #240 (Mi240) [8]. Using the peptide sequence for Mi131, a tBLASTn search of the Expressed Sequence Tag (est) database identified six ESTs with 100% identity to the Mi131 sequence (Genbank sequence IDs: JK291082. 1, JK291081. 1, JK267125. 1, JK298994. 1, JK303909. 1, JK306024. 1). These sequences contain an open reading frame of 381 bp that encodes a protein of 126 aa (S1 Fig). The protein sequence is 100% identical to the Mi131 peptide sequence, and it contains a profilin domain (pfam00235) (S1 Fig). When this protein sequence was used in a BLASTp search of C. elegans Sequencing Consortium genome project, the best hit was to C. elegans Profilin 3 (CePFN3), with 63. 5% amino acid identity (Figs 1A and S2). In NCBI BLASTp search of Genbank’s non-redundant (nr) database with Mi131, the top four hits were to profilin 3 in parasitic and free-living nematodes: Trichinella spiralis PFN3,65. 6% identity; C. elegans PFN3,63. 5% aa identity; Caenorhabditis remanei, -PFN-3,63. 5% identity, and Caenorhabditis brenneri -PFN-3,63. 5% identity. Based on this sequence similarity to profilin 3 in C. elegans and other nematodes (Fig 1A), we refer to the gene as MiPFN3. According to Bellafiore et al. (2008), Mi240 had a BLASTp top hit to C. elegans Profilin 1 (CePFN1). Using this information, we found significant tBLASTn hit in the M. incognita genome to one predicted full-length M. incognita gene, Minc11290, which encodes a 133 aa protein. Using primers based on this sequence, we amplified a 402 bp sequence. When this 402 bp sequence was used in a nucleotide blast search of the NCBI EST database, it was 100% identical to four M. incognita ESTs (Genbank Sequence IDs: JK295768. 1, CF802658. 1, CF802625. 1, CK984306. 1). This sequence encodes a 133 aa protein that is 95. 5% identical to the protein encoded by Minc11290 (Fig 1B). In addition, a BLASTp search using the non-redundant (nr) database showed that the protein encoded by our amplified sequence had highest sequence homology to the hookworm profilin [Necator americanus, 70% aa identity]. There were also hits to Toxocara canis profilin [TC-PFN1 70% identity], Caenorhabditis remanei profilin, [CRE-PFN-1 protein 64% aa identity], and Caenorhabditis brenneri profilin [CBN-PFN-1 63% aa identity] (Figs 1B and S2). Based on this sequence homology to PFN1 in parasitic and free-living nematodes, we refer to the Mi240 protein as MiPFN1. Because genes involved in pathogenicity may be upregulated during the parasitic life stages of the nematode, we asked if either gene was up-regulated during nematode developmental life-stages associated with parasitism. Using primers specific for MiPFN1 and MiPFN3, we measured the genes’ transcript levels in eggs, second stage juveniles (J2), and in nematodes-infected tomato roots at 7,14, and 21 days post-inoculation, which represent the parasitic life-stages. The expression at the egg stage was set to 1 and used to calculate the fold change of the expression at the other time points. The expression of MiPFN1 was not differentially expressed in the egg, J2 and in the parasitic nematodes life stages (Fig 2A). In contrast, MiPFN3 expression was up-regulated in the J2 compared to the egg stage. In a later developmental life-stage (7 dpi), the MiPFN3 expression level was equivalent to its expression in eggs (Fig 2B). Expression then decreased in nematodes in the 14 and 21 dpi samples. To study earlier time points of nematode infection, we collected infected roots from Arabidopsis grown on MS-media at 4 dpi. We also collected infected roots at 7,14, and 42 dpi. Acid fuchsin staining of the infected roots showed that at 4 dpi, the J2s penetrated the roots and were migrating as parasitic J2 (S3 Fig). At 7 dpi, visible galls had formed, but the parasitic nematodes still look like slim, non-feeding J2. By 14 dpi, we noticed that some nematodes had begun to looked fatter, an indication that the nematodes had initiated feeding. By 42 dpi, the nematode females had laid eggs in a gelatinous matrix on the surface of the root (S3 Fig). The expression of MiPFN1 and MiPFN3 was measured in the eggs, J2, and in the parasitic nematodes (4,7, 14,42 dpi). The expression at the egg stage was set to 1 and used to calculate the fold change of the expression at the other time points. Overall, MiPFN1 was not differentially expressed at any time point (Fig 2C). However, MiPFN3 expression was significantly up-regulated compared to the egg stage in the J2 and early parasitic stages (4 and 7 dpi) (Fig 2D). Because MiPFN3 was strongly expressed in the J2, the infective stage of the nematode, and during early parasitic stages, it may be playing a role in the initial stages of plant parasitism, and therefore, we focused on MiPFN3 for further characterization. To determine where the MiPFN3 transcript is expressed in the J2, we performed in situ hybridization using a DIG-labelled antisense MiPFN3 cDNA probe on fixed juveniles. The probe hybridized to the esophageal glands (Fig 3A and 3B). The sense probe for MiPFN3 did not hybridize to the nematodes (Fig 3C and 3D). Thus, we found MiPFN3 transcripts present in a nematode secretory organ, which indicates that it encodes a protein secreted by M. incognita. We next wanted to investigate if MiPFN3 has a role in nematode parasitism. Arabidopsis Col-0 was transformed with full-length MiPFN3 driven by the Cauliflower Mosaic Virus 35S promoter. We obtained two T2 lines for plants with a dwarf phenotype, in which the rosettes were significantly smaller than Col-0 (lines G and M) (Fig 4). However, we were also able to generate two homozygous transgenic lines with single transgene insertion (MiPFN3 B. 2 and I. 3) and which did not exhibit size defects (Fig 4). The size of the plants corresponded to the amount of MiPFN3 transcript measured in the plants, with lines G and M having significantly higher levels of MiPFN3 than plants from lines B. 2 and I. 3, which had wild-type root growth (S4 Fig) and rosette sizes (Fig 4). Because the dwarf plants had smaller roots that can affect the number of nematode infection sites, we only tested lines that had wild-type root length phenotype in the root-knot nematode infection assays. Therefore, the T3 generation of MiPFN3 lines B. 2 and I. 3 were infected with M. incognita. Both independent transgenic lines showed increased levels of galling (Figs 5A and S4). The size of the galls in B. 2 and I. 3 were not significantly different to the control (Figs 5B and S5). There also was no obvious qualitative difference in the giant cells formed in the wild-type and transgenic lines (S6 Fig). Overall, the expression of MiPFN3 in the plant leads to increased number of galls, suggesting that MiPFN3 promotes nematode infections. A BLASTp search against the five profilin proteins in Arabidopsis showed that MiPFN3 had highest homology to Arabidopsis Profilin 4 AtPRF4 (38% amino acid identity) (S7 Fig). AtPRF4 can bind to actin monomers, and is specifically expressed in reproductive tissues (pollen and flowers) [35]. Based on its organ-specific expression pattern, AtPRF4 forms a complex with actin monomers that are also expressed in the plant’s reproductive organs (ACT1, ACT3, ACT4, ACT11and ACT12) [36]. A previous report showed that mis-expression of the reproductive actin ACT1 leads to an aberrant actin architecture in the plant, causing severe dwarfing of the plants. However, co-expression of AtPRF4 in these plants could suppress the ACT1-mediated dwarf phenotype [37]. Because MiPFN3 has similarity to AtPRFN4, we investigated whether MiPFN3 could also suppress the ACT1-induced dwarf phenotype. Arabidopsis Col-0 and two transgenic MiPFN3 lines (B. 2 and I. 3) were transformed with 35S: : AtACT1. When the T1 seedlings started to produce inflorescences, the rosette size and the leaf morphology were graded into three categories: 1) small rosette (dwarf), 2) intermediate rosette size and 3) wild-type-like rosette size. When AtACT1 was ectopically expressed in Col-0, approximately 30% of the T1 population exhibited a dwarf phenotype (Fig 6). When the transgenic MiPFN3 lines were transformed with 35S: : AtACT1, none of the T1 plants exhibited a small rosette (Fig 6), indicating that MiPFN3 could suppress the AtACT1-induced dwarf phenotype. To clarify the effects of MiPFN3 on actin in more detail, in vitro actin sedimentation assays were performed using non-muscle actin and recombinant His-tagged MiPFN3. The His- MiPFN3 was added to soluble G actin prior to actin polymerization. After actin polymerization and sedimentation by centrifugation, we measured the ratio of soluble G actin to filamentous (F) -actin. When buffer or BSA (bovine serum albumin) were added to the G actin before polymerization, there was significantly more actin in the pellet fraction compared to the supernatant, indicating that most of the G actin had polymerized into F actin. On the other hand, when G actin was incubated prior to actin polymerization with purified recombinant MiPFN3, there was an increase in the G actin observed in the supernatant fraction after ultracentrifugation (Fig 7). In other words, relatively less F-actin polymerized if the actin monomers were incubated with MiPFN3 prior to polymerization. To study the effects MiPFN3 expression on the actin cytoskeleton in plant cells, we expressed MiPFN3 in Arabidopsis leaf protoplasts constitutively expressing 35S: : ABD2-GFP. The 35S: : ABD2-GFP construct encodes a fibrin protein fused to GFP, and it can bind to and fluorescently label actin filaments [38]. We found that protoplasts expressing 35S: : RFP-MiPFN3 showed disrupted actin filaments and reduced the visible levels of ABD2-GFP compared to untransformed ABD2-GFP protoplasts, which showed dense, GFP-labeled F-actin (Fig 8). To determine whether the effects on the actin cytoskeleton was specific for MiPFN3 expression, the M. incognita Peptidyl-prolyl cis-trans isomerase (Minc06346) [39] was expressed in the ABD2-GFP protoplasts. The peptide corresponding to Minc06346 coding sequence was present in the M. incognita secretome [8], and the protein does not contain any predicted actin binding domains (Interpro Scan). When we transiently expressed full length cDNA Minc06346 driven by the 35S promoter fused at the N-terminus to RFP in the ABD2-GFP leaf protoplasts, the RFP fusion protein could be detected, and these cells had fluorescently labeled, dense actin filaments, similar to the un-transformed control (Fig 8). Therefore, MiPFN3 expression in the protoplasts can specifically affect the organization and structure of the actin filaments. We have identified a nematode profilin gene called MiPFN3 as a novel nematode effector that is up-regulated in expression during parasitic life stages and when expressed in plants, it enhanced plant susceptibility to nematodes. MiPFN3 has homology to C. elegans profilin 3 (CePFN3) (Fig 1A). There are three profilin genes in the C. elegans genome. All three profilins are expressed in the worm and all behave as classical actin binding proteins [40]. In the CePFN1 knockdown, cytokinesis of embryonic cells was affected [41], but the gene knockouts of CePFN2 and CePFN3 did not have phenotypes, indicating that these genes are non-essential [40]. The profilins in other free living and parasitic nematodes have not been characterized. Interestingly, the parasite Toxoplasma gondii possesses a profilin-like protein that is released by the parasite to facilitate its invasion of host cells [42,43]. The profilin-like protein can bind actin, but it has also evolved a role at the host-parasite interface [42]. MiPFN3 was originally found in the root-knot nematode secretome [8]. Surprisingly, MiPFN3 lacks a canonical secretion signal sequence. A root-knot nematode effector protein devoid of canonical secretion signal is not without precedent, and there are several examples of root-knot nematodes effectors, such as MI-14-3-3 and Mi-GSTS-1, which do not have canonical signal peptides but play key roles in plant-nematode interactions [20,44–46]. The secretion of these proteins may be though a non-canonical secretory pathway that functions independently of the endoplasmic reticulum -Golgi network [20,45,46]. MiPFN3 may play a role in the secretome, and is expressed in the nematode esophageal glands, indicating that MiPFN3 is secreted from the nematode. Since these glands are connected by the esophagus to the stylet, it is possible for MiPFN3 to be delivered through the stylet directly into the plant [5]. In an effort to functionally characterize this profilin from root-knot nematodes, we performed in vitro actin polymerization assays. In these assays, actin-binding proteins bind and sequester actin monomers, preventing them from polymerizing into actin filaments. We found that pre-incubation of actin monomers with purified MiPFN3 prevented the formation of new actin filaments. Our results suggest that MiPFN3 sequesters G actin and inhibits nucleation of actin polymers in vitro. Profilins are found in all eukaryotes, including plants [47]. Arabidopsis has five profilin genes that can be divided into two groups based on the tissues in which they are expressed: vegetative and reproductive [33,37]. MiPFN3 has highest identity (38%) to AtPFN4 (At4g29340), a profilin expressed in plant reproductive organs. A previous report showed that the mis-expression of the reproductive actin AtACT1 in plants caused dwarf plants, but overexpression of AtPRF4 in these plants suppressed this phenotype [37]. By binding and sequestering the AtACT1 monomers, AtPRF4 prevented the deleterious effects on cytoskeletal architecture caused by reproductive actin misexpression [37,48]. Because a root-knot nematode profilin that is secreted into the plant and may functionally mimic plant profilins, we tested MiPFN3 for its ability to suppress the ACT1-induced dwarf phenotype. Expression of MiPFN3 in the 35S: : ACT1 plants resulted in plants with wild-type rosette size. Because MiPFN3 could suppress the dwarf morphological phenotype caused by ACT1 mis-expression, we conclude that MiPFN3 could bind to ACT1 in planta and sequester the excess reproductive actin monomers to a level that allowed for normal growth and development. When MiPFN3 was expressed at high levels in Col-0 (wild type), the resulting transgenic lines (G and M) showed growth defects (Fig 4). This indicates that MiPFN3 can also bind to the vegetative class of actin, which is found in vegetative organs such as leaves and roots. Overall, this data indicate MiPFN3 binds to both reproductive and vegetative classes of actin; there is no actin-class specific interaction between MiPFN3 and actin monomers. We also wanted to investigate the effects of a nematode profilin on the actin filament dynamics of the plant. When MiPFN3 was expressed in protoplasts, the GFP-labeled actin filaments appeared fragmented. In plant cells that were not transfected or transfected with another nematode gene, the actin cytoskeleton appeared intact. Thus, 35S: : MiPFN3 expression in Arabidopsis protoplasts affected the plant actin filaments. The fragmented actin phenotype was similar to the actin phenotype of Tradescantia stamen hair cells injected with birch profilin. The excess of birch profilin led to actin filament depolymerization and a reduction of actin microfilamants in the stamen hair cells [49,50]. It was proposed that injected birch profilin sequestrated monomeric actin, leading to an inhibition actin polymerization and a depletion of F actin. Paradoxically, later studies showed that profilin facilitates actin polymerization by interacting with cytoskeletal proteins like formin and promoting the turnover of actin monomers. Profilin was also shown to promote actin filament assembly at the barbed-ends, competing with barbed end regulators and filament branching machinery [25,27,51,52]. Thus, profilins regulate actin homeostasis through its roles in actin polymerization and depolymerization. The effect on actin depends on the concentration of profilin and other actin binding proteins [53]. The strong expression of MiPFN3 in the protoplasts may be disrupting the balance of profilin in the cell, and this leads to aberrant actin filaments. The transgenic lines with the highest level of MiPFN3 expression showed aberrant, small rosettes, suggesting that the quantity of MiPFN3 correlates with developmental defects [54,55]. The two transgenic lines that had the lowest levels of MiPFN3 transcript (B. 2 and I. 3) had roots that looked similar to wild type plants. Interestingly, although the lines B. 2 and I. 3 showed increased galling compared to the control, the overall gall phenotypes morphologies were not obviously different. Thus, MiPFN3 has no role in gall expansion. Based on the expression of MiPFN3 in the nematode during early parasitism, we postulate that MiPFN3 facilitates early infection and feeding processes that lead to a higher percentage of nematodes that are successful in forming galls. One possibility is that MiPFN3 plays a role in facilitating multinucleate giant cell formation and possibly maintenance. Previous cytological work showed that phragmoplasts, which act as scaffolding to support the newly formed cell wall between divided nuclei, are disordered and do not fully develop in giant cells [56–58]. The malformed phragmoplast results in aborted cell division and this leads to multinucleate giant cells [22,23,56,59]. Work looking at the actin filaments associated with the phragmoplasts in giant cells showed that these actin filaments are disorganized [22]. Because the nematode effector MiPFN3 is linked to actin reorganization, MiPFN3 may be injected into the plant cell to play a role in the phragmoplast failure causing a blockage of cytokinesis in giant cells. Interestingly, cross sections of galls showed that giant cells in the wild-type and transgenic (B. 2 and I. 3) lines did not exhibit any obvious phenotypic differences (S6 Fig). The transgenic lines had relatively low levels of MiPFN3 expression, and the lack of any obvious giant cell irregularities may reflect the delicate balance between the level of MiPFN3 and actin monomers in the infected-transgenic plants. Our data showed that high levels of MiPFN3 could tip the balance, leading to abnormal plant phenotypes. For example, high levels of MiPFN3 in two transgenic lines (G and M) resulted in stunted growth, and strongly expressing MiPFN3 in protoplasts affected the actin filaments. The MiPFN3 in the transgenic lines B. 2 and I. 3 did not have a negative effect on galls size or giant cell phenotype, and MiPFN3 may help nematodes to establish giant cells so that a higher percentage of infective juveniles are successful in infections and making galls. In giant cells, the appearance of fragmented cytoplasmic actin filaments has been shown to be accompanied by transcriptional activation of actin and actin-related genes [22,60–62]. Two representative of the actin gene family, ACT2 and ACT7, are transcriptionally upregulated during giant cell development [22]. The up-regulation of these genes may be in response to wounding by nematode feeding or it may be suggestive that a pool of G-actin is necessary in feeding cells [22,59]. In Arabidopsis, there is also an upregulation of formins (AtFH1, AtFH6 and AtFH10), which are involved in actin remodeling [56,60,61,63]. The actin dynamics in the giant cells have also been linked with an increase in Arabidopsis actin-depolymerizing factor (ADF) gene expression. ADF/cofilins sever actin filaments and increase the rate at which actin monomers fall off the pointed end of the actin filaments [64]. Recently, specific ADF genes were also shown to be up-regulated in M. incognita infected cucumber (Cucumis sativus L) roots. The up-regulation of specific cucumber ADF genes correlated with the changes in plant actin structure that occurred during root-knot nematode infection [65]. Since ADFs can be involved in severing and depolymerizing actin filaments from their pointed ends, the increase ADF family gene expression in giant cells may be related to the fragmentation of the actin filaments that is observed in the feeding sites. In Arabidopsis, AtADF2 RNAi knockdowns exhibited an accumulation of actin bundles, and in these plants, feeding cell expansion was inhibited [60], indicating the important role of ADFs for nematode feeding site development. Considering the roles of ADFs and profilins, it may be possible that the nematode is enhancing the expression of endogenous ADFs to increase the pool of ADP–G-actin that can bind to MiPFN3. This data suggest that diminishing the actin network density is important for facilitating nematode feeding. Up to now, the data has shown that the expression of plant genes, such those encoding formins and actin depolymerizing factors, can affect the organization of the actin filaments in giant cells. Because MiPFN3 is found in the secretome [8] and the transcript localizes to the esophageal glands, the protein is likely secreted into the plant. Here we have shown that a presumably secreted MiPFN3 can bind actin monomers to manipulate plant actin in conjunction with changes in plant gene expression. Meloidogyne incognita (Morelos) was used in all experiments. To collect nematode eggs, roots from infected tomato (Solanum lycopersicum Green Zebra) were mixed vigorously in 10% commercial bleach for 5 min. The eggs were collected on a 25 μm sieve and were further surface sterilized by vigorously shaking them in 10% bleach for 5 min. followed by three washes with sterile H2O. The bleach and wash steps were performed twice. After the last wash, the eggs were pelleted by a final centrifugation (4,000 rpm for 5 min) and re-suspended in 5 ml water with 0. 1% SDS and 0. 2% Plant Preservative Mixture (Plant Cell Technology). Freshly hatched J2 were collected on a modified Baermann Funnel as described [66]. Sequences for MiPFN1 and MiPFN3 were obtained by BLAST searches of databases available online, such as WormBase, WormBase ParaSite [67] and NCBI, www. ncbi. nlm. nih. gov. M. incognita J2 cDNA was the template for amplifying the coding sequences of MiPFN3 and the coding sequence for Minc06346 [39] by PCR. (See S1 Table for primer sequences). The amplified products were cloned into the Gateway pENTR Directional vector (Invitrogen) and then into the Gateway vectors pB2GW7, to generate constructs for Arabidopsis plant transformation, into the Gateway vector pB7WGR2, to generate the 35S: : RFP-N terminal fusions for protoplast transformation [68], or into pDEST17 for expression in Escherichia coli for protein purification. For stable plant transformation with 35S: : MiPFN3, the construct was introduced to the A. tumefaciens strain GV3101 by heat shock transformation [69], and this was used to transform Arabidopsis thaliana Col-0 (N1093), using the floral dip method [70]. The seeds of the primary transformants were selected for BASTA resistance (Bayer CropScience, Wolfenbüttel, Germany). In the T2 generation, we selected lines segregating 3: 1 (BASTA-resistant/BASTA-susceptible). At least seven BASTA resistant plants for each segregating T2 line were transferred to new pots. We found two lines that showed dwarf rosettes and two lines that had plants with normal growth and developmental phenotypes. The two wild-type looking lines were grown for seed, and homozygous lines were confirmed by 100% survival on BASTA-containing media in the T3 generation. For the cloning of AtACT1, Arabidopsis Col-0 cDNA was used as the template for PCR, and the product was cloned into the entry vector pDONR207 and then destination vector pK2WG7. This construct was introduced to the A. tumefaciens strain GV3101 by heat shock transformation [69], which was then used to transform Arabidopsis thaliana Col-0 [70]. Seedlings (T1) from each background were first grown on plant media containing kanamycin to select for transformants containing the 35S: : AtACT1 construct. At 10 days post germination on selective media, healthy plants were transferred to soil. When seedlings started to produce inflorescences (approx. 4 weeks at 14h light/10h dark, 22°C), the rosette size and the leaf morphology were graded into three categories: 1) severe abnormal leaf curling/small rosette, 2) intermediate rosette size and 3) wild-type-like rosette size. Arabidopsis seeds were surface sterilized in 70% ethanol for 10 minutes, washed in 95% ethanol and allowed to air-dry. Seeds were placed on Murashige and Skoog media [71] with 2% sucrose and incubated in a growth chamber at 22°C/ 18°C, 80–100 μmol Photons/m2/s, 14h light/10h dark. The 14 day old seedlings were inoculated with 100–200 J2 of M. incognita. The inoculated plants were kept in the dark at 22°C as this facilitates infection for root-knot nematode bioassays [72]. Galls per root were counted at 4 weeks post-inoculation. For sectioning, galls at 23 dpi were dissected from plants, fixed overnight at 4°C in 2% PFA, 2% GA 0. 1M Cacodylate buffer. After the overnight incubation, fresh fixative was added to the samples and the samples were microwaved at 200 w until they reached 30°C. The samples were then incubated for 5 minutes at room temperature, rinsed then post-fixed with 1% OsO4 for overnight 4°C. The samples were dehydrated in an ethanol series (30% - 100%), then propylene oxide (PO). Galls were infiltrated with Spurrs resin prior to embedding and polymerization at 70°C overnight. Thick sections between 500 and 1000 nm were cut on a Leica EM UC7, stained for 45 seconds with 1% toluidine blue in 1% borax aq. , then mounted for light microscopy. Samples were observed using a Zeiss Axio Observer A1 microscope. Total RNA was extracted from eggs, freshly hatched J2, and gall enriched tissue. The gall enriched tissue was collected from infected tomato (Rutgers) roots grown in sand at greenshouse conditions at 7,14, and 21 days post-inoculation (dpi). To monitor earlier time points, 2 week old Col-0 seedlings grown on MS were inoculated with freshly hatched nematodes. Root tissue was collected at 4,7, 14, and 42 dpi. To monitor the nematode life stage in the Arabidopsis plants, the roots were stained with acid fuschin [73]. Nematodes and infected plant tissue from each time point was pooled for RNA extraction. cDNA synthesis and qRT-PCR was performed as previously described [74]. MiPFN1 and MiPFN3 expression was normalized to reference gene MiGAPDH [75]. Calculations were done according to the 2–ΔCT method [76]. For qRT-PCR analysis of transgene expression in the stable transgenic lines, RNA extraction from Arabidopsis seedlings or whole plants (dwarves). qRT-PCR analysis for transgenic plants were performed as described [74,77]. Calculations were done according to the 2–ΔCT method. AtUBQ5 served as a reference gene [78]. Primers for the qRT-PCR are listed in S1 Table. Using MiPFN3 purified PCR product as a template, an asymmetric PCR was performed in the presence of of DIG-labelled deoxynucleotide triphosphates (dNTPs) (Roche) to generate sense and anti-sense cDNA probes. In brief, the PCR contained 1x Advantage 2 buffer, 1x DIG-labelled nucleotides, 0. 4 μM forward or reverse primer (The sense probe primer 5’-AACTGGCCATGTCTCAAAGG-3’; the anti sense probe primer 5’-TTAATAATTGATGCTTCGAAAGTAA-3’), approximately 200–800 ng PCR product, 1x Advantage 2 Polymerase Mix. The reaction performed for 1 cycle 95°C, 1 min and then 35 cycles at 95°C for 30 seconds, 59°C for 30 seconds, and 68°C for 30 seconds. The PCR probes were precipitated by mixing in 1 volume of 3M sodium acetate and 3 volumes 100% ethanol and kept at -20°C for at least one hour. The probe was centrifuged and the pellet resuspended in 300 μl hybridization buffer 50% deionized formamide, 4X SSC buffer, 1X Blocking Reagent, 2% SDS, 1X Denhardt' s, 1 mm EDTA, pH 8,200 μg/ml Fish sperm DNA, 3. 125 yeast tRNA). Freshly hatched J2 were fixed and probed following the protocol of de Boer et al. , 1988 [79]. The DIG-labeled probes were detected by incubation with the Alkaline phosphatase-conjugated anti-digoxigenin antibody (Roche Molecular Biochemicals) and the alkaline phosphatase substrate. Representative images were collected with a digital camera on a Leica microscope. The E. coli strain BL21 was transformed with either pDEST17-Mi131 (6xHis-Mi131). BL21 was cultured in 3 ml of LB + 100 μg/μl of ampicillin (Amp) overnight. The overnight culture was transferred into 30 ml of LB-Amp and grew until OD600 = 0. 5. The protein expression was induced by adding 1 mM Isopropyl β-D-1-thiogalactopyranoside (IPTG) and incubating cells transformed with pDEST17-Mi131 construct at 37°C with 200 rpm shaking for 2 hours. Cells were harvested by centrifugation and resuspended in the lysis buffer and lysed by sonication at 60% power input for 5 minutes on ice. For the His-Mi131 purification, columns were prepared by adding 200 μl of Profinity IMAC resin into a Micro Bio-spin column. The spin columns were centrifuged at 1000g for 15 sec and washed with 250 μl deionized water. Columns were equilibrated by twice adding 250 μl of His purification wash buffer and centrifuging at 1000 x g for 15 sec. 200 μl of the bacterial lysate was added onto equilibrated columns and gently mixed by pipette. Lysates were incubated with resin for at least 5 minutes before centrifugation. The excess unbound proteins were removed by washing the column 3 more times with 250 μl of wash buffer. The bound protein was eluted with 100 μl of His purification elution buffer. Prior to the actin assays, the lysates, BSA, α-actin and purified His-Mi131 were prepared by ultracentrifugation at 150,000 x g for 60 min at 4°C and the supernatants were transferred into new Eppendorf tubes. The G actin sequestration and F actin binding assays were performed following the manufacturer’s protocol (Cytoskeleton #BK013). In brief, a G actin solution was prepared by diluting 1 mg/ml of non-muscle actin with 225 μl of general actin buffer. The G actin solution was mixed by pipetting up and down several times and incubated on ice for 60 min prior to the assay. After the incubation, 40 μl of G actin solution was added into each tube with either 10 μl of test proteins or 10 μl of actin buffer. The mixture was mixed several times by pipetting up and down and incubated at RT for 30 mins. After the incubation, 2. 5 μl of 10x polymerization buffer was added into each tube, mixed and incubated at room temperature for 30 min. To separate F actin from G actin, the mixtures were centrifuged at 150,000 x g for 90 min at 24°C. The supernatant was carefully removed and 5x reducing Laemmli buffer was added to each sample. The samples were centrifuged and the pellets were resuspended in 30 μl of Milli-Q water and incubated on ice for 10 min. Then 30 μl of 2 x Laemmli buffer was added to each sample. Samples were run on 4–20% SDS-gels and visualized by Coomassie staining. Arabidopsis Col-0 was transformed with pCAMBIA2300-ABD2 (Department of Cell Biology, Goettingen) using the floral dip method [70]. Approximately 10–15 leaves from 4–6 weeks old 35S: : ABD2-GFP plant (T2), grown at 22°C/ 18°C, 80–100 μmol Photons/m2/s, 12h light/12h dark, 60% humidity, were collected. The leaf tissue was lysed using a' Tape-Arabidopsis Sandwich' technique [80], in which pealed leaves were placed into 10 ml of enzyme solution (1. 25% (w/v) Cellulase R-10,0. 3% Macerozyme R-10,0. 4 M mannitol, 20 mM KCl, 10 mM CaCl2,20 mM MES (pH 5. 7). The leaves were incubated at room temperature for 2 hours with constant slow rotation until the protoplasts were released into the enzyme solution. Then the protoplasts were carefully collected by centrifugation at 750 rpm for 5 minutes. The pellet was washed twice with 10 ml W5 buffer (2 mM MES (pH 5. 7), 154 mM NaCl, 125 mM CaCl2,5 mM KCl). The cells were chilled on ice for 30 minutes prior. Prior to PEG transformation, the W5 buffer (2 mM MES (pH 5. 7), 154 mM NaCl, 125 mM CaCl2,5 mM KCl) was removed by centrifugation, and the pellet was gently resuspended in 5 ml MMG buffer (4 mM MES (pH 5. 7), 0. 4 M mannitol, 15 mM MgCl2). For PEG transfection of the protoplasts, up to 15. 0 μg of the plasmid DNA was placed in a 2 ml Eppendorf tube containing 300 μl of 40% PEG 4000 solution and gently mixed with Protoplasts resuspended in 300 μl MMG buffer. The solution was gently mixed and incubated at 22°C for 30 minutes. At the end of the incubation, 800 μl of W5 buffer was added and gently mixed. The supernatant was removed after centrifugation at 750 rpm for 2 minutes and protoplasts were washed with 800 μl of WI buffer (4 mM MES (pH 5. 7), 0. 5 M mannitol, 20 mM KCl). The supernatant was removed and the pellet was suspended in 300-μl WI buffer, mixed gently and incubated at 22°C, overnight. On the next day, the incubated protoplasts were transferred onto a glass slide for the observation under the confocal laser scanning microscope (x40).

Root-knot nematodes are microscopic plant pests that infect plant roots and significantly reduce yields of many crop plants. The nematodes enter the plant roots and modify plant cells into complex, multinuclear feeding sites called giant cells. The formation and maintenance of giant cells is critical to nematode survival. During giant cell organogenesis, the progenitor plant cells undergo many morphological changes, including a re-organization of the cytoplasmic actin cytoskeleton. As a result, the giant cell cytoplasmic actin appears fragmented and disorganized. Plant cells can regulate their actin filament assembly, in part, through the expression of actin binding proteins such as profilins. Here we show that infectious nematode juveniles express a profilin called MiPFN3. Expression of MiPFN3 in Arabidopsis plants made the plants more susceptible to root-knot nematodes, indicating that MiPFN3 acts as an effector that aids parasitism. We show evidence that the expression MiPFN3 in plant cells causes the fragmentation of plant actin filaments. The work here demonstrates that nematode effector MiPFN3 can directly affect plant actin filaments, whose reorganization is necessary for giant cell formation.

Abstract Introduction Results Discussion Materials and methods

biotechnology cell motility medicine and health sciences actin filaments cell processes brassica parasitic diseases nematode infections plant science model organisms genetically modified plants experimental organism systems plants genetic engineering research and analysis methods arabidopsis thaliana sequence analysis contractile proteins actins actin polymerization genetically modified organisms animal cells proteins bioinformatics biological databases biochemistry cytoskeletal proteins eukaryota plant and algal models sequence databases cell biology database and informatics methods biology and life sciences cellular types giant cells plant biotechnology organisms

2018

The root-knot nematode effector MiPFN3 disrupts plant actin filaments and promotes parasitism

Polycystic ovary syndrome (PCOS) is a disorder characterized by hyperandrogenism, ovulatory dysfunction and polycystic ovarian morphology. Affected women frequently have metabolic disturbances including insulin resistance and dysregulation of glucose homeostasis. PCOS is diagnosed with two different sets of diagnostic criteria, resulting in a phenotypic spectrum of PCOS cases. The genetic similarities between cases diagnosed based on the two criteria have been largely unknown. Previous studies in Chinese and European subjects have identified 16 loci associated with risk of PCOS. We report a fixed-effect, inverse-weighted-variance meta-analysis from 10,074 PCOS cases and 103,164 controls of European ancestry and characterisation of PCOS related traits. We identified 3 novel loci (near PLGRKT, ZBTB16 and MAPRE1), and provide replication of 11 previously reported loci. Only one locus differed significantly in its association by diagnostic criteria; otherwise the genetic architecture was similar between PCOS diagnosed by self-report and PCOS diagnosed by NIH or non-NIH Rotterdam criteria across common variants at 13 loci. Identified variants were associated with hyperandrogenism, gonadotropin regulation and testosterone levels in affected women. Linkage disequilibrium score regression analysis revealed genetic correlations with obesity, fasting insulin, type 2 diabetes, lipid levels and coronary artery disease, indicating shared genetic architecture between metabolic traits and PCOS. Mendelian randomization analyses suggested variants associated with body mass index, fasting insulin, menopause timing, depression and male-pattern balding play a causal role in PCOS. The data thus demonstrate 3 novel loci associated with PCOS and similar genetic architecture for all diagnostic criteria. The data also provide the first genetic evidence for a male phenotype for PCOS and a causal link to depression, a previously hypothesized comorbid disease. Thus, the genetics provide a comprehensive view of PCOS that encompasses multiple diagnostic criteria, gender, reproductive potential and mental health. Polycystic ovary syndrome (PCOS) is the most common endocrine disorder in reproductive aged women, with a complex pattern of inheritance [1–5]. Two different diagnostic criteria based on expert opinion have been utilized: The National Institutes of Health (NIH) criteria require hyperandrogenism (HA) and ovulatory dysfunction (OD) [6] while the Rotterdam criteria include the presence of polycystic ovarian morphology (PCOM) and requires at least two of three traits to be present, resulting in four phenotypes (S1 Fig) [6,7]. PCOS by NIH criteria has a prevalence of ~7% in reproductive age women worldwide [8]; the use of the broader Rotterdam criteria increases this to 15–20% across different populations [9–11]. PCOS is commonly associated with insulin resistance, pancreatic beta cell dysfunction, obesity and type 2 diabetes (T2D). These metabolic abnormalities are most pronounced in women with the NIH phenotype [12]. In addition, the odds for moderate or severe depression and anxiety disorders are higher in women with PCOS [13]. However, the mechanisms behind the association between the reproductive, metabolic and psychiatric features of the syndrome remain largely unknown. Genome-wide association studies (GWAS) in women of Han Chinese and European ancestry have reproducibly identified 16 loci [14–17]. The observed susceptibility loci in PCOS appeared to be shared between NIH criteria and self-reported diagnosis [17], which is particularly intriguing. Genetic analyses of causality (by Mendelian Randomization analysis) among women of European ancestry with self-reported PCOS suggested that body mass index (BMI), insulin resistance, age at menopause and sex hormone binding globulin contribute to disease pathogenesis [17]. We performed the largest GWAS meta-analysis of PCOS to date, in 10,074 cases and 103,164 controls of European ancestry diagnosed with PCOS according to the NIH (2,540 cases and 15,020 controls) or Rotterdam criteria (2,669 cases and 17,035 controls), or by self-reported diagnosis (5,184 cases and 82,759 controls) (Tables 1 and S1). We investigated whether there were differences in the genetic architecture across the diagnostic criteria, and whether there were distinctive susceptibility loci associated with the cardinal features of PCOS; HA, OD and PCOM. Further, we explored the genetic architecture with a range of phenotypes related to the biology of PCOS, including male-pattern balding [18–21]. We identified 14 genetic susceptibility loci associated with PCOS, adjusting for age, at the genome-wide significance level (P < 5. 0 x 10−8) bringing the total number of PCOS associated loci to nineteen (Tables 2 and S2 and Fig 1). Three of these loci were novel associations (near PLGRKT, ZBTB16 and MAPRE1, respectively; shown in bold in Table 2). Six of the 11 reported associations were previously observed in Han Chinese PCOS women [14,15]. Eight loci have been reported in European PCOS cohorts [16,17]. Obesity is commonly associated with PCOS and in most of the cohorts, cases were heavier than controls (Table 1). However, adjusting for both age and BMI did not identify any novel loci; and the 14 loci remained genome-wide significant. All variants demonstrated the same direction of effect across all phenotypes including NIH, non-NIH Rotterdam, and self-report (Fig 2 and S2 Table). Only one SNP near GATA4/NEIL2 showed significant evidence of heterogeneity across the different diagnostic groups (rs804279, Het P = 2. 6x10-5; Fig 2 and S3 Table). For this SNP, the largest effect was seen in NIH cases and the smallest in self-reported cases. Credible set analysis, which prioritises variants in a given locus with regards to being potentially causal, was able to reduce the plausible interval for the causal variant (s) at many loci (S4 Table). Of note, 95% of the signal at the THADA locus came from two SNPs. Examination of previously published genome-wide significant loci from Han Chinese PCOS [14,15] demonstrated that index variants from the THADA, FSHR, C9orf3, YAP1 and RAB5B loci were significantly associated with PCOS after Bonferroni correction for multiple testing in our European ancestry subjects (S5 Table). We assessed the association of the PCOS susceptibility variants identified in the GWAS meta-analysis with the PCOS related traits: HA, OD, PCOM, testosterone, FSH and LH levels, and ovarian volume in PCOS cases (Tables 3 and S6 and S2 Fig). We found four variants associated with HA, eight variants associated with PCOM and nine variants associated with OD. Of the eight loci associated with PCOM, seven were also associated with OD. Three of the four loci associated with HA were also associated with OD and PCOM. Two additional loci were associated with OD alone, one of which was the locus near FSHB (S6 Table). This locus was also associated with LH and FSH levels. There was a single PCOS locus near IRF1/RAD50 associated with testosterone levels (S6 Table). We repeated this analysis with susceptibility variants reported previously in Han Chinese PCOS cohorts [14,15]. In this analysis, there was one association with HA (near DENND1A), three with PCOM and three with OD (S2 Fig and S5 Table). A limitation of these analyses is the variable sample size across the phenotypes analysed. Additionally, the known referral bias for the more severely affected NIH phenotype (patients having both OD and HA) may result in more PCOS diagnoses than the other criteria [22], and may have contributed to the number of associations between the identified PCOS risk loci and these phenotypes. In the analyses looking at the weighted genetic risk score in the Rotterdam cohort, we observed an increase in the risk for PCOS (S3 Fig). Compared to individuals in the third quintile (reference group), individuals in the top 5th quintile of risk score have an OR of 1. 9 (1. 4–2. 5; 95% CI) for PCOS based on NIH criteria and an OR of 2. 1 (1. 7–2. 5; 95% CI) for Rotterdam criteria based PCOS. Of the associations, only the effect estimate for the Rotterdam criteria was significant, possibly due to the smaller size available with cases diagnosed according to the NIH criteria. When looking at the area under the ROC curves at SNPs with different P-value thresholds, we found a maximum AUC of 0. 54 using SNPs with a P-value < 5x10-6 for both diagnostic criteria. While this is significantly better than chance, it is unlikely that a risk score generated from the variants discovered to date would represent a clinically relevant tool. LD score regression analysis revealed genetic correlations with childhood obesity, fasting insulin, T2D, HDL, menarche timing, triglyceride levels, cardiovascular diseases and depression (Table 4) suggesting that there is shared genetic architecture and biology between these phenotypes and PCOS. There were no genetic correlations with menopause timing or male pattern balding. Mendelian randomization suggested that there was a causal role for BMI, fasting insulin and depression pathways (Table 5). Interestingly, while there was no genetic correlation detected for male pattern balding or menopause timing with PCOS, the Mendelian randomization analyses were significant. The difference in the genetic correlation compared to the Mendelian randomization result suggests that there may be a small number of key biological process that are common between the phenotypes, and that the common genetic causal variants are limited only to the variants shared by the subset of key biological processes. The importance of BMI pathways on reproductive phenotypes was further demonstrated by the attenuation of significance of Mendelian randomization analysis for age-at-menarche when BMI-associated variants were excluded from the analysis. We found 14 independent loci significantly associated with the risk for PCOS, including three novel loci. The 11 previously reported loci implicated neuroendocrine and metabolic pathways that may contribute to PCOS (1. 1 Note in S1 Data). Two of the novel loci contain potential endocrine related candidate genes. The locus harbouring rs10739076 contains several interesting candidate genes; PLGRKT, a plasminogen receptor and several genes in the insulin superfamily; INSL6, INSL4 and RLN1, RLN2 which are endocrine hormones secreted by the ovary and testis and are suspected to impact follicle growth and ovulation [23]. ZBTB16 (also known as PLZF) has been marked as an androgen-responsive gene with anti-proliferative activity in prostate cancer cells [24]. PLZF activates GATA4 gene transcription and mediates cardiac hypertrophic signalling from the angiotensin II receptor 2 [25]. Furthermore, PLZF is upregulated during adipocyte differentiation in vitro [26] and is involved in control of early stages of spermatogenesis [27] and endometrial stromal cell decidualization [28]. The third novel locus harbours a metabolic candidate gene; MAPRE1 (interacts with the low-density lipoprotein receptor related protein 1 (LRP1), which controls adipogenesis [29] and may additionally mediate ovarian angiogenesis and follicle development [30] (1. 2 Note in S1 Data). Thus, all the new loci contain genes plausibly linked to both the metabolic and reproductive features of PCOS. We found that there was no significant difference in the association with case status for the majority of the PCOS-susceptibility loci by diagnostic criteria. All susceptibility variants demonstrated the same direction of effect for the NIH phenotype, non-NIH Rotterdam phenotype and self-report, with only one variant demonstrating significant heterogeneity among the groups. It is of considerable interest that the cohort of research participants from the personal genetics company 23andMe, Inc. , identified by self-report, had similar risks to the other cohorts where the diagnosis was clinically confirmed. Our findings suggest that the genetic architecture of these PCOS definitions does not differ for common susceptibility variants. Only one locus, GATA4/NEIL2 (rs804279), was significantly different across diagnostic criteria: most strongly associated in NIH compared to the Rotterdam phenotype and self-reported cases. Deletion of GATA4 results in abnormal responses to exogenous gonadotropins and impaired fertility in mice [31]. The locus also encompasses the promoter region of FDFT1, the first enzyme in the cholesterol biosynthesis pathway [32], which is the substrate for testosterone synthesis, and is associated with non-alcoholic fatty liver disease [33]. The major difference between the NIH phenotype and the additional Rotterdam phenotypes is metabolic risk; the NIH phenotype is associated with more severe insulin resistance [34]. rs804279 does not show association with any of the metabolic phenotypes in the T2D diabetes knowledge portal {Type 2 Diabetes Knowledge Portal. type2diabetesgenetics. org. 2015 Feb 1; http: //www. type2diabetesgenetics. org/variantInfo/variantInfo/rs804279} so it may represent a PCOS-specific susceptibility locus. The significant association of PCOS GWAS meta-analysis susceptibility variants with the cardinal PCOS related traits OD, HA and PCOM further strengthened the hypothesis that specific variants may confer risk for PCOS through distinct mechanisms. Three variants at the C9orf3, DENND1A, and RAB5B were associated with all PCOS related traits. The findings were consistent with the Han Chinese DENND1A variant association with HA, as suggested previously [35]. Thus, these loci, along with GATA4/NEIL2 (as discussed above) may help identify pathways that link specific PCOS related traits with greater metabolic risk. In contrast, the variants at the ERBB4, YAP1, and ZBTB16 loci were strongly associated with OD and PCOM, and therefore, might be more important for links to menstrual cycle regularity and fertility. In addition, the FSHB variant was associated with the levels of FSH and LH [16,17], suggesting that it may act by affecting gonadotropin levels. This variant maps 2kb upstream from open chromatin (identified by DNase-Seq) and an enhancer (identified by peaks for both H3K27Ac and H3K4me1) in a lymphoblastoid cell line from ENCODE, indicating a potential role for a regulatory element ~25kb upstream from the FSHB promoter. Furthermore, the association between the IRF1/RAD50 variant and testosterone levels may indicate a regulatory role in testosterone production. Of note, results of the follow-up analysis show a high level of shared biology between PCOS and a range of metabolic outcomes consistent with the previous findings [17]. In particular, there is genetic evidence for increased BMI as a risk factor for PCOS. There is also genetic evidence that fasting insulin might be an independent risk factor. This study also confirmed a causal association with the pathways that underlie menopause [17], suggesting that PCOS has shared aetiology with both classic metabolic and reproductive phenotypes. Furthermore, there was an apparent effect of depression-associated variants on the likelihood of PCOS, suggesting a role for psychological factors on hormonally related diseases. However, the links between PCOS and depression might be complicated by pathways that are also related to BMI, as BMI pathways are causal in both PCOS and depression [36]. In addition, male-pattern balding-associated variants showed strong effects on PCOS, suggesting that this might be a male manifestation of PCOS pathways, as has been previously suggested [18,20,21,37]. This observation may reflect the biology of hair follicle sensitivity to androgens, seen in androgenetic alopecia, a well-recognised feature of HA and PCOS [38,39]. The Mendelian randomization results for male-pattern balding and menopause are significant despite non-significant genetic correlation results, suggesting that the shared aetiology may be specific to only a few key pathways. In conclusion, the genetic underpinnings of PCOS implicate neuroendocrine, metabolic and reproductive pathways in the pathogenesis of disease. Although specific phenotype stratified analyses are needed, genetic findings were consistent across the diagnostic criteria for all but one susceptibility locus, suggesting a common genetic architecture underlying the different phenotypes. There was genetic evidence for shared biologic pathways between PCOS and a number of metabolic disorders, menopause, depression and male-pattern balding, a putative male phenotype. Our findings demonstrate the extensive power of genetic and genomic approaches to elucidate the pathophysiology of PCOS. All research involving human participants has been approved by the authors' Institutional Review Board (IRB) or an equivalent committee, and all clinical investigation was conducted according to the principles expressed in the Declaration of Helsinki. Written informed consent was obtained from all participants. The Boston cohort was approved by the Partners IRB (# 2002P001924) and the University of Utah IRB (IRB_00076659). The deCODE cohort was approved by the National Bioethics Committee of Iceland (VSN 03–007), which was conducted in agreement with conditions issued by the Data Protection Authority of Iceland. Personal identities of the participants’ data and biological samples were encrypted by a third-party system (Identity Protection System), approved and monitored by the Data Protection Authority. The UK cohort was approved by the Parkside Health Authority (Now—NHS Health Research Authority, NRES Committee—West London & GTAC, UK, London, UK) under EC2359" The Molecular Genetics of Polycystic Ovaries." The Rotterdam PCOS cohort, the COLA study, was approved by institutional review board (Medical Ethics Committee) of the Erasmus Medical Center (04–263). Controls from the Rotterdam Study were approved by the Medical Ethics Committee of the Erasmus MC (registration number MEC 02. 1015) and by the Dutch Ministry of Health, Welfare and Sport (Population Screening Act WBO, license number 1071272-159521-PG). The Rotterdam Study Personal Registration Data collection is filed with the Erasmus MC Data Protection Officer under registration number EMC1712001. The Rotterdam Study has been entered into the Netherlands National Trial Register (NTR; www. trialregister. nl) and into the WHO International Clinical Trials Registry Platform (ICTRP; www. who. int/ictrp/network/primary/en/) under shared catalogue number NTR6831. The Chicago PCOS cohort was approved by the Northwestern IRB (#STU00008096). The control subjects from the NUgene study were approved by the Northwestern IRB (# STU00010003). The Estonia cohort was approved by the Research Ethics Committee of the University of Tartu approved the study (198T-18). The Twins UK study was approved by the St Thomas' Hospital Research Ethics Committee (EC04/015). The Nurses' Health Study (NHS I and II) was approved by the Partners Human Research Committee (#1999-P-011114). The meta-analysis included 10,074 cases and 103,164 controls from seven cohorts of European descent. For the analysis of PCOS related traits three additional cohorts, the Northern Finnish Birth Cohort (NFB66) [40], Twins UK [41] and the Nurses’ Health Study (NHS) [42] were included. Cases were diagnosed with PCOS based on NIH or Rotterdam Criteria or by self-report. The NIH criteria require the presence of both OD and clinical and/or biochemical HA for a diagnosis of PCOS [6]. The Rotterdam criteria require two out of three features 1) OD defined by oligo- or amenorrhea (chronic menstrual cycle interval >35 days in all cohorts), 2) clinical and/or biochemical hyperandrogenism (HA) and/or 3) PCOM for a diagnosis of PCOS [7]. Non-NIH Rotterdam was defined by OD and PCOM or clinical and/or biochemical hyperandrogenism (HA) and PCOM. Self-reported female cases from research participants in the 23andMe, Inc. (Mountain View, CA, USA) cohort either responded “yes” to the question “Have you ever been diagnosed with polycystic ovary syndrome? ” or indicated a diagnosis of PCOS when asked about fertility (“Have you ever been diagnosed with PCOS? ” or “What was your diagnosis? Please check all that apply. ” Answer = PCOS), hair loss in men or women (“Have you been diagnosed with any of the following? Please check all that apply. ” Answer = PCOS) or research question (“Have you ever been diagnosed with PCOS? ”) [17]. 23andMe controls were female, only. HA was defined as hirsutism and quantified by the Ferriman-Gallwey (FG) score. The FG score assesses terminal hair growth in a male pattern in females, and a score above the upper limit of normal controls (>8) is considered hirsutism [43]. Hyperandrogenemia was defined as testosterone, androstenedione or DHEAS greater than the 95% confidence limits in control subjects in the individual population. OD was defined as cycle interval <21 or >35 days [44]. PCOM was defined as 12 or more follicles of 2–9 mm in at least one ovary or an ovarian volume >10 mL [7]. The quantitative PCOS traits included levels of total testosterone (T), follicle-stimulating hormone (FSH), and luteinizing hormone (LH) and ovarian volume (S1 Table). An overview of the cohorts, diagnostic criteria and number of subjects included in each subphenotype or trait analysis are summarized in Tables 1 and S1. Each study provided summary results of genetic per-variant estimates produced in either case-control or trait association analyses. Adjustment for principle components was performed at the study level. The collected files underwent quality control (QC) by two independent analysts using the EasyQC pipeline [45]. Variants were excluded based on minor allele frequency (MAF) < 1%, imputation quality (R2) < 0. 3 or info < 0. 4 for MACH and IMPUTE2 respectively [46,47]. Per-cohort QC results from EasyQC are shown (S7 Table), and allele frequency spectrum for each cohort, and the combined cohort after meta-analysis is shown (S4 Fig). The per-variant estimates collected from the summary statistics of contributing studies were meta-analysed using a fixed-effect, inverse-weighted-variance meta-analysis that employed either GWAMA [48] or METAL [49]. In addition to the overall meta-analysis, we performed meta-analyses for studies with available data for the separate PCOS diagnostic criteria: NIH, non-NIH Rotterdam [7] and self-report [17], as well as for the PCOS related traits of HA, OD and PCOM. The meta-analysis of PCOS status was performed using two models; (1) age-adjusted, (2) age and BMI-adjusted, given the high prevalence of obesity in affected women that resulted in cases being significantly heavier than controls in most cohorts (Table 1). We removed any variants that were not present in more than 50% of the effective sample size prior to combining with 23andMe as this was the largest cohort in the meta-analysis, providing approximately 51% of the PCOS cases and 80% of controls. We also removed any variants only present in one study. The meta-analysis of PCOS related traits was performed adjusting for age and BMI. Identified variants were annotated for insight into their biological function using ANNOVAR [50] to assign refGene gene information, SIFT score [51], PolyPhen2 scores [52], CADD scores [53], GERP scores [54] and SiPhy log odds [55]. In order to compare different PCOS diagnostic criteria [ (1) NIH, (2) non-NIH Rotterdam and (3) self-reported] included in the PCOS meta-analysis, an additional meta-analysis was performed to test for heterogeneity across these independent PCOS case groups. These three PCOS case groups were combined in an inverse variance weighted fixed meta-analysis and the heterogeneity statistics (Cochran’s Q and I2) were obtained using GWAMA [48]. Any variant with a statistically significant Cochran’s Q p-value (P<0. 05/14 = 0. 0036 corrected for multiple testing) and I2>70% were considered exhibiting heterogeneity across the PCOS case groups. Further analysis of the heterogeneity included comparison of the 95% confidence intervals for the direction of effect and overlaps. In order to understand biology relevant to identified PCOS susceptibility, we assessed the association between index SNPs at each genome-wide-significant locus and the PCOS related traits HA, OD, PCOM as well as the quantitative traits testosterone, LH and FSH levels and ovarian volume. The threshold for significance in this analysis was p<4. 5×10−4 (Bonferroni correction [0. 05/ (14 independent loci x 8 traits) ]. In order to identify shared risk loci between the previously reported GWAS in Han Chinese PCOS cases and our European ancestry cohort, 13 independent signals (represented by 15 SNPs) at 11 genome-wide significant loci reported by Chen et al. [14] and Shi et al. [15] were investigated for association in our meta-analyses of PCOS and PCOS related traits. The adjusted P-value for this analysis was <0. 00048 (Bonferroni correction [0. 05/ (13 independent signals x 8 traits) ]). Information on the biological function of the nearest gene (or genes, if variants were equidistant from more than one coding transcript and annotated as such by ANNOVAR [49] to the index SNP of each identified risk locus) was collected by performing a search of the Entrez Gene Database [56], and collecting the co-ordinates of the gene (genome build 37; hg19) as well as the cytogenetic location and the summary of the gene function. In addition to the EntrezGene Database queries, the gene symbol was used as a search term in the PubMed database [57], either alone or combined with the additional search term “PCOS” to identify relevant published literature in order to obtain information on putative biological function and involvement in the pathogenesis of PCOS (summarized in 1. 1 Note in S1 Data). One potential use of genetic risk scores is prediction of disease. The ability of genetic risk scores calculated from loci discovered in analysis of the different diagnostic criteria to discriminate cases from alternative criteria was measured. We constructed a weighted genetic risk score based on a meta-analysis excluding the Rotterdam Study subjects. The weighted genetic risk score was divided into quintiles and tested for association with PCOS in the Rotterdam cohort. The middle quintile was used as the reference and the odds for having PCOS based on both Rotterdam and NIH criteria was then calculated. Additionally, the 23andMe results were used to select independent SNPs with cut-offs of p<5×10−4 to p<5×10−8. The Rotterdam cohort was then used to calculate risk scores and the area-under-the curve (AUC) for both NIH and Rotterdam diagnostic criteria. Analyses were performed using PLINK v1. 9 and SPSS v21 (IBM Corp, Armonk, NY) [58]. To assess the level of shared etiology between PCOS and related traits, we performed genetic correlation analysis using LD-score regression [59]. Publicly available genome-wide summary statistics for body mass index (BMI) [60], childhood obesity [61], fasting insulin levels (adjusted for BMI) [62], type 2 diabetes [63], high-density lipoprotein (HDL) levels [64], menarche timing [65], triglyceride levels [64], coronary artery disease [66], depression [36], menopause [17] and male pattern balding [67] were used to estimate the genome-wide genetic correlation with PCOS. The adjusted P-value for this analysis was p<0. 0045 after a Bonferroni correction (0. 05/11 traits). Phenotypes of interest, both where there was evidence of shared genetic architecture and where there was previous evidence for genetic links, were assessed using Mendelian randomization methods. Mendelian randomization differs from LD score regression in that one phenotype is analysed as a potential causal factor for another. Mendelian randomization was performed using both inverse weighted variance and Egger’s regression methods [68], with inverse weighted methods being more powerful, but Egger’s methods being resistant to directional pleiotropy (where there are a set of SNPs that appear to have an alternative pathway of effect). We report here the results of the IVW methods as none of the analysis suggested that the MR-EGGERs results were more appropriate given that none of the EGGERs intercepts were significant (Table 5). In addition to the phenotypes implicated by the LD-score regression measures, male pattern balding has a strong biological rationale and was therefore included. The genetic score for childhood obesity substantially overlaps with the score for adult BMI (such that the INSIDE violation—where the effect of SNPs on a confounding factor scales with that on the trait of interest—of Mendelian randomization would likely occur [69], so only a score for BMI was used, with the proviso that this represents BMI across the whole of the life course after very early infancy. The SNPs for depression were drawn from the results of a more recent analysis, for which there was not, at time of analysis, publicly available genome-wide data. We defined a locus as mapping within 500kb of the lead SNP. For each locus, we first calculated the posterior probability, πCj, that the jth variant is driving the association, given by: πcj=ΛjΣkΛk where the summation is over all retained variants in the locus. In this expression, Λj is the approximate Bayes’ factor [70] for the jth variant, given by Λj=Vj+ωVjexp[−ωβ2j2Vj (Vj+ω) ] where βj and Vj denote the estimated allelic effect (log-OR) and corresponding variance from the meta-analysis. The parameter ω denotes the prior variance in allelic effects, taken here to be 0. 04 [70]. The 99% credible set [71] for each signal was then constructed by: (i) ranking all variants according to their Bayes’ factor, Λj; and (ii) including ranked variants until their cumulative posterior probability of driving the association attained or exceeded 0. 99.

We performed an international meta-analysis of genome-wide association studies combining over 10,000,000 genetic markers in more than 10,000 European women with polycystic ovary syndrome (PCOS) and 100,000 controls. We found three new risk variants associated with PCOS. Our data demonstrate that the genetic architecture does not differ based on the diagnostic criteria used for PCOS. We also demonstrate a genetic pathway shared with male pattern baldness, representing the first evidence for shared disease biology in men, and shared genetics with depression, previously postulated based only on observational studies.

Abstract Introduction Results Discussion Methods

genome-wide association studies body weight medicine and health sciences statistics metaanalysis computational biology cancers and neoplasms oncology mathematics physiological parameters genome analysis polycystic ovary syndrome research and analysis methods genomics mathematical and statistical techniques gynecological tumors genetic loci phenotypes physiology genetics body mass index biology and life sciences physical sciences human genetics genetics of disease statistical methods

2018

Large-scale genome-wide meta-analysis of polycystic ovary syndrome suggests shared genetic architecture for different diagnosis criteria

Schistosoma japonicum still causes severe parasitic disease in mainland China, but mainly in areas along the Yangtze River. However, the genetic diversity in populations of S. japonicum has not been well understood across its geographical distribution, and such data may provide insights into the epidemiology and possible control strategies for schistosomiasis. In this study infected Oncomelania snails were collected from areas in the middle and lower (ML) reaches of the Yangtze River, including Hubei, Hunan, Anhui, Jiangxi and Jiangsu provinces, and in the upper reaches of the river, including Sichuan and Yunnan provinces in southwest (SW) China. The adult parasites obtained from experimentally infected mice using isolated cercariae were sequenced individually for several fragments of mitochondrial regions, including Cytb-ND4L-ND4,16S-12S and ND1. Populations in the ML reaches exhibited a relatively high level of diversity in nucleotides and haplotypes, whereas a low level was observed for populations in the SW, using either each single fragment or the combined sequence of the three fragments. Pairwise analyses of F-statistics (Fst) revealed a significant genetic difference between populations in the ML reaches and those in the SW, with limited gene flow and no shared haplotypes in between. It is rather obvious that genetic diversity in the populations of S. japonicum was significantly correlated with the geographical distance, and the geographical separation/isolation was considered to be the major factor accounting for the observed difference between populations in the ML reaches and those in the SW in China. S. japonicum in mainland China exhibits a high degree of genetic diversity, with a similar pattern of genetic diversity as observed in the intermediate host snails in the same region in China. Schistosomiasis is one of the most neglected tropical diseases, with six species in the Schistosoma still infecting more than 200 million people in the world [1]–[3]. Schistosomiasis japonica is distributed in Indonesia, Philippines, and China. In mainland China, this parasitic disease is the most severe zoonosis infecting about 360,000 people and about 1% buffalo and/or cattle in endemic regions, particularly in lake/marshland and hilly areas of Hubei, Hunan, Anhui, Jiangxi and Jiangsu provinces and mountainous areas of Sichuan and Yunnan provinces [4]. Over the last 50 years, continuous efforts involving various measures, such as health education, snail control, community chemotherapy and environmental management have contributed significantly to the dramatic reduction in infection levels and epidemic areas of this parasitic disease in China, setting China as one of the most successful countries in control of schistosomiasis in the world [5]–[8]. However, recently available data have suggested that schistosomiasis has re-emerged over the last decade, probably as a severe threat once again to human health especially in rural areas of mainland China [9], [10]. The drastic pathogenesis, the number of reservoir hosts involved in epidemiology and the vast endemic areas of schistosomiasis japonica have inevitably resulted in a less investigated status for S. japonicum in respect with its genetic diversity, host immune response etc. when compared with other schistosomes [6], [11], [12]. The genus Oncomelania, which is the intermediate host of S. japonicum, was classified into different species and/or subspecies according to their morphology, biogeography and phylogeny [13], [14]. With the distinct diversity of snails in the genus Oncomelania which has been verified using various markers [14]–[16], the diversity of the parasite S. japonicum is of great interest for research from a co-evolutionary point of view. How diverse the digenean S. japonicum really is in such a large geographical range has not been well assessed especially in mainland China. An accurate measure of its population genetic diversity is certainly needed to clarify our understanding on the epidemiology of schistosomiasis [17], which may be also useful for implementing control measures, and for developing drugs or potential vaccines, as worms of different genetic backgrounds may respond differently to such treatments [18], [19]. In recent years, several molecular markers have been used to detect the variability of S. japonicum populations. Gasser et al. [20] found the variability among 7 geographical isolates across mainland China using the random amplified polymorphism DNA (RAPD) technique and suggested a potential strain complex for S. japonicum. Sorensen et al. [21] reported differences between S. japonicum populations from 6 localities in mainland China using NADH dehydrogenase subunit 1 (ND1) gene, but could not detect variability conclusively at the intrapopulation level. Bøgh et al. [22] did find 15 types of ND1 conformations and 23 types of cytochrome c oxidase subunit 1 (CO1) conformations in 9 populations from 7 provinces across mainland China by single-strand conformational polymorphism (SSCP). These results did in fact suggest the significant polymorphism among S. japonicum in mainland China, but provided very limited information relating to the population genetic diversity of this species. Upon the identification of polymorphic microsatellite loci, Shrivastava et al. [6] investigated the genetic variation of S. japonicum populations from 8 geographical locations in 7 endemic provinces across mainland China, and a high level of polymorphism was reported between and within populations. They considered that populations of S. japonicum in mainland China could be separated mainly into the populations in Sichuan and Yunnan provinces as being in southwest (SW) China and those in low-lying lake regions along the middle and lower (ML) reaches of Yangtze River. With three partial mitochondrial genes (cox3, nad4 and nad5) from 28 individual adult worms, Zhao et al. [23] reported recently that all parasites from SW China were grouped together, whereas those from the ML reaches of Yangtze River were not clustered together. However, the reports by Shrivastava et al. [6] and Zhao et al. [23] both contained limited specimens from relatively few localities, which may not represent the geographical distribution of this schistosome, and thus not its exact population genetic diversity, in mainland China. A comprehensive analysis is therefore needed using more molecular markers to examine more populations of S. japonicum from a wide range of its geographical distributions, especially in severely endemic areas along the ML reaches of Yangtze River in China. In this study, mitochondrial DNA sequences including Cytb-ND4L-ND4,16S-12S and ND1 were examined for S. japonicum collected from localities in seven provinces of China, where schistosomiasis is geographically endemic. The diversity in nucleotides and haplotypes was analyzed for different populations based on each of the three mitochondrial sequences and their combined sequences. Phylogenetic tree and parsimony network were constructed for observed haplotypes, and the genetic distance was examined against the geographical distance in order to understand the genetic diversity in populations of S. japonicum in mainland China. The procedures involving animals were carried out in accordance with the guidelines of the Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC). The animal study protocol was approved by the Institutional Animal Care and Use Committee of Wuhan University. The intermediate host, Oncomelania hupensis, from 18 localities of 7 schistosomiasis endemic provinces in mainland China, including Hubei, Hunan, Anhui, Jiangxi, Jiangsu provinces in the ML reaches of Yangtze River, and Sichuan and Yunnan provinces which are in the higher reaches of the river in SW China, but separated from the ML reaches by mountain ranges (Fig. 1 and Table 1), were collected and transported to laboratory from October 2005 to October 2006. After one month captivity, snails were washed and exposed individually in water for 3 h in a vial under light at 25°C to stimulate the emergence of cercariae for identifying the S. japonicum infection. Overall, snails from different localities had an infection rate ranging from 0. 1% to 4%. To generate adult worms, the best source of DNA, 10 field-collected infected snails from each locality, with the exception of Zongyang in Anhui province (AHzy) and Pengze in Jiangxi province (JXpz) where only three and four infected snails were obtained respectively, were exposed to light for 4 hours to stimulate the emergence of cercariae. Five laboratory mice were infected percutaneously with 30 cercariae per mouse for each geographical locality. 6 weeks following the infection, adult worms were retrieved by perfusion from mesenteric veins using 0. 9% NaCl, and worms from each mouse infected with cercariae were pooled together, and washed extensively in saline before being preserved in 95% ethanol at 4°C. The total genomic DNA was extracted individually from both male and female schistosomes using a standard sodium dodecyl sulfate-proteinase K procedure [24]. Each worm was incubated and thawed in 200 µl extraction buffer containing 50 mM Tris-HCl, 50 mM EDTA, 100 mM NaCl, 1% SDS and 100 µg/ml proteinase K, at 56°C for 2 h with gentle mixing. DNA in solution was extracted using standard phenol/chloroform purification, followed by 3 M sodium acetate (pH 5. 2) and ethanol precipitation. Pellets of DNA were washed in 70% ethanol, air-dried, and resuspended in 10 µl TE (pH 8. 0). For each adult worm, three fragments, i. e. Cytb-ND4L-ND4, ND1 and 16S-12S of the mitochondrial genome were sequenced. For the Cytb-ND4L-ND4 fragment, the forward primer ND4F (5′- TTGGGGGTTGTCATGCGGAGTA -3′) and the reverse primer ND4R (5′- CAAATACCCAATAGCAACGGAACAC -3′) were used based on available GenBank sequence AF215860. For the ND1 fragment, the forward primer ND1F (5′- TAGAGGGTTTGTTGGTTGTTTTG -3′) and the reverse primer ND1R (5′- ACCATACTTTCATACTACTGCC -3′) were used based on available GenBank sequence AF215860. For the 16–12S fragment, the forward primer 16S-12SF (5′- GATTATTTCTAGTTCCCGAATGG -3′) and the reverse primer 16–12SR (5′- TGTAACGCACAACAACCTATACC -3′) were used based on available GenBank sequence AF215860. The PCR protocols were 94°C for 3 min followed by 30 cycles of 94°C for 30 s, 58°C (for ND1) or 63°C (for Cytb-ND4L-ND4 and 16S-12S) for 30 s, and 72°C for 90 s and then a final elongation step at 72°C for 10 min. The amplified products were purified on 1. 0% agarose gel stained with ethidium bromide, using the DNA gel extraction kit (Omega Bio-Tek). The purified PCR products were sequenced using ABI PRISM BigDye Terminators v3. 0 Cycle Sequencing (ABI). The DNA sequences were deposited in the GenBank database under accession numbers FJ851893–FJ852573. Sequences were aligned using ClustalX1. 83 [25] at default settings followed by manual correction in SEAVIEW [26] for each molecular marker. DNAsp version 4. 0 [27] was used to define the haplotype. The three parts, i. e. Cytb-ND4L-ND4, ND1 and 16S-12S, of mitochondrial data were also combined and aligned into a new combined mitochondrial data set, with this combined sequence named as combined mtDNA. Nucleotide divergences within and between populations were calculated in Arlequin3. 11 [28] and DNAsp. Genetic variation within different populations was estimated by calculating nucleotide diversity (π) and haplotype diversity (h) values. Selective neutrality was tested with Tajima' s D [29] and Fu' s F test [30]. The pairwise genetic difference was estimated for all populations by calculating Wright' s F-statistics (Fst) based on gene flow (Nm). A Mantel g-test to compare the correlation between pairwise distance and geographical distance among localities was analyzed in Arlequin, with geographic distances (km) for the correlation analysis between geographical distance and genetic distance calculated using the great circle distance between localities. The phylogenetic analysis for 96 haplotypes generated using combined mitochondrial DNA data was performed with Bayesian inference (BI), which was carried out with MrBayes 3. 1 [31] under the best-fit substitution model. Analyses were run for 1×106 generations with random starting tree, and four Markov chains (with default heating values) sampled every 100 generations. Posterior probability values were estimated by generating a 50% majority rule consensus tree following the discard of first 3000 trees as part of a burn-in procedure. The HKY+I+G model was determined as the best-fit model of sequence evolution by using the hierarchical likelihood ratio tests implemented in Modeltest 3. 7 [32]. The phylogenetic tree was rooted using Schistosoma mansoni as outgroup. The genetic structure was phylogenetically evaluated by constructing unrooted parsimony network of haplotypes for combined mtDNA data sets using TCS version 1. 21 [33]. The primary sequence data were obtained by amplifying and sequencing three partial regions of the mitochondrial genome, i. e. Cytb-ND4L-ND4 with 793–794 bp, ND1 with 767 bp, and 16S-12S with 1463–1466 bp. Measures of diversity of haplotypes and nucleotides within populations on the basis of the three mitochondrial regions are presented in Tables S1, S2 and S3, respectively. The highest values for the diversity were all observed for populations in the ML reaches, and the lowest all in populations from the SW (for details regarding each fragment, see Tables S1, S2 and S3). The pairwise genetic distance among all 18 populations showed a high degree of variation, as revealed respectively from the three different mitochondrial regions (for details, see Tables S4, S5 and S6). A significant correlation was observed between geographical distance and genetic distance (pairwise Fst) for all 18 populations for Cytb-ND4L-ND4 (R = 0. 642, P<0. 001) and 16S-12S (R = 0. 746, P<0. 001), respectively, which indicates that genetic distance increased with the increase in geographical distance (Fig. 2a, b). No significant correlation was detected when ND1 was used, with the correlation coefficient R = 0. 080 (P>0. 05) (Fig. 2c). However, among 15 populations in the ML reaches, the value of the correlation coefficient decreased to 0. 119 (P>0. 05) and 0. 061 (P>0. 05) for Cytb-ND4L-ND4 and 16S-12S, respectively (Fig. 2d, e), implying that the genetic distance was not correlated with the geographical distance for populations in the ML reaches of Yangtze River. Although some base substitutions were observed, selective neutrality of the observed nucleotide polymorphisms was suggested for S. japonicum, as indicated either by Tajima' s D or Fu' s F test (P>0. 05) in each of the three regions. As many studies have shown that longer genes contain generally more variable characters with proportionally more signals, and hence yield accurate phylogenetic estimates than shorter ones [34]–[36], the combined mitochondrial data sets were then deduced from 169 specimens by aligning combined Cytb-ND4L-ND4, ND1 and 16S-12S sequences (combined mtDNA), which had a range of 3024 to 3027 bp, resulted in 3028 characters, including gaps, and 166 variable sites (113 parsimony informative sites). A total of 96 mitochondrial haplotypes was observed (Table 1). Measures of haplotype and nucleotide diversity based on combined mtDNA are presented in Table 2. The highest values in the diversity of haplotype and nucleotide were all observed for populations in the ML reaches, and the lowest were all in populations from the SW, which is consistent with the findings from the three separate mitochondrial DNA sequences. 88 haplotypes were isolated from 143 specimens in five provinces along the ML reaches, with the mean haplotype and nucleotide diversity being 0. 987±0. 003 and 0. 0036±0. 0001, respectively. However, only 8 haplotypes were isolated from 26 specimens in the SW, with the haplotype and nucleotide diversity being 0. 766±0. 075 and 0. 0017±0. 0003, respectively. The Fst of all pairwise analyses varied from 0. 482 to 0. 870 between populations in the ML reaches and those in the SW (Table 3), showing highly significant difference (P<0. 001). Among the 3 populations in the SW, the Fst between SCxc and two Yunnan populations (YNey and YNhq) showed highly significant differences (P<0. 001), whereas no significant difference was observed between YNey and YNhq (P>0. 05). Among the 15 populations in the ML reaches, the Fst varied from 0. 014 to 0. 807 (Table 3), with most of them being significantly different (P<0. 05). When all specimens were classified into two populations according to whether they were from above or below the three Gorges region, i. e. population in the ML reaches of the Yangtze River and population in Sichuan and Yunnan provinces of the SW China, the value of genetic distance (Fst) and the gene flow (Nm) between them was 0. 381 (P<0. 001) and 0. 410, respectively. Significant correlation was also observed between geographical distance and genetic distance (pairwise Fst) among all 18 populations for combined mtDNA (R = 0. 670, P<0. 001), indicating that genetic distance increased with the increase in geographical distance (Fig. 2f). Among 15 populations in the ML reaches, the value of the correlation coefficient decreased to 0. 077 (P>0. 05) (Fig. 2g), implying that the genetic distance was not correlated with the geographical distance for populations in the ML reaches of Yangtze River. As shown in the Bayesian phylogenetic tree (Fig. 3), two clades can be clearly separated. Clade A contains almost all haplotypes from all five provinces in the ML reaches of the Yangtze River. Although various divergence and some subclades were observed within this clade, support probabilities for each clade were generally very low. Haplotypes in the ML reaches were clustered in various subclades, and no obvious lineage was observed for haplotypes from different provinces along the ML reaches. However, subclades A1 and A2 include most haplotypes from Hubei, Hunan, Anhui, and Jiangxi provinces, and subclade A6 includes haplotypes from Hubei, Hunan, Anhui, and Jiangsu provinces. It is apparent that clade B can be separated into two distinct subclades, B1 and B2, with clade B1 having a high support probability and containing only haplotypes from Sichuan and Yunnan provinces in SW China, and B2 containing three haplotypes from three provinces in the ML reaches. Surprisingly, other trees (NJ, ML, MP; not shown), although inconsistent in their respects, all had such two clades containing haplotypes from SW China, and three from the ML reaches, despite a relatively low level of support probabilities. The network constructed by statistical parsimony from 96 haplotypes on the basis of combined mtDNA sequences showed some characters as observed in the phylogenetic tree. The haplotype network was rather complicated, without any obvious lineages for those haplotypes from localities in the ML reaches (Fig. 4). However, all haplotypes from SW (from H26 to H33) were clustered together (Fig. 4), which corresponded exactly to clade B1 in Fig. 3, and this clade contained no haplotypes from the ML reaches of Yangtze River, but was related with a few haplotypes from the ML reaches (Fig. 4), as also indicated in clade B2 which formed, together with B1, into clade B (Fig. 3). A relatively large network containing haplotypes (from H71 to H93) from about 10 localities (Fig. 4) showed some similarity with clade A1 in Fig. 3, in composition of haplotypes. It is, however, impossible to detect any other patterns of haplotype networks, and impossible to find other geographical relationships or characteristic lineages in other network branches, which is largely consistent with the complex structure of clade A in Fig. 3. The difference in genetic diversity of S. japonicum populations was demonstrated in samples collected from currently epidemic areas of schistosomiasis in mainland China, with the use of three mitochondrial fragments, Cytb-ND4L-ND4, ND1 and 16S-12S, respectively, and the combined sequences of these three fragments. The present study contains the mostly widespread and the largest number of S. japonicum populations in any attempts so far to examine the parasite genetic diversity in China. Overall, populations of S. japonicum in mainland China showed a relatively large degree of variation in terms of nucleotide and haplotype diversity. However, it is apparent that across the geographical distribution of schistosomiasis endemic areas in China, the genetic distance was correlated significantly with geographical distance when Cytb-ND4L-ND4,16S-12S, and combined mtDNA were used, although non-significance was observed for ND1. It is even more obvious that as revealed through analyses of nucleotide and haplotype diversity, populations in Hubei, Hunan, Anhui, Jiangxi and Jiangsu provinces, namely in the ML reaches of Yangtze River showed a much larger degree of genetic variation than those in Sichuan and Yunnan provinces of the SW China in the upper reaches of the river, and no haplotypes were shared between populations in the ML reaches and those in the SW. Significant difference was also observed in genetic distance between populations in the ML reaches and populations in the SW, as revealed in pairwise analyses using individual and/or combined mitochondrial sequences. Along the Yangtze River, are the endemic areas of schistosomiasis, and severe epidemic areas are mainly in the ML reaches [5]. However, in the Three Gorges area that is from Yichang going upwards to Yibin (Fig. 1), human schistosomiasis has never been reported [10]. It is quite obvious that the distribution of S. japonicum is geographically separated by the gorge area of the river. This apparent geographical separation may account for the observed difference in no-shared haplotypes, and in the genetic distance for S. japonicum between areas in the ML reaches of Yangtze River and areas in the SW China. When populations from the ML reaches and from the SW were further grouped separately, the Fst value (0. 381) was greater than 0. 25, a value which was considered to be ‘very great’ by Wright [37] for genetic differentiation between populations. It is therefore all indicated that a large level of genetic differentiation has evolutionarily occurred for S. japonicum, due to at least the geographical separation by the Three Gorges area and mountains. Phylogenetic analyses and haplotype network may support this conclusion, as parasites from Sichuan and Yunnan provinces in the SW were all closely clustered in the phylogenetic tree and the haplotype network. Using different molecular markers, other authors [6], [23] have also, to some extent, detected the genetic difference between S. japonicum populations in the SW and those in the flood plain of the ML reaches of the Yangtze River. Despite the finding that the mean nucleotide and haplotype diversity of populations in the SW were rather low when compared with the same parameters in the ML reaches, the genetic distance had some significant difference between the population from Sichuan, SCxc, and the two populations from Yunnan, YNey and YNhq, as revealed by Fst of pairwise analyses using ND1,16S-12S, and the combined mtDNA sequences, with the exception of Cytb-ND4L-ND4. Sichuan and Yunnan provinces are both distributed in Hengduan Mountains, and schistosomiasis was reported historically in various localities in these two provinces [38]. As various mountain ranges and rivers, as well as intermountain basins, are the general features in Hengduan Mountains [39], there must be some degree of geographical isolation in the distribution of S. japonicum in this region at a large geographical scale. However, only three populations were included in the present study and efforts to obtain more parasite samples have been unsuccessful, although the intermediate host snails were collected in a much wider range (unpublished data), due possibly to the continuous and extensive practices in either snail control or human chemotherapy in the two provinces. Thus, whether there is an effect of geographical isolation on populations of S. japonicum in this mountainous area will likely remain unknown, and whether the observed low level of genetic variation in these populations resulted from a recent bottleneck effect as a consequence of intensive control practices may also remain to be answered. Ecological habitats were thought to affect population genetic diversity of S. japonicum in mainland China [40]. The mountainous habitats in Sichuan and Yunnan provinces may differ obviously from the habitats for the intermediate host in the ML reaches, in several aspects such as in hydrology, altitude and soil etc. [41], [42], but the difference should mostly be attributed to the geographical separation, rather than a simple impact from habitat difference. In the ML reaches of Yangtze River, it was impossible to clarify any patterns of haplotype clustering in relation to types of sample localities or to provinces, as haplotypes from a single locality were generally clustered in different clades. It can thus be speculated that S. japonicum might have experienced frequent gene flows in most populations in this region (Table 3). The localities for O. hupensis in the ML reaches have extensive physical connections through channels with the Yangtze River. With frequent occurrence of floods in the Yangtze River basin, especially in its ML reaches, snails in these habitats can be dispersed and subsequently deposited widely in various localities, and this naturally occurred instance was, in a previous research, proposed to explain the high genetic diversity of O. hupensis in the ML reaches [16]. It was further considered that this distinct genetic diversity in snail intermediate hosts may have strong implications in genetic diversity of schistosomes in mainland China [16], as demonstrated clearly in the present study. The accumulation of mixed sources of snails, especially infected snails can reconstitute the parasite population, leading to the existence of various haplotypes within a single population, and also to the limited degree of genetic distance between populations in the ML reaches as observed in the present study, which supports the speculation by Davis et al. [43] that floods may be the cause of the widespread mixing and dispersal of snails, leading to greater genetic diversity in O. hupensis populations along the Yangtze River plains compared with populations in SW China. Surprisingly, the number of haplotypes, being 80 and 13 for the intermediate host snails in the ML reaches, and in Sichuan and Yunnan provinces [16], matches roughly, if not coincidently, with the number of haplotypes, 88 and 8, for S. japonicum in the ML reaches and in Sichuan and Yunnan provinces in this study, respectively. The intermediate host snails and the schistosome in China exhibit a lesser degree of genetic diversity in the SW, but a relatively larger degree in the ML reaches of the Yangtze River, as reported in a previous study on the intermediate host snails [16] and in this study. No shared haplotypes were observed either in the intermediate host snails or in the schistosomes between localities from the ML reaches and from the SW. Zhao et al. [44] recently reported that the intermediate host snails O. hupensis robertsoni in Sichuan and the snail O. hupensis hupensis in the ML reaches had a 10. 3% genetic distance, strongly indicating that the two subspecies may differ at the species level. In a phylogenetic study on the Schistosomatidae, Lockyer et al. [45] considered that schistosomes in east Asia and their intermediate hosts in the Pomatiopsidae may be considered as the only co-evolutionary model between schistosomes and their intermediate host snails. Davis et al. [46] also speculated, as snail population forms have diverged genetically, so must their associated schistosomes or else become regionally extinct. However, it would be only possible to examine such relationship if the intermediate host snails and schistosomes are collected from a large geographical range in east Asia. In a very small-scale area in Anhui province of China, Rudge et al. [40] detected strong genetic differentiation in S. japonicum between two types of habitats, lake/marshland region and hilly region, and suggested that contrasting host reservoirs may be associated with the genetic differentiation, with rodents and dogs being important infection reservoirs in hilly regions and bovines in lake/marshland regions. On the other hand, they found little or no parasite genetic differentiation among host species within most villages; but in another study, Wang et al. [47] reported that schistosomes were separated into two clades representing the parasites from different definitive hosts. It seems likely that S. japonicum has undergone genetic differentiation in a relatively small-scale area, as in a large geographical region reported in this study. In the above two studies, miracidia from definitive hosts were examined with microsatellite markers. In the present study, adult parasites were obtained through infecting mice with cercariae. As definitive host-based genetic variation in S. japonicum has been noticed [40], [47], the selection pressure through definitive host may need to be further investigated. Unexpectedly, three haplotypes representing some schistosomes from three localities, each in Hubei, Hunan, Anhui provinces, were actually clustered together within another clade containing all haplotypes from Sichuan and Yunnan provinces. It is, however, at present impossible to explain this mixed cluster. As the movement of people has been frequent in China [48], the possible transmission through definitive host cannot be ruled out as a possible interpretation. In conclusion, substantial genetic diversity was demonstrated in populations of S. japonicum in schistosomiasis endemic areas in mainland China. Overall, a significant correlation was observed between the genetic distance and the geographical distance among the populations. It is apparent that the populations from Sichuan and Yunnan provinces in SW China exhibited a relatively low level of genetic variation, and were genetically different from the populations in the ML reaches of the Yangtze River, which had a much more complicated genetic diversity. Such obvious genetic diversity should be taken into consideration in guiding any strategic control programmes and/or vaccine development/trials in the future.

Despite the existing threat of schistosomiasis in some rural areas along the Yangtze River, the genetic diversity of Schistosoma japonicum has not been investigated across its wide geographical distribution in China, and such information may provide insight into the disease epidemiology and the development of its control measures. In this study, the adult parasites, obtained through infecting mice with cercariae from snails of the genus Oncomelania collected from a wide range of localities in currently endemic areas of schistosomiasis in the middle and lower (ML) reaches of the Yangtze River, and in Sichuan and Yunnan provinces in the upper reaches of the river in southwest (SW) China, were sequenced individually for mitochondrial genes. In general, a relatively high degree of genetic variation was observed in populations in the ML reaches in terms of nucleotide and haplotype diversity, but a low level was observed in populations in the SW. The significant difference in genetic diversity as revealed by F-statistics, and the existence of no shared haplotypes, were observed between populations in the ML reaches and those in the SW, indicating the effect of geographical separation/isolation upon the schistosomes and probably the parasite-snail system in China.

Abstract Introduction Materials and Methods Results Discussion

zoology ecology genetics biology genomics evolutionary biology population biology genetics and genomics

2012

Diversification of Schistosoma japonicum in Mainland China Revealed by Mitochondrial DNA

Integrating genetic perturbations with gene expression data not only improves accuracy of regulatory network topology inference, but also enables learning of causal regulatory relations between genes. Although a number of methods have been developed to integrate both types of data, the desiderata of efficient and powerful algorithms still remains. In this paper, sparse structural equation models (SEMs) are employed to integrate both gene expression data and cis-expression quantitative trait loci (cis-eQTL), for modeling gene regulatory networks in accordance with biological evidence about genes regulating or being regulated by a small number of genes. A systematic inference method named sparsity-aware maximum likelihood (SML) is developed for SEM estimation. Using simulated directed acyclic or cyclic networks, the SML performance is compared with that of two state-of-the-art algorithms: the adaptive Lasso (AL) based scheme, and the QTL-directed dependency graph (QDG) method. Computer simulations demonstrate that the novel SML algorithm offers significantly better performance than the AL-based and QDG algorithms across all sample sizes from 100 to 1,000, in terms of detection power and false discovery rate, in all the cases tested that include acyclic or cyclic networks of 10,30 and 300 genes. The SML method is further applied to infer a network of 39 human genes that are related to the immune function and are chosen to have a reliable eQTL per gene. The resulting network consists of 9 genes and 13 edges. Most of the edges represent interactions reasonably expected from experimental evidence, while the remaining may just indicate the emergence of new interactions. The sparse SEM and efficient SML algorithm provide an effective means of exploiting both gene expression and perturbation data to infer gene regulatory networks. An open-source computer program implementing the SML algorithm is freely available upon request. Genes in living organisms do not function in isolation, but may interact with each other and act together forming intricate networks [1]. Deciphering the structure of gene regulatory networks is crucial for understanding gene functions and cellular dynamics, as well as for system-level modeling of individual genes and cellular functions. Although physical interactions among individual genes can be experimentally deduced (e. g. , by identifying transcription factors and their regulatory target genes or discovering protein-protein interactions), such experimental approach is time-consuming and labor intensive. Given the explosive number of combinations of genes involved in any possible gene interaction, such an approach may not be practically feasible to reconstruct or “reverse engineer” gene networks. On the other hand, technological advances allow for high-throughput measurement of gene expression levels to be carried out efficiently and in a cost-effective manner. These genome-wide expression data reflect the state of the underlying network in a specific condition and provide valuable information that can be fruitfully exploited to infer the network structure. Indeed, a number of computational methods have been developed to infer gene networks from gene expression data. One class leverages a similarity measure, such as the correlation or mutual information present in pairs of genes, to construct a so-termed co-expression or relevance network [2], [3]. Another approach relies on Gaussian graphical models with edges being present (absent) if the corresponding gene pairs are conditionally dependent (respectively independent), given expression levels of all other genes [4], [5]. While the approach based on Gaussian graphical models entails undirected graphs, directed acyclic graphs (DAGs) or Bayesian networks have also been employed to infer the dependency structure among genes [6], [7]. The fourth approach employs linear regression models and associated inference methods to find the dependency among genes and to infer gene networks [8]–[11]. Finally, while these approaches use gene expression data in the steady-state, several methods exploiting time-series expression data have also been reported; see e. g. , [12], [13] and references therein. Recently, gene expression data from gene-knockout experiments have been combined with time series comprising gene expression data with perturbations to considerably improve the accuracy of network inference [14]. When a gene is knocked out or silenced, expression levels of other genes are perturbed. Different from using gene expression levels of the original network alone, comparing gene expression levels in the perturbed network with those in the original network reveals extra information about the underlying network structure. Gene perturbations can be performed with other experimental approaches such as controlled gene over-expression and treatment of cells with certain chemical compounds [8], [9]. However, these gene perturbation experiments may not be feasible for all genes or organisms. To overcome this hurdle, one can exploit naturally occurring genetic variations that can be viewed as perturbations to gene networks [15]. More importantly, such genetic variations enable inference of the causal relationship between different genes or between genes and certain phenotypes. Several approaches are available to capitalize on both genetic variations and gene expression data for inference of gene networks. The first approach models a gene network as a Bayesian network, and then infers the network by incorporating prior information about the network obtained from expression quantitative trait loci (eQTLs) [16]–[18]. In the second approach, a likelihood test is employed to search for a casual model that “best” explains the observed gene expression and eQTL data [19]–[23]. The third approach relies on the structural equation model (SEM) to infer gene [24]–[27] or phenotype networks [28]–[34]. While these approaches focus on inference of gene networks incorporating information from eQTL, another approach employs both phenotype and QTL genotype data to jointly decipher the phenotype network and identify eQTLs that are causal for each phenotype [35]. Logsdon and Mezey [26] proposed an adaptive Lasso (AL) [36] based algorithm to infer gene networks modeled with an SEM. They compared the performance of a number of methods using simulated directed acyclic or cyclic networks. Their simulations showed that the AL-based algorithm outperformed all other methods tested. Despite its superiority over other methods, the AL-based algorithm does not fully exploit the structure of the SEM. Therefore, it is expected that a more systematic inference algorithm may significantly improve the performance of the SEM-based approach. Motivated by the fact that gene networks or more general biochemical networks are sparse [8], [37]–[39], a sparse SEM is advocated in this paper to infer gene networks from both gene expression and eQTL data. Incorporating network sparsity constraints, a sparsity-aware maximum likelihood (SML) algorithm is developed for network topology inference. The core technique used is to maximize the likelihood function regularized by the -norm of the parameter vector determining the network structure. The -norm controls complexity of the SEM, and thus yields a sparse network. The key innovative element of the SML algorithm is a block coordinate ascent method derived to maximize the -regularized likelihood function, which makes the SML algorithm computationally efficient. The simulations provided demonstrate that the novel SML algorithm offers significantly better performance than the two state-of-the-art algorithms: the AL [26], and the QDG algorithm [21]. The SML algorithm is further applied to infer a human network of 39 human genes related to the immune function. Consider expression levels of genes from individuals measured using e. g. , microarray or RNA-seq. Let denote the vector collecting the expression levels of these genes of individual. Suppose that a set of perturbations to these genes has been also observed. These perturbations can be due to naturally occurring genetic variations near or within the genes, gene copy number changes, gene knockdown by RNAi or controlled gene over-expression. In this paper, focus is placed on genetic variations observed at eQTLs, although the network model and the inference method described in the next section are also applicable to cases where other perturbations are available. As in [26], it is assumed that each gene has at least one cis-eQTL so that the structure of the underlying gene network is uniquely identifiable. Let denote the genotype of eQTLs of individual. The goal is to infer the network structure of the genes from the available gene expression measurements, , and eQTL observations, . As in [25], [26], the gene network is postulated to obey the SEM (1) where matrix contains unknown parameters defining the network structure; matrix captures the effect of each eQTL; vector accounts for possible model bias; and vector captures the residual error, which is modeled as a zero-mean Gaussian vector with covariance, where denotes the identity matrix. It is assumed that no self-loops are present per gene, which implies that the diagonal entries of are zero. As mentioned in [26], lack of self-loops and a diagonal covariance matrix of are commonly assumed in almost all graph-based network inference methods. It is further assumed that the loci of eQTLs have been determined using an existing eQTL method, but the effective size of each eQTL is unknown. Therefore, has unknown entries whose locations are known and remaining zero entries (for instance is a diagonal matrix when). The network inference task is to estimate unknown entries of, and as a byproduct, the unknown entries of. Without any knowledge about the network, no restriction is imposed on the structure specified by. Therefore, the network is considered as a general directed graph that can possibly be a directed cyclic graph (DCG) or a DAG. Network inference is challenging since the number of unknowns to be estimated is very large for a moderately large. Note that under the assumption that each gene has at least one cis-eQTL, the “Recovery” Theorem in [26] guarantees that the network is identifiable for both DCGs and DAGs. As discussed in [8], [37]–[39], gene regulatory networks or more general biochemical networks are sparse meaning that a gene directly regulates or is regulated by a small number of genes relative to the total number of genes in the network. Taking into account sparsity, only a relatively small number of the entries of are nonzero. These nonzero entries determine the network structure and the regulatory effect of one gene on other genes. The SEM in (1) under the aforementioned sparsity assumption will be henceforth referred to as the sparse SEM. Exploiting the sparsity inherent to the network, an efficient and powerful algorithm for network inference will be developed in the ensuing section. Upon defining, , and, the SEM in (1) can be compactly written as, where 1 is the vector of all-ones. Given and, the log-likelihood function can be written as (2) where denotes matrix determinant, and denotes the Frobenius norm. As mentioned earlier, is a sparse matrix having most entries equal to zero. In order to obtain a sparse estimate of, the natural approach is to maximize the log likelihood regularized by the weighed term, where denotes the th entry of. In a linear regression model, it is well known that the -regularized least-squares estimation also known as Lasso [40] can yield a sparse estimate of the regression coefficient vector. Similarly, the -regularized maximum likelihood (ML) approach used here is expected to shrink most of the entries of toward zero, thereby yielding a sparse matrix. It is easy to show that maximizing with respect to (w. r. t.) yields, where and. Upon defining, , , , and substituting for in (2), the proposed -penalized ML estimation approach yields (3) where denotes the set of row and column indices of the entries of known to be zero. As assumed earlier, each phenotype has at least one cis-eQTL that has been identified, which implies that the locations of nonzero entries of or equivalently the set is known. However, our sparse SEM and inference method are also applicable to more general cases where some or all phenotypes have cis-eQTLs that have not been identified. In these cases, the locations of nonzero entries of corresponding to the unidentified cis-eQTLs are unknown. We can form a weighted -norm of the entries of excluding those corresponding to the identified cis-eQTL and then add a penalty term involving this -norm to the objective function in (3). This new optimization problem can be solved efficiently using a method modified from the one solving (3), as it is described in the supporting text S1. Weights in the penalty term are introduced to improve estimation accuracy in line with the AL [36]. They are selected as, where is found using a preliminary estimate of obtained via ridge regression as (4) The sparsity-controlling parameters in (3) and in (4) are selected via cross validation (CV), while is estimated as the sample variance of the error using and. In adaptive Lasso based linear regression [36], Zou suggested using the ordinary least squares (OLS) estimate to determine the weights; if the OLS estimate does not exist due to, e. g. , collinearity, Zou suggested the estimate obtained from ridge regression, although it remains to show if the ridge regression estimate is consistent in this case and if the resulting adaptive Lasso yields the desired oracle properties. If OLS is used for estimating and in the SEM, the solution usually does not exist since the number of unknowns is typically larger than the number of samples. However, even in this case the solution can always be obtained from ridge regression as in (4). Moreover, every entry of the solution is typically nonzero, which yields a finite weight for every variable, and thus every variable will be included in the following -penalized ML procedure. An alternative approach is to replace the weighed -norm in (3) with an unweighted -norm to obtain a preliminary estimate of and then calculate the weights from this preliminary estimate, as in [26]. However, the unweighted -penalized ML procedure may shrink many variables to zero and exclude them from the weighted -penalized ML estimator, possibly yielding a biased estimate. For this reason, the inference method in this paper uses ridge regression to determine, with the additional advantage of (4) admitting a closed-form solution. A block diagram of the novel inference algorithm, abbreviated as the sparsity-aware maximum likelihood (SML) algorithm, is depicted in Figure 1. The first and third blocks in Figure 1 perform cross-validation to select optimal parameters and to be used in (3) and (4), respectively (see the description of the cross-validation procedure in the supporting text S1.) The third block produces weights and error-variance estimate after solving (4). Finally, the fourth block takes data and together with, and and solves (3) to yield, representing the SML estimator for in (1) and revealing the genetic-interaction network. As it will be described in the Methods section, (4) is separable across rows of and, and each row of and becomes available in closed form [cf. (8) – (9) ]. The -regularized ML problem (3) is solved efficiently using a novel block coordinate ascent iterative scheme given by (11) – (16) in the Methods section. Precise description of the overall SML algorithm is also presented in the Methods section as Algorithm 1, which was used to yield an executable computer program. In their simulation studies, Logsdon and Mezey [26] compared the performance of their AL-based algorithm with that of several other algorithms including the PC-algorithm [41], [42], the QDG algorithm [21], the QTLnet algorithm [35], and the NEO algorithm [22]. In two out of four simulation setups, the AL outperformed all other algorithms; and in the other two simulation setups, the AL and QDG algorithms exhibited comparable performance, but consistently outperformed the other two algorithms. Logsdon and Mezey [26] also considered other existing algorithms [25], [43], but these were deemed either computationally too demanding [43] or prohibitively complex [25]. For these reasons, the AL and QDG algorithms are regarded as state-of-the-art in the field. Their performance was compared against this paper' s SML algorithm. Following the setup of Logsdon and Mezey [26], two types of acyclic gene networks were simulated first: one with 10 genes and another with 30 genes. Specifically, a random DAG of 10 or 30 nodes with an expected edges per node was generated by creating directed edges between two randomly picked nodes. Care was taken to avoid any cycle in the simulated graph. If an edge from node to node was emerging, was generated from a random variable uniformly distributed over the interval or; otherwise, . The genotype per eQTL was simulated from an F2 cross. Values 1 and 3 were assigned to two homozygous genotypes, respectively, and 2 to the heterozygous genotype. Hence, was generated as a ternary random variable taking values with corresponding probabilities. Matrix was the identity matrix, was sampled from a Gaussian distribution with zero mean and variance, and was set to zero. Finally, was calculated from For each type of gene network, 100 realizations or replicates of the network were generated, and then the SML, the AL and the QDG algorithms were run to infer the network topology. When running the SML algorithm, 10-fold CV was employed to determine the optimal values of parameters and and then use these values to infer the network. An edge from gene to was deemed present if. The AL algorithm also automatically ran using CV to determine the values of its parameters. For 100 replicates of the network, counted the total number of edges, denoted the total number of edges detected by the inference algorithm. Among detected edges, stands for the number of true edges presented in the simulated networks, and for the number of false edges. The power of detection (PD) was then found as, and the false discovery rate (FDR) as. The PD and the FDR of the SML, AL, and QDG algorithms for different sample sizes are depicted in Figure 2. It is seen from Figures 2 (a) and (c) that the PD of the SML algorithm exceeds 0. 9 for both networks across all sample sizes, whereas the PD of the AL algorithm is about 0. 65 for and 0. 35 for. The PD of the QDG algorithm is even lower ranging from 0. 22 to 0. 33. As shown in Figures 2 (b) and (d), the FDR of the SML algorithm is on the order of for most sample sizes, and is much lower than that of the AL and QDG algorithms, which is about 0. 3 for and over the range from 0. 31 to 0. 6 for. Two types of cyclic networks were subsequently simulated: one with 10 genes and the other with 30 genes. The average number of edges per gene is again equal to 3. The same procedures used in simulating acyclic networks described earlier were employed, except that DCGs instead of DAGs were simulated. Again, 100 replicates for each type of the networks were randomly generated. The PD and the FDR of three algorithms are depicted in Figure 3. As shown in Figure 3 (a) and (c), the PD of the SML algorithm is between 0. 83 and 0. 9, whereas the PD of the AL algorithm is about 0. 52 for and 0. 29 for, and the PD of the QDG algorithm is between 0. 16 and 0. 28. As shown in Figures 3 (b) and (d), the FDR of the SML algorithm is, which is much smaller than that of the AL and QDG algorithms over the range from 0. 33 to 0. 68. For the convenience of comparison, the results in Figures 2 and 3 at sample size 500 are summarized in Table 1. As confirmed by Figures 2 and 3, the SML algorithm offers much better performance in terms of PD and FDR than the AL and QDG algorithms. However, these results were obtained for gene networks of small size. To test performance of the SML algorithm for networks of relatively large size, an acyclic network of 300 genes was simulated with an expected edge per node, and 10 replicates of the network were randomly generated. PD and FDR of the SML and AL algorithms obtained from these replicates are depicted in Figure 4. The PD of SML exceeds across all sample sizes from 100 to 1,000, whereas that of the AL algorithm is about 0. 04 for sample sizes from 100 to 500, and gradually increases to 0. 42 at the sample size of 1,000. The FDR of SML stays below for sample sizes from 400 to 1,000, whereas the FDR of the AL algorithm is on the order of for the same sample size. When the sample size is relatively small (in the range from 100 to 300), the FDR of SML is higher than that of the AL algorithm, but it is still relatively small (). Note that the AL algorithm essentially does not work for sample sizes, since its power is too small. All simulation results show that the novel SML algorithm significantly outperforms the AL and QDG algorithms in terms of PD and FDR. An extra set of simulations assessing the stability of SML is described in the section of “Stability of model selection under CV with different folds” in supporting text S1, and in Figures S1 and S2. As an alternative to CV, stability selection (STS) [44] provides a means of selecting an appropriate sparsity level to guarantee that the FDR is less than a theoretical upper bound. The STS procedure was applied to the SML algorithm as described in the supporting text S1, and was used with the selection probability cutoff and an upper bound or target FDR = 0. 1 in simulations for the networks in Figures 2[ (c) and (d) ] and 3 [ (c) and (d) ]. As shown in Figure S3, the FDR of the STS is indeed much smaller than the target FDR and almost uniform across different sample sizes, but the PD of the STS is smaller than that of CV. In fact, the FDR of the STS is on the same order as that of the CV except at the sample size of 100 for the DAG. As seen from these simulation results, although the STS guarantees a FDR upper bound, this upper bound is loose for the simulation setups tested, which may sacrifice detection power. Nevertheless, the STS procedure can select a set of stable variables as described in [44] and verified by our simulations. So far, all the simulated data were generated with noise variance. Next, the performance of SML was analyzed for simulated networks of 30 genes, when was increased to 0. 05 and was changed from 3 to 1 or 5. Reducing from 3 to 1 improved the performance of SML for most of the sample sizes, as it is depicted in Figure 5, withstanding the increase in the noise variance. Increasing at constant, or increasing at constant degraded the performance, most notably in the later case. Comparing Figure 5 with Figures 2 and 3 [ (c) and (d) ] demonstrates that in both cases the SML estimates still achieve higher detection power and lower FDR than those estimates obtained with the AL algorithm for and. Pickrell et al. [45] used RNA-Seq technology to sequence RNA from 69 lymphoblastoid cell lines derived from unrelated Nigerian individuals extensively genotyped by the International HapMap Project [46]. For each gene, they evaluated possible associations between its gene expression level calculated from RNA-Seq reads and all 3. 8 million single nucleotide polymorphisms (SNPs) using the genotypes from phases II and III of the HapMap Project. At FDR = 0. 1, they identified 929 genes or putative new exons that have eQTLs within 200 kb of the gene or the exon. From these 929 genes, 39 genes that are related to immune functions were selected manually by an expert as mentioned in the Acknowledgements section; expression levels and the genotypes of the eQTLs of these 39 genes in 69 individuals were used to infer the underlying regulatory network. Pickrell et al. normalized expression values using quantile normalization before performing eQTL mapping. They also provided a data set that contains the number of reads mapped to each of 929 genes. This data set was obtained and the number of reads for each of 39 genes was normalized with the length of the gene to yield expression value. Such kind of values may better reflect the real expression values than the values normalized with quantile normalization, and thus they were used to infer the network. To ensure the quality of the data, the SAS ROBUSTREG procedure was applied to 69 expression values of each of 39 genes to detect outliers. The default M estimation method of the ROBUSTREG procedure was employed and the outliers were detected at a significance level of 0. 05. Several gene expression values were identified as outliers since they are much larger than the remaining values that were classified as non-outliers. The outliers were replaced with the largest non-outlier. More sophisticated means of revealing and imputing outliers are possible using robust statistical schemes; see e. g. , [47]. The genotypes of the eQTLs of the 39 genes were downloaded from HapMap database using the SNP IDs for the eQTL provided by Pickrell et al. . About 12% genotypes are missing. These missing genotypes were imputed using the program IMPUTE2 [48]. The name and a brief description of each gene were obtained from DAVID [49] using the Ensembl gene IDs provided by Pickrell et al. Information of these 39 genes including their Ensembl gene IDs and names, a brief description of each gene, and HapMap SNP IDs of the associated eQTLs can be found in Table S1 in the supporting information. The SML algorithm was run with the expression levels and genotypes of eQTLs of these 39 genes. An edge from gene to was detected if. To improve the reliability of the detected edges, the SML algorithm was run with stability selection at an FDR using 100 random subsamples, yielding 13 directional edges as shown in Figure 6. The frequency of each edge detected in 100 runs is given in Table S2. It is interesting to see from Figure 6 that only 9 genes are involved in the network, and the remaining 30 genes are not connected with any other genes and thus not shown in the figure. AL and QDG algorithms were also run with stability selection at an FDR using 100 random subsamples. The edges detected by AL and QDG algorithms and their frequencies are included in Table 4. The AL algorithm detected only one edge that was not detected by the SML algorithm. The QDG yielded 3 edges, one of which was also detected by the SML algorithm. The relatively small number of edges detected by three algorithms was likely due to relatively low signal-to-noise ratio (SNR) in this data set. The estimated noise variance was and the estimated SNR was dB, which was much lower than that (about 25. 8 dB) in the case of in Figure 5. However, comparing the results of three algorithms shows that our SML algorithm detected more edges than the other two algorithms at the same FDR due to its higher detection power as confirmed also by the simulations. When the FDR was increased to, the SML algorithm with stability selection yielded a network of 16 genes that have 42 edges as shown in Figure S4 in the supporting information. Since only 39 genes were used to construct the network, an edge between two genes may not necessarily imply a direct regulatory effect, but may reflect the fact that two genes are either directly linked or very close to each other in the real network that consists of all genes. Particularly, if two genes are co-regulated by another gene which is not included in the 39 genes, these two genes may have a unidirectional or bidirectional edge. Most edges in Figure 6 are between major histocompatibility complex (MHC) genes (HLA-A, HLA-DPA1, HLA-DQA2, HLA-DQB1, HLA-DRB4 and HLA-DRB5), which is expected since these genes may interact with each other and/or be co-regulated. FCRLA is a member of Fc receptor-like family of genes. It is expressed in B cells and interacts with IgG and IgM [50], [51]. IGH, encoding the heavy chain of immunoglobulin, characterizes the B-cell origin of the samples. Hence, it is not surprising to see an edge between FCRLA and IGH. Interleukin-4-induced gene 1 (IL4I1) was first described in the mouse [52] and subsequently characterized in human B cells [53]. Human IL4I1 is expressed by antigen-presenting cells [54], which may allude to the edge between HLA-A and IL4I1, but this may be speculative since there is no edges between IL4I1 and MHC class II genes in the network. The edges between IGH and HLA-A and between IGH and HLA-DRB4 may reflect the coordinated effect of antibody and MHC as a response to antigens. In fact, IGH is connected to most of MCH genes in Figure S4, which may imply the wide coordination between the two classes of molecules. Integrating genetic perturbations with gene expression data for inference of gene networks not only improves inference accuracy, but also enables learning of causal regulatory relations among genes. Although much progress has been made recently on the development of inference methods that integrate both types of data, a truly efficient algorithm is missing. The SEM provides a systematic framework to integrate both types of data, and offers flexibility to model both directed cyclic as well as acyclic graphs. However, there is no systematically designed inference method for SEMs of relatively high dimension, which is particularly true for gene networks typically including hundreds or thousands of genes. Traditionally, inference for SEMs has relied on the ML or generalized least-squares methods implemented with a numerical optimization algorithm [55], [56]; but recently, Bayesian alternatives [57] have emerged too, based on Markov chain Monte Carlo simulations [58], [59]. These methods not only are computationally intensive, but also may be inaccurate for sparse SEMs of relatively high dimension, since they do not account for sparsity present in the model. In the context of QTL mapping, Newton' s method is employed in [27] to implement the ML method, while the genetic algorithm [60], [61] is used in [24], [25] to maximize the likelihood function, and in conjunction with a model selection method using a test or Occam' s window to search for the best network topology. These methods are not scalable to SEMs of relatively high dimension. The AL-based algorithm proposed in [26] is more efficient because it automatically incorporates variable selection into the inference process, and also takes into account the sparsity present in gene networks. However, the AL-based scheme borrows the adaptive Lasso [36] optimally designed for the linear regression model instead of the SEM. In contrast, the SML algorithm proposed in this paper directly maximizes the -regularized likelihood function of the SEM, which fully exploits the information present in the data and therefore improves inference accuracy. Moreover, the novel block coordinate ascent method combined with discarding rules can efficiently maximize the -regularized likelihood function, rendering the SML algorithm applicable to SEMs of high dimension. However, unlike the AL-based algorithm, the SML algorithm maximizes a non-convex objective function as given in (3). Although the “Recovery” Theorem in [26] guarantees the identifiability of the network, the algorithm can converge to a local maximum that may not necessarily be coincident with the global maximum corresponding to the optimal network. A common technique for alleviating this problem is to use multiple random initial values. We tested multiple initial values in our simulations and observed that the algorithm converged to the same solution. In Algorithm 2, we used the pathwise coordinate optimization strategy as used in [62], where the solution of (3) obtained with was used as the initial point for the run with. The pertinence of this strategy is corroborated by simulated numerical tests, showing significant performance gains of the SML algorithm in terms of detection power and FDR when compared to the AL-based algorithm. Comparisons in the Simulation Studies section, as summarized in Figures 2–5, demonstrated that the SML algorithm markedly outperforms two state-of-the-art algorithms: the AL [26] and QDG [21] algorithms. For three directed acyclic networks with number of genes and 300, respectively, the PD of the SML algorithm exceeds 0. 9 for all sample sizes from 100 to 1,000, and is greater than 0. 99 for most sample sizes. This is much greater than the PD of the AL and QDG algorithm that ranges from 0. 004 to 0. 67. In fact, The QDG algorithm was too time-consuming to obtain results for. The FDR of SML is on the order of for most sample sizes, which is much smaller than those of the AL and QDG algorithms, that are between 0. 25 and 0. 6 for and 30. The FDR of the AL algorithm for is between 0. 02 and 0. 1. The only case where the FDR of SML exceeds that of the AL algorithm is when, and the sample size. However, the AL algorithm essentially does not work in this case, since its PD is about 0. 04. In the case of directed cyclic networks, all algorithms offer slightly degraded performance when compared to that of directed acyclic networks. However, the SML algorithm still considerably outperforms the AL and QDG algorithms. Using a limited amount of available data [45], 39 genes related to the immune system and having one eQTL per gene were selected to infer a possible network among these genes. At an FDR 10% for the detected edges, a network of 9 out of 39 genes containing 13 edges were obtained. An edge between two genes in the inferred network may be an indication of the direct regulator effect, or indirect interaction or co-regulation mediated by some other genes that are not among the 39 genes. The majority of the edges were reasonably expected from the experimental results in the literature, while the remaining edges may represent new interactions to be elucidated. Structural equation modeling has a long history of about a century, with well-documented contributions to various fields including biology, psychology, econometrics and other social sciences [55], [56], [63], [64]. The model considered in this paper belongs to a class of SEMs with observed variables [55]. The SML algorithm is the first one that is systematically developed for inferring sparse SEMs with observed variables. It is expected to accelerate the application of high-dimensional SEMs not only in biology, but also in other fields. The overall SML approach described in the Methods section, including the ridge regression weights, the discarding rules, and the coordinate descent cycle is depicted step-by-step in Algorithm 1. The for-loop starting from line 8 and ending at the last line is the -regularized ML method for computing and in (3), which comprises the block coordinate ascent algorithm and discarding rules. In our computer program, these lines were written as a subroutine. Since the CV on line 7 needs to solve (3), the subroutine is also called on line 3 with varying from to. An additional subroutine implementing ridge regression was written to solve (4), and subsequently called on lines 1 and 2. In the supporting text S1, three relevant extensions to the SML algorithm are described. First, stability selection [44] is applied to the SML, as an alternative to CV, to select the sparsity level so that the FDR is controlled. Second, the SML is extended to handle heteroscedasticity in the SEM error. Third, the SML is modified to enable inference of unknown eQTLs. In addition, supporting text S1 gives a description of the state-of-the-art AL-based and QDG algorithms that were considered for comparison with SML.

Deciphering the structure of gene regulatory networks is crucial for understanding gene functions and cellular dynamics, as well as system-level modeling of individual genes and cellular functions. Computational methods exploiting gene expression and other types of data generated from high-throughput experiments provide an efficient and low-cost means of inferring gene networks. Sparse structural equation models are employed to: i) integrate both gene expression and genetic perturbation data for inference of gene networks; and, ii) develop an efficient sparsity-aware inference algorithm. Computer simulations corroborate that the novel algorithm markedly outperforms state-of-the-art alternatives. The algorithm is further applied to infer a real human gene network unveiling possible interactions between several genes. Since gene networks can be perturbed not only by genetic variations but also by other means such as gene copy number changes, gene knockdown or controlled gene over-expression, this paper' s method can be applied to a number of practical scenarios.

Abstract Introduction Results Discussion Methods

regulatory networks biology computational biology

2013

Inference of Gene Regulatory Networks with Sparse Structural Equation Models Exploiting Genetic Perturbations

Molecular interactions between male and female factors during mating profoundly affect the reproductive behavior and physiology of female insects. In natural populations of the malaria mosquito Anopheles gambiae, blood-fed females direct nutritional resources towards oogenesis only when inseminated. Here we show that the mating-dependent pathway of egg development in these mosquitoes is regulated by the interaction between the steroid hormone 20-hydroxy-ecdysone (20E) transferred by males during copulation and a female Mating-Induced Stimulator of Oogenesis (MISO) protein. RNAi silencing of MISO abolishes the increase in oogenesis caused by mating in blood-fed females, causes a delay in oocyte development, and impairs the function of male-transferred 20E. Co-immunoprecipitation experiments show that MISO and 20E interact in the female reproductive tract. Moreover MISO expression after mating is induced by 20E via the Ecdysone Receptor, demonstrating a close cooperation between the two factors. Male-transferred 20E therefore acts as a mating signal that females translate into an increased investment in egg development via a MISO-dependent pathway. The identification of this male–female reproductive interaction offers novel opportunities for the control of mosquito populations that transmit malaria. In many organisms, male–female molecular interactions occurring during sex shape reproductive success and may drive the rapid evolution of reproductive phenotypes [1]. While in species where females mate multiple times these reproductive interactions are often antagonistic due to the different reproductive strategies utilized by males and females 2–5, in monandrous species—that is, species where females mate a single time—they are believed to benefit both sexes [6]. Indeed this hypothesis has been proven experimentally in Drosophila melanogaster: removing sexual selection in this naturally promiscuous species through “imposed” monogamy induced the evolution of less antagonistic traits, where males became less harmful and females less resistant to induced harm [7]. In the malaria mosquito Anopheles gambiae, females rarely mate more than once during their lifetime [8]. As yet unknown male–female molecular interactions occurring during this single copulation regulate a series of postmating events that profoundly change the physiology and behavior of females. While in copula, females receive sperm, which are stored in a dedicated store organ named the spermatheca, and seminal secretions produced by the male accessory glands (MAGs). MAG secretions coagulate during mating to form a gelatinous mating plug that is transferred to the uterus (atrium), where it is digested in 1–2 d [9], [10]. Following this copulation event, blood-fed females increase their egg production [11] and start laying eggs [12], [13]. The regulation of egg production in A. gambiae is a particularly intricate process that depends on two main signals: one derived from blood feeding and one triggered by mating. While all females need to feed on blood to develop eggs, virgins in general have a pregravid state where they require two or more consecutive feedings to complete the first gonotrophic cycle 14–16. This has profound implications for malaria transmission, as it increases the likelihood of contact with the human host. Pregravid behavior may be caused by insufficient metabolic reserves at emergence due to nutritional deprivation during larval stages [14], [17]. This, in turn, may drive the need to optimize resource allocation between highly energy-demanding processes like flight and reproduction [18]. Indeed smaller A. gambiae mosquitoes tend to produce fewer eggs [19], [20] and appear to feed as virgins [21], perhaps to build up energy reserves for mating. The cascade of events triggered by blood feeding and leading to egg development, partially described in A. gambiae [22], has been well characterized in another mosquito species, the yellow fever and dengue vector Aedes aegypti. In these mosquitoes, after a blood meal the ovaries are released from their previtellogenic arrest and start vitellogenesis, the process of synthesis and secretion of yolk protein precursors (YPPs) by fat body cells. Upon secretion into the hemolymph, the YPP Vitellogenin (Vg) and the lipid transporter Lipophorin (Lp) become internalized into the ovaries via receptor-mediated endocytosis [23], [24], leading to the maturation of 50–150 oocytes in approximately 2–3 d (reviewed in [25]). The transcription of YPPs is under endocrine regulation. After blood feeding the brain-secreted ovarian ecdysteroidogenic hormone (OEH) stimulates the ovaries to produce the steroid hormone ecdysone (E) [26], [27], which in turn is hydroxylated into 20-hydroxy-ecdysone (20E) by the fat body cells. 20E synthesis releases the state of arrest of the fat body, activating the transcription of YPPs [25], [28]–[30] by binding to the nuclear hormone receptor heterodimer Ecdysone Receptor (EcR) /Ultraspiracle (USP), prompting it to function as a transcriptional activator [31]. A similar role of 20E in vitellogenesis after blood feeding has been demonstrated also in A. gambiae [22], where titers of 20E in blood-fed females correlated to Vg expression, suggesting a conservation of this pathway between Anopheles and Aedes mosquitoes. No information is instead available on the factors regulating the mating-induced stimulation of oogenesis observed in A. gambiae. Mating increases the rate of egg production in a number of insects, and in some cases this effect has been attributed to the transfer of MAG secretions (reviewed in [32]). The D. melanogaster Sex Peptide increases production of YPPs and oocyte maturation by inducing the female corpora allata to synthetize the sesquiterpenoid Juvenile Hormone III-bisepoxide (JHB3) [33]–[35]. In Photinus fireflies, seminal secretions translocated to ovaries positively influence female fecundity [36]. In mosquitoes, a role of MAG products in egg development has been suggested by a number of studies where injections of MAG extracts into the hemolymph of Aedes females stimulated Vg synthesis and/or oogenesis [37]–[40]. In A. gambiae, indirect evidence suggests that MAG secretions act as master regulators of female postcopulatory behavior and physiology [41]–[45]. Thus far more than 100 A. gambiae MAG genes have been identified [46], [47], and a number of them encode proteins that are packaged in the mating plug and transferred to females [10]. The A. gambiae MAGs, so far uniquely among mosquitoes, also produce large amounts of 20E [48], and delivery of this potent regulator of gene expression during sex may at least partly explain the vast transcriptional response that females undergo after mating [49]. This hypothesis is strengthened by the finding that among the genes regulated by mating is the 20E-responsive gene Vg, which is strongly induced in the female reproductive tract at 6 h after copulation [49]. Here we show that the 20E steroid hormone produced by the male and transferred to the female reproductive tract during copulation triggers a series of molecular events leading to the increased egg production observed in blood-fed A. gambiae mosquitoes after mating. We identify an atrial-specific Mating-Induced Stimulator of Oogenesis (MISO) that is regulated by and interacts with 20E. This interaction translates the male hormonal signal into an increased expression of a major vitellogenic lipid transporter, facilitating oocyte development via the accumulation of lipids in the ovaries. Our previous studies had identified a gene (AGAP002620, henceforth referred to as MISO) that is highly upregulated in the atrium during the first day after mating [49]. This gene encodes a glycine-rich protein of 152 aminoacids with no known functional domains. After confirming the atrium-specific, mating-induced expression of this gene (Figure S1A, B), we decided to examine whether MISO is involved in the regulation of two female postmating responses, oogenesis and oviposition. Consistent with a possible role in these processes, immunofluorescence and confocal microscopy analyses on virgin and mated atria at 12 and 24 h postmating (hpm) identified the protein in the ampullae, the tissues that connect the anterior part of the atrium to the oviducts (Figure S1C). To study the function of MISO, we performed RNA interference (RNAi) –mediated gene silencing by injecting females with double-stranded RNAs (dsRNAs) targeting this gene (dsMISO) (transcript mean reduction = 74. 4%±19. 9%, one-sample t test: t14 = 14. 45, p<0. 0001, range 95%–31%; this knock-down completely abolished protein expression; Figure S1D, E). When injected females were mated, blood-fed, and allowed to lay eggs, a higher proportion of dsMISO females did not oviposit (29 out of 125,23%) compared to control females injected with an unrelated control dsRNA (dsLacZ) (13 out of 138,9%) (χ2 = 9. 281, p = 0. 0023) (Table S1). Additionally, females injected with dsMISO laid a significantly smaller number of eggs (dsLacZ, 82. 5 eggs; dsMISO, 65. 4 eggs; Poisson regression, χ2 = 236. 6, p<0. 0001) (Figure 1A). Dissection of the ovaries from both groups, however, revealed that this difference was due to a larger proportion of dsMISO females (16%) failing to develop eggs compared to controls (4%) (χ2 = 11. 68, p = 0. 0006) (Table S1). The percentage of females with fully developed ovaries that did not oviposit was instead similar in both groups (dsLacZ, 6%; dsMISO, 9%; χ2 = 0. 5781, p = 0. 4470), suggesting that MISO is important for egg development rather than for egg laying. No difference was detected in the fertility of the two groups (dsLacZ, 97%, n = 125; dsMISO, 96%, n = 96; Mann–Whitney test, U = 5089, p = 0. 3985) (unpublished data). To further investigate the impact of MISO on oogenesis, dsRNA-injected females were mated and blood fed, and 3 d later, when oogenesis is normally completed, the ovaries of fully engorged females were dissected without allowing egg laying, and eggs were counted. Virgin dsLacZ females were included as controls to verify that under our experimental conditions virgins are less likely to produce eggs than mated females, as demonstrated by others [11]. Control females showed a 20% increase in egg production after mating (mated dsLacZ, 77. 8 eggs; virgin dsLacZ, 62. 3 eggs) (Figure 1B); however, this increase in egg production was completely abolished in mated dsMISO females (mated dsMISO, 60. 4 eggs) (Figure 1B) (Poisson regression, χ2 = 306. 6, p<0. 0001; Bonferroni multiple comparison post hoc test: virgin dsLacZ versus mated dsLacZ, p = 0. 002; mated dsLacZ versus mated dsMISO, p<0. 03; virgin dsLacZ versus mated dsMISO, p>0. 05). Silencing of MISO before copulation therefore decreased egg development to levels observed in virgins, suggesting that this gene is required for the increase in oogenesis observed in A. gambiae females after mating. After assessing the role of MISO in determining the increase in oogenesis induced by mating, we next analyzed the progression of oocyte development in mated dsMISO and control virgin and mated females at two time points (24 h and 60 h) after a blood meal. At 24 h postblood feeding, dsMISO follicles showed delayed development compared to mated dsLacZ controls, similar to what observed in the ovaries of virgin dsLacZ females (Figure 2A). By 60 h postblood feeding, oogenesis was completed in all three groups (Figure 2B); however, dsMISO (and virgin dsLacZ) ovaries showed a number of undeveloped primary follicles (indicated by asterisks in Figure 2B) in agreement with the finding that MISO silencing reduces egg development. A time course of five time points (12,24,36,48, and 60 h) after blood feeding in virgin and mated females confirmed that, similar to virgin dsLacZ controls, mated dsMISO females exhibited a statistically significant delay in egg development, and only achieved oocytes of the size exhibited by mated dsLacZ individuals at 60 h postblood feeding (Figure S2 and Table S2). These results suggest that the effects of MISO on egg development are due to delayed or impaired accumulation of lipids into the growing oocytes. The effects on oocyte growth observed in dsMISO females prompted us to analyze whether MISO plays a role in regulating the lipid transport to the oocyte. We therefore analyzed the expression levels of the vitellogenic lipid transporter Lp (AGAP001826) and the major YPP Vg (AGAP004203) in the fat body of blood-fed females at their peak of expression. In five different experiments, Lp transcript levels at 24 h after blood feeding were strongly reduced (54% mean reduction) in mated dsMISO females compared to mated controls, similar to virgin control levels (50% mean reduction) (Figure 2C) (Repeated Measures ANOVA, F2,4 = 8. 142, p = 0. 0118; Tukey' s Multiple Comparison post hoc test: virgin dsLacZ versus mated dsLacZ, p<0. 05; mated dsLacZ versus mated dsMISO, p<0. 05; virgin dsLacZ versus mated dsMISO, p>0. 05). Vg instead was not significantly affected by MISO silencing (Repeated Measures ANOVA, F2,4 = 1. 362, p = 0. 3098) (Figure 2C). Taken together, these results indicate that mating increases the blood feeding–induced expression of Lp and that this regulation is dependent on MISO. In A. aegypti mosquitoes, the expression of the Vg and Lp after blood feeding is induced by the function of 20E produced by the female [29], [30], and a similar regulation occurs also in A. gambiae [22]. We tested whether the MISO-mediated upregulation in the expression of Lp in mated females after a blood meal was caused by an increased production of this hormone. We measured ecdysteroid levels secreted in vitro by the ovaries of virgin and mated dsLacZ and mated dsMISO females before and 18 h after a blood meal, at their peak of secretion [48]. As expected, blood feeding strongly increased the steroidogenic capacity of the ovaries (Figure 3A) (one-way ANOVA, F5,42 = 11. 17, p<0. 0001; post hoc Tukey' s multiple comparison, non–blood-fed versus blood-fed groups, p<0. 01). However, no differences between virgin and mated females were observed, and silencing of MISO did not affect ecdysteroid secretion levels (p>0. 05) (Figure 3A). Besides being produced by the female after blood feeding, in A. gambiae 20E is also synthesized in the MAGs and transferred to females during mating [48]. We therefore hypothesized that sexually transferred 20E may play a role in the MISO-mediated regulation of female physiology after mating. As a first step, we determined that the MAG-produced 20E is transferred to the female as part of the mating plug (Figure S3A). By 12 hpm, 20E localization was restricted to the anterior portion of the plug that is enclosed within the ampullae (Figure S3A), where MISO also localizes (Figure S1C). The amount of 20E detected in the MAGs corresponded to a mean of 632 pg (±17 pg), consistent with previous findings by others (Figure S3B) [48]. Interestingly, no 20E could be detected in the male reproductive tissues of two mosquito species, Anopheles albimanus and A. aegypti, which do not produce mating plugs (Figure S3B). We next investigated whether MISO affects the activity of 20E transferred by males during copulation. To this aim, we analyzed steroid hormone levels in the atria of dsLacZ and dsMISO females at five time points after mating (0. 5,6, 12,18, and 24 hpm) to monitor 20E release from the mating plug over time. Immediately after mating (0. 5 hpm), the atria of control and dsMISO females contained similar hormone titers (Figure 3B). Ecdysteroid levels in the atria of controls were statistically significantly decreased at the four later time points (Wilcoxon test, p<0. 001) and reached about 3 pg per individual by 24 hpm, suggesting that 1 d after copulation the steroids have been fully released from the mating plug and have circulated out of the atrium. Interestingly, ecdysteroid titers declined more slowly in the atria of dsMISO females (P-mixed effects model, p = 0. 055) (Figure 3B). No 20E was detected in the atria of virgin females (unpublished data), confirming that this hormone in the female is only produced after blood feeding. These results suggest that silencing of MISO impairs the release of ecdysteroids from the plug and/or their diffusion from the atrium, possibly affecting their function. To confirm the latter hypothesis, we analyzed the transcription levels of five 20E-responsive genes at three time points after mating (6,12, and 18 hpm) in the two RNAi-injected groups. If MISO impairs the release of 20E from the atrium, then the expression levels of these genes in surrounding tissues should be altered in dsMISO females compared to controls. Besides Vg and Lp [28], [29], we analyzed Ecdysone Receptor (EcR, AGAP012211) [31], Ultraspiracle (USP, AGAP002095) [50], [51], and Hormone Receptor 3 (HR3, AGAP009002) [52]. As mentioned above, EcR is a nuclear receptor that in conjunction with USP activates transcription of downstream genes upon binding of 20E [31], [50], [51], while HR3 is known to interact directly with EcR [52]. Three genes exhibited a significant reduction in postmating expression in dsMISO females over the time frame analyzed: HR3 was downregulated by 50% at 6 hpm (t test, t6 = 2. 431, p = 0. 0256), Vg was reduced by 54% at 12 hpm (t test, t6 = 2. 785, p = 0. 0159), while EcR was decreased by 44% at 18 hpm (t test, t6 = 1. 876, p = 0. 0587) (Figure 3C). The expression levels of Lp and USP did not significantly differ between control and experimental females (Figure 3C). All together, these results show that MISO silencing impairs both the titers of 20E in the atrium and the expression of 20E-responsive genes after mating, reinforcing the hypothesis that MISO influences the function of male-derived ecdysteroids delivered by the mating plug. We next investigated whether the effects of MISO silencing on 20E titers and on the expression of 20E-responsive genes were caused by a possible interaction between MISO and 20E. To this aim, Western blot analyses were performed under native (i. e. , nondenaturing) conditions. An anti-20E antibody detected a band of approximately 40 kDa in the atria of mated female (8 hpm) that was not detected in virgin extracts (Figure 4A). This band reacted also with anti-MISO antibodies, suggesting that the two factors are part of the same complex (Figure 4A). Moreover, immunoprecipitation of MISO in extracts of virgin and mated atria at 8 hpm followed by an ELISA coupled with anti-20E antibodies detected significant amounts of 20E co-immunoprecipitating in mated females, while no signal was observed in virgins (Figure 4B). All together these results suggest an interaction between MISO and 20E in the atrium of females after mating. As 20E is known to regulate the expression of genes that are ultimately responsible for its function (reviewed in [53]), we next analyzed whether this steroid hormone plays a role in the expression of MISO in the atrium. To this aim, we injected three 10-fold dilutions of 20E into the hemolymph of virgin females, and analyzed MISO transcript levels specifically in the atrium (where the gene is not normally expressed in virgin females) at 24 h postinjection. At the highest concentration, 20E significantly induced MISO expression to levels similar to those achieved by mating (178- and 349-fold induction, respectively) (one-way ANOVA, F6,23 = 14. 79, p<0. 0001; post hoc Dunnett' s multiple comparison against virgins, p<0. 01), while the ethanol and cholesterol controls had no effect (Figure 4C). At lower dilutions, 20E injections increased MISO expression levels relative to controls, however this effect was not statistically significant. No effect on MISO expression was seen in tissues other than the atrium, confirming the tissue-specific restriction of expression of this gene (unpublished data). The expression of AGAP009584, an atrial gene that is not modulated by mating [10], [49], was not induced by the injection of any of the 20E dilutions (Figure 4C) (one-way ANOVA, F6,23 = 0. 5089, p = 0. 7947). Only the highest concentration of injected 20E achieved physiological atrial concentrations similar to those transferred during mating (Figure S4), explaining the observed titration-dependent upregulation of MISO expression. Finally, to further confirm that 20E induces MISO expression in the atrium, we tested MISO induction levels in the absence of the 20E receptor EcR. We injected virgin females with dsRNA targeting EcR, and analyzed levels of MISO induction after mating. In four different experiments, injection of dsEcR (transcript mean reduction = 45%; one-sample t test, t3 = 7. 069, p = 0. 0058, range 63%–36%) impaired MISO induction at 24 hpm by an average of 30-fold compared to injected controls (t test, t6 = 2. 466, p = 0. 0244) (Figure 4D), reinforcing the notion that the expression of this gene after mating is regulated by male-transferred 20E. Interestingly, EcR silencing also reduced transcript levels of Vg (24 hpm: t test, t6 = 2. 106, p = 0. 0399), as expected as this gene is under the control of 20E and its expression is induced by both blood feeding and mating in A. gambiae (Figure 4D) [22], [49]. These data demonstrate that the mating-induced expression of MISO is under the control of sexually transferred 20E, and that EcR mediates this regulation. In this study we unravel a major male–female molecular interaction that switches females to a mated state in terms of egg development and modulates their postmating physiology. We identify a female atrial protein, MISO, which is responsible for the increase in egg production after mating. Silencing of MISO reverts fecundity of mated females back to virgin levels, completely abolishing the effects of mating on oogenesis (Figure 1). Moreover we demonstrate that MISO is induced by and interacts with the steroid hormone 20E transferred by the male (Figure 4). Sexually transferred 20E therefore acts as a “mating signal” that regulates female postmating physiology, and its interaction with MISO translates this signal into increased oogenesis in blood-fed females. To our knowledge, this is the first demonstration of an interaction between a male allohormone and a female protein in insects. The identification of this novel interaction in A. gambiae expands our knowledge of male–female molecular partnerships important for reproductive success, to date limited to few examples from Drosophila (reviewed in [54]). The mating-induced increase in egg development seen in our experimental settings only partially reflects the deep impact that mating has on oogenesis in field conditions. Blood-fed virgins from natural mosquito populations rarely develop eggs after a single blood meal [14]–[16], presumably because of limited nutritional reserves from larval stages [17]. MISO may therefore represent a mating sensor that directs precious resources towards oogenesis only when females are inseminated. Indeed in two different phenotypic assays, MISO influenced pregravid behavior, and similar to virgin females, approximately 15% of dsMISO mated females completely failed to develop eggs compared to 4% of mated controls (Figure 1 and Table S1). It is reasonable to speculate that this effect would be much more pronounced in conditions of limiting resources such as those possibly available in field settings. The interaction between MISO and 20E affects the function of the steroid hormone, as demonstrated by the effects of MISO silencing on 20E titers in the atrium and on the expression of a number of 20E-responsive genes (Figure 3B, C). Although the protein does not have any known functional domains that suggest a role as a sterol carrier, our data indicate that MISO facilitates the release of 20E from the mating plug and its diffusion from the atrium (Figure 3B). Further studies may help elucidating the mechanism by which this female atrial protein regulates 20E function. On the other hand, the finding that sexually transferred 20E induces the atrial-specific expression of MISO via the EcR receptor shows a remarkable mutual cooperation between the two factors (Figure 4C, D). Preventing males from producing and transferring 20E will clarify the full extent of the role that this ecdysteroid plays in regulating female postmating physiology and behavior. A number of hypotheses can be formulated on the downstream events triggered by the interaction of MISO and 20E that lead to increased fecundity. One possibility is that this interaction may prime the fat body to respond to the female-derived ecdysteroids synthesized after a blood meal. This hypothesis is strengthened by the observations that mated dsMISO females experienced a reduced induction in Lp expression after blood feeding compared to controls, paralleled by delayed or impaired oocyte growth (Figure 2, Figure S2, and Table S2). The higher level of Lp expression seen in control mated females is not due to an increased release of ecdysteroids from the ovaries after blood feeding, as ecdysteroid titers were similar in control and dsMISO females (Figure 3A). Interestingly, MISO silencing affects the expression of Vg and Lp differentially: while the most prominent effect on Lp occurs after blood feeding (Figure 2C), Vg transcript levels are repressed only after mating (Figure 3C). This observation suggests a bimodal role for the MISO–20E interaction: a local effect on the expression of mating-responsive genes such as Vg that may regulate the function of reproductive tissues and possibly the remodeling of atrial cells observed after mating [49], and a later effect due to 20E release from the atrium that may control the response of the fat body to blood feeding, thereby affecting Lp transcript levels and egg development. Importantly, these results are consistent with a recent report that identified Lp rather than Vg as the factor most relevant for egg development in A. gambiae [55]. Another possible mechanism is that sexually transferred 20E may regulate resorption of ovarian follicles. In A. aegypti the interplay between JH and 20E influences the fate of follicular resorption during the previtellogenic and vitellogenic stages [56]. Low JH titers during the previtellogenic stage result in higher follicular resorption that can be prevented by the application of methoprene, a JH mimic [57]. 20E can also stimulate resorption of “poor quality” follicles that express low levels of Vg and Lp receptors [58], probably by a caspase-mediated cell death mechanism [59]. In A. gambiae male transferred 20E may therefore act cooperatively with female-derived JH in determining correct follicular resorption. Alternatively, the large amount of 20E transferred from the MAGs, that as confirmed here exceeds the concentration produced by the ovaries after blood feeding [48], may increase the number of developing oocytes by causing yolk accumulation in secondary follicles already during the first blood meal. This process has been observed in A. aegypti [60] and A. stephensi [61] after 20E injection. Mating does not modulate egg development in all anopheline species. For instance, oogenesis is not affected by copulation in the central American malaria vector A. albimanus [62], and interestingly, we could not detect any 20E in the MAGs of this mosquito species (Figure S3B). This result suggests that the effect of mating on fecundity in anophelines might be directly linked to the presence of 20E in the male reproductive tract. Intriguingly, secretion of lower 20E titers in A. gambiae compared to A. albimanus females after a blood meal [22], [63] may be due to the availability of 20E from males in the former species. An increase in egg development following mating is also seen in A. aegypti [39], however the absence of 20E in the MAGs of this species suggests that this effect is caused by a different mechanism (Figure S3B). This increase may be regulated by MAG proteins stimulating the synthesis of growth hormones, as in the case of the stimulation of JH synthesis by Sex Peptide in Drosophila [33]. Indeed the existence of a Sex Peptide–like factor inducing postcopulatory changes in A. aegypti is supported by the observation that MAG extracts injected into virgin females trigger oviposition after blood feeding [64], [65], contrary to A. gambiae where they have no effect [13]. Alternatively, hormones other than 20E produced by the male and transferred during mating may play this role. JH has been detected in the MAGs of A. aegypti [66], and the application of the JH analog methoprene to virgin A. aegypti females enhances oogenesis [39]. No evidence of JH synthesis exists in the MAGs of A. gambiae, and unlike A. aegypti, application of methoprene to blood-fed females inhibits egg maturation and vitellogenesis [22], suggesting differences in the mechanism of oogenesis in the two species. The analysis of the synthesis of 20E in the MAGs of other mosquito species, facilitated by the sequencing of an additional 16 anopheline genomes (http: //www. vectorbase. com), will clarify the existence of a possible correlation between mating plug formation and 20E synthesis in the male, two reproductive features that are both present in A. gambiae but not in A. albimanus and A. aegypti, and between the sexual transfer of 20E and the occurrence of mating-induced oogenesis. Finally, the identification of a previously uncharacterized reproductive pathway in A. gambiae has promise for the development of tools for the control of malaria-transmitting mosquito populations. The effects of the 20E-MISO partnership are likely to be more prominent in field mosquitoes, where nutritional resources are limited and egg development rarely occurs in virgins. Manipulation of this interaction with specific inhibitors or with genetically manipulated males impaired in 20E synthesis might therefore offer an attractive option for reducing the reproductive output of natural Anopheles populations. Moreover, interfering with the mating-induced pathway of oogenesis may have an effect on the development of Plasmodium malaria parasites. A recent study has shown that the expression of Vg and Lp reduces the mosquito Plasmodium-killing efficiency mediated by TEP-1, the principal antiparasitic factor in A. gambiae [55]. As YPPs are regulated after a blood meal via a MISO-dependent mechanism, the 20E–MISO interaction may play a role in the modulation of Plasmodium development in A. gambiae. Mosquitoes from a laboratory colony of the A. gambiae G3 strain were reared under standard conditions [26–28°C, 65%–80% relative humidity, 12 h∶12 h Light/Darkness (L∶D) photoperiod]. For mating experiments, mosquitoes were separated by sex as pupae and raised in cages supplied with sucrose ad libitum. Matings were performed as described previously, and couples were captured in copula [49]. A 397 bp region corresponding to the coding sequence of MISO (AGAP002620) was amplified from atrial cDNA 24 hpm using specific primers FWD: 5′GGTGTTGCCATTGTGTGTGT-3′ and REV: 5′AGTACTCGGCCAGCTGAATG -3′ and cloned into the pLL10 plasmid [67]. A 435 bp region corresponding to AGAP012211 (EcR) was amplified from female abdomen cDNA using the primers FWD: 5′CTGCTCCAGTGAGGTGATGA-3′ and REV: 5′GGCAGCTTACGGTTCTTCAG-3′, while a 495 bp portion of the eGFP control gene was amplified using the primers FWD: 5′TGTTCTGCTGGTAGTGGTCG-3′ and REV: 5′ACGTAAACGGCCACAAGTTC-3′; both amplicons were cloned into pCR2. 1 (Invitrogen). These constructs were then used to synthetize dsRNAs targeting the different genes, following established protocols [10], [67], [68]. Females were sexed as pupae and injected with 69 nl of dsRNA (4 mg/ml) within 24 h of eclosion. Surviving females were allowed to mate with 4-d-old virgin males 3 d after injection. Mated females were then used for phenotypic assays or dissected for qRT-PCR analysis. RNA extraction, cDNA synthesis, and SYBR-green based qRT-PCR were performed as described previously [49] using the primers listed in Table S3. The ribosomal protein gene RpL19 (AGAP004422) was used for normalization, using previously described primers [49]. Three days after dsRNA injections, females were captured during mating and kept in isolation until blood feeding. Females were blood fed ab libitum on human blood. Partially fed or unfed mosquitoes were removed. For oviposition and fertility assays, 3 d after the blood meal, females were put into individual oviposition cups for 4 nights. After completion of oviposition, eggs were counted under the microscope and those that hatched into a larva were scored as fertile. For the egg development assay, abdomens were dissected 3 d after blood feeding, and eggs developed inside the ovaries were counted under the microscope. Affinity-purified polyclonal antibodies against MISO were raised in rabbit against the peptide epitope CSNGPSSSYGPPRNT by a commercial supplier (GenScript Corp. , Piscataway, NJ). Female tissues were homogenized in 20 µl RIPA buffer (10 mM Tris/HCl pH 7. 6,100 mM NaCl, 10 mM EDTA, 0. 5%, Nonidet P40,0. 5% Triton ×100,1× proteases inhibitor from Roche). Samples were centrifuged at 13,000 rpm for 15 min at 4°C. The supernatant was diluted into NuPAGE reducing agent and sample buffer (Invitrogen), heated at 70°C for 10 min, and applied to precast NuPAGE gels (Invitrogen) under reducing conditions according to the manufacturer' s instructions. For native conditions, protein extraction was performed by homogenizing the tissues in a hypotonic solution (10 mM Tris/HCl pH 7. 6,10 mM NaCl, 10 mM EDTA, 1× protease inhibitor from Roche) followed by centrifugation at 13,000 rpm for 15 min at 4°C. The soluble phase was then loaded onto an acrylamide gel in the absence of SDS. Proteins were transferred to a Hybond ECL membrane using the XCell II Blot module (Invitrogen). Membranes were immunostained using standard protocols with the following primary antibody titres: anti-MISO, 0. 96 mg/ml; anti-20E (1∶10 dilution, Cayman Chemicals); and anti-β-actin (1∶1,000 dilution, Santa Cruz Biotechnologies). HRP-conjugated secondary antibodies (Santa Cruz Biotechnologies) were used at a dilution of 1∶10,000. Bands were visualized using ECL Western blotting detection reagents (GE Healthcare). Reprobing with additional primary antibodies was performed after incubating membranes in stripping solution (10 mM Tris/HCl PH 6. 8,100 mM DTT, SDS 2%) at 50°C for 30 min. Before adding the new primary antibody, incubation with the secondary antibody used in the first analysis was tested by ECL to exclude any signal from the previous incubation. MAGs or female reproductive tracts from 3–4-d-old mosquitoes (virgins and mated) were dissected on ice, fixed in 4% formaldehyde, washed in PBS, then blocked and permeabilized in PBS with 1% BSA and 0. 1% saponin. Samples were incubated in either 3 mg/ml anti-MISO or a 1∶10 dilution anti-20E (Cayman Chemicals), then a 1∶1,000 dilution of anti-rabbit Alexa-Fluor 488 (Invitrogen). Alternatively, ovaries were stained with 1∶1,000 dilution of Nile-Red (10 mg/ml in DMSO, Sigma-Aldrich). Tissues were then mounted in DAPI-containing Vectashield medium (Vector Laboratories, Inc.) and visualized using a Point Scanning Confocal microscope Nikon TE2000 or a Zeiss Axio Observer inverted fluorescent microscope with apotome. Ovaries of dsRNA-injected females were dissected from virgin and mated mosquitoes before or after 18 h after a blood meal. Blood feeding was performed 1 h after mating. Ovaries of mated non-blood-fed females were dissected 19 h after copulation. After dissection in Schneider medium (Sigma-Aldrich), individual pairs of ovaries were separately transferred to 50 µl of Schneider medium and incubated for 5 h at 25°C. After incubation, culture medium was stored at −80°C until ecdysteroid quantification. Atria from groups of three virgin females or from groups of three mated females at different time points after mating, previously injected with dsMISO or dsLacZ, were placed in 50 µl methanol and frozen at −80°C. Alternatively, MAGs or testes from 10 A. gambiae, A. albimanus, and A. aegypti males were dissected and placed in 50 µl methanol. Tissues were then homogenized and loaded into separate wells of a 96-well plate pre-coated with mouse anti-rabbit IgG (Cayman Chemical). For the analysis of the in vitro ovarian ecdysteroid secretion, 50 µl of Schneider medium where the ovaries have been incubated were directly loaded into the gel. A standard curve was prepared from 18 ng 20E (Sigma-Aldrich) in methanol or Schneider medium (Sigma), with a series of seven 3-fold dilutions. After evaporation of the methanol, 50 µl of each of the following solutions were added: Enzyme ImmunoAssay Buffer (0. 1 M phosphate solution containing 0. 1% BSA, 0. 4 M sodium chloride, 1 mM EDTA, and 0. 01% sodium azide); 20E acetylcholinesterase (AChE) Tracer, which is a covalent conjugate of 20E and AChE; and anti-20E rabbit IgG (Cayman Chemical). The plate was incubated with the solutions overnight at 4°C, washed with PBS 1× containing 0. 05% TWEEN20, incubated with 200 µl Ellmans reagent (5,5′-dithiobis- (2-nitrobenzoic acid) ) (Cayman Chemical), and finally developed for 90–120 min and measured in an ELISA reader at 420 nm. Three-day-old females were injected with different quantities (2. 5 µg, 0. 25 µg, and 0. 025 µg) of 20E (138 nl of 10% ethanol solution). As controls, either the same volume of 10% ethanol or 0. 25 µg of water-soluble cholesterol (which is the maximum soluble concentration) (Sigma-Aldrich) were injected. Female lower reproductive tracts (LRT, atrium, spermatheca, and parovarium) were dissected 24 h after injection and analyzed by qRT-PCR. Three replicates were performed containing 6–8 tissues per replicate. LRTs were also dissected and analyzed by qPCR from noninjected virgin females and from mated females at 24 hpm. Fifty atria from virgin and mated (8 hpm) females were dissected and homogenized in 15 µl of hypotonic solution (10 mM Tris/HCl pH 7. 6,10 mM NaCl, 10 mM EDTA, 1× protease inhibitor from Roche) and centrifuged at 13,000 rpm for 15 min at 4°C. The soluble phase was then incubated for 1 h at 4°C under gentle rocking with 2 mg of anti-MISO rabbit IgG that had been previously linked to Dynabeads protein A (Invitrogen) in a 10 min incubation at 25°C under gentle rocking followed by three PBS 1× washes. The immunoprecipitate was washed three times with PBS 1× and split in two aliquots: one-third of the total volume was utilized in a Western blot incubated with anti-MISO, while the remaining two-thirds were diluted with 100 µl of methanol, to extract 20E, and kept at −80°C. The methanol solution was then analyzed with an anti-20E ELISA. As controls, 25 ng of 20E were incubated under the same conditions with 2 mg of Rabbit anti-MISO linked to Dynabeads protein A to measure the unspecific binding of 20E to the antibody or to the Dynabeads. All samples were also immunoprecipitated using pre-immune rabbit IgG to control for unspecific bindings. ELISA quantification was performed normalizing the signal to anti-MISO rabbit IgG-Dynabeads protein A incubated in methanol. To examine the effects of MISO on oviposition and egg development, we utilized a generalized linear model approach where the number of eggs was modeled with a log link function and Poisson distribution function using SAS Proc GenMod (SAS, Inc. , Cary, NC). Replicate was also included as a covariate in each of these analyses. Post hoc comparisons for fecundity were made using the Bonferroni Multiple Comparison Procedure in SAS (SAS, Inc.). Differences in the number of females that fail to lay or to develop eggs (Table S1) between different groups were analyzed with a chi-square test using Prism 5. 0 (GraphPad Software, Inc. , La Jolla, CA). To test for difference in gene expression between two or more treatments (Figure 2C, Figure 3C, Figure 4), we used t test or ANOVA test, respectively, using Prism (GraphPad Software, Inc). Similarly, ecdysteroid secretion by ovaries and oocyte lengths between different groups were compared using ANOVA test. Differences in fertility between dsLacZ and dsMISO were examined through Mann–Whitney (Prism, GraphPad Software, Inc.). For 20E titers in mated atria (Figure 3B), a Wilcoxon test was used to compare the natural logarithm transformed ecdysteroid levels of each group at different time points. Furthermore, we compared trajectories of steroid hormone levels of dsMISO and dsLacZ female groups through a mixed model, with natural logarithm transformed steroid levels and a random intercept to accommodate within female correlations measured at the five time points after mating (0. 5,6, 12,18, and 24 hpm). Since we did not expect to find any differences in the mean levels of steroid at the first time point, we forced a common intercept for dsLacZ and dsMISO females by including in these models only a fixed effect for time. Statistical significance in the trajectory of the geometric mean of steroid levels (exp {mean[log (steroid) ]}) between the two female groups was tested through an interaction term between time and female group (S. plus 8. 0, TIBCO Software).

Anopheles gambiae mosquitoes are the most deadly vectors of human malaria. The reproductive ability of these mosquitoes contributes to their role as disease vectors as it ensures high population densities for malaria transmission. The number of eggs developed by females after blood feeding depends on whether they have previously mated. Indeed in natural mosquito populations, virgin females rarely develop eggs when blood fed. Here we report on the identification of a molecular interaction between 20-hydroxy-ecdysone (20E), a steroid hormone transferred by the male during sex, and the Mating-Induced Stimulator of Oogenesis (MISO), a female reproductive protein, expression of which is triggered by mating and leads to increased egg production. We show that the expression of MISO after mating is regulated by 20E via the Ecdysone receptor (EcR). Experimental silencing of MISO reduces the ability of mated females to develop eggs after blood feeding, by reducing expression of a vitellogenic lipid transporter. By showing how male mosquitoes contribute to oogenesis in females, we identify a molecular pathway that can be targeted to reduce the reproductive success of natural mosquito populations to aid malaria control.

Abstract Introduction Results Discussion Materials and Methods

2013

The Interaction between a Sexually Transferred Steroid Hormone and a Female Protein Regulates Oogenesis in the Malaria Mosquito Anopheles gambiae

Familial recurrent hydatidiform mole (RHM) is a maternal-effect autosomal recessive disorder usually associated with mutations of the NLRP7 gene. It is characterized by HM with excessive trophoblastic proliferation, which mimics the appearance of androgenetic molar conceptuses despite their diploid biparental constitution. It has been proposed that the phenotypes of both types of mole are associated with aberrant genomic imprinting. However no systematic analyses for imprinting defects have been reported. Here, we present the genome-wide methylation profiles of both spontaneous androgenetic and biparental NLRP7 defective molar tissues. We observe total paternalization of all ubiquitous and placenta-specific differentially methylated regions (DMRs) in four androgenetic moles; namely gain of methylation at paternally methylated loci and absence of methylation at maternally methylated regions. The methylation defects observed in five RHM biopsies from NLRP7 defective patients are restricted to lack-of-methylation at maternal DMRs. Surprisingly RHMs from two sisters with the same missense mutations, as well as consecutive RHMs from one affected female show subtle allelic methylation differences, suggesting inter-RHM variation. These epigenotypes are consistent with NLRP7 being a maternal-effect gene and involved in imprint acquisition in the oocyte. In addition, bioinformatic screening of the resulting methylation datasets identified over sixty loci with methylation profiles consistent with imprinting in the placenta, of which we confirm 22 as novel maternally methylated loci. These observations strongly suggest that the molar phenotypes are due to defective placenta-specific imprinting and over-expression of paternally expressed transcripts, highlighting that maternal-effect mutations of NLRP7 are associated with the most severe form of multi-locus imprinting defects in humans. The most common form of complete hydatidiform mole (CHM) is sporadic and androgenetic diploid in origin. These products of conception frequently result from the fertilization of an oocyte from which the maternal chromosomes are lost and endoreduplication of a single sperm genome, or the fertilization by two sperm, to give a diploid DNA content of entirely paternal origin [1]. Occasionally HM can be recurrent and familial in nature (OMIM 231090) [2]. Detailed homozygosity mapping and gene mutation screening has identified two loci, 19q13. 4 and 6q13, which harbor the causative genes, NLRP7 (NACHT, leucine rich repeat and PYD containing 7) and KHDC3L (also known as C6ORF221) respectively [3,4]. Approximately 70% of women affected by familial recurrent HM (RHM) are associated with recessive mutations of NLRP7 [5,6], whereas genetic aberrations of KHDC3L are much less frequent, and present in only ~10% of patients without an NLRP7 involvement [7,8]. In both cases the mutations cause the RHM by maternal-effect. Definitive evidence that a defective oocyte is responsible for the pathophysiology of RHM comes from the observations that assisted reproductive cycles using donated oocytes in three patients with recessive NLRP7 mutations resulted in normal offspring [8,9]. This maternal-effect model is consistent with transcript abundance of both NLRP7 and KHDC3L, which accumulate in the developing oocytes and are present in early pre-implantation embryos [10,11]. Such expression profiles are coherent with an involvement in the control of maternally derived epigenetic programing or early developmental events in the zygote that occur before embryonic genome activation. Paternal transmission of NLRP7 mutations does not interfere with spermatogenesis, since males homozygous for NLRP7 mutations can father children [5,12]. Epigenetic studies in these abnormal pregnancies have revealed aberrant DNA methylation profiles at a limited number of imprinted genes [13,14]. Imprinted genes are expressed in a parent-of-origin specific fashion, which is coordinated by differentially methylation regions (DMRs) inherited from the gametes [15]. NLRP7 does not have an orthologue in mouse, but is thought to have originated from an evolutionary duplication of its nearest family member, NLRP2 [16]. Curiously, mutations of NLRP2 are responsible for a single familial case of Beckwith-Wiedemann syndrome with methylation defects at multiple loci, including KvDMR1 (also known as ICR2) and MEST DMR [17]. Consistent with their androgenetic composition, our recent genome-wide methylation profiling of sporadic HMs revealed the paternalization of all known imprinted DMRs, with maternally-methylated DMRs being devoid of methylation and paternally-derived DMRs being fully methylated [18]. Similar analyses on RHM biopsies tissues with known underlying genetic causes are difficult to conduct, partially hampered by the fact that genetic diagnosis takes place in the phenotypically normal affected women, with the molar tissues discarded following pathological examination. In this study, we determine the genome-wide methylation profiles of RHMs from four females with NLRP7 mutations using the high-density Illumina Infinium HumanMethylation450 (HM450k) BeadChip arrays, which simultaneously quantify methylation at ~2% of all CpG dinucleotides in the human genome. The RHM samples were from women with a variety of genetic lesions including siblings carrying the same homozygous non-synonymous missense mutation (c. 2078G>C; p. R693P), an individual homozygous for a deletion that removes exons 2–5 (c. -39-1769_2129+228del) and a female with compound heterozygous mutations (c. 2018C>G, c. 2161C>T; p. S673X, p. R721W) (Fig 1A). Our initial analysis focused on comparing the methylation profiles of four androgenetic moles and seven normal placental samples (three first trimester and four third trimester) with those obtained for the NLRP7-mutated samples (Fig 1B and S1A Fig). A total of 616 probes mapping to 36 known ubiquitous imprinted DMRs were assessed, with observations confirmed by both pyrosequencing and standard allele-specific bisulphite PCR and sub-cloning (Fig 1C, S1C and S2 Figs). These comprehensive analyses revealed that, while normal placental biopsies had partial methylation consistent with allelic methylation (with the exception of the fully methylated NNAT and GNAS-AS1 promoters) [18], the majority of maternally methylated DMRs presented with lack-of-methylation (LOM) in both androgenetic and NLRP7-associated HMs. The only exceptions were for the IGF1R and RB1 DMRs that maintain allelic methylation in both types of mole, whereas the SNURF DMR was maintained in some of the NLRP7-mutated samples. In addition, we observe some inter-individual differences. The FAM50B DMR maintained a partially methylated state in two androgenetic CHMs and in a RHM from one of the sisters with the NLRP7 p. R693P mutation. Surprisingly, this same RHM sample also showed imprinted methylation at the PLAGL1 and PEG10 DMRs (Fig 1C and S2 Fig). Furthermore a comparison of two different RHMs from patient 4 revealed a similar methylation profile with the exception that the PEG10 and SNURF DMRs presented allelic methylation in one of the moles (S1 Fig). The PEG10 DMR was previously reported to be largely unaffected in three familial RHM samples [14]. The only paternal DMR with probes present on this array that acquires methylation in the male germline and is partially methylated in placenta is the H19 DMR (also known as ICR1). Consistent with the two copies of the sperm genome, the androgenetic CHMs are fully methylated at this locus, whereas the RHM are partially methylated. In 3 cases allele-specific bisulphite PCR revealed that the methylation was on the paternal allele (Fig 1C). Quantitative pyrosequencing of bisulphite PCRs targeting the IG-DMR on chromosome 14, which also acquires methylation from sperm but does not have probes on the HM450k array platform, revealed a partially methylated profile in both control placenta and RHMs (S2 Fig). Similarly the MEG3 DMR, which is regulated in-cis by the IG-DMR, shows a similar partially methylated profile consistent with allelic methylation (Fig 1B). The ZBDF2 and ZNF597/NAA60 promoters were fully methylated in both androgenetic CHM and NLRP7 mutated RHMs. This is consistent with the presumption that these regions acquire methylation on the paternal allele during early development under the hierarchical influence of the maternally methylated GPR1-AS and ZNF597 DMRs, respectively [18]. To determine if the lack of methylation at imprinted DMRs results in altered expression we performed allelic-specific RT-PCR on the RHM samples. We confirm that HYMAI, PEG10 and PEG3 transcripts are paternally expressed in control placenta samples but expressed from both alleles in RHMs (Fig 1D). Biallelic expression of NAP1L5 was associated with LOM in RHM1, but imprinted expression was preserved in RHM3 that had allelic methylation at this DMR (Fig 1C and 1D). Recently, using genome-wide methylation profiling in normal placental biopsies and androgenetic moles, we identified 18 placenta-specific maternally methylated DMRs [18]. Genome-wide methylation analysis utilizing methyl-seq in human gametes revealed that these loci inherit methylation from oocytes and maintain allelic methylation during pre-implantation reprograming [19]. We interrogated the 153 probes mapping to these placenta-specific imprinted DMRs and confirmed observations by both pyrosequencing and allele-specific bisulphite PCR (Fig 2, S1 and S2 Figs). This revealed that, while first trimester and term placental biopsies had partial methylation indicative of maternally methylated DMRs, all androgenetic and NLRP7-associated HMs presented with robust LOM. In several cases the RHM samples were heterozygous for single base pair polymorphisms (SNPs) that confirmed that methylation was absent from the maternal alleles (Fig 2B). Consistent with the lack of allelic methylation, the normally paternally expressed imprinted genes MCCC1, LIN28B and GLIS3 are expressed from both parental alleles in RHMs (Fig 2C). Furthermore qRT-PCR revealed an increased expression of DNMT1 and AGBL3 compared to normal placenta samples coherent with biallelic over-expression. Expression was within the normal range for H19 which in consistent with the maintained paternally derived methylation at this DMR (Fig 2D). Furthermore, our previous results revealed that approximately 12% of all CpG methylation is contained within LINE-1 sequences [20]. Pyrosequencing analysis of these retrotransposable elements, as well as α-satellites and Alu-Yb8 sequences, in NLRP7-mutated RHMs revealed a profile indistinguishable from normal placenta (S3 Fig) [18]. Together these observations suggest that only maternally derived methylation is affected in RHM, consistent with oocyte epigenetic aberration and not somatic imprint maintenance. To expand our methylation analysis we performed an unbiased screen for additional loci with abnormal methylation in the RHM samples with underlying NLRP7 mutations. The identified regions were characterized by at least 3 Infinium probes and relative distance between consecutive probes below 500 bp, requiring runs with a consistent change (same direction and p-value < 0. 01) and an absolute average methylation change >20% (β 0. 2). This analysis identified 61 regions, 56 of which are CpG islands, 88% mapping to transcript promoters (Fig 3A; S1 Table). Surprisingly all candidate regions identified were partially methylated in normal placental biopsies and devoid of methylation in androgenetic CHMs and somatic tissues (Fig 3A; S1 Table). This profile suggests the existence of further placenta-specific maternally methylated regions that could regulate imprinted expression. Consistent with the regions being maternally methylated, all regions were unmethylated in sperm (S1 Table). To confirm if the observed methylation was restricted to the maternal allele we developed a methylation-sensitive genotyping assay in which polymorphic allele calling is performed on genomic DNA before and following digestion with the methylation-sensitive HpaII endonuclease (Fig 3B). Allelic methylation is confirmed when a heterozygous genomic DNA sample is reduced to homozygosity following digestion with the remaining allele representing the methylated chromosome. Twenty-eight of the 61 candidate regions had highly informative SNPs that allowed parental origin of methylation to be determined. In 22 cases we confirmed the presence of maternal methylation in multiple placenta samples, with a further six regions being allelically methylated with parental genotypes being uninformative (Fig 3C and 3D, S4 and S5 Figs; S2 Table). Fifteen of these samples were subsequently shown to be allelically methylated using bisulphite PCR and subcloning with an additional five regions associated with RHOBTB3, PURA, FGF8, CCDC71L and WIF1 presenting with both fully methylated and unmethylated DNA strands (Fig 3D and S5 Fig). The main biological significance of allele-specific methylation is allele-specific RNA expression, which in the case of maternally methylated regions is predicted to dictate paternal expression. We subsequently determined allelic expression for a subset of transcripts that contained highly polymorphic exonic SNPs. Allele-specific RT-PCR confirmed paternal expression of RHOBTB3, SCIN, ZNF396, ST8SIA1, ZFP90, CCDC71L, RASGRF1, HECW1 and CMTM3 with monoallelic expression of CD83 in a polymorphic fashion in multiple placental biopsies (Fig 4A and S6A Fig; S3 Table). In situations where monoallelic expression was uninformative due to maternal DNA also being heterozygous, the methylated allele was always the repressed one, suggesting a functional link between methylation and expression. In addition, we identified a maternally methylated CpG islands overlapping the promoter of SNCB, a transcript that has previously been described as paternally expressed in placenta [21]. Furthermore, we also identified a maternal DMR within the TTC39A gene that is adjacent to EPS15, a transcript also reported to be imprinted in placenta [22] (S5 Fig). To determine whether these placenta-specific DMRs can orchestrate allelic silencing of gene clusters similar to ubiquitous imprinted DMRs, we performed allele-specific RT-PCR for 20 flanking genes associated with loci containing imprinted transcripts (Fig 4B and S6B Fig). Surprisingly, with the exception of ADAM23 within the GPR1-AS domain on chromosome 2, we observe that the remaining 19 transcripts analyzed are expressed equally from both parental chromosomes, indicating that placenta-specific DMRs do not possess the ability to regulate allelic expression of surrounding genes. Despite evolutionary conserved imprinting of GPR1-AS and ZDBF2 [23], paternal expression of ADAM23 is not observed in mouse placenta (Fig 4C). Together this suggests that this locus is regulated in a different manner to the majority of placenta-specific imprinted loci identified and that subtle species differences exist [24]. We have compared the DNA methylation profiles of RHMs with underlying NLRP7 mutations with androgenetic CHM biopsies. This has revealed not only widespread methylation defects at imprinted loci, but has facilitated in the identification of novel maternally methylated loci. As a result of unbiased bioinformatic analyses, the number of genes associated with placenta-specific maternally methylated DMRs has increased from 18 to 43, with a further eight regions of allelic methylation indicative of an imprinted DMR. In addition, a further 28 candidates did not contain SNP allowing allelic methylation to be determined. Our observations indicate that there are more imprinted domains in the human placenta than in somatic tissues. Interestingly, 10 of these loci (TMEM17, NUDT12, RHOBTB3, CD83, ARMC3, AIFM2, ST8SIA1, PCK2, RASGRF1, CMTM3) show opposing methylation profiles in diandric (two paternal plus one maternal haploid genomes) compared to digynic (extra maternal chromosomes) triploid biopsies [25], consistent with the maternal methylation profile we describe. It is currently unknown which imprinted genes are responsible for the HM phenotype, but aberrant expression from the maternal allele of these placenta-specific genes is likely to play an important role, since they include the essential epigenetic gene DNMT1 and the micro-RNA processor LIN28B. In addition there are several strong candidates for influencing trophoblast development, including the cytochrome P450 (CYP) subfamily member CYP2J2 that has previously been shown to be up-regulated in preeclampsia and THSD7A, a placental and endothelial protein that mediates cellular migration [26,27]. In addition, the deregulation and over-expression of the C19MC pri-miRNA will lead to the concomitant increased abundance of 50 mature miRNAs that have recently been shown to regulate trophoblast invasion [28,29]. Our genome-wide analysis revealed aberrant methylation profiles in RHM associated with maternal-effect NLRP7 defects at imprinted loci. It is not possible to determine if the methylation anomalies are restricted to imprinted loci, as many regions inheriting methylation from oocytes undergo epigenetic reprograming during pre-implantation development [30,31], resulting in an unmethylated state that is indistinguishable from an epimutation. Similar epigenetic profiling of blood-derived DNA from the women carrying biallelic recessive genetic mutations of NLRP7 failed to identify any methylation anomalies when compared to healthy controls (S7 Fig), endorsing the hypothesis that the methylation defect arises by maternal-effect either in the developing oocyte or in early pre-implantation stages. Our results imply that the epigenetic aberration observed in RHM arise early in the female germline since paternally methylated DMRs are unaffected, maintaining the correct methylation profiles at the H19 and IG-DMR loci. This suggests that NLRP7 has a different function to ZFP57 [32] or DPPA3 [33], which both protect multiple imprints from TET3-associated 5mC to 5hmC reprogramming at the zygotic stage when the pronuclei have yet to breakdown and the parental genomes fuse [34]. Endorsing this theory, detailed immunostaining for NLRP7 in early human embryos revealed that this protein is exclusively localized to the cytosekeleton and not in the nucleus where it could associate with chromatin and influence methylation [11]. In addition, NLRP7 is not observed in the nucleus of developing oocytes (germinal vesicle (GV) stage oocytes through to those arrested in metaphase II) [5,11]. Curiously, immunostaining for the de novo methylatransferases DNMT3A and DNMT3B revealed a similar cytoplasmic localization [35], indicating that a NLRP7-complex may ensure the correct cellular localization and nuclear translocation of these epigenetic factors during a yet to be identified period of oocyte development. Once in the nucleus, this low abundance complex may associate to specific DNA sequences by direct interaction with chromatin regulator YY1 or ZBTB16 [36,37]. This, and the absence of DNMT3L in human GV-metaphase II oocytes, highlights the fact that the process of imprint acquisition in humans and mouse differ greatly. The NLRP family of proteins is known to play a direct role in inflammasome activation, which results in the secretion of interleukin-1β (IL-1β) [38]. These observations offer an indirect mechanism explaining how NLRP7 influences maternally derived methylation. Prenatal oogenesis produces hundreds of thousands of oocytes, most of which are discarded before birth. During fetal development this phenomenon is associated with oocyte apoptosis, acting as a quality control measure, eliminating cells with meiotic anomalies. Interestingly, strong NLRP7 staining has been reported in human blastomeres undergoing apoptosis [11]. The process of oocyte selection can be influenced by pro-survival factor IL-1β, a cytokine known to be involved in oocyte nuclear maturation in many mammalian species [39]. This reduction in oocyte number occurs at the approximate time (14–20 weeks gestation) when they presumably acquire the methylation signatures at imprinted regions [40]. It is therefore plausible that disruption to the selection mechanism through defective NLRP7 may allow for the survival, and eventual dominant follicle recruitment, ovulation and fertilization decades later, of an oocyte with an inappropriate methylation state. Whatever the underlying mechanism, we show that maternal-effect mutations of NLRP7 are associated with the most severe cases of multi-locus imprinting defects in humans. All women had presented with multiple RHMs (patients 1–4 had 3,6, 4 and 3 previous RHMs, respectively) and provided informed consent to use their tissues for research. The ethical approval was granted by the Bellvitge Institute for Biomedical Research (PR096/10) and the Tissue Management Committee of the Imperial College Healthcare NHS Trust Research Tissue Bank (R15048), which is approved by NRES to provide deemed ethics for projects accessing material and data stored within the Research Tissue Bank. All mothers provided informed consent for themselves and their child prior to participating in the study. Ethical approval for collecting blood and placental samples was granted by the ethical committees of Hospital St Joan De Deu Ethics Committee (Study number 35/07), Bellvitge Institute for Biomedical Research (PR006/08) and the National Center for Child Health and Development (project 234). Peripheral blood samples were obtained from healthy volunteers and tissue samples were obtained from BrainNet Europe/Barcelona tissue bank. Mouse work was approved by the Institutional Review Board Committees at the National Center for Child Health and Development (approval number A2010-002). Animal husbandry and breeding were conducted according to the institutional guidelines for the care and the use of laboratory animals. Five molar biopsies from four different women with two mutated copies of the NLRP7 gene who were referred to the Trophoblastic Tumour Screening and Treatment Centre, Charing Cross Hospital (London, UK) were used in this study. The mutations were identified using standard PCR and sequencing as previously described [5]. All women had presented with multiple RHMs and provided informed consent to use their tissues for research. A cohort of 72 human placenta biopsies with corresponding maternal blood samples were collected at Hospital St Joan De Deu (Barcelona, Spain) and the National Center for Child Health and Development (Tokyo, Japan). All placenta biopsies were collected from the fetal side around the cord insertion site. The placenta-derived DNA samples were free of maternal DNA contamination based on microsatellite repeat analysis. Both DNA and RNA extractions and cDNA synthesis were carried out as previously described [20,41]. Wild type mouse embryos and placentae were produced by crossing C57BL/6 (B) with Mus musculus molosinus (JF1) mice and collected at embryonic day 9. 5. We generated methylation datasets using the Illumina Infinium HumanMethylation450 BeadChip arrays, which simultaneously quantifies ~2% of all CpG dinucleotides. Bisulphite conversion of 600 ng of DNA was performed according to the manufacturer’s recommendations for the Illumina Infinium Assay (EZ DNA methylation kit, ZYMO, Orange, CA). The bisulphite-converted DNA was used for hybridisation following the Illumina Infinium HD methylation protocol at genomic facilities of the Cancer Epigenetics and Biology Program (Barcelona, Spain) or the Barts and The London School of Medicine and Dentistry Genome Centre (London, UK). The resulting data for the NLRP7-mutated familial RHMs and the corresponding maternal blood samples have been deposited in the GEO database with the accession number GSE66247. In addition we used the androgenetic CHMs, control placenta and leukocyte datasets from GSE52576. Before analysing the data, we excluded possible sources of technical biases that could influence results. We applied signal background subtraction and inter-plate variation was normalized using default control probes in BeadStudio (version 2011. 1_Infinium HD). We discarded probes with a detection p-value >0. 01. We also excluded probes that lack signal values in one or more of the DNA samples analysed. For the analysis of known imprinted domains, probes mapping to the DMRs identified by Court and colleagues were directly analysed. Prior to screening for novel imprinted DMRs we excluded all X chromosome CpG sites. An in-house bioinformatic pipeline (using R-package) was utilized to tests the difference of a minimum of 3 consecutive Infinium probes within 500bp windows via a linear model (empirical Bayes moderated p-value < 0. 01) that provides a t-statistic, with an absolute methylation change of > 20% (beta 0. 2). The circular heatmaps used to display the DNA methylation profiles were generated using Circos software. Genotypes of potential SNPs identified in the UCSC hg19 browser were obtained by PCR and direct sequencing. Sequence traces were interrogated using Sequencher v4. 6 (Gene Codes Corporation, MI) to distinguish heterozygous and homozygous samples. Heterozygous sample sets were analyzed for either allelic expression using RT-PCR, methylation-sensitive genotyping or bisulphite PCR, incorporating the polymorphism within the final PCR amplicon so that parental alleles could be distinguished (for primer sequence see S4 Table). Expression of the transcripts of interest was analyzed by quantitative real-time RT-PCR with a fluorochrome (SYBR Green) assay and normalized against RPL19. Primer sequences are listed in S4 Table. The assays were run in triplicate in 384 well plates in 7900HT Fast Real Time PCR System (Applied Biosystems). Dissociation curves were obtained at the end of each reaction to rule out the presence of primer dimers oabbrevr unexpected DNA species in the reaction. Non-template controls and a calibrator cDNA were included in each assay. Results were analyzed with the SDS 2. 3 software (Applied Biosystems). Data analysis was performed using the RQ method and final graphs generated in prism5. Approximately 1 μg DNA was subjected to sodium bisulphite treatment and purified using the EZ DNA methylation-Gold kit (ZYMO, Orange, CA) and was used for all bisulphite PCR analysis. Approximately 2 ul of bisulphite converted DNA was used in each amplification reaction using Immolase Taq polymerase (Bioline) at 45 cycles and the resulting PCR product cloned into pGEM-T easy vector (Promega) for subsequent subcloning and sequencing (for primer sequence see S4 Table). Approximately 50 ng of bisulphite converted DNA was used for pyrosequencing. Standard bisulphite PCR was used to amplify the imprinted DMRs with the exception that one primer was biotinylated (see S4 Table for primer sequences). Previously published primers targeting LINE-1, α-satellites and ALU-Yb8 were used for amplification and sequencing [20]. In all cases the entire biotinylated PCR product (diluted to 40 μl) was mixed with 38 μl of Binding buffer and 2 μl (10 mg/ml) streptavidin-coated polystyrene beads. After washing in 70% ethanol, DNA was denaturated with 50 μl 0. 5M NaOH. The single-stranded DNA was hybridized to 40-pmol sequencing primers dissolved in 11 μl annealing buffer at 80°C. For sequencing, forward primers were designed to the complementary strand. The pyrosequencing reaction was carried out on a PyroMark Q96 instrument. The peak heights were determined using Pyro Q-CpG1. 0. 9 software (Biotage). Approximately 500 ng of heterozygous genomic DNA was digested with 10 units of HpaII restriction endonuclease for 4 hours at 7°C. The digested DNA was subject to ethanol precipitation and resuspended in a final volume of 20 μl TE or water. Approximately 2 μl of digested DNA was used in each amplification reaction using Bioline Taq polymerase for 40 cycles. The resulting amplicons were sequenced and the sequences traces compared to those obtained for the corresponding undigested DNA template. The Illumina Infinium HumanMethylation450 BeadChip array data has been deposited in the GEO repository and assigned the accession number GSE66247 and GSE52576.

Complete hydatidiform moles (CHMs) are abnormal human conceptsus characterized by excessive trophoblast proliferation that commonly result from the absence of a maternal genetic contribution compensated by two copies of the paternal genome. In a few rare cases HMs maybe recurrent (RHM), characterized by a biparental genetic contribution and underlying NLRP7 mutations. It is speculated that aberrant genomic imprinting plays a key role in HM formation, but to date no studies have determined the extent of imprint defects in molar biopsies. By comparing the methylation profile of CHMs and RHMs with normal placentas, we confirm widespread absence of allelic methylation at imprinted loci and identify many aberrantly methylated regions, all of which have profiles consistent with imprinting.

Abstract Introduction Results Discussion Material and Methods

2015

Absence of Maternal Methylation in Biparental Hydatidiform Moles from Women with NLRP7 Maternal-Effect Mutations Reveals Widespread Placenta-Specific Imprinting

Trypanosomiasis is regarded as a constraint on livestock production in Western Kenya where the responsibility for tsetse and trypanosomiasis control has increasingly shifted from the state to the individual livestock owner. To assess the sustainability of these localised control efforts, this study investigates biological and management risk factors associated with trypanosome infections detected by polymerase chain reaction (PCR), in a range of domestic livestock at the local scale in Busia, Kenya. Busia District also remains endemic for human sleeping sickness with sporadic cases of sleeping sickness reported. In total, trypanosome infections were detected in 11. 9% (329) out of the 2773 livestock sampled in Busia District. Multivariable logistic regression revealed that host species and cattle age affected overall trypanosome infection, with significantly increased odds of infection for cattle older than 18 months, and significantly lower odds of infection in pigs and small ruminants. Different grazing and watering management practices did not affect the odds of trypanosome infection, adjusted by host species. Neither anaemia nor condition score significantly affected the odds of trypanosome infection in cattle. Human infective Trypanosoma brucei rhodesiense were detected in 21. 5% of animals infected with T. brucei s. l. (29/135) amounting to 1% (29/2773) of all sampled livestock, with significantly higher odds of T. brucei rhodesiense infections in T. brucei s. l. infected pigs (OR = 4. 3,95%CI 1. 5-12. 0) than in T. brucei s. l. infected cattle or small ruminants. Although cattle are the dominant reservoir of trypanosome infection it is unlikely that targeted treatment of only visibly diseased cattle will achieve sustainable interruption of transmission for either animal infective or zoonotic human infective trypanosomiasis, since most infections were detected in cattle that did not exhibit classical clinical signs of trypanosomiasis. Pigs were also found to be reservoirs of infection for T. b. rhodesiense and present a risk to local communities. Tsetse transmitted African trypanosomiasis poses a severe socio-economic impact throughout sub-Saharan Africa with losses to production estimated at over US$ 1. 3 billion annually in terms of meat and milk yield in cattle [1]. Animal trypanosomiasis, is a serious constraint to productivity in Busia District in Western Province, Kenya, where there are also sporadic cases of human sleeping sickness reported [2]. An estimated 70% of the potential labour force of the district is engaged in subsistence mixed crop-livestock farming [3] in this poor rural area. Trypanosomiasis related losses include both direct livestock out-put (weight-loss, decrease in milk, decreased reproductive rate) as well as lost opportunity in terms of integration of livestock into crop production and the potential for crop-improvement (loss of draught power and manure) [1], [4]. Trypanosoma congolense (T. congolense), T. vivax and to a lesser extent T. brucei s. l. are the species that affect local African cattle in this region. Small ruminants are generally reported to be less susceptible to clinical trypanosomiasis [5], however they can harbour low grade chronic trypanosome infections, which can induce severe pathology when transmitted to cattle [6]. Pigs are moderately susceptible to T. congolense and T. brucei s. l. infections [7], [8]. T. brucei s. l. infections are generally less pathogenic in indigenous livestock than either T. vivax or T. congolense [9]. However in areas endemic for Rhodesian sleeping sickness, livestock play an important role as a reservoir for the human infective subspecies T. brucei rhodesiense (T. b. rhodesiense), frequently without displaying overt clinical signs of infection. Traditionally, control of trypanosomiasis in Kenya was state-run. Until the late 1980s, large scale aerial and ground-spraying campaigns had been used by public agencies as the mainstay of tsetse and thus of trypanosomiasis control [10]. Over the last two decades, ongoing cuts in the budget of the Veterinary Department, concentrated the remaining available funds on the provision of public-goods services [11]. Trypanosomiasis was no longer perceived as an acute risk to human health in Kenya, but viewed as a livestock production disease, the control of which was in the interest of the individual livestock-owners. This shift in responsibility radically changed the scale of control efforts from area wide programmes, to small-scale community based interventions [4]. Behavioural and geographical risk factors for human sleeping sickness have been identified at the local scale in a neighbouring district in south-east Uganda (Tororo District) [12], but less is known about the epidemiology of livestock trypanosomiasis at the local scale in areas endemic for trypanosomiasis and human sleeping sickness in Kenya. This study examines the biological and management risk factors associated with trypanosomiasis infections (both animal infective and human infective, zoonotic infections), in a range of domestic livestock in Busia District, Kenya. Making use of a unique census set of blood samples from the local livestock population in two study sites in Busia, in combination with sensitive polymerase chain reaction (PCR) technology for identification of trypanosome infections, this study aimed to establish animal inherent and management related risk factors for trypanosomiasis at the household level to identify the parasite reservoir and to aid identification of infected animals and inform control. Furthermore this study aimed to reassess the public health significance of trypanosomiasis in Busia, by investigating the presence and distribution of the human infective T. b. rhodesiense in the livestock reservoir as proxy for transmission risk for human sleeping sickness. The study used samples of blood stored in long term storage from a number of livestock species collected from the ear vein. This non invasive approach requiring minimal restraint of the animals was approved by both the University of Edinburgh Ethics Review Committee and the Kenyan Department of Veterinary Services. The study was performed in two sites within Busia District, Western Province, Kenya. Site 1 comprised nine adjacent villages and was located in Funyula Division between 0. 249°–0. 281° North and 34. 087°–34. 124° East (Datum WGS84) and Site 2 comprised ten adjacent villages and was located in Butula Division between 0. 317°–0. 358° North and 34. 201°–34. 240° East (Datum WGS84). These two sampling areas were established field sites, which were well characterised in terms of livestock-keeping dynamics and veterinary care seeking behaviour [13], [14]. Figure 1 shows a map of the study sites. Census sampling targeting the entire livestock population (cattle, pigs and small ruminants) of the two sampling sites was performed in July (Funyula site) and October (Butula site) 2004, by visiting all livestock keeping homesteads in all 19 sampling villages. The geographic co-ordinates of each livestock keeping homestead, linked to a unique identification number, were recorded using a handheld global positioning system 12 (GPS 12) Personal Navigator (Garmin Ltd, Kansas, USA). Whole blood samples from ear-veins were collected from all cattle, pigs and small ruminants at each livestock keeping homestead. Samples (100 µl) were applied to FTA Cards (Whatman, Maidstone, Kent, UK) and allowed to air dry prior to storage at room temperature [15]. A total of 2773 livestock samples from 549 livestock-keeping homesteads were collected (see table 1). Ear-vein sampling was attempted in all animals other than those below two weeks of age. In a number of animals (mainly goats and sheep) ear vein puncture failed to draw sufficient blood due to small or collapsing ear veins. Several pigs were excluded as owners were reluctant to give permission for sampling of pregnant or lactating sows and their piglets, for fear of stress causing abortion or cessation of lactation. Household identification number, animal species (cattle, pig, small ruminant) and gender (male, female) were recorded for each blood sample collected. In addition, age group, body condition score and anaemia score were recorded for all sampled cattle. Age was recorded as identified by owner or by tooth eruption pattern (category a: milk teeth, under 18 months of age; category b: one pair of permanent incisors, between 18 months and 3 years of age; category c: more than one pair of permanent incisors, over 3 years of age). Body condition score of the animal was initially scored on a scale of nine categories according to Nicholson and Butterworth (1986) [16] whereby animals are assessed as lean (L), medium (M) or fat (F), with each category subdivided into three classes, for example M-/M/M+, according to muscle mass and extent of fat deposition. Due to the low number of animals in some sub-categories, they were subsequently collapsed to the three main categories (L, M, F) for statistical analysis. Anaemia score was coded as normal (N) or anaemic (N+) as assessed by the veterinarian according to the colouration and perfusion of the mucous membranes in eyes and mouth of the animal. Anaemia score was assessed by the same experienced veterinarian throughout the whole study to exclude inter-observer bias. Anaemia scoring in cattle by visual assessment of the mucous membranes of the mouth and eyes has been previously shown to have a good correlation with blood haemoglobin levels [13]. For each homestead, the respondent was asked how many animals of each livestock species were owned by the household, and where these animals were grazed and watered. For statistical analysis, grazing and watering management for the animals of the respective species were coded into two categories (home/away): The category “home” was defined as animals that were fed/watered within the immediate compound of the homestead, either through grazing within the confines (usually tethered), or through feed/water being brought to the animal, whereas animals that were taken beyond the confines of the homestead compound for grazing/watering were assigned to the category “away”. Compounds were usually delineated by shrubbery or hedges and varied in size, with the majority of compounds being between 5 m and 20 m in diameter. When grazing and watering were analysed jointly as a combined variable (overall management) the management practice was categorised as “home” only when both feeding and watering practice were coded as “home”, otherwise the category “away” was assigned. All blood samples were analysed by PCR for the presence of the African animal pathogenic trypanosome species T. brucei s. l. , T. vivax, T. congolense and T. simiae using two established PCR protocols on each sample. The first, the internal transcribed spacer region PCR (ITS-PCR), detects and differentiates the trypanosome species affecting livestock [17]. The second PCR was specific for T. brucei s. l. [18]. Any sample that was positive by at least one PCR for T. brucei s. l. (cumulative results for ITS-PCR and Trypanozoon specific PCR run in parallel), was considered positive for T. brucei s. l. and further screened in pentaplicate for the presence of the human infective subspecies T. b. rhodesiense using a multiplex PCR targeting the SRA gene [19] and the product visualized using southern blotting. For each PCR reaction one 2 mm disc was cut from the samples on the FTA Card and prepared according to the manufacturer' s instructions. Briefly, the discs were washed twice in FTA purification reagent to remove PCR inhibitors from the sample, followed by two washes with 1xTE buffer to remove residual FTA purification reagent. Once dried, the discs were used to seed the reactions. Standard PCR amplifications were carried out in 25 µl reaction mixtures. PCR reaction conditions, primer sequences and adapted cycling conditions are shown in table 2. One positive control [genomic deoxyribonucleic acid (DNA) ] and one negative control (blank FTA disc) were run with each reaction. PCR products were separated by electrophoresis in a 1. 5% (w/v) agarose gel containing 0. 5 µg/ml ethidium bromide and visualised by ultraviolet light. To obtain DIG-labeled probe, genomic DNA of a known T. b. rhodesiense stock (LIRI024) (PLC and SRA) was amplified by multiplex PCR [19]. Products were separated by electrophoresis, extracted using a MiniElute Gel Extraction Kit (Qiagen) and labeled using DIG-High Prime labeling mixture (Roche, Mannheim, Germany) according to the manufacturer' s instructions. After probe yield estimation, labeled probe was used at a concentration of 25 ng/cm3 in hybridisation buffer, for each Southern blot. After DNA denaturation [20 min in denaturing solution (0. 5 M NaOH, 1. 5 M NaCl) 20 min in neutralising solution (0. 5 M Tris-HCl at pH 7. 5,3 M NaCl) ], transfer of the DNA from the agarose gel onto the nitrocellulose membrane, was performed on a vacuum blotter (QBiogene, Cambridge, UK), followed by UV-cross linking to the membrane. Hybridisation with the probe and visualisation was performed according to the DIG standard protocol (Roche) provided by the manufacturer. Questionnaire data were recorded in a Microsoft Excel spreadsheet (Microsoft Corporation, Redmond, USA). Test results were appended to this spreadsheet and samples classified as trypanosome positive if they were positive for any of the detectable trypanosome species by either one of or both the ITS-PCR and the T. brucei s. l. specific PCR. Agreement between the two PCRs for the detection of T. brucei s. l. was assessed by calculating Cohen' s Kappa coefficient and marginal homogeneity was assessed by the McNemar' s test [20]. A number of risk factors were considered for entry into a multivariable logistic regression model, on the basis of their potential biological significance, including administrative division, animal species, sex, and management practice (see factors under investigation). The relationships between the factors of interest and trypanosome infection status were initially examined using univariable logistic regression. Multicollinearity of risk factors was assessed using Pearson' s Chi-squared tests. Risk factors/explanatory variables were screened for each response variable. Factors with a likelihood ratio p-value of <0. 2 were passed forward for inclusion in the multivariable model for each response variable. The multivariable model was constructed by first including all variables that passed the initial screening and then dropping variables manually in a backwards elimination procedure based on the likelihood ratio test. Only variables that were significant at the 5% level in the likelihood ratio test were retained. Accurate data on age were only available for cattle and in order to allow this to be included, livestock species for cattle was further divided for each age group of cattle. The Wald test p-values were used to compare the effect of factor levels within the variables. The potential confounding effects of those variables not retained in the final model were assessed by refitting each variable in succession into the final model and inspecting the percentage change in the odds ratio of the retained variables, with a change greater than 20% being considered evidence of confounding [20]. Administrative division was forcibly retained in the multivariable model as this variable simultaneously represented samples collected at different time points (July and October 2004) and in different study sites. A significant two-way interaction of division with the species and cattle age variable was demonstrated for the final model for overall trypanosome status, and T. vivax infection status, but not for the T. brucei s. l. model, from which division was thus ultimately dropped. Finally the effect of the study design was taken into account by adding household as the random effect into the final model, examining the impact on the parameter estimates in the single-level model, and estimating the percentage of total variance occurring at the level of the random effect, using the latent variable approach [20]. The fit of the final fixed-effects model was assessed using the Pearson Chi-squared goodness-of-fit test, and its predictive ability was determined through the generation of receiver operator characteristics (ROC) curve. Statistical analysis was performed in R version 2. 8. 1 (The R foundation for Statistical Computing at http: //CRAN. R-project. org). Separate analyses were performed using T. vivax, T. brucei s. l. and T. b. rhodesiense status of the samples as the respective response variables. It was not possible to construct a separate multivariable logistic regression for T. b. rhodesiense infection status, due to few infection events. A total of 2773 livestock samples in 549 households (representing 85% of the targeted population) were collected in the two sampling sites (Funyula and Butula Division) in Busia District (see table 1). The total population of each livestock species was calculated from the number of animals which household respondents stated the household owned. A small fraction of livestock-keeping homesteads were excluded from sampling due to absence of the owner on sampling days (Funyula: 5/196 (2. 6%); Butula 9/367 (2. 5%) ). Household herdsizes ranged from one to 47 animals but in general the study area was characterized by small herdsizes with the majority of households (60%) owning five animals or less and only 5. 9% of households owned over 15 animals (figure 2). Approximately three-quarters (74. 3%) of livestock-owning households had cattle, close to two-thirds (63. 6%) owned small ruminants and just over one third (37. 9%) kept pigs (figure 2). At the univariable level, the chance of overall trypanosomiasis infection was significantly affected by the factors host species, grazing regime, overall management as well as cattle age and cattle anaemia score. The chances of infection with T. vivax infection or T. brucei s. l. infection were each significantly affected by host species, overall management, as well as cattle age. The chance of a T. brucei s. l. infected animal to be infected with T. b. rhodesiense was significantly affected by host species, with T. b. rhodesiense being significantly more likely to be detected in T. brucei s. l. infected pigs (47. 4%; 9/19) than in T. brucei s. l. infected cattle (17. 3%; 19/110). The results of the univariable analysis of the effects of the original and combined variables on trypanosome infection status as determined by PCR in livestock are presented for overall trypanosome infection status (table 3) as well as separately for T. vivax (table 4), T. brucei s. l. (table 5), and human infective T. b. rhodesiense (table 6). Due to low density of infection events in sheep and goats, these livestock species were combined into the small ruminant category for data analysis. Of the variables collected only for cattle (age group, anaemia status and condition score), age group remained as the only variable with a significant effect on trypanosome infection status when a multivariable model was fitted by backward selection to the cattle data. After adjustment for cattle age, cattle anaemia status was no longer significant in the multivariable analysis and was thus not included in the final model. In the separate multivariable mixed-effect model for overall trypanosome infections and for T. vivax infections, the chance of infection was significantly affected by the combined “species and cattle age” variable, and a significant interaction effect between this factor and division was observed. Division in itself was not a significant factor. For T. brucei s. l. only the combined “species and cattle age” factor remained significant in the final model. The overall management variable, which was significant at the univariable level was no longer significant in any of the three final models, after adjustment for the “host species and cattle age” variable by which it was confounded. The effect of the variables included in the final multivariable mixed-effect model constructed for overall trypanosome infection and separately for T. brucei s. l. and T. vivax are summarised in table 7. In the final model for overall trypanosomiasis, infection was significantly associated with the “species and cattle age” variable. The odds of overall trypanosome infection were significantly increased in cattle in the intermediate age group B (18–36 months) (OR = 2. 11,95% CI: 1. 18–3. 76) and the oldest group C (>36 months) (OR = 2. 04,95% CI: 1. 36–3. 06) as compared to the youngest cattle age group A (<18 months), which served as the reference. The odds of overall trypanosome infection in pigs (OR = 0. 51,95% CI: 0. 27–0. 96) and small ruminants (OR = 0. 23,95% CI: 0. 14–0. 39) were significantly decreased as compared to the youngest cattle age group. Whilst division in itself did not have a significant effect on the odds of overall trypanosome infections, there was a significant interaction effect between division and the “species and cattle age” variable. This interaction effect could be attributed to a significantly lower infection prevalence detected in pigs in Butula division, as compared to Funyula Division, and some variation in the infection prevalence in the different cattle age groups, which were not statistically significant (figure 3, table 7 multivariable model). Grazing and overall management, which were significant variables at the univariable level, were confounded by the “species and cattle age” variable and were no longer significant in the multivariable analysis and therefore dropped from the final model. In the final model fitted for T. vivax, the “species and cattle age” variable was significantly associated with T. vivax infection. With cattle in the youngest age category as the reference, the odds of T. vivax infection were significantly increased for cattle in the intermediate age group (OR = 2. 5,95% CI: 1. 27–4. 92) and significantly decreased for pigs (OR = 0. 34,95% CI: 0. 13–0. 88) and small ruminants (OR = 0. 35,95% CI: 0. 19–0. 66), however the comparative increase in odds of T. vivax infections in the oldest cattle age group was not statistically significant (OR = 1. 57,95% CI: 0. 93–2. 65). Again, division in itself did not have a significant effect on the odds of trypanosome infections, but there was a significant interaction effect between division and the “species and cattle age” variable, and division was therefore retained in the final model for T. vivax (table 7). Overall management, which had been significantly associated with T. vivax at the univariable level, was dropped from the final model, as it was no longer significant after adjustment for the “species and cattle age” variable. Of 138 samples positive for T. brucei s. l. by PCR in total, 110 (79. 7%) were detected by the species specific PCR [18] and 96 (69. 6%) were detected using ITS-PCR [17] (table 8). Cohen' s kappa test showed substantial agreement between the two PCRs (κ = 0. 65; 95% CI: 0. 57–0. 73). A test of marginal homogeneity, testing whether the disagreement is spread evenly (McNemar' s Test, χ2 = 2. 8, df = 1, p = 0. 09) was not significant, indicating that there was no significant systematic bias in the detection of T. brucei s. l. by either PCR method. The combined “species and cattle age” variable was the only variable significantly associated with T. brucei s. l. infection, and was therefore the only fixed effect retained in the final model for T. brucei s. l. (table 7). The odds of T. brucei s. l. infection were significantly increased for cattle of the oldest age group C (OR = 4. 42,95% CI: 2. 24–8. 73), and significantly decreased for small ruminants (OR = 0. 17,95% CI: 0. 06–0. 47), but there was no significant difference in odds of T. brucei s. l. infection in the intermediate cattle age group or in pigs as compared to cattle in the youngest age group. There was no significant interaction effect with the division variable, which was therefore dropped from the final model for T. brucei s. l. , along with the overall management variable which was no longer significant, after adjustment for the “species and cattle age” variable. Household was included as the random effect in the final models to account for the hierarchical structure of the data. Based on the multivariable mixed-effect models fitted for overall trypanosomiasis, T. brucei s. l. and T. vivax, 11. 8%, 44. 8% and 7. 0% of the total variance respectively occurred at the household level, as estimated using the latent variable approach [20]. The Pearson Chi-squared test statistic, calculated to assess the goodness-of-fit of the respective multivariate models, was ≪0. 001 (df = 7; p>0. 99) for the fixed-effects multivariable model for overall trypanosomiasis and ≪0. 001 (df = 7, p>0. 99) for the fixed effects multivariable model for T. vivax. This indicated that there was no evidence that the respective models did not fit the data well. The area under the ROC curve was 0. 74 for the overall trypanosomiasis model and 0. 72 for the T. vivax model indicating that both models have an acceptable predictive ability. In the densely populated agro-pastoral study district of Busia domestic livestock are the only trypanosome reservoir of epidemiological significance. The extensive cross-sectional data set collected for the present study through census sampling, achieved coverage of over 85% of the livestock population in the sampling area permitting detailed analysis of the trypanosome infections at the household level. PCR is a sensitive molecular tool, which can increase the number of trypanosome infections detected at least two-fold when compared directly to the microscopy results from the same sample set of cattle [21], [22]. Cattle were identified as the livestock species with the highest prevalence of all trypanosome infections (20. 1%), whereas the prevalence in small ruminants was low (<5%). The prevalence in pigs differed significantly between the two sampling sites, with a prevalence of 17. 4% in pigs in Funyula Division as compared to only 8. 4% in Butula Division (figure 3). Considering individual trypanosome species, T. congolense and T. simiae infections were detected in under 0. 3% and 1. 5% of all samples, respectively, and these infections were not separately analysed. Several studies conducted in Busia, have also shown low T. congolense infection rates of under 3% in cattle and close to 0% in small ruminants and pigs [23], [24]. In cattle and small ruminants, T. vivax was the most prevalent trypanosome species detected. In pigs, T. brucei s. l. infections predominated, with a similar prevalence to that found in cattle. There was substantial agreement for detection of T. brucei s. l. infections between the T. brucei s. l. specific PCR [18] and the ITS-PCR [17]. Both methods were employed in parallel to increase the sensitivity of detection for T. brucei s. l. , to allow for subsequent detection of the zoonotic subspecies, T. b. rhodesiense. Previous observations of a high prevalence of PCR detected trypanosomiasis in small ruminants (20–25%) in Busia District by Ng' ayo and colleagues (2005) were not observed in this study [25]. The results presented here support microscopy and PCR studies in Western Kenya and Eastern Uganda, in which cattle were identified as the most important reservoir of trypanosomiasis, low levels of infection were detected in small ruminants and highly variable infection prevalence depending on sampling sites were seen in pigs (2–20%) [8], [23], [26]. Differences in trypanosome prevalence between livestock species have previously been attributed to reduced susceptibility of small ruminants resulting in a low or transient parasitaemia [27], or lower exposure of small ruminants to tsetse bites [28]. The latter was supported by the identification of cattle and pigs as the major source of blood meals of both Glossina fuscipes fuscipes and Glossina pallidipes in this region [29]–[31]. Protection of cattle from pathogenic trypanosome infections is at the centre of productivity-motivated control strategies in this region. Trypanosome prevalence in cattle in this study was shown to significantly increase with age. A comparatively low prevalence of trypanosomiasis, observed in young cattle has previously been explained by either an inherent resistance to trypanosome infections in young animals [32], tsetse feeding preferences for adult cattle due to size and olfactory cues [33]–[35], or lower tsetse exposure of young cattle due to separate management from the rest of the herd [36]. However as animals remain infected unless treated, the higher prevalence of trypanosomiasis observed in adult cattle in the present study may simply be a result of older cattle having been exposed to tsetse for a longer time-span and thus having a higher cumulative risk of infection. The two key indicators commonly used for the clinical diagnosis of trypanosomiasis, namely anaemia and poor body condition [37], performed poorly in the present study. There was no significant difference in trypanosome prevalence according to cattle condition score. Whilst the overall chance of a trypanosome infection was significantly increased in cattle classified as anaemic this was no longer the case after adjustment for age group and only a minority of infected animals were classified as anaemic. Over 80% of trypanosome infected cattle did not display pallor of mucous membranes. The chance of T. brucei s. l. infection was not significantly increased in anaemic cattle, confirming that anaemia is more commonly associated with T. congolense and T. vivax rather than T. brucei s. l. infections [9]. However, even the sensitivity for detecting T. vivax infections based on anaemia status was low in the present study, with only 20% of infected cattle being identified as anaemic by visual inspection of mucous membranes. This may either be attributed to insufficient sensitivity of visual examination of mucous membranes to detect lower grade anaemia or a certain degree of trypanotolerance in zebu cattle in Busia, resulting in sub-clinical infections. Anaemia status as classified by more elaborate methods such as packed cell volume (PCV) has been shown to be moderately sensitive for the detection of trypanosome infections, with a sensitivity of 56% being recorded using a cut-off point of PCV below 24% as indicator of anaemia [38]. However, whilst measurements of PCV or haemoglobin allow a more precise and (depending on the selected cut-off point) more sensitive determination of anaemia status, such techniques require either the use electricity (centrifuge for PCV) or fairly expensive equipment and disposables (hand held haemoglobinometer and haemocuvets), neither of which are an option for routine pen-side testing in a poor rural area such as Busia. In the present study pallor of the mucous membranes was elected as the indicator of anaemia to reflect the criteria on which veterinary clinicians or animal health workers would base their treatment decisions. PCR will detect a significant proportion of sub-clinically infected cattle, which contribute to the reservoir of trypanosome infection in this endemic area. With very low average profit margins on livestock production in Busia [14], there is limited scope for sophisticated diagnostic procedures and block treatment of cattle. Treatment of visibly ill cattle with trypanocidal drugs, as practiced at present [13], limits immediate economic losses at the household level but is unlikely to impact on the reservoir of infections and impact on transmission of the parasite, which would be necessary for sustainable control. Human infective T. b. rhodesiense were detected by PCR in a total of 19/1260 cattle (1. 5%) and 9 out of the 312 pig samples (2. 9%). PCR can detect sub-clinical infections with very low parasitaemia. However even low-grade infections must still be regarded as transmissible as only a single trypanosome is required to infect a tsetse fly [39] and it has been demonstrated that even during chronic, low parasiaemic phases of T. brucei s. l. infections in cattle, sufficient parasites are present to infect tsetse [40]. Cattle are the most important reservoir of T. b. rhodesiense in this region [41], [42], with up to 18% of cattle infected in an epidemic focus in Uganda [43]. The comparatively low prevalence of T. b. rhodesiense detected in cattle and pigs during the current study nevertheless still poses a threat to human health in this area of Western Kenya, as was demonstrated by a case of sleeping sickness reported from Busia District in early 2006 [2] and the last recorded case from neighbouring Teso District, diagnosed in 2008 (Alupe Hospital, Western Kenya, pers. comm.). Only sporadic cases of sleeping sickness cases have been reported from Busia over the last ten years. It has been suggested that anthropogenic changes, especially increased cultivation, played a role in reducing the tsetse habitat and tsetse densities and thus reducing the overall probability of transmission to humans [44]. However, a degree of under-detection of human cases, as has been reported for Uganda [45], may also play a role in the low number of sleeping sickness cases reported from Busia. Previous studies have demonstrated that the trypanosome prevalence detected in livestock varied significantly according to the grazing routes and type of watering places frequented. Natural river watering sites were transmission hotspots [46], [47] and cattle and small ruminants tethered for grazing within the village showed a lower probability of becoming infected [24], [26]. In the present study the majority of livestock (61%) were confined to the immediate surroundings of their respective homestead, with feed and water being provided in situ. There appeared to be a significant protective effect of this strategy when data were analysed at the univariable level. However, management practice was confounded by livestock species: cattle (with the highest infection prevalence) were more likely to be taken out of the compound for feeding and watering than the other livestock species, creating the impression of lower odds of infection in animals managed within the compound. Overall, data collected on management regimes in the present study did not provide evidence that confining animals within the homestead compounds decreased the likelihood of animals becoming infected. Of a total of 329 livestock samples detected to be trypanosome infected, over 50% (171) were taken from animals that did not leave the immediate vicinity of their homestead. The management practice of maintaining livestock within the immediate vicinity of the homestead rather than taking animals for grazing on communal land and watering at the river is widespread in the sampling areas of Busia, in particular for small herds. However such management appeared to provide only a very limited protective effect against trypanosome infections in livestock. Evidence of a considerable proportion of infections having been acquired by livestock maintained on homestead compounds, pointed towards an important element of transmission in the vicinity of the homestead compounds in the epidemiology of trypanosomiasis in Busia. Due to the higher probability of exposure of humans to tsetse bites, such transmission would also increase the risk of transmission of the human infective T. b. rhodesiense from its livestock reservoir to the human population.

Rhodesian sleeping sickness caused by Trypanosome brucei rhodesiense is a parasitic disease transmitted by tsetse flies which is fatal in humans if it is not treated. The parasites also infect a range of animal species in which they do not cause acute disease and may co-exist with other non human infective parasites. Busia District (Western Kenya) is a historic sleeping sickness focus. Human cases of this disease are still reported occasionally in Busia and neighbouring Teso District, most recently in 2008, showing that the human infective parasite species are still present in the area. However, trypanosomes in this region are mainly regarded as a threat to the productivity of domestic livestock and the responsibility for trypanosomiasis control has shifted from the state to livestock holders. To examine whether farmer-based control strategies can be successful, this study assessed the factors that influence trypanosomiasis in livestock at the local level. The study showed that cattle are the livestock species most frequently affected by trypanosomes. However infection in cattle was not necessarily associated with signs of disease; furthermore pigs were shown to be important carriers of the human infective parasite. The treatment of only visibly diseased cattle to avoid losses in productivity will not successfully control the parasite in the long term. Keeping livestock in the vicinity of the homesteads also did not protect the animals from trypanosome infection. This indicated that the tsetse fly transmits the parasite in close proximity to human habitation, which could increase the risk of humans being infected.

Abstract Introduction Materials and Methods Results Discussion

infectious diseases/neglected tropical diseases infectious diseases/protozoal infections

2011

Factors Associated with Acquisition of Human Infective and Animal Infective Trypanosome Infections in Domestic Livestock in Western Kenya

Kaposi’s sarcoma herpesvirus (KSHV) causes Kaposi’s sarcoma and certain lymphoproliferative malignancies. Latent infection is established in the majority of tumor cells, whereas lytic replication is reactivated in a small fraction of cells, which is important for both virus spread and disease progression. A siRNA screen for novel regulators of KSHV reactivation identified the E3 ubiquitin ligase MDM2 as a negative regulator of viral reactivation. Depletion of MDM2, a repressor of p53, favored efficient activation of the viral lytic transcription program and viral reactivation. During lytic replication cells activated a p53 response, accumulated DNA damage and arrested at G2-phase. Depletion of p21, a p53 target gene, restored cell cycle progression and thereby impaired the virus reactivation cascade delaying the onset of virus replication induced cytopathic effect. Herpesviruses are known to reactivate in response to different kinds of stress, and our study now highlights the molecular events in the stressed host cell that KSHV has evolved to utilize to ensure efficient viral lytic replication. Kaposi’s sarcoma-associated herpesvirus (KSHV) is a human tumor virus in the family of gamma2-herpesviruses. KSHV is the etiologic agent of Kaposi’s sarcoma (KS) and other KSHV-associated lymphoproliferative diseases such as primary effusion lymphoma (PEL) [1,2]. KSHV genome consists of linear double-stranded DNA (dsDNA), and like other herpesviruses, the virus displays two modes of infection in the infected cells, the latent and lytic replication phase. Upon entry into the host cell nucleus, the linear dsDNA genome circularizes forming a non-integrated viral episome that persists as multiple copies in the latently infected cells [3]. The latent infection (latency) provides an immunologically silent mode of persistence, whereas the lytic replication phase allows replication and production of new virions to be shed and transmitted to new cells and hosts. The switch between the latency and lytic replication (virus reactivation) is a critical step in viral pathogenesis. Although the KSHV-associated tumors typically show low level of virus reactivation [4,5], epidemiological studies support the importance of lytic replication in the initiation and progression of KS [6–8]. Despite of active research, the regulation of viral reactivation is not completely understood. However, significant advances have been made in recent years, and the reported mechanisms of KSHV reactivation involve hypoxia [9–11], reactive oxygen species [12], inflammation [13–15], activation of cellular kinases [16–20] and epigenetic mechanisms [21–25]. KSHV reactivation can also be chemically induced e. g. with certain kinase agonists (TPA) and chemical inhibitors affecting histone acetylation (HDAC inhibitors) or DNA methylation (reviewed in [26]). Recent studies by several groups have demonstrated that the intracellular viral genome has chromatin structures similar to that of the host chromosome (reviewed in [22]). The latent KSHV genome is epigenetically modified with methylation at CpG dinucleotides as well as mutually exclusive activating and repressive histone modifications [27–29]. The bivalent chromatin structure represents a poised state of repression during viral latency, which can be rapidly reversed once the lytic cycle is induced, and enables the virus to fine-tune its gene expression patterns in response to changes in virus infected cells. Further support for the importance of epigenetic regulation in the switch from latency to lytic replication was provided by the demonstration of the cohesin subunits as major repressors of KSHV lytic gene activation suggesting that cohesins could be a direct target of butyrate-mediated lytic induction [30]. Other recently identified epigenetic regulators of KSHV reactivation include the H3K27me3 histone methyltransferase of the Polycomb group proteins, EZH2 [28], HDAC class I and II [25], and the histone demethylase JMJD2A [31]. To discover novel mechanisms regulating KSHV reactivation we designed and performed a small interfering RNA (siRNA) screen using a library of siRNAs specific for human genes involved in epigenetic processes. In this screen we assessed which epigenetic enzymes help the virus to maintain latency. We identify MDM2, an E3 ubiquitin ligase, as a novel modulator whose depletion by siRNA accelerates KSHV reactivation. We also show that MDM2 down-regulation leads to subsequent activation of p53 and p21 as well as induction of a p21-dependent cell cycle arrest, which are required for the induction of efficient viral lytic replication cascade. To identify novel regulators of KSHV reactivation we performed a siRNA screen using a custom-made siRNA library targeting 615 human genes with Gene Ontology (GO) annotations related to epigenetics, chromatin remodeling/maintenance, and co-regulatory functions, and consisting of two independent siRNAs for each gene [32]. Before reverse transfection, the siRNAs, mixed with transfection reagent and components of the extracellular matrix, were spotted onto a microplate-sized array plate and analyzed by the cell-spot microarray technique (CSMA; [33]. For the screen, we used SLK cells stably infected with a recombinant KSHV (rKSHV. 219) which during latency constitutively expresses green fluorescent protein (GFP) under the control of the cellular EF-1α promoter, and upon reactivation the red fluorescent protein (RFP) from the promoter of the viral early lytic gene PAN [34]. The workflow of the primary screen is depicted in S1A Fig (for details see S1 Methods). Among the siRNAs that enhanced reactivation, one of the strongest and the most reproducible ones were the siRNAs against MDM2 (siMDM2). The depletion of MDM2 led to RFP levels that were three SDs above the mean RFP value for the screen (S1 Fig). In this study, we pursued validation and characterization of the role of MDM2 in KSHV lytic replication. To validate the results of the screen, we tested two additional MDM2 siRNAs targeting different, non-overlapping sequences of the MDM2 transcript. The siRNAs were pooled and transfected into iSLK. 219 cells, a recently developed cell clone derived from SLK cells stably infected with rKSHV. 219 [35]. These cells contain an exogenous copy of the viral gene RTA that can be induced by doxycycline. Addition of low doses of doxycycline results in efficient RTA synthesis (S2A Fig) that triggers virus reactivation in approximately 5–15% of cells at 24 hours post induction (hpi) (S2B and S2C Fig). This can be substantially augmented by the combination of doxycycline (Dox) with TPA (TPA/Dox) or NaB (NaB/Dox) which increase the fraction of RFP-positive cells to 80–90% within 24 hours of treatment (S2B and S2C Fig). In the absence of Dox, treatments with TPA or NaB alone do not result in lytic reactivation (S2B and S2C Fig), thereby providing an invaluable control to monitor possible pleiotropic effects of the two compounds. To assess the effect of MDM2 depletion in the iSLK. 219 cells [35] we incubated the siRNA transfected cells for 72 hours (h), and treated the cells by suboptimal doses of doxycycline that alone produce no or very low level of RFP induction. After fixation, cells were imaged and analyzed for RFP expression using automated high-content fluorescence microscopy. The depletion efficiency of the MDM2 siRNAs was 62% as monitored by quantitative real time PCR (qRT-PCR). As shown in Fig 1A and 1B, and confirming the results of the screen, depletion of MDM2 led to a 3-fold increase in the number of RFP—positive cells compared to the control siRNA transfected cells. To further validate the results of the screen in cells that are naturally infected by KSHV, we silenced MDM2 expression in BC-3 cells, a patient-derived primary effusion lymphoma (PEL) cell line[36]. BC-3 cells were transduced with lentiviruses expressing shMDM2, or shCtrl, and 72 h later virus reactivation was induced by TPA treatment. We first assessed whether the depletion of MDM2 induced spontaneous reactivation by analyzing the mRNA levels of the lytic genes ORF50 (encoding for RTA; immediate early), ORF57 (encoding for MTA; delayed early), vGPCR (intermediate) and K8. 1 (late) in non-induced BC-3 cells. Whole cell extracts were collected at indicated times and viral lytic transcripts were analyzed by qRT-PCR (Fig 1C and 1D). Depletion of MDM2 induced the expression of the viral transcripts by 3. 2 to 5. 6 fold compared to controls (Fig 1C) indicating spontaneous viral reactivation. Similarly, during TPA-induction MDM2 depletion led to a rapid increase of all tested viral lytic transcripts that ranged from 2. 5 (ORF50) up to 15 fold (ORF57) at 4 hpi (Fig 1D). Interestingly, over time, the magnitude of the up-regulation gradually decreased, reaching control levels at 48 hpi. Thus, silencing of MDM2 in PEL cells led to spontaneous reactivation and accelerated the kinetics of lytic gene expression, without altering their maximal levels during the time course studied. MDM2 is the major regulator of the tumor suppressor p53 and targets p53 to proteasomal degradation through its ubiquitin E3 ligase activity [37]. While p53 is inactivated during KSHV latency [38–45] its potential role during the lytic phase of the virus replication cycle has not been investigated. The possibility that the transcription factor p53 could be required for efficient viral lytic gene expression prompted us to conduct a comprehensive, unbiased analysis of the p53 chromatin binding sites during the reactivation of BC-3 cells. To this end, we carried out chromatin immunoprecipitation using antibodies against p53, followed by genome-wide deep sequencing (ChIP-seq) analysis of the associated DNA. Non-specific IgG antibody was used as a negative control. BC-3 cells were subjected to ChIP-seq after TPA induction for 0 or 24 h. As a positive control for detection of the canonical p53 targets, non-induced BC-3 cells were treated with Nutlin-3 (here referred to as Nutlin), a small molecule that binds and inhibits MDM2, stabilizing p53 and inducing a potent p53 response. Compared to the DMSO-treated cells, the TPA treatment induced a global activation of p53 that could be visualized by averaging the sequencing signal over the whole cellular genome (S3A Fig). Specific sequencing signal was found at regions preceding known p53-target genes, such as MDM2, CDKN1A (p21Cip1), P53R2 and PAG608 (Fig 2A). For some genes, the sequencing peak obtained from TPA-treated cells was comparable to that of Nutlin treated cells (Fig 1A, p53R2; S3B Fig, red arrow heads). Pathway analysis indicated that the genes close to the 300 most significant p53 binding sites (p<0. 01) were involved in Cell cycle arrest, Apoptosis, DNA repair and p53 regulation (i. e. MDM2) (S3C Fig). 94 of the 99 peaks with the lowest p-values from the TPA sample overlapped with the peaks identified from Nutlin-treated cells (S3D Fig), confirming that in BC-3 cells TPA treatment induced a bona fide p53 response. Notably, the ChIP-seq analysis of TPA- or Nutlin-treated cells did not produce statistically significant p53 binding events on the viral genome. The sequencing signal from TPA- or Nutlin- treated cells (Fig 2B red and blue lines, respectively) was comparable to the background noise obtained with the nonspecific IgG controls (Fig 2B, light blue and green lines, respectively) despite the high copy number of viral genomes per cell in BC-3 and PEL cells in general [46]. The absence of p53-binding sites on the viral genome prompted us to perform a quality control to confirm that the p53 binding sites identified in the cellular DNA corresponded to the expected consensus sequences [47]. A de novo identification of the p53 DNA-binding sequence returned its known consensus sequence (Fig 2C) [47], for both TPA and Nutlin treated cells. Thus, if p53 was involved in virus reactivation, this effect occurred through the activation of cellular rather the viral genes. Of the different pathways induced by p53, we investigated the role of cell cycle arrest, a phenomenon known to occur during the lytic phase of herpesviruses and other DNA viruses [48]. Of particular interest was p21Cip1 (p21), a strong inhibitor of cell cycle progression [49], which was also identified in our ChIP-seq analysis of TPA induced cells (Fig 2A and S3B and S3C Fig). To validate the results of the ChIPseq experiment, we determined the level of p53 and its target p21 by immunoblotting of cell extracts collected at indicated times after reactivation with TPA or Nutlin treatment (Fig 2D). Compared to the DMSO controls (0 h), induction with TPA caused a small increase in the levels of p53 and a strong induction of p21 (Fig 2D). The up-regulation of p21 was fast and could be detected already 4 h after TPA treatment in BC-3 cells (S4E Fig). The p21 levels also increased in iSLK. 219 cells treated with TPA/Dox or NaB/Dox (Fig 2E). To test if this effect was due to viral lytic gene expression, we compared by WB the levels of p21 in iSLK. 219 cells treated with DMSO (non induced), Dox, TPA or TPA/Dox (S3E Fig). Viral lytic gene expression was monitored with antibodies against MTA (ORF57, early lytic gene). In iSLK. 219 cells, TPA treatment is not sufficient to trigger lytic reactivation (see also S2B and S2C Fig). However, the levels of p21 in cells treated with TPA alone for 4h were comparable to those in cells treated with TPA/Dox that results in lytic reactivation (S3E Fig). No increase in p21 levels was observed after 4h treatment with Dox that leads to the synthesis of RTA (ORF50) from a plasmid stably maintained in the iSLK. 219 cells. Four hours after Dox treatment the synthesis of lytic genes downstream of RTA is still not detectable by WB. Thus, the fast increase in p21 levels was not due to viral lytic gene expression but appeared to be an intrinsic effect of the TPA treatment. To confirm this, we incubated non-infected SLK cells with similar doses of TPA/Dox or NaB/Dox and monitored by WB the levels of p21 over time. Nutlin treatment was used as a positive control. Although with slower kinetics, TPA and NaB induced an increase of p21 levels by 12 h (S3F Fig). These results are consistent with numerous earlier reports obtained in other cell types [50–53]. To test whether the induction of p21 was p53-dependent we depleted p53 in iSLK. 219 cells and then monitored the levels of p21 after treatment with DMSO, Nutlin or TPA/Dox. At 48 h after transduction with lentiviruses expressing shRNA against p53 (sh-p53) or nonspecific non-targeting controls (sh-Ctrl), iSLK. 219 were treated with indicated compounds for 4 h and the levels of p53 and p21 were monitored by WB of whole cell extracts (Fig 2F). The specificity of the p21 antibody was confirmed by including iSLK. 219 cells treated with siRNAs against p21 (si-p21). Compared with DMSO controls, treatment with Nutlin increased the levels of p53 and p21. As in our previous results in BC-3 cells (Fig 2D), the levels of p21 in Nutlin and TPA/Dox treated cells were comparable despite the large difference in the levels of p53 (Fig 2F, Nutlin and TPA/Dox, sh-Ctrl lanes). Depletion of p53 in DMSO- and Nutlin-treated cells was very efficient and abrogated the induction of p21 (Fig 2F, DMSO and Nutlin). Upon TPA/Dox treatment, however, p21 was induced even in the absence of detectable p53 (Fig 2F, TPA/Dox, sh-p53 lanes), indicating that upon TPA/Dox induction the levels of p21 are regulated by both p53-dependent and -independent mechanisms. To test if an increase in p53/p21 levels would favor virus lytic gene expression we induced iSLK. 219 cells with Dox or TPA/Dox in the presence of Nutlin for 24 h, and measured the levels of RFP (virus reactivation) by automated microscopy. In each experiment, we monitored the stabilization of p53 by Nutlin using immunofluorescence (S4A Fig). In parallel experiments using non-induced iSLK. 219 cells, the decrease in the number of cells after Nutlin treatment was used to measure the proliferation arrest in response to p53 stabilization (S4B Fig). Consistent with the results obtained by depletion of MDM2, inhibition of MDM2 by Nutlin led to a two-fold increase in virus reactivation (Fig 3A and 3B). Conversely, depletion of p53 using RNAi decreased RFP expression induced by TPA/Dox in iSLK. 219 cells (Fig 3C and 3D). In these experiments the silencing of p53 was monitored by WB analysis (Fig 3D, insert). Similar results were obtained in TPA and NaB treated BC-3 cells, where the levels of lytic gene expression were monitored by qRT-PCR and WB (S4C–S4E Fig). We then addressed the relevance of p21 in mediating the lytic reactivation. Similarly to the depletion of p53, siRNA-mediated depletion of p21 in TPA/Dox induced iSLK. 219 cells decreased fluorescence intensity of RFP (Fig 3E and 3F) and other viral lytic genes (Fig 3G). To address the role of p21 for efficient viral lytic replication in a biologically relevant KSHV-infection system, we chose to use two PEL cell lines, TPA-treated BC-3 and TREx BCBL1-Rta (here referred to as BCBL1RTA) cells that can be induced to lytic replication through doxycycline-inducible expression of RTA [54]. We stably depleted p21 in both cell lines by transducing them with lentiviruses expressing a control unspecific shRNA (sh-Ctrl) or two different sh-p21 constructs (sh1-p21, sh2-p21) followed by puromycin selection. Twelve days after selection, both BC-3 and BCBL1RTA cell lines expressing the sh-p21 constructs had growth kinetics undistinguishable from their respective sh-Ctrl transduced cells (S5A and S5B Fig). As shown for BC-3 and iSLK. 219 cells (Fig 2D and 2E), the levels of p21 mRNA also increased in these new cell lines upon TPA (Fig 4A, BC-3 sh-Ctrl) or doxycycline (Fig 4B, BCBL1RTA shCtrl) treatments. The silencing efficiencies of p21 were monitored by qRT-PCR (S5C and S5D Fig) and WB (Fig 4C). Stable depletions of p21 had little or no effect on the expression levels of the latent gene ORF73 (Fig 4D) at 24h after TPA or Dox induction in BC-3 or BCBL1RTA cells, respectively (Fig 4D). However, the expression levels of early, intermediate and late lytic genes were significantly decreased at 24 h (Fig 4E) or 48 h (Fig 4F) after induction of reactivation in BC-3 and BCBL1RTA, respectively. Of the two cell lines used, BC-3 cells were more sensitive to the depletion of p21 than the respective BCBL1RTA cells (compare Fig 4E and 4F). Thus the expression of p53 and p21 favored efficient viral lytic gene expression in both TPA-treated and RTA-overexpressing naturally KSHV-infected cells. Moreover, the decreased lytic gene expression in the p21-depleted cells resulted in a reproducible and significant delay in the onset of cytopathic effect (cell lysis), a known consequence of efficient virus replication (Fig 4G and 4H). Interestingly, despite the significant reduction of viral lytic gene expression, we saw only a modest decrease (up to approximately 30%) in the amount of infectious viruses released in the supernatant of p21-depleted BCBL1RTA cells compared with their respective shCtrl expressing cells after reactivation by doxycycline for 24h (S5E and S5F Fig). p21 can arrest the cell cycle at G1/S or G2 phase by inhibiting the cyclin/Cdk complexes [49]. To assess how these properties of p21 contribute to the KSHV lytic reactivation, we developed an automated, image-based assay in which we used iSLK. 219 cells and monitored viral reactivation and cell cycle progression at single-cell level. The expression of RFP served as a marker of virus reactivation, while cell cycle progression was monitored by immunofluorescence analysis using antibodies against histone H3 phosphorylated on serine 10 (pH3-S10). In dividing cells that enter the G2 phase, this antibody gives a punctate signal that labels the sites of chromatin condensation. As cells progress towards prometaphase, chromatin condensation continues and the levels of pH3-S10 increase proportionally, reaching a maximum at metaphase and anaphase, when the cellular DNA is tightly packed into chromosomes [55]. At the end of mitosis, histone H3 is rapidly dephosphorylated, and is undetectable in G1 and S phases. These stages could be easily distinguished in the iSLK. 219 cells with automated fluorescence microscopy (Fig 5A). Using the fluorescence-intensity of the pH3-S10 staining and automated image analysis, cells were assigned to G1/S (no pH3-S10 signal, panel' a' ), G2 (weak and punctate pH3-S10 signal, panels' b' and' c' , respectively) or M phase (bright pH3-S10 signal; panels' d-f' ) (Fig 4A and S6A Fig). To further demonstrate that cells displaying pH3-S10 fluorescence are indeed in G2 or M-phase, we co-stained non-induced iSLK. 219 cells with cyclin B1, another cellular marker of G2 and M-phase. As expected, the two markers were perfectly correlated (S6B Fig). Cells in G2 displayed a weak and punctate pH3-S10 signal and cytoplasmic cyclin B1, while cells in M-phase had bright and diffused pH3 S10 fluorescence and nuclear cyclin B1 (S6B Fig). To test the accuracy of our image analysis method, we used as controls Nutlin, that arrests cells in G1/S, and Etoposide, an inhibitor of cellular topoisomerase-II that causes DNA damage and a G2/M cell cycle arrest (S6C and S6D Fig). Compared to DMSO controls, both Nutlin and Etoposide treatments strongly decreased the number of M phase cells (S6D Fig). Consistent with a robust G1/S arrest, Nutlin also decreased the number of cells in G2, while Etoposide increased the fraction of cells in G2 by more than six fold compared to DMSO treated samples (S6C and S6D Fig). Treatment with TPA/Dox led to an increase in the fraction of cells in G2, suggesting that during reactivation iSLK. 219 cells arrest in G2 (S6C and S6D Fig, TPA/Dox). In parallel experiments, we excluded cross-talk between the RFP and Alexa488 fluorescence signal used to detect pH3 S10 (S7A Fig). Using similar drug treatments, the image analysis based cell cycle measurements were confirmed by traditional FACS analysis using propidium iodide to stain DNA (S8 Fig). Also in this case, the induction of iSLK. 219 cells with TPA/Dox induced an increase in the fraction of S- and G2/M-phase cells, similar to the effect of etoposide treatments (S8B–S8G Fig). In these experiments the settings of the FACS detection were adjusted such that the RFP fluorescence did not contribute to the detection of PI (S8D Fig). Although very sensitive and in agreement with the results of the image analysis, the PI FACS analysis method could not distinguish between cells in G2- or in M-phase. Based on the pH3 S10 imaging method, about 85% of non-induced iSLK. 219 cells were at G1/S phase, 10% at G2 and 5% at M phase (Fig 5B and 5C, DMSO). At 24 h after TPA/Dox treatment, 62% of cells expressed RFP (compare with S2C and S2D Fig). The number of cells in M phase decreased from 4. 6% to 1. 5%, and the majority of reactivated cells (77% of all cells and 86% of RPF positive cells) displayed the weak pH3-S10 signal indicating an arrest in G2 phase (Fig 5B and 5C, TPA/Dox). To confirm that reactivated cells arrest in G2 we repeated the experiment using antibodies against cyclin B1. Indeed, compared to DMSO controls (S7B Fig), the vast majority of RFP positive cells also contained cyclin B1 (S7C Fig). In the absence of Dox, TPA treatment did not induce G2 arrest, but only slightly decreased the number of M phase cells (Fig 5C and 5D, TPA). Thus viral lytic gene expression coincided with the G2 arrest. To test whether depletion of p21 affected the G2 arrest observed during KSHV reactivation we silenced p21 for 48 h and subsequently monitored the levels of pH3-S10 in iSLK. 219 cells treated with TPA/Dox for 24 h (Fig 5E–5G). As in previous experiments, induction with TPA/Dox induced a G2 arrest in cells treated with si-Ctrl (Fig 5E–5G, si-Ctrl). However, depletion of p21 in the TPA/Dox treated cells resulted in a significant decrease in the fraction of cells in G2 and a correspondent increase in the number of mitotic profiles to a level approaching the non-induced cells (Fig 5E–5G). Upon lytic reactivation, the levels of p21 rise within the first 4 h, and increasing its levels with Nutlin did result in a complete G1/S arrest in latently infected cells. Why didn' t reactivated cells then arrest in G1/S? If a mechanism existed to inactivate the G1/S checkpoint during the lytic phase, then reactivated cells would move on to the S-phase and reach the G2. To test this possibility, we pre-treated non-induced iSLK. 219 cells with Nutlin for 18 h to induce a complete G1/S arrest, and then triggered virus reactivation by TPA/Dox in the presence of Nutlin. DMSO served as a control (Fig 6A). Cells were then fixed 18 h after reactivation and, the cell cycle progression was monitored by image analysis of pH3-S10 as described before. As expected, after Nutlin treatment the non-induced cells arrested in G1/S (Fig 6B and 6C, Nutlin/DMSO). Again, TPA/Dox treatment arrested most of the cells in G2 (Fig 6B and 6C, DMSO/TPA/Dox). Strikingly, despite the Nutlin pre-treatment, the addition of TPA/Dox led to an increase in the fraction of cells in G2 while those in G1/S phase decreased (Fig 6B and 6C, Nutlin/TPA/Dox). This was not observed in the non-infected SLK cells used as a control (S9 Fig). In the absence of Dox, TPA treatment alone does not reactivate the virus in iSLK. 219 cells and did not allow the cells to overcome the Nutlin-induced G1/S arrest (Fig 6C, Nutlin/TPA). Thus, during virus reactivation the G1/S checkpoint is inactivated while the G2/M checkpoint remains active. The G2 arrest in lytic cells suggested activation of a DNA damage response (DDR). We and others have previously reported DNA damage induced by KSHV [56]. We therefore next assessed whether TPA/Dox or NaB/Dox activated DDR in iSLK. 219 cells. DDR was monitored by immunofluorescence using antibodies against the phosphorylated forms of the Ataxia telangiectasia mutated (pATM, S1981) kinase, histone H2AX (γ-H2AX), and checkpoint kinase 1 (pChk1, S317) proteins (Fig 6D–6F). Non-induced iSLK. 219 cells treated with Etoposide (10μM) served as positive controls (Fig 6F, dashed red line). Based on automated quantitative immunofluorescence analyses TPA/Dox-induced RFP-positive cells expressed robustly all DDR markers to a similar degree as the Etoposide-treated cells (Fig 6D–6F). The kinetics of DDR response was slower in the NaB/Dox treated cells but became comparable to those of TPA/Dox treated cells at 48 hpi (Fig 6F). Treatments with TPA or NaB in the absence of Dox, which is not sufficient to induce virus reactivation in this cell line, did not result in detectable DDR at 24 h (Fig 6G). Thus, reactivated cells accumulated DNA damage and activated DDR. Similar to the G2 arrest, the DDR was not due to pleiotropic effects of the TPA of NaB treatments but required viral lytic gene expression. When cells accumulate DNA damage, cell cycle arrest is necessary to allow DNA repair and prevent the propagation of damaged DNA to dividing cells, which would otherwise induce cell death during the M phase. Given the extent of DDR accumulated during the lytic phase, we hypothesized that in iSLK. 219 the G2 arrest could prevent premature cell death. The G2 arrest is enforced by the effectors of DDR, such as the kinases Chk1 and 2. We therefore tested the ability of different inhibitors of DDR (S10A Fig) to impair the G2 arrest and cause death in cells reactivated by TPA/Dox (lytic reactivation). The assay was optimized using increasing concentrations of Etoposide and caffeine, a broad inhibitor of ATM, ATR and other cellular kinases. The extent of DNA damage and cell cycle progression were monitored by immunofluorescence using antibodies against γH2AX and pH3-S10, respectively. Cells were treated with caffeine 1 h prior to addition of Etoposide, fixed 48 h later and processed for image analysis. Through inhibition of the upstream signals of the DDR (ATM/ATR), caffeine was very effective in inhibiting the Etoposide-induced G2 arrest and induced ‘mitotic catastrophe’ (S10B and S10C Fig; compare insets 1,2 and 3). Unlike caffeine, a specific inhibitor of ATM (KU-55933) was not effective in overcoming the G2 arrest caused by Etoposide (S10D Fig). A specific inhibitor of ATR, VE-821 (S10D Fig), and inhibitors of Chk1/2 (ADZ-7762) or Chk1 (MK-8776), did restore the number of M phase cells when low concentrations of Etoposide were used (S10D and S10E Fig). Having established the appropriate experimental conditions and the most effective concentrations for each of the inhibitors, we tested their effect in reactivated iSLK. 219 cells. Drugs were added 1 h prior reactivation, and maintained in the culture medium throughout the experiment. Cells were fixed 48h after reactivation (TPA/Dox). The extent of viral lytic gene expression and of cell cycle progression was monitored by image analysis of indicated viral lytic genes and pH3- S10, respectively. As shown in Fig 7A, only caffeine partially inhibited the viral lytic gene expression. Consistently, none of the drugs was able to overcome the G2 arrest and restore progression to M-phase (Fig 7B). Our results are consistent with other reports in which the ability of these compounds to sensitize cancer cells to radiation-induced cell death failed if the cells had active p53 and p21 [57]. Through silencing the E3 ubiquitin ligase MDM2 in a siRNA screen we here identify the p53/p21 axis as an important positive regulator of viral reactivation, and demonstrate that cellular stress, an inducer of herpesvirus reactivation, favors KSHV lytic replication. Our data further show that lytic replication leads to severe and sustained DNA damage response and lead to a prominent G2 arrest, indicative of a virus-activated cellular checkpoint. Depletions of p21 in iSLK. 219 cells reactivated by TPA/Dox treatment significantly decreased the kinetics of early (RFP, MTA), intermediate (ORF59) and late (K8. 1) lytic gene expression and alleviated the G2 cell cycle arrest. Similar results were obtained in biologically relevant PEL cells, BC-3 and BCBL-1. Interestingly, despite the significant reduction in viral lytic gene expression, we observed only a modest decrease in the amounts of infectious virions produced from BCBL1RTA cells after p21 depletion. The modest decrease, however, could be misleading, as compared to p21-depleted cells, the sh-Ctrl expressing cells lysed significantly faster, which could affect the infectivity of the released viruses. Other DNA viruses, including human papilloma virus, adeno and parvoviruses, have been shown to induce a G2/M arrest during viral DNA amplification [48]. Why would KSHV then benefit from halting the cell cycle at G2? One possibility is that the virus needs to access cellular factors expressed at the G2 phase which could include factors for recombination, repair etc. [58]. Interestingly, we found that reactivated cells were able to bypass the G1/S arrest induced by treatments with Nutlin prior to reactivation. As the G1/S checkpoint is inactivated, the infected cells progress to the S-phase and eventually arrest in G2. While the detailed mechanism of this inactivation is under investigation, we and other groups have demonstrated that the restoration of p53 activity by Nutlin leads to efficient cell death in latent PEL cells and KSHV-infected endothelial cells [44,56,59,60], but fails to kill reactivated cells [61]. The described inactivation of the G1/S checkpoint in cells undergoing lytic replication now explains why the same treatment was not effective in eliminating the reactivated cells [61]. Another advantage of the G2 arrest could be to avoid factors active in M phase that could impair virus amplification. Similar to cellular DNA, the viral DNA genome in infected cells is complexed with nucleosomes [22]. The viral genome may therefore also undergo some degree of modifications (e. g. condensation) when the cell enters mitosis, which could disturb viral lytic gene expression and DNA replication once the lytic cycle is initiated. While more experiments are required to address this possibility, a recent report showed that KSHV genome can be co-immunoprecipitated using antibodies against pH3-S10 [17], a modification present during chromatin condensation at the onset of mitosis [55,62–65]. Interestingly, depleting the enzymes that are responsible for H3 phosphorylation at S10 led to robust spontaneous reactivation and decreased levels of pH3-S10 on the viral genome [17]. Both p21 and the DNA-damage-activated Chk1, which we found active during reactivation, can inhibit the G2/M transition [66–68]. p21 has been shown to efficiently inhibit CDK1 in response to DNA damage, which was sufficient to cause a permanent G2 arrest [69]. The early induction of p21 described here could therefore inhibit Cdk1, support the G2 arrest and increase the efficiency of virus reactivation. Intriguingly and supporting this hypothesis, Cdk1 inhibition by small molecule inhibitors has been shown induce spontaneous KSHV reactivation in PEL cells [70]. Although the activation of p21 seems to be an early event during KSHV reactivation, the activation of a DNA damage response (e. g. pATM, pChk1) observed at later stages of the lytic cycle could reinforce the inhibition of the cell cycle progression [66]. We therefore attempted to inhibit the DDR by small molecule inhibitors, hoping to restore the cell cycle progression and activate cell death responses during M phase, a process known as ‘mitotic catastrophe’ [71,72]. This strategy has been successfully used to sensitize cancer cells to radiation-induced DNA damage [73]. However, the DDR inhibitors have not been effective in cells with functional p53/p21 axis, because cell death is prevented by the p53-dependent cell cycle arrest [74,75]. Our attempt of using inhibitors of DDR to induce cell death in reactivated cells was also unsuccessful. Similar to other cancer cells, this resistance could be due to p21-induced cell cycle arrest. The tumor suppressor p53 is often mutated and inactivated in human cancers. However, this is not the case in KSHV-associated malignancies, where p53 mutations are rarely found [39,59,76]. Instead, the virus has co-evolved with wild type p53 and utilizes a large repertoire of mechanisms to bypass its growth restrictive functions during the latency [38–42,44,45,59]. Our work now provides a mechanistic explanation as to why the virus has evolved to retain p53 and p21 activities to support the lytic phase. Our ChIPSeq cell analyses indicated that p53 does not bind KSHV genome during the reactivation. Work on murine gammaherpesvirus 68 (MHV68) and Epstein Barr virus (EBV), two other gammaherpesviruses, has demonstrated that p53 contributes to the expression of the lytic genes, by stimulating transcription of RTA or RTA and an immediate-early viral Zta gene, respectively [77–79]. Our results now demonstrate a different role for p53 in KSHV reactivation, where this transcription factor is required to enhance viral replication through modulation of the host cell stress responses induced upon viral lytic replication. Lentiviral expression plasmids pLKO. 1-sh-Scr, sh5-MDM2, sh1-p53, sh2-p53, sh1-p21 and sh2-p21 were obtained from Open Biosystems and Biomedicum Functional Genomics Unit (Helsinki, Finland). Lentivirus stocks were produced as described [56]. To establish BC-3 cell lines stably expressing sh-p53, cells were selected with medium containing 3. 5 μg/ml puromycin (Sigma) and to establish TREx BCBL1-Rta (here referred as BCBL1RTA) and BC-3 cell lines stably expressing sh-p21, cells were selected with medium containing 2 μg/ml puromycin (Sigma). The nonspecific siRNAs control (ON-TARGETplus, D-001810-10) and siRNA against p21 (ON-TARGETplus, L-003471-00) were from Dharmacon (SMARTpool siRNA). Cells were reverse-transfected (2000 cells per well) for 48 h in 96-well view plates (Perkinelmer) using RNAiMax (Invitrogen), and reactivated by indicated treatments for 24 h. For more details see supplemental information. The Primary screen was performed with the cell spot microarray technique [33]. 48 h after reverse transfection, rKSHV-SLK cells were reactivated for 30h, fixed and imaged with an Olympus Scan-R high-content microscope. For more details see S1 Methods. For immunofluorescence in 96 well imaging plates (Perkin Elmer) cells were fixed in PBS containing 4% paraformaaldehyde, (20 min, room temperature), permeabilized for 10 min in PBS containing 0. 1% Triton-X 100 (Sigma). All antibody incubations were performed in PBS containing 2% BSA for 45 min. Alexafluor-conjugated secondary antibodies were from Invitrogen. For western blot analysis cells were lysed in RIPA buffer containing protease (Thermo scientific; Cat # 88666) and phosphatase inhibitors (Thermo scientific; Cat # 88667) and whole cell extracts were then clarified by centrifugation, mixed with Laemmli buffer and boiled for 5 min. Protein concentration was determined by Bio-Rad protein assay (Bio-Rad) and 20–40 μg of protein were loaded in each lane of a Criterion TGX midi gel 4–15% (Bio-Rad) and transferred to nitrocellulose membranes (Protran nitrocellulose membrane 0,45 um, Perkin Elmer). Membranes were immunostained in TBST buffer containing 5% fat-free milk, and HRP-conjugated secondary antibodies detected with the Enhanced Chemiluminescence kit (Western Bright Sirius, Advansta) on Fuji films. Densitometry analysis was performed using the ImageJ software (National Institutes of Health, USA). ChIP and sequencing were carried out as described [80] using a monoclonal antibody against p53 (Clone DO-1, GeneSpin) and an Illumina HiSeq 2000 (single 36 bp reads) system. The analysis of the sequencing data was performed as described [81]. For details see S1 Methods. Caffeine (Sigma) was dissolved in water. All the other drugs were dissolved in DMSO and stored at -20 C. Nutlin-3, Etoposide, KU-55933 and UCN-01 (Sigma); MK-8776 (also called SCH900776, Chemie Tech); VE-821 and AZD-7762 (Selleckchem). All experiments were performed in 96-well imaging plates (PerkinElmer). For details see S1 Methods. All experiments were performed in 96-well imaging plates (Perkin Elmer) and images acquired using the automated fluorescence microscope Cellinsight (Thermo Scientific). Image analysis was performed with the Cell Profiler 2 software [82]. For the analysis of the pH3 S10 staining we modified a premade imaging pipeline (" Percent positive" ) available in the Cell Profiler web site (www. cellprofiler. org). After detection of nuclei, the G1/S, G2 and M-phase cells were classified by thresholding the mean fluorescence intensity of pH3 S10 signal in each nucleus. For FACS analysis cells were fixed/permeabilixed in 70% Ethanol at -20°C for 3 h, washed once with PBS and incubated for 45 min in PBS containing 30 μg/ml RNase and 10 μg/ml propidium iodide. Cells were washed two times with PBS before FACS analysis. For each condition, 10. 000 cells were analyzed. MDM2 (Gene ID: 4193); TP53 (Gene ID: 7157); p21 (CDKN1A; Gene ID: 1026); P53R2 (Gene ID: 50484); PAG608 (Gene ID: 64393); Cyclin B1 (CCNB1; Gene ID: 891); H2AX (Gene ID: 3014); ATM (Gene ID: 472); ATR (Gene ID: 545); CHK1 (Gene ID: 1111); CHK2 (Gene ID: 11200) KSHV genes: ORF73/LANA (Gene ID: 4961527); ORF50/RTA (Gene ID: 4961526); ORF57/MTA (Gene ID: 4961525); ORF74/vGPCR/K14 (Gene ID: 4961465); ORF25 (Gene ID: 4961452); ORF29 (Gene ID: 4961443); K8. 1 (Gene ID: 4961469)

Herpesviruses are known to wake up and reactivate in response to different kinds of stress. Our study now highlights the key molecular host cell events that KSHV has evolved to utilize for efficient viral lytic replication: the activation of p53 and upregulation of p21, which slows down the cell cycle, but promotes viral replication and transcription of viral lytic genes. Mutations in TP53 gene are rarely found in KSHV-associated malignancies. Therefore, our work now provides a mechanistic explanation as to why the virus has evolved to retain p53.

Abstract Introduction Results Discussion Materials and Methods

fluorescence imaging gene regulation cell cycle and cell division cell processes microbiology dna damage dna immunologic techniques research and analysis methods small interfering rnas imaging techniques gene expression immunoassays viral replication immunofluorescence biochemistry rna cell biology nucleic acids virology genetics biology and life sciences non-coding rna image analysis

2016

Oncogenic Herpesvirus Utilizes Stress-Induced Cell Cycle Checkpoints for Efficient Lytic Replication

Protein inference, the identification of the protein set that is the origin of a given peptide profile, is a fundamental challenge in proteomics. We present DeepPep, a deep-convolutional neural network framework that predicts the protein set from a proteomics mixture, given the sequence universe of possible proteins and a target peptide profile. In its core, DeepPep quantifies the change in probabilistic score of peptide-spectrum matches in the presence or absence of a specific protein, hence selecting as candidate proteins with the largest impact to the peptide profile. Application of the method across datasets argues for its competitive predictive ability (AUC of 0. 80±0. 18, AUPR of 0. 84±0. 28) in inferring proteins without need of peptide detectability on which the most competitive methods rely. We find that the convolutional neural network architecture outperforms the traditional artificial neural network architectures without convolution layers in protein inference. We expect that similar deep learning architectures that allow learning nonlinear patterns can be further extended to problems in metagenome profiling and cell type inference. The source code of DeepPep and the benchmark datasets used in this study are available at https: //deeppep. github. io/DeepPep/. The accurate identification of proteins in a proteomics sample is a key challenge in life sciences. Proteins, the final gene product and the fundamental blocks of all cellular processes, are elusive to detect. The standard technology that provides fast, high-throughput characterization of complex protein mixtures is mass spectrometry-based shotgun proteomics. Initially, proteins are fragmented in small amino acid chains that are called peptides that then pass through a mass spectrometer. This results in a specific mass spectrum signature for each peptide, which correlates current intensity with a peptide’s weight and charge. Next, this signature is matched to a peptide database to identify which peptides are present in the sample (peptide identification step). Finally, the peptide profile is used to predict which proteins were more likely to produce the observed peptide set (protein inference step) [1]. More precisely, the challenge in protein inference is to infer the proteins (output) that give rise to the peptides observed in the sample. Each peptide has been identified after a database search of the sample mass spectrum, with a certain confidence level, also known as the “peptide probability” [2]. The protein inference problem has been particularly challenging due to existence of degenerate peptides and ‘one-hit wonders’ [1]. A degenerate peptide is one that can be generated by multiple proteins. A ‘one-hit wonder’ is a protein that has only one peptide match. The tools that have been developed for tackling these challenges use a wide arsenal of algorithmic methods, including optimization and parsimonious techniques [3], non-parametric [4] and parametric models [5] (in particular, ensemble of machine-learning methods [6]), among others [7–9] (see [1] for an extensive review). Most recent advances [8,9] rely on the quantification of peptide detectability, a measure of the detection probability for a given peptide from a standard sample mixture by a standard proteomics routine given its parent protein. That is, the peptide detectability can be considered as an inherent characteristic of a peptide that is primarily determined by its sequence and its parent protein sequence [1]. Such information is typically predicted from peptide features such as amino acid composition, N- and C- terminal residues. An intrinsic problem in estimating peptide detectabilities is the existence of different biological conditions, the unequal concentrations of proteins before preparation of the sample and the existence of errors, all of which complicate protein inference and introduce noise [7]. To address these challenging aspects of protein inference, we have developed a deep learning technique, DeepPep, that uses the sequence information of proteins and peptides as inputs. Artificial neural networks have been applied in the past in proteomics, with applications such as the prediction of the retention time in LC-MS data [10,11], prediction of peptide detectability [7,12,13], prediction of peak intensity in a MS/MS spectrum [14] and more recently, prediction of protein secondary structure [15]. Notably, [13] used ensemble of 30 neural networks with one hidden layer where the number of hidden nodes ranges from 1 to 4 for prediction of peptide detectability and protein quantity. In a recent work, a neural network with one hidden layer is trained to differentiate target proteins from decoy proteins in target decoy dataset [16]. DeepPep differs from previous work as it uses convolution layers for capturing the sequence information of proteins and peptides, hence allowing for more complex nonlinear relationships. This introduces new computational challenges that we solve using advanced techniques ranging from regularization, efficient optimization method Section 2. 5, and sparsity-aware, redundancy-aware computations Section 2. 7. In the following sections we present the DeepPep’s architecture, the inference algorithm and its application in seven benchmark datasets. DeepPep is a deep learning-based framework that predicts the list of scored proteins (output) that are more likely to generate a given peptide profile (input). Each candidate protein is scored based on its effect on peptide probability predictions of the deep learning model when it is present/absent Section 2. 6. The framework is composed of four sequential steps as shown in Fig 1. The training data consist of two components: (a) the observed peptide sequences with associated probabilities returned by mass spectra database search [17], and (b) the whole set of proteins (and their sequences) for the respective organism. This data are imported to the convolutional neural network (CNN) for identifying complex patterns between the probability of observed peptides and the positional information of the peptide in the protein sequences. There are two key ideas here that are translated to tasks for the CNN to (indirectly) learn: first, that a protein that matches multiple peptides in the sample is more likely to be part of the protein set that gave rise to that particular sample; second, a protein with non-overlapping peptide hits is more likely to exist in the sample than a protein with the same amount but overlapping peptide hits. “Non-overlapping peptide hits” are the peptide matches to a specific protein that do not coincide, that is, there is no position in the protein sequence where two or more peptides match. The more frequent occurrence of non-overlapping hits stems from the chemical treatment of the proteins; proteins are typically degraded to peptides by trypsin, which deterministically cleaves at the carboxyl side of the amino acids lysine or arginine (except when either is followed by proline [18]) and therefore, this deterministic process in protein degradation gives rise to non-overlapping peptides. In the next step, all peptide probabilities are predicted with and without the presence of each protein. We finally score the proteins based on the difference in peptide probability between being present and absent. This differential ranking is a key concept behind DeepPep. Conceptual basis of this idea is rooted from feature selection in machine learning [19], which is performed to find informative attributes in a prediction task. Protein inference methods using feature selection can be vulnerable to cases where attributes contain similar information. Although this can make such methods give low or zero weights to proteins containing homologous peptides, we observe that such cases are very limited in real datasets. For seven separate proteomic datasets (Section 2. 2), the population of protein pairs having similar peptide matches (PCC > 0. 7) represent on average below 0. 05% among all possible pairwise comparisons of candidate proteins (S1 Table in S1 Text). We used seven separate MS/MS datasets for evaluation purpose as follows: The summary statistics and source information of the seven datasets are shown in S2 Table (in S1 Text). Four datasets of Sigma49, UPS2,18Mix, and Yeast were the ones available with information of the true protein set that gave rise to the respective peptide profiles. For three remaining datasets (DME, HumanMD, HumanEKC) however the true protein set is unknown. To mitigate, we used the target decoy strategy [26] for evaluation in such datasets instead. A target decoy strategy adds a set of incorrect (i. e. decoy) proteins to the search space to be considered as true negatives during evaluation. More specifically we used the default decoy strategy in Trans Proteomic Pipeline (TPP v5. 0. 0) [27], which performs random shuffling of tryptic peptides of a real protein from database to generate a new decoy protein. For each dataset, given all available mass spectrometry files and the related reference protein database, peptide probabilities are calculated using TPP. For database search and for estimation of peptide identification probabilities, X! Tandem and PeptideProphet are used, respectively. The peptide detectabilities (as needed for running MSBayesPro and ProteinLasso) are estimated using a pre-trained model provided by DQmodel [7] team. For optimizing hyper-parameters of Fido, we used the target decoy strategy over each dataset where the performance is measured based on how well the method differentiates target proteins from decoy proteins without looking into the true protein set. This doubles the size of each search dataset by adding decoy protein sequences for the amount of target proteins. For investigation of hyper-parameters for DeepPep, we employ the same approach used for Fido but only on 18Mix dataset. We would like to note that this strategy doesn’t use the information of the true protein set that gives rise to the observed profiles and thus can be used for evaluation purpose with regards to prediction of the true proteins. The training dataset consists of training pairs {xi, yi} for each peptide ppi. The actual number of peptides that are present in any proteomics profile differ and can be from a few hundred to thousands. The input xi is a set of binary vectors that has been constructed by (a) first preparing the amino-acid sequences of all proteins that have at least one match for any of the observed peptides in the profile, (b) then replacing their amino-acid sequences of each protein with ones in any place where there is a full match of peptide ppi and zeros otherwise. In other words, xi is a set of vectors that contains the location where the peptide ppi matches to the organism’s proteome. The input yi is the probabilistic score of a specific peptide-spectrum match for peptide ppi, calculated from PeptideProphet [2]. In other words, it is the identification probability of a peptide in mass spectra returned by a database search. We choose the highest probability in case there exist multiple spectra matches to the same peptide. The construction of input sequence (xi) is described in Algorithm 1 and it has to be applied for each of the observed peptides. Algorithm 1 Build CNN Training Data Comment: p ^ is the set of sequences for the candidate proteins p. p p ^ i is the sequence of peptide ppi. xj[m, n] is the output string from position m to n for protein pj. Input: p ^, p p i ^ Output: x for j = 1 to |p| do Initialize xj to zeros where | x j | = | p ^ j | for m = 1 to | p ^ j | - | p p ^ i | + 1 do if equal (p ^ j [ m, m + | p p ^ i | - 1 ], p p ^ i) then x j [ m, m + | p p ^ i | - 1 ] = 1 end if end for end for The deep convolutional network (CNN) is organized by a series of layers (Fig 2). Unlike typical CNNs, the CNN in the DeepPep framework has distinctive binary vectors as its input layer to represent sequences of proteins encoding input peptides. The organization of layers in the CNN architecture is similar to what has been used before in other fields [28], alternating a convolution layer and a pooling layer four times, followed by a fully connected layer and finally overlaying an output layer that predicts the peptide probability. Each layer computes a linear transformation of the output from the previous layer multiplied by a weight matrix, followed by a nonlinear transformation (ReLU). The only exception is at output layer where it produces linear value from the previous layer without applying the transformation. The CNN uses three different types of layers: the convolution layer, the pooling layer and the fully connected layer. The convolution layer computes output by one-dimensional convolution operation with a specified number of filters (weight matrices) and all convolution operation outputs are then transformed by the rectified linear activation function (ReLU): X i, f p, l = ReLU (∑ j = 1 w ∑ k = 1 n W j, k f, l - 1 X i + j, k p, l - 1) where X is the input, p is the index of protein, l is the index of the convolution layer, i is the index of the output position and f is the index of filters. Each convolution filter Wf, l is an w × n weight matrix of filter f at layer l with w being the window size and n being the number of input channels. A pooling layer computes the function pool (max-pooling) in a window with the specified length (w) for each filters computed from preceding convolutional layers. The step size is equal to the size of the pooling window. This reduces the size of the output to that of the window size and thus allows sequence learning patterns of higher abstraction in the next convolution layer: X i, f p, l = pool ({ X i w, f p, l - 1, X i w + 1, f p, l - 1. . . X i w + w - 1, f p, l - 1 }) where Xp, l is the input of protein p from preceding convolution layer l − 1, i is index of output position, f is index of filter, w is pooling window size. The number of filters for each of the four convolution layers increase at each level (5,10,15 and 20 respectively), so that higher-level convolution layers can produce more complex patterns. We added a fully connected layer over the fourth pooling layer, which is the high-level representation of information computed from the binary proteome sequence that encodes the input peptide profile. This fully connected layer computes the product WlXl, where Xl is the combined input of all proteins from convolution layer l and Wl is the weight matrix for the fully connected layer. There is one weight matrix Wl corresponding to the output node of the output layer that calculates the predicted peptide probability. The optimal configuration with respect to the pooling function pool, number of filters, window sizes in convolution layer and pooling layer and number of nodes in fully connected layer was empirically determined (Section) and the final architecture is depicted in Fig 2. The objective function is the minimization of the sum of squared errors between predicted and measured peptide probability: minimize ∑ i = 1 (y i - y ^ i) 2 where i is the peptide index of all observed peptides and yi indicates the measured probability of peptide i, y ^ i represents predicted probability of peptide i. Derivatives of the objective function with respect to the model parameters are calculated and used in standard backpropagation [29]. For updating weights in CNN, we used a gradient decent optimization algorithm called RMSprop, which was developed to deal with radically diminishing learning rates [30]. This algorithm converged significantly faster than conventional approaches such as Stochastic Gradient Descent (SGD), requiring only 30 epochs with the learning rate of 0. 01 to reach below root mean squared error (RMSE) of 0. 01 in all seven datasets we examined. After computation of each convolutional and fully connected layer, a dropout layer where a fraction (20% at each layer) of the model parameters are set to zero is applied to prevent overfitting [31]. We didn’t impose any regularization constraints as dropout already discarded a sufficient number of the parameters. The key idea behind DeepPep is that if a protein is the origin of the input peptide, the peptide probability will depend on the presence or absence of the protein from the input. To quantify this, we define the normalized change in the probability of peptide ppj due to protein pi as: c i, j = | y j - CNN (x j, p i) | n i j where CNN (xj, pi) represents the predicted probability of peptide ppj in the absence of protein pi. The normalization quantity n i j denotes the number of positions in protein pi that were reset to zero to declare the absence of the protein. In other words, these are the number of amino acids in the protein that have a perfect match with the peptide ppj and it is always a multiple of its length. This normalization is necessary as the probability difference will increase with the number of zeros that we impose. Finally, the score of protein pi is assigned to be the average of ci, j for all j. The scores of all proteins p are the outputs of DeepPep. Given the way DeepPep represents peptide-protein matches, the memory requirement for loading inputs can be massive. For example, the input for the Yeast dataset takes 26GB. Even worse, the memory that a deep network needs is several times more than the actual input size (for storing gradients in each layer). Furthermore, in terms of CPU, computation of convolution, linear transformations and gradients in the network given such data, can pose a limit to scalability of DeepPep. However, given that any protein only matches with a handful of peptides, the input to DeepPep is largely sparse (between 95% to 99% depending on the dataset) and therefore taking advantage of this property can greatly reduce the memory and computational overhead of DeepPep (Section 3. 4). Next we explain how to take advantage of this sparsity in the context of Deep Neural Networks. We used the following four popular methods that are based on optimization, Bayesian, and constrained regression approaches to compare with DeepPep. The running time of protein inference methods including prerequisite steps of each method (e. g. DQModel, TPP pipeline) was all measured and compared on Two Intel E5-2630 v3 2. 4GHz CPUs with eight cores with 64GB of RDIMM RAM. Hyper-parameter optimization of DeepPep was performed on parallel under the CPU environment on the NCSA Blue Waters supercomputer (a petascale machine with 22,500 nodes of AMD 6276 Interlagos 2. 3GHz processors, 64GB memory per node). The computing environment used for comparison of running time of protein inference methods was described in Section 2. 8. The “Data Preparation” step is done in Python while the training and protein inference using CNN architecture are implemented with the torch7 framework. For efficient implementation of convolutional layers, sliding windows between neighboring proteins are omitted to avoid any biases coming from concatenation of strings amongst heterogeneous proteins. The source code of DeepPep and the benchmark datasets used in this study are available at https: //deeppep. github. io/DeepPep/. We first performed empirical hyper-parameter optimization by measuring the effect of different parameters to the prediction performance. The optimal configuration (e. g. max pooling) was investigated with respect to the pooling function pool, number of filters, window sizes in convolution layer and pooling layer and number of nodes in fully connected layer. The performance of each configuration was evaluated using the target decoy approach on 18Mix dataset (Section 2. 2). In this approach, the performance is measured based on how well the method differentiates target proteins from decoy proteins and therefore, this evaluation does not use the information of the true protein set. As shown in S4 Table (in S1 Text), DeepPep remains robust with a high AUC/AUPR value (0. 94±0. 009/0. 93±0. 008) in the spectrum of the experiments we performed (final selection is shown in Fig 2). Fig 3 depicts the performance of the six methods for the seven independent datasets with respect to ROC curve and PR curve. Overall, DeepPep shows competing performance across different datasets, ranking first by a narrow margin in overall AUC and AUPR. It is noticeable that DeepPep outperforms other methods for HumanEKC dataset. Although the AUC/AUPR performance of DeepPep are below those of other methods for DME dataset, we noticed that the performance based on the final list of inferred proteins (i. e. F1-measure) is comparable as shown in Fig 4. Furthermore, the hyper-parameters of DeepPep learned from the decoy-added 18Mix dataset might not be optimal for DME dataset, which could be improved once hyper-parameter optimization is performed separately for each of the seven datasets using the target decoy approach Section 2. 2 as done for Fido Section 2. 8. 1. DeepPep ranks first by a small margin or ties with others in first place for four (18Mix, Sigma49, Yeast and HumanEKC) out of seven datasets. Next, we assess the degree that the convolution layers impact the performance of DeepPep. For this purpose, we compare the performance of DeepPep against traditional Artificial Neural Networks without convolution layers but otherwise similar settings (ANN-Pep). ANN-Pep uses fully connected layers to connect inputs to outputs. The results (Fig 3 and S5 Table in S1 Text) show that overall, DeepPep (AUC/AUPR: 0. 80/0. 84) outperforms ANNs with 18 different architectural configurations for seven datasets (max AUC/AUPR: 0. 74/0. 77), which suggests spatial dependencies within input sequences are crucial to maximize the capacity of protein inference. The performance of all 18 configuration are tabulated in S5 Table (in S1 Text). We also evaluated DeepPep with other methods based on the list of predicted proteins (Fig 4). DeepPep shows comparable performance with the other four methods across different datasets with regards to positive prediction and negative prediction of inferred proteins. MSBayesPro shows top performance in positive prediction of inferred proteins for HumanMD dataset, whereas its performance is degraded on other datasets (e. g. Sigma49 and HumanEKC). Second, DeepPep is particularly sensitive in prediction of degenerate proteins, which have peptides with multiple protein matches. It is particularly notable that the performance of other methods for degenerate proteins fluctuates across the datasets of Sigma49,18Mix, UPS2, and Yeast whereas DeepPep shows consistently competitive performance overall. Dealing with such proteins has been considered more challenging than others because there are multiple options to select protein origins of a peptide. Overall, DeepPep outperforms other methods in terms of precision for degenerate proteins. We next investigated visually underlying processes in DeepPep. As shown in Fig 5A, the mean change in peptide probability with and without a protein (left bar) highly correlates with a list of gold standard proteins (right bar). As expected, the proteins with more peptide matches undergo more changes in peptide probability in general (heat map). This is because our CNN learns the source of observed peptide probability with respect to proteome sequences and a protein with multiple peptides matches will have higher weights in the CNN than a protein with few matches. This will result in increased change in the CNN’s output when a protein that is the origin of the peptide is nullified. In the underlying processes of computing output from input in the CNN, the deeper convolution layers increase the difference between positive and negative samples (Fig 5B). One explanation is that the changes are transmitted to adjacent neurons as a pair of pooling/convolution undergoes and proteins with more peptide matches (which is likely be in the protein set) will overall impact more in neighboring neurons than proteins with few peptide matches. This interpretation is more visible in Fig 5C, which shows that in the first convolution layer, the changes are visible in regions matching with input peptides only when it is propagated to the whole segment of the protein as it gets into a deeper layer. We examined computational efficiency of DeepPep in comparison to other methods over seven datasets. The results (Table 1) show that the efficient implementation of DeepPep (Section 2. 7) enables it to run between 2. 5 minutes and 90 minutes depending on the size of dataset (S2 Table in S1 Text). We observed that MSBayesPro shows a significant delay when the dataset size becomes larger (i. e. Yeast, DME, HumanMD, and HumanEKC). Please note that the computational efficiency of Fido can be enhanced further with its advanced version (FidoCT, [33]) although its impact should be minimal in the rank as it already shows top performance overall. Apparently, although DeepPep is not the most efficient method among five methods in terms of running time of the protein inference method, we would like to point out that many other methods need prerequisite steps before execution ranging from estimation of peptide detectability (ProteinLasso and MSBayesPro) to hyper-parameter optimization using target decoy strategy (Fido), which all affect the overall running time (S6 Table in S1 Text). Specifically, Fido’s prerequisite step to optimize hyper-parameters based on a grid search over a decoy-added dataset necessitates to almost double the running time of TPP by adding decoy proteins on the search database, which consumes from 15 min to 29 hours more depending on the size of dataset. Considering all these hidden steps required before running actual methods, DeepPep ranks in second or third place among five methods in overall running time comparisons. We described DeepPep, a convolutional neural network method for deep protein inference. Our results provide evidence that using sequence-level location information of a peptide in the context of proteome sequence can result in more accurate and robust protein inference. DeepPep demonstrated a competitive predictive ability (AUC of 0. 80±0. 18, AUPR of 0. 84±0. 28) in inferring proteins without need of peptide detectability on which recent methods mostly rely. This has significant implications in proteomics pipelines, where peptide detectability quantification is a major step. We also demonstrated the predictive value of the convolutional layers, by comparing DeepPep to various other ANNs, highlighting the importance of spatial dependencies in peptide/protein sequences. In performance comparison, while DeepPep was trained on the same datasets that required protein inference, the detectability-based methods (ProteinLasso and MSBayesPro) were executed with peptide detectabilities predicted using models trained on totally different datasets. As shown in [7], detectability predictions impact the protein inference performance, therefore, training detectability prediction models on the datasets to infer the protein set might alter the reported performance of ProteinLasso and MSBayesPro at the expense of having a longer overall computation time. The architecture of DeepPep can be extended to predict quantities of proteins beyond identification of proteins. For example, one can use the concentration or count of each peptide as an informative feature for predicting the concentration of each protein in the original sample. Furthermore, although we have addressed scalability issues in DeepPep by employing the sparsity of proteome datasets, other advances to tackle computational complexity in deep learning, for example, distributed training [35] and optimization of memory use [36], can be integrated, which will make the tool more available, anticipating the method can be deployed for most practical applications, similar to the way it was demonstrated in this work. In addition, the proposed method relies on the preceding method (PeptideProphet [2]) that identifies peptides from a given set of mass-spectra. This method has shown high precision, achieving AUC from 0. 96 to 0. 97 across different datasets [37]. To minimize any noise produced in identifying peptides, the proposed framework can be extended to directly take mass spectra as input (e. g. input encodes short peptide corresponding to each mass spectra and its intensity becomes desired output of CNN.). We would like to emphasize that application of a similar architecture to the one present in DeepPep can be introduced to solve biological problems beyond protein inference. For example, it can be applied in metagenome sequencing where genetic profiles are derived directly from environmental samples and there the task will be to identify the microbial consortia [38]. Similarly, cell type inference from short RNA reads of fragmented heterogeneous cells [39] can benefit from such architecture. Doing so will add one more area where the power of deep learning can be harvested to increase prediction performance in computational biology [28,40,41].

The accurate identification of proteins in a proteomics sample, called the protein inference problem, is a fundamental challenge in biomedical sciences. Current approaches are based on applications of traditional neural networks, linear optimization and Bayesian techniques. We here present DeepPep, a deep-convolutional neural network framework that predicts the protein set from a standard proteomics mixture, given all protein sequences and a peptide profile. Comparison to leading methods shows that DeepPep has most robust performance with various instruments and datasets. Our results provide evidence that using sequence-level location information of a peptide in the context of proteome sequence can result in more accurate and robust protein inference. We conclude that Deep Learning on protein sequence leads to superior platforms for protein inference that can be further refined with additional features and extended for far reaching applications.

Abstract Introduction Methods Results Discussion

neural networks database searching neuroscience artificial neural networks optimization mathematics artificial intelligence computational neuroscience convolution research and analysis methods sequence analysis computer and information sciences mathematical functions bioinformatics proteins mathematical and statistical techniques biological databases proteomics biochemistry proteomic databases sequence databases proteomes database and informatics methods biology and life sciences physical sciences computational biology

2017

DeepPep: Deep proteome inference from peptide profiles

Area-wide integrated pest management strategies that include a sterile insect technique component have been successfully used to eradicate tsetse fly populations in the past. To ensure the success of the sterile insect technique, the released males must be adequately sterile and be able to compete with their native counterparts in the wild. In the present study the radiation sensitivity of colonised Glossina brevipalpis Newstead (Diptera; Glossinidae) males, treated either as adults or pupae, was assessed. The mating performance of the irradiated G. brevipalpis males was assessed in walk-in field cages. Glossina brevipalpis adults and pupae were highly sensitive to irradiation, and a dose of 40 Gy and 80 Gy induced 93% and 99% sterility respectively in untreated females that mated with males irradiated as adults. When 37 to 41 day old pupae were exposed to a dose of 40 Gy, more than 97% sterility was induced in untreated females that mated with males derived from irradiated pupae. Males treated as adults with a dose up to 80 Gy were able to compete successfully with untreated fertile males for untreated females in walk-in field cages. The data emanating from this field cage study indicates that, sterile male flies derived from the colony of G. brevipalpis maintained at the Agricultural Research Council-Onderstepoort Veterinary Institute in South Africa are potential good candidates for a campaign that includes a sterile insect technique component. This would need to be confirmed by open field studies. Tsetse flies (Diptera; Glossinidae), the cyclical vectors of trypanosome parasites that cause human African trypanosomosis (HAT) and African animal trypanosomosis (AAT), infest 10 million km² of sub-Saharan Africa [1–3]. In South Africa, Glossina brevipalpis Newstead and Glossina austeni Newstead are the vectors of Trypanosoma congolense and Trypanosoma vivax [4–7] that cause AAT in an area of about 16 000 km2 in the north eastern parts of the KwaZulu-Natal Province [8,9]. Because of the limited options available for controlling the parasite, i. e. no efficient vaccines and increased resistance to the commonly used trypanocidal drugs [1,10], vector control remains an effective and economical option for managing AAT in KwaZulu-Natal. The use of sterile males to control field populations of insect pests, as conceptualised by E. F. Knipling in the 1940s [11–13], has been applied successfully to control tsetse fly populations. In the 1970 and 1980’s, the sterile insect technique (SIT) was used in feasibility studies against populations of Glossina morsitans morsitans Westwood in Zimbabwe and Tanzania, Glossina tachinoides Westwood in Chad and Glossina palpalis gambiensis Vanderplank in Burkina Faso [13–19]. The first eradication campaign that integrated the use of radiation-sterilized adult male tsetse with suppression methods such as insecticide-impregnated targets, was implemented against Glossina morsitans submorsitans Newstead, G. p. gambiensis and G. tachinoides in Burkina Faso in the 1980’s [20]. A similar approach was used around the same period to target a population of Glossina palpalis palpalis (Robineau-Desvoidy) in an agro-pastoral area of Lafia in Nigeria [21]. All targeted populations of the four species were eradicated from the controlled zones. These programmes, although successful, were however not conducted area-wide, and their pest free status was lost due to reinvasion from neighbouring areas [13]. The most successful AW-IPM programme with a SIT component against a tsetse fly population to date was implemented on Unguja Island of Zanzibar, Tanzania in the 1990’s [22]. Suppression of G. austeni populations with insecticide-treated screens and cattle started in 1988 and from August 1994 up to December 1997,8. 5 million sterile flies were released on the island [22]. The last wild G. austeni was trapped in September 1996 and to date, Unguja Island is still free of tsetse flies and trypanosomosis. Initial results of an AW-IPM campaign with a SIT component currently underway against a G. p. gambiensis population in the Niayes of Senegal, seems likewise promising [23–25]. The SIT entails the use of radiation to sterilise males of the target species reared in a mass-rearing facility [11,26], followed by the sequential and area-wide release of these males in sufficient numbers to outcompete their wild counterparts [13,27]. The mating of sterile males with wild fertile females results in no progeny, which leads to reduction in the size of the targeted population and, in some cases, to local eradication [27]. Because the released sterile males must be able to outcompete the local wild males in mating with wild females, the biological quality and sexual competitiveness of the sterile males are of utmost importance [28,29]. Releasing low quality sterile males will necessitate higher release rates and might prolong the duration of the programme, which will require more funding and the potential for failure will be higher [28]. Ionizing radiation can potentially influence the mating competiveness of the released males. The effect of gamma radiation on the reproduction and competitiveness of several tsetse species was investigated in the past [27–34]. Radiation doses ranging from 50 Gy for G. brevipalpis up to 170 Gy for G. tachinoides induced acceptable or complete sterility in these two species [30,31]. Doses of 110 Gy to 120 Gy induced sterility in G. austeni [32], G. tachinoides [33], G. p. palpalis [34], Glossina pallidipes Austen [35] and G. p. gambiensis [36]. It was showed that a dose of 120 Gy did not affect mating competitiveness of G. pallidipes males in walk-in field cages and keeping the irradiated pupae for 24–72 h at a low temperature of 15°C even increased the competitiveness of the emerged males [37]. The successful and sustained removal of G. austeni from the Island of Unguja [22] demonstrated the feasibility of using the SIT to eradicate a population of this species. However, the SIT has never been used or evaluated against the second tsetse fly species present in South Africa i. e. G. brevipalpis, and no data is available on the mating performance of irradiated flies. The distribution of G. brevipalpis extends from Ethiopia in the north to KwaZulu-Natal, South Africa in the south and is present in Somalia, Uganda, Kenya, Rwanda, Burundi, Tanzania, Malawi, Zambia, Zimbabwe and Mozambique [38]. Although this species belongs to the fusca group of tsetse flies, considered of less or no epidemiological importance for livestock, it was recently shown that the role of this species in transmitting AAT in South Africa should not be underestimated [9]. It has become evident that removing only G. austeni from KwaZulu-Natal would not solve the AAT problem and any potential AW-IPM campaign with a SIT component should include G. brevipalpis [9]. With the exception of a study in the 1990s, that showed that a radiation dose of 50 Gy administered to adult G. brevipalpis males induced 95% sterility [31] very little is known on the biology of irradiated G. brevipalpis flies. We therefore assessed the radiation sensitivity of adult G. brevipalpis males when treated as pupae and adults, as well as their mating performance in walk-in field cages. Materials used in the study posed no health risk to researchers and no vertebrate animals were involved. Permission to do research in terms of Section 20 of the animal diseases act South Africa (ACT no. 35 of 1984) has been granted for tsetse fly colony maintenance, Ref 12/11/1/1. Glossina brevipalpis adults and pupae were obtained from a laboratory colony housed at the Agricultural Research Council-Onderstepoort Veterinary Institute (ARC-OVI), Pretoria, South Africa. This colony was established in 2002 using seed material from the FAO/IAEA Insect Pest Control Laboratory in Austria. The colony flies were maintained on abattoir collected defibrinated bovine blood using an artificial in vitro membrane feeding system under standardised environmental conditions (23–24°C, 75–80% RH and subdued/indirect illumination, 12h light/12h dark) [39,40]. To determine the radiation sensitivity of G. brevipalpis, adult males and pupae (both males and females) were irradiated with either 40 Gy, 80 Gy, 100 Gy, 120 Gy or 140 Gy using a 137Cs source (Gammacell 40 S/N50) at a dose rate of 0. 69 Gy/min in air. Adult males were irradiated four days after emergence. To synchronise adult emergence from irradiated pupae, pupae were collected from the colonies in 24-hour increments. Pupae were irradiated on day 41 (group 1), 39 (group 2) or 37 (group 3) post-larviposition (PL), i. e. three-, five- and seven days before expected emergence. Only the males from the irradiated pupae were used. To determine the effect of radiation on reproduction, 15 six-day-old treated (un-irradiated for the control or irradiated as adults or pupae) males were mated with 30 three-day-old un-irradiated virgin females at a 1: 2 male: female ratio. Males and females were allowed to mate for four days in standard colony holding cages (Ø 20 cm). Thereafter they were kept separately under standard colony conditions [39,40]. Each treatment dose was replicated three to four times. Male and female survival was monitored daily and pupae production recorded. Fecundity was defined as the number of pupae produced per mature female day [31,32]. Mature female days were calculated for each treatment by adding the number of flies alive each day, from day 18 (first larviposition day) after emergence until the end of the experiment on day 60 [32]. All pupae produced were mechanically sorted into one of five class sizes (A (3. 5 mm) to E (4. 3 mm) ) according to the standards used by the FAO/IAEA Insect Pest Control Laboratory. The pupal size classes had the following associated weight (mg): A (<56), B (56– <68), C (68– <76), D (76– <84), and E (≥84) and adult emergence was recorded [39,40]. Aborted eggs and immature larval stages were monitored daily. After 60 days, all surviving females were dissected to determine their reproductive status, insemination rate and spermathecal fill [39,40]. The spermathecae were removed and their fill microscopically scored as empty (0), quarter full (0. 25), half (0. 5), three quarters full (0. 75) or full (1) [41]. Male mortality was monitored until all the males had died. Comparative assessment of the mating performance of gamma sterilised G. brevipalpis males was conducted during February 2013 in walk-in field cages [37,42] deployed in a small forest under near-natural conditions at the ARC-OVI [43]. The cylindrical walk-in field cages (Ø 2. 9 m x 2. 0 m) were made of cream polyester netting with a flat floor and ceiling. Black nylon bands, connecting the ceiling and floor with the sides, encircled the top and bottom of the cage. A 1. 5 m potted Schotia brachypetala was placed in the middle of the cage. Throughout the experiment, temperature and relative humidity were recorded every 10 minutes using a DS1923-F5# Hygrochron iButton data logger (Fairbridge technologies, South Africa). Light intensity at the top and bottom of the cage and at tree level was recorded every 15 minutes using a Major Tech MT940 light meter. It was previously determined that the optimal mating age for G. brevipalpis males was nine days [43]. Therefore, 30 nine-day-old males irradiated four days after emergence with either 40 Gy or 80 Gy and a group of 30 nine-day old un-irradiated males competed for 30 three-day-old virgin females, giving a male to female ratio of 2: 1 in each field cage at the commencement of the experiment. The male groups were distinguished by a dot of different coloured polymer paint placed on the notum 24 hours before being released in the field cage [37]. Experiments were conducted from 12: 00 h to 15: 00 h. Females were released 5 minutes before the males in the middle of the cage, and the time of the first mating recorded to determine mating latency The observer remained inside the cage for the 3-hour duration of the experiment and movements were kept to a minimum. The mating pairs were collected individually into small vials, and duration of the mating recorded. Although no direct adverse effect on mating behaviour was observed when the pairs were collected, its potential influence on mating behaviour cannot be ruled out. To minimise this effect mating pairs were collected in the same way in all experiments. They were not replaced. Females that did not mate in the field cage by the end on the 3-hour duration were collected, immobilised at -5°C and dissected to confirm virginity. The mated females were transferred from the small vials to individual holding cages and kept under standard colony conditions. The females were kept for 60 days and monitored daily for survival and production similar to that of the radiation sensitivity experimental flies, except that the pupae produced were not classed by size. Females were dissected after 60 days to determine reproductive status. Mating performance was assessed by the following mating indices: the propensity of mating (PM), relative mating index (RMI) and relative mating performance (RMP). Propensity of mating (PM) was defined as the overall proportion of released females that had mated. Relative mating index (RMI) was defined as the number of pairs of one treatment group as a proportion of the total number of matings [37]. Relative mating performance (RMP) was defined as the difference between the number of matings of two treatments of males as a proportion of the total number of matings [37]. In addition, the mating latency time, mating duration, insemination rate and the spermathecal fill of each mated female were determined. Data were analysed using the statistical software GraphPad Instat [44] and R [45]. Proportional differences in adult emergence rates were determined with Chi-square (χ2) analysis with the Yate’s continuity correction. The P value was two-sided. Linear regression analysis was used to assess effect of radiation dose on fecundity. The effect of time of irradiation and dose on male survival was analysed with multivariate linear regression model: Estimated survival=β0+β1⋅age (days) +β2⋅dose (Gy) A one-way analysis of variance (ANOVA) was used to differentiate between the relative mating index, average mating latency, mating duration and spermathecal fill. If the data were normally distributed standard (parametric) methods were used and the Tukey’s test was applied otherwise a nonparametric Kruskal-Wallis test was used. Tests were done at the 5% significance level. The SIT component of AW-IPM programmes can only be successful if the released sterile males are able to compete with their native counterparts. The selection of the optimal radiation dose to sterilise the insects is important: a dose below the optimal will result in insects that are not adequately sterile and a too high dose can negatively impact on the quality of these insects and may result in insects that are not competitive with wild flies [46–48]. Factors such as the developmental stage, age of the insect and the atmosphere used during irradiation can influence the level of sterility achieved with a specific dose [46,47]. As was previously shown, the proportions of dominant lethal mutations induced in the sperm of G. brevipalpis increased with increasing radiation dose [31,32], and as a result the rate of induced sterility in mated untreated females likewise increased proportionally with increasing radiation dose. In the present study, mating of males treated as adults with 40 and 80 Gy induced 93% and 99% sterility, respectively in untreated G. brevipalpis females that mated with the treated males. The induced sterility increased to 97% or more when the males were irradiated with 40 Gy as pupae. These results are in accordance with those of Vreysen et al. [31], who reported that a dose of 50 Gy administered to 4- to 6-day-old males in air induced about 95% sterility in untreated females. The dose required to induce more than 93% dominant lethal mutations in the sperm of G. brevipalpis is much lower as compared with species such as Glossina fuscipes fuscipes Newstead, G. tachinoides [31], G. pallidipes [35], G. morsitans [49] and G. austeni [50]. Unlike the other species tested, G. brevipalpis belongs to the fusca group, which have a different chromosome structure than the members of the morsitans or palpalis groups. Although it was suggested that this higher radiation susceptibility might be chromosome related, the main reasons for this still needs to be elucidated [31]. The longer emergence periods and high variation in emergence rate of irradiated pupae obtained in the present study may be an artefact of the basic rearing conditions in the colony. Due to logistical constrains the two species are kept under the same rearing conditions at the ARC-OVI, which may not be the most optimal. Furthermore, due to the variation in emergence rates obtained for the different irradiation doses and time of irradiation, no clear effect of these variables could be detected. The relatively higher radiation sensitivity of G. brevipalpis pupae compared to that of adults obtained in the present study are supported by previous studies, i. e. the production rate relative to the untreated control of G. p. palpalis irradiated with 80 Gy was 7. 1% and 4. 08% when treated as adults and pupae, respectively [34]. Dissection results revealed a clear abortion pattern for female flies that mated with treated males irrespective of life stage, with the rate of abortions increasing with an increase in radiation dose. The uterus of females that mated with irradiated males either contained an egg or was empty due to abortion of the egg in embryonic arrest or an immature larva. It was suggested that these reproductive abnormalities could be used to monitor the impact of sterile male releases on a natural tsetse population [51]. The imbalance between uterus content and the follicle next in ovulation sequence [31] was indeed used to monitor induced sterility in the wild female G. austeni population in the eradication campaign on Unguja Island, Zanzibar [22]. To assess the rate of induced sterility in the native females of the target population (as a result of a mating with a released sterile male), the natural abortion rate of a population needs to be determined and deducted from the abortion rates observed during a sterile male release programme. Our dissection results clearly indicate that this method of assessing reproductive abnormalities in females in a population under sterile male releases can be used to monitor induced sterility in a targeted G. brevipalpis population, and hence, as an indicator of programme progress [28]. In addition, the applied radiation doses did not affect G. brevipalpis males’ insemination ability whether irradiated as adults or as pupae. The reduction in average longevity of irradiated males is a manifestation of the somatic damage caused by irradiation [22]. Irradiating younger pupae of Anopheles (Diptera; Culicidae) mosquitoes and the date moth, Ectomyelois ceratoniae (Lepidoptera; Pyralidae) reduced adult survival [52,53] and pupae emergence [54–56] more as compared with irradiating older pupae. This phenomenon was also clearly observed in our study as irradiation as pupae reduced adult longevity more than a treatment as adults. Data on survival of irradiated insects in the laboratory provides relative information regarding the impact of different doses on male survival as compared with the survival of untreated males. Male survival in the laboratory is most likely not a true reflection of their survival in the field, as the latter will be influenced by many factors other than radiation, and care needs to be taken when extrapolating laboratory survival data to a field situation in an operational sterile male release programme. Although it is important that released males live as long as possible in the wild, the period that they remain sexually active and be able to transfer sperm and outcompete their native counterparts, is the more important aspect. Lifespan and capability of sperm transfer may not be directly related and this requires further research. Exposure to ionising radiation may affect the biological quality of the released insects [57], and the effect is often dose dependent. With some mosquito species the sterilising dose may be very low (e. g. 35 Gy induced > 95% sterility in Aedes albopictus) with minimal impact on male competitiveness (male Ae. albopictus treated with 35 Gy showed a sexual competitiveness of 0. 53 in walk-in field cages five days after emergence) [58]. Lepidoptera on the other hand, have a special chromosome structure with diffuse centromeres, and a much higher dose (250 Gy or more) is usually required to sterilise these males [45]. Our study showed that irradiation doses of up to 80 Gy did not affect the ability of sterilised colony G. brevipalpis males to compete with fertile colony males for untreated colony females in walk-in field cages. This is in accordance with field cage evaluations that showed that the competitiveness of G. pallidipes males irradiated with a dose of 120 Gy did not differ from that of untreated ones [37]. In our study, untreated fertile G. brevipalpis males did form mating pairs sooner and mated for longer than the irradiated males. Although not statistically significant in the present study, any delay in initial mating by irradiated males may potentially reduce their competitiveness in the field. Females that mated with untreated fertile G. brevipalpis males did have a larger spermathecal fill than those that mated with irradiated males. The results of the present mating performance studies in walk-in field cages indicated that the colonised G. brevipalpis seemed to be well suited for use in programmes that have a SIT component. The absence of any noteworthy differences in the quality of flies when irradiated as adults or pupae indicated that G. brevipalpis can be irradiated either as adults or as late stage pupae, i. e. ≥ 37 days PL. The mating performance of G. brevipalpis males irradiated as pupae still needs to be compared with that of fertile males. There are positive and negative aspects to take into consideration in deciding to irradiate adult flies or pupae. Treating pupae has the operational advantage that larger numbers can be irradiated at a time, handling and transport of the immobile pupa are less cumbersome and pupae are less fragile than adult flies. However, the current relative crude sex separation protocol, based on development temperature [59,60] results a certain proportion of females being irradiated and these can therefore not be used for further production in the colony. This is counterproductive because of the limited offspring produced by tsetse flies and all available females are needed. Our results showed that G. brevipalpis males were highly sensitive to irradiation and could be sterilised at a much lower dose than some other tsetse fly species. As a result, sterilised colony males proved to be very competitive with un-irradiated colony males for untreated colony females in walk-in field ages. However, the mating performance of irradiated colonised G. brevipalpis males when competing with fertile wild males under open field conditions still needs to be assessed.

The radiation sensitivity and mating performance of colonised G. brevipalis adult and pupae were evaluated. It was showed that G. brevipalpis males were highly sensitive to irradiation and could be sterilised at a much lower dose, between 40 Gy and 80 Gy, than some other tsetse fly species. As a result, males sterilised with a dose of up to 80 Gy performed similar as that of un-irradiated males for untreated females in walk-in field ages. The data from the study indicate that, under the experimental field cage conditions, the G. brevipalpis males derived from the colony at ARC-OVI should be well suited for use in AW-IPM programmes that have a SIT component.

Abstract Introduction Materials and methods Discussion

invertebrates medicine and health sciences insemination animals glossina insect pests developmental biology pupae pest control tsetse fly population biology insect vectors pests fecundity infectious diseases fertilization life cycles disease vectors insects agriculture arthropoda population metrics biology and life sciences species interactions organisms

2017

Evaluation of radiation sensitivity and mating performance of Glossina brevipalpis males

The parasitic flagellate Trypanosoma vivax is a cause of animal trypanosomiasis across Africa and South America. The parasite has a digenetic life cycle, passing between mammalian hosts and insect vectors, and a series of developmental forms adapted to each life cycle stage. Each point in the life cycle presents radically different challenges to parasite metabolism and physiology and distinct host interactions requiring remodeling of the parasite cell surface. Transcriptomic and proteomic studies of the related parasites T. brucei and T. congolense have shown how gene expression is regulated during their development. New methods for in vitro culture of the T. vivax insect stages have allowed us to describe global gene expression throughout the complete T. vivax life cycle for the first time. We combined transcriptomic and proteomic analysis of each life stage using RNA-seq and mass spectrometry respectively, to identify genes with patterns of preferential transcription or expression. While T. vivax conforms to a pattern of highly conserved gene expression found in other African trypanosomes, (e. g. developmental regulation of energy metabolism, restricted expression of a dominant variant antigen, and expression of ‘Fam50’ proteins in the insect mouthparts), we identified significant differences in gene expression affecting metabolism in the fly and a suite of T. vivax-specific genes with predicted cell-surface expression that are preferentially expressed in the mammal (‘Fam29,30,42’) or the vector (‘Fam34,35,43’). T. vivax differs significantly from other African trypanosomes in the developmentally-regulated proteins likely to be expressed on its cell surface and thus, in the structure of the host-parasite interface. These unique features may yet explain the species differences in life cycle and could, in the form of bloodstream-stage proteins that do not undergo antigenic variation, provide targets for therapy. African trypanosomes are unicellular vector-borne hemoparasites of humans, domestic livestock and wild animals. They cause African trypanosomiasis, an endemic disease of sub-Saharan Africa otherwise known as sleeping sickness in humans and nagana in animals, and are transmitted between vertebrate hosts by the bite of tsetse flies (Glossina spp.). This endemic disease causes considerable morbidity in livestock herds and associated losses in animal productivity. The threat of Animal African trypanosomiasis in tsetse-infested areas also prevents effective exploitation of available pasture, thereby impeding economic development in the world’s poorest nations. There are several species of African trypanosome that vary in life cycle, host range and pathology. Trypanosoma brucei is predominantly an animal pathogen that has evolved the ability to infect humans on multiple occasions [1], while T. congolense and T. vivax are exclusively animal pathogens. During their life cycles, T. brucei and T. congolense exist as procyclic forms in the mid-gut of the tsetse fly before migrating into the salivary glands and proventriculus respectively, where they develop into epimastigotes and then metacyclic trypomastigotes that are able to infect vertebrates (see Fig 1). In contrast, T. vivax lacks a procyclic stage in the insect mid-gut and has no complex migration within the insect; rather, T. vivax develops directly into epimastigote forms within the insect proboscis [2] (Fig 1). This difference might explain why T. vivax can be transmitted by other kinds of biting insect [3–4] and has therefore spread beyond the sub-Saharan distribution of the tsetse fly into northern Africa and South America [5–6]. In recent years our understanding of trypanosome biology has progressed substantially through the determination of genome sequences for T. brucei [7–10] and for T. congolense and T. vivax [11], as well as numerous analyses of gene expression, (largely confined to T. brucei, except for three studies [12–14]), using transcriptomic [15–20] and proteomic [21–23] approaches. Consequently, we now know that the complex life cycle of African trypanosomes is facilitated by considerable developmental regulation of gene expression. Developmental regulation in T. brucei is particularly apparent in the expression of major surface glycoproteins belonging to the procyclic, epimastigote and bloodstream forms respectively, i. e. procyclin [24], Brucei Alanine-Rich Protein (BARP; [25]) and the Variant Surface Glycoprotein (VSG). There are compelling reasons for supposing that gene expression in T. vivax will be different to T. brucei in important ways, not least due to differences in life cycle development (Fig 1), but also because the T. vivax genome contains substantially different repertoires of VSG and BARP-like genes (and no procyclin at all), as well as numerous gene families that appear to be unique [26]. As in vitro cultivation of insect stages has not previously been possible, gene expression in T. vivax has only been analyzed in the bloodstream form, and then only through transcriptomic analysis [14]. Moreover, given that gene regulation is achieved largely through post-transcriptional modifications in trypanosomes (reviewed in [27]), differences between transcript and peptide abundances across the life cycle are expected. We recently established in vitro cultures of the insect stages of T. vivax [28], and so a comparison of gene expression across African trypanosome species is now possible. Using transcriptome sequencing and proteomics, we have analyzed differences in gene expression between T. vivax epimastigote, metacyclic and bloodstream forms. Our results show that the numerous T. vivax-specific genes predicted to function on the parasite cell surface are transcribed and often developmentally regulated. Genome-wide patterns of developmental regulation are conserved across African trypanosome species, with some notable exceptions concerning pyruvate metabolism in T. vivax, which might indicate an important species difference in energy metabolism. Comparative genomics suggests that T. vivax differs quite considerably from the model T. brucei; by illuminating the expression of distinctive features in the T. vivax genome, this study moves us closer to understanding their phenotypic effects. All mice were housed in the Institut Pasteur animal care facilities in compliance with European animal welfare regulations (European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes CETS No. : 123). Institut Pasteur is a member of Committee #1 of the Comité Régional d’Ethique pour l’Expérimentation Animale (CREEA), Ile de France. Animal housing conditions and the protocols used in the work described herein were approved by the ‘‘Direction des Transports et de la Protection du Public, Sous-Direction de la Protection Sanitaire et de l’Environnement, Police Sanitaire des Animaux” (#B 75-15-28), in accordance with the Ethics Charter of animal experimentation that includes appropriate procedures to minimize pain and animal suffering. Authorization (to PM) to perform experiments on vertebrate animals is granted by license #75–846 issued by the Paris Department of Veterinary Services, DDSV. Trypanosoma (Duttonella) vivax IL 1392 was originally derived from the Zaria Y486 Nigerian isolate. Bloodstream form parasites were maintained in vivo by continuous passage in mice, as previously described [29]. Once parasitemia reached at least 5x108 parasites per ml blood was collected by cardiac puncture onto heparin (2500 IU/kg), and was then diluted 1: 10 (v/v) with PBS 0. 5% glucose to 5x107 parasites per ml. Parasites were separated from red blood cells by differential centrifugation using a swing-out rotor (Jouan GR412, Fisher Bioblock Scientific, Strasbourg, France). Diluted blood was processed by one round of centrifugation (5 min at 200 g) and the supernatant withdrawn with a pipette without disturbing the red blood cell layer and the thin interface containing the white blood cells. Parasite enriched suspension was submitted to a second round of centrifugation (5 min at 200 g) to eliminate all residual cells. The supernatant was then centrifuged for 10 min at 1800 g and bloodstream form-containing pellets devoid of host cells were submitted to two further PBS washes under the same centrifugation conditions. Bloodstream form-containing pellets were further treated for RNA or protein extractions. T. vivax epimastigote cultures have been previously described [28]. Briefly, bloodstream forms purified as described above from infected mice differentiated into epimastigotes in TV3 media: IMDM 50%, DMEM (without glucose) 10% heat-inactivated fetal calf serum (FBS, MP Biomedicals or Invitrogen) and/or 10% heat-inactivated goat serum (GS, Invitrogen), 0. 03 mM bathocuproinedisulfonic acid, 0. 45 mM L-cysteine, 0. 2 mM hypoxanthine, 0. 14 mM ß-mercaptoethanol, 4mM L-proline, 0. 05 mM thymidine, and 25 mM HEPES pH7. 4. All supplements were obtained from Sigma Aldrich except HEPES (Invitrogen, Cergy Pontoise). Epimastigote growth cultures were maintained in vitro by serial passages. Epimastigotes attached to the surface of the culture flask formed micro-colonies and covered the entire surface after two weeks; the number of cells in the supernatant increased proportionally to the density of the adherent cell layer. Adherent epimastigotes were recovered from the flask by scraping and washed three times with PBS. As previously described, metacyclic forms are produced during in vitro growth and are found in the cell culture supernatant [28]. Metacyclic forms were isolated from the cell culture using an approach derived from “bovine plasma aggregation method” [30]. Supernatant from a dense culture (14 days) was remove from the flask, 30% non-inactivated goat serum was added to the cells and incubated at 27°C for 30 min. During the incubation period, epimastigotes aggregate into cell clumps, while metacyclic forms continue to swim freely. The metacyclic forms were then separated from the epimastigote clumps by passing the trypanosome suspension through a 5 μm pore size filter (Millipore Cat. Bedford, MS, USA). The metacyclic forms were then concentrated and washed by centrifugation at 750g for 15 min in 14 ml conical centrifuge tubes and RNA or protein prepared from the resultant cells pellets. Total RNA was isolated using an RNeasy Mini Kit (Qiagen, Courtaboeuf, France) in accordance with the manufacturer' s instructions. RNA purity and concentration were evaluated by spectrophotometry using NanoDrop ND-2000 (ThermoFisher). RNA quality and the relative contributions of total and small RNA were assessed by the Agilent 2100 Bioanalyzer microfluidics-based platform (Agilent Technologies, Santa Clara, USA). Four biological replicates were prepared for bloodstream form and metacyclic cells each. Five replicates were produced for epimastigote cells. For each replicate, poly-adenylated RNA (mRNA) was purified from total RNA using an oligo-dT magnetic bead pull-down, using TruSeq RNA Sample Prep v2 kits (Illumina). The mRNA was then fragmented using metal ion-catalyzed hydrolysis. A random-primed cDNA library was synthesized and double-strand cDNA was used as the input to a standard Illumina library preparation, with a fragment size of 400bp. The libraries were amplified with 10 cycles of PCR using KAPA Hifi Polymerase. Samples were quantified and pooled based on a post-PCR Agilent Bioanalyzer, followed by size-selection using the LabChip XT Caliper. The multiplexed library was sequenced on the Illumina HiSeq 2000 with forward and reverse primers, according to the manufacturers standard protocol, resulting in 100-nucleotide paired-end reads. Sequenced data was analyzed and quality controlled and individual indexed library BAM files created. Paired-end RNA-seq data were mapped to the T. vivax Y486 reference strain [11] (downloaded from TritrypDB release 6. 0) using Bowtie2 [31] under the default parameters and within the Galaxy bioinformatics platform [32]. Transcript abundance for each replicate was estimated across the genome using Cufflinks [33] and measured in Fragments Per Kilobase Mapped (FPKM). The option for quartile normalization within Galaxy was applied to maximize our ability to detect preferential expression of low abundance transcripts against the background of highly abundant species. The option for bias detection and correction was enforced. The option for multi-read correction was applied because some of our genes of interest are multi-copy and may map to multiple locations. Fold change in transcript abundance, and significance of differential expression, was estimated using Cuffdiff [33] for three pairwise comparisons of T. vivax life stages, combining all replicates in each case. Cuffdiff applies the Benjamini-Hochberg correction for multiple testing when assessing the significance of fold changes. To ensure accurate assessment of differential expression, transcript abundance was corroborated using a second method, edgeR [34]. Correlations for fold change in transcript abundance returned by Cufflinks and edgeR displayed high congruence when comparing life stages (r2 = 0. 89–0. 91). Significant differences in transcript expression were defined as at least 2-fold enrichment between conditions and q < 0. 05, where q is the p value corrected for false discovery rate (FDR). Protein from cell lysates was dispensed into low protein-binding microcentrifuge tubes (Sarstedt, Leicester, UK) and made up to 160 μl by addition of 25 mM ammonium bicarbonate. The proteins were denatured using 10 μl of 1% (w/v) RapiGest (Waters MS Technologies, Manchester, UK) in 25 mM ammonium bicarbonate followed by three cycles of freeze-thaw, and two cycles of 10 min sonication in a water bath. The sample was then incubated at 80°C for 10 min and reduced with 3 mM dithiothreitol (Sigma-Aldrich, Dorset, UK) at 60°C for 10 min then alkylated with 9 mM iodoacetamide (Sigma-Aldrich, Dorset, UK) at room temperature for 30 min in the dark. Proteomic grade trypsin (Sigma-Aldrich, Dorset, UK) was added at a protein: trypsin ratio of 50: 1 and samples incubated at 37°C overnight. Three biological replicates were prepared for each cell type. Peptide mixtures were analyzed by on-line nanoflow liquid chromatography using the nanoACQUITY-nLC system (Waters MS technologies, Manchester, UK) coupled to an LTQ-Orbitrap Velos (ThermoFisher Scientific, Bremen, Germany) mass spectrometer equipped with the manufacturer’s nanospray ion source. The analytical column (nanoACQUITY UPLCT BEH130 C18 15 cm x 75 μm, 1. 7 μm capillary column) was maintained at 35°C and a flow-rate of 300nl/min. The gradient consisted of 3–40% acetonitrile in 0. 1% formic acid for 90 min then a ramp of 40–85% acetonitrile in 0. 1% formic acid for 3 min. Full scan MS spectra (m/z range 300–2000) were acquired by the Orbitrap at a resolution of 30,000. Analysis was performed in data-dependent mode. The top 20 most intense ions from MS1 scan (full MS) were selected for tandem MS by collision induced dissociation (CID) and all product spectra were acquired in the LTQ ion trap. Ion trap and Orbitrap maximal injection times were set to 50 ms and 500 ms, respectively. Thermo RAW files were imported into Progenesis LC–MS (version 4. 1, Nonlinear Dynamics, UK). Runs were time aligned using default settings and using an auto selected run as reference. Peaks were picked by the software and filtered to include only peaks with a charge state of between +2 and +6. Peptide intensities were normalized against the reference run by Progenesis LC-MS and these intensities are used to highlight differences in protein expression between control and treated samples with supporting statistical analysis (ANOVA and q-values) calculated by the Progenesis LC-MS software. Spectral data were transformed to mgf files with Progenesis LC-MS and exported for peptide identification using the Mascot (version 2. 3. 02, Matrix Science) search engine. Tandem MS data were searched against a custom database that contained the common contamination and protein sequences predicted for the T. vivax reference genome (downloaded from TriTrypDB v-6. 0). Search parameters were as follows; precursor mass tolerance set to 10ppm and fragment mass tolerance set to 0. 5 Da. One missed tryptic cleavage was permitted. Carbamidomethylation (cysteine) was set as a fixed modification and oxidation (methionine) set as a variable modification. Mascot search results were further processed using the machine learning algorithm Percolator. The false discovery rates were set at 1% and at least two unique peptides were required for reporting protein identifications. Protein abundance (iBAQ) was calculated as the sum of all the peak intensities (from Progenesis output) divided by the number of theoretically observable tryptic peptides [35]. Protein abundance was normalized by dividing the protein iBAQ value by the summed iBAQ values for that sample. The reported abundance is the mean of the biological replicates. All cDNA sequence data are available from the European Nucleotide Archive (http: //www. ebi. ac. uk/ena), accession number ERP001753. Details of the transcriptomic experiments are also available from the Array Express website (https: //www. ebi. ac. uk/arrayexpress/), accession number E-ERAD-100. The mass spectrometry proteomics data have been deposited with the ProteomeXchange Consortium via the PRIDE partner repository (http: //www. ebi. ac. uk/pride/archive/) with the dataset identifier PXD001617. By exploiting new protocols for the in vitro cultivation of T. vivax epimastigote and metacyclic forms, we have produced comparative transcriptomic and proteomic data for the whole T. vivax life cycle using RNAseq and LC-MS/MS approaches respectively. Transcripts were detected for 10116 T. vivax Y486 genes (85. 1% of all genes); 8994 of these transcripts (88. 9%) were observed with at least 10 FPKM. The abundance of each transcript, as estimated using Cufflinks [33], is described in S1 Table. The most abundant transcripts in the bloodstream form were derived from tubulins, diverse ribosomal proteins and VSG-like sequences TvY486_0009580 (17331 FPKM) and TvY486_0018880 (9669 FPKM), which are assumed to have been the active VSG at the time of sequencing. Besides these, abundant transcripts encoding named proteins concern glyceraldehyde 3-phosphate dehydrogenase (TvY486_0603710; 1512 FPKM), a receptor-type adenylate cyclase (TvY486_0029610; 1210 FPKM), cathepsin B-like cysteine peptidase (TvY486_0600060; 794 FPKM), and an uncharacterized gene specific to T. vivax (TvY486_0900440; 1646 FPKM). The most abundant transcripts in the epimastigote and metacyclic cells encoded the same set of highly abundant tubulins and ribosomal proteins, but not the putative VSG, displaying instead an abundance of BARP-like proteins TvY486_0012620 (975 FPKM) and TvY486_1114940 (847 FPKM). Abundance estimates across our independent replicates were consistent, with strong positive correlations of replicates (ranging from 0. 94 to 0. 99) across all life stages (S1 Fig); and when fold change in transcript abundance is compared between life stages using edgeR, replicates cluster by stage illustrating their consistency (S2 Fig). Peptide abundance, as defined by quantitative analysis with MASCOT, is described in S2 Table. 11099 peptides were counted corresponding to 1952 proteins (16. 3% of predicted proteome). Of these, 1245 were sufficiently abundant to be quantified by iBAQ (i. e. two unique peptides were observed with a FDR of 0. 01). Of these, 798 had a q value < 0. 05, meaning that differential expression can be reliably inferred. The most abundant peptides in the bloodstream form were alpha and beta tubulin, various histones, putative VSG (TvY486_0009580/TvY486_0018880; i. e. coinciding with the most abundant VSG-like transcripts), and metabolic enzymes such as fructose-bisphosphate aldolase, enolase, glutamate dehydrogenase, arginine kinase, phosphoglycerate mutase, succinyl-coA: 3-ketoacid-coenzyme A transferase and glycerol-3-phosphate dehydrogenase. The most abundant peptides in the epimastigote and metacyclic stages largely belonged to the same set of proteins, except that VSG were not observed and glycolytic enzymes were less abundant. As with transcript abundance, the proteome was consistent between independent replicates for each life stage, as illustrated in a principle component analysis in which replicates cluster tightly by stage (S3 Fig). The degree to which relative abundance of transcripts and peptides concur throughout the life cycle is an important question with implications for regulation of gene expression, especially in trypanosomatids in which regulation is thought to be mostly post-transcriptional [27]. Correlations of transcript and peptide abundance across the genome (a-c) and for differentially expressed genes in each life stage (d-f) are shown in S4 Fig. These graphs show that the correlation is poor for all genes (r2 between 0. 22 and 0. 36) but improved for genes with evidence of developmental regulation (r2 between 0. 38 and 0. 65). Differential expression is defined as significant where transcript abundance displays at least two-fold enrichment and where q < 0. 05. We found that 11. 2% (1137) of transcripts showed significant differential expression in one or more stage comparison; we refer to these as ‘developmentally regulated’ and they are listed in S3 Table. In bloodstream forms, 518 transcripts were significantly more abundant relative to epimastigotes, and 382 transcripts were significantly more abundant in bloodstream forms relative to metacyclics. The greatest enrichment in favor of bloodstream forms concerned the putative active VSG (TvY486_0018880; fold-change (FC) = 110. 9); other large fold-changes that implicated named sequences concerned three receptor-type adenylate cyclases (TvY486_0026190, TvY486_0003180 and TvY486_0029610; FC = 49. 8,18. 9 and 10. 8 respectively), a glycerol-3-phosphate dehydrogenase (TvY486_0802930; FC = 9. 4) and a phospholipase A1 (TvY486_0102170; FC = 15. 6). Besides these instances, the majority of transcripts (83%) preferentially expressed in bloodstream forms encode hypothetical proteins. Among these are hypothetical proteins belonging to T. vivax-specific families that are included in a Cell-Surface Phylome (CSP) that we published previously [26] for gene families predicted to be expressed on the cell surfaces of the three principal African trypanosome species; i. e. Fam30 (e. g. TvY486_0003670; FC = 29. 4), Fam28 (e. g. TvY486_0030920; FC = 27. 5), Fam34 (e. g. TvY486_0009950; FC = 26. 5) and Fam31 (e. g. TvY486_0000210; FC = 23. 4). Yet another family of uncharacterized genes, unique to T. vivax but not included in the CSP presently, show greater differential expression in bloodstream forms than any other family except for VSG. This gene family occurs 25 times among transcripts up-regulated in bloodstream forms relative to epimastigotes (S3 Table) and provides four of the 20 largest fold-changes in favor of bloodstream forms (e. g. TvY486_0033680, FC = 38. 1). A BLASTp analysis shows that this family has at least 44 members across the T. vivax genome but none of these paralogs were observed to be preferentially expressed in either epimastigote or metacyclic form. In epimastigotes, we identified 393 transcripts that were developmentally regulated, 387 of which are significantly more abundant in epimastigotes relative to bloodstream forms, while 8 transcripts are significantly more abundant in epimastigotes relative to metacyclic forms (see S3 Table). The dearth of preferential expression in epimastigotes relative to metacyclics was only slightly relieved by analysis with edgeR, which reported 23 cases. Since it was necessary to grow epimastigote cultures to high density in order to achieve a high proportion of metacyclic cells, it is possible that the lack of significant differences between these cell types is due to the effects of high density on growth. As with bloodstream forms, most developmentally regulated transcripts encode hypothetical proteins (65. 1%), including those with the greatest fold changes in expression, i. e. TvY486_1110640 (FC = 73. 1). Transcripts preferentially expressed in epimastigotes also concern further T. vivax-specific, CSP gene families, namely Fam35 (11 paralogs; FC = 5. 4–43. 7) and Fam43 (three paralogs; FC = 35. 0–53. 8). However, these gene families were more abundant still in metacyclic forms (see below), indicating that their main focus of expression was not the epimastigote. Aside from these uncharacterized gene families, transcripts implicated in cellular respiration were also seen, for example, components of the electron transfer chain such as cytochrome c1 (TvY486_0801280; FC = 13. 0), cytochrome c (TvY486_0804690; FC = 13. 2) and cytochrome c oxidase subunits (FC between 4. 3–20. 9). Also transcripts for multiple cation transporters (FC between 3. 8–24. 7) and a meiotic recombination protein DMC1 (TvY486_0904120; FC = 10. 5). In the final, metacyclic life stage, 357 transcripts were significantly more abundant relative to bloodstream forms, and these gave a very similar picture to the enriched transcripts in epimastigotes. A further 136 transcripts were significantly more abundant relative to epimastigotes (see S3 Table) including the T. vivax-specific, CSP families 34,35 and 43 (see above), and other transcripts encoding DNA polymerase kappa (TvY486_1109280; FC = 2. 8), an adenylate cyclase (TvY486_0029610; FC = 2. 5), and various reverse transcriptases derived from SLACS elements (average FC = 6. 8). With respect to these observations, it should be noted that further analysis using edgeR produced very similar results to Cufflinks, with only 8. 6% of the gene set displaying significant differential expression. Also, there was substantial overlap in the identities of developmentally regulated transcripts between comparisons; thus, of 518 transcripts significantly enriched in bloodstream forms relative to epimastigotes, 372 of these were also enriched relative to metacyclics; similarly, of 387 transcripts significantly enriched in epimastigotes relative to bloodstream forms, 285 of these were also enriched in metacyclics relative to the latter. From this it should be clear that, of the three life stages, the epimastigotes and metacyclic transcriptomes were most alike. We examined the developmental regulated transcripts for Gene Ontology (GO) terms that were significantly enriched using a Fishers Exact test in BLAST2GO [36]. This confirmed that transcripts preferentially expressed in bloodstream forms are enriched for terms associated with glycolysis (GO: 0006096; q = 3. 38e-04) and glycosome (GO: 0020015; p = 1. 83e-03), while those preferentially expressed in epimastigotes are enriched for cytochrome-c oxidase activity (GO: 0004129; p = 2. 01e-13) and ATP synthesis coupled proton transport (GO: 0015986; p = 1. 43e-03). While suggestive of consistent differences in energy metabolism between life stages, differences in transcript abundance do not guarantee disparity in protein expression; thus, we sought to corroborate these observations with our proteomic data. A protein shows significant differential expression if a constitutive peptide displays at least 2-fold enrichment and q < 0. 05. Under these criteria, 595 or 30. 5% of observed proteins (74. 6% of quantifiable peptides) were developmentally regulated and these are listed in S4 Table. In the bloodstream form, 131 proteins were significantly more abundant relative to epimastigotes, while 63 proteins were significantly more abundant relative to metacyclics. Compared to differentially expressed transcripts, there is a smaller proportion of hypothetical proteins (29%), suggesting that our proteomic analysis has captured the most abundant components of cellular physiology that are relatively well understood. Accordingly, various structural proteins and diverse enzymes are listed, most notably components of the glycolytic pathway such as phosphoglycerate mutase (TvY486_0302920; FC = 6. 6), fructose-bisphosphate aldolase (TvY486_1005670; FC = 3. 5) and glycerol-3-phosphate dehydrogenase (TvY486_0802930; FC = 5. 5). Analysis of functional terms associated with peptides preferentially expressed in bloodstream forms shows that glucose metabolism is significantly enriched (GO: 0006006; FDR = 5. 6e-05) relative to epimastigotes, while the citric acid cycle is significantly enriched (KEGG; FDR = 4. 6e-04) relative to metacyclics. The multi-copy, T. vivax-specific transcripts described above as being the most highly abundant family in bloodstream forms were also identified among our peptides. Four members of this family are preferentially expressed in bloodstream forms, one relative to epimastigotes (TvY486_0008730; FC = 15. 9) and three more relative to metacyclics. None were expressed in either insect stage, further indicating that this is a novel and very prominent feature of bloodstream forms. In the epimastigote form, 147 and 104 proteins were significantly more abundant relative to bloodstream forms and metacyclics respectively. GO terms associated with oxidation-reduction processes (GO: 0055114; FDR = 1. 1e-03) and amino acid metabolism (GO: 0006520; FDR = 2. 3e-02) were found to be significantly enriched. Peptides suggestive of oxidative phosphorylation, although no other elements of mitochondrial energy metabolism, were also significantly more abundant relative to metacyclics. In the metacyclic stage, 62 peptides were significantly more abundant relative to bloodstream forms, and 89 peptides were significantly more abundant relative to epimastigotes. In both cases, the greatest fold changes pertain to hypothetical proteins encoded by members of CSP gene families (see below). Among the 88 peptides, various proteins with roles in intracellular trafficking were implicated, for example, the vesicle formation protein Sec24C (TvY486_0300580; FC = 11. 5; [37]) and signal recognition particle receptor (TvY486_1110560; FC = 4. 9), as well as several dynein heavy chains (FC between 2. 2 and 7. 1). A Fishers Exact test shows that functions associated with intracellular transport (GO: 0046907; FDR = 2. 0e-02) are significantly enriched. The same test shows that purine ribonucleotide catabolism (GO: 0009154; FDR = 2. 1e-03), (which refers here to the ATP-binding requirements of the same intracellular transport processes), is also significantly enriched. It should be noted that, unlike the cohorts of stage-specific transcripts, there was no overlap in the membership of these various stage-specific peptide groups, i. e. there were no peptides significantly enriched in both epimastigotes and metacyclics relative to bloodstream forms. Transcriptomic and proteomic data from diverse African trypanosomes indicate consistently that life stages differ with respect to primary energy metabolism. Previous studies of global gene expression T. brucei and T. congolense have shown developmental regulation of such genes, as well as those encoding the principal cell surface glycoproteins [12–13,16–19,21–22]. We would like to know how conserved such developmental regulation is through evolutionary time, and thus, how regulatory evolution might have contributed to phenotypic differences between species. Comparative genomics predicts that T. vivax lacks certain enigmatic components of T. brucei and T. congolense cell surfaces, such as the VSG-related transferrin receptor of bloodstream forms and procyclin, as well as some elements of Fam50 (see below) [26]. To relate global gene expression across the African trypanosomes, we compared our proteome with existing data sets for T. brucei [21–22] and T. congolense [13], calculating the fold change in abundance from the non-infective insect stage (i. e. epimastigote for T. vivax and procyclic form for T. brucei and T. congolense) to the bloodstream form, for each protein observed in all species (N = 128). Fig 2 compares relative peptide abundance across four proteomic datasets from three species. Subset a contains proteins that are preferentially expressed in the bloodstream form in all species; the GO term for glycolysis (GO: 0006096; FDR = 6. 4e-06) is enriched among these proteins. This indicates that the use of substrate-level phosphorylation as the dominant process for ATP generation in the bloodstream form is a consistent feature of African trypanosomes. Subset b contains proteins with the opposite expression profile, i. e. preferentially expressed in the non-infective insect stage in all species. Analysis of GO terms associated with these proteins shows that proton-transporting ATP synthase activity (GO: 0046933; FDR = 1. 8e-03), ATP synthesis coupled proton transport (GO: 0015986; FDR = 1. 8e-03) and oxidation-reduction process (GO: 0055114; FDR = 3. 1e-03) are enriched. This is consistent with the widespread use oxidative phosphorylation in the low-glucose environment of the insect vector to generate ATP via a proton-motive force across the mitochondrial membrane. Hence, at the broadest level, developmental regulation is conserved across all species, as their shared insect host will have predicted. However, there are obvious differences also. Against this background of conserved developmental regulation, we are interested in genes that are regulated differently in T. vivax, and which might contribute to its unique phenotypes. Subset c contains proteins that are significantly more abundant in the insect stages of T. brucei and T. congolense than the vertebrate stage, but preferentially expressed in T. vivax bloodstream forms. In the larger dataset of Fig 3, this cohort is expanded to 714 proteins by excluding T. congolense (for which proteome coverage is lowest); the number of proteins falling into subset c is increased to 27 and these are listed in Table 1. The expression profile of these proteins in T. vivax is not simply a lack of regulation or low expression generally, since many are highly abundant. Ten of the proteins in Table 1 appear in the top 10% of our proteome when ranked by abundance. Analysis of the GO terms associated with these proteins shows enrichment for succinate-CoA ligase (GDP-forming) activity (GO: 0004776; FDR = 1. 8e-02), pyruvate dehydrogenase (acetyl-transferring) activity (GO: 0004739; FDR = 1. 8e-02) and glucose metabolic process (GO: 0006006; FDR = 2. 8e-02). Hence, comparison of differentially expressed genes across African trypanosomes shows that much is conserved at the regulatory level but that important differences exist, even in the most essential physiology. We have previously identified several gene families, known as Fam27-45, that are predicted to encode cell surface proteins and which, being unique to T. vivax, might distinguish the parasite from T. brucei and T. congolense [26]. Fam27-45 are among the most highly expressed genes and these data are extracted in Table 2. Many of the CSP families unique to T. vivax also appear to be developmentally regulated at the transcript level. For example, Fam27 (five paralogs), Fam35 (11 paralogs) and Fam 43 (five paralogs) are preferentially transcribed in insect stages. Conversely, Fam29 (13 paralogs), Fam30 (38 paralogs) and Fam32 (seven paralogs) are preferentially transcribed in bloodstream forms. Indeed, rarely are transcripts belonging to one of the families found throughout the lifecycle; Fam34 (25 paralogs) being one such case. Fig 4 summarizes the evidence for differential expression of Fam27-45 at the transcript and peptide level. In four cases, (Fams 33,40,41 and 45), both transcriptomic or proteomic evidence for expression is lacking and we conclude that these sequence families do not encode protein-coding genes (and so should be removed from the CSP). As expected, proteomic evidence for gene expression is not as prevalent as transcriptomic data, although it generally corroborates the latter when it is available. The best supported cases for developmental regulation concern the metacyclic-specific expression of Fam34,35 and 43. We observed 34 distinct Fam34 transcripts and 24 of these were differentially expressed; 12 in the metacyclic and another 12 in the bloodstream form. However, the proteomic evidence indicates more selective developmental regulation; of 11 Fam34 proteins that were observed, five were preferentially expressed and all in the metacyclic stage (FC between 2. 2–10. 8). Of 17 distinct Fam35 transcripts, 11 were differentially expressed; all were significantly more abundant in the metacyclic stage (FC between 13. 2–92. 0). Proteomic data support this view; of the six Fam35 peptides observed, all were most abundant in the metacyclic stage and two significantly so (TvY486_0041300 (FC = 6. 55) and TvY486_0039920 (FC = 22. 19) ). We observed seven distinct Fam43 transcripts and of these five were differentially expressed, all most abundant in the metacyclic (FC between 19. 7–86. 3). All three Fam43 peptides that were observed were preferentially expressed in the metacyclic stage (FC between 19. 8–30. 9). Hence, these results indicate that the putative T. vivax-specific gene families are (mostly) genuine protein-coding sequences, and are often developmentally regulated at the transcript and (where observed) protein level. Fam50 is a CSP gene family that includes the BARP genes of T. brucei and the GARP and CESP genes of T. congolense, known to be preferentially expressed on their respective cell surfaces during the insect stages [25,38–40], as well as various, currently uncharacterized, genes that may also be transcribed preferentially during insect stages [41]. The genomic complement of Fam50 genes in T. vivax is smaller and less diverse than those of the other species, which may reflect the simpler existence of T. vivax in the tsetse fly [26]. Our transcriptomic data include all 17 T. vivax Fam50 genes, 13 of which are transcribed most abundantly in the insect stages (see S1 Table). Six transcripts are significantly more abundant in the epimastigote or metacyclic stage relative to bloodstream forms (Table 3a) and one of these was confirmed by the proteomic data (i. e. TvY486_0012620). In total, five proteins were detected and all were differentially expressed (Table 3b); four were most abundant in the insect (FC between 2. 2 and 4. 2). A single protein (corresponding to TvY486_0001140) was significantly more abundant in the bloodstream form. Thus, while transcriptomic data seems to be a poor predictor of differential expression of Fam50 proteins, perhaps suggesting the highly dynamic promotion and repression of Fam50 variants, there is good evidence for developmental regulation of both Fam50 transcripts and peptides, largely in preference for the insect stages and so consistent with observations in other species. The bloodstream forms of African trypanosomes are defined partly by the expression of a VSG coat on the cell surface. In our analysis, we observed 89 distinct VSG transcripts with q < 0. 05 (S1 Table), of which 61 were most abundant in bloodstream forms; however, most of these were observed at very low levels. There were 12 transcripts displaying significant preferential expression in the bloodstream stage (FC between 2. 5–110. 9; Table 4a). In our proteomic analysis we recorded nine distinct VSG sequences, of which three are represented by a single peptide and so unquantified (S2 Table). Of the remaining six (Table 4b), three were most abundant in bloodstream forms, including the two putative active VSG and a third sequence (TvY486_0000810, or identical paralog) that was expressed at a much lower level but still preferentially in bloodstream forms (FC = 5. 7). A fourth VSG was expressed preferentially in the metacyclic stage (TvY486_0027560; FC = 2. 8). Finally, two VSG were most abundant in the epimastigote; one of these (TvY486_0001860; FC = 4. 3) was differentially expressed and the second nearly so (TvY486_0041140; FC = 2. 1; q = 0. 066). Notably, these low abundance VSG expressed in epimastigotes belong to a T. vivax-specific VSG-like family (Fam25), which is not seen in other African trypanosomes. We have produced transcriptomes and proteomes for three different developmental forms of T. vivax, and identified the transcripts and peptides that are significantly enriched in each. These data provide the first profile of global gene expression and developmental regulation throughout the complete T. vivax life cycle. The profile suggests a situation broadly similar to that already observed in T. brucei and T. congolense, though with significant distinctions, not least further evidence for developmental regulation of species-specific cell surface glycoproteins in both the vertebrate and insect stages of T. vivax. Such are their sensitivities, transcriptomic methods typically provide much greater coverage of the genome than do proteomic methods. This might be more pronounced for trypanosomatids since they constitutively express all genes within polycistronic transcripts and regulate protein expression post-transcriptionally [27]. Thus, 85. 1% of T. vivax genes are represented in our transcriptome, but few transcripts are unique to a particular life stage and the proportion of differentially expressed transcripts is only 11. 2%. By contrast, the proteome represents only 16. 3% of all genes, but in 798 cases where differential expression could be assessed, 74. 6% of peptides show significant differential expression and these are unique to one life stage. It may be that developmentally regulated proteins are also particularly abundant, certainly this is true for the components of stage-specific cell surface coats, and this would cause differentially expressed peptides to be overrepresented within the proteome. Previous proteomic studies for T. brucei have reported more proteins than we have found here for T. vivax (i. e. 3553 [21] and 3458 [22]) but a smaller proportion with significant differential expression, i. e. 24. 8%/39. 2% respectively. A proteomic analysis of T. congolense identified 1291 proteins, of which 21. 5% displayed significant differential expression [13]. Hence, it may be that greater sensitivity would reduce the proportion of cases showing differential expression if low abundance proteins are more likely to be constitutively expressed. Previously, comparison of the T. vivax genome sequence with those of T. brucei and T. congolense demonstrated that there are more than 2000 genes that are only present in T. vivax [11]. Their specificity, and typically the absence of any recognizable protein domains, make these genes obscure. Nonetheless, they appear to be genuine, since we found a lack of protein-coding evidence in only a few cases. One defining feature is that they comprise multi-copy gene families (Fam27-45 of the CSP) that are thought to be expressed on the cell surface based on the presence of a putative signal peptide, transmembrane domain and/or glycerophosphosinositol anchor in their predicted protein sequence [26]. Comparative genomics also showed that T. vivax lacked procyclin, the canonical cell surface glycoprotein of T. brucei and T. congolense insect stages [42]. This is consistent with T. vivax lacking a procyclic stage in the insect mid-gut and raises the question of what coats the T. vivax surface if not procyclin. Clearly, the abundant T. vivax-specific gene families offer plausible candidates for the role, but it could also be filled by Fam50; which has been shown to include various surface glycoproteins expressed during the insect stages of T. brucei and T. congolense [26]. Certainly, our data indicate that multiple Fam50 proteins are expressed in T. vivax and preferentially expressed in the epimastigote, while several transcripts (belonging to different loci) were significantly enriched in both epimastigotes and metacyclics. The fact that the transcripts and peptides are not derived from the same loci may suggest that different Fam50 genes became activated in the period between our preparation of RNA and protein. This is not the situation we observe for VSG, for which the identity of enriched transcripts and peptides largely match, suggesting that regulation of Fam50 gene expression is highly dynamic with multiple isoforms being promoted and repressed over short intervals. The presence of transcripts in the metacyclic stage at levels comparable to the epimastigote levels may be an artefact (i. e. residual epimastigotes in the metacyclic preparation), since Fam50 peptides in metacyclics are sparse and comparable in abundance to bloodstream forms. In short, the expression of multiple Fam50 proteins in the T. vivax epimastigote supports the view derived from T. brucei and T. congolense that this is a conserved family of glycoproteins performing diverse roles in the insect stages of the life cycle. Given that BARP in T. brucei and CESP in T. congolense are cell-surface glycoproteins [25,40], Fam50 homologs are therefore probably a prominent component of the epimastigote cell surface in T. vivax. Although Fam50 is preferentially expressed in epimastigotes, other multi-copy families, including the various T. vivax-specific cases, seldom are. This study has confirmed that most T. vivax-specific gene families are expressed. In the cases of Fam33,40,41 and 45, which should now be discounted from the CSP, the apparent lack of transcription raises the question of what function these repetitive non-coding sequences might perform. Three T. vivax-specific gene families (Fam34,35 and 43) are very strongly enriched in the metacyclic stage, which is intriguing because in T. brucei and T. congolense the metacyclic coat is characterized by VSG. While metacyclic VSG are replaced by other VSG upon differentiation into bloodstream forms, and so are temporally distinct, there is no metacyclic-specific cohort of VSG sequences [8,11]. VSG may also be present on T. vivax metacyclics, since we observed a low abundance VSG protein preferentially expressed in metacyclics (TvY486_0027560). However, assuming that Fam34,35 and 43 are expressed on the cell surface as predicted, it is clear that the infective form of T. vivax has a qualitatively different surface architecture to the other species, with a considerable non-VSG component. The same could be claimed for bloodstream forms also. Three families show exclusive enrichment in bloodstream forms at the transcript level (Fam28-30), though without proteomic support. This could indicate that our analysis lacked the sensitivity to detect them, perhaps because in bloodstream forms the superabundant VSG dominates the sequencing effort, making the detection of lower abundance proteins less effective than for either metacyclic or epimastigote. The presence of numerous non-VSG surface proteins might account for the observation that the T. vivax VSG coat is less dense than that of T. brucei [43], and VSG comprise a smaller proportion of cell surface-expressed T. vivax transcripts [14]. Assuming that their surface role is correct and they are confirmed as having preferential expression in bloodstream forms, these families are particularly interesting they have properties as surface antigens that could be targeted for vaccine development. Immune responses to the VSG do not provide lasting and comprehensive protection because of antigenic variation and the considerable structural diversity of the VSG repertoire. By contrast, Fam28-30 number not more than 30 paralogs, are less structurally diverse, and multiple transcripts are expressed at relatively high abundance, indicating that these families are not subject to antigenic variation, if monoallelic expression is a diagnostic feature of that process. The VSG genes themselves present an expression profile typical of other African trypanosomes. VSG expression is regulated to produce a succession of structural variants that can evade specific immune responses but also prevent exposure of the total VSG structural repertoire to the host immune system, which would lead to a comprehensive immune response. Thus, VSG genes are expressed in a monoallelic fashion from the highly regulated context of a dedicated VSG expression site [44]. In their analysis of expressed sequence tags (EST) from different T. congolense life stages, Helm et al. (2009) recorded 13 distinct VSG transcripts in metacyclic cells, with the most abundant comprising 24% of the total number, and 26 distinct VSG transcripts in bloodstream forms, with the most abundant contributing 62% of the total [12]. This supports the established experimental model in which most individuals of a T. congolense population express the same active VSG, while a few individuals express a range of low abundance alternatives. In fact, when combined, the 12 least abundant VSG EST were only 0. 5% of all VSG transcripts in bloodstream forms [12]. Similarly in T. brucei, Jensen et al. (2009) identified in a microarray-based study cohorts of less abundant transcripts in addition to the known, active VSG, some of which were expressed most abundantly in the insect stages [16]. In contrast, a previous analysis of VSG transcripts in T. vivax using 454 sequencing technology concluded that only one VSG was expressed [14]. Proteomic analyses have presented a similar picture. In T. congolense, 11 different VSGs were identified across all life stages [13]. Four were confidently associated with the metacyclic stage while two others were significantly enriched in bloodstream forms (including the known, active VSG). In proteomic comparisons of procyclic and bloodstream forms of T. brucei, one proteome identified 10 canonical VSGs [22], while another only two [21], although different T. brucei strains were used. These did not include the active VSG because neither study used the reference strain (927), and so the active VSG did not map to a VSG gene in the reference genome. Consequently, the 10 VSGs identified by Butter et al. (2013) are all low abundance alternatives, represented by few peptides (< 7) and achieving poor coverage (< 9%) [22]. Three of these low abundance VSGs were preferentially expressed in procyclic forms [22]. Taking these previous data together, their obvious methodological variations notwithstanding, low abundance alternatives to the dominant VSG are observed at both the transcript and protein levels in both T. brucei and T. congolense. The role, if any, of these ‘accessory’ VSGs is unclear; some are very likely metacyclic VSGs and it is known that expression of these can continue several days after transmission [45] and so could be present in bloodstream forms. Alternatively, ‘accessory’ VSGs may simply be rare antigens expressed by low frequency subpopulations, or produced by inefficiency in the mechanism silencing inactive VSG expression sites. In contrast to the previous result [14], we observed several low abundance VSGs in T. vivax consistent with the expression profiles of VSG observed in T. brucei and T. congolense. The two dominant VSG sequences were superabundant at both transcript and peptide levels. Therefore the identity of the active VSG remained constant in the period between RNA and protein preparation, meaning that this is unlikely to represent a transition between two VSGs and that T. vivax strain IL1392 probably a mixture of parasites expressing one of two different VSGs. Both of these active VSG belong to Fam24, the subtype homologous to canonical b-type VSG in T. brucei and T. congolense [11]. In a similar fashion to T. congolense, the less abundant VSG in T. vivax may represent metacyclic VSGs. One VSG, TvY486_0027560, may be a metacyclic VSG in this strain as it was preferentially expressed in the metacyclic form (its transcript was not recorded). Finally, two VSG-like sequences belonging to Fam25, a T. vivax-specific subtype [11], are preferentially expressed at low levels in epimastigotes. The roles of Fam25 and 26 genes remain mysterious, and there is no definitive evidence that they encode variant antigens. Beyond the differences in genetic repertoire that are evident from comparative genomics, it is presumed that differences in the regulation of conserved genes will contribute to phenotypic differences between African trypanosomes. The cohort of conserved genes identified in Figs 2 and 3 that are regulated conversely in T. vivax relative to T. brucei (and probably T. congolense) indicate that this is so. In the bloodstream stage, African trypanosomes exclusively employ glycolysis to exploit abundant glucose in host plasma to generate ATP via substrate-level phosphorylation in the glycosome [46]. In the tsetse fly, where glucose is limited but amino acids such as proline are present in the host hemolymph, the parasites generate ATP through gluconeogenesis and oxidative phosphorylation in the mitochondrion [46]. In this regard, T. vivax is consistent; all glycolytic enzymes are preferentially expressed in the bloodstream form, where they are among the most abundant transcripts and peptides, and all components of the electron transfer chain are preferentially expressed in the epimastigote (S3 and S4 Tables). The species differences highlighted in Fig 3 concern the metabolic steps linking glycolysis and events in the mitochondrion, i. e. pyruvate metabolism (Fig 5). Experimental evidence indicates that T. brucei produces ATP during its insect stage by further substrate-level phosphorylation in the glycosome, by catabolizing phosphoenolpyruvate (PEP), and in the mitochondrion by catabolizing pyruvate. This results in insect forms excreting succinate and acetate, while bloodstream forms excrete pyruvate [47]. Accordingly, the enzymes for converting PEP into succinate and pyruvate into acetate are preferentially expressed in the procyclic form of T. brucei [21,22]. Fig 5 describes the points in this pathway where differential expression is reversed in T. vivax. We see that enzymes for the catabolism of PEP, such as glycosomal malate dehydrogenase and glycosomal phosphoenolpyruvate carboxykinase, and for the conversion of pyruvate to acetate, i. e. multiple components of the pyruvate dehydrogenase complex and of the succinyl-CoA synthetase complex, are significantly more active in the bloodstream form than in the epimastigote. Additionally, the fumarase responsible for reaction 13 in Fig 5, while upregulated in procyclic form T. brucei, is constitutively expressed in T. vivax (TvY486_1105200; FC = 0. 04). However, the final enzyme in the pathway (NADH-dependent fumarate reductase; reaction 14) is preferentially expressed in the insect stages in both species. We speculate that some other genes in Table 1 support this function; for example, TvY486_0901260, which possesses a mitochondrial pyruvate carrier protein domain homologous to mt1 in Humans, which is required to import pyruvate across the inner mitochondrial membrane [48], and TvY486_0702860, which encodes a bacterial-type nitro-FMN oxidoreductase that might serve to regenerate NAD+ [49]. Thus, we would predict that T. vivax excretes fumarate, acetate and perhaps succinate in its bloodstream stage rather than in the insect. It is not clear why T. vivax would benefit from pyruvate metabolism in the bloodstream when substrate-level phosphorylation using glucose should suffice. However, in the insect stage, when the parasite remains in the proboscis and without access to the hemolymph, it could be that such metabolism serves little purpose. Therefore, this may reflect a lack of upregulation in the epimastigote rather than adaptive upregulation in the bloodstream form, illustrating how life cycle variation has affected the regulation of energy metabolism in these organisms. The first global perspective on gene expression in T. vivax has confirmed that a broadly similar process of developmental regulation occurs in all African trypanosome species. However, subtle differences, for instance in energy metabolism and putative cell surface molecules, offer new insights into the molecular basis for the life cycle differences that exist between species. Beyond the background of conservation, this study has confirmed the presence of numerous T. vivax-specific gene families and shown that these are developmentally regulated, indicating that the surface of T. vivax differs quite substantially from the model derived from other African trypanosomes.

Trypanosoma vivax is a single-celled parasite that infects cattle and non-domesticated animals through the bite of the tsetse fly. The parasite causes animal trypanosomiasis, a chronic condition resulting in severe anemia, muscle wastage and ultimately death if untreated. This disease is endemic across sub-Saharan Africa but has also spread to South America and causes considerable losses in animal productivity, impeding economic development in the world’s poorest nations. To develop new ways of preventing and treating animal trypanosomiasis, we need an accurate understanding of how the parasite causes disease. In this study, we present an analysis of gene expression throughout the T. vivax life cycle that compares the abundance of gene transcripts (mRNA) and proteins in the mammalian and insect hosts. We have identified genes that are preferentially expressed in each life stage, including many that are unique to T. vivax and probably expressed on its cell surface. Our findings provide a comprehensive understanding of how gene expression is regulated in T. vivax and further refine a pool of T. vivax-specific genes that could be exploited to prevent and treat animal trypanosomiasis.

Abstract Introduction Methods Results Discussion

2015

Global Gene Expression Profiling through the Complete Life Cycle of Trypanosoma vivax

Speech production involves the movement of the mouth and other regions of the face resulting in visual motion cues. These visual cues enhance intelligibility and detection of auditory speech. As such, face-to-face speech is fundamentally a multisensory phenomenon. If speech is fundamentally multisensory, it should be reflected in the evolution of vocal communication: similar behavioral effects should be observed in other primates. Old World monkeys share with humans vocal production biomechanics and communicate face-to-face with vocalizations. It is unknown, however, if they, too, combine faces and voices to enhance their perception of vocalizations. We show that they do: monkeys combine faces and voices in noisy environments to enhance their detection of vocalizations. Their behavior parallels that of humans performing an identical task. We explored what common computational mechanism (s) could explain the pattern of results we observed across species. Standard explanations or models such as the principle of inverse effectiveness and a “race” model failed to account for their behavior patterns. Conversely, a “superposition model”, positing the linear summation of activity patterns in response to visual and auditory components of vocalizations, served as a straightforward but powerful explanatory mechanism for the observed behaviors in both species. As such, it represents a putative homologous mechanism for integrating faces and voices across primates. When we speak, our face moves and deforms the mouth and other regions [1], [2], [3], [4], [5]. These dynamics and deformations lead to a variety of visual motion cues (“visual speech”) related to the auditory components of speech and are integral to face-to-face communication. In noisy, real world environments, visual speech can provide considerable intelligibility benefits to the perception of auditory speech [6], [7], faster reaction times [8], [9], and is hard to ignore—integrating readily and automatically with auditory speech [10]. For these and other reasons, it' s been argued that audiovisual (or “multisensory”) speech is the primary mode of speech perception and is not a capacity that is simply piggy-backed onto auditory speech perception [11]. If the processing of multisensory signals forms the default mode of speech perception, then this should be reflected in the evolution of vocal communication. Naturally, any vertebrate organism (from fishes and frogs, to birds and dogs) that produces vocalizations will have a simple, concomitant visual motion in the area of the mouth. However, in the primate lineage, both the number and diversity of muscles innervating the face [12], [13], [14] and the amount of neural control related to facial movement [15], [16], [17], [18] increased over time relative to other taxa. This ultimately allowed the production of a greater diversity of facial and vocal expressions in primates [19], with different patterns of facial motion uniquely linked to different vocal expressions [20], [21]. This is similar to what is observed in humans. In macaque monkeys, for example, coo calls, like the /u/ in speech, are produced with the lips protruded, while screams, like the /i/ in speech, are produced with the lips retracted [20]. These and other homologies between human and nonhuman primate vocal production [22] imply that the mechanisms underlying multisensory vocal perception should also be homologous across primate species. Three lines of evidence suggest that perceptual mechanisms may be shared as well. First, nonhuman primates, like human infants [23], [24], [25], can match facial expressions to their appropriate vocal expressions [26], [27], [28], [29]. Second, monkeys also use eye movement strategies similar to human strategies when viewing dynamic, vocalizing faces [30], [31], [32]. The third, indirect line of evidence comes from neurophysiological work. Regions of the neocortex that are modulated by audiovisual speech in humans [e. g. , 8,33,34,35,36,37], such as the superior temporal sulcus, prefrontal cortex and auditory cortex, are similarly modulated by species-specific audiovisual communication signals in the macaque monkey [38], [39], [40], [41], [42], [43]. However, none of these behavioral and neurophysiological results from nonhuman primates provide evidence for the critical feature of human audiovisual speech: a behavioral advantage via integration of the two signal components of speech (faces and voices) over either component alone. Henceforth, we define “integration” as a statistically significant difference between the responses to audiovisual versus auditory-only and visual-only conditions[44]. For a homologous perceptual mechanism to evolve in monkeys, apes and humans from a common ancestor, there must be some behavioral advantage to justify devoting the neural resources mediating such a mechanism. One behavioral advantage conferred by audiovisual speech in humans is faster detection of speech sounds in noisy environments—faster than if only the auditory or visual component is available [8], [9], [45], [46]. Here, in a task operationalizing the perception of natural audiovisual communication signals in noisy environments, we tested macaque monkeys on an audiovisual ‘coo call’ detection task using computer-generated monkey avatars. We then compared their performance with that of humans performing an identical task, where the only difference was that humans detected /u/ sounds made by human avatars. Behavioral patterns in response to audiovisual, visual and auditory vocalizations were used to test if any of the classical principles or mechanisms of multisensory integration [e. g. 47,48,49,50,51,52,53] could serve as homologous computational mechanism (s) mediating the perception of audiovisual communication signals. We report two main findings. First, monkeys integrate faces and voices. They exhibit faster reaction times to faces and voices presented together relative to faces or voices presented alone —and this behavior closely parallels the behavior of humans in the same task. Second, after testing multiple computational mechanisms for multisensory integration, we found that a simple superposition model, which posits the linear summation of activity from visual and auditory channels, is a likely homologous mechanism. This model explains both the monkey and human behavioral patterns. All experiments and surgical procedures were performed in compliance with the guidelines of the Princeton University Institutional Animal Care and Use Committee. For human participants, all procedures were approved by the Institutional Review Board at Princeton University. Informed consent was obtained from all human participants. Nonhuman primate subjects were two adult male macaques (Macaca fascicularis). These monkeys were born in captivity and provided various sources of enrichment, including cartoons displayed on a large screen TV as well as olfactory, auditory and visual contact with conspecifics. The monkeys underwent sterile surgery for the implantation of a head-post. The human participants consisted of staff or graduate students (n = 6,4 males, mean age = 27) at Princeton University. Two of the subjects were authors on the paper (CC, LL). The other four human subjects were naïve to the purposes and goals of the experiment. We would like to briefly explain here why we chose to use avatars. First, it is quite difficult to record monkey vocalizations which only contain mouth motion without other dynamic motion components such as arbitrary head motion and rotation— which themselves may lead to audiovisual integration [54]. Second, start and end positions of the head from such videos of vocalizations, at least for monkeys, tend to be very variable which would add additional visual motion cues. Third, we wanted constant lighting and background and the ability to modulate the size of the mouth opening and thereby parameterize visual stimuli. Fourth, the goal of this experiment was to understand how mouth motion integrated with the auditory components of vocalizations and we wanted to avoid transient visual stimuli. Real videos would not have allowed us to control for these factors; avatars provide us with considerable control. Experiments were conducted in a sound attenuating radio frequency (RF) enclosure. The monkey sat in a primate chair fixed 74 cm opposite a 19 inch CRT color monitor with a 1280×1024 screen resolution and 75 Hz refresh rate. The 1280×1024 screen subtended a visual angle of ∼25° horizontally and 20° vertically. All stimuli were centrally located on the screen and occupied a total area (including blank regions) of 640×653 pixels. For every session, the monkeys were placed in a restraint chair and head-posted. A depressible lever (ENV-610M, Med Associates) was located at the center-front of the chair. Both monkeys spontaneously used their left hand for responses. Stimulus presentation and data collection were performed using Presentation (Neurobehavioral Systems). Experiments were conducted in a psychophysics booth. The human sat in a comfortable chair approximately 65 cm opposite a 17 inch LCD color monitor with a 1280×1024 screen resolution and 75 Hz refresh rate. The 1280×1024 screen subtended a visual angle of 28 degrees horizontally and 24 degrees vertically. All stimuli were centrally located on the screen and occupied an area of 640×653 pixels. All stimulus presentation and data collection were performed using Presentation (Neurobehavioral Systems). Our audiovisual detection experiment is an extension of the classical redundant signals paradigm. In such experiments, it is common to observe that RTs to multisensory targets presented simultaneously are faster than unisensory RTs. This effect is usually termed the “redundant signals effect”. One important class of explanations for the redundant signals effect is the “race model”. According to the race model (or a “parallel first terminating” model), redundancy benefits are not due to an actual integration of visual and auditory cues. To illustrate, assume that the time to detect and respond to a single target—e. g. , the facial motion--varies according to a statistical distribution. Similarly, the time to detect and respond to the auditory-only vocalization also varies according to a statistical distribution. Whenever, both facial motion and the vocalization are presented together, the stimulus that is processed faster in a given trial determines the response time. As the RT distributions typically overlap with one another, slow processing times are removed. Thus, RTs to redundant stimuli are always faster than those for the single stimuli. A standard way to test whether this principle can explain RT data is to use the race model inequality [57], which posits that the cumulative RT distribution for the redundant stimuli never exceeds the sum of the RT distributions for the unisensory stimuli. That is, if FAV (t), FV (t) and FA (t) are the estimated cumulative distributions (CDF) of the RTs for the three different modalitiesthen one cannot rule out race models as an explanation for the facilitation of RT. On the other hand, if this inequality is violated in a given data set, then parallel processing cannot completely account for the benefits observed for multisensory stimuli and an explanation based on co-activation is necessary. We computed the CDFs of our conditions and then computed the difference between the actual CDF of the audiovisual condition and the CDF predicted by the race model. The maximum positive point of this difference was taken as the index of violation, R. Positive values of R means that the race model is rejected. If this value is 0, then the race model is upheld. Several models of audiovisual integration have been proposed over the years, but superposition models are simple and possess considerable explanatory power. Here we briefly describe the model, and detailed explanations are available elsewhere [63], [64]. We first consider the case of single modality trials (visual or auditory). We assume that the onset of the stimulus (i. e. visual mouth motion or the auditory vocalization) induces a neural renewal counting process (for examples, action potentials or spikes, but it could be any event which is counted) that counts up to a critical number of events, c. The assumption is that, as soon as a critical number of events, c, have been registered at some decision mechanism, a response execution process, M, which consumes an amount of time with a mean µM, is started. The main postulate of the superposition model is that in the audiovisual condition the renewal processes generated by either the visual and the auditory signals are superposed, thereby reducing the waiting time for the critical count. Specifically, if NV (t) and NA (t) are themselves counting processes for the visual-only and auditory-only conditions, and the two stimuli are presented simultaneously, that is with 0 lag, then the new process for the audiovisual stimulus is given as It is immediately apparent that this audiovisual process will reach the critical level of c counts faster than the individual auditory and visual processes. If the auditory-only and visual-only stimuli are presented with a lag of say τ, as in our case with visual mouth motion preceding the auditory vocalization by τ milliseconds, then the process becomes, To specify this model fully and test and fit to data, one must specify an inter-arrival distribution. Usually this is assumed to be exponential in nature that leads to a homogenous Poisson counting process. For τ = 0, the waiting time for the cth event is well defined and follows a gamma distribution with mean c/λ and variance c/λ2, where λ (λ>0) is the intensity parameter of the Poisson process. For example, the auditory and visual-only RT would then be The mean audiovisual RT would be given by the following simple expression. It is the waiting time for the cth with the visual and auditory rates summed and is given as follows. When this model is to be applied when there are differences in the SOAs, that is, τ>0, the waiting time for the cth event is no longer gamma distributed and instead follows a more complicated distribution. Fortunately, this model has been completely explicated and published expressions are already available [63], [64]. The audiovisual RT in this case is the expected value of the waiting time to the cth count. We measured the accuracy of the monkeys and humans detecting vocalizations in the audiovisual, auditory-only and visual-only conditions. Figure 2A shows the detection performance of Monkey 1 averaged over all sessions (both coo calls) as a function of SNR for the three conditions of interest. In the case of the visual-only condition, the size of mouth opening has a constant relationship with the auditory SNR and it is thus shown on the same x-axis. In the auditory-only condition, as the coo call intensity increased relative to the background noise, the detection accuracy of the monkey improved. In contrast, modulating the size of the mouth opening in the visual-only condition had only a weak effect on the detection accuracy. Finally, the detection accuracy for audiovisual vocalizations was mildly enhanced relative to the visual-only condition and with very little modulation as a function of the SNR. The same pattern was seen for Monkey 2 (Figure 2B). When the data was pooled over all the SNRs, accuracy was significantly better for both monkeys in the audiovisual condition compared to either unisensory condition (paired t-tests, Monkey 1: AV vs V, t (47) = 3. 77, p<. 001, AV vs A, t (47) = 19. 94, p<. 001; Monkey 2: AV vs V, t (47) = 15. 85, p<. 001, AV vs A, t (47) = 8. 1, p<. 001). This general pattern was replicated in humans (n = 6). Figure 2C shows the performance of a single human observer on this same task detecting the /u/ sound. Excluding the lowest SNR value in the auditory-only condition, accuracy is almost at ceiling for all three stimulus conditions. The average accuracy over the 6 human subjects as a function of SNR is shown in Figure 2D. Performance pooled across all SNRs was maximal for the audiovisual condition and was enhanced relative to the auditory-only condition (t (5) = 2. 71, p = 0. 04). It was not significantly enhanced relative to the visual-only condition (t (5) = 0. 97, p = 0. 37). The lack of enhancement relative to the visual-only condition is likely because the visual-only performance itself was close to ceiling for humans. In both species, the similarities in detection accuracy for visual-only and audiovisual conditions (Figures 2A–D) suggest that they were perhaps not integrating auditory and visual signals but instead may have adopted a unisensory (visual) strategy. According to this strategy, they used visible mouth motion only for both the visual and audiovisual conditions, and used the sound only when forced to do so in the auditory-only condition. We therefore examined the reaction times (RTs) to distinguish between a unisensory versus an integration strategy. Figures 3A, B show the mean RT as a function of the SNR and modality computed by pooling RT data from all the sessions for Monkeys 1 and 2. RTs for the auditory-only vocalization increased as the SNR decreased (i. e. the sound was harder to hear relative to the background). In contrast, RTs to the visual-only condition only increased weakly with an increase in the mouth opening size — a result consistent with the accuracy data. Although the audiovisual accuracy was only modestly better than the visual-only accuracy (Figure 2A, B), audiovisual RTs decreased relative to both auditory-only and visual-only RTs for several SNR levels. To illustrate, a non-parametric ANOVA (Kruskal-Wallis) computed for Monkey 1, which compared the ranks of the RTs for the auditory-only, visual-only and audiovisual conditions for the highest SNR (+22dB), was significant (χ2 = 490. 91, p<. 001). Post-hoc Mann-Whitney U tests revealed that the RT distribution for the audiovisual condition was significantly different from the auditory-only distribution and the visual-only distribution for all SNRs; that is, RTs in the audiovisual condition were faster than visual and auditory RTs. In Monkey 2, the audiovisual RT distribution was different from the auditory-only distribution for all SNRs (p<0. 001), and was significantly different from the visual-only distribution for all but the lowest SNR (−10 dB, p = 0. 68). It is notable that at the highest SNR (+22 dB; largest mouth opening), the RTs of both monkeys seem more like the auditory-only RTs, while at the lowest SNR (−10 dB; smallest mouth opening), the RTs seem to be more similar to the visual-only RTs. Humans also show a RT benefit in the audiovisual versus unisensory conditions, with a similar, but not identical, pattern to that observed in monkeys. Figure 3C shows the average RTs of a single human subject as a function of the SNR. Similar to monkeys, decreasing the SNR of the auditory-only condition leads to an increase in the RTs, and RTs for the visual-only condition were only weakly modulated by the size of the mouth opening. For a range of SNRs, the audiovisual RTs were faster than auditory- and visual-only RTs. Figure 3D shows the average RTs over all 6 subjects. Paired t-tests comparing audiovisual RTs to auditory-only and visual-only RTs reveal that they were significantly different in all but the lowest SNR condition (p = 0. 81 for the −10 dB condition, p<0. 05 for all other conditions, df = 5). Though the RT patterns from human participants seem dissimilar to the monkey RT patterns (e. g. , in monkeys the auditory-RT curve crossed the visual-only RT curve but for humans there was no cross over), we can show that the two species are adopting a similar strategy by exploring putative mechanisms. We do so in the next sections. Our analysis of RTs rules out the simple hypothesis that monkeys and humans are defaulting to a unisensory strategy (using visual in all conditions except when forced to use auditory information). Another hypothesis is that a “race” mechanism is at play [59]. A race mechanism postulates parallel channels for visual and auditory signals that compete with one another to terminate in a motor or decision structure and thereby trigger the behavioral response. We chose to test this model to ensure that the observers were actually integrating the faces and vocalizations of the avatar. A simple physiological correlate of such a model would be the existence of independent processing pathways for the visual mouth motion and an independent processing pathway for the auditory vocalization. In the race scenario, there would be no cross-talk between these signals. Race models are extremely powerful and are often used to show independent processing in discrimination tasks [70], [71], [72]. In our task, independent processing would mean that in the decision structure, two populations of neurons received either auditory or visual input. These two independent populations count spikes until a threshold is reached; the population that reaches threshold first triggers a response. Such a model can lead to a decrease in the RTs for the multisensory condition, not through integration, but through a statistical mechanism: the mean of the minimum of two distributions is always less than or equal to the minimum of the mean of two distributions. Figure 4A shows a simulation of this race model. The audiovisual distribution, if it is due to a race mechanism, is obtained by taking the minimum of the two distributions and will have a lower mean and variance compared to the individual auditory and visual distributions. Typically, to test if a race model can explain the data, cumulative distributions of the RTs (Figure 4B) are used to reject the so-called race model inequality [51], [57]. The inequality is a strong, conservative test and provides an upper bound for the benefits provided by any class of race models. Reaction times faster than this upper bound mean that the race model cannot explain the pattern of RTs for the audiovisual condition; the RT data would therefore necessitate an explanation based on integration. Figure 4C plots the cumulative distributions for RTs collected in the intermediate SNR level and for ISIs between 1000 and 1400 ms for Monkey 1; the prediction from the race model is shown in grey. We used this ISI interval because, in monkeys only, the ISI influenced the pattern of audiovisual benefits (see Text S1 and Figure S2). Maximal audiovisual benefits were for ISIs in the 1000–1400 ms range. The cumulative distribution of audiovisual RTs is faster than can be predicted by the race model for multiple regions of RT distribution, suggesting that the RTs cannot be fully explained by this model. To test whether this violation was statistically significant, we compared the violation from the true data to one using conservative bootstrap estimates. Several points for the true violation were much larger than the violation values estimated by bootstrapping (Figure 4D). Audiovisual RTs are therefore not explained by a race model. For the entire range of SNRs and this ISI for the monkeys, maximal race model violations were seen for the intermediate to high SNRs (+5, +13 and + 22 dB; Figure 4E). For the softer SNRs (−10, −4 dB), a race model could not be rejected as an explanation. The amount of race model violation for the entire range of ISIs and SNRs is provided in Figure S3. For both monkeys, longer ISIs resulted in weaker violations of the race model and rarely did the p-values from the bootstrap test reach significance. For humans, we observed similar robust violations of the race model. Figure 4F shows the average amount of race model violation across subjects as a function of SNR. Since humans showed much less dependence on the ISI, we did not bin the data as we did for monkeys. Similar, to monkeys, maximal violation of the race model was seen for loud and intermediate SNRs. For 3 out of the 5 SNRs (+22, +13, +5 dB), a permutation test comparing maximal race model violation to a null distribution was significant (p<0. 05). In conclusion, for both monkeys and humans, a race model cannot explain the pattern of RTs at least for the loud and intermediate SNRs. These results strongly suggest that monkeys do integrate visual and auditory components of vocalizations and that they are similar to humans in their computational strategy. In the next sections, we therefore leveraged these behavioral data and attempt to identify a homologous mechanism (s) that could explain this pattern of results. Our search was based on the assumption that classical principles and mechanisms of multisensory integration [48], [49], [50], [51], [73], originally developed for simpler stimuli, could potentially serve as starting hypotheses for a mechanism mediating the behavioral integration of the complex visual and auditory components of vocalizations. The first mechanism we tested was whether the integration of faces and voices demonstrated in our data followed the “principle of inverse effectiveness” [49], [50]. This idea, originally developed to explain neurophysiological data, suggests that maximal benefits from multisensory integration should occur when the stimuli are themselves maximally impoverished [49], [50], [74], [75]. That is, the weaker the magnitude of the unisensory response, the greater would be the gain in the response due to integration. In our case with behavior, this principle makes the following prediction. As the RTs and accuracy were the poorest for the lowest auditory SNR, the benefit of multisensory integration should be maximal when the lowest auditory SNR is combined with the corresponding mouth opening. Our metric for multisensory benefit was defined as the speedup for the audiovisual RT relative to the fastest mean RT in response to the unisensory signal (regardless of whether it was the auditory- or visual-only condition). The principle of inverse effectiveness would thus predict greater reaction time benefits with decreasing SNR for both monkeys and humans. Figures 5A and B plot this benefit as a function of SNR for Monkeys 1 and 2. For monkeys, the maximal audiovisual benefit occurs for intermediate SNRs. The corresponding pattern of benefits for humans is shown in Figure 5C. For humans, this benefit increases as the SNR increases and starts to flatten for the largest SNRs. This pattern of benefits reveals that the maximal audiovisual RT benefits do not occur at the lowest SNRs. This is at odds with the principle of inverse effectiveness [49], [50]. If our results had followed this principle, then the maximal benefit relative to both unisensory conditions should have occurred at the lowest SNR (lowest sound intensity coupled with smallest mouth opening). Neither monkey nor human RTs followed this principle and therefore it cannot be a homologous mechanism mediating the integration of faces and voices in primates. One potential caveat is that we are testing the principle of inverse effectiveness using absolute reaction time benefits whereas the original idea was developed using proportional referents. Thus, we re-expressed the benefits as a percent gain relative to the minimum of the auditory and visual reaction times for each SNR. We observed that, even when converted to a percent benefit relative to the minimum reaction time for each SNR, the inverted U-shape pattern of gains for monkeys (Figures S4A, B), as well as increasing gain with SNR for humans (Figure S4C), was replicated. Thus, whether one uses raw benefits or a proportional measure, RT benefits from combining visual and auditory signals could not be explained by invoking the principle of inverse effectiveness. If inverse effectiveness could not explain our results, then what other mechanism (s) could explain the patterns of reaction time benefits? Monkey performance at intermediate SNRs (where the maximal benefits were observed; Figures 3A, B), the visual-only and auditory-only reaction time values were similar to each other. Similarly, for humans at intermediate to large SNRs (where maximal benefits were observed for humans), the visual-only and auditory-only reaction time values were similar to one another. This suggests a simple timing principle: the closer the visual-only and auditory-only RTs are to one another, the greater is the multisensory benefit. A similar behavioral result has been previously observed in the literature, albeit with simpler stimuli, and a mechanism explaining this behavior was (somewhat confusingly) dubbed “physiological synchrony” [51], [73]. According to this mechanism, developed in a psychophysical framework, performance benefits for the multisensory condition are modulated by the degree of overlap between the theoretical neural activity patterns (response magnitude and latency) elicited by the two unisensory stimuli [51], [73]. Maximal benefits occur during “synchrony” of these activity patterns; that is, when the latencies overlap. To put it another way, maximal RT benefits will occur when the visual and auditory inputs arrive almost at the same time. To test this idea, we transformed the benefit curves shown in Figures 5A-C by plotting the benefits as a function of the absolute value of the difference between visual-only and auditory-only RTs. That is, instead of plotting the benefits as a function of SNR (as in Figures 5A–C), we plotted them as a function of the difference between the visual-only and auditory-only RTs for each SNR. If our intuition is correct, then the closer the auditory- and visual-only RTs are (i. e. , the smaller the difference between them), then the greater would be the benefit. Figure 6A plots the benefit in reaction time as a function of the absolute difference between visual- and auditory-only RT for monkeys 1 & 2. The corresponding plot for humans is shown in Figure 6B. By and large, as the difference between RTs increase, the benefit for the audiovisual condition decreases with the minimum benefit occurring when visual- and auditory-only RTs differ by more than 100 to 200 milliseconds. Thus, physiological synchrony can serve as a homologous mechanism for the integration of faces and voices in both monkeys and humans. Although the original formulation of the principle suggested “synchrony”, it seemed too restrictive. The reaction time data—at least for integrating faces and voices—suggest that there is a range of reaction time differences over which multisensory benefits can be achieved. That is, there is a “window of integration” within which multisensory benefits emerge. We use the term “window of integration” as typically defined in studies of multisensory integration: It is the time span within which auditory and visual response latencies must fall so that their combination leads to behavioral or physiological changes significantly different from responses to unimodal stimuli. Such windows have been demonstrated in physiological [49], [76] as well as in psychophysical studies of multisensory integration[48], [77]. To explore the extent of this “window of integration”, we elaborated upon the analysis shown in Figures 6A and B to the whole dataset of sessions and SNRs. For all the sessions and SNRs (48 sessions and 5 SNRs for 2 monkeys), we computed a metric that was the difference between the mean visual-only and auditory-only RTs. This gave us 480 values where there was a difference between visual and auditory RTs and, corresponding to this value, the benefit for the audiovisual condition. After sorting and binning these values, we then plotted the audiovisual benefit as a function of the difference between the mean visual-only and auditory-only RTs. Figure 6C shows this analysis for monkeys. Only in an intermediate range, where differences between unisensory RTs are around 100 – 200 ms, is the audiovisual benefit non-zero—with a maximal benefit occurring at approximately 0 ms. In addition, this window is not symmetrical around zero. It is 200 ms long when visual RTs are faster than auditory RTs and around 100 ms long when auditory-only RTs are faster than visual-only RTs. We repeated the same analysis for humans and the results are plotted in Figure 6D. For humans, a similar window exists: when visual reaction times are faster than auditory reaction times then the window is approximately 160 ms long. We could not determine the extent of the window because, in humans, auditory RTs were never faster than visual RTs. To summarize, combining visual and auditory cues leads to a speedup in the detection of audiovisual vocalizations relative to the auditory-only and visual-only vocalizations. Our analysis of the patterns of benefit for the audiovisual condition reveals that maximal benefits do not follow a principle of inverse effectiveness. However, the principle of physiological synchrony that incorporates a time window of integration provided a better explanation of these results. The principle of physiological synchrony with a time window of integration provides an insight into the processes that lead to the integration of auditory and visual components of communication signals. The issue however is that although this insight can be used to predict behavior, it does not have any immediate mechanistic basis. We therefore sought a computational model that could plausible represent the neural basis for these behavioral patterns. We specified two criteria for the model based on our results. First, audiovisual RTs should be faster than auditory- and visual-only RTs. Second, it should be consistent with, and perhaps subsume, the principle of physiological synchrony with a time window of integration—benefits accrued by combining visual and auditory cues should occur when the visual- and auditory-only RTs are almost equal to one another. If these two criteria are validated, then the model would be a straightforward homologous mechanism. Superposition models are one class of integration models that could incorporate our criteria [53], [63], [64]. According to these models, activation from different sensory channels is linearly combined until it reaches a criterion/threshold and thereby triggers a response. We will use a model formulation based on counters for simplicity [63]. According to this counter model, the onset of a stimulus would lead to a sequence of events occurring randomly over time. Let N (t) denote the number of events that have occurred by time t after stimulus presentation. After the number of counts reaches a criterion, c, it triggers a response. Let us assume that there are separate counters for visual and auditory conditions, NV (t) and NA (t). During the audiovisual condition, a composite counter, NAV (t) = NA (t) + NV (t), comprised of both the visual and auditory signals, counts to the criterion, c (Figure 7A). This composite, multisensory counter would reach the criterion faster than either of the unisensory counters alone. Figure 7B shows that a computer simulation of a counter composed of superposed activity from both visual and auditory cues would reach criterion faster than the unisensory ones alone. Using the RT data from Monkey 1, we set the parameters of the superposition model for the auditory- and visual-only RTs and then used the model to estimate the audiovisual RTs (Figure 7C). From this, the model produced audiovisual RTs that were faster than both the auditory-only and visual-only RTs—like the pattern of results we observed for monkeys (Figures 3A, B). As Figure 7C shows, except for the lowest SNR, there is a good one to one correspondence between the model' s prediction of audiovisual RTs and the actual raw data. Thus, this model can at least generate the patterns of reaction times observed in response to audiovisual vocalization. We next estimated the benefits in RT for the audiovisual condition relative to the visual-only and auditory-only condition from the simulated model (Figure 7D). The benefit curve has the same inverted U-shaped profile as the real patterns of benefit shown in Figure 5A. We repeated this analysis for the human RTs and the pattern of results is shown in Figure 7E–F. Figure 7E shows the predicted reaction time of the average participant as a function of SNR along with actual data. The predicted reaction times are very similar to the actual RTs observed for humans in Figure 3D. As with the monkey behavioral data, the fits performed worst for the softest SNR. Like the benefit patterns shown in Figure 5C, the benefit for the AV condition increases as SNR increases (Figure 7F). This replication by the model of the pattern of monkey and human data—faster audiovisual RTs and maximal benefit when visual and auditory RTs are well matched—suggests that a superposition model is a viable homologous mechanism. Monkeys and humans share many homologous mechanisms for the production of vocalizations [22]. In humans, these production mechanisms deform the face in such a manner that facial motion enhances the detection and discrimination of vocal sounds by receivers [6], [7], . Often this enhanced behavior takes the form of decreased reaction times to audiovisual versus unisensory presentations of speech [8], [9]. While nonhuman primates could theoretically use the same or very similar facial motion to enhance their auditory perception, there has been no evidence of this to date. Several studies demonstrated that, like human infants, monkeys and apes can match facial expressions to vocal expressions [26], [27], [28], [29], and that eye movement patterns generated by viewing vocalizing conspecifics is similar between monkeys and humans [30], [31], [32]. None of these nonhuman primate studies, however, demonstrated a behavioral advantage for perceiving audiovisual vocalizations over unisensory expressions. Demonstration of such an advantage is necessary to invoke the hypothesis that a multisensory integration mechanism for communication signals is homologous across species. In the current study, we provide the first demonstration that monkeys exhibit a behavioral advantage for audiovisual versus unisensory presentations of vocalizations. The patterns of both accuracy and reaction time benefits were similar to humans performing an identical task. Although we have emphasized throughout the similarities in the patterns of behavior for monkeys and humans, it is important to note that there were also differences. The most important difference was that humans were consistently faster for the visual-only vocalization compared to the auditory-only vocalization across the range of auditory intensities. Monkeys, on the other hand, responded faster to some auditory-only conditions versus visual-only conditions across the range of intensities. These differences ultimately led to differences in the amount of integration. Such differences could potentially arise due to the differences in auditory stimuli (/u/ sounds in humans vs coo calls in monkeys) or the amount of attentional engagement. We have suggested acoustic equivalence of “coos” and /u/ vocalizations, but they are not communicatively equivalent. Coos are common vocalizations in monkeys with behavioral significance including a positive emotional valence. In contrast, the /u/ sound we used with humans does not have any behavioral significance. With regard to engagement, we trained our monkeys using standard operant conditioning techniques. This meant the use of timeouts as negative reinforcement whenever the monkeys made false alarms. As a result, when compared to human performance, monkeys may adopt a more conservative criterion for the detection of these sounds to avoid false alarms. Despite these caveats, it is worth emphasizing that positing a linear superposition of visual and auditory signals reconciled these dissimilar results from the two species. Two other design features of our study are worth pointing out before we discuss the broader implications of our results. First, we used a fixed delay between the mouth motion and the onset of the vocalization. Under natural conditions, delays between onset of mouth opening and sound onset, which we term time-to-voice (TTV), are wide ranging and can vary from utterance to utterance and speaker to speaker [5]. At the neural level, different TTVs modulate the degree of integration in local field potential signals recorded from the upper bank of the superior temporal sulcus of monkeys [40]. Thus, how this variable would affect behavioral integration of faces and voices in monkeys and humans is not tested in our experiments or in any other study. A second design feature that we used consisted of the presence of a static face on the screen during the auditory-only vocalization. This face was also identity-incongruent with the auditory vocalization. Thus, both of these features could potentially slow down auditory-only RTs by creating confusion: the face doesn' t move when it should during a vocalization and/or the face doesn' t match the identity of the voice. However, we believe this concern is mitigated by the more naturalistic conditions that our design mimics and more pressing problems that it avoids. Our paradigm is naturalistic in the following sense: faces in noisy, cocktail-party like scenarios do not appear and disappear. Furthermore, monkeys like humans can recognize indexical cues in vocalizations (cues that indicate body size, age, identity, etc) and match them to faces [82], [83]. Thus, in our paradigm, it is not odd to hear one individual' s voice while seeing another individual' s face, a typical occurrence under natural conditions. The key to the face-voice integration is combining motion of the face to the correct, corresponding voice. If we did not present a static face during the auditory-only condition and observed an audiovisual benefit, then the benefits could be attributed to differences in overall attention or arousal (a frequent criticism of physiological studies of AV integration). Moreover, if we adopted a design where audiovisual vocalizations involved the sudden onset of a face followed by its mouth motion, then any RT benefits for audiovisual compared to auditory-only vocalizations would be uninterpretable: we could not be sure if it was due to the integration of facial motion with the sound or from the integration of the sound with the sudden onset of the face. Whatever influences our design may actually have on our participants' RTs; we can model the outcome of hypothetically faster RTs that may arise with a study design that did not use a static, incongruent face in the auditory-only conditions. Since our data demonstrate that the principle of physiological synchrony with a time window of integration, we can actually perform a thought experiment to see what would happen if our auditory RTs are sped up. Simply put, the result would be that the point at which visual and auditory RT curves cross will be at a different SNR and this point of crossing would be the new point of maximal integration. Figure S5 shows that if we sped up all auditory RTs by 40,80 and 120 ms in the model relative to the original data, the point of maximal integration shifts to lower SNRs. We demonstrated that combining visual mouth motion with auditory vocalizations speeds up reaction times in monkeys and humans. Faster reaction times to multisensory signals compared to unisensory signals are a frequent outcome in human psychophysical studies [51], [57], [58], [59], [84], [85], [86], [87]. The first such demonstration, nearly a hundred years ago, showed that there was a speedup in responses for bi- and tri-modal stimuli compared to unimodal stimuli [87]. Since then, this seminal result has been replicated in a variety of settings almost always with the use of simple stimuli [51], [57], [85], [88], [89], [90], [91]. In particular, shortened reaction times are observed in response to multimodal stimuli using both saccades and lever presses as dependent measures [85], [92]. Physiologically, there are similar results. Neurons in the superior colliculus of anaesthetized cats respond faster to audiovisual compared to auditory and visual stimuli [93]. Our results confirm that similar behavioral advantages exist when combining the visual and auditory components of complex social signals encountered in everyday settings. While there are certainly similarities in the integration processes for simple and complex signals like speech, there are also differences. An important issue which has been repeatedly demonstrated is that there are differences in the window of integration for simple versus complex stimuli[94]. For the integration of simple stimuli, tolerance of asynchrony between visual and auditory cues is very small leading to a narrow window of integration [94]. In contrast, for speech stimuli, observers are able to tolerate very large asynchronies and still bind them into a common percept[47]. We return to this issue later in the Discussion. For both monkeys and humans, we found that the maximal benefit obtained by combining visual and auditory cues was for intermediate values of SNR. This is at odds with the principle of inverse effectiveness [49], [50]. This idea was originally formulated in the context of electrophysiological experiments and suggests that the maximal benefit (greater proportional response magnitude) from multisensory stimulus inputs would be achieved by combining visual and auditory cues that, individually, elicit weak responses. Support for the inverse effectiveness rule is also evident at the behavioral level in both monkeys and humans in detection tasks involving simple stimuli [85], [92], [95], [96]. If this principle held true for detecting vocalizations, then we would have observed maximal reaction time savings for the lowest SNR, with the benefit decreasing with increasing SNR. On the contrary, monkeys' detection of vocalizations generated a non-monotonic curve with peak multisensory benefits occurring at intermediate SNRs. For humans, the multisensory benefit increased with increasing SNRs. Thus, for the multisensory integration of vocalizations (with reaction times as a behavioral measure), neither in monkeys nor in humans does the principle of inverse effectiveness explain the behavior. Other results from the speech processing literature support our assertion. For example, in studies of speech intelligibility, maximal benefits gained by integration of auditory speech with visual speech are found when the auditory speech is presented in an intermediate, versus high, level of noise [7], [81]. Similarly, the McGurk effect occurs even under clear listening conditions (i. e. , noisy signals aren' t required to generate the illusory percept) [10], and vision can boost the comprehension of extended auditory passages even under excellent listening conditions [97]. As mentioned before, there are several studies which claim to support this principle in behavior [75], [85], [92], [95], [96], [98], [99], [100], so why do we not see support for the principle of inverse effectiveness in our data or in other studies [51], [57], [88], [92], [101]? We think that this principle is sensitive to the way multisensory stimuli are parameterized and tested in different experiments. In particular, the choice of stimuli, levels of intensity and the pairing of stimuli could all affect whether this principle will be apparent in the resultant data. To illustrate what we mean, we tested two hypothetical scenarios, where inverse effectiveness can be observed using RTs and compare it to a scenario resembling our experimental data. For each scenario, we constructed auditory and visual RTs to have a certain profile with respect to different intensity levels. Then, given that the superposition model is an excellent explanation of our RT data, as well as RTs to simple stimuli[53], [62], [63], [64], [86], we used it to simulate the expected audiovisual reaction times for these same intensity levels. We then examined if the multisensory benefits were consistent with the principle of inverse effectiveness or not. The first scenario is a case wherein RTs to both senses increase with decreases in intensity level, but at every intensity level, they are still roughly equal to one another (Figure S6A). In this scenario, RTs to visual and auditory stimuli increase with decreasing intensity and visual and auditory RTs are largely similar at every intensity level. Keeping with multisensory integration, audiovisual RTs are faster than both auditory-only and visual-only RTs. Critically, in line with our intuition, the multisensory benefit increases with the decrease in SNR — and is thus consistent with the principle of inverse effectiveness (Figure S6B). We can also outline a second scenario where this principle would be observed to be in action. This is the case when the stimuli are such that the RT of one modality approaches the RT of the other modality only for the lowest intensity levels. Figure S6C shows a simulation of this scenario. The auditory-only RTs are much faster than the visual-only RTs for the highest intensity levels. However, as the stimulus intensity decreases, the auditory- and visual-only RTs approach each other. Again, audiovisual RTs are faster than auditory- and visual–only RTs. Like the previous scenario, as intensity decreases, the benefit increases and is thus consistent with the principle of inverse effectiveness (Figure S6D). A recent study showing support for inverse effectiveness had visual and auditory RTs closely following this scenario [99]. The third scenario is one that is a simulation of our data (Figure S6E). In this case, visual RTs do not change much with intensity level, but auditory RTs increase with a decrease in intensity. Audiovisual RTs are again faster than auditory and visual-only RTs. Critically, these data result in a pattern of benefits that is non-monotonic and takes the form of an inverted U; it is not consistent with the principle of inverse effectiveness (Figure S6F). In summary, given that the superposition model is an excellent fit to data, simulations of this model using the scenarios above suggest that observing the principle of inverse effectiveness in behavior is to some extent dependent upon the way the parameters of the stimuli that are used in an experiment. Different multisensory stimuli (speech versus non-speech) as well as the choice of intensity levels are bound to have different effects on multisensory benefit. Thus, the principle of inverse effectiveness may be operational only under some situations. We would however note that, this framework of superposition only explains the inconsistencies about inverse effectiveness in RT output. A similar careful analysis is needed to explain accuracy of subjects as well as performance in tasks such as localization [75], [98]. We showed that maximal benefits from integration of visual and auditory components of communication signals occurred when the reaction times to visual and auditory cues are themselves very similar to one another. This is consistent with the idea of “physiological synchrony”, a principle proposed to explain behavioral data. The principle of physiological synchrony was first formulated based on psychophysical experiments using punctate, simple stimuli [51], [73]. In these experiments, it was noted that maximal multisensory benefits occurred when the stimulus-onset asynchrony between visual and auditory stimuli was adjusted to be equal to the difference between visual-only and auditory-only RTs. That is, “synchrony” was defined by theoretical neurophysiological activity (with reaction times as a proxy) rather than physical synchrony defined by the stimulus-onset asynchrony. According to this idea, performance benefits for the multisensory condition are modulated by the degree of temporal overlap between the theoretical neurophysiological activity patterns elicited by the two unisensory stimuli [51], [73]. Maximal benefits occur during synchrony of these neural activity patterns; that is, when their latencies over-lap. It is worth repeating that this notion of physiological synchrony is a behavioral construct derived by considering RTs. RTs are a simple but powerful metric for indexing this behavior. However, they are the output of a complex mixture of sensory processing, motor preparation, temporal expectation, attention and other cognitive processes. Thus, the physiological synchrony mechanism, although it explains patterns of behavior using RTs to sensory stimuli does not necessarily predict that the integration is occurring in a purely sensory circuit. The neural locus where integration is taking place is not known. Sensory, premotor and/or motor circuits involved in multisensory processing are very likely all involved in generating behavioral responses during this task. We found that there was a time window within which differences in reaction times between visual and auditory signals could lead to integration. This notion of a “temporal window of integration” is a recurring concept in behavioral and neurophysiological experiments of multisensory integration [48], [76], [77], [84], [89], [102], [103]. For example, participants perceive the McGurk effect when the stimulus-onset asynchrony between visual and auditory cues is in a window approximately 400 milliseconds wide, beyond which the illusion disappears [48]. Similarly, studies of orienting responses to audiovisual stimuli using saccades show that speedup of saccadic RTs occur in a variety of experimental settings within a time window of 150–250 ms [77], [84], [89], [104], [105]. Finally, neurophysiologically, maximal integration in multisensory neural responses in the superior colliculus is observed when the stimulus onset asynchrony is adjusted such that the discharge patterns to visual and auditory signals themselves overlap with each other [76]. We showed that a simple computational model of integration—a linear superposition model—explained the behavioral patterns observed for the integration of audiovisual vocalizations by monkeys and humans. The main tenet of this model is that the information from the two unisensory channels is integrated at a specific processing stage by the linear summation of channel-specific activity patterns. Superposition models have been successfully used to predict the reaction times of observers in other multisensory detection tasks, albeit with much simpler stimuli [53], [62], [63], [64], [86]. Physiologically, support for this principle was suggested in studies of the sensitivity of multisensory neurons in superior colliculus [76]. Our results suggest that this model can be readily extended to the integration of visual and auditory components of vocalizations, at least during behaviors involving speeded detection. Indeed, invoking this mechanism reconciled the observed dissimilarity in RTs from monkeys and humans. In addition, it automatically subsumes the principle of physiological synchrony and generates appropriately asymmetric time windows of integration. Whether this model works well for other tasks such as multisensory spatial orientation [75], [98], is an open question. Nevertheless, for the task presented in this study, i. e. the detection of vocalizations in noise, it is a parsimonious homologous mechanism. That a linear, additive model could provide a good explanation for the detection of audiovisual vocalizations might seem irreconcilable with typical notions of multisensory integration that emphasize “super-additivity” or non-linear responses [49], [50]. Recent studies, however, report that multisensory neurons can integrate their inputs in an additive manner both in terms of spiking activity [See for e. g. 50,52,106], as well at the level of synaptic input [107]. Our emphasis on the superposition model as a homologous mechanism has another important implication. First, there are a remarkable number of nodes on which visual and auditory inputs that are sensitive to faces and voices, respectively, could converge. Any or all of these sites could be responsible for the behavioral advantage we report here. For example, neurons in the amygdala and association areas such as the upper bank of STS and prefrontal cortex respond to both the visual and auditory components of vocalizations. In some cases, we know that they integrate these vocalization-related cues [40], [42], [43], [108], [109]—at least during the passive reception of these signals. For example, in keeping with the linear superposition model we posited here, approximately 7% of ventrolateral prefrontal cortical neurons integrate visual and auditory components of vocalizations linearly [43]. The superposition model subsumes the time window of integration. The basis of superposition models is that they require activity patterns to overlap with one another and add together to generate benefits. Thus, activity patterns that overlap with one another have a higher probability of leading to integration, whereas activity patterns that do not overlap will not lead to integration. This implies that the measured window of integration is going to depend on the inherent statistics of the visual and auditory signals and the response profiles to the two signals in some neural structure on which they converge. The narrowness and the latency of these response profiles will thus determine the window of integration. Thus, in any given experiment, choices of the strength and duration of these visual and auditory signals would automatically result in corresponding changes in latencies and response profiles. A flash is highly likely to be processed in primary visual cortex and a moving face through a combination of face- and motion-sensitive neural structures. A similar argument can be made for auditory stimuli. Thus, unless the response profile (s) in some integrative structure (s) mediating detection of these various stimuli are identical, the windows of integration are bound to be different for simple stimuli such as flashes and tone pips versus more complex audiovisual vocalizations and speech signals. This might be a partial explanation for one of the best known findings in the multisensory literature — asymmetric broad windows for speech [47], [48], versus the small windows for simple stimuli [94]. Finally, the superposition model is similar in many respects to a Bayesian model of bimodal integration. For example, in models developed by Ernst and colleagues [110], [111], maximal benefit due to bimodal discrimination occurs when the difficulty of each modality is roughly equated [112]. This is remarkably similar to the notion of physiological synchrony. Thus, Bayesian models could, presumably, be adapted to explain the reaction times and would also subsume the time window of integration concept. However, the advantage the superposition model has is that its neurophysiological implementation is immediately apparent. Bayesian models, in contrast, are usually more abstract, and it is unclear what their neural implementation would look like.

The evolution of speech is one of our most fascinating and enduring mysteries—enduring partly because all the critical features of speech (brains, vocal tracts, ancestral speech-like sounds) do not fossilize. Furthermore, it is becoming increasingly clear that speech is, by default, a multimodal phenomenon: we use both faces and voices together to communicate. Thus, understanding the evolution of speech requires a comparative approach using closely-related extant primate species and recognition that vocal communication is audiovisual. Using computer-generated avatar faces, we compared the integration of faces and voices in monkeys and humans performing an identical detection task. Both species responded faster when faces and voices were presented together relative to the face or voice alone. While the details sometimes appeared to differ, the behavior of both species could be well explained by a “superposition” model positing the linear summation of activity patterns in response to visual and auditory components of vocalizations. Other, more popular computational models of multisensory integration failed to explain our data. Thus, the superposition model represents a putative homologous mechanism for integrating faces and voices across primate species.

Abstract Introduction Materials and Methods Results Discussion

neuroethology auditory system behavioral neuroscience psychology social and behavioral sciences psychophysics biology sensory perception sensory systems neuroscience animal cognition

2011

Monkeys and Humans Share a Common Computation for Face/Voice Integration

Live-attenuated strains of simian immunodeficiency virus (SIV) routinely confer apparent sterilizing immunity against pathogenic SIV challenge in rhesus macaques. Understanding the mechanisms of protection by live-attenuated SIV may provide important insights into the immune responses needed for protection against HIV-1. Here we investigated the development of antibodies that are functional against neutralization-resistant SIV challenge strains, and tested the hypothesis that these antibodies are associated with protection. In the absence of detectable neutralizing antibodies, Env-specific antibody-dependent cell-mediated cytotoxicity (ADCC) emerged by three weeks after inoculation with SIVΔnef, increased progressively over time, and was proportional to SIVΔnef replication. Persistent infection with SIVΔnef elicited significantly higher ADCC titers than immunization with a non-persistent SIV strain that is limited to a single cycle of infection. ADCC titers were higher against viruses matched to the vaccine strain in Env, but were measurable against viruses expressing heterologous Env proteins. In two separate experiments, which took advantage of either the strain-specificity or the time-dependent maturation of immunity to overcome complete protection against SIVmac251 challenge, measures of ADCC activity were higher among the SIVΔnef-inoculated macaques that remained uninfected than among those that became infected. These observations show that features of the antibody response elicited by SIVΔnef are consistent with hallmarks of protection by live-attenuated SIV, and reveal an association between Env-specific antibodies that direct ADCC and apparent sterilizing protection by SIVΔnef. The development of a vaccine against HIV-1 continues to be hampered by our limited understanding of the types of immune responses needed for protection. Although safety considerations preclude the use of live-attenuated HIV-1 in people [1]–[5], live-attenuated strains of simian immunodeficiency virus (SIV) afford the most reliable protection achieved to date in non-human primate models, often providing apparent sterilizing immunity against closely related challenge viruses [6]–[9]. Thus, identifying the immune responses that mediate protection by live-attenuated SIV and understanding how to elicit them by vaccination may provide important insights for the development of a safe and effective HIV-1 vaccine [10]. Antibody, T cell, and innate immunity have evolved to operate synergistically as an integrated system [11]–[13], and a combination of these immune responses may be necessary for complete protection by live-attenuated SIV. However, the efficacy of at least one of these immune responses increases over time, since animals challenged with pathogenic SIVmac251 months after inoculation with live-attenuated SIV are protected from infection, whereas animals challenged at early time points become infected [7], [8]. Although live-attenuated SIV elicits virus-specific T cells [14]–[16], and the quality of these T cell responses may change over time, the frequency of virus-specific CD8+ T cells declines after the acute peak of live-attenuated SIV replication [17]. In contrast, antibodies capable of neutralizing virus infectivity develop over time through affinity maturation [18]–[21]. An essential role for the affinity maturation of antibody responses could account for the time-dependent development of protection by live-attenuated SIV [22]. However, SIVmac251 is inherently resistant to neutralization [23], and antibodies capable of neutralizing this challenge virus are typically undetectable among completely protected animals [8], [9]. We therefore reasoned that functions of antibodies other than neutralization may contribute to protection by live-attenuated SIV. In addition to virus neutralization, the antiviral functions of antibodies include complement fixation and numerous consequences of Fc receptor crosslinking, such as ADCC [13], [24]–[29]. Since ADCC represents a potential effector mechanism and a proxy for other activities of the same antibodies, we developed a novel assay for quantifying the ability of antibodies to direct ADCC. This assay measures ADCC against virus-infected target cells expressing native conformations of the viral envelope glycoprotein (Env), and is therefore more physiologically relevant than methods based on coating target cells with recombinant gp120, gp140, or peptides [30]–[37]. We used this assay to investigate the induction of antibodies with ADCC activity, and to test the hypothesis that higher ADCC activity against cells infected by the challenge virus is associated with protection. Our results indicate that persistent infection with SIVΔnef elicits Env-specific ADCC titers that develop over time, are cross-reactive with Env proteins expressed by heterologous SIV strains, are proportional to vaccine strain replication, and are higher among animals protected against SIVmac251 infection. Plasma samples collected at longitudinal time points after inoculation with SIVmac239Δnef were tested for their ability to neutralize SIVmac239 and to direct ADCC against SIVmac239-infected cells. Only four of ten macaques developed neutralizing antibody titers, and these were not detectable until thirteen weeks after inoculation with SIVmac239Δnef (Figure 1A). In contrast, ADCC titers were detectable in all animals just three weeks after inoculation with SIVmac239Δnef (Figure 1B). These ADCC titers were Env-specific, since none of the plasma samples had detectable ADCC activity against target cells infected with SHIVSF162P3, which expresses the Env protein of HIV-1SF162. To quantify ADCC titers, we calculated the plasma dilution that reduces the luciferase signal from virus-infected cells by 50%, and to measure differences in the extent of target cell elimination over all dilutions tested, we calculated values for the area under the curve (AUC). By both measures, progressive increases in ADCC were observed over 21 weeks. Thus, antibody titers capable of directing ADCC against SIVmac239-infected cells increased over time, but unlike neutralizing antibodies, emerged early and were detectable in all animals. Neutralizing antibodies were only detectable in the plasma samples with high ADCC titers. A 50% ADCC titer of approximately 104 emerged as a threshold, below which neutralization of SIVmac239 was not detectable (Figure 1C). Among all of the plasma samples collected after inoculation with SIVmac239Δnef, the odds of detecting neutralization were 1,966-fold higher per log10 increase in 50% ADCC titer (95% CI = 1. 8 to 2,192,451, P = 0. 034), as estimated by logistic regression (Figure 1C). Likewise, there was a trend towards a higher probability of detecting neutralization by plasma with higher AUC values for ADCC (odds ratio: 67,788-fold higher per 1 AUC unit, 95% CI = 0. 835–5. 5×109, P = 0. 054) (Figure 1D). ADCC titers therefore predicted and were correlated with neutralization. The contribution of ongoing vaccine strain replication to the development of antibody responses was evaluated by comparing ADCC in animals immunized with SIVmac239Δnef to ADCC in animals immunized with an SIV strain that is limited to a single cycle of infection. Plasma samples collected two or twelve weeks after a series of inoculations with single-cycle SIV [38] were tested for ADCC against SIVmac239-infected cells (Figure 2A). Since the geometric mean peak viral RNA loads in plasma for SIVmac239Δnef and single-cycle SIV were within two-fold of each other, 1. 3×105 and 7. 4×104 copies per ml respectively (Figure 2B), differences in antibody responses relate to differences in the persistence of SIVmac239Δnef versus single-cycle SIV. Five weeks after inoculation with SIVmac239Δnef, median 50% ADCC titers were 51-fold higher than those elicited by single-cycle SIV, and this difference expanded to 233-fold by week 21 (Figure 2C). The 50% ADCC titers (Figure 2C) and the AUC values for ADCC (Figure 2D) at any time point after inoculation with SIVmac239Δnef were significantly higher than at either time point after inoculation with single-cycle SIV (2-tailed Mann-Whitney U tests, P = 0. 0062 to P<0. 0001). Thus, in contrast to persistent infection with SIVmac239Δnef, repeated stimulation of antibody responses with SIV limited to a single cycle of infection did not elicit high ADCC titers. ADCC titers measured after immunization with SIVmac239Δnef and single-cycle SIV were compared with antibody titers that bind recombinant forms of SIVmac239 Env in enzyme-linked immunoadsorbent assays (ELISAs). Relationships among these measures of Env-specific antibody responses were evaluated by calculating Spearman correlation coefficients (RS). ELISA titers against SIVmac239 gp120 correlated with 50% ADCC titers against SIVmac239-infected cells (RS = 0. 8190, P<0. 0001) (Figure 2E), and with AUC values for ADCC (RS = 0. 7809, P<0. 0001) (Figure 2F). Although a linear relationship was observed between ADCC and gp120-binding titers in the animals persistently infected with SIVmac239Δnef, plasma samples from the animals immunized with single-cycle SIV belonged to an out-group, which was displaced towards higher gp120-binding titers, relative to ADCC (Figure 2E and F). ELISA titers against gp140 also correlated with 50% ADCC titers (RS = 0. 9015, P<0. 0001) (Figure 2G), and with AUC values for ADCC (RS = 0. 8836, P<0. 0001) (Figure 2H). However, any displacement of the single-cycle SIV-immunized animals towards higher gp140-binding, relative to ADCC titers, was more subtle than that observed for gp120-binding titers (Figure 2E–H). This may reflect the occlusion of surfaces in gp140 oligomers that are exposed on gp120 monomers [39]. In comparison to persistent infection with SIVmac239Δnef, immunization with single-cycle SIV may therefore stimulate a higher proportion of gp120-specific antibodies with low or undetectable ADCC activity against virus-infected cells, due to recognition of epitopes that are occluded in the native Env trimer. The ADCC activity against SIV strains that were matched or mismatched with the vaccine strain in Env was compared. Sera were collected from twelve macaques inoculated with SIVmac239Δnef (Figure 3A), and twelve inoculated with a recombinant form of SIVmac239Δnef containing the env gene of SIVsmE543-3 [40], designated SIVmac239Δnef/E543-3env (Figure 3B). Sera from all 24 animals were tested for ADCC activity against target cells infected with SIVmac239 or SIVmac239/E543-3env. On average, the 50% ADCC titers were seven-fold higher when the vaccine and test viruses were matched in Env than when they were mismatched (2-tailed Wilcoxon matched pairs test, P<0. 0001). The 50% ADCC titers were also approximately seven-fold higher at week 22 than at week 6 (2-tailed Wilcoxon matched pairs test, P<0. 0001). Thus, the 50% ADCC titers against the Env-matched virus at week 6 and the Env-mismatched virus at week 22 were comparable. Therefore, ADCC titers against Env-mismatched viruses were lower and required more time to develop than ADCC titers against Env-matched viruses. The extent of vaccine strain replication was estimated by calculating AUC values for log10-transformed SIVΔnef viral RNA loads in plasma over the first 21 or 22 weeks after inoculation. AUC values for viral loads among animals inoculated with SIVmac239Δnef and SIVmac239Δnef/E543-3env were similar, averaging 65 and 67 log10-transformed RNA copies per ml×weeks, respectively. The extent of vaccine strain replication by the end of this time period correlated with 50% ADCC titers against Env-matched (RS = 0. 68, P<0. 0001) and Env-mismatched (RS = 0. 55, P = 0. 006) viruses (Figure 4A), and also with AUC values for ADCC against Env-matched (RS = 0. 64, P<0. 0001) and Env-mismatched (RS = 0. 42, P = 0. 0421) viruses (Figure 4B). These relationships suggest that the development of antibodies that direct ADCC is driven by the extent of antigenic stimulation provided by vaccine strain replication. Twelve animals were challenged intravenously with SIVmac251NE 46 weeks after inoculation with SIVΔnef. The SIVΔnef strain in six of these twelve animals was SIVmac239Δnef, whereas the other six were inoculated with SIVmac239Δnef/E543-3env. All twelve of these animals resisted two intravenous challenges with SIVmac239 on weeks 22 and 33, while three naïve control animals challenged on week 22 and two challenged on week 33 all became infected. When the twelve SIVΔnef-immunized animals were subsequently re-challenged with SIVmac251NE on week 46, three became infected, as did both naïve control animals challenged at the same time. Although all three immunized animals that became infected were among those inoculated with SIVmac239Δnef/E543-3env, the trend toward more infections in this group was not significant (2-tailed Fisher' s exact test, P = 0. 18). Comparisons of SIVΔnef viral loads for animals that became infected versus remained uninfected, or animals immunized with SIVmac239Δnef versus SIVmac239Δnef/E543-3env, did not reveal significant differences (Figure S1A–D). Likewise, sera collected on the day of challenge from the animals that remained uninfected by SIVmac251NE did not have significantly higher binding titers against gp120 (2-tailed Mann-Whitney U test, P = 0. 2091) (Figure S1E), or gp140 (2-tailed Mann-Whitney U test, P = 0. 3727) (Figure S1F), in comparison to the animals that became infected. Sera drawn the day of intravenous challenge with SIVmac251NE were tested for neutralization of SIVmac251NE and for ADCC against SIVmac251NE-infected cells. Neutralizing antibody titers were low to undetectable (Figure 5A), and differences among the infected versus uninfected animals were not significant at the highest serum concentration tested, a 1∶8 dilution (2-tailed Mann-Whitney U test, P = 0. 3727). However, all of these samples had measureable ADCC activity (Figure 5B). Whereas differences in 50% ADCC titers were not significant (Figure 5C), the animals that remained uninfected by SIVmac251NE had higher AUC values for ADCC than those that became infected (2-tailed Mann-Whitney U test, P = 0. 0091) (Figure 5D). Thus, more complete elimination of the SIVmac251NE-infected target cells, as measured by AUC values for ADCC, was associated with protection against infection by intravenous challenge with SIVmac251NE. To address the temporal association between the development of antibody responses and protective immunity, we studied animals that were challenged at different time points after inoculation with SIVmac239Δnef. Groups of six female macaques were challenged by high-dose vaginal inoculation with SIVmac251UCD at weeks five, twenty, or forty after immunization with SIVmac239Δnef (Reeves et al. , manuscript in preparation). All six animals challenged at week five became infected, as did three of six animals challenged at week twenty, and four of six animals challenged at week forty. Three naïve control animals challenged at each time point all became infected, except one animal challenged at week twenty. Peak SIVmac239Δnef viral loads were unrelated to the outcome of challenge (Figure S2A). Although the total extent of vaccine strain replication, as estimated from AUC values for SIVmac239Δnef viral loads, tended to be higher among the animals that remained uninfected compared to those that became infected, this trend was not significant (2-tailed Mann-Whitney U test, P = 0. 0939) (Figure S2B). The animals that remained uninfected by SIVmac251UCD also did not have significantly higher binding antibody titers than those that became infected, as measured by ELISA against gp120 (2-tailed Mann-Whitney U test, P = 0. 4304) (Figure S2C) or gp140 (2-tailed Mann-Whitney U test, P = 0. 1148) (Figure S2D). Sera collected on the day of challenge with SIVmac251UCD were evaluated for neutralization of SIVmac251UCD (Figure 6A–C). However, neutralization of SIVmac251UCD was not detectable for any of these serum samples (Figure 6A–C). The absence of detectable neutralizing antibodies is a consequence of the inherent resistance of SIVmac251UCD to neutralization (Figure S3). Indeed, SIVmac251UCD is even more resistant to neutralization than SIVmac251NE by plasma from animals chronically infected with SIVmac239 (2-tailed Wilcoxon matched pairs test, P = 0. 0098) (Figure S3A), and by soluble CD4 (Figure S3B). The relative resistance of SIVmac251UCD to antibody may have been a factor in the small number of animals that were protected against infection with this challenge strain. The capacity of the same sera to direct ADCC against SIVmac251UCD-infected cells was evaluated (Figure 6D–F). In contrast to neutralization, all had measurable ADCC activity (Figure 6D–F). Statistically significant outcomes could not be reached at individual time points, or for a group of animals that combines just those challenged on weeks twenty and forty (Table S1). However, when the animals challenged five, twenty and forty weeks after inoculation with SIVmac239Δnef were analyzed together, those that remained uninfected had higher 50% ADCC titers on the day they were challenged than those that became infected (2-tailed Mann-Whitney U test, P = 0. 0487) (Figure 6G). A similar trend was observed for AUC values for ADCC, although these differences were not significant (Figure 6H). Also, among the animals that became infected, there was a trend towards higher ADCC activity in animals with lower peak SIVmac251UCD viral loads (Figure 6I and J). Therefore, higher 50% ADCC titers present at later challenge time points after inoculation with SIVmac239Δnef were associated with protection against infection by high-dose vaginal challenge with SIVmac251UCD. Identifying the immune responses that mediate protection by live-attenuated SIV and understanding their induction may inspire strategies for engineering a safe and effective vaccine against HIV-1. We hypothesized that antibody functions other than neutralization contribute to the protective immunity provided by live-attenuated SIV against wild-type pathogenic SIV challenge. Here we demonstrate that properties of the antibody response reflected in ADCC titers mirror hallmarks of protection by live-attenuated SIV. The protective immunity conferred by live-attenuated SIV increases over time [7], [8], is usually incomplete against heterologous challenge [41], [42], and is greater for vaccine strains that replicate at higher levels [7], [9]. In accordance with these observations, our data indicate that ADCC titers increase progressively over time, are lower against viruses expressing heterologous Env proteins, and are proportional to the extent of vaccine strain replication. Furthermore, in two different challenge experiments, measures of ADCC activity were associated with protection against infection by SIVmac251. In one experiment, macaques inoculated with SIVmac239Δnef or SIVmac239Δnef/E543-3env that remained uninfected after intravenous challenge with SIVmac251NE had higher AUC values for ADCC than those that became infected. In another experiment, animals that remained uninfected after high-dose vaginal challenge with SIVmac251UCD at different time points after inoculation with SIVmac239Δnef had higher 50% ADCC titers than those that became infected. Differences in AUC values for ADCC were significant in one experiment, whereas differences in 50% ADCC titers were significant in the other, perhaps reflecting the limited power to detect differences using small numbers of infected versus uninfected animals. Additional differences between the two studies, including the greater resistance of SIVmac251UCD than SIVmac251NE to antibodies, the greater length of time allowed for the maturation of antibody responses prior to challenge with SIVmac251NE (46 weeks) than SIVmac251UCD (5,20, and 40 weeks), and the effect of mismatches in Env between SIVmac239Δnef/E543-3env and SIVmac251NE, may also have contributed to the detection of differences in AUC values for ADCC against SIVmac251NE. Although differences between the two SIVmac251 challenge experiments may have favored one method of data analysis over the other, measures of ADCC activity were associated with complete protection in both experiments. While the relationship between ADCC activity and the outcome of challenge suggests that these antibodies contribute to protection, correlation does not establish causation. In addition to ADCC, Fc receptor crosslinking stimulates the secretion of molecules that promote lymphocyte homing and activation, and that may inhibit virus replication [26], [27]. The antibodies that direct ADCC may also mediate effector functions through complement fixation [25], [28]. Furthermore, ADCC assays may measure antibodies that block virus infection at concentrations present in vivo, but are undetectable using conventional neutralization assays. Therefore, although ADCC may be an important effector mechanism [43], ADCC could also be a surrogate for other effector mechanisms that contribute to protection. Mechanisms of immunity not mediated by antibodies may also covary with ADCC activity. For instance, T cell, antibody, and innate immune responses may all be affected by the extent of antigenic stimulation. It is conceivable that the observed relationships are due to differences that exist among animals inoculated with SIVmac239Δnef versus SIVmac239Δnef/E543-3env, or among animals challenged five, twenty, and forty weeks after inoculation with SIVmac239Δnef. Thus, while our findings implicate antibodies in protection by live-attenuated SIV, they do not preclude a role for other immune responses. More than one type of immune response elicited by live-attenuated SIV may be necessary for complete protection against SIVmac251 challenge. Passive transfer experiments in different live-attenuated SIV vaccine models have yielded mixed results regarding the ability of antibodies alone to protect against SIV infection, demonstrating complete protection in one study [44], and no protection in another [45]. In contrast to the absence of detectable neutralizing antibodies in the completely protected animals in this study, relatively high concentrations of neutralizing monoclonal antibodies were necessary to protect macaques against SHIV infection in passive transfer experiments [43], [46]–[48]. T cell responses present in macaques inoculated with SIVΔnef [14]–[16], but absent in macaques that received antibodies passively, may help to explain these differences in neutralizing antibody titers required for complete protection. Our observations are in agreement with other reports that have associated antibody responses with protection. The Robert-Guroff laboratory, and others, have associated lower viral loads after infection with higher ADCC activity measured using target cells that were coated with monomeric gp120 [32], [36], [37], recombinant gp140 [33], or infected with T cell line-adapted SIV [49]. However, in these studies, ADCC was not associated with protection from infection, or measured using target cells infected with neutralization-resistant viruses. Nevertheless, antibodies that bound recombinant forms of gp120 by ELISA and that neutralized neutralization-sensitive SIV strains were associated with a reduced rate of infection [37]. Consistent with these observations on vaccine protection, a recent study on mother-to-child transmission of HIV-1 found that the breast milk of mothers whose newborns remained uninfected contained antibodies with higher ADCC activity against gp120-coated target cells [50]. In the context of vaccination with different live-attenuated strains of SIV, antibody avidity was also associated with resistance to infection and lower post-challenge viral loads after vaginal challenge with SIVmac251NE [9]. Likewise, neutralization of SIVmac251NE at a 1∶4 dilution of serum was associated with protection in a combined group of animals that remained uninfected or strongly controlled SIVmac251NE viral loads [7]. Taken together, these studies support a role for antibodies in protective immunity. Interest in antibody functions other than neutralization has recently increased as a result of the RV144 trial, in which a modest reduction in the rate of HIV-1 infection was reported among recipients of a recombinant canarypox vector prime and gp120 protein boost vaccine [51]. Virus-specific CD8+ T cell responses were not measurably different between vaccinated and unvaccinated trial participants. Whereas antibodies capable of neutralizing primary HIV-1 isolates were also undetectable among vaccinated individuals, gp120-binding titers were consistently detectable by ELISA. Functions of antibodies other than neutralization have therefore been postulated to potentially be responsible for protection in the RV144 trial [52]. Among six primary variables in the immune correlates analysis of the RV144 trial, IgG titers to the V2 region of gp120 were associated with protection, whereas Env-specific IgA antibodies were associated with a higher risk of infection [53]. There was also a non-significant trend towards a lower risk of HIV-1 infection among vaccine recipients with higher ADCC activity using the assay described here. This relationship reached borderline statistical significance after excluding subjects with Env-specific IgA in plasma [53]. These observations further support a role for antibodies in vaccine protection against immunodeficiency virus infection. Persistent expression of Env may be essential to elicit protective antibody responses. The progressive increases in ADCC activity over time, and the considerably higher ADCC activity elicited by SIVmac239Δnef versus single-cycle SIV, imply that the persistent antigenic stimulation provided by ongoing SIVmac239Δnef replication is important for the development of high ADCC titers. Differences in the maturation of antibody responses may also contribute to the better protection provided by SIVΔnef in comparison to single-cycle SIV [38]. Furthermore, a longer period of persistent infection with SIVΔnef was required for ADCC titers against SIV strains expressing heterologous Env proteins to reach the levels observed at an earlier time point against the Env-matched strain. Persistent Env expression may therefore be required to elicit antibodies with high and broadly reactive ADCC activity against circulating HIV-1 strains with diverse neutralization-resistant Env proteins. A vaccine against HIV-1 must contend with a degree of sequence variation that typically renders neutralizing sera ineffective against heterologous HIV-1 strains isolated from other people [19], [54]. The Env proteins of SIVmac239 and SIVsmE543-3 differ in amino acid sequence by 18%, which approximates the median difference between the Env proteins of individual HIV-1 isolates within a clade [54], [55]. Therefore, the ADCC titers observed against cells infected with neutralization-resistant Env-mismatched target viruses suggest that antibodies may have broader efficacy against heterologous HIV-1 isolates than is generally revealed by neutralization assays. In summary, we show that properties of the antibody response elicited by SIVΔnef mirror hallmarks of protection by live-attenuated SIV, and that ADCC activity is associated with apparent sterilizing protection against SIVmac251. These observations support a role for antibodies in protection by live-attenuated SIV, despite the paradoxical absence of detectable neutralizing antibody titers against the challenge virus in most fully protected animals. The temporal analyses of ADCC activity against both Env-matched and Env-mismatched viruses, and the significantly higher ADCC titers observed in SIVΔnef-infected animals than in animals repeatedly immunized with single-cycle SIV, suggest that persistent Env expression may be necessary to drive the maturation of high-titer, broadly reactive antibody responses. Therefore, strategies designed to persistently stimulate Env-specific antibodies may significantly improve the efficacy of vaccines against HIV-1. Due to the variability and limited scalability of assays dependent upon primary cells, we engineered a pair of cell lines to serve as targets and effectors in ADCC assays. The target cells were derived from CEM. NKR-CCR5 CD4+ T cells [56], [57] (AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH, contributed by Dr. Alexandra Trkola). These were transduced with a pLNSX-derived retroviral vector to express firefly luciferase under the transcriptional regulation of the SIV LTR promoter [23]. Target cells were maintained in “R10” cell culture media consisting of RPMI (Invitrogen) supplemented with 10% fetal bovine serum (FBS) (Invitrogen), 25 mM HEPES (Invitrogen), 2 mM L-glutamine (Invitrogen), and 0. 1 mg/ml Primocin (InvivoGen). Effector cells were derived from the CD16-negative human NK cell line KHYG-1 [58] (Japan Health Sciences Foundation) by stable transduction with a pQCXIN-derived retroviral vector expressing rhesus macaque CD16 (FCGR3A variant 7) [59]. KHYG-1 effector cells were maintained at a density of 1×105 to 4×105 cells per ml in R10 media supplemented with IL-2 at 10 U per ml (Roche) and cyclosporine A at 1 µg per ml (Sigma). Four days prior to each ADCC assay, the target cells were infected by spinoculation [60]. On the day of the assay, infected target cells were washed three times with R10 immediately before use. ADCC assays were conducted in round-bottom 96-well plates. Each well contained 105 effector cells and 104 target cells. Effector and target cells were incubated together for eight hours in the presence of triplicate serial two-fold dilutions of serum or plasma before the luciferase activity was measured using the luciferase substrate reagent BriteLite Plus (Perkin Elmer). Relative light units (RLU) indicate luciferase expression by infected target cells. Wells containing uninfected target cells plus effector cells defined 0% RLU, and wells containing infected target cells plus effector cells with no serum or plasma defined 100% RLU. The %ADCC activity was defined as 100 minus %RLU. Antibodies from some macaques bound uninfected CEM. NKR-CCR5 cells. To deplete these antibodies, 107 uninfected target cells were resuspended in the sample and incubated for twenty minutes at room temperature. This process was repeated twenty times for the animals immunized with single-cycle SIV, and twelve times for the animals in the Env-mismatch and vaginal challenge studies. Neutralization was measured as previously described [23], [61]. The sensitivity of the virus neutralization assay was maximized by minimizing the amount of virus input required to obtain a consistent level of infected C8166-secreted alkaline phosphatase (SEAP) reporter cells. These amounts were 0. 5 ng p27 SIVmac239,5 ng p27 SIVmac251NE, and 0. 5 ng p27 SIVmac251UCD per well. However, the resistance of SIVmac251NE, SIVmac251UCD, and SIVmac251TCLA to neutralization by sCD4-IgG was evaluated using 5 ng p27 per well for all 3 viruses. Each well contained 15,000 C8166-SEAP cells. Plasma or serum dilutions were pre-incubated with virus for one hour at 37°C before adding C8166-SEAP cells. After three days, SEAP activity was determined using a luminescent assay (Applied Biosystems). Recombinant 6-His tagged SIVmac239 gp120 or gp140 protein (Immune Technology) diluted to 0. 5 µg/ml in 0. 1 M sodium bicarbonate pH 9. 5 was coated onto Maxisorb ELISA plates (NUNC). Plates were blocked using PBS containing 0. 5% TWEEN-20 (Sigma) and 5% blotting-grade non-fat dry milk (NFDM) blocker (BioRad). Antibody dilutions were made in PBS containing 0. 5% TWEEN-20 and 5% NFDM. Bound IgG was detected using a horseradish peroxidase conjugated goat anti-monkey/human IgG antibody (Santa Cruz Biotechnology). The upper limit of a 95% confidence interval calculated using plasma from 10 naïve macaques diluted 1∶100 served as the endpoint titer [62]. The animals were Indian-origin rhesus macaques (Macaca mulatta) housed in a biocontainment facility at the New England Primate Research Center (NEPRC), and given care 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. The animal samples used here were collected under experimental protocols approved by the Harvard Medical Area Standing Committee on Animals, and conducted in accordance to the Guide for the Care and Use of Laboratory Animals [63]. Additional analyses using these animals will be published separately by R. C. Desrosiers and by R. K. Reeves. The intravenous SIVmac239 challenge dose used on week 22 consisted of twenty animal-infectious doses of virus produced by transfection of 293T cells. The week 33 intravenous challenge with SIVmac239 contained ten animal-infectious doses of a rhesus PBMC-derived virus stock used previously [38]. The intravenous SIVmac251NE challenge was ten animal-infectious doses (32 pg p27) of a rhesus PBMC stock prepared in February of 1991, used in other studies [7]–[9]. Vaginal challenges consisted of two inoculations on one day of 1 ml undiluted SIVmac251UCD [64] (100 ng p27), prepared at the California National Primate Research Center in June of 2004. Neutralization and ADCC assays were done using SIVmac239 and SIVmac239/E543-3env produced by transfection of 293T cells, and SIVmac251NE and SIVmac251UCD expanded from the corresponding uncloned challenge stocks in rhesus macaque PBMC. SHIVSF162P3 was also expanded in rhesus PBMC (AIDS Research and Reference Reagent Program, NIAID, NIH, contributed by Drs. Janet Harouse, Cecilia Cheng-Mayer, Ranajit Pal and the DAIDS, NIAID). The rhesus PBMC used to expand virus stocks were depleted of CD8+ cells using Dynal anti-CD8 magnetic beads (Invitrogen), activated with phytohemagglutinin (PHA) (Sigma), and then cultured in IL-2 (Roche). SIVmac251TCLA was grown in MT4 cells. Challenge viruses were detected using primers specific for the nef sequences of SIVmac239 or SIVmac251 within the deletion in SIVmac239Δnef [65]. The primers specific for wild-type SIVmac239 were GAATACTCCATGGAGAAACCCAGC and ATTGCCAATTTGTAACTCATTGTTCTTAG, and the labeled probe had the sequence CTTTTGGCCTCACTGATACCCCTAC. To reflect the polymorphic nature of the uncloned SIVmac251 virus stocks, the primer set designed to amplify SIVmac251 contained degenerate bases P and K, which mimic mixtures of C and T or A and G, respectively (GlenResearch). The SIVmac251-specific primers were GAATACPCCATGGAKAAACCCAGC and TGCCAATTTGTAA (C, T, G) TCATTGPTCTTAGG, and the SIVmac251-specific probe sequence was TAGAPGAGGAAGATGATGACTTGKTAGGG. Complete or apparent sterilizing protection was defined as the absence of detectable wild-type viral RNA from plasma at every post-challenge time point using the above primer/probe sets in a real-time RT-PCR assay with a nominal threshold of detection of 10–30 copies of RNA per ml [66]. Fifty percent titers were calculated as the dilution at which a line connecting the values above and below 50% RLU would intercept the 50% RLU line. AUC values for ADCC were calculated such that they would be proportional to 50% ADCC titers, and represent the areas between 100% RLU and the titration curves as they appear in the figures. Whereas %ADCC, defined as 100% minus %RLU, appears asymptotic as it approaches 100%, minimum %RLU values are inversely proportional to 50% ADCC titers. Therefore, AUC values for ADCC were calculated from values for log10100 minus log10%RLU, which were summed over all dilutions. This sum was multiplied by the log10-transformed dilution factor of two to find an area. The ability of ADCC activity to predict neutralization was evaluated by logistic regression using SPSS (IBM). The statistical significance of other comparisons was evaluated in Prism version 4. 1b (GraphPad Software) using 2-tailed Mann-Whitney U tests, 2-tailed Fisher' s exact tests, 2-tailed Wilcoxon matched pairs tests, and Spearman correlation coefficients.

Live-attenuated vaccines can prevent simian immunodeficiency virus (SIV) infection upon experimental challenge of rhesus macaques. Although safety considerations preclude vaccinating humans with live-attenuated HIV-1, it may be possible to replicate the types of immunity induced by live-attenuated SIV through an alternative approach. Thus, identifying the immune responses underlying protection by live-attenuated SIV and understanding their induction would provide guidance for HIV-1 vaccine design. An important role for the maturation of virus-specific antibody responses could explain the time-dependent development of protection by live-attenuated SIV. However, antibodies that block the entry of the challenge virus into cells are usually undetectable. Antibodies can also direct the killing of virus-infected cells by antibody-dependent cell-mediated cytotoxicity (ADCC). Here we show that live-attenuated SIV induces progressive increases in ADCC over time, and that the development of these antibodies is dependent upon the persistent replication of the vaccine strain. In two different experiments, the animals immunized with live-attenuated SIV that remained uninfected after pathogenic SIV challenge had higher measures of ADCC than those that became infected. Our results suggest that antibodies contribute to protection by live-attenuated SIV, and that persistent stimulation of antibody responses may be essential for HIV-1 vaccines to induce high ADCC activity.

Abstract Introduction Results Discussion Materials and Methods

medicine infectious diseases model organisms immunology biology microbiology

2012

ADCC Develops Over Time during Persistent Infection with Live-Attenuated SIV and Is Associated with Complete Protection against SIVmac251 Challenge

Lysine acetylation has recently emerged as an important post-translational modification in diverse organisms, but relatively little is known about its roles in mammalian development and stem cells. Bromodomain- and PHD finger-containing protein 1 (BRPF1) is a multidomain histone binder and a master activator of three lysine acetyltransferases, MOZ, MORF and HBO1, which are also known as KAT6A, KAT6B and KAT7, respectively. While the MOZ and MORF genes are rearranged in leukemia, the MORF gene is also mutated in prostate and other cancers and in four genetic disorders with intellectual disability. Here we show that forebrain-specific inactivation of the mouse Brpf1 gene causes hypoplasia in the dentate gyrus, including underdevelopment of the suprapyramidal blade and complete loss of the infrapyramidal blade. We trace the developmental origin to compromised Sox2+ neural stem cells and Tbr2+ intermediate neuronal progenitors. We further demonstrate that Brpf1 loss deregulates neuronal migration, cell cycle progression and transcriptional control, thereby causing abnormal morphogenesis of the hippocampus. These results link histone binding and acetylation control to hippocampus development and identify an important epigenetic regulator for patterning the dentate gyrus, a brain structure critical for learning, memory and adult neurogenesis. Lysine acetylation involves covalent addition of an acetyl moiety to the ε-amino group of a lysine residue and is important for modification of both prokaryotic and eukaryotic proteins [1–3]. Proteomic analyses have detected this modification in thousands of mammalian proteins with important roles not only in chromatin-templated nuclear processes but also various cytoplasmic pathways [4–7]. In addition, it is abundant in bacteria [8,9]. Although some of the modification events in bacteria are dependent on acetyl-phosphate [10], this modification is exclusively enzymatic in eukaryotes. In humans, at least 15 known lysine acetyltransferases (KATs) catalyze the forward reaction [2,11–13]. These enzymes are divided into three families, one of which is the MYST family, composed of TIP60 (HIV Tat-interacting protein of 60 kDa), MOZ (monocytic leukemia zinc finger protein), MORF (MOZ-related factor), HBO1 (HAT bound to ORC1) and hMOF (homolog of Drosophila males absent on the first), which are also known as KAT5, KAT6A/MYST3, KAT6B/MYST4, KAT7/MYST2 and KAT8/MYST1, respectively [14–16]. Although mainly referred to as histone acetyltransferases, members of this family also acetylate non-histone substrates, including the tumor suppressor p53 [17–19] and the DNA-damage response regulator DBC1 (deleted in breast cancer 1) [20]. TIP60 and hMOF also carry out autoacetylation essential for their activation [21–25]. Furthermore, four recent studies have revealed that tyrosine phosphorylation of TIP60 links chromatin sensing to ATM signaling and that both TIP60 and hMOF regulate autophagy [26–29]. Thus, this family of acetyltransferases is important in diverse cellular programs. Molecular and cell-based studies have firmly established that three members of this family, MOZ, MORF and HBO1, form tetrameric complexes with BRPF1 (bromodomain- and PHD finger-containing protein 1), along with two other subunits [30–32]. Within the complexes, BRPF1 functions as a scaffold to bridge subunit interaction, stimulate acetyltransferase activity and restrict substrate specificity [30–32]. Moreover, BRPF1 possesses two PHD fingers for binding to unmodified histone H3 [32], one bromodomain for acetyllysine-recognition [33] and a PWWP domain for specific interaction with methylated histone H3 [34,35]. Thus, BRPF1 is a unique multivalent histone binder able to activate different acetyltransferases. BRPF1 is highly conserved from Drosophila to humans [reviewed in 16]. In C. elegans, a distantly related protein, Lin-49, regulates neuron asymmetry, hindgut development and fecundity [36–38]. In vertebrates, BRPF1 has two paralogs, BRPF2 and BRPF3 [30,31]. Inactivation of zebrafish Brpf1 alters pharyngeal segmental identity [39], and disruption of medaka fish Brpf1 affects craniofacial and caudal skeletons [40], indicating that fish Brpf1 regulates skeletal development. These studies suggest that mammalian BRPF1 may also play an important role in development. Of relevance, loss of mouse Brpf2 leads to embryonic lethality, with growth retardation, neural tube defects, abnormal eye development and faulty erythropoiesis [41], supporting that BRPF1, BRPF2 and BRPF3 have non-redundant functions in vivo. As key partners of BRPF1, MOZ and MORF are important in different types of normal and pathological stem cells. Mouse Moz plays a key role in self-renewal and maintenance of hematopoietic stem cells [42,43]. Consistent with this, human MOZ and MORF are rearranged in leukemia and other hematological malignancies [reviewed in 14,16]. One of the resulting fusion proteins is crucial for self-renewal of leukemic stem cells [44,45]. In addition, the MORF gene is frequently altered in castration-resistant prostate cancer [46] and its mutations have been detected in breast cancer [47], although it remains unclear whether related cancer stem cells are affected. The gene is also mutated in four developmental disorders, Noonan syndrome-like disorder [48], Ohdo syndrome [49] and genitopatellar syndrome [50,51] and blepharophimosis-ptosis-epicanthus inversus syndrome [52]. One common characteristic of these disorders is intellectual disability. Related to this, mice with residual MORF expression display neocortical defects [53] and abnormal neural stem cells [54], raising an interesting question whether BRPF1 plays a role in the brain. The interaction of BRPF1 with MOZ and MORF, as indicated by molecular and cell-based studies [30–32], suggests the exciting possibility that BRPF1 may regulate mammalian development. Related to this, we have recently found that it is essential for mouse embryo survival [55]. Our expression survey has also identified high-level expression of Brpf1 in the brain [55]. Here we examine this further and demonstrate that forebrain-specific inactivation of the mouse Brpf1 gene leads to dentate gyrus hypoplasia, reduces expression of key genes involved and deregulates neural stem cells and progenitors. Upon Brpf1 loss, Sox2+ neural stem cells and Tbr2+ intermediate neuronal progenitors fail to settle at the subgranular zone of the dentate gyrus, one of the two major sites known to harbor adult neural stem cells. This is the first epigenetic regulator to be identified with such an important role in the dentate gyrus. By using a mouse strain containing a LacZ knockin cassette inserted at the Brpf1 locus (S1A Fig), we have recently detected high β-galactosidase activities in the neocortex and hippocampus [55]. To gain further insights, we examined frozen sections prepared from Brpf1l/+ pre- and post-natal brains more carefully than previously reported [55]. As shown in Fig. 1A (left), strong expression was detected at the marginal zone of the developing neocortex at embryonic day (E) 14. 5. At E17. 5, very weak expression was present in different regions of the neocortex (Fig. 1B, left). At postnatal day (P) 3, P14 and the adult stage, much stronger expression was found in all six layers of the neocortex, with some enrichment in layers II-III (Fig. 1C-E, left). Similarly, we examined the expression in the hippocampus. At E14. 5 and E17. 5, no expression was detectable in the hippocampal primordium (Fig. 1A-B, right). Interestingly, at both stages, sparse signals were present in the dentate migration stream (Fig. 1A-B, right). At P3, relatively strong expression appeared in the cornum amonni (CA) regions (Fig. 1C, right), indicating that Brpf1 expression increases dramatically from E17. 5 to P3. At P3, only sparse signals were detected in the developing dentate gyrus (Fig. 1C, right). At P14 and the adult stage, strong expression was detected in the dentate gyrus (Fig. 1D-E, right). These results support a potential role of Brpf1 in neocortex and hippocampus development. To determine the function of Brpf1 in the forebrain, we crossed heterozygous Brpf1f/+ mice (S1A Fig) with the Emx1-Cre strain, which expresses the Cre recombinase from the Emx1 locus and allows LoxP excision in the neocortex and hippocampus [56]. The resulting Brpf1f/+; Emx1-Cre mice appeared normal and intercrosses between them yielded Brpf1f/f; Emx1-Cre (or bKO, short for forebrain conditional knockout) animals. Genomic PCR (S1B Fig) and RT-PCR (S1C Fig) confirmed specific excision in the forebrain but not cerebellum. The specific knockout had minimal effects on the expression of Brpf2, Brpf3, Moz and Morf (S1D Fig). Most of the bKO pups died prior to weaning at P21 [57]. Systematic histological analysis of the mutant brain identified three defective areas: the neocortex, hippocampus and corpus callosum, whereas other brain regions such as the cerebellum were normal [57]. Among these defects, the one in the hippocampus is the most striking and thus investigated here. Nissl staining of brain sections revealed that when compared to the control, the suprapyramidal blade of the dorsal hippocampus in the P10 mutant brain was shorter, with one end remaining attached to the ventricular zone, whereas the infrapyramidal blade was completely missing (Fig. 2A, right). The nuclear layers of CA1 and CA3 appeared more diffusely packed than those in the control (Fig. 2A, right). In the mutant, the junction of CA1 with the subiculum was not as clear-cut as that in the control and the subiculum itself was expanded. Similar changes were found in the mutant brain at P24 (Fig. 2B). More importantly, these defects also appeared in serial sagittal sections and similar abnormalities were found in the ventral hippocampus (Fig. 2C-D), indicating that the entire hippocampal formation is affected. The mouse dentate gyrus develops from the cortical hem around mid-gestation and involves dynamic neuron migration and differentiation, both of which continue in the first two weeks after birth [58,59]. To determine the developmental point when the defects start to occur, we applied Nissl staining to brain sections from E17. 5 fetuses and P0 neonates. As shown in Fig. 3A-B, the developing dentate gyrus was underdeveloped at both time points, indicating that the defects originate from prenatal development. To characterize the defects further, we performed Golgi-Cox staining to assess formation of axons and dendrites. This staining revealed abnormal axon and dendritic trees in the hippocampus (Fig. 3C). Interestingly, the disorganization was not just limited to the dentate gyrus but also occurred in the CA regions, indicating that Brpf1 is required for proper development of the entire hippocampus. Another method to examine the axons of the dentate granule cells, the so-called mossy fibers, is the Timm’s stain. The first mossy fibers invade CA3 at around P0, and their number gradually increases during postnatal development [60]. We examined mossy fiber development at P8 by Timm’s stain and found that the fiber projection was defective from the very beginning of its development. The suprapyramidal bundles (spb) and the dentate hilum of mossy fibers were virtually missing in the mutant (S2 Fig). To shed light on the underlying mechanisms for the defects, we examined the sections by immunohistochemical staining with an anti-Ki67 antibody. As shown in Fig. 3D-E, Ki67+ neuronal precursors were enriched in the subgranular zone of the wild-type dentate gyrus at both P10 and P24 (left), but such precursors were missing in the mutant (right), indicating a lack of proliferation in the subgranular zone. These results indicate that Brpf1 is essential for hippocampus development and regulates neuronal proliferation in the dentate gyrus. To gain mechanistic insights into the observed defects in the hippocampus, we asked whether there are other mutant mice displaying similar phenotypes in the dentate gyrus. Literature search revealed that loss of several transcription factors cause similar hypoplasia in the dentate gyrus, including Sox2 [61], Tlx (tailless) [62], Tbr2 (T-box brain protein 2, also known as eomesodermin) [63], NeuroD1 [64], Emx2 (empty spiracles homeobox 2) [65], neurogenin 2 [66] and FoxG1 (also known as BF1, for brain factor 1) [67]. Among these, Sox2 and Tlx are two well-known neural stem cell markers [68,69], whereas Tbr2 is important for intermediate neuronal progenitors [70]. The unexpected finding that upon loss, Brpf1 shares phenotypes with these three transcription factors in dentate gyrus development suggests a potential link to neural stem cells and progenitors. The subgranular zone of the dentate gyrus is one of two major sites harboring adult neural stem cells [68]. As noted above, Sox2 is a neural stem cell marker and its loss leads to dentate gyrus hypoplasia [61]. In addition, Ki67+ neuronal precursors disappeared in the subgranular zone of the mutant dentate gyrus (Fig. 3D-E). These observations suggest that Brpf1 loss may deregulate neural stem cells. To investigate this possibility, we performed immunostaining of brain sections with an antibody against Sox2. At P0, Sox2+ neural stem cells were enriched in the wild-type dentate gyrus (Fig. 4A, left two panels). This population was smaller in the mutant dentate gyrus (Fig. 4A, right two panels & Fig. 4D). In support of this, when compared to the wild-type dentate gyrus, the mutant contained a much smaller population of neurons expressing Ctip2 (Fig. 4E-F), a transcription factor important for dentate gyrus development [71]. As development progressed to P10 and P14, wild-type Sox2+ neural stem cells became enriched in the subgranular zone (Fig. 4B-C, left two panels). By contrast, no such enrichment was present in the mutant (Fig. 4B-C, right two panels), suggesting the requirement of Brpf1 for development of Sox2+ neural stem cells in the dentate gyrus. We also analyzed doublecortin (Dcx) -expressing neuroblasts. At P0, Dcx+ cells were enriched in both wild-type and mutant dentate gyri (Fig. 4G), and the distribution was rather uniform in the developing dentate gyrus. Different from Sox2+ neural stem cells, Dcx+ immature neurons were also present in the pryramidal layers of the CA1 and CA3 regions (Fig. 4G). As the development progressed, Dcx+ immature neurons became enriched in the subgranular zone of the wild-type dentate gyrus at P10 and P24 (Fig. 4H-I, left panels). A similar trend of enrichment was observed in the mutant, but the cell number was greatly reduced (Fig. 4H-I, right panels), supporting that Brpf1 loss comprises Dcx+ neuroblasts. Intermediate neuronal progenitors are essential for dentate gyrus development [72,73]. Tbr2 is a marker of these progenitors. The similar phenotype of Tbr2 inactivation in the dentate gyrus [63] suggests that these progenitors may also be affected. Intermediate neuronal progenitors are derived from radial glial cells [70], so we first analyzed the expression of Gfap (glial fibrillary acidic protein), a well-known marker of radial glial cells [73]. At E16. 5, distribution of Gfap+ glial cells was slightly altered in the mutant dentate gyrus (Fig. 5A). At P0, such cells were enriched at the outer rim of the wild-type dentate gyrus (Fig. 5B, left two panels), but this distribution became disorganized in the mutant (right two panels). At P10, Gfap+ radial glial cells and astrocytes were nicely organized along the granular cell layer of the dentate gyrus and the hippocampal fissure, respectively (Fig. 5C, left two panels), but this pattern became virtually missing in the mutant (right two panels). In particular, there were fewer Gfap+ radial glial cells in the granular cell layer of the mutant dentate gyrus and these cells were disoriented (Fig. 5D). We next analyzed intermediate neuronal progenitors with an anti-Tbr2 antibody. At E13. 5, Tbr2 expression in the hippocampal primordium was rather similar between the wild-type and mutant (Fig. 6A). By E16. 5, Tbr2+ progenitors were present in the wild-type dentate neuroepithelium and migrated along the dentate migratory stream to the forming dentate gyrus (Fig. 6B, left two panels). Such migration was also found in the mutant, but the population was smaller in the forming dentate gyrus (Fig. 6B, right two panels). At P0, a similar difference was found between the wild-type and mutant dentate gyri (Fig. 6C). Quantification confirmed this (Fig. 6E). As the development progressed, Tbr2+ progenitors translocated to the subgranular zone and the hilum of the wild-type dentate gyrus (Fig. 6D, left two panels). In stark contrast, no such translocation was found in the mutant (right two panels), with some of Tbr2+ progenitors stayed at the molecular cell layer (Fig. 6D, right two panels; marked with yellow arrowheads). These results indicate that the abnormalities start at a fetal stage and Brpf1 inactivation impairs the migration of Tbr2+ progenitors. We performed additional analyses with antibodies against three neurogenic transcription factors, NeuroD1, neurogenin 2 and FoxG1. At E16. 5, NeuroD1+ neuroblasts were enriched in both wild-type and mutant developing dentate gyri, rather uniformly distributed across the granular layer and the hilum (Fig. 6F). As the development progressed to P10, NeuroD1+ neuroblasts became enriched in the granular cell layer in the wild-type dentate gyrus (Fig. 6G, left panels). By contrast, few such cells were present in the mutant dentate gyrus (right panels). At P24, NeuroD1+ neuroblasts were restricted to the wild-type subgranular zone (Fig. 6H, left panels), but almost none was observed in the mutant (right panels). These results indicate that Brpf1 is required for proper development and migration of NeuroD1+ neuroblasts. Neurogenin 2 is essential for dentate gyrus formation [66]. At P0, distribution of neurogenin 2+ progenitors was ubiquitious in the wild-type and mutant hippocampi (S3A Fig). At P24, all granular neurons in the dentate gyrus and the pyramidal neurons in the CA1 and CA3 regions expressed neurogenin 2 (S3B Fig, left two panels). This expression pattern was also observed in the mutant, but the cells were not as tightly packed in the pyramidal layers of the CA regions and the suprapyramidal blade of the dentate gyrus (right two panels). As expected, the infrapyramidal blade was missing. For immunostaining to detect the brain-specific transcription factor FoxG1, we performed immunostaining with antibodies against FoxG1 and Tuj1, a neuron-specific β-tubulin. This analysis revealed no obvious defects at E13. 5 (S4A Fig). At P0, the hippocampal region became slightly disorganized (S4B Fig). At P24, for the wild-type hippocampus, FoxG1 was expressed in pyramidal neurons of the CA1 region and the granular layer of the dentate gyrus, but not in pyramidal neurons of the CA3 region (S4C Fig, left two panels). For the mutant hippocampus, FoxG1 was also expressed in these two regions (S4C Fig, right two panels), but the neurons were much less tightly packed than those in the wild-type hippocampus (S4C Fig). Moreover, unlike the wild-type, FoxG1 expression was also detected in pyramidal neurons of the CA3 region (S4C Fig). As expected, the infrapyramidal blade was absent in the mutant hippocampus (S4C Fig, right two panels). These results support that Brpf1 inactivation alters neurogenesis in the hippocampus. It is interesting to note that the impact on Sox2, Tbr2 and NeuroD1 (Figs. 4A-C & 6) is quite different from that on neurogenin 2 and FoxG1 (S3–S4 Figs). With the former three, the positively expressing neural stem cells or neuronal precursors displayed active migration in the dentate gyrus and settled at the subgranular zone (Figs. 4A-C & 6). By contrast, neurogenin 2-expressing progenitors or FoxG1-expressing neurons were rather uniformly distributed in the granular layer (S3–S4 Figs). These two groups of transcription factors appear to regulate different stages of dentate gyrus development. Thus, Brpf1 may regulate different stages of hippocampus development. At the prenatal and postnatal stages, dentate gyrus development involves two waves of neuronal precursors migrating from the neuroepithelium at the ventricular or subventricular zone to the dentate gyrus [58,59,74,75], so we next investigated whether and how Brpf1 loss affects the migration. For this, BrdU was injected into pregnant dams at E12. 5, E14. 5 and E16. 5. The dams were then sacrificed at P0 to isolate the brain for immunohistochemical analysis with an anti-BrdU monoclonal antibody. BrdU+ cells were quantified as three populations, outlined as the primary matrix (1ry), secondary matrix (2ry) and tertiary matrix (3ry) (Fig. 7A, top left) [66]. Representative images of the hippocampal regions are shown in Fig. 7A and the quantification results of the three BrdU+ populations are presented in Fig. 7B. In the first and second matrices, no difference was found between the control and mutant for all three labeling time points. In the tertiary matrix (corresponding to the developing dentate gyrus), there were fewer BrdU+ cells when labeled at E12. 5 and E14. 5 (Fig. 7A-B), indicating that Brpf1 loss affects migration of neuronal progenitors from the dentate neuroepithelium to the developing dentate gyrus. At both P10 and P24, Ki67+ neuronal precursors virtually disappeared in the subgranular zone of the mutant dentate gyri (Fig. 3D-E). To substantiate this, we performed BrdU labeling and sacrificed the pups 1 h later. Different from the birthdating analysis just described above (Fig. 7A-B), this short labeling protocol was to assess cells with active DNA synthesis. As shown in Fig. 7C-D, this protocol identified dramatic reduction of BrdU+ cells at the mutant subgranular zone at P12. In addition, immunostaining for cleaved caspase 3 failed to evident apoptosis in the wild-type or mutant dentate gyrus. Together, these results indicate that Brpf1 impairs cell cycle progression. To investigate this further, we analyzed cell cycle progression of the progenitors. E15. 5 pregnant mice were sacrificed after 1 h pulse of BrdU labeling to retrieve the fetal brain for immunofluorescence microscopy with anti-Ki67 and-BrdU antibodies. Representative images of the hippocampal regions are shown in Fig. 8A and the quantification of the immunostained cells in two regions (outlined with dotted lines) is presented in Fig. 8B. In the dentate neuroepithelium, no difference was detected (Fig. 8B, left). In the dentate migration stream, the number of BrdU+ progenitors was normal but the Ki67+ cycling cell population (at G1, S, G2 and M, but not G0) increased significantly, thereby decreasing the ratio of S-phase (BrdU+) over proliferating (Ki67+) cells (Fig. 7B, right). Moreover, immunostaining analysis of the related sections with an antibody specific to phospho-Ser10 of histone H3 revealed no difference between the wild-type and mutant (Fig. 8C-D), indicating that the M phase of the cell cycle is normal in the mutant. These results indicate that Brpf1 loss impairs cell cycle progression of the dentate migration stream at E15. 5, most likely through the G1 phase. Notably, the defects at E15. 5 (Fig. 8) were smaller than those at or after P10 (Figs. 3D-E & 7C-D). Consistent with difference, Brpf1 expression in the developing hippocampus was low at E14. 5 and E17. 5, but increased dramatically after P3 (Fig. 1). Having identified the cellular mechanisms for the observed defects in the dentate gyrus, we then investigated the underlying molecular mechanisms. Brpf1 activates Moz, Morf and Hbo1 [30–32]. These acetyltransferases function as transcriptional coregulators [19,76–79]. Thus, we considered whether Brpf1 inactivation deregulates transcription. For this, we first performed RT-PCR. Brpf1 deletion occurred efficiently (Fig. 9A) [57]. The inactivation did not affect mRNA levels of Brpf2, Brpf3, Moz and Morf (S1D Fig). Similarly, neither Hbo1 nor hMof was altered (Fig. 9B). Importantly, mRNA levels of NeuroD1, Tbr2 and FoxG1 were reduced in the mutant (Fig. 9C). This is consistent with the immunofluorescence microscopic results Figs. (6 & S4). The transcript levels of Emx2 and Tlx were also reduced (Fig. 9C). As these transcription factors are known to be important for dentate gyrus development, these results nicely explain hypoplasia of the bKO dentate gyrus. As Brpf1 loss affects cell cycle progression (Figs. 3D-E, 7C-D & 8), we analyzed expression of four cell cycle inhibitors, p16, p19, p21 and p15. Among them, p16 and p21 are two known targets of Moz [19,80]. Microarray-based gene expression analysis identified the Cdkn2a (encoding both p16 and p19) and Cdkn2b (encoding p15) genes as two of the top 35 candidates whose transcription was upregulated in the mutant dorsal cortex at P4 [57]. As shown in Fig. 9D, RT-qPCR identified increase in transcripts of p19 and p15, but not p16 or p21. These results indicate that Brpf1 loss promotes p15 and p19 transcription, which may then contribute to deregulated cell cycle progression. The microarray analysis [57] also identified transcriptional reduction in six genes related to hippocampus development (Fig. 9D). RT-qPCR confirmed this (Fig. 9D), supporting that Brpf1 is required for gene expression important for hippocampus development. As a scaffold, Brpf1 bridges subunit interaction, stimulates enzymatic activity and restricts substrate specificity of Moz, Morf and Hbo1 acetyltransferase complexes [30–32,41], so we asked whether these acetyltransferases contribute to the defects observed in the Brpf1 bKO mice. Related to this, mice with residual Morf expression display defects in the neocortex but not the hippocampus [53], suggesting that either Morf is not the sole mediator or it is not involved at all. Using a knockin LacZ reporter, we have recently detected Moz expression in the hippocampus, with a pattern similar to that of Brpf1 [55]. To investigate the role of Moz in the hippocampus and neocortex, we crossed Mozf/+ mice with the Emx1-Cre strain to obtain Mozf/+; Emx1-Cre mice. Intercrosses yielded Mozf/f; Emx1-Cre mice. They appeared normal and Nissl staining of the brain sections revealed none of the defects observed in the Brpf1-deficient brain, suggesting that like Morf, Moz is either not the sole mediator of Brpf1 or not involved at all. We then asked whether Hbo1 is involved. For this, we determined its expression in the hippocampus by immunostaining. At P10, Hbo1 was enriched in pyramidal layers of CA1 and CA3, as well as in the granular cell layer of the dentate gyrus (S5A Fig, left two panels). A similar expression pattern was also detected at P4 (S5B Fig, left two panels). When Brpf1 was deleted, the expression of Hbo1 in these three regions became weak at P4 and almost disappeared at P10 (S5A-B Fig, right two panels). The difference was not obvious at E16. 5 (S5C Fig). At P12, the total Hbo1 protein level (S5D Fig) at the dorsal cortex was not affected, suggesting that Brpf1 deletion may destabilize Hbo1 specifically at the pyramidal layers of the CA regions and the granular cell layer of the dentate gyrus. The MYST family of mammalian acetyltransferases is composed of five members: MOZ, MORF, HBO1, TIP60 and hMOF in mammals [16]. MOZ and MORF are almost interchangeable in cell-based assays [31], so we compared the interactions of MOZ, HBO1, TIP60 and hMOF with BRPF1 under the same experimental conditions. As shown in S6 Fig (lanes 1–2 & 5–6), co-expression of BRPF1, ING5 and EAF6, which are known to form a trimeric complex [31], increased the expression levels of MOZ and HBO1, possibly through tetrameric complex formation and subsequent stabilization. This may explain the loss of Hbo1 expression by immunofluorescence microscopy (S5A-B Fig). HBO1 was slightly more efficient than MOZ in co-precipitation of BRPF1 (S6 Fig). Interestingly, HBO1 co-precipitated both ING5 and EAF6 much more efficiently than MOZ, indicating that HBO1 forms a better tetrameric complex than MOZ. By contrast, TIP60 was much less efficient and hMOF displayed no affinity for BRPF1, ING5 and EAF6 (S6 Fig). These cell-based results indicate that although both MOZ and HBO1 interact efficiently and specifically with BRPF1, HBO1 is better than MOZ in forming a tetrameric complex with BRPF1, ING5 and EAF6. Herein we have demonstrated that Brpf1 is dynamically expressed during forebrain development (Fig. 1) and that forebrain-specific inactivation of mouse Brpf1 led to abnormalities in the hippocampus, esp. the dentate gyrus (Fig. 2), highlighting its importance in hippocampal neurogenesis. This is the first report of an epigenetic regulator whose loss exerts such profound effects on hippocampus development. Loss of two other groups of proteins has been reported to yield similar phenotypes. The first group includes eight DNA-binding transcription factors, Sox2 [61], Tlx (tailless) [62], Tbr2 [63], NeuroD1 [64], Emx2 [65], neurogenin 2 [66], Prox1 [81] and FoxG1 (also known as BF1, brain factor 1) [67]. Within the second group, there are three signaling regulators, Cxcr4, smoothened and Kif3a [70,82]. The latter two are important for Hedgehog signaling, while Cxcr4 is a receptor for the chemokine Cxcl12 (also known as stromal cell-derived factor 1). The Cxcl12-Cxcr4 signaling pair is not only important for granular neuron migration during dentate gyrus development [70], but also for the homing and self-renewal of hematopoietic stem cells [83]. Our unexpected finding that Brpf1 plays a similar role in the dentate gyrus suggests a potential link of the DNA-binding transcription factors and the signaling molecules to epigenetic and acetylation regulation by Brpf1. The similar phenotypes suggest that these two groups of regulators and Brpf1 define novel pathways required for patterning the dentate gyrus. The dentate gyrus develops from the cortical hem around mid-gestation, involving subsequent neuron migration and specification until the second week after birth [58,59,74,75]. Brpf1 appeared to regulate developmental processes at the late gestation and postnatal stages (Figs. 4–7). The dentate gyrus is also key to learning and memory, so it will be important to map out how each regulator is involved. The outcome will shed light on neurogenesis in the subgranular zone, a key area with adult neural stem cells [68,73]. The bKO mice are particularly similar to those lacking Tbr2 in terms of dentate gyrus hypoplasia [63]. Specific to intermediate neuronal progenitors, Tbr2 is required for transition from neural stem cells to these progenitors [72]. Since Tbr2+ cells were compromised by Brpf1 inactivation (Fig. 6A-E), Brpf1 may be essential for progression from neural stem cells to the progenitors. Mouse Cxcl12 and its receptor Cxcr4 are important for migration of granular cells to the dentate gyrus [84,85] and Tbr2 controls Cxcr4 expression and regulates Cxcl12 signaling [70], so compromised Cxcl12-Cxcr4 signaling may also contribute to dentate gyrus hypoplasia. In addition, Hedgehog signaling may be involved. First, Smoothened loss causes dentate gyrus hypoplasia [82]. Second, the neural stem cell marker Sox2 is required for Hedgehog signaling and its loss results in similar hypoplasia [61]. Finally, Brpf1 inactivation reduced Sox2 expression (Fig. 4A-D). Thus, Brpf1 is important for production of Sox2+ neural stem cells and Tbr2+ intermediate neuronal progenitors. As for additional cellular mechanisms, Brpf1 loss led to abnormal migration of neuronal progenitors from the dentate neuroepithelium to the dentate gyrus (Fig. 7A-B). The cell cycle progression was also impaired (Figs. 3D-E, 7C-D & 8). At the molecular level, Brpf1 loss reduced transcription of NeuroD1, Tbr2, Tlx, FoxG1 and Emx2 (Fig. 9C). In addition, transcription of the cell cycle inhibitors p15 and p19 was elevated, whereas the transcript levels of six genes important for hippocampal development were reduced in the mutant (Fig. 9D). Further studies are needed to investigate how Brpf1 loss confers histone acetylation and chromatin changes at these specific loci and across the entire genome. Brpf1 interacts with Moz, Morf and Hbo1, and stimulates their acetyltransferase activities [30–32], so an interesting question is how these acetyltransferases may mediate the influence of Brpf1, regulate transcription of genes such as Tbr2 and Sox2, and contribute to hippocampus development. Morf is expressed in the hippocampus, but its inactivation does not affect the hippocampus [53]. In the hippocampus, Moz and Hbo1 display expression patterns similar to Brpf1 (Figs. 1 & S5) [55]. All three acetyltransferases interact efficiently with Brpf1 [30–32], although Hbo1 is better than Moz and perhaps also Morf (a Moz paralog) in forming a tetrameric complex with Brpf1, Ing5 and Eaf6 (S6 Fig). Inactivation of either Moz or Morf does not exert the same effects on the hippocampus as Brpf1 deletion [53]. Interestingly, immunofluorescence microscopy identified reduction of Hbo1 expression in the granular cell layer of the dentate gyrus (S5A-B Fig) even though the total protein level in the dorsal cortex was not altered (S5D Fig). The H3K14 acetylation level was not altered in the mutant dorsal cortex (S5D Fig). Related to this, deletion of the Hbo1 gene in mouse embryos or its fibroblasts dramatically reduces the H3K14 acetylation level [41,79]. Thus, even if it has a role, Hbo1 may not be the sole mediator. Instead, Moz, Morf and Hbo1 may have redundant roles in the hippocampus, so Brpf1 may act through all of them. As human HBO1 was slightly better in forming a tetrameric complex with BRPF1, ING5 and EAF6 (S6 Fig), mouse Hbo1 may be a major contributor in mediating Brpf1 functions. Alternatively, Brpf1 may function independently of Moz, Morf and Hbo1. Related to this, in addition to a small region required for interaction with the three acetyltransferases, BRPF1 possesses two PHD fingers for binding to unmodified histone H3 [32], one bromodomain for acetyllysine-recognition [33] and a PWWP domain for interaction with methylated histone H3 [34,35]. It also possesses a motif for interaction with ING5 and EAF6 to form a trimeric complex [31], and ING5 has its own histone binding ability [86]. Global inactivation of Hbo1 is embryonically lethal [79], but specific deletion in different tissues has not been carried out. Further analysis of single and compound brain-specific knockouts of Moz, Morf and Hbo1 will shed light on how Brpf1 may, or may not, act through these acetyltransferases during hippocampus development. In summary, we have taken a mouse genetic approach and demonstrated that Brpf1 is important for development of the hippocampus, especially the dentate gyrus, one of the two major areas with active adult neurogenesis. Mechanistically, Brpf1 is important for proper development of related neural stem cells and intermediate progenitors by regulating neuronal migration, cell cycle progression and transcription of related genes. Further studies are needed to investigate its molecular and genetic interaction with related transcription factors, histone acetyltransferases and other chromatin regulators, and to determine the genome-wide action during hippocampus development. Mouse strains were maintained in a newly established animal facility at McGill University and all procedures involving the use of mice were performed according to guidelines and protocols approved by the McGill Animal Use Committee. Heterozygous Brpf1l mice have been described [55]. A promoterless LacZ cassette is between two FRT sites, while two loxP sites flank exons 4–6 of Brpf1 (S1A Fig). For genotyping, genomic PCR with primers Brpf1-F1 and-R1 generated a 227-bp band for wild-type, whereas genomic PCR with primers Brpf1-F1 and-mR1 produced a 162-bp fragment for the knock-in allele (S1A Fig). The LacZ cassette was deleted by breeding with PGK1-FLPo mice (Jackson Laboratory) to obtain the conditional allele Brpf1f (S1A Fig). Cross of Brpf1f/+ mice with Emx1-Cre mice (Jackson Laboratory) resulted in the heterozygote Brpf1f/+; Emx1-Cre (or Brpf1+/-, S1A Fig), and subsequent intercrosses yielded the homozygote Brpf1f/f; Emx1-Cre (or Brpf1-/-, referred to as bKO). The lines were maintained on the C57BL/6J background. Heterozygous Mozl/+ mice have been described [55] and Mozf/+ mice were generated as described above. Cross of Mozf/+ mice with Emx1-Cre mice (Jackson Laboratory) resulted in Mozf/+; Emx1-Cre, and subsequent intercrosses yielded Mozf/f; Emx1-Cre. Moz lines were maintained on a mixed C57BL/6J-CD1 genetic background. Other experimental procedures are presented in S1 Text.

Lysine acetylation refers to addition of the acetyl group to lysine residues after protein synthesis. Little is known about how this modification plays a role in the brain and neural stem cells. It is catalyzed by a group of enzymes known as lysine acetyltransferases. A novel epigenetic regulator called BRPF1 acts as a master activator of three different lysine acetyltransferases and also contains multiple structural domains for histone binding. In this study, we show that forebrain-specific inactivation of the mouse Brpf1 gene causes abnormal development of the dentate gyrus, a key component of the hippocampus. We trace the developmental origin to compromised neural stem cells and progenitors, and demonstrate that Brpf1 loss deregulates neuronal migration and cell cycle progression during development of the dentate gyrus. This is the first report on an epigenetic regulator whose loss has such a profound impact on the hippocampus, especially the dentate gyrus, a brain structure critical for learning, memory and adult neurogenesis.

Abstract Introduction Results Discussion Methods

2015

The Lysine Acetyltransferase Activator Brpf1 Governs Dentate Gyrus Development through Neural Stem Cells and Progenitors

A complete description of the transcriptome of an organism is crucial for a comprehensive understanding of how it functions and how its transcriptional networks are controlled, and may provide insights into the organism' s evolution. Despite the status of Saccharomyces cerevisiae as arguably the most well-studied model eukaryote, we still do not have a full catalog or understanding of all its genes. In order to interrogate the transcriptome of S. cerevisiae for low abundance or rapidly turned over transcripts, we deleted elements of the RNA degradation machinery with the goal of preferentially increasing the relative abundance of such transcripts. We then used high-resolution tiling microarrays and ultra high–throughput sequencing (UHTS) to identify, map, and validate unannotated transcripts that are more abundant in the RNA degradation mutants relative to wild-type cells. We identified 365 currently unannotated transcripts, the majority presumably representing low abundance or short-lived RNAs, of which 185 are previously unknown and unique to this study. It is likely that many of these are cryptic unstable transcripts (CUTs), which are rapidly degraded and whose function (s) within the cell are still unclear, while others may be novel functional transcripts. Of the 185 transcripts we identified as novel to our study, greater than 80 percent come from regions of the genome that have lower conservation scores amongst closely related yeast species than 85 percent of the verified ORFs in S. cerevisiae. Such regions of the genome have typically been less well-studied, and by definition transcripts from these regions will distinguish S. cerevisiae from these closely related species. Twelve years ago, in a landmark study resulting from the collaborative work of hundreds of scientists around the world, the budding yeast Saccharomyces cerevisiae became the first eukaryote to have its genome fully sequenced [1]. The initial analysis of the genome utilized the following (necessarily) arbitrary rules for defining whether an Open Reading Frame (ORF) was a protein-coding gene (a “genic ORF”) or not: 1) a genic ORF had to start with ATG and have at least 100 sense codons, and 2) if two ORFs of more than 100 sense codons overlapped one another by more than 50% of their lengths, then the longer was picked as being a genic ORF, while the shorter was discarded. In this way, it was determined that the sequence of 12,068 kilobases contained 5,885 potential protein-coding genes. In addition, non-protein-coding genes consisting of approximately 140 ribosomal RNA genes, 40 small nuclear RNA genes, and 275 transfer RNA genes were identified using various criteria, resulting in a total of approximately 6,340 genes. Early analyses of the predicted protein-coding genes showed that about 35% had no known function or homolog [2], leading to questions about the validity of the rules used to identify genic ORFs. Various algorithmic methods have predicted fewer genes in the yeast genome than the originally predicted number of 6,340, based on a variety of criteria [3]–[7], while other methods have found and verified new ones, especially non-coding genes [8], [9]. Comparative genomics [10]–[12], and various experimental methods [13]–[17] have also resulted in significant changes to the primary annotation of the yeast genome, introducing hundreds of newly predicted genic ORFs, while marking many others as ‘dubious’. However, new genes added by one study are frequently marked as ‘dubious’ by another, as recorded within the Saccharomyces Genome Database (SGD) [18], indicating the speculative nature of many of these annotations. Additionally, a recent study [19] has shown that the use of comparative genomics alone to determine whether or not a genomic region is likely to harbor a genic ORF can result in false negatives, since many transcribed elements may not be conserved across even closely related species. It has been suggested that such ORFs may be important for the micro-evolutionary divergence between species. Clearly, even in a genome as simple as, and containing as few introns as that of S. cerevisiae, it is still not straightforward to identify all of the genes simply based on the DNA sequence. Hybridization of RNA to tiling microarrays (microarrays containing overlapping, offset probes that tile across the entire genome) has been used to generate genome-wide transcript profiles and to detect previously unannotated transcripts. While this technique has its own caveats, it overcomes the limitations of many previous attempts to find undiscovered transcripts, by providing direct experimental support with high-resolution data. Tiling array studies have revealed more than 5,000 novel transcripts in Arabidopsis [20] and rice [21], and more than 10,000 previously unknown transcripts in human cells [22]–[24]. In yeast, tiling array experiments performed by David et al. [25], using RNA isolated from a single experimental condition, identified almost 800 novel (i. e. , not annotated in SGD [18]) transcripts. Recently, Miura et al. [26], also working with S. cerevisiae, performed large-scale sequencing of vector-capped cDNA clones [27], [28] from two cDNA libraries to accurately map over 11,000 transcriptional start sites (TSSs). Of these predicted transcripts, 667 were novel (many of which were also identified by David et al.), and contained ORFs corresponding to 100 amino acids or less and thus would have been missed in the original annotation. Furthermore, they discovered 45 new introns, 367 novel antisense transcripts, and showed that most yeast genes have two or more TSSs, demonstrating that the transcriptional potential of the yeast genome is more complex than previously thought. In total, their analysis detected only 3,599 of the more than 6,000 currently annotated genic ORFs, suggesting either that many genes were missing from their cDNA library, or that many of the annotated genic ORFs are not correct. Recent advances in sequencing technology [29]–[32] have allowed an unprecedented look at the transcriptome, using a method known as RNA-Seq [33]. This method can yield millions of sequence reads from cDNA libraries, and has been used to discover and validate transcribed regions of the genome in various organisms [34]–[36]. Most recently, RNA-Seq has been used to identify additional transcripts expressed in S. cerevisiae growing in rich medium [37], and transcripts expressed in S. pombe growing under several different conditions, including a meiotic time course [38]. From tens of millions of sequence reads, 204 novel transcripts were identified in S. cerevisiae, and 453 novel transcripts in S. pombe; additionally, many transcript boundaries were refined, and novel introns identified. The functions of these novel transcripts remain unknown, with few expected to be protein-coding [38]. There exist various mechanisms by which RNA is processed, surveyed, and turned over. In S. cerevisiae, there are two major pathways that play a role in the decay of mRNAs in the cytoplasm, both of which involve deadenylation (Figure 1). In the first pathway, deadenylation is followed by the removal of the 5′ m7G cap by Dcp1p and Dcp2p, which is then followed by degradation in the 5′ to 3′ direction by Xrn1p [39]–[44]. In addition to Dcp1p and Dcp2p, there exists a group of proteins that function as activators for decapping, including Pat1p, the Lsm1-7p complex, and Dhh1p [45]–[49]. In the second pathway, deadenylated mRNAs are degraded in the 3′ to 5′ direction by the exosome and the Ski complex (consisting of Ski2p, Ski3p, and Ski8p) [50], [51]. In the nucleus, mRNAs that are unspliced, improperly processed, and/or otherwise unable to leave the nucleus are degraded in pathways using the same machinery [52]–[55]. Rrp6p, a nuclear-only component of the exosome which has 3′ to 5′ exonuclease activity [56], [57], plays a major role in the nuclear degradation of mRNAs as well as CUTs ([58] and reviewed in [59], [60]). As described above, genome-wide screens for novel transcripts have revealed the existence of many non-coding, intergenic, and/or antisense RNAs. Such RNAs are poorly understood, sometimes being referred to as ‘transcriptional noise’, whose expression may be initiated from inadvertent binding of RNA polymerase complexes to DNA sequences that bear resemblance to ‘real’ transcriptional promoters. In S. cerevisiae, some of these transcripts are rapidly degraded and have been labeled as cryptic unstable transcripts or CUTs (Figure 1; [58] and reviewed in [59], [60]). While the roles of these CUTs are unclear, the mechanism by which these RNAs are degraded has been elucidated and it has been shown that they are specifically targeted for degradation via polyadenylation by the non-canonical polyadenylation protein Trf4p, a component of the TRAMP complex [58], [61], [62]. Why these RNAs are transcribed at all, and why a specific degradation pathway exists for them in the budding yeast remains speculative. To identify additional novel transcripts in the yeast S. cerevisiae, we have employed both tiling microarrays and RNA-Seq, with the explicit goal of identifying those transcripts that are either short-lived and/or occur in low abundance. Such transcripts may include previously unrecognized protein-coding transcripts and non-coding transcripts, as well as cryptic unstable transcripts and ‘transcriptional noise’. To allow better detection of these types of transcripts, we have analyzed RNA isolated from three strains containing various combinations of deletions of six genes that play a role in RNA processing (RRP6, XRN1, PAT1, LSM1, SKI2 and SKI8), with the hypothesis that the most unstable and/or least abundant transcripts would show the greatest relative change in abundance in such mutants. The mutant-derived RNA was compared to RNA from wild-type cells, using Affymetrix strand-specific tiling microarrays. Novel strand-specific transcripts were identified by segmentation of the relative expression measures from the tiling arrays and subsequently validated using Illumina' s Solexa sequencing platform. Using a combined tiling array and RNA-Seq approach, we have identified a total of 365 transcripts that are currently unannotated in SGD. Comparison of our data to various recently published transcriptome studies [25], [26], [37], [63] reveals that of these unannotated transcripts, 185 are novel and unique to our study. Our primary goal was the discovery of novel transcripts based on comparing RNA from mutants deficient in RNA degradation pathways to RNA from a wild-type strain. We wanted to provide, in a high-throughput fashion, distinct and complementary lines of evidence for the existence of each putative transcript. We thus selected two technologies as being appropriate for this aim: tiling arrays and high-throughput sequencing. We used the tiling arrays to discover novel transcribed segments, with their strand of origin information. This approach has been used successfully in previous studies [25] and there are well-established computational and statistical methods for analyzing tiling array data. Tiling arrays, as opposed to high-throughput sequencing, provide an even spacing of measurements across the entire genome, making them more amenable to off-the-shelf segmentation algorithms. In addition, an entire population of molecules is hybridized to the microarray, whereas a sequencing based approach is inherently a sampling strategy, limited by the depth to which one can afford to sequence, and by the complexity of the sample being sequenced. However, high-throughput sequencing provides an independent experimental platform well-suited for transcript validation as each read provides distinct evidence for the presence of a transcribed segment. Tiling microarray analysis of mRNA from yeast grown under a diverse set of several different conditions suggested that the greatest fraction of known transcripts are detectable in the presence of high salt (0. 8 M NaCl) (our unpublished results); we thus chose high salt as the growth condition used in the experiments described herein. All our deletion strains (the ‘mutant’ strains) and the wild-type strain (see Table 1 for strain details) were shocked with high salt for 30 minutes; total RNA was isolated from each strain, from which a poly A+ RNA sample was also purified, resulting in two different RNA preparations per strain. These RNAs were then labeled and hybridized to both forward and reverse strand Affymetrix yeast genome tiling microarrays (see Materials and Methods). Only perfect match (PM) probes mapping uniquely to the genome were used in the analysis; mismatch probes were discarded. In order to correct for probe-specific effects and to detect only those transcripts that were differentially expressed between a mutant and the wild-type, we used as expression measures the log ratio of mutant PM intensities to wild-type PM intensities. We segmented the log ratios using a piecewise constant change point model as implemented in the ‘segment’ function in the R package ‘tilingArray’ [64] from Bioconductor [65]. Following Huber et al. , we utilized the Bayesian information criterion (BIC) penalized likelihood to select the number of transcribed segments. Poly A+ RNA and total RNA microarray data were segmented separately. Based on a visual assessment of the resulting segmentation it appeared that BIC overestimated the number of segments (also noted by Huber et al.). Oversegmentation makes downstream validation of the segments more challenging, as putative segments are judged in pieces as opposed to their entirety. Thus, we post-processed the segmented data to: (1) join adjacent segments with similar expression measures, (2) drop segments that are not differentially expressed, using a threshold of <0. 5 on the log2 scale, (3) remove segments overlapping known annotation on the same strand, (4) remove segments containing fewer than 5 probes, and (5) remove segments opposite known annotation if they had a log2 fold change less than 2, or there was detectable transcription on the opposite strand (see Materials and Methods for a detailed discussion). For the sake of consistency, we will now refer to our post-processed segments as clusters, as they may refer to one or more original segments. After segmentation and post-processing of the tiling microarray data, we identified 892 candidate clusters in the poly A+ RNA data (826 of which were intergenic) and 338 from the total RNA data (324 of which were intergenic). Our criteria in analyzing the microarray data were somewhat liberal, with the aim of being as inclusive as possible; however, we coupled this with more stringent criteria for subsequent validation by sequencing, with the expectation that many of these clusters identified from the tiling microarrays would not be subsequently validated. All subsequent analyses were done at the cluster level. Following identification of these clusters from the tiling array data, we sought to validate them using sequencing. The same RNAs harvested for the tiling microarray experiments were used to generate cDNA libraries for ultra high-throughput sequencing on a Solexa 1G Genome Analyzer. Libraries were generated from double polyA purified RNAs (see Materials and Methods) from both the wild-type and the mutant strains, and were run on four lanes each of a Solexa flow cell. Reads that passed Solexa' s software filters were aligned to the genome using ELAND, allowing up to two mismatches per read. For subsequent analyses, we retained only reads mapping to a unique location, and in total, we generated more than 50 million uniquely mappable reads across all four strains. The wild-type library generated a total of 14,103,067 uniquely mapped reads from four lanes, the Δrrp6Δlsm1Δpat1 mutant library generated 14,745,813 reads, the Δski2Δski8Δrrp6 mutant library 14,973,577 reads, and the Δxrn1Δrrp6Δlsm1Δpat1 mutant library 10,714,094 reads. Following an assessment of the inter-lane variation we combined data across lanes for each strain (see Materials and Methods and Figure S2). In order to determine whether the sequence reads generated from the cDNA libraries contained sufficient coverage and depth of the transcriptome, we determined the coverage at each base within the following classes: Verified ORFs, Uncharacterized ORFs, Dubious ORFs, Introns, and Background regions. Background regions were defined as regions that were intergenic on both strands, with the following additional regions removed: novel regions identified in David et al. , Davis and Ares, Miura et al. , and Nagalakshmi et al. [25], [26], [37], [63], as well as putative novel regions identified in this study using the tiling array. For each of these categories we determined the percentage of total bases sequenced to a depth of 3 or greater (see Figure 2 and Figures S3 and S4). For comparison, we have included the publicly available data from Nagalakshmi et al [37]. Figure 2 demonstrates that with an increase in sequencing effort there would be a diminishing return in terms of percentage of bases sequenced to a certain depth. Figure 2 also illustrates that an increase in sequencing effort results in an increase in the percentage of bases sequenced from both background and intronic regions (see discussion). This is the case in our data as well as those of Nagalakshmi et al. This implies that any method for declaring a gene as “detected” must evaluate the data in the context of the reads observed in these regions. Figure 3 shows ROC-like curves depicting the tradeoff between detecting ORFs and detecting background regions, as we vary the detection cutoff. These plots demonstrate that the choice of a detection cutoff imposes a sample specific tradeoff between detecting annotated ORFs and background regions. For subsequent analyses, we chose a cutoff corresponding to calling 20% of background regions detected. Using this cutoff, we detected on average 75% of the Verified ORFs across all four experiments. A GO analysis [66] of the Verified ORFs that were not detected above background indicated a significant enrichment for ORFs whose gene products are involved in the cell cycle and sporulation. The lack of sporulation gene expression is not surprising, as the cells would not be expected to be undergoing sporulation under these conditions; as for cell cycle gene expression, presumably the salt shock shuts off the cell cycle, and those transcripts are no longer detected at these thresholds by the time we collected the cells (30 minutes after exposure to salt). In addition, we also analyzed our sequence reads to look at the dynamic range of detected transcripts. By considering Verified ORFs (>50 unique bp) that were detectable above background in the sequence data, the most abundantly expressed transcript in every mutant, and the wild-type, in terms of number of mapped reads per unique base was that of HSP12 (YFL014W), which is known to be induced under conditions of osmotic stress. Its average number of reads per unique nucleotide was ∼400 in every case. The least abundant transcript was different in each mutant, but with an average number of reads per base of less than 1. Thus, transcript abundances of the Verified ORFs (as measured by sequencing) span at least 3 orders of magnitude (see Table S3 for read counts and RPKMs [33] for all annotated ORFs). As another measurement of the validity of the sequenced libraries, we determined how many known introns we were able to detect by looking for reads that spanned exon-exon junctions. To detect these intron spanning reads, we identified those reads that mapped to the set of spliced genic ORFs but did not map to the unspliced genome. The wild-type and mutant libraries each generated sequence reads that map to exon-exon junctions, which, when combined, confirm splice junctions in 244 (86%) of the 284 known spliced ORFs reported in the current SGD annotation. In the most extreme case (RPL28) we saw 1399 reads that mapped to the exon-exon junction in the data from the Δrrp6Δlsm1Δpat1 mutant. Of those forty genes whose exon-exon junctions we failed to detect, two were in mitochondrial genes, and 16 were in Dubious or Uncharacterized ORFs. Of the remaining 22, six of the genes are expressed in meiosis, and fourteen have an initial exon of only a few residues. These were less likely to have been detected by our strategy, as we looked for reads that matched the ORF sequence and not the genome, which would have had to start at a few specific residues to be detected. Subsequent analysis, by inclusion of 5′ UTR sequence to capture such exon boundary spanning reads, was able to identify these remaining introns. Thus, only two Verified ORFs, YER014C-A/BUD25 and YPL075W/GCR1, which were not meiosis specific, failed to have reads detected that spanned their exon junctions. BUD25 is opposite two other Verified ORFs in the genome, while Nagalakshmi et al [37] also noted that they were unable to identify exon-exon boundary spanning reads for GCR1. Indeed, we were able to identify reads that spanned the 5′ exon-intron junction, and the 3′ intron-exon junction, suggesting that the intron is misannotated. We then examined an integrated dataset consisting of our tiling array and sequencing data as well as data from other published high-resolution studies. Various statistics of the potentially novel transcripts were computed to determine our proposed changes to the set of transcripts produced from the yeast genome. Firstly, we required that a cluster had to contain at least 50% uniquely mappable bases. For every potential novel transcript identified by our microarray data in a particular mutant, we employed the following criteria to Solexa data originating from the same mutant to validate the transcript: (A) the transcript detectable above background level, (B) the transcript differentially expressed between the mutant and the wild-type, and (C) the transcript differentially expressed when compared to its surrounding regions (see Materials and Methods for detailed explanation of precise criteria and cutoffs used for determination of validity). In addition, we analyzed our data for the presence of reads containing a putative poly A+ tail, which would allow us to infer both the strand of origin as well as a precise 3′ boundary, however, very few such reads were present in our dataset most likely due to our use of random priming as opposed to oligo dT priming. Following validation of individual clusters, we determined which clusters were common across the different mutants and as well as our poly A+ and total RNA hybridizations. 240 of our validated clusters were found in data from only one microarray, 79 were found in 2,26 in 3, and 20 were found in 4 or more of the six microarrays, resulting in 365 validated transcripts (see Table S1), identified by virtue of differential transcript abundance between one or more mutants and the wild-type strain. Of these, 204 were found exclusively in the poly A+ RNA, 86 were found exclusively in the total RNA fraction, and 75 were detected in both. Several of these overlap with novel transcripts identified in recent studies: 67 with David et al. [25], 116 with Miura et al. [26], 46 with Nagalakshmi et al. [37], and 43 with Davis et al [63]. Beyond these, our 365 validated transcripts includes 185 additional previously undescribed transcripts, which we were able to discover by down-regulating RNA degradation. The majority of these novel transcripts (140 of 185) were found and validated in a single mutant only, with only 45 of them being identified and validated on two or more mutants (Figure 4). For each of the potential novel transcripts, their immediate surrounding regions were plotted (e. g. see Figures 5 through 10 and Figures S5 and S6) along with a track of the current annotation from SGD [18], and data from David et al. [25], Miura et al. [26], and Nagalakshmi et al. [37]. Additional tracks representing nucleosome positioning [67] and the degree of conservation between Saccharomyces cerevisiae and other closely related yeast species [68] were also plotted. In addition, the transcript' s chromosome and its strand of origin are shown at the bottom of each plot. Six examples of transcripts unannotated in SGD and identified in this study can be seen in Figures 5 through 10, all of which are located in regions currently described as intergenic. Plots for all 365 currently unannotated transcripts identified in this study can be found in Figures S5 and S6. Of the 185 transcripts novel to this study, more than 80% have an average conservation score lower than 85% of the Verified ORFs (see Figure 11, as well as Figures 5 through 9 for five such examples; see also Figure S7). This implies that the vast majority of these transcripts could not have been found using comparative genomics. Figures 5 through 8 show four novel transcripts unique to this study that are all located in regions of the genome that show poor conservation across different Saccharomyces species, as indicated by the conservation track at the bottom of each plot. Both our tiling microarray data and our UHTS data clearly show that the transcripts in Figures 5 through 8 are only seen in the one or more of the mutant strains and not in the wild-type, which was the criterion that enabled us to identify them. Prior transcript discovery studies, however, were only able to identify transcripts that are present in the wild-type, and in Figures 5 through 8, there are no data from David et al. , Miura et al. , or Nagalakshmi et al. to suggest that they could detect these novel transcripts. In some cases the nucleosome track is suggestive of transcriptional potential, due to there being low occupancy immediately upstream of the potential transcript. In Figures 5 through 8 there is a nucleosome dip immediately upstream of the identified segment, which is frequently observed in connection with transcribed regions [67]. Figures 9 and 10 illustrate two examples of intergenic transcripts found in this study that have been found in at least one other study (we considered a transcript to be one found by another study if there was a 25% overlap between the transcripts on the same strand); one of these falls in a conserved region (Figure 10), while the other does not (Figure 9). Additionally, in both examples, it is clear in our UHTS data that these transcripts were present in the wild-type strain, though at lower levels than within our mutants, indicating that they could readily be detected in the other studies, as they indeed have been. Figure 9 shows a transcript on the Crick strand that is upstream of a verified ORF and is seen in all three of the other studies (though Nagalakshmi et al. do not call it). There is a large region of low nucleosome occupancy just upstream of it, suggesting that the region is indeed transcribed, and the transcript itself overlaps with the nucleosome dip of the downstream ORF, suggesting that this new transcript may play a role in the transcriptional regulation of the ORF downstream of it. Figure 10 shows a relatively long transcript (1,721 bp) on the Watson strand that is also seen in David et al. and Nagalakshmi et al. It is highly conserved and the presence of a nucleosome dip upstream suggests that this region is transcribed. We analyzed all of our novel transcripts for potential open reading frames, to determine if any were likely to be protein-coding. In each case, the longest open reading frame was translated and blasted against the non-redundant protein dataset (nr) from GenBank. The shortest novel transcript identified was 47 nucleotides long (intergenic), while the longest was 1,869 nucleotides in length (also intergenic), though the longest ORF that it contains only has the potential to encode a peptide 80 amino acids in length. The longest ORF that we discovered within all of our novel transcripts was within an ∼438 bp transcript on the Watson strand of chromosome 7 (coordinates 23,339–23,777), with the potential to encode an 87 amino acid polypeptide. However, this potential peptide showed no significant similarity when BLASTed against the GenBank non-redundant protein dataset. The remaining longest ORFs within each novel transcript were all shorter, with no significant similarities to any known proteins. It is not clear whether this means they do not encode proteins, or whether they encode novel, short proteins, which are currently uncharacterized due to their low conservation. We also analyzed each of our novel transcripts for any matches to known RNA structures present in the RFAM database [69], [70], but none of the sequences showed matches to any RFAM entries. Using our detected above background statistic, we sought to determine the percentage of recently published novel transcripts present in our sequencing data. It should be noted that non-detection based on our data does not imply non-existence of these transcripts due to the differing experimental conditions as well as the distinct assays. Using our wild-type data, we detected 18. 1% of the 487 Nagalakshmi et al. transcripts, 43. 7% of the 784 David et al. transcripts, and 16. 3% of the 667 Miura et al. transcripts. Using our Δrrp6Δlsm1Δpat1 data, we detected 65. 3% of the 176 Davis and Ares transcripts (see Table S2 and Table S2 for a discussion of which transcripts were used from each study). In this study, we have clearly demonstrated that there is still much we do not know about the transcriptome of S. cerevisiae, despite its deserved reputation as the most well-characterized eukaryote. Unbiased genome-wide studies of the budding yeast transcriptome [25], [26], [37] have yielded a remarkable amount of information, regarding new transcripts, new introns, the presence and location of antisense transcripts, and corrections to the current annotation. As described here, we have utilized tiling microarrays in conjunction with “next-generation” technologies to sequence cDNA libraries, with which we generated more than 50 million uniquely mappable reads from a wild-type and four mutant strains. Using these data, we have identified and validated 365 transcripts, the majority of which are more abundant in one or more of the RNA turnover mutants than in the wild-type strain (with a minority being less abundant), all of which are currently unannotated in SGD. The functions of these new RNAs remain unknown, though it is possible that many of the newly discovered transcripts correspond to CUTs, which normally would have been targeted for degradation by the TRAMP complex, but have been stabilized in the mutant background. Others may correspond to novel functional transcripts. These novel transcripts do not contain long ORFs capable of encoding proteins with recognizable similarity to known proteins; it is not clear whether this means they do not encode proteins or whether they code for hitherto unknown proteins with no known homologs. They also do not contain any recognizable RNA structures found in the RFAM database. While our work described here has much in common with the work described in David et al. and Nagalakshmi et al. , our use of RNA turnover mutants resulted in the finding of an additional 185 novel transcripts that may have otherwise remained undiscovered. Miura et al.' s use of vector-capped cDNA clone libraries is powerful in that it has a single nucleotide resolution, as opposed to our tiling microarray resolution of 4 nucleotides, allowing these authors to map transcriptional start sites to the exact nucleotide, in a high throughput manner. The use of overlapping, but non-identical, techniques among all these studies (including this one) has resulted in an ever more detailed knowledge of the yeast transcriptome. In our approach, we utilized a high-throughput discovery and validation pipeline. Clearly, much work needs to be done to characterize and understand the transcripts discovered here as well as those discovered in previous studies, however a first step in characterizing the transcripts is localization and then validation. In our computational analysis we employed a strategy of being lenient in identification of putative novel transcripts (differentially expressed at 0. 5 on the log2 scale). This was followed by a strict validation step (at our thresholds, on average 75% of annotated Verified ORFs were detected in our 3 mutant experiments as described by the ROC-like curves in Figure 3). Many (∼55%) of the clusters found in the microarray analysis were not validated by these stringent thresholds. These tended to be shorter, be less differentially expressed and included many clusters that were less abundant in the mutants as compared to wild-type. By using distinct assays with rigorous criteria for transcript validation, we have elucidated more of the regions of the yeast genome that are transcribed. In our attempt to find low abundance and transient transcripts by restricting our search to transcripts that were present in differential relative abundance in our RNA processing mutants, we may have missed transcripts that are present in the mutant and the wild-type at the same abundance. This was a caveat we had to consider in the pursuit of transcripts that we believed would otherwise be difficult to detect, and the discovery of 185 novel transcripts despite the work of other comprehensive genome-wide transcriptome studies shows that our strategy was a fruitful one. By utilizing the strand-specific tiling array were able to localize transcripts to their strand of origin, something that was not possible (without introducing a 3′ bias to the data by priming the labeling reaction with oligo-dT) with the current protocols for RNA-Seq using the Solexa 1G Genome Analyzer. It is likely that modified protocols will soon address this shortcoming, and indeed such protocols for the ABI SOLiD sequencing system have been recently published [71]. We can now ask the important and obvious question: has the yeast transcriptome been completely described, and what does completion mean? It is possible that if we sequence deeply enough, we may observe that every nucleotide within the genome is transcribed at some level (see Figure 2), though clearly this is not a strict enough criterion to allow us to identify a transcribed segment. The genome-wide studies that have set out to discover new transcripts in yeast in an unbiased fashion have so far used a limited set of experimental conditions. Thus, it seems likely that deep sequencing of RNA from dozens of possible conditions (which must be carefully chosen to span as much of the “expression space” as possible) will yield yet more new transcripts, or show new variations in existing ones. It will be of particular interest to profile all of these novel transcripts under a variety of conditions to see how they are regulated and co-regulated, as well as to determine whether they encode proteins or functional RNAs, and whether their absence results in a detectable phenotype. Since many of the recently discovered transcripts (including those in this study) have been found in regions of the genome where there is little or no sequence conservation (though the conservation scores from Siepel et al. [68] do not indicate whether these regions are evolving neutrally, or under positive selection), it will be informative to profile different and diverse strains of S. cerevisiae to determine if these transcripts are ubiquitous within the species, and to determine whether the syntenic (but non-conserved) regions within closely related species within the Saccharomyces sensu stricto are also transcribed. With such data, we can hope to discover and hopefully appreciate not only how each of these species are related to one another, but also how their transcriptional potential and networks have diverged. Since the landmark publication of the S. cerevisiae genome sequence 12 years ago, more than 25,000 research publications on yeast have appeared, yet we are still adding to our knowledge of the transcriptome of S. cerevisiae. While arguably the most well-understood eukaryote, we still do not have a complete understanding of such a fundamental concept as “what and where are all of its genes. ” New technologies such as high resolution tiling microarrays and ultra high-throughput sequencing are opening up new avenues of research, and it is clear that the quantity of data that these technologies allow us to generate will only increase. This study (and others like it) underscores how much work remains to be done in understanding and cataloging the transcriptomes of even the most well-studied model organisms. All deletions were created in a diploid Saccharomyces cerevisiae strain which was created by crossing strains FY23 and FY86 [72], which are isogenic to the sequenced strain S288C and carry the auxotrophic markers: his3-Δ200, leu2-Δ1, trp1-Δ63, and ura3-52. All deletions were created using the Geneticin antibiotic resistance marker, utilizing the system described in [73]. Specifically, primers specific to regions to be deleted by homologous recombination were designed to utilize the plasmid pFA6-kanMX6 as a PCR template in order to replace the regions of interest with the gene encoding for resistance against the antibiotic Geneticin. PCR was performed (see Table 2 for primers), generating approximately 1. 5 kb DNA fragments in agreement with the size of the Geneticin resistance gene, which were then transformed using standard lithium acetate transformation techniques into the diploid cells grown in YPD at 30°C at mid-log phase. Cells were selected on YPD agar plates with 300 µg/ml working concentration of Geneticin. Deletions were confirmed by PCR (see Table 2 for primers) and the diploids were sporulated and their tetrads dissected to generate haploid segregants carrying the deletions of interest. Different deletions strains were mated to generate diploids, which were then sporulated and tetrads were dissected. Because only the Geneticin marker was used to generate these deletions, PCR analysis was used to confirm all newly generated double mutant strains. The process was repeated to generate the triple and quadruple mutants (see Table 1 for resulting strains used in this study). Some deletion combinations could not be generated, suggesting they are synthetically lethal, and thus were not used in this study. For instance, Δxrn1 and Δski8 are synthetically lethal, as any attempt to combine strains with these deletions was unsuccessful. Haploid strains exhibiting phenotypes suggesting the accumulation of suppressor mutations were not used for further study. Originally the decapping factor DHH1 and the Ski complex component SKI3 were selected to be included, but strains carrying either Δdhh1 or Δski3 showed a propensity to accumulate suppressor mutations when combined with other deletions from this study and thus were dropped from the analysis. The Affymetrix tiling array data as well as the sequencing data confirmed that there was no expression signal corresponding to the genetic loci of the deleted genes. Our unpublished studies suggested that among two dozen or so different conditions that we have assayed, exposure to high salt (0. 8 M NaCl) results in the expression of the greatest fraction of known and novel transcripts, and thus was chosen as the experimental condition to use to find previously unannotated and low abundance transcripts. Cells were grown at 30°C in YPD to approximately 1×107 cells/ml as determined by a Beckman Coulter Z2 Particle Count and Size Analyzer. 1. 6 M NaCl (in YPD) was added in an equal volume of YPD prewarmed to 30°C (final concentration 0. 8 M). Cells were harvested after 30 minutes by filtration, frozen in liquid nitrogen, and kept at −80°C until RNA extraction and purification. RNA was extracted from the cells using a slightly modified version of the traditional hot phenol protocol [74] followed by ethanol precipitation and washing. Briefly, 5 ml of lysis buffer (10 mM EDTA pH 8. 0,0. 5% SDS, 10 mM Tris-HCl pH 7. 5) and 5 ml of acid phenol were added to frozen cells and incubated at 60°C for 1 hour with occasional vortexing, then placed on ice. The aqueous phase was extracted after centrifuging and additional phenol extraction steps were performed as needed, followed by a chloroform extraction. Total RNA was precipitated from the final aqueous solution with 10% volume 3 M sodium acetate pH 5. 2, and ethanol, and resuspended in nuclease-free water. All microarray analyses were carried out using Affymetrix GeneChip S. cerevisiae Tiling 1. 0R Array (Reverse) (part number: 900645) for Watson strand expression or GeneChip S. cerevisiae Tiling 1. 0F Array (Forward) (part number: 520286) for Crick strand expression. The arrays each contain more than 2. 5 million perfect match probes, which are offset from one another by 4 bases across the genome (21 bp overlap). Thus, each residue in the genome is interrogated on average by 6 oligonucleotide probes. Total RNA samples were prepared following the protocol exactly as described in David et al. [25]. PolyA RNA samples were prepared as follows. 500 µg of total RNA were PolyA purified using Qiagen Oligotex suspension to produce approximately 10 µg of PolyA RNA as determined by OD260/280. 2 µg of the PolyA purified RNA were then used in the generation of cDNA as per Affymetrix First Strand and Second Strand Synthesis protocols utilizing a T7-Oligo (dT) as the primer for the First Strand, followed by in vitro transcription to generate biotin labeled cRNA, as outlined by Affymetrix protocols. The cRNA was fragmented as described by Affymetrix, and then sent for hybridization and scanning by the PAN facility at Stanford (http: //cmgm. stanford. edu/pan/) according to standard Affymetrix protocols. Our goal was to identify short-lived transcripts based on measured intensities of probes tiling the genome. It is well known that probe affinities significantly bias the relationship between measured intensity and actual transcript abundance. In David et al. this was addressed by effectively forming log ratios between wild-type and genomic DNA hybridization. In order to highlight the changes between mutants and wild-type transcription and to reduce the effect of probe affinities we formed log ratios between mutant and wild-type intensities. This approach has the same effect on probe affinities as the approach used by David et al. , see Figures S1 and S2. The probes on the tiling array were mapped to the yeast genome, as downloaded from the Saccharomyces Genome Database on May 19th 2008, using MUMmer [75]. Only perfect match (PM) probes mapping to a unique region were retained for further analysis. For each mutant RNA hybridization, log ratios of mutant PM intensities to wild-type PM intensities were calculated. The resulting data were segmented using the ‘segment’ function in the R package ‘tilingArray’ [64] from Bioconductor release 2. 1, which performs a simple change-point analysis. The log ratios of mutant compared to wild-type for total RNA and poly A+ purified mRNA extractions for each mutant and chromosome strand were segmented separately. An open question in any segmentation analysis is the selection of the number of segments. We followed Huber et al. (2006) in using the Bayesian information criterion (BIC) penalized log-likelihood, noting that this tends to overestimate the number of segments (see below). Following the segmentation we were left with a set of segments for each of the three mutants and two RNA sample types (total RNA or poly A purified RNA). Our analyses indicated that transcripts are often split into a number of segments due to various artifacts of the array data (outliers, incomplete probe-affinity correction, cross-hybridization). At this stage, we wished to both join appropriate segments into adjacent co-expressed segments (clusters) as well as filter out a priori uninteresting clusters. The pipeline for constructing clusters from segments and producing a set of putative clusters to be validated using the sequencing data worked as follows: This process resulted in a set of putative clusters that were subsequently considered for validation by Solexa sequencing. In order to generate libraries for the Solexa platform, various reagents and kits were required. At the time that these experiments were performed, Illumina did not have an RNA-Seq specific kit, and thus parts of various kits were utilized. Note that not all of the reagents from the kits provided by Illumina were used, as these kits were adapted for use in the protocol below and not necessarily used as described in the instructions that came with the kit. They are as follows: For protocols desiring PolyA purified RNAs: Also required was a magnetic stand that can accommodate 1. 5 ml microcentrifuge tubes. The protocol as described below was done using DNase/RNase certified free siliconized 1. 5 ml microcentrifuge tubes. Strains used for our UHTS experiments are GSY147 and GSY1289 (see Table 1). GSY147 was derived from DBY10146 (a gift from David Botstein) (which itself was derived from an FY background [72]) which was backcrossed by Katja Schwartz to FY2 and FY3 [72] to generate a wild-type S288C strain that had no auxotrophies or mutations. Two consecutive purifications using oligo dT conjugated magnetic beads were performed as follows. 100 µg of Total RNA were diluted in a final volume of 100 µl water and heated at 65°C for two minutes and then placed on ice. 200 µl of beads were equilibrated by two consecutive 100 µl washes in binding buffer (mixed gently by hand), using a magnetic stand to separate the beads from the buffer. The beads were then resuspended in 100 µl of binding buffer. The RNA was added to the beads, and the tube was mixed gently by hand for 5 minutes at room temperature and then placed on the magnetic stand to separate the beads from the supernatant. The supernatant was discarded, and the beads underwent two consecutive washes with 200 µl washing buffer. The beads were resuspended in 10 mM Tris-HCl pH 7. 5, and the tube was heated at 80°C for two minutes and then immediately placed on the magnetic stand where the supernatant was transferred to a new tube. The beads were saved and prepared for the second round of PolyA purification by washing them once with 200 µl washing buffer. The entire process was then repeated once for a second round of purification, beginning with the dilution of the RNA and the denaturing of the RNA secondary structure. PolyA purified treated RNA samples were then fragmented to ensure an unbiased binding of the random hexamers during cDNA synthesis. 5× Fragmentation Buffer (200 mM Tris Acetate pH 8. 2,500 mM Potassium Acetate, 150 mM Magnesium Acetate) was made, of which 5 µl was added to the RNA sample, and the total reaction was brought up to 25 µl. The sample was heated at 94°C for 2. 5 minutes and immediately placed on ice. The sample was then run through a G-50 spin column that has been equilibrated with 3×400 µl of nuclease free water to remove ions from the fragmentation. The sample was concentrated to 10. 5 µl with a Micron filter. First Strand Synthesis: 10. 5 µl of fragmented RNAs were transferred to a PCR tube and 1 µl of random hexamer (3 µg/µl) was added. The tube was heated to 65°C for 5 minutes and then placed on ice. The following reagents from the Illumina kit were then added: 4 µl 5×1st strand buffer, 2 µl 100 mM DTT, 1 µl 10 mM dNTP, and 0. 5 µl RNaseOUT (40 U/µl). The tube was mixed and left at room temperature for 2 minutes. 1 µl SuperScript III (200 U/µl) was added, and the sample was placed in a thermocycler with the following program: 25°C for 10 minutes, 42°C for 50 minutes, 70°C for 15 minutes, 4°C hold. Second Strand Synthesis: The first strand synthesis reaction was transferred to a 1. 5 ml siliconized microcentrifuge tube and placed on ice. 61 µl nuclease free water was added to the sample, along with the following reagents from the Illumina kit: 10 µl 2nd strand buffer, 3 µl 10 mM dNTPs, 1 µl RNase H (2 U/µl), and 5 µl DNA Pol I (10 U/µl). The sample was vortexed and placed in an Eppendorf Thermomixer R (set at 16°C and programmed to spin at 1400 rpm for 15 seconds and stand for 2 minutes) overnight (minimum 2. 5 hours). The newly synthesized cDNA was purified with a QIAquick PCR spin column as per Qiagen protocols and eluted in 30 µl EB solution. The following reagents from the Illumina kit were added to the 30 µl sample as follows: 45 µl nuclease free water, 10 µl T4 DNA ligase buffer with 10 mM ATP, 4 µl 10 mM dNTPs, 5 µl T4 DNA polymerase (3 U/µl), 1 µl Klenow DNA polymerase (5 U/µl), 5 µl T4 PNK (10 U/µl). The sample was vortexed and incubated at 20°C for 30 minutes. Afterwards, the sample was purified with a QIAquick PCR spin column as per Qiagen protocols and eluted in 32 µl EB solution. The following reagents from the Illumina kit were added to the 32 µl sample as follows: 5 µl Klenow buffer, 10 µl 1 mM dATP, and 1 µl Klenow 3′ to 5′ exonuclease (5 U/µl). The sample was vortexed and incubated at 37°C for 30 minutes. Afterwards, the sample was purified with a MinElute spin column as per Qiagen protocols and eluted in 10 µl EB solution. The following reagents from the Illumina kit were added to the 10 µl sample as follows: 25 µl DNA ligase buffer, 2 µl adaptor oligo mix, and 5 µl DNA ligase (1 U/µl). The sample was vortexed and incubated at 25°C for 15 minutes. Afterwards, the sample was purified with a MinElute spin column as per Qiagen protocols and eluted in 10 µl EB solution. The 10 µl sample was loaded onto a 1% TAE agarose gel at least one lane away from a 100 bp ladder. The sample was run sufficiently far enough and a gel slice corresponding to approximately 200 bp+/−50 bp was excised out of the gel with a scalpel (note that no cDNA may be visible on the gel). The cDNA was purified using a Zymo Research Zymoclean Gel DNA Recovery Kit and eluted in 10 µl nuclease free water. The 10 µl sample was transferred to a PCR tube. The following reagents from the Illumina kit were added to the 10 µl sample as follows: 27 µl nuclease free water, 10 µl 5× cloned Phu buffer, 1 µl oligo 1. 1,1 µl oligo 2. 1,0. 5 µl 25 mM dNTPs, 0. 5 µl Phu polymerase. The sample was then run on a thermocycler using the following program: 98°C hold for 30 seconds, 98°C for 10 seconds, 65°C for 30 seconds, 72°C for 30 seconds, 72°C hold for 5 minutes, 4°C hold, for 50 cycles. The sample was purified with a QIAquick PCR spin column as per Qiagen protocols and eluted in 30 µl EB solution. The sample was then run through a G-50 spin column that had been equilibrated with 3×400 µl of nuclease free water to remove any remaining unincorporated nucleotides that would interfere with the concentration determination of the library. The DNA was concentrated through the use of a Speed Vac until the final volume of the library was 10 µl. The cDNA was quantified using a Nanodrop. A concentration range between 10–100 ng/ml final concentration of an RNAseq library is required for good quality sequencing. The sample was then sent for sequencing in the Genetics Department Solexa machine at Stanford. Sequence reads that passed Solexa' s quality filters were aligned to both the yeast genome and the spliced yeast ORF set (allowing up to 2 mismatches), downloaded from the Saccharomyces Genome Database (SGD) [76] on May 19th, 2008, using ELAND, which is part of the Solexa analysis pipeline [77] (we used version 0. 3. 0). Only reads mapping uniquely to the genome were retained. We examined the goodness of fit for a simple Poisson model described below, using the chi-square goodness of fit statistic (see [78]). QQ-plots of the observed statistic for each known gene against the theoretical distribution are shown in Figure S2 and show a remarkably good fit. Based on this model, we aggregated data for each strain across the multiple lanes on the Solexa flow cell. In order to validate each putative transcript identified by tiling array data analysis, we investigated the following three criteria: An important consideration in all subsequent analyses was that certain areas of the genome are unmappable due to repeated sequences. We defined a base as non-unique if the 25mer starting at that position occurs elsewhere in the genome. We excluded all such bases from consideration in subsequent analyses. We applied the detected above background statistic described above with a cutoff of. 8. Results are available in Table S2. All raw data have been deposited in the GEO database with accession number GSE11802.

The budding yeast Saccharomyces cerevisiae, because of the relative ease of its genetic manipulation and its ease of handling in the laboratory, has long served as a model on which studies in higher organisms have been based. To more fully understand how eukaryotic cells express their genomes, we sought to identify RNA species that are transcribed at very low levels or that are rapidly degraded. We created mutants deficient in the ability to degrade RNA, with the expectation that this would increase the relative abundance of such RNAs, and then used high-resolution microarrays and sequencing technologies to locate and identify from where these RNAs are transcribed. Using this approach, we have identified 365 transcripts that do not appear in the most current list of annotated S. cerevisiae RNA transcripts; of these, 185 are unique to our study. Many of these novel transcripts derive from regions of the genome that are poorly conserved between S. cerevisiae and other closely related yeast species, suggesting that these RNAs may play an important role in the divergent microevolution of S. cerevisiae.

Abstract Introduction Results Discussion Materials and Methods

genetics and genomics/genomics genetics and genomics/gene discovery genetics and genomics/gene expression molecular biology/mrna stability molecular biology/bioinformatics genetics and genomics/bioinformatics

2008

Novel Low Abundance and Transient RNAs in Yeast Revealed by Tiling Microarrays and Ultra High–Throughput Sequencing Are Not Conserved Across Closely Related Yeast Species

Drug resistant strains of the malaria parasite, Plasmodium falciparum, have rendered chloroquine ineffective throughout much of the world. In parts of Africa and Asia, the coordinated shift from chloroquine to other drugs has resulted in the near disappearance of chloroquine-resistant (CQR) parasites from the population. Currently, there is no molecular explanation for this phenomenon. Herein, we employ metabolic quantitative trait locus mapping (mQTL) to analyze progeny from a genetic cross between chloroquine-susceptible (CQS) and CQR parasites. We identify a family of hemoglobin-derived peptides that are elevated in CQR parasites and show that peptide accumulation, drug resistance, and reduced parasite fitness are all linked in vitro to CQR alleles of the P. falciparum chloroquine resistance transporter (pfcrt). These findings suggest that CQR parasites are less fit because mutations in pfcrt interfere with hemoglobin digestion by the parasite. Moreover, our findings may provide a molecular explanation for the reemergence of CQS parasites in wild populations. Drug resistance is a critical issue facing worldwide malaria control. The spread and persistence of chloroquine-resistant (CQR) Plasmodium falciparum has rendered chloroquine, an inexpensive and potent drug, ineffective throughout most of the world [1]. In sub-Saharan Africa [2] and the island of Hainan (China) [3], where CQR parasites formerly accounted for 85–98% of the population, the coordinated cessation of chloroquine treatment resulted in a dramatic reduction (40–100%) in the prevalence of CQR parasites over 10 years. Although the reemergence of chloroquine sensitive (CQS) parasites is a major development with regard to human health, the underlying molecular mechanisms behind this phenomenon are unknown. The importance of drug resistance to world health has prompted a half century of intensive research into the parasite' s mechanisms of resistance [4]. These efforts identified the predominant gene responsible for chloroquine resistance, the P. falciparum chloroquine resistance transporter (pfcrt, pf3D7_0709000) [5]. PfCRT is a multiple pass membrane protein that is localized to the digestive vacuole [5], [6]. Mutations associated with chloroquine resistance have been mapped [7]–[10] and a single polymorphism in the first transmembrane domain (K76T) has been shown to be essential for drug resistance [11]. Recently, the resistant form of PfCRT was found to transport chloroquine under physiologically relevant conditions. Wildtype PfCRT is also assumed to function as a transporter [12], but its native substrates are unclear and the impact of CQR alleles on PfCRT' s normal function remains a mystery. Although mutations in pfcrt are necessary and sufficient to confer chloroquine resistance, several other genes have also been implicated in drug tolerance. The interactions between pfcrt and these other loci, including the P. falciparum multiple drug resistance gene (pfmdr1) [9], [13]–[15] and P. falciparum multiple resistance protein (pfmrp1) [16], are not clearly understood. One possibility is that mutations at secondary loci interact with pfcrt to modulate drug resistance. Alternatively, mutations at other loci may compensate for loss of function associated with CQR forms of PfCRT. Understanding how pfcrt mutations affect parasite physiology is an essential step towards unraveling these polygenic contributions to chloroquine resistance. Given PfCRT' s transmembrane structure, its localization to the digestive vacuole, and its ability to transport chloroquine in vitro, we hypothesized that wildtype PfCRT functions as a transporter. Furthermore, we predicted that mutations in pfcrt, and other CQR-associated genes, alter steady-state metabolite levels in PfCRT-associated pathways. Identifying these phenotypes and linking them to specific genes is difficult because 1) metabolites are often derived from multiple pathways, 2) steady-state levels of compounds can be affected by small perturbations far upstream or downstream of a particular compound and 3) metabolic regulation often involves complex interactions between nonlinear factors, such as covalent modification of enzymes, feedback inhibition, and allosteric regulation. Quantitative trait locus (QTL) mapping is a powerful tool for unraveling complex metabolic networks and tracing metabolic regulation to specific genes [17], [18]. QTL mapping uses the segregation of alleles across a phenotypically and genetically diverse population to measure the genome-wide contribution of individual alleles to a phenotype (e. g. a metabolite concentration) [19]–[24]. Recently, this approach has been integrated with untargeted metabolomics to study metabolic regulation on a genome-wide scale [17]. This emerging metabolic QTL (mQTL) strategy is of obvious applicability to malariology. However, only three genetic crosses of P. falciparum have ever been completed because of serious logistical challenges [25]. One of these efforts crossed the CQS HB3 parasite clone with the CQR Dd2 clone. The haploid progeny from this cross provide a unique opportunity to investigate the metabolic consequences of drug resistance and the role of compensatory mutations in maintaining metabolic homeostasis. In this study, we use high resolution mass spectrometry to measure the global metabolic profiles of progeny from the HB3×Dd2 genetic cross. Using mQTL mapping, we identify a family of hemoglobin-derived peptides that accumulate in parasites carrying CQR pfcrt alleles. We show that this phenotype can be recapitulated in transgenic parasite lines in which the native pfcrt gene has been replaced with a recombinant CQR or CQS pfcrt allele. In addition, we show that two independently evolved CQR alleles of pfcrt confer a fitness cost. From these data, we propose that CQR imparts a fitness cost on parasites by disrupting hemoglobin catabolism. We combined genome-wide QTL mapping with mass spectrometry (MS) -based metabolomics to identify genetic loci in P. falciparum that have a significant influence over steady-state metabolite levels. To achieve this, synchronous trophozoite-stage cultures (24 hour post invasion) of the 34 haploid progeny and two parental lines from the HB3×Dd2 genetic were grown using established in vitro methods [26]. Metabolites were harvested from each of the cultures and high-resolution mass data were collected on an LTQ-Orbitrap. For maximum sensitivity, mass data were peak-picked near the noise threshold (minimum signal/noise = 3) and biologically relevant data were identified using a two-stage assignment routine. In the first stage, promising signals in the untargeted mass list (N = 124,020) were identified by QTL analysis and genetic linkages with LOD scores greater than 3 (N = 1,707) were manually curated to remove artifacts and correct for errors in the automated MS data peak picking algorithm. Untargeted mass data are highly redundant because electrospray ionization generates numerous adducts and in-source fragments for each input metabolite. Consequently, we employed a second assignment stage to condense redundant data into a single representative parent mass for each compound. Curated signals (N = 279) were clustered by coelution, signal covariance, and mass difference relative to common adducts/isotopomers/fragments. The most intense signal from each group was designated as the parent mass. A total of 15 signals passed this two-stage filtering routine. Each of these signals had LOD scores above their permutation-established thresholds for genome-wide significance (α =. 05), and all but 3 of these signals were significant after Bonferroni correction (α =. 05/24) for multiple hypothesis testing (Table 1). Notably, all but one of these 15 signals have LOD scores above the 5% false discovery threshold (LOD = 5. 5) established by Q-value analysis for the original unfiltered mass list. Surprisingly, all Bonferroni-compliant signals were linked to a single 22 cM genomic region on chromosome 7 (Figure 1) containing pfcrt, the gene responsible for conferring chloroquine resistance [5]. Tentative metabolite assignments were generated for each of the PfCRT-linked masses by submitting observed signals to the Madison Metabolomics Consortium Database [27] and the Human Metabolome Database [28]. The resulting list of putative IDs was evaluated by analyzing ms/ms fragmentation spectra from enriched parasite extracts (Figure 2). Eleven of the 15 significant compounds (α =. 05) had exact masses and fragmentation spectra consistent with small peptides. Peptide assignments were empirically validated by co-elution of the parasite-derived signal with synthetic peptide standards. These experiments were conducted using single reaction monitoring (SRM) mass spectrometry, a robust analytical method for confirming specific metabolite assignments in complex mixtures [29]. Each of the PfCRT-linked peptides co-eluted with their respective synthetic standards. Moreover, the intensity of the parasite-derived signal changed in a concentration dependent manner when standards were added to parasite extracts (Figure 2). QTL analysis demonstrated that the peptide accumulation phenotype observed in CQR parasites maps to a 36 kb locus on chromosome 7 containing pfcrt and eight other genes (Figure 3). To determine if the pfcrt gene is responsible for the peptide accumulation phenotype, we analyzed transgenic parasites in which the native pfcrt gene has been replaced with either a CQS (C2; HB3 allele) or CQR (C4, Dd2 allele; C6,7G8 allele) variant of pfcrt [11]. The two CQR alleles we tested have distinct evolutionary origins but all of the transgenic parasites share the CQS GC03 background (a progeny clone from the HB3×Dd2 cross). MS analysis demonstrated that parasites carrying either of the CQR pfcrt alleles accumulate peptides over the 48-hour intraerythrocytic life cycle to much higher levels (>32-fold) than parasites carrying the CQS allele (Figure S1). Furthermore, a survey of diverse parasite genotypes showed that all of the parasites that accumulate peptides carry the critical PfCRT-K76T polymorphism that is required for chloroquine resistance [30] (Figure 4). Hemoglobin catabolism is an essential activity that provides amino acids and the physical space parasites need to grow [31]. All of the PfCRT-linked peptides identified by QTL analysis are found in, but not unique to, hemoglobin. Because PfCRT is located on the digestive vacuole membrane, which is the organelle where hemoglobin metabolism occurs, we hypothesized that PfCRT polymorphisms directly affect hemoglobin catabolism. To test this hypothesis, we conducted a comprehensive peptidomics analysis of parasites to monitor PfCRT-related effects on the hemoglobin digestion pathway. Erythrocytes infected with either CQS (C2) or CQR (C4, C6) parasites were purified by Percoll density gradient and endogenous peptides present in parasite extracts were analyzed by high-resolution nanospray LC-MS/MS. This analysis identified 362 endogenous peptides ranging from trimers to 32-mers that exactly correspond to sequences found in either the α or β chain of hemoglobin (Figure 5, Figure S3, and Figure S4). The majority of these peptides (e. g. VHLTPEE) have sequences that are unique to hemoglobin (i. e. have no other exact matches in either the P. falciparum or human genomes) and exist in overlapping clusters of structurally related peptides. The peptide clustering we observed is consistent with the parasite' s semi-ordered hemoglobin digestion cascade, which involves protein degradation via a series of proteases (plasmepsins, falcipains, and falcilysin) and aminopeptidases [32]. The boundaries of most of the peptide clusters we observed coincide with established proteolytic cleavage sites [33]. In addition, we observed several sequence breaks that may suggest previously unmapped cut sites (Figure 5). Quantitative analysis identified 87 peptides that show evidence for differential accumulation between CQS and CQR lines (|z-score|>4; Figure 5). These peptides include the mQTL-linked peptides (e. g. PEE and DLS), other structurally related peptides (e. g. VHLTPEEK and HFDLS), and novel classes of peptides that fell below the analytical limit of detection in the original mQTL analysis (e. g. DPENFR in β-hemoglobin). We interpret peptide accumulation in CQR parasites as evidence for impaired hemoglobin catabolism. Because hemoglobin catabolism is essential to P. falciparum [31], we anticipated that CQR parasites may have alterations in other metabolic pathways. To determine the degree to which CRQ pfcrt alleles affect metabolic homeostasis, we quantified metabolites present in extracts from density-purified samples of RBCs infected with transgenic CQS (C2) and CQR (C4, C6) allele exchange parasites. Using high resolution HPLC-MS, we quantified 80 metabolites, including representative central carbon metabolites, nucleotides, cofactors, amino acids, and peptides. Surprisingly, the only compounds showing consistent steady-state metabolic differences between the CQS and CQR lines were hemoglobin-derived peptides (Figure S5). These data suggest that altered hemoglobin catabolism is the most significant metabolic consequence of CQR mutations. Given the significance of hemoglobin catabolism in the parasite' s blood stage development, we hypothesized that CQR-induced peptide-accumulation would be associated with a fitness cost. To test this hypothesis, we conducted long-term competition experiments between CQS and CQR transgenic parasites grown for 70 days in mixed cultures containing either two (C2, C4; C2, C6) or three (C2, C4, C6) allele-exchange parasite lines in the same flask. This experiment differs from previous in vitro studies in our use of the transgenic pfcrt allele exchange parasites, which controls for the polygenic contributions from genetic background, and the long timeframe over which cultures were allowed to compete (∼35 generations). DNA was harvested every 48 hours and the abundance of each pfcrt allele was quantified by Sanger sequencing (Figure S6 and Figure S7). Quantitative DNA sequencing showed that mixed populations of CQS/CQR parasites converted to nearly pure populations of CQS parasites after 70 days (Figure 6). This result was consistent across both Asian (Dd2, C4) and South American (7G8, C6) CQR pfcrt alleles and the outcome was not influenced by the starting ratio of the mixed populations (Figure 6 and Figure S8). Our competition assay showed a transient increase in CQR allele abundance that peaked at ∼20 days. This transient peak is attributable to a difference in cell cycle length between CQR and CQS parasites. Differences in cycle length accumulate across generations and thus progressively offset cycle stages of competing populations. Since parasites amplify their DNA midway through their 48 hour cell cycle, and daughter progeny do not all successfully re-invade, the population with the longer generation-to-generation replication time will have more DNA (but fewer infected cells) when populations are offset across generations (i. e. the leading population is at the ring stage whereas the trailing population is at the schizont stage). To quantify this phenomenon, we constructed a computer model of in vitro P. falciparum competition. Our model showed that a 2 hour difference in cell cycle length and a 13% overall fitness cost per generation can explain both the transient increase in CQR DNA and the subsequent disappearance of CQR alleles from mixed cultures (R2 =. 98, Figure 6). This computational assessment is empirically supported by competition experiments involving asynchronous populations of parasites, which showed the same long-term population dynamics but lack the transient increase in CQR allele abundance (Figure S8). Since CQR parasites are less fit (Figure 6 and Figure S8) and have altered hemoglobin metabolism when compared to CQS parasites (Figure 6), we predicted that CQR parasites would have more difficulty in using hemoglobin-derived amino acids for biomass production. To test this, growth assays were conducted in a modified RPMI medium lacking all amino acids except isoleucine (which is not present in hemoglobin), which forces parasites to use hemoglobin to supply its amino acid needs. In agreement with the literature [31], [34] parasites incubated in isoleucine-only medium grew more slowly than those incubated in rich medium. However, CQR parasites were significantly more impaired than CQS parasites (p = 0. 0075, Figure S9), suggesting that CQR-induced fitness changes are linked to hemoglobin catabolism. This study provides four independent lines of evidence linking chloroquine resistance to hemoglobin catabolism: I) mQTL analysis demonstrated that elevated hemoglobin-derived peptides co-segregate with the CQR-encoding pfcrt locus (Figure 1 and Figure 3), II) levels of hemoglobin-derived peptides observed in parasite extracts predict chloroquine susceptibility in genetically diverse parasite strains from around the world (Figure 4), III) genetically identical parasite lines that differ only by CQR versus CQS isoforms of PfCRT recapitulate the peptide phenotypes observed in wild isolates (Figure S1 and Figure S2), and IV) forcing parasites to rely on hemoglobin as an amino acid source affects CQR parasites more severely than CQS parasites. In addition, we show that chloroquine resistant alleles affect the levels of 87 of the 362 observable peptides in the parasite' s hemoglobin catabolism pathway (Figure 5). Finally, we demonstrate that the peptide accumulation phenotype is associated with a 2 hour increase in cell cycle duration and a 13% overall fitness cost in transgenic parasites that only differ by their pfcrt allele. Together, these data argue that the significant fitness disadvantage observed in CQR parasites is attributable to impaired hemoglobin metabolism. These results provide the first molecular explanation for the reemergence of CQS parasites in wild populations following the cessation of chloroquine treatment. Parasites degrade ∼75% of the host cell hemoglobin over the course of their 48 hour intraerythrocytic developmental cycle [35], [36]. Any impairment in the hemoglobin digestion pathway directly affects I) the amino acid pool available for new protein synthesis [32], II) the osmotic stability of infected cells [37], and III) may reduce the number of developing merozoites that can fit within the physical confines of the infected erythrocyte. Any of these mechanisms could account for the increased cycle length and lower generation-to-generation fecundity we observed [37]. Although this is the first report of a metabolic perturbation inherent to CQR alleles of pfcrt, similar fitness-linked phenomena are associated with other drug resistance genes [13], [38]. Epidemiological analysis of the parasite population changes in Malawi and Hainan estimated the fitness cost of CQR to be ∼5% [39], [40]. Our in vitro competition studies support this conclusion and show that the fitness cost may actually be much higher in the absence of compensatory mutations (Figure 6). While PfCRT isoforms are clearly the main contributor to the peptide-accumulation phenotype, our data also show that genetic background modulates PfCRT' s effects on hemoglobin metabolism. The wide distribution of peptide levels observed across the HB3×Dd2 progeny (Figure 3), and the consistent secondary peaks observed in the mQTL analysis (Figure 1), suggest that loci other than PfCRT are contributing to peptide accumulation. Similarly, the CQR allele-exchange parasite lines (C4 and C6) both showed more extreme phenotypes than their respective parental lines (Dd2 and 7G8; Figure 4) despite having identical pfcrt sequences. The PfCRT/hemoglobin catabolism link we describe here, along with our peptidomics approach, provide a framework for investigating compensatory mutations elsewhere in the genome. Identifying the native function of PfCRT is a subject of considerable interest to the parasitology community. One possible interpretation of the peptide accumulation phenotype is that wildtype PfCRT functions as a peptide transporter and that CQR mutations interfere with this activity [41]. This interpretation is supported by a recent report of glutathione transport in Xenopus oocytes expressing PfCRT [42], and by work in Arabidopsis, which showed that a plant PfCRT homolog mediates glutathione transport [43]. However, the broad diversity of sizes (2–32mers) and physical properties of peptides accumulated by CQR parasites are inconsistent with the relatively narrow range of substrates carried by most peptide transporters [44]. An alternative interpretation of the peptide phenotype is that CQR-associated mutations affect hemoglobin catabolism indirectly by altering the permeability of the digestive vacuole membrane. Resistance mutations in PfCRT, and perhaps other membrane proteins, may cause the digestive vacuole to leak protons [45], glutathione [42], heme, or other osmolytes, which thereby alter the solution conditions of the vacuolar compartment. Given that protease activity can be very sensitive to solution conditions [46], even modest changes in vacuolar conditions could interfere with hemoglobin catabolism by the parasite. Similarly, perturbations in solution conditions may affect protein-protein interaction and thereby disrupt the recently described hemoglobin degradation complex [47]. In conclusion, this study demonstrates that chloroquine resistance, impaired hemoglobin catabolism, and reduced parasite fitness are linked to polymorphisms in PfCRT. This surprising linkage provides a molecular explanation for the reemergence of CQS parasites in Africa and Asia. Our results suggest that co-formulating chloroquine with a P. falciparum protease inhibitor [48] may be an effective strategy for slowing the emergence of resistant parasites. Synchronous parasites were grown using established methods [26] in RPMI 1640 supplemented with 25 mM HEPES, 100 µM hypoxanthine (all from Sigma), 10 µg/mL gentamycin (Gibco) and 2. 5 g/L Albumax II (Gibco). A total of 47 parasite strains were analyzed in this study; these strains include 36 lines from the HB3×Dd2 cross (34 progeny and 2 parental), three pfcrt allele swap lines (C2, C4, and C6) prepared in an isogenic GC03 background [11], and 8 out-group lines (V1/S, PAD, 7G8, GB4,3D7, D10, SL/D6, and 106/1) used to measure metabolic phenotypes in diverse genetic backgrounds. All cultures were maintained at 5% hematocrit in a 37°C incubator with an atmosphere of 5% CO2,6% O2, and 89% N2. Cultures were triple synchronized using consecutive treatments with 5% sorbitol at 0,48, and 56 h of culture. Invasion time was determined by preparing blood smears every two hours starting 34 h after the last sorbitol treatment. The zero time point was designated when cultures reached 95% rings. Samples were harvested at 24 h post invasion for the HB3×Dd2 cross study, 38 h for the parasite enrichment studies, and at several time points throughout the cyclic 48-hour asexual blood stage (12,24, and 36 h) for the out-group analysis. To confirm our PfCRT-related phenotypes and improve our analytical sensitivity, erythrocytes infected with late trophozoite-stage parasites (38 h post invasion) were separated from uninfected erythrocytes using an established density gradient method [49]. Briefly, bulk cultures were suspended at 30% hematocrit in RPMI. Cultures were layered over dual Percoll layers (70% lower, 30% upper) diluted in 1× RPMI (final concentration). Samples were centrifuged (2,000× g, 15 min) and the infected erythrocyte layer was collected from the 30%/70% interface. Infected cells were washed with 50 volumes of RPMI, then suspended at 0. 4% hematocrit in RPMI. Parasitemias of the purified samples were checked by blood smear and the purified samples were allowed to recover for 4 h in a 37°C incubator prior to metabolite extraction (Text S1 provides a step-by-step protocol). Our metabolite extraction protocol is adapted from a previously established method [50]. Metabolites were extracted by suspending 50 µL packed cells in 1 mL 4°C 90% methanol. Samples were vortexed and briefly sonicated, if necessary, to disrupt the cell pellet and generate a uniform homogenate. Homogenates were centrifuged (13,000× g, 10 min) and the supernatants were harvested. Samples were stored at −80°C as 90% methanolic extracts until metabolite analysis. Just prior to analysis, extracts were dried under a stream of N2 gas and resuspended in 200 µL H2O (Text S1 provides a step-by-step protocol). Metabolite extracts were analyzed by high performance liquid chromatography (HPLC) mass spectrometry (MS). The chromatographic conditions used in this study have been described in detail elsewhere [50]. Briefly, metabolites were separated by reverse phase C18 chromatography run over a 50 minute (HB3×Dd2 cross study and coelution assays) or 25 minute gradient (all other studies) using tributylamine as an ion pairing agent. General metabolite analyses were conducted using negative-mode electrospray ionization on a Thermo Scientific LTQ-Orbitrap (HB3×Dd2 cross) or Thermo Exactive (all other studies). Metabolite assignments were validated by single reaction monitoring (SRM) on a Thermo TSQ Quantum Discovery Max triple quad. For peptidomics analysis, aliquots of each metabolite extract were harvested, diluted 1∶4 in a 3% acetonitrile and 0. 1% formic acid solution (final concentrations), and analyzed in positive mode on an LTQ-Orbitrap using nanospray from a 120 minute hydrophilic interaction liquid chromatography (HILIC) gradient. Scans were conducted at both low (150–500 m/z) and high (450–1800 m/z) mass ranges to accommodate multiple charge states and MS2 scans were automatically conducted on fragments from each of the top seven signals observed at any given time. Both the original cross dataset and the peptidomic analyses were collected at the Princeton mass spectrometry facility; all other data were collected in house. Genome-wide scans were performed using pseudomarker [51] to detect QTLs associated with metabolite levels in the HB3×Dd2 genetic cross. Intensities of mass signals were log-transformed and the median signal for each mass across replicates was used as a phenotype. Batch number was included as an independent covariate [52] to correct for run-to-run changes in MS instrument sensitivity. Genome-wide significance thresholds were determined by permutation testing (n = 1000 permutations) [53] and the strength of each linkage was expressed as a LOD score. We accounted for multiple hypothesis testing using established methods [21]. Briefly, false discovery rates were calculated from p-values using the q-value approach [54]. QTL-based significance scores were used to filter the large untargeted mass list to a manageable subset of putative signals. Final LOD scores and significance thresholds were computed using R/qtl [55] (interval mapping parameters step = 1, n. draws = 64; QTL mapping method = hk). The custom R code (Text S2 and Text S3) and data tables used for this analysis (Table S2 and Table S3) are included in the supplemental materials. Mass data were visualized and analyzed using MAVEN, a freely available software package for MS-based metabolomics [56]. Data were peak picked using a permissive threshold (S/N = 3) and raw LOD scores generated by QTL mapping were then used to identify the most promising subset of signals. The extracted ion chromatogram of each signal with a LOD score greater than 3 was visually inspected and data originating from peak picking errors, thermal noise, elution artifacts, or associated with the void and wash volumes were excluded. Coeluting adducts, fragments, and isotopomers were condensed into their respective parent masses and the intensities for each of the final parent masses were hand-verified to correct for peak picking errors. A list of potential compound identities was generated by searching the Madison Metabolomics Consortium Database [27] and the Human Metabolome Database [28] for metabolites matching the QTL-identified masses. Putative IDs were evaluated by ms/ms fragmentation analysis. Final compound identities were confirmed by coelution of the parasite-derived compounds with standards purchased from Sigma and the Proteomics Resource Center at Rockefeller University. The final assignment and quantification steps were conducted by single reaction monitoring (SRM) on a triple quadrupole mass spectrometer. Peptide assignments were conducted using a comprehensive hemoglobin digestion library loaded into Mascot proteomics software (Matrix Science). Only assignments with mass defects of less than 10 ppm, matching scores greater than nine, and observable peptide signals in all nine of the extracts (N = 3 per genotype) were included. All assignments based on parent masses that mapped to adducts or fragments of hemoglobin peptides were excluded. A custom MAVEN-format standards library was generated using the Mascot results and the extracted ion chromatogram of each assignment was visually inspected and quantified in MAVEN. Peak intensities for each peptide were compiled and aligned to both the hemoglobin α and β primary sequences using custom software written in R. Synchronous cultures of isogenic pfcrt allele exchange parasites (C2, C4, C6) were grown to the late trophozoite phase and magnetically purified from uninfected cells using a MACS column. The parasitemia of each enriched sample was determined by microscopy and cell counts were determined by hemocytometer. Mixed culture flasks containing either two (C2, C4; C2, C6) or three (C2, C4, C6) genotypes were constructed at mixing ratios of 50∶50 (C2/C4, C2/C6), 25∶75 (C2/C6), or 50∶15∶35 (C2, C4, C6). Each two-way competition experiment was run as a single biological replicate whereas the three-way competition was replicated in three independent flasks (established from a single seed culture) run in parallel. The entire experimental procedure was repeated a second time using asynchronous populations of parasites. Flasks were maintained continuously for 70 days under standard culturing conditions. Culture flasks were cut back 1∶10 every 48 h (to ∼0. 5% parasitemia) and DNA was harvested from the excess parasites via Saponin lysis (0. 1%) followed by genomic DNA isolation (DNeasy kit, Qiagen). The pfcrt allele present in each sample was PCR amplified (primers: CGAGCGTTATAGAGAATTAG, ACAACATCACCGGCTAAGAA). Products were then Sanger sequenced using independent diagnostic primers (GGCTCACGTTTAGGTGGAGG, ACAACATCACCGGCTAAGAA). Sequencing results were analyzed using online tools from Genewiz and allelic abundances in each flask were quantified using diagnostic single nucleotide polymorphisms in PfCRT amino acid positions 74–76, and 98 (C2: ATG AAT AAA AAC, C4: ATT GAA ACA AGC, C6: ATG AAT ACA GAC, Figure S6). Allele frequencies observed in long-term competition experiments were fit using a custom model of P. falciparum in vitro growth. All modeling and regression analyses were conducted using custom software written for the R statistical software environment. Our model makes the following assumptions: 1) long-term changes in allele frequencies follow exponential kinetics, 2) parasite clones can differ with respect to life cycle length, 3) DNA abundance follows a sigmoidal accumulation over the life cycle with peak DNA synthesis occurring mid lifecycle, 4) most of the DNA synthesized in one generation is not amplified in the following generation because not all merozoites successfully reinvade, 5) parasite life cycle synchronicity follows a Gaussian distribution that becomes progressively broader with each generation. Using these assumptions, the relative DNA content expected in mixed culture flasks was modeled for each point in the 70 day competition experiment. Initial differences in allele frequencies were set according to the empirical mixing ratio, then life cycle lengths and exponential growth rates were sampled by grid search. A best-fit multiple regression model was identified by iterative grid searches with progressively finer increments of cycle lengths and growth rates. The custom R code (Text S2 and Text S3) and data table (Table S4) used for this analysis is provided in the supplemental materials.

Chloroquine was formerly a front line drug in the treatment of malaria. However, drug resistant strains of the malaria parasite have made this drug ineffective in many malaria endemic regions. Surprisingly, the discontinuation of chloroquine therapy has led to the reappearance of drug-sensitive parasites. In this study, we use metabolite quantitative trait locus analysis, parasite genetics, and peptidomics to demonstrate that chloroquine resistance is inherently linked to a defect in the parasite' s ability to digest hemoglobin, which is an essential metabolic activity for malaria parasites. This metabolic impairment makes it harder for the drug-resistant parasites to reproduce than genetically-equivalent drug-sensitive parasites, and thus favors selection for drug-sensitive lines when parasites are in direct competition. Given these results, we attribute the re-emergence of chloroquine sensitive parasites in the wild to more efficient hemoglobin digestion.

Abstract Introduction Results Discussion Materials and Methods

parastic protozoans genetics of disease biochemistry protein metabolism parasite evolution genetics protozoology biology microbiology plasmodium falciparum metabolism parasitology

2014

Metabolic QTL Analysis Links Chloroquine Resistance in Plasmodium falciparum to Impaired Hemoglobin Catabolism

Mycobacteria of the Mycobacterium tuberculosis complex (MTBC) greatly affect humans and animals worldwide. The life cycle of mycobacteria is complex and the mechanisms resulting in pathogen infection and survival in host cells are not fully understood. Recently, comparative genomics analyses have provided new insights into the evolution and adaptation of the MTBC to survive inside the host. However, most of this information has been obtained using M. tuberculosis but not other members of the MTBC such as M. bovis and M. caprae. In this study, the genome of three M. bovis (MB1, MB3, MB4) and one M. caprae (MB2) field isolates with different lesion score, prevalence and host distribution phenotypes were sequenced. Genome sequence information was used for whole-genome and protein-targeted comparative genomics analysis with the aim of finding correlates with phenotypic variation with potential implications for tuberculosis (TB) disease risk assessment and control. At the whole-genome level the results of the first comparative genomics study of field isolates of M. bovis including M. caprae showed that as previously reported for M. tuberculosis, sequential chromosomal nucleotide substitutions were the main driver of the M. bovis genome evolution. The phylogenetic analysis provided a strong support for the M. bovis/M. caprae clade, but supported M. caprae as a separate species. The comparison of the MB1 and MB4 isolates revealed differences in genome sequence, including gene families that are important for bacterial infection and transmission, thus highlighting differences with functional implications between isolates otherwise classified with the same spoligotype. Strategic protein-targeted analysis using the ESX or type VII secretion system, proteins linking stress response with lipid metabolism, host T cell epitopes of mycobacteria, antigens and peptidoglycan assembly protein identified new genetic markers and candidate vaccine antigens that warrant further study to develop tools to evaluate risks for TB disease caused by M. bovis/M. caprae and for TB control in humans and animals. Mycobacterium tuberculosis has infected more than 2. 5 billion people worldwide with approximately 9 million new tuberculosis (TB) cases reported every year [1]. Animal TB is caused by infection with Mycobacterium bovis and closely related members of the M. tuberculosis complex (MTBC) such as M. caprae. Although cattle are the main concern regarding animal TB in industrialized countries, several other mammals including humans are also infected [2,3]. Eurasian wild boar (Sus scrofa) are a natural reservoir for M. bovis in some regions and thus vaccination strategies are being developed for TB control in this species [4–8]. Several other domestic and wild animals are also infected with M. bovis and may act as reservoir species [9–13]. The life cycle of mycobacteria is complex and the mechanisms resulting in pathogen infection and survival in host cells are not fully understood [14]. Nevertheless, it is generally accepted that after inhalation into the lung or entry to the oropharyngeal cavity, the principal entry routes, mycobacteria of the MTBC are phagocytized by macrophages, which constitute the main host cell. As with other intracellular bacteria, mycobacteria survive inside macrophages by escaping host immune response, which results in the formation of a granuloma that effectively contains infected cells. A change in the host-bacterial equilibrium of granulomas is thought to result in the release of infected cells outside containment and onward transmission of mycobacteria to susceptible hosts [14]. The association between M. bovis spoligotypes and TB lesions in cattle has been used to correlate bacterial genotype with virulence [15]. However, these genotyping methods cover only a small portion of the approximately 4,000 genes contained in the 4. 4 Mb mycobacterial genome [16]. Recently, whole-genome sequencing and comparative genomics analyses have provided new insights into the evolution and adaptation of the MBTC to survive inside the host and explained phenotypic traits related with transmissibility and virulence [16–21]. Although the first M. tuberculosis genome sequence was reported in 1998 [16], it was not until 2003 when the first M. bovis genome was sequenced [19]. Presently, a large number of M. tuberculosis but few M. bovis (except for BCG strains) genome sequences are available [18]. The relative paucity of M. bovis genome sequence information limits the possibility of characterizing mycobacterial evolution and correlation with virulence at the whole-genome level. In this study, the genome of three M. bovis (MB1, MB3, MB4) and one M. caprae (MB2) field isolates with different lesion score, prevalence and host distribution phenotypes were sequenced. Genome sequence information was used for whole-genome and protein-targeted comparative genomics analysis with the aim of finding correlates of phenotypic variation with potential implications for TB disease risk assessment and control. All animal sampling was post-mortem. Wildlife samples came from hunter-harvested individuals that were shot during the legal hunting season independently and prior to our research while livestock samples were obtained at the slaughterhouse where they were being processed as part of the normal work and submitted to the reference laboratory by the slaughterhouse veterinarian. According to EU and National legislation (2010/63/UE Directive and Spanish Royal Decree 53/2013) and to the University of Castilla–La Mancha guidelines no permission or consent is required for conducting this type of study. Field isolates used come from the EU Reference Laboratory for Animal Tuberculosis (VISAVET). Three M. bovis (MB1, MB3, MB4) and one M. caprae (MB2) field isolates were selected for this study (Table 1). These isolates were originally obtained from wild boar (MB3, MB4), cattle (MB1) and goat (MB2). The study focused on Ciudad Real Province, Spain. This is a high ungulate density area, the west side of the province composed by interspersed game ranges and protected nature areas, with persistent TB infection in extensive livestock farms [22]. Nine hundred MTBC isolates collected from wild ungulates and livestock from 2000 to 2011 were spoligotyped resulting in 62 different spoligotypes ([23] and S1 Fig). The criteria for selection of the MTBC spoligotypes included in the study were based on: Following these three criteria, the M. bovis spoligotypes SB0339 (MB1, MB4) and SB0134 (MB3) were selected. The M. caprae spoligotype SB0157 (MB2) was included in the study as an outgroup but closely related species [29] and for the increased proportion of M. caprae isolated from bovine samples during 2004–2009 [30]. The MB4 isolate was included because although it has the same spoligotype as MB1, it served as a model to characterize possible differences between MTBC isolates otherwise classified with the same spoligotype. The four isolates were grown in 15 ml of Middlebrook 7H9 liquid media supplemented with 0. 36% sodium pyruvate and 10% OADC (Oleic Albumin Dextrose Catalase) for 5 weeks. Chromosomal DNA samples were obtained as described by van Soolingen et al. [31]. Briefly, cultures were centrifuged and pellets were washed twice in 5 ml water. Mycobacteria were heated at 100°C for 15 min to kill the cells. After centrifugation, the cells were resuspended in 5 ml TE buffer (0. 01 M Tris-HCl, 0. 001 M EDTA, pH 8. 0). Lysozyme was added to a final concentration of 1 mg/ml and the tube was incubated over night at 37°C. Eighty hundred and seventy-five microliters of 10% sodium dodecyl sulfate (SDS) with 62. 5 μl of proteinase K (at a 10-mg/ml concentration) were added, and the mixture was incubated for 1 h at 60°C. The extract was transferred to a phase lock gel tube (prime5, Fisher Scientific SL, Madrid, Spain) for a phenol/chloroform DNA extraction. Genomic DNA (2–5 μg) was subjected to mechanical fragmentation using a BioRuptor (Life Technologies, Carlsbad, CA, USA). The number of cycles was adjusted to obtain DNA fragments of a final average size of about 500 pb. Samples were then used to prepare sequencing-amenable TruSeq libraries (NEB-Next, New England Biolabs, Ipswich, MA, USA). Briefly, DNA fragments were made blunt-ended, phosphorylated, adenylated and Illumina-compatible adapters were ligated. After purification, barcoded sequences as well as Illumina-specific sequences were introduced by PCR, followed by quantitation of individual libraries. Libraries were then pooled and quantified again. A quality control of the pooled library made in bioanalyzer is shown in the figure, including an estimation of the percentage of non-overlapping reads that could be obtained using a 2x250 paired-end sequencing protocol. Library was qPCR-quantitated and brought to a final concentration of 10 nM. DNA was then denatured and equilibrated so that a final concentration of 18 pM of library was loaded onto a MiSeq v. 3 flowcell (Illumina, San Diego, CA, USA) and sequenced using a 2x250 paired-end sequencing protocol to obtain more than 400x high quality coverage (1. 9–2. 5 Gb) with 84% of the bases showing a Q30 factor > 30. Reads were finally split according to barcodes and used for bioinformatics analysis. High quality overlapping reads were merged using FLASH (Magoc et al. , 2011) and then assembled using Velvet [32] with k-value = 97 (S1 Table). Contigs were annotated using BG7 [33,34] (S2 Table). For annotation, a set of 191,017 reference proteins was used including (a) all Uniprot proteins from M. bovis and M. tuberculosis, (b) a set of bacterial antibiotic resistance related Uniprot proteins selected using the GO annotation terms “antibiotic resistance” and “response to antibiotic” and a selection of proteins based on similarity to the proteins of ARDB [35], and (c) all Uniprot proteins with Enzyme Code (EC) from MTBC. For whole genome comparative analysis, the 4 genomes were then aligned against the M. bovis reference genome sequence (AF2122/97; http: //www. ncbi. nlm. nih. gov/nuccore/31791177) using Differences program (Turrientes et al. , 2010) that allows comparisons at the whole genome level and particularly the detection of substitutions, insertions or deletions of any length and at any region of the genome. Genome sequence information and annotation was deposited in GenBank under the accession numbers CDHF01000001-CDHF01000049, CDHG01000001-CDHG01000059, CDHH01000001-CDHH01000094 and CDHE01000001-CDHE01000118 for MB1-MB4 isolates, respectively [36]. A phylogenetic tree was constructed based on the SNPs found at the core genome sequence shared with a similarity over a threshold between all genomes included in the analysis using Harvest [37] and visualized with EvolView (v 198. 3) [38]. Harvest defines the SNPs aligning whole assembled genomes and not reads like in other approaches, thus allowing the identification of the gene or the intergenic region where differences are allocated. The SNPs are included in the. vcf file provided by Harvest suite. The genomes used in the SNP phylogenetic analysis using Harvest included MB1-MB4, M. bovis AF2122/97 (http: //www. ncbi. nlm. nih. gov/nuccore/31791177), M. bovis ATCC BAA-935 (http: //www. ncbi. nlm. nih. gov/nuccore/690294709), M. bovis BCG Pasteur 1173P2 (http: //www. ncbi. nlm. nih. gov/nuccore/121635883), and M. tuberculosis H37Rv (http: //www. ncbi. nlm. nih. gov/nuccore/448814763). The multilocus sequence analysis was conducted using 18 genes coding for the proteins linking stress response with lipid metabolism (S3 Table). The nucleotide sequences of the genes were obtained from the genomes of MB1-MB4 isolates used in this study. For comparison, the same sequences were obtained from the reference M. bovis BCG Pasteur 1173P2, M. bovis AF2122/97 and M. tuberculosis H37Rv. M. canettii (NCBI reference sequence NC_019950) was used as outgroup. The nucleotide sequences were concatenated and then aligned with MAFFT (v7), configured for the highest accuracy [39]. After alignment, regions with gaps were removed and 20080 gap-free sites were used in maximum likelihood phylogenetic analysis as implemented in PhyML (v3. 0 aLRT) [40,41]. The reliability for the internal branches was assessed using the approximate likelihood ratio test (aLRT–SH-Like) [41]. Graphical representation and editing of the phylogenetic trees were performed with EvolView (v 198. 3) [38]. The alignments obtained by MAFFT were used to perform codon alignments using HIV database website (www. hiv. lanl. gov; [42]). Non-synonymous (dN) and synonymous (dS) nucleotide substitutions were classified using the SNAP method [43] implemented in HIV database website [42]. SNPs were identified by pairwise comparison of MB1-MB4 and M. bovis AF2122/97 using SNAP. The dN/dS ratio was calculated for the 18 genes coding for the proteins linking stress response with lipid metabolism (S3 Table) and for 81 antigen-coding genes (S4 Table) present on each of the MB1-MB4 isolates included in the study. Under the Datamonkey server (http: //www. datamonkey. org; [44]), the algorithm SLAC [45] was used to detect which nucleotide substitution site were positively or negatively selected. For each dS and non-synonymous dN substitution site, four measurements were made: normalized expected (ES and EN) and observed numbers (NS and NN). The SLAC algorithm then calculated: dN = NN/EN and dS = NS/ES. If dN < dS a codon was negatively selected and if dN > dS a codon was positively selected. A two-tailed extended binomial distribution set at P<0. 05 was used to assess significance of the algorithm. The SLAC algorithm uses a neighbor-joining tree with a maximum likelihood for branch lengths and substitution rates. To identify non-synonymous mutations that may be associated with virulence and/or transmission, the orthologs of peptidoglycan assembly protein locus Rv0050 (H37Rv) in strains MB1-MB4 were compared to the equivalent locus in animal isolates (BCG Pasteur 1173P2, AF2122/97,09–1192) and human isolates (Bz 3115, B2 7505) of M. bovis available in the GenBank. The M. bovis strains were selected among all strains for having distinct Rv0050 locus and/or distinct source of isolation. The presence of a putative signal peptide and cleavage site in the Rv0050 locus was analyzed with a previously validated program for their prediction in M. tuberculosis (SignalP; http: //www. cbs. dtu. dk/services/SignalP-3. 0/) [46]. Results were also confirmed with two other programs, Signal-Blast (http: //sigpep. services. came. sbg. ac. at/signalblast. html) and Phobious (http: //phobius. sbc. su. se/). To confirm selected SNPs identified in the mycobacteria genomes sequenced in this study, sequence-specific oligonucleotide primers were design using reference genomes M. bovis AF2122/97 (BX248333. 1) or M. tuberculosis H37Rv (AL123456. 3) for PCR and sequencing of the amplicons. Selected loci and direct and reverse primers used for analysis included Rv0050 (ponA1; 5´-GACTTTCCCCAAACCGACCGAGG-3’ and 5’-GATCGGTCCCCCGACCACCATT-3’), Rv0589 (MCE2a; 5´-GTGCCAACGCTGGTGACGAG-3’ and 5’-AGAACACGATCAACCCATGA-3’), Rv1198 (ESAT-6; 5´-ATGACCATCAACTATCAATT-3’ and 5’-TCGGCTCCAGCTGGGCCTGA-3’), and Rv1860 (FAP-B; 5´-ATGCATCAGGTGGACCCCAA-3’ and 5’- AGCGGACCTTACCGGCCTGA-3’). The PCR was conducted using 2 μl of DNA with 20 pmol of each primer in a 50 μl reaction PCR Master Mix (Promega, Madison, WI, USA) using a GeneAmp PCR System 2700 thermocycler (Applied Biosystems, Carlsbad, CA, USA). PCR products were electrophoresed on 1. 5% agarose gels to check the size of the amplified fragments by comparison to molecular weight marker GeneRuler 1kb DNA ladder (Thermo Scientific, Waltham, MA, USA). Amplified DNA fragments were purified with a PureLink Quick PCR Purification Kit (Thermo Scientific, Waltham, MA, USA) and sequenced using the reverse primer on each locus. Amplicons from at least two independent PCR reactions were sequenced. Among the 900 MTBC field isolates analyzed, 62 different spoligotypes were identified suggesting that the study area is one of the regions with the highest diversity of MTBC spoligotypes described in the literature [47]. The high genetic diversity of MTBC in this area supported an important natural scenario where MTBC and particularly M. bovis have diversified, thus offering an interesting epidemiological and evolutionary context where new genotypes can emerge and diversify in terms of adaption to host under a range of environmental and human driven factors. The spoligotypes SB0121, SB0134, SB0339, SB0120, SB1263 and SB0157 were among the most frequent ones found in both livestock and wild ungulates in the study area (S1 Fig). The spoligotypes found at a higher frequency than expected given a predicted mutation rate were identified (Fig 1A) and represented in a hierarchical tree (Fig 1B). The spoligotypes SB0134 (MB3) and SB0339 (MB1 and MB4), which clustered in space and time and showed low mutation rates but high abundance, were selected as emergent under study conditions. The output hierarchical tree suggested a history of mutation events and a relationship between these spoligotypes with different levels of delection for spoligotypes SB0134 (MB3) and SB0339 (MB1 and MB4) (Fig 1B). These spoligotypes were therefore chosen for this study as fulfilling selection criterion (i) and (ii) described above. Additionally, findings in Iberian red deer showed a higher lesion score caused by spoligotype SB0134 when compared to spoligotype SB0339 (Mann-Whitney U test; p = 0. 04) while these spoligotypes and particularly SB0134 and SB0157 suggested an association with high TB severity in Eurasian wild boar thus also fulfilling selection criteria (iii) (Fig 1C). In summary, three M. bovis (MB1, MB3, MB4) and one M. caprae (MB2) field isolates with different prevalence, lesion score and host distribution phenotypes were selected for this study (Table 1). MB3 showed high distribution and lesion score while MB1 and MB4 were highly distributed but with low lesion score. The M. caprae (MB2) isolate had moderate distribution and high lesion score and was selected for comparison with the M. bovis isolates. These phenotypic variations are relevant for pathogen transmission and virulence and could be correlated with genome sequence information with implications for TB disease risk assessment and control. The results of the phylogenetic analysis showed that the MB1 and MB4 isolates with the same spoligotype were the most closely related isolates (Figs 2A and S2). The MB2 isolate (M. caprae) clustered separately from the other isolates, which clustered together with M. bovis sequences. Nevertheless, M. caprae (MB2) was closely related to M. bovis when compared to M. tuberculosis (Fig 2A). The genome of M. bovis BCG Pasteur 1173P2 was the most similar to all sequenced mycobacteria genomes (Figs 2A, 2B and 3). It has been proposed that during evolution, a clone of M. tuberculosis that was originally adapted to cause human TB evolved to infect a non-human mammal and thus began the transition into non-human ecotypes such as M. bovis, which in turn spread to cattle, goats, oryx, seals and pigs [20]. Our results supported a close relationship between M. bovis isolates and suggested that M. caprae is one of the M. bovis-related mycobacteria adapted to infect goats and sheep as well as other hosts such as wild boar, red deer, cattle and humans [29,48,49]. To better understand the relationship between these isolates, a comparative genomics approach was used. The results showed the presence of translocations, deletions of small genomic regions and SNPs between genomic sequences (Figs 2B, 3 and S2). Large-scale polymorphism studies have demonstrated that the MTBC shows a large number of deletions of small genomic regions consistent with the reductive evolution typical of intracellular bacteria [20]. Nevertheless, sequential chromosomal nucleotide substitutions are considered to be the main driver in the M. tuberculosis genome evolution [20]. The results reported here supported these findings for M. bovis isolates and suggested that the isolates with high lesion score, MB2 and MB3, contained the largest number of polymorphisms when compared to the MB1 and MB4 isolates with low lesion score (Figs 2B and 3). However, a clear correlation between phenotype and genome sequence requires a protein-targeted comparative analysis between the different isolates. For protein-targeted analysis, the study was focused on proteins that are known to play an important role in mycobacterial viability or virulence. Protein-targeted comparative analysis was conducted for (a) the ESX or type VII secretion system, (b) proteins linking stress response with lipid metabolism, (c) host T cell epitopes of mycobacteria, (d) antigens and (e) peptidoglycan assembly protein to define possible correlates with bacterial virulence and viability or distribution.

Mycobacteria belonging to the Mycobacterium tuberculosis complex infect humans and animals since pre-history and are a serious health problem worldwide. Whole-genome sequencing and comparative genomics generate information on the evolution and molecular basis of pathogenicity and transmissibility. However, while genomic information is increasingly available for the main human pathogens such as Mycobacterium tuberculosis, little is known about closely related bacteria, Mycobacterium bovis and Mycobacterium caprae. These mycobacteria infect humans causing zoonotic tuberculosis and are the main causative agents of animal tuberculosis. Although human-to-human transmission of zoonotic tuberculosis is limited, the infection often causes extra-pulmonary disease in humans and is still a major public health concern in developing countries, causing not only human disease but also severe effects on livelihoods. In this study, whole-genome sequences and targeted comparative genomics of three Mycobacterium bovis and one Mycobacterium caprae field isolates generated new information on the evolution and phenotypic variation of these mycobacteria. The results identified new genetic markers and candidate vaccine antigens that warrant further study to develop tools to evaluate risks for tuberculosis caused by M. bovis/M. caprae and for disease control in humans and animals.

Abstract Introduction Materials and Methods Results and Discussion

2015

Comparative Genomics of Field Isolates of Mycobacterium bovis and M. caprae Provides Evidence for Possible Correlates with Bacterial Viability and Virulence

The contribution of individuals with subclinical infection to the transmission and endemicity of cutaneous leishmaniasis (CL) is unknown. Immunological evidence of exposure to Leishmania in residents of endemic areas has been the basis for defining the human population with asymptomatic infection. However, parasitological confirmation of subclinical infection is lacking. We investigated the presence and viability of Leishmania in blood and non-invasive mucosal tissue samples from individuals with immunological evidence of subclinical infection in endemic areas for CL caused by Leishmania (Viannia) in Colombia. Detection of Leishmania kDNA was conducted by PCR-Southern Blot, and parasite viability was confirmed by amplification of parasite 7SLRNA gene transcripts. A molecular tool for genetic diversity analysis of parasite populations causing persistent subclinical infection based on PCR amplification and sequence analysis of an 82bp region between kDNA conserved blocks 1 and 2 was developed. Persistent Leishmania infection was demonstrated in 40% (46 of 114) of leishmanin skin test (LST) positive individuals without active disease; parasite viability was established in 59% of these (27 of 46; 24% of total). Parasite burden quantified from circulating blood monocytes, nasal, conjunctival or tonsil mucosal swab samples was comparable, and ranged between 0. 2 to 22 parasites per reaction. kDNA sequences were obtained from samples from 2 individuals with asymptomatic infection and from 26 with history of CL, allowing genetic distance analysis that revealed diversity among sequences and clustering within the L. (Viannia) subgenus. Our results provide parasitological confirmation of persistent infection among residents of endemic areas of L. (Viannia) transmission who have experienced asymptomatic infection or recovered from CL, revealing a reservoir of infection that potentially contributes to the endemicity and transmission of disease. kDNA genotyping establishes proof-of-principle of the feasibility of genetic diversity analysis in previously inaccessible and unexplored parasite populations in subclinically infected individuals. Asymptomatic dermal or visceral leishmaniasis (VL) constitute a variable and sometimes high proportion of the naturally exposed population in endemic foci of Leishmania transmission, ranging from 17 to 91% of incident infections [1–3]. Although xenodiagnosis has shown that sand flies can acquire infection from asymptomatic dogs in different settings of L. infantum transmission [4–6], and even from vaccinated dogs [7], the epidemiological impact of asymptomatic infection in the transmission of leishmaniasis is unknown. Parasite persistence and viability have been demonstrated after treatment and clinical resolution of disease [8,9], supporting the possibility that subclinically infected individuals can accumulate to constitute an important, unrecognized proportion of the population in endemic foci. Prospective population-based studies of the natural history of cutaneous leishmaniasis (CL) in the Pacific coast and North-central regions in Colombia, and the Peruvian Andes showed that leishmanin skin test (LST) reactivity, and presence of scars compatible with history of CL, were risk factors for development of new active lesions [1,3, 10]. Hence, re-activation of prior infection, whether clinically apparent or subclinical, is a potential source of incident disease. Attention has only recently focused on understanding host, parasite, entomological and epidemiological determinants of subclinical infection in order to enlighten the development of strategies for disease prevention and control [2]. Consensus criteria for subclinical or asymptomatic human infection are currently unavailable primarily due to lack of a means to differentiate immunologically sensitizing exposure to Leishmania, from persistent infection. Screening for delayed type hypersensitivity response to Leishmania antigen or in vitro expansion of memory T cells, and serological reactivity (in the case of asymptomatic VL) are used to define the infected population in endemic settings. However, immunological reactivity may not be indicative of a persistent infection. Parasitological demonstration of asymptomatic infection has not been achieved in the context of endemic exposure to transmission of cutaneous leishmaniasis. Access to parasitological evidence of infection and quantitative data on infection may allow modeling to make projections of potential impact [11] and assessment of the epidemiological contribution of persistent subclinical infection (following asymptomatic infection or successful treatment of symptomatic infection) to transmission. Detection of parasite DNA and RNA by molecular methods have demonstrated respectively, the presence and viability of Leishmania after clinical cure [8,9], and in clinically normal mucosal tissues and peripheral blood monocytes during active disease [12]. To date, the development and use of molecular detection methods has been focused on determining the presence of Leishmania, and quantifying parasite burden in clinical samples. Robust molecular tools for in-depth analysis of Leishmania genome/transcriptome or genetic diversity such as next generation sequencing and multilocus microsatellite typing (MLMT) require isolation of, or abundant parasites in tissue samples for adequate performance and in order to obtain informative results. These requirements have impeded the analysis of samples with low parasite burden such as those from subclinically infected individuals. PCR-based methods targeting polymorphic high copy number coding or non-coding DNA sequences including kDNA, GP63, and cysteine peptidase b, among others [13], have partially overcome this impediment. However, standardized side-to-side comparisons between novel and validated methods for diversity analysis including MLMT or isoenzyme typing have not been conducted. This investigation sought to demonstrate the presence and viability of Leishmania in blood and non-invasive mucosal tissue samples from LST positive individuals without active disease from areas endemic for the transmission of Leishmania species of the Viannia subgenus, and to design a molecular tool for genetic diversity analysis of Leishmania involved in subclinical infection. Parasitological demonstration of subclinical infection in individuals residing in foci of transmission, and the feasibility of molecular characterization of Leishmania populations found in persistent subclinical infection, provide bases to evaluate the contribution of the human population in the persistence and dissemination of CL. This study was approved and monitored by the Institutional Review Board for Ethical Conduct of Research Involving Human Subjects of the Centro Internacional de Entrenamiento e Investigaciones Médicas (CIDEIM) with approval code CIEIH1104, in accordance with national and international guidelines. All individuals voluntarily participated in the study. Written informed consent was obtained from each participant over 18 years of age. For children, informed consent was signed by the accompanying parent, and assent was signed by children over 7 years of age. This descriptive study was conducted in two phases (Fig 1). An initial exploratory phase (Phase 1) was performed to determine the feasibility of molecular detection of persistent subclinical infection, which evaluated the presence of Leishmania kDNA in blood and mucosal swab samples from members of five family households of rural areas of Tumaco (Nariño, Colombia), endemic for infections caused by L. (Viannia) species. Families/households were selected based on the presence of one household member (index case) with a parasitologically confirmed prior history of active disease as well as presence of compatible scars and positive leishmanin skin test reaction (LST), at least one co-habitant with asymptomatic infection based on LST reactivity in the absence of active or healed lesions, and a minimum of four co-habitants. In the second phase a larger scale assessment was designed to 1) establish the demonstrable presence of Leishmania in immunologically defined asymptomatic infection, and 2) to evaluate the persistence, viability and genotypic diversity of Leishmania among asymptomatically infected individuals and those with subclinical infections following prior symptomatic CL. This study was conducted in areas of endemic transmission of cutaneous leishmaniasis caused by L. (Viannia) species in the departments of Nariño and Risaralda in Colombia. Following LST evaluation, the presence of parasites was assessed by PCR-Southern Blot for Leishmania kDNA in blood monocytes, aspirates of lesion scars, and swab samples from nasal, tonsil, and conjunctival mucosa [14]. Parasite viability was determined by qRT-PCR of the Leishmania 7SLRNA transcript [12], and genetic diversity of Leishmania was assessed by genotyping of the conserved region of the kDNA as described below. 184 subjects participated in this study. Phase 1 comprised a total of 30 individuals of which 25 were residents of endemic areas from rural communities of the municipality of Tumaco, department of Nariño and 5 were residents of non-endemic areas. Study groups were defined as follows: 1) Individuals with clinical history of CL (n = 5) confirmed by a typical scar of CL [1] and positive LST result, 2) asymptomatic infection (n = 15), defined as residents of an endemic area of transmission of dermal leishmaniasis, having a positive LST and no evidence or history of dermal lesions or typical scars; and 3) healthy LST negative participants without history of leishmaniasis from endemic (n = 5) or non-endemic areas (n = 5) (Fig 1). Phase 2 included 154 subjects: 103 participants were residents of a rural community in Pueblo Rico, Risaralda and 51 participants resided in rural communities of the municipality of Tumaco, department of Nariño. Study groups were as follows: 1) Individuals with clinical history of CL (n = 116), 2) individuals with asymptomatic infection (n = 17), 3) healthy LST negative participants from endemic areas (n = 18), and 4) individuals with active parasitologically confirmed CL (n = 3) that served as positive controls (Fig 1). Blood monocytes, lesion/scar aspirates and duplicate swab samples from tonsil, conjunctiva and nasal mucosa were obtained from the study participants on one occasion. Duplicate samples were obtained for the purpose of independent RNA and DNA extraction procedures. Blood monocytes were separated from a 10 mL sample of peripheral blood using 1-Step Monocytes following the manufacturer´s protocol (Accurate Chemical & Scientific Co.). Monocytes and lesion/scar aspirate samples were stored in TRIzol Reagent at -70°C until processing. RNA and DNA were extracted from blood monocytes using the AllPrep DNA/RNA Minikit (Qiagen). Mucosal swab samples were stored at -20°C until processing, and DNA extracted from one of the two replicate swabs as previously reported [12,14]. RNA was extracted from the second swab sample using TRIzol followed by RNA cleanup using the RNeasy extraction Kit (Qiagen). Purified RNA was treated with DNAse I. All RNA samples were suspended in a final volume of 35μl of nuclease free water. Quantity and quality of nucleic acids were evaluated using a NanoDrop2000 spectrophotometer. Reference strains and clinical strains isolated from patients with CL were obtained from the CIDEIM BioBank (S1 Table). All strains were previously typed by monoclonal antibodies and/or isoenzyme electrophoresis. Promastigotes were maintained at 26°C in RPMI medium supplemented with 10% heat-inactivated foetal bovine serum (Gibco), 1% glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin. Logarithmic phase promastigotes were harvested by centrifugation, washed in phosphate-buffered saline (PBS), and solubilized in lysis buffer for DNA extraction. For the analyses of genetic diversity we also included kDNA sequences obtained from NCBI Genbank as summarized in S2 Table. A 242 bp product of the human GAPDH gene was targeted for amplification from all samples as a quality control procedure, using the primers Fw (5’- CTG GCC CTC TGC CCT CCT ACC A -3’) and Rv (5’- TTC CAT CCA GCC TGG GGC GAA -3’). L. (Viannia) minicircle kDNA was amplified from 100ng of DNA samples by PCR using the LV-B1 primers followed by Southern blot hybridization as previously described [15]. kDNA positive samples were evaluated by real time reverse transcriptase PCR to confirm the viability of parasites and estimate parasite burden using the Leishmania 7SLRNA transcript as previously described [12] and using 10μl of the RNA sample. The single copy gene coding for the human TATA Box Binding protein (TBP) was amplified for quantitation of human nucleated cells. Amplification was performed with the primer set TBP Fw (5’- CAC GAA CCA CGG CAC TGA TT -3’) and TBP Rv (5`- TTT TCT TGC CAG TCT GGA -3’). PCR efficiency was determined in each individual run by inclusion of purified L. (V). panamensis DNA as a positive control. Potential DNA carry-over or contamination was evaluated by inclusion of blank (water) and negative control (DNA of PBMCs from a healthy donor) samples. Standard curves for quantitation of parasite and human nucleated cells were constructed by ten-fold serial dilution of cDNA products obtained from 1x107 L. (V.) panamensis promastigotes and from the human U-937 promonocytic cell line (1 x 107 cells), respectively. Specificity of the qRT-PCR products was assessed by analysis of the melt peak curve. Parasite burden was calculated by extrapolation to a standard curve and normalized to the number of human cells based on TBP expression. Real time detection of amplification products was performed using SYBR Green Master Mix (Applied Biosystems) on a BioRad CFX-96 detection platform. For those kDNA positive samples that were below the limit of detection of the 7SLRNA qRT-PCR, a maximum likelihood estimate of 0. 0001 parasites per reaction (0. 00357 parasites per swab) was calculated based on the assumption of 10,000 minicircle kDNA copies and 250 copies of the 7SLRNA transcript per organism [16,17], and the samples volumes for each assay [18]. Leishmania strains and kDNA positive samples from study participants were processed for genetic diversity analyses based on sequence comparison of the conserved region of Leishmania minicircle kDNA. A Nested PCR was designed to amplify the conserved block of Leishmania kDNA. The external primers LVp1-Fw (5´- GAC ATG CCT CTG GGT AGG GGC GTT C -3´) and LVp1-Rv (5´- GGG TGG TAC GAT TTT GAC CCT AA -3´) were used for the first PCR reaction. Internal primers LVp1-Fw and LVp5-Rv (5´- CTG GGA TGC GCG GCC CAC TAT -3´) were used in the second PCR reaction (S1 Fig). Each 25 μL of the first PCR reaction mixture contained 0. 8 mM of dNTP, 0. 04 U/μL high fidelity Platinum Taq polymerase (Invitrogen), 2 μL template DNA, 2. 2 mM MgCl2,1 X PCR buffer, and 0. 4 nM of LVp1-Fw and LVp1-Rv primers. The cycling reaction was as follows: 95°C for 5 min, followed by 35 cycles, each of 1 min at 95°C, 62°C for 30 sec and 72°C for 30 sec, and a final extension of 1 min at 72°C. The products of the first PCR were diluted 1: 10 with ultrapure water and 2 μL of this dilution was used as template for the second PCR, which was performed under the conditions described above using LVp1-Fw and LVp5-Rv primer and set and an annealing temperature of 59°C. Specificity of the tool was evaluated using total DNA extracted from human PBMCs obtained from healthy donors. Negative controls included within each reaction were PCR mix and DNA from human PBMCs. PCR products were separated in 1. 3% agarose gels and products of approximately 180 bp were extracted and purified for the sequencing reaction using the QIAquick gel extraction kit (Qiagen). An 82bp fragment spanning conserved blocks 1 to 2 from the kDNA was selected for the analysis based on sequence variability in the inter-block regions. High resolution sequences were obtained by bi-directional Sanger sequencing (Macrogen-Korea) using LVp1-Fw and LVp5-Rv primers, and sequences edited and analyzed using BioEdit v7. 2. 5. Genetic distances were calculated using MEGA 6. 0. Selection of the nucleotide substitution model was based on MEGA 6. 0 outputs for the model that best fitted the sequence data. Models analyzed included Jukes-Cantor, Hasegawa-Kishino-Yano and Tamura-Nei among others. Results for model use, based on the full sequence alignment indicated that the Jukes-Cantor model and G distribution best fitted the data. MEGA6 was also used to construct trees from resulting distance matrices. DNA was extracted from log-phase promastigotes using DNeasy Blood & Tissue Kit (Qiagen, USA). Fourteen microsatellite loci distributed in 13 Leishmania chromosomes were amplified by PCR, as previously described [19]. The size of the microsatellites was determined by mobility of the PCR products in 4. 5% agarose gels. Genetic distances were estimated using MSA4. 05 software and Populations-1. 2. 32, and neighbor joining and UPGMA trees were constructed using MEGA6. D' Agostino and Pearson omnibus test [20] was used to test for departures of quantitative data from normal distributions. For parasite burdens, this was done on log-transformed data. The Mann Whitney U-Test was employed for statistical comparisons. Binomial proportion confidence intervals were calculated with the Wilson method using the' binom' package in R [21]. Statistical significance was defined as p<0. 05. Data were analyzed using Prism 5 software (GraphPad Software, Inc. , La Jolla, CA). Amplification of human GAPDH gene in 95% (114 of 120) of the DNA samples obtained from the 30 participants in the exploratory phase (Phase 1) corroborated the quality of the extracted DNA. The six GAPDH negative samples corresponded to swab samples from conjunctiva, reflecting difficulty in obtaining samples of a suitable quality from this potentially informative tissue. GAPDH negative samples were excluded from further analysis. Four of the 5 individuals (80%) with prior history of CL (index cases), and 6 of 15 individuals (40%) with asymptomatic infection had one or more kDNA positive samples (Table 1). kDNA positive samples from the index case and from one or more asymptomatic cohabitants were found in three of the five households. A representative blot of samples from members of one household is shown in Fig 2. kDNA positive samples were not found in any of the healthy individuals from either endemic (n = 5) or non-endemic (n = 5) areas. Based on the results of the exploratory analyses we proceeded to evaluate samples obtained from 154 individuals residing in CL endemic communities in Nariño and Risaralda for the presence of Leishmania kDNA. At least one sample was obtained from each study participant, however, 27 individuals did not consent to provide blood samples, 9 refused nasal swabs and 25 refused tonsil swabs. Amplification of the human GAPDH gene was achieved in at least one sample from each of the 154 participants, representing a total of 78% of samples (504 of 647 total samples). Sixty-two percent of GAPDH negative samples (88 of 143) corresponded to lesion scar aspirates, 33% were from conjunctival swabs and the remaining 5% were from blood monocytes, nasal and tonsil swabs. The one time sampling of study participants coupled to PCR of Leishmania kDNA and Southern blot hybridization, revealed parasite persistence in 40% (46/114) of LST+ individuals without active disease, composed of 41 of 97 LST+ participants with a clinical history of CL and 5 of 17 asymptomatic infected individuals (Table 2). Nineteen (12%) individuals had more than one kDNA positive sample, for a total of 84 kDNA positive samples within the study population. Molecular demonstration of the presence of Leishmania was more frequent among individuals who had clinical evidence of prior symptomatic infection (44%, 51 of 116) than among LST positive asymptomatic infected individuals (29%, 5 of 17). kDNA positive samples (from blood monocytes, tonsil and nasal mucosa) were also detected in 3 of 18 skin test negative healthy individuals from endemic areas, and in the three positive control cases with active disease. Blood monocytes were the most frequently positive sample for kDNA, being positive in 26% of individuals without active disease, followed by nasal and tonsil swabs that were positive in 14% of sampled individuals. We quantified the parasite burden and assessed the viability of Leishmania among kDNA positive samples by detection of RNA transcripts of the Leishmania 7SLRNA gene. Transcripts were amplified from 59% of LST+ individuals in whom Leishmania kDNA had been detected; from 25 of the 41 participants with clinical history of CL, and from 2 of the 5 asymptomatic infected individuals. Transcripts were also amplified from 6 of 10 individuals with a history of CL but with negative or unavailable LST results, and from 2 of the 3 LST- healthy individuals in the study communities from whom Leishmania kDNA was amplified (Table 2). Amplification of the human TATA box binding protein (TBP) gene transcript was successful in 90% of the 84 kDNA positive samples, supporting the quality and reliability of the amplified RNA. Twenty-eight kDNA positive samples were below the limit of detection of the 7SLRNA qRT-PCR (10 from blood monocytes and 18 from mucosal swab samples). Parasite loads ranged between 0. 2 to 22 parasites per reaction. Absolute parasite loads and those normalized to the number of human nucleated cells per sample were slightly higher in blood monocytes than those from mucosal swab samples (Fig 3A–3C). No apparent differences in parasite loads of different mucosal tissues were observed. The presence of parasite RNA transcripts substantiates the viability of Leishmania within the sampled tissues and among a high proportion of individuals with subclinical infection. To explore relationships among Leishmania in individuals with subclinical infection and parasite strains isolated from patients with active disease within the same foci of transmission, we developed a nested PCR approach targeting a 180bp segment of the conserved region of Leishmania minicircle kDNA spanning blocks 1 to 3. The lower limit of detection of the nested PCR was 10−3 promastigotes per reaction (Fig 4A). No cross-reactivity with human DNA was detected (Fig 4B). Amplification products were obtained from Leishmania species pertaining to both L. (Viannia) and L. (Leishmania) subgenera and a faint amplification product of ~180bp was detected in T. cruzi DNA (Fig 4C). Although multiple sequencing attempts were performed on T. cruzi amplification products, quality chromatograms were never obtained, reflecting the low-sensitivity cross reaction due to low homology of the LVp5 primer annealing site (S1 Fig). Alignment of L. donovani, L. infantum, L. mexicana, L. tarentolae, L. (V.) braziliensis, L. (V.) guyanensis, L. (V.) panamensis and T. cruzi kDNA sequences (S2 Table) showed that regions between conserved blocks 1 and 2 were polymorphic among species and strains of the same species (Fig 5). To explore the resolution and assess the reliability of the strain grouping achieved by kDNA genotyping, we performed a comparative analysis against multilocus microsatellite typing (MLMT). Genetic distances were calculated from MLMT data (S3 Table) and from kDNA sequences of a panel of 34 L. (V.) panamensis and 9 L. (V.) guyanensis strains isolated from diverse locations within the Colombian territory (S1 Table). Results showed that both methods concurred in clustering of strains obtained from individuals with CL in foci of active transmission during outbreaks of CL, thus delimiting strains by time and location (Fig 6). Nevertheless, subgroups defined by MLMT such as zymodemes 2. 2 and 2. 3 of L. (V.) panamensis were not clustered by kDNA genotyping, in line with the higher variability and rate of evolution of kDNA compared to microsatellite sequences [13,22]. To estimate the best method for data analysis and clustering, we run in parallel Neighbor-Joining, UPGMA and Maximum Likelihood methods of kDNA sequences. As shown in S2 Fig, all clustering algorithms provided the same group distribution with the exception of clustering of database sequences obtained from the L. mexicana complex. Although it is recognized that bootstrap values are indicators of robustness for definition of the most appropriate clustering method, bootstrap values were low when using any of the above methods, potentially due the small sequence size (82bp) and the low overall variability among sequences (where the most divergent strains—L. (V) panamensis 2363 and L. infantum LLM735—showed 83% sequence homology). That all of the methods generated similar clustering and that clusters are biologically concordant, supports their usefulness in these analyses. UPGMA clustering was selected for data interpretation given that clustering of the subgenera represented more accurately the Leishmania taxonomy. Considering the reported variability of kDNA sequences among clinical strains [16,23] and possible variations introduced during the PCR or sequencing reactions, we assessed the reproducibility of the method by re-amplifying and re-sequencing DNA samples from 13 strains belonging to different species of the Viannia subgenus obtained from the CIDEIM BioBank. As shown in S3 Fig, 100% sequence identity was obtained in two independent technical replicas in sequences from 10 of the 13 strains, and 99% sequence identity (corresponding to a 1bp change) in sequences from the remaining 3 strains. These results support the reproducibility of the method for analysis of Leishmania genetic diversity using clinical samples and strains. The intra-species variability of the target sequence and the sensitivity and specificity of the method for Leishmania kDNA, support the potential of kDNA genotyping for the analysis of clinical samples with low parasite burden. We attempted to sequence nested PCR amplification products from 84 kDNA positive samples from 59 individuals. Fifty-four samples from 45 subclinically infected participants were successfully amplified by the nested PCR. However, PCR products from 33 of these samples, corresponding to 28 individuals with immunological or molecular evidence of subclinical infection (26 with prior history of CL, 2 with asymptomatic infection and one from a “healthy” LST- individual in the endemic site) were of sufficient quality and quantity to provide accurate sequences. kDNA sequences were also obtained from L. (V.) panamensis strains isolated from the three patients with active CL. Sequences were analyzed alongside a panel of representative L. (V.) panamensis, L. (V.) guyanensis and L. (V.) braziliensis strains from diverse geographical origins within the Colombian territory S1 Table), and database sequences from L. donovani and L. infantum as outgroups (S2 Table). kDNA sequences from individuals with subclinical infection grouped within the L. (Viannia) cluster (Fig 7). However, the relationship between L. (V.) panamensis strains currently circulating in the focus of transmission and parasites involved in subclinical infections that have persisted over many years could not be discerned. Asymptomatic infections are common to all transmissible agents, constituting a variable segment of the spectrum of outcomes. The frequency of these inapparent infections varies according to the pathogenicity of the agent and the susceptibility of the exposed population. Investigations of the incidence of infection based on LST conversion revealed that 90% of incident infections in a focus of L. (V.) panamensis transmission in the municipality of Tumaco, Colombia, were asymptomatic [1], whereas only 17% of infections were asymptomatic in a focus of L. (V.) peruviana transmission in Peru [3]. In the current study, LST reactivity provided an immunological marker for asymptomatic and subclinical infections, which were defined respectively as infections that result in skin test conversion but not disease, and persistent infection following clinical resolution of disease. Leishmanin reactivity induced by infection, whether symptomatic or asymptomatic is typically long-lived and presumably sustained by antigen exposure. The amplification of kDNA and 7SLRNA transcripts from 40% and 24% of LST+ individuals, respectively, supports this assumption and demonstrates the technical feasibility of molecular detection of Leishmania in mucosal and blood monocyte samples from individuals with subclinical infections. Although the relevance of asymptomatic and subclinical infections to public health is poorly understood, the detection of Leishmania in the absence of active disease in 40% of LST+ and a small proportion of LST- residents of endemic foci of L. (Viannia) transmission provides an indication of the substantial magnitude of the population harboring parasites. Amplification of Leishmania DNA and RNA from 2 of 18 “healthy” LST negative participants could reflect waning delayed-type hypersensitivity responses [24] or a lower sensitivity threshold of this immune-reactivity based method to detect previous parasite exposure. LST negativity in at least 12 of 116 participants with prior history of CL is also consistent with either of these scenarios. Considering that sampling was conducted only one time and that tissue distribution and burden of parasites are likely to vary over time as shown with sequential samples of asymptomatic L. infantum infection [25], and with the health status of each individual, the actual infected population is probably underestimated. Subclinical infections represent a risk for reactivation and the triggering of pathogenesis. The recurrent behavior of dermal leishmaniasis in Latin America was clinically recognized in mucosal disease, long considered a secondary manifestation of prior cutaneous disease [26] or even prior asymptomatic infection [27]. Direct evidence of re-activation as a cause of recurrent leishmaniasis was provided by biochemical and genotypic analyses of Leishmania strains isolated from primary lesions and from recurrent or new lesions following complete resolution of primary lesions [23]. Additionally, activation of CL following local trauma [28] and immunosuppression [29–31] in individuals with asymptomatic infection or healed lesions provides evidence for the participation of immunological and inflammatory triggers in the activation/re-activation of CL. Population-based studies in areas of endemic transmission have identified both LST reactivity and the presence of scars compatible with history of CL as important risk factors for development of new incident lesions [1,3, 10], underscoring the role of asymptomatic and subclinical infection in the epidemiology and natural history of CL. Our results documenting the long-term presence of viable Leishmania in blood monocytes and unaffected mucosal tissues in individuals with asymptomatic and subclinical infection support the indefinite persistence of Leishmania in the human host. Through mechanistic mathematical modeling, Miller et. al. have recently shown that a small proportion of asymptomatically infected individuals (3. 2%) with the highest parasitemias in VL endemic areas in Ethiopia, were responsible for infection of an average 62% of infected sand fly vectors [11]. Although parasite loads in unaffected mucosal tissues and blood monocytes (0. 2 to 22 parasites per reaction, equivalent to 7 to 770 parasites per sample) of individuals with subclinical and asymptomatic infection in our study population are relatively low compared to those with asymptomatic infection in VL endemic areas [11], these are within the range of parasite loads found in unaffected mucosal tissues in individuals with active CL [12]. Together with lessons learned from investigations of the infectivity of asymptomatic and even vaccinated dogs for Lu. longipalpis, the principal vector of VL in the Americas, and from malaria elimination initiatives showing that asymptomatic carriers, in addition to being at risk of developing disease via reactivation of infection [32] can transmit infection to mosquitos [33,34], support the possibility that subclinically infected individuals may act as reservoirs for anthroponotic transmission of L. (Viannia) species, which have traditionally been considered zoonoses. The importance of understanding the role of subclinical infection in the incidence and propagation of leishmaniasis has been recently recognized as a priority research area by the World Health Organization Expert Committee on the Control of the Leishmaniases [35]. A limiting factor in the study of subclinical infection is the technical feasibility for phenotypic and genotypic characterization of Leishmania from samples having low parasite burdens. To address this limitation we developed a strategy for analysis of genetic diversity based on nested PCR amplification and sequence genotyping of the conserved region of Leishmania minicircle kDNA. The sensitivity of this strategy allowed Leishmania kDNA sequences to be obtained from mucosal swab samples and blood monocytes from 28 of individuals with immunological (LST) or molecular (kDNA detection) evidence of subclinical infection, demonstrating the plausibility of approaching genotypic characterization of Leishmania causing subclinical infection. Although the sensitivity of nested PCR was comparable to that of kDNA amplification and detection by southern blot (10−3 promastigotes per reaction), partial degradation of DNA samples after prolonged storage, low concentration of target DNA for the sequencing reaction and DNA loss during the band purification process could have contributed to reduced sensitivity of the amplification of clinical samples. Regions within the conserved blocks of Leishmania kDNA minicircles were selected as targets for genetic diversity analysis based on limited yet potentially informative polymorphic characteristics. The methodology developed revealed diversity among Leishmania causing subclinical infection in endemic foci of transmission of L. (V.) panamensis, and clustering within the L. (Viannia) subgenus. Nevertheless, the limited number of active cases (n = 3) and the level of heterogeneity in the amplified sequences did not allow relationships to be discerned between strains isolated from individuals with active disease at the time of the study and Leishmania persisting in the absence of disease. This could be due to variation in parasite populations during decades of transmission within the sites, or limitations of the genotyping methodology. Comparative analysis of kDNA and MLMT genotyping in a panel of strains of L. (Viannia) species showed clustering of strains isolated from disease outbreaks to be achieved by both methodologies, suggesting that kDNA genotyping could be exploited for future studies of the cycle and dynamics of transmission in active foci and during disease outbreaks. However, because phenotypically distinguishable strains (eg. L. (V.) panamensis strains pertaining to zymodemes 2. 2 and 2. 3) clustered by MLMT but not kDNA typing, micro-heterogeneity of kDNA sequences could impede the discernment of relationships among closely related parasites [36,37]. This outcome illustrates some of the limitations of kDNA genotyping, which include analysis of a single target sequence and the sequence length. Multi-target sequence analysis of polymorphic and high copy number sequences such as miniexon, Cytochrome B, GP63 or Cysteine proteinase B genes [38], and improved sensitivity for minicircle kDNA amplification could optimize the robustness of this approach to accessing subclinical infection and the parasites involved. Our results provide parasitological confirmation of persistent infection in the absence of disease among residents of endemic areas of CL and a methodological approach to investigate the epidemiology and public health impact of subclinical infections. The novel exploitation of kDNA genotyping establishes proof-of-principle of the feasibility of genetic diversity analysis in parasite populations previously inaccessible and unexplored, and provides bases for more robust analyses of the relationships among these parasite populations.

A variable and often high proportion of individuals residing in areas where cutaneous leishmaniasis is endemic are exposed to Leishmania parasites, yet do not develop symptoms of disease. The role of this asymptomatic population in the transmission of disease is unknown and could interfere with the effectiveness of community or population-based control measures. Exposure to Leishmania is indirectly assessed by immunological tests; however, immunological evidence does not discriminate between historical exposure to the parasite and actual presence of parasites without causing clinical manifestations. We sought to determine whether viable Leishmania are present in individuals with immunological evidence of asymptomatic infection. Our results showed that at least 24% of individuals having immunological evidence of subclinical or asymptomatic infection harboured live Leishmania. These individuals may be at risk of activation of disease, or could represent an unperceived reservoir of parasites for vector-borne transmission. Characterization of Leishmania causing asymptomatic infection has not been possible due to technical limits of detection of parasites in low grade infections. We developed a molecular method that allows genotypic analysis of parasites involved in subclinical infection and potentially provides a means to assess their involvement in transmission.

Abstract Introduction Methods Results Discussion

2015

Parasitological Confirmation and Analysis of Leishmania Diversity in Asymptomatic and Subclinical Infection following Resolution of Cutaneous Leishmaniasis

Inherited prion disease (IPD) is caused by autosomal-dominant pathogenic mutations in the human prion protein (PrP) gene (PRNP). A proline to leucine substitution at PrP residue 102 (P102L) is classically associated with Gerstmann-Sträussler-Scheinker (GSS) disease but shows marked clinical and neuropathological variability within kindreds that may be caused by variable propagation of distinct prion strains generated from either PrP 102L or wild type PrP. To-date the transmission properties of prions propagated in P102L patients remain ill-defined. Multiple mouse models of GSS have focused on mutating the corresponding residue of murine PrP (P101L), however murine PrP 101L, a novel PrP primary structure, may not have the repertoire of pathogenic prion conformations necessary to accurately model the human disease. Here we describe the transmission properties of prions generated in human PrP 102L expressing transgenic mice that were generated after primary challenge with ex vivo human GSS P102L or classical CJD prions. We show that distinct strains of prions were generated in these mice dependent upon source of the inoculum (either GSS P102L or CJD brain) and have designated these GSS-102L and CJD-102L prions, respectively. GSS-102L prions have transmission properties distinct from all prion strains seen in sporadic and acquired human prion disease. Significantly, GSS-102L prions appear incapable of transmitting disease to conventional mice expressing wild type mouse PrP, which contrasts strikingly with the reported transmission properties of prions generated in GSS P102L-challenged mice expressing mouse PrP 101L. We conclude that future transgenic modeling of IPDs should focus exclusively on expression of mutant human PrP, as other approaches may generate novel experimental prion strains that are unrelated to human disease. Prion diseases are a closely related group of neurodegenerative conditions which affect both humans and animals [1,2]. They are both experimentally and, in some cases, naturally transmissible within and between mammalian species. Cross-species transmission is generally much less efficient than within-species transmissions, being limited by a ‘species’ or transmission barrier [2,3]. Prion diseases in humans include Creutzfeldt-Jakob disease (CJD), Gerstmann-Sträussler-Scheinker disease (GSS), fatal familial insomnia (FFI), kuru and variant CJD (vCJD) [1,4, 5]. According to the widely accepted ‘protein-only’ hypothesis [6], the central feature of prion disease is the conversion of host-encoded cellular prion protein (PrPC) to alternative isoforms designated PrPSc [1,2, 7]. It is proposed that PrPSc is the infectious agent acting to replicate itself with high fidelity by recruiting endogenous PrPC and that the difference between these isoforms lies purely in the monomer conformation and its state of aggregation [1,2, 8] although it is now clear that infectivity can also be associated with protease-sensitive disease-related PrP assemblies distinct from classical PrPSc [9–11] and that infectious and neurotoxic PrP species can be uncoupled [12,13]. Inherited prion disease (IPD) is caused by autosomal-dominant mutations in the human PrP gene (PRNP) and constitute about 15% of all human prion disease [4,14]. Over 40 mutations have been identified, but the precise biochemical mechanisms that lead to disease remain unknown. Within the framework of the protein-only hypothesis, pathogenic mutations in PrP are thought to predispose the mutant proteins to adopt disease-causing conformations and assembly states [2–4]. A proline to leucine substitution at codon 102 (P102L) of human PrP is the most common mutation associated with the GSS phenotype and was first reported in 1989 [15]. Many other kindreds have now been documented worldwide [16], including the original Austrian family reported by Gerstmann, Sträussler, and Scheinker in 1936 [17,18]. Progressive ataxia is the dominant clinical feature, with dementia and pyramidal features occurring later in a disease course typically much longer than that of classical CJD. However, marked variability at both the clinical and neuropathological levels is apparent, with some patients developing a classical CJD-like phenotype with early and rapidly progressive dementia [18–30]. A significant part of this phenotypic variability appears to be contributed by variable propagation of distinct disease-related PrP species generated from either PrP 102L [21,22] or wild type PrP [24,27]. Two distinct abnormal conformers of PrP 102L that generate proteinase K (PK) -resistant fragments of either ~21–30 kDa or ~8 kDa [21,22,24,27,31] have distinct prion transmission properties in 101LL PrP gene knock-in mice [25], while the potential transmissibility or neurotoxicity of abnormal conformers of wild type PrP (that generate PK-resistant fragments of 21–30 kDa [24,27]) remains unknown. Such heterogeneity in disease-related PrP isoforms present in IPD P102L patient brain severely complicates interpretation of transmissions in both conventional and transgenic mice. The conformational selection hypothesis [2,32] predicts that heterogeneous prions formed from PrP in distinct conformations would be differentially selected by hosts expressing different PrP primary sequences. In this regard expression of the homotypic human mutant protein in the host may be critical to accurately model the disease, as only the human mutant protein may be conformationally susceptible to the prion strain involved [2,3, 33]. Much of the transgenic modelling of inherited prion disease has however focused on superimposing human PrP mutations onto rodent PrP in order to establish whether infectious prions can be generated de novo. An extremely important consideration in such studies is whether superimposition of pathogenic human PrP mutation into mouse PrP will have the same structural consequences [2,3, 33,34]. The possibility of propagating novel prion strains that do not recapitulate the molecular and neuropathological phenotype of the original human disease appears probable [2,3, 33] and indeed has been documented with variant CJD transmissions [35]. Recently we established that IPD P102L patient brain isolates could transmit disease with 100% clinical attack rates and short incubation periods to transgenic mice expressing human PrP 102L on a mouse PrP null background (designated 102LL Tg27 mice) [33]. In these transmissions we observed the propagation of the abnormal conformer of PrP 102L that generates protease-resistant fragments of ~21–30 kDa [33]. We also demonstrated that such mice were susceptible to infection with classical CJD prions leading to the generation of prions with altered PrPSc glycoform ratios [33]. The availability of these prions from 102LL Tg27 mice, in which disease-related PrP is entirely composed of PrP 102L (as opposed to the heterogeneous PrP in primary human GSS brain inoculum), now permits direct testing of their host range and in particular the ability of these prions to propagate using wild type human PrP or mouse PrP as substrate. Our findings show that human PrP 102L can support the propagation of distinct prion strains and that human PrP 102L prions have transmission properties strikingly different from those generated in transmission models in which the 102L mutation was superimposed onto mouse PrP. Prions originating from the primary transmission of three different IPD P102L patient brains to 102LL Tg27 mice [33] (hereafter designated GSS-102L prions) transmitted clinical prion disease with 100% attack rates and short mean incubation periods (~165 days) when passaged in further 102LL Tg27 mice (Table 1). Brain samples of all mice in these transmissions were positive for PK-resistant PrP 102L by immunoblotting using ICSM 18 (Fig 1A) (Table 1). Both the PK-resistant PrP fragment size (~21–30 kDa) and predominance of the di-glycosylated PrP glycoform mirrored that seen in the 102LL Tg27 mouse brain inoculum (Fig 1A). Immunohistochemistry with ICSM 18 showed extensive abnormal PrP deposition throughout the brain (thalamus shown in Fig 1C) accompanied by prominent astrocytosis and spongiosis. Collectively, these findings establish that IPD P102L prions propagate with high efficiency when serially passaged in 102LL Tg27 mice. Much of the previous modeling of IPD P102L has involved superimposing the mutation onto the wild type mouse PrP sequence (reviewed in ref [3]). However, it is unclear if challenge of mouse PrP 101L with human P102L prions would lead to the generation of authentic human prion strains or conversely would lead to the generation of experimental prion strains with different transmission characteristics. Notably, after human P102L prions were passaged once in 101LL PrP gene knock-in mice the resultant prions were shown to readily infect wild type mice [36]. We therefore inoculated the GSS-102L prion isolates that transmitted efficiently on passage in 102LL Tg27 mice to wild type mice and also to transgenic mice expressing wild type human PrP on a mouse PrP null background. Strikingly, in complete contrast to findings with the mouse 101LL PrP gene knock-in model we found that GSS-102L prions failed to produce clinical prion disease or any evidence of sub-clinical prion infection when inoculated into wild type mice (Table 1). Even more remarkably the same GSS-102L prions produced no clinical prion disease or evidence of sub-clinical prion infection when inoculated into transgenic mice expressing wild type human PrP (Table 1). In all of these negative transmissions examination of brain showed no detectable PrPSc by high sensitivity immunoblotting (Fig 1A and 1B, lanes 4 and 5) or abnormal PrP deposition by immunohistochemistry (Fig 1E–1H). In addition, no evidence for elevated levels of spongiosis or gliosis in comparison to the brain of uninoculated age-matched control mice was observed. Collectively these data establish that GSS-102L prions which replicate with high efficiency in a host expressing human PrP 102L are unable to propagate using wild type human PrP or wild type mouse PrP as substrate. In comparison to IPD P102L prions, transmission of classical CJD prions to 102LL Tg27 mice appears to be limited by a transmission barrier [33]. In primary transmissions, although nearly all CJD prion-challenged 102LL Tg27 mice showed evidence for prion infection, only a proportion of mice developed clinical prion disease and then only after prolonged incubation periods [33]. In addition, a change in propagated PrPSc type was observed (which itself is indicative of a transmission barrier [2]) with PrPSc glycoform ratios switching from those present in the CJD inocula to ones that more closely resemble those seen in the brain of IPD P102L patients and IPD P102L prion-challenged Tg27 mice [33]. From these primary transmissions, we were unable to distinguish whether the 102L mutation in the host PrP had directly dictated the strain characteristics of the propagated prions (to essentially become congruent with GSS-102L prions) or whether CJD-like prion strain properties were retained. To investigate this, we passaged prions from CJD-challenged 102LL Tg27 mice (hereafter designated CJD-102L prions) in further 102LL Tg27 mice, in transgenic mice expressing wild type human PrP and in wild type mice (Table 2). In 102LL Tg27 mice we observed that the barrier to development of clinical prion disease seen at primary transmission of classical CJD prions was not abrogated at secondary passage (Table 2). Although nearly all CJD-102L prion-inoculated mice developed prion infection, as evidenced by detection of PrPSc (Fig 2) and abnormal PrP deposition throughout the brain (Fig 3), clinical prion disease was again only observed in a proportion of inoculated recipients and then only at prolonged incubation periods (Table 2). PrPSc typing showed that the altered PrPSc glycoform ratio of CJD-102L prions generated after primary transmission of classical CJD prions to 102LL Tg27 mice was not maintained after further passage in the same mice. The PrPSc type now appeared to more closely resemble the original classical CJD inoculum with a predominance of mono-glycosylated PrP and was readily distinguishable from the di-glycoslyated PrP dominant glycoform pattern seen after secondary passage of GSS 102L prions in 102LL Tg27 mice (Fig 2A, 2C and 2E) (Table 3). From these transmissions we concluded that GSS-102L prions and CJD-102L prions have incongruent transmission properties after further passage in 102LL Tg27 mice. Importantly, the disparate nature of CJD-102L prions and GSS-102L prions became obvious after examining the transmission properties of CJD-102L prions in transgenic mice expressing wild type human PrP. In complete contrast to GSS-102L prions, all CJD-102L prion isolates transmitted clinical prion disease to mice expressing wild type human PrP in a fashion analogous to the original CJD inoculum (Table 2). In these transmissions PrPSc was readily detected in brain by immunoblotting (Fig 2) and abnormal PrP deposition was observed throughout the brain by immunohistochemistry (Fig 3). Humanised transgenic mice expressing human PrP 129 valine on a Prnp null background are highly susceptible to sporadic CJD prions regardless of the PrPSc type or codon 129 genotype of the inoculum [37–43]. These transmissions are typically characterised by 100% attack rates of prion infection producing uniform clinical prion disease after similar short incubation periods of around 200 days. The absence of a transmission barrier to sporadic CJD prions is not however uniformly observed in transgenic mice expressing human PrP 129 methionine on a Prnp null background. Here mismatch at residue 129 between the inoculum and host can significantly affect transmission [41,44–46] as evidenced by more prolonged and variable incubation periods and reduced attack rates [41,43,44]. Remarkably, we observed that CJD-102L prions behaved in a closely similar fashion that corresponded with the codon 129 status of the original CJD inoculum (Table 2). This was striking because all of the CJD-102L prion isolates have PrP with residue 129 methionine. Consistent with the CJD-like transmission properties of CJD-102L prions in transgenic mice expressing wild type human PrP, PrPSc typing of the recipient mouse brain showed that the di-glycosylated dominant PrPSc glycoform ratio of CJD-102L prions in the inoculum had switched to a mono-glycosylated PrPSc dominant pattern which more closely resemble CJD prions (Table 3; Fig 2B, 2D and 2F, lanes 5). Collectively, these data show that CJD-102L prions are distinct from GSS-102L prions and retain the transmission properties of the original CJD prion strains. Notwithstanding these observations, all the CJD-102L prion isolates were obtained after a single passage of classical CJD prions in 102LL Tg27 mice and it remains to be seen whether serial passage on the mutated sequence would lead to similar conservation of CJD phenotype. Consistent with the finding that classical CJD prions transmit prion infection only occasionally to wild type mice with long and variable incubation periods [37,39,40,42,47] we found that CJD-102L prions were also unable to propagate efficiently in wild type mice (Table 2). We found that only one out of eighteen CJD-102L prion-inoculated wild type mice became infected (Table 2) with all other mice showing no evidence of subclinical prion infection by either PrP immunoblotting or immunohistochemistry. Co-propagation of distinct disease-related PrP conformers in IPD brain, combined with differences in their neuropathological targeting, abundance and potential neurotoxicity, provides a general molecular mechanism underlying phenotypic heterogeneity in patients with the same PRNP mutation. Previously we and others have reported the propagation of distinct isoforms of protease-resistant PrP with divergent properties in IPD P102L patient brain and such molecular heterogeneity severely hampers interpretation of primary transmissions to both conventional and transgenic mice (for review see [3]). In the present study we have investigated the properties of prions generated in transgenic mice expressing human PrP 102L following the intracerebral inoculation of IPD P102L or classical CJD brain isolates. The resultant prion isolates from these transgenic mouse brain were designated GSS-102L or CJD-102L prions, respectively, and because they are associated exclusively with disease-related conformers of human PrP 102L this enables unequivocal examination of the effects that this point mutation has on prion transmission barriers. Our findings show that GSS-102L and CJD-102L prions are distinct from one another with divergent prion strain transmission properties following further passage in transgenic mice expressing either human PrP 102L or wild type human PrP (Fig 4). Thus human PrP 102L is capable of supporting the propagation of distinct lethal prion strains and these data establish that the point mutation does not restrict PrP 102L to a single dominant pathogenic assembly state when templated by an exogenous prion strain. Importantly our model has enabled us to isolate and investigate the transmission properties of prions originating from IPD P102L patient brain following amplification exclusively on human 102L PrP. Our data show that 102L PrP prions from IPD P102L patient brain that generate PK-resistant PrP fragments of ~21–30 kDa have prion strain transmission properties distinct from all other prion strains propagated in acquired or sporadic human prion disease. The most outstanding feature of this prion strain is its inability to propagate in transgenic mice expressing wild type PrP. Significantly, the inability of GSS-102L prions to also propagate in wild type mice clearly shows that this prion strain is distinct from prions generated in IPD P102L prion-challenged 101LL PrP gene knock-in mice [36]. The remarkable ease of transmission of 101L-passaged IPD P102L prions to wild type mice [36] contrasts strikingly with our data and suggests that a novel prion strain was propagated by the mutant mouse PrP rather than faithful replication of the authentic human PrP 102L prion strain. We therefore recommend that future transgenic modeling of inherited prion disease should focus exclusively on using models that express the homotypic mutant human PrP primary sequence. We and others have reported that variable involvement of disease-related conformers of wild type human PrP may contribute to phenotypic heterogeneity in IPD P102L [24,27]. However the mechanism by which abnormal wild type PrP is deposited in P102L patient brain remains ill-defined. Wild type PrP may be recruited by a seeded reaction with 102L PrPSc or may accumulate independently as a consequence of pathological changes associated with disease progression. Notably, the glycoform ratios of proteinase K-resistant fragments of 102L PrP and wild type PrP from P102L patient brain are distinct from each other [24] [27] suggesting that the 102L point mutation powerfully dictates thermodynamic preferences for disease-related PrP assembly states that cannot be adopted by wild type PrP and that a significant transmission barrier may be associated with conversion of wild type PrP by a 102L PrPSc seed. This idea is supported by the observation that PK-resistant wild-type PrP in P102L patient brain does not appear to exceed approximately 10% of total PK-resistant PrP [24,27]. Here we show that GSS-102L prions that propagate efficiently in further 102LL Tg27 transgenic mice fail to produce prion infection in transgenic mice expressing wild type human PrP. Based upon the strength of this transmission barrier we conclude that seeded conversion of wild type PrP by abnormal conformers of 102L PrP that generate proteolytic fragments of ~ 21–30 kDa may, at best, be highly inefficient. From these data it is tempting to speculate that abnormal conformers of 102L PrP that generate protease-resistant fragments of 8 kDa might instead be responsible for variable recruitment of wild type PrP in IPD P102L patient brain. However other explanations may be equally possible. In particular, our transmission experiments do not mirror the situation in IPD P102L patient brain where both PrP 102L and wild type PrP are co-expressed. Thus in IPD P102L patient brain, wild type PrP will be exposed throughout the disease time course to all propagating 102L PrPSc species (rather than only at inoculation) and such prolonged exposure in vivo may be required for the generation of misfolded isoforms of wild type PrP. Alternatively because prion strains appear to comprise a quasispecies maintained under host selection pressure (rather than constituting a single molecular clone) [2,48–50] minor subtypes of 102L PrPSc may be populated differently in individual P102L patients leading to variable degrees of recruitment of wild type PrP. Notwithstanding such possibilities, at present we cannot conclusively resolve whether wild type PrP in IPD P102L patient brain misfolds through a directly seeded conversion reaction with an abnormal 102L PrP template or as a consequence of other pathological changes in the brain. In this regard, transmission experiments in heterozygous transgenic mice expressing both 102L PrP and wild type PrP would not be able to differentiate between these possibilities. Although the mechanism that leads to the accumulation of abnormal wild type PrP continues to remain ill-defined, this remains a potentially important contributor to phenotypic variation, not only in IPD P102L, but also in IPD associated with other PRNP mutations [51–56]. Storage and biochemical analyses of post-mortem human brain samples and transmission studies to mice were performed with written informed consent from patients with capacity to give consent. Where patients were unable to give informed consent, assent was obtained from their relatives in accordance with UK legislation and Codes of Practice. Samples were stored and used in accordance with the Human Tissue Authority Codes of Practice and in line with the requirements of the Human Tissue Authority licence held by UCL Institute of Neurology. This study was performed with approval from the National Hospital for Neurology and Neurosurgery and the UCL Institute of Neurology Joint Research Ethics Committee (now National Research Ethics Service Committee, London—Queen Square) —REC references: 03/N036,03/N038 and 03/N133. Work with mice was performed under approval and licence granted by the UK Home Office (Animals (Scientific Procedures) Act 1986; Project Licence number 70/6454) and conformed to University College London institutional and ARRIVE guidelines (www. nc3rs. org. uk/ARRIVE/). Transgenic mice homozygous for a human PrP102L, 129M transgene array and murine PrP null alleles (Prnpo/o) designated Tg (HuPrP102L 129M+/+ Prnpo/o) -27 mice (102LL Tg27) [33] have been described previously and were used without modification. Transgenic mice homozygous for a wild type human PrP129M transgene array and murine PrP null alleles (Prnpo/o) designated Tg (HuPrP129M+/+ Prnpo/o) -35 congenic (129MM Tg35c) were derived by subjecting previously described 129MM Tg35 mice [35,44,57] to commercially available speed congenic backcrossing on FVB/N genetic background (Charles River UK). Similarly, transgenic mice homozygous for a wild type human PrP129V transgene array and murine PrP null alleles (Prnpo/o) designated Tg (HuPrP129V+/+ Prnpo/o) -152 congenic (129VV Tg152c) were derived by subjecting previously described 129VV Tg152 mice [37,39,42] to the speed congenic scheme (Charles River UK). Inbred FVB/NHsd mice were supplied by Harlan UK Ltd. Strict bio-safety protocols were followed. Inocula were prepared, using disposable equipment for each inoculum, in a microbiological containment level 3 laboratory and inoculations performed within a class 1 microbiological safety cabinet. Ten mice per group from three transgenic lines, 102LL Tg27,129MM Tg35c, 129VV Tg152c and FVB/N wild type mice were inoculated with a panel of prion isolates, all previously passaged in 102LL Tg27 transgenic mice and therefore adapted to human 102L PrP. The primary inocula comprised human brain homogenates from three IPD P102L patients, one sporadic CJD patient and three iatrogenic CJD patients. Diagnosis of all cases had been neuropathologically confirmed. The genotype of each mouse was confirmed by PCR of DNA prior to inclusion and all mice were uniquely identified by sub-cutaneous transponders. Disposable cages were used and all cage lids and water bottles were also uniquely identified by transponder and remained with each cage of mice throughout the incubation period. Care of the mice was according to institutional and ARRIVE guidelines. Mice were anaesthetised with a mixture of halothane and O2, and intra-cerebrally inoculated into the right parietal lobe with 30 μl of 1% (w/v) brain homogenate prepared in Dulbecco’s phosphate buffered saline lacking Ca2+ or Mg2+ ions (D-PBS). All mice were thereafter examined daily for clinical signs of prion disease. Mice were killed if they exhibited any signs of distress or once a diagnosis of prion disease was established. At post-mortem brains from inoculated mice were removed, divided sagittally with half frozen and half fixed in 10% buffered formol saline. Anti-PrP monoclonal antibodies ICSM 18 and ICSM 35 were supplied by D-Gen Ltd, London, UK. ICSM antibodies were raised in Prnpo/o mice against α or β PrP as described elsewhere [58]. ICSM 18 is an IgG1 with an epitope spanning residues 142–153 of human PrP [58]. ICSM 35 is an IgG2b with an epitope spanning residues 93–105 of human PrP [58,59]. ICSM 18 recognizes both human PrP 102L and wild type human PrP whereas ICSM 35 recognizes wild type human PrP but not human PrP 102L [24]. Brain homogenates (10% (w/v) ) were prepared in D-PBS and aliquots analysed in duplicate with or without proteinase K digestion (50 μg/ml final protease concentration, 1h, 37°C) by electrophoresis and immunoblotting as described previously [60,61]. Duplicate blots were blocked in PBS containing 0. 05% v/v Tween-20 (PBST) and 5% w/v non-fat milk powder and probed with ICSM 18 or ICSM 35 anti-PrP monoclonal antibodies (0. 2 μ g/ml final concentration in PBST) in conjunction with anti-mouse IgG-alkaline phosphatase conjugated secondary antibody and chemiluminescent substrate CDP-Star (Tropix Inc, Bedford, MA, USA) and visualized on Biomax MR film (Kodak) as described [60,61]. For analysis of PrP glycoforms, blots were developed in chemifluorescent substrate (AttoPhos; Promega) and visualized on a Storm 840 phosphoimager (Amersham) using ImageQuaNT software (Amersham) [31,61]. Fixed brain was immersed in 98% formic acid for 1 h and paraffin wax embedded. Serial sections of 4 μm nominal thickness were pre-treated with Tris-Citrate EDTA buffer for antigen retrieval [61]. PrP deposition was visualized using ICSM 35 or ICSM 18 as the primary antibody, using an automated immunostaining system (www. ventana. com). Visualization was accomplished with diaminobenzidine staining. Bright field photographs were taken on an ImageView digital camera (www. soft-imaging. de) and composed with Adobe Photoshop.

Inherited prion disease (IPD) is caused by pathogenic mutations in the human prion protein (PrP) gene leading to the formation of lethal prions in the brain. To-date the properties of prions causing IPD and their similarities to prions causing other forms of human prion disease remain ill-defined. In the present study we have investigated the properties of prions seen in patients with Gerstmann-Sträussler-Scheinker (GSS) disease associated with the substitution of leucine for proline at amino acid position 102 (GSS P102L). We examined the ability of these prions to infect transgenic mice expressing human mutant 102L PrP, human wild-type PrP or wild-type mice. We found that GSS-102L prions have properties distinct from other types of human prions by showing that they can only infect transgenic mice expressing human PrP carrying the same mutation. Mice expressing wild-type human PrP or wild-type mouse PrP were entirely resistant to infection with GSS-102L prions. We conclude that accurate modeling of inherited prion disease requires the expression of authentic mutant human PrP in transgenic models, as other approaches may generate results that do not mirror the human disease.

Abstract Introduction Results Discussion Methods

2015

Transmission Properties of Human PrP 102L Prions Challenge the Relevance of Mouse Models of GSS

The sleeping sickness parasite Trypanosoma brucei has a complex life cycle, alternating between a mammalian host and the tsetse fly vector. A tightly controlled developmental programme ensures parasite transmission between hosts as well as survival within them and involves strict regulation of mitochondrial activities. In the glucose-rich bloodstream, the replicative ‘slender’ stage is thought to produce ATP exclusively via glycolysis and uses the mitochondrial F1FO-ATP synthase as an ATP hydrolysis-driven proton pump to generate the mitochondrial membrane potential (ΔΨm). The ‘procyclic’ stage in the glucose-poor tsetse midgut depends on mitochondrial catabolism of amino acids for energy production, which involves oxidative phosphorylation with ATP production via the F1FO-ATP synthase. Both modes of the F1FO enzyme critically depend on FO subunit a, which is encoded in the parasite’s mitochondrial DNA (kinetoplast or kDNA). Comparatively little is known about mitochondrial function and the role of kDNA in non-replicative ‘stumpy’ bloodstream forms, a developmental stage essential for disease transmission. Here we show that the L262P mutation in the nuclear-encoded F1 subunit γ that permits survival of ‘slender’ bloodstream forms lacking kDNA (‘akinetoplastic’ forms), via FO-independent generation of ΔΨm, also permits their differentiation into stumpy forms. However, these akinetoplastic stumpy cells lack a ΔΨm and have a reduced lifespan in vitro and in mice, which significantly alters the within-host dynamics of the parasite. We further show that generation of ΔΨm in stumpy parasites and their ability to use α-ketoglutarate to sustain viability depend on F1-ATPase activity. Surprisingly, however, loss of ΔΨm does not reduce stumpy life span. We conclude that the L262P γ subunit mutation does not enable FO-independent generation of ΔΨm in stumpy cells, most likely as a consequence of mitochondrial ATP production in these cells. In addition, kDNA-encoded genes other than FO subunit a are important for stumpy form viability. The parasitic protist Trypanosoma brucei undergoes a complex life cycle involving stages within the mammalian bloodstream and its tsetse fly vector. In the bloodstream of the mammalian host, the cell population exhibits two major morphotypes: the proliferative long slender bloodstream form (BSF) and the cell cycle-arrested stumpy form. Differentiation from the slender BSF to the stumpy form is triggered upon high slender parasite numbers [1]. The emergence of cell cycle-arrested stumpy forms prevents parasitaemia increasing further, prolonging host survival, and results in the characteristic waves of parasitaemia seen in bloodstream infections in rodents. This density dependent differentiation has been shown to be induced by a stumpy induction factor (SIF) via a form of quorum sensing [2]. The stumpy form is insect-transmissible and is preadapted to survive within the low glucose environment of the tsetse fly midgut, where it differentiates to the procyclic (PCF) tsetse midgut form of the parasite. PCF are able to generate ATP using mitochondrial energy production pathways, involving both oxidative and substrate-level phosphorylation [3–6]. In contrast, ATP production in the slender BSF is thought to solely involve non-mitochondrial glycolysis, utilising the glucose-rich environment found within the mammalian bloodstream [7,8]. Comparatively little is known about the metabolic requirements of stumpy forms, but studies have demonstrated an increase in the abundance of many mitochondrial proteins in the stumpy life cycle form compared to the slender BSF, including subunits of the mitochondrial respiratory complexes and key mitochondrial metabolic enzyme activities such as pyruvate dehydrogenase, α-ketoglutarate (α-KG) dehydrogenase, acetate: succinate CoA-transferase (ASCT) and succinyl-CoA synthetase (SCoAS) [9–12]. Accordingly, stumpy forms can utilise both glucose and α-KG as carbon sources for mitochondrial substrate level phosphorylation, at least in vitro [9,10,13]. Cytochromes have not been detected in stumpy forms [10], but the presence of an abbreviated oxidative phosphorylation pathway consisting of respiratory complexes I (cI; NADH: ubiquinone oxidoreductase) and V (F1FO-ATP synthase) and the trypanosome alternative oxidase (AOX) has been proposed [14]. The single mitochondrion of T. brucei contains a complex genome, termed kinetoplast DNA or kDNA and comprising of maxicircles and minicircles [15]. The maxicircle corresponds to the mitochondrial DNA of other organisms and encodes subunits of the respiratory chain and mitoribosome. Messenger RNAs for 12 of the 18 protein-coding genes require RNA editing by uridylyl insertion and deletion for maturation, a post-transcriptional process directed by guide RNAs, which are encoded in the minicircles. The parasite’s kDNA is essential in the long slender BSF and PCF stages of the life cycle [16–19]. In the latter, this is presumably due to a requirement for kDNA-encoded subunits of respiratory complexes III (cytochrome bc1 complex; subunit b encoded in kDNA), IV (cytochrome oxidase; subunits I, II and III encoded in kDNA) and the F1FO-ATP synthase (subunit a encoded in kDNA). These complexes are key constituents of the oxidative phosphorylation pathway that is required to generate ATP in that stage of the life cycle [4,6]. In the slender BSF, the F1FO-ATP synthase operates as an ATP hydrolysis-driven proton pump to maintain the essential electrical potential across the inner mitochondrial membrane (ΔΨm) [20–24]. As a consequence, the F1FO-ATP synthase, along with its kDNA-encoded subunit a, is essential in long slender BSF T. brucei. It is not known which, if any, mitochondrial genes are essential in the stumpy form. T. b. evansi and T. b. equiperdum are naturally occurring ‘dyskinetoplastic’ subspecies of T. brucei that survive in the complete (akinetoplastidy; kDNA0) or partial (kDNA-) absence of kDNA [25,26] (it should be noted that the taxonomic status of these organisms is controversial, see [27,28]). Interestingly, T. b. evansi and T. b. equiperdum are generally considered to be “monomorphic”: they remain in the slender form within the host, and, respectively, are transmitted mechanically via the mouthparts of hematophagous flies or sexually in horses. Stumpy forms have only been reported occasionally in field samples of kDNA- and kDNA0 strains [29–31]. However no surviving laboratory-grown strains of kDNA0 trypanosomes show any ability for pleomorphism [18,32]. It is not known whether the lack of kDNA influences the monomorphism displayed by these kDNA0 parasites, or whether the loss of pleomorphism and kDNA are independently selected. This question is potentially relevant for the spread of drug resistance, as kDNA independence is associated with reduced susceptibility to anti-trypanosomatid compounds belonging to the phenanthridines and diamidines [33]. We decided to investigate kDNA function and mitochondrial metabolism in stumpy forms by utilising a mutation in the nuclear-encoded γ subunit of the F1FO-ATPase, L262Pγ, that allows kDNA-independence in slender BSF T. b. brucei [16]. We generated pleomorphic slender BSF T. b. brucei cell lines that express L262Pγ, allowing the deletion of kDNA by acriflavine to produce clonal kDNA0 cells, and studied the within-host dynamics of these cell lines in mice and their physiology in vitro. Mouse infections showed that kDNA0 T. b. brucei are in fact capable of differentiating to the stumpy form in vivo. However, we found that kDNA0 stumpy cells have a shortened lifespan in vivo and in vitro. These kDNA0 stumpy cells are unable to sustain viability with α-KG as carbon source and do not have a Δψm. Treatment of kDNA+ stumpy form with the F1FO-ATPase inhibitor azide abolished Δψm. Our findings suggest that, like in slender BSF, the F1FO-ATP synthase generates the Δψm in early stumpy forms by acting as a proton-pumping ATPase. However, in contrast to slender BSF, the L262Pγ mutation is not sufficient to maintain full viability of kDNA0 stumpy forms. Our study provides new insight into stumpy form energy metabolism and kDNA function. In order to be able to investigate the requirement for functional kDNA in the differentiation of T. brucei from slender to stumpy form, we replaced one allele of the nuclear-encoded F1FO-ATPase subunit γ with a version with the L262P mutation (L262Pγ) in the pleomorphic cell line EATRO 1125 (AnTat1. 1 90: 13) [34] (generating cell line WT/L262Pγ; see Table 1 for a list of cell lines used in this study). This mutation fully compensates for the requirement for kDNA in slender bloodstream form T. brucei [16]. We introduced a wild type version (WTγ) into the same parental cell line to generate an otherwise isogenic control (WT/WTγ). We generated two cell lines lacking kDNA (kDNA0 #1 and #2) from two distinct clones of genotype WT/L262Pγ (#1 and #2) by treatment with acriflavine [16]; we obtained a third WT/L262Pγ (kDNA0) cell line (#3) fortuitously after spontaneous loss of kDNA. We confirmed absence of kDNA in all three cell lines by PCR and microscopically (S1A and S1B Fig). In vitro, WT/L262Pγ kDNA+ and kDNA0 cell lines grew at the same rate as the WT/WTγ cell line, showing that both modifications had no effect on the viability of the cells under these conditions (S1C Fig). As expected, cells expressing an L262Pγ allele, regardless of the presence of kDNA, were resistant to 10 nM EtBr [33], unlike cells expressing solely WTγ, which died within 4–5 days of treatment (S1C Fig). To test the capacity for differentiation to stumpy forms, we infected mice of strain MF1 with cell lines WT/WTγ, WT/L262Pγ #2 and WT/L262Pγ (kDNA0 #1 and #2) via IP injection. Accurate measures of parasitaemia level and morphology during the first peak of infection were recorded over time for each cell line in four replicate infections. Parasites that were morphologically stumpy were seen in all infections as they progressed (Fig 1A, days 7–8) and were found to express the stumpy-specific protein PAD1 [35] (Fig 1B). This demonstrated that kDNA0 populations were capable of generating stumpy forms. We next carried out a detailed comparison between cell lines in terms of the efficiency of the slender to stumpy transition, and the length of time that the parasitaemia was maintained, to judge the effect of kDNA loss on the in vivo dynamics of a mouse infection. Cell lines WT/WTγ and WT/L262Pγ had first peaks of parasitaemia that were very similar to each other (Fig 2A–2E) and to published data [36]. In contrast, we observed three main differences for the kDNA0 cell lines: (i) a delayed rise in parasitaemia (Fig 2A–2C); (ii) a more rapid decline in parasitaemia once peak density had been reached (Fig 2A); and (iii) absence of a smaller, second peak in slender form parasitaemia evident in kDNA+ cells on days 7–8 post infection (Fig 2D). Both kDNA0 clones showed a delayed rise in cell numbers (Fig 2A, days 3-6), at least in part caused by a slower growth rate up to day 4 (Fig 2B and 2C), suggesting that a lack of kDNA affects the parasite’s ability to proliferate in vivo and/or to persist during the transitions they undergo through a mouse infection. To investigate this observation further, we compared the rates of differentiation to intermediate and stumpy cells (Fig 2F–2H). Here, no consistent differences were observed between kDNA0 and kDNA+ cells. At peak parasitaemia, populations of all cell lines consisted of 80–90% stumpy cells, again demonstrating that kDNA is not required for the differentiation of T. brucei from slender to stumpy forms. Although differentiation of kDNA0 cell line #1 was delayed by approximately half a day, this was not the case for the other kDNA0 cell line and therefore unlikely to be a consequence of kDNA loss. The slower growth (and therefore the delayed rise) of these cells could therefore be due to longer cell-cycle times or an increased death rate. Having reached maximum parasitaemia, total cell numbers for kDNA0 parasites declined more rapidly than for kDNA+ parasites (Fig 2A, days 6–9). When we assessed stumpy form densities, it was evident that kDNA0 stumpy cells maintained high densities for a shorter period of time than kDNA+ stumpy cells (Fig 2E and 2I). To investigate this further, we assessed the lifespan of stumpy forms in vitro. In this experiment we included a third kDNA0 cell line (#3) that had lost its kDNA spontaneously to address the possibility that stumpy lifespan might have been affected by any non-kinetoplast related mutagenic effects of acriflavine. We harvested populations enriched for the stumpy form parasites from mice and incubated them in HMI-9 medium in the presence of the cytostatic agent α-difluoromethylornithine (DFMO); this prevents contaminating slender cells from proliferating [37–39]. We sampled cells every 8 h and determined numbers with a particle counter. We also stained cells with carboxyfluorescein diacetate succinimidyl ester (CFDA-SE) to analyse the proportion of dead cells over time. Consistent with what we observed in vivo (Fig 2I), cell numbers for kDNA0 stumpy cells dropped more quickly for kDNA0 than for kDNA+ stumpy cells (Fig 3A), with kDNA0 stumpy cells reaching a threshold of 50% dead cells 40–50 h earlier than the kDNA+ populations (Fig 3B). All kDNA0 cell lines behaved in a very similar manner, confirming that the decrease in stumpy life span was not due to secondary mutations caused by acriflavine treatment. Finally, kDNA+ cells had a second peak in slender parasitaemia around days 7–8 (Fig 2D). This second peak was absent in kDNA0 cells. The presence of the second peak in slender density in kDNA+ parasites was confirmed by quantifying the percentage of cells in G2 phase of the cell cycle in samples taken across the time course, using flow cytometry (S2 Fig). In summary, our mouse infection data demonstrated differences in the within-host infection dynamics for T. brucei parasites with and without a kinetoplast. Most importantly, stumpy cells lacking kDNA had a reduced life span. Mathematical models allow complex biological systems to be deconstructed into individual components and parameters, and as such are suitable for quantitation, hypothesis generation and testing. We used an existing mathematical model for T. brucei infection dynamics [36] to interpret the experimentally obtained data presented in Fig 2 and to provide us with testable hypotheses as to the differences in the infection dynamics observed between kDNA+ and kDNA0 cells. The cell types in this model are (i) non-committed replicating slender cells (i. e. slender cells not yet committed to stumpy formation), (ii) committed replicating slender cells, and (iii) non-proliferating differentiated cells, including both stumpy and intermediate cells (Fig 4A). We first compared two models: the published model, where the differentiation rate was proportional to SIF concentration [36], and a modified version of this model, where slender to stumpy differentiation rates are additionally influenced by a SIF-independent differentiation term [40]. The latter reflects a constant background level of slender form differentiation, independent of the concentration of SIF. Hence, each slender cell has a fixed probability of differentiating per cell cycle independently of SIF. The two terms are summed, with both terms acting to affect differentiation, with the SIF-dependent term only having a significant effect at high slender form concentrations due to the accumulation of high SIF levels. We inferred model parameter estimates (mean and confidence interval) by fitting both models to the data for all 16 mice using a Bayesian MC-MC method (S3 Fig). We next assessed the fit of each model by residual analysis (Fig 4B and S4A Fig). In the model containing SIF-independent differentiation, there was a lower number of outlying residuals, with most within the 95% predictive interval more of the time than in the model containing only SIF-dependent differentiation. This difference was apparent for the slender proportion on day 4. The model with additional SIF-independent differentiation captured the drop in slender proportion from 100% to 90% by day 4 (Fig 4B; S3 Fig, panels A, C, E and G, dark blue curves). In contrast, the model with only SIF-dependent differentiation could not capture this drop as there was an insufficient accumulation of SIF by this time to induce such a large amount of differentiation to stumpy forms, resulting in larger residuals (S4A Fig; S3 Fig, panels B, D, F and H, dark blue curves). Hence, this model overestimated the slender proportion on day 4. We also assessed the models by the Akaike information criterion (AIC), which measures the quality of a fit of a mathematical model to a set of data, taking into account the goodness of fit and the number of parameters estimated in the model [41]. The models with and without SIF-independent differentiation had AIC values of 2659 and 3033, respectively, hence the former is preferred as it has the lower AIC. When we reanalysed the infection data from an earlier study [36] by including the additional SIF-independent term, that model was also preferred when mathematically assessed by the AIC (S4B Fig). In conclusion, the mathematical model for within-host infection dynamics of T. brucei provided a better fit to experimental data when it included an additional term for SIF-independent slender-to-stumpy differentiation. This is consistent with the recent identification of a quorum sensing-independent path to stumpy development in this parasite [42]. We next used the optimised mathematical model to identify and quantitate the parameters predicted to be responsible for these differences between kDNA+ and kDNA0 cell lines. This resulted in two key observations. First, infection with kDNA+ parasites resulted in a broader peak of high cell density than infection with kDNA0 parasites (Fig 5A). The model predicts that the rise in parasitaemia levels off due to SIF-induced differentiation to the stumpy form: as slender forms proliferate, SIF begins to rise (S3 Fig, pink curves), which increases the rate of differentiation, and stumpy forms begin to emerge (S3 Fig, light orange curves). As the number of slender forms decrease, SIF concentration falls, causing the differentiation rate to fall. Stumpy cells disappear due to an intrinsically limited life span or immune clearance (Fig 4A). The estimates for total committed lifespan of kDNA+ and kDNA0 cells were 72–77 h and 41–51 h, respectively (Table 2). The total committed lifespan can be broken down into duration of the committed slender form and the duration of the stumpy form. Committed slender kDNA+ cells were estimated to survive longer than kDNA0 cells; they were predicted to have gone through at least one further cell cycle division than kDNA0 cells before they entered cell cycle arrest as the intermediate form (Table 2, ‘Committed slender replications’). Stumpy forms with kDNA lived on average for 56–62 h, whereas for kDNA0 cells the calculated average stumpy lifespan was predicted to be considerably lower, 36–49 h. Interestingly, the model predicted a clear difference between the immune responses to kDNA+ and kDNA0 stumpy cells. While the model estimated similar and consistent clearance rates for WT/WTγ and WT/L262Pγ stumpy cells, at ~20% per hour (Table 2), it estimated that immune clearance was not required to explain the disappearance of kDNA0 stumpy cells. We note that immune response against trypanosomes, although multifactorial, highly complex and incompletely understood, is represented by a simple step function in our model. There is insufficient data to support more realistic representations of the immune response. Nonetheless, our model fitted the experimental data very well and predicted that the narrower peak of high parasitaemia in kDNA0 parasites (see Fig 5A) was largely due to accelerated cell death of stumpy cells lacking kDNA. Second, in kDNA+ cells, a second peak in parasitaemia emerged around day 7, when a reduced SIF concentration allowed slender cells to proliferate again (Fig 5B). According to the model, the density fell again due to onset of immune killing (S3A and S3C Fig, yellow curves). Without immune killing the model predicts a continued rise in slender density to a much higher level. The absence of this second peak in kDNA0 parasites (Fig 5B) was explained by our model with an onset of immune killing about 1. 5 days earlier (S3E and S3G Fig, yellow curves; Table 2, ‘Start time of immune response’), which completely suppressed the second rise in parasitaemia. In summary, an optimised mathematical model for within-host infection dynamics that included an additional SIF-independent parameter for slender-to-stumpy differentiation provided a very good fit to experimental data and captured experimentally observed differences between kDNA+ and kDNA0 parasites. A narrower peak of high parasitaemia in the latter was predicted to be largely due to accelerated cell death of stumpy forms lacking kDNA. The shortened lifespan of kDNA0 stumpy forms (Fig 2 and Fig 3) pointed to loss of critical mitochondrial functions. A hallmark of functional mitochondria is the presence of ΔΨm. In BSF T. brucei ΔΨm is primarily generated by ATP hydrolysis-driven proton pumping of the F1FO-ATPase, whereas PCF T. brucei generate ΔΨm by proton pumping of respiratory complexes III and IV and, potentially, cI. It is not clear how the ΔΨm is generated in stumpy forms, although it has been reported to be sensitive to cI inhibitors but insensitive to the F1FO-ATPase inhibitor oligomycin [14]. To explore the role of kDNA in maintaining ΔΨm in stumpy forms, we harvested parasites from all five cell lines (WT/WTγ, WT/L262Pγ, and the three WT/L262Pγ kDNA0 cell lines) from infected mice at maximum parasitaemia, with a proportion of approximately 90% stumpy cells (S5A Fig), and stained them with the ΔΨm probe tetramethylrhodamine ethyl ester (TMRE) under various experimental conditions. First, we assessed ΔΨm in WT/WTγ cells in the presence or absence of azide, a specific inhibitor of the F1 moiety that disrupts ΔΨm production in slender BSF cells and kills these cells in 36–48 h [23]. Treatment with 0. 1–2 mM azide completely abolished ΔΨm (Fig 6A), indicating that F1 has an essential role in generating the ΔΨm in the stumpy form. Although treatment with 0. 5 mM azide eliminated ΔΨm, it did not reduce the viability of stumpy forms: in the absence of azide, the percentage of dead cells in the population increased from 0. 4% to 17. 2% after 24 h, and to 25. 3% after 48 h, and these percentages were not significantly increased in the presence of azide (Fig 6B and S5C Fig). This suggested that maintaining ΔΨm is not critical for the viability of stumpy forms. The role of the F1 ATPase in generating ΔΨm could be direct, as described above, or indirect, as in slender BSF kDNA0 cells. In the latter, the ATP/ADP carrier (AAC) acts to generate ΔΨm via the electrogenic exchange of matrix ADP3- for cytosolic ATP4-. F1 acts independently of FO by hydrolysing ATP4- to maintain an ATP/ADP ratio across the inner mitochondrial membrane that can sustain AAC activity [16,23]. We investigated whether this FO-independent pathway can function in stumpy forms by assessing ΔΨm in WT/L262Pγ (kDNA0) cells purified from mice. Live, freshly isolated kDNA0 stumpy cells were found to not have a ΔΨm (Fig 6C), demonstrating that the kDNA-encoded a subunit of the Fo-proton pore is required for ΔΨm generation and that this requirement cannot be circumvented by the L262Pγ mutation. Hence, the alternative, FO-independent mechanism of ΔΨm generation that functions in kDNA0 slender T. brucei and in subspecies T. b. evansi and T. b. equiperdum cannot operate in stumpy forms. The stumpy life cycle stage is preadapted to differentiation to the PCF in the midgut of the tsetse fly. Stumpy forms can use glycolysis or, alternatively, mitochondrial catabolism of α-KG as energy sources [9,10,13], which reflects the shift in metabolism towards the glucose-deficient environment of the tsetse midgut. We assessed the ability of kDNA0 stumpy forms to survive in the presence of glucose or α-KG. Although kDNA0 stumpy forms in the presence of glucose showed normal viability after 24 h, more than 70% of cells had died after 24 h of incubation with α-KG as sole major carbon source (Fig 7A). When α-KG was provided in addition to glucose it had little, if any, detrimental effects on kDNA0 cells (S5D Fig). The addition of N-acetyl glucosamine (GlcNAc), a non-metabolized glucose analog, to prevent uptake of residual glucose present in fetal calf serum (FCS) [43] further reduced the number of surviving cells (Fig 7A). The viability of kDNA+ control cells was comparable for medium with α-KG vs. glucose as main carbon source, and addition of GlcNAc to medium with α-KG had no negative effects on WT/WTγ cells (S5D Fig), confirming that GlcNAc only interferes with glucose-based energy metabolism. These results demonstrate that, unlike kDNA+ stumpy forms, kDNA0 parasites are unable to use α-KG to sustain viability. Interestingly, stumpy WT/WTγ cells treated with azide to inhibit generation of a ΔΨm died within 24 h if α-KG was the sole carbon source, but azide had little effect in the presence of glucose (Fig 7B), indicating that ΔΨm may be required for the entry of α-KG into the mitochondrion or for efficient export of mitochondrially produced ATP. In summary, these experiments suggest that the lack of a ΔΨm in stumpy cells without kDNA precludes the use of α-KG to satisfy the energy needs of these cells. Our mouse infections with cell lines with the genotypes WT/WTγ, WT/L262Pγ and WT/L262Pγ (kDNA0) and quantification of within-host dynamics using mathematical modelling showed that lack of kDNA did not affect the rate of differentiation into stumpy forms. The morphological changes we observed during slender to stumpy differentiation were similar for kDNA+ and kDNA0 cells. These results confirm conclusions from an earlier study that had investigated differentiation of kDNA-depleted T. brucei cells obtained by treatment with acriflavine for 24 h [18], but that study could not rule out that some kDNA-encoded factors had persisted after treatment. As kDNA0 slender BSF T. brucei are able to transition to the stumpy form as efficiently as cells that have their kDNA intact, we conclude that the absence of kDNA is not the primary reason why the dyskinetoplastic subspecies T. b. evansi and T. b. equiperdum are generally monomorphic [29,31]. A different molecular mechanism must therefore prevent naturally occurring kDNA0 or kDNA- T. brucei subspecies from differentiating to the stumpy life cycle form, such as loss of function in components of the SIF secretion or stumpy induction pathways [1], in a fashion similar to monomorphic T. b. brucei BSF forms [2]. SIF-dependent differentiation to the stumpy form places a limit on the parasitaemia level, presumably in part to extend the lifespan of the host [1]. As the probability of mechanical transmission of trypanosomes increases with the levels of parasitaemia in the blood [45], preventing slender to stumpy differentiation could thus have been a key event in the evolution of T. b. evansi, as has been discussed elsewhere [46,47]. Genome sequences from a number of T. b. evansi isolates are now available [26,48] and could be mined for candidate mutations in differentiation pathways. A few T. b. evansi strains have historically been reported to have some limited capacity to produce stumpy forms (for example [30]); this could be due to SIF-independent background differentiation or residual and inefficient SIF-dependent differentiation. We found that a mathematical model including SIF-independent differentiation provided a better fit to the experimental data from both the current study and a previous infection study [36] than a model only including SIF-dependent differentiation. A recent study provided biological evidence for SIF-independent differentiation in trypanosome infections [42] and there is emerging evidence for a background level of random differentiation in Plasmodium and Theileria infections [49–51]. Thus, our study supports the view that stochastic, low level differentiation events occurring in parallel to a signal transduction-type of differentiation could be a more broadly conserved aspect of infections with protist parasites. Although kDNA0 T. brucei differentiated into stumpy forms with the same efficiency as control cells, we observed some important differences in other aspects of their within-host dynamics. Firstly, kDNA0 cells showed a slightly lower growth rate up to day 4 of infection, resulting in a delayed rise in cell numbers leading up to the first peak of parasitaemia. In contrast, growth rates of kDNA0 and kDNA+ parasites were very similar when cultured in rich medium in vitro. Potential explanations are that kDNA0 parasites are more affected by the more limiting growth conditions in the host environment, or that they are more sensitive to attack by the host’s immune system, or both. Our mathematical model predicts that the immune response does not significantly affect infection dynamics until day 6 or 7 (although there appear to be differences for kDNA0 vs. kDNA+ parasites, see below), but as the action of the immune system during a trypanosome infection is not fully understood, our modelling of this aspect is necessarily an oversimplification. Understanding these differences will require further investigation, for example by comparing growth rates in minimal medium [52] and by investigating infection dynamics in immunosuppressed mice. Secondly, a second peak in slender parasitaemia was evident in kDNA+ cells around days 7–8 post infection, but completely absent in kDNA0 cells. According to the mathematical model, this second slender peak in infections with kDNA+ cells was due to SIF-dependent differentiation causing slender density, and therefore SIF density, to fall around day 6, allowing the remaining slender cells to begin proliferation rather than entering cell cycle arrest. A strong immune response around day 8 then prevented a further rise in parasitaemia. The model predicts that an earlier onset of immune killing in infections with kDNA0 cells was responsible for completely suppressing this second peak in these parasites. This surprising result requires further investigation; we speculate that kDNA0 cells could be less efficient at VSG switching or production, allowing more efficient clearance of slender cells in the earlier stage of the infection. Alternatively, kDNA0 cells could be less able to access potentially immune privileged body compartments [53,54], or they could swim more slowly or in a different way, preventing the efficient clearance of antibody that is mediated by swimming [55]. Finally, and most importantly, we observed a substantially shorter lifespan of kDNA0 stumpy forms. For kDNA+ T. brucei we determined an average value for ‘duration of stumpy form’ of 56–62 h. This is in good agreement with other reports using a mouse infection model [36,38]. There are no reports on stumpy cell lifespan in other hosts, or how it might be affected by parasite distribution in different tissues [53], but our in vitro survival assay showed that >90% of kDNA+ stumpy cells had perished 70 h after isolation from a mouse, i. e. within a time span comparable to the one observed in vivo. In marked contrast, we determined substantially shorter lifespans for kDNA0 stumpy cells, both in vivo (duration of stumpy forms 36–49 h) as well as in vitro. This indicated that the underlying cause was intrinsic to the parasites, rather than due to faster immune clearance. The mechanism of cell death in stumpy forms is not understood, but an early event in programmed cell death in other organisms can be loss of ΔΨm [56–58]. Furthermore, ΔΨm is a key indicator of mitochondrial health [59], it is essential for mitochondrial protein import and other transport processes [60], and its generation in both BSF and PCF T. brucei depends on kDNA-encoded proteins [23,44]. In the present study we show that the F1-ATPase inhibitor azide completely abolished ΔΨm in kDNA+ stumpy cells, suggesting its generation by the F1FO-ATP synthase functioning as a proton pump, as in slender BSF cells [20,21,23,24]. We propose that the switch in directionality of this enzyme from ATPase to ATP synthase activity occurs during the transition from stumpy BSF parasites to PCF parasites. This is also consistent with the increase of the IF1 protein during that transition measured in a recent proteomics study [61]. IF1 (Tb927. 10. 2970) is a specific inhibitor of the ATP hydrolase activity of the F1FO-ATP synthase [62] and shows strict developmental regulation in T. brucei, with repression in slender BSF and expression in PCF [63]. An earlier study had reported that ΔΨm in stumpy forms was sensitive to the cI inhibitor rotenone but insensitive to the F1FO-ATP synthase inhibitor oligomycin [14]; the authors of that study had concluded that cI generates ΔΨm in stumpy forms, with the F1FO-ATP synthase acting in ATP synthesis mode, driven by the proton motive force. One possible explanation for this apparent discrepancy is that the relatively high concentration of rotenone used in the earlier study had caused non-specific effects, as has been argued by others [64]. Future studies with genetic mutants for specific subunits of cI and the F1FO-ATP synthase in a pleomorphic T. brucei strain will be required to investigate this apparent discrepancy further. We did not detect a ΔΨm in kDNA0 stumpy cells, indicating that the alternative, FO-independent mechanism for generating ΔΨm enabled by the L262Pγ mutation in slender T. brucei [16] cannot operate in the stumpy life cycle stage. This alternative mechanism depends on electrogenic exchange of matrix ADP3- for cytosolic ATP4- by the AAC and continued ATP hydrolysis by F1, perhaps in vicinity of the AAC, to maintain a suitable ATP/ADP ratio across the inner mitochondrial membrane [16,23,65]. Significant mitochondrial ATP production would be expected to thwart this mechanism, and indeed there is evidence for this occurring in stumpy form T. brucei [10,12–14]. Conceivably this could occur via F1FO-ATP synthase activity, as mentioned above, or, more consistent with our data, via substrate level phosphorylation involving SCoAS and, depending on the carbon source, ASCT [3,66,67]. Pyruvate from glycolysis can be catabolised by stumpy cells to acetate, with ATP production via the ASCT / SCoAS cycle (Fig 8) [12]. It was also demonstrated that motility of stumpy cells, but not of slender BSF cells, can be sustained in vitro with α-KG as sole carbon source [9], with mainly succinate as end product [10] and ATP production via SCoAS (Fig 8). A putative mitochondrial α-KG transporter, termed MCP12 (Tb927. 10. 12840), has been identified and functionally characterised in T. brucei [68,69], and a proteomics study reported ~20-fold upregulation of this protein in stumpy cells compared to slender cells [61]. Potentially, pyruvate could be converted to α-KG via L-alanine aminotransferase (Fig 8), an enzyme expressed in BSF and PCF T. brucei [70]. This step would require glutamate as co-substrate, which could be obtained directly from the medium or via proline catabolism. We confirmed that kDNA+ stumpy cells maintain viability for at least 24 h when incubated in minimal medium supplemented with glucose or α-KG. Nearly 100% of kDNA0 stumpy cells survived for at least 24 h when medium was supplemented with glucose, but the survival rate dropped to ~20% when provided with α-KG instead of glucose, and suppressing uptake of residual glucose with GlcNAc resulted in a further drop to less than 10% survivors. We also found that azide, which abolished ΔΨm in kDNA+ stumpy cells, prevented survival of these cells in minimal medium supplemented with α-KG, while it did not affect survival in the presence of glucose. At least two scenarios that are not mutually exclusive could explain these results. Firstly, ΔΨm could be required for mitochondrial uptake of α-KG. The transporter identified in T. brucei was proposed to be an α-KG/malate antiporter [69], analogous to the mammalian enzyme, although this has not yet been confirmed experimentally. In that case α-KG import would not be directly dependent on ΔΨm. The E. coli enzyme is an α-KG/proton symporter that depends on a proton motive force [71], but its closest homolog in T. brucei is a myo-inositol/proton symporter in the Golgi [72]. Secondly, α-KG import could be ΔΨm-independent and still drive mitochondrial substrate phosphorylation in the absence of kDNA or presence of azide (Fig 8), but in the absence of ΔΨm, ATP may not reach the cytosol in sufficient quantities to sustain viability: ATP4-/ADP3- exchange by the AAC is driven by the concentration gradient of the substrates as well as ΔΨm [73,74]. Resolving which of these scenarios, if any, is correct will require further experimental evidence. In summary, these experiments demonstrate clear differences in physiology and metabolic capacity of stumpy cells with and without kDNA. Although we found clear evidence for deficiencies in mitochondrial function in kDNA0 stumpy cells, correlating any of these deficiencies to the reduced lifespan was not straightforward. The most prominent defect of kDNA0 stumpy cells that we identified in this study was lack of a ΔΨm. However, the ability of stumpy cells lacking a ΔΨm (i. e. kDNA0 cells or kDNA+ cells in the presence of azide) to survive for at least 48 h in medium provided with glucose suggests that sufficient amounts of ATP can be produced via glycolysis in the absence of a ΔΨm, at least in the short term. In the long term, ΔΨm-dependent mitochondrial transport processes such as continued import of the alternative oxidase, are vital for sustained glycolysis in proliferating parasites [7], but this may be less relevant for cell-cycle arrested stumpy forms with their intrinsically limited life span. If loss of ΔΨm does not affect viability of kDNA+ stumpy cells, what is the cause of the reduced lifespan in kDNA0 stumpy cells? One possibility is an impaired redox balance in the mitochondrial matrix caused by loss of kDNA. At least seven subunits of cI are kDNA-encoded [75], and therefore kDNA0 cells will be cI-deficient. Activity of this enzyme is dispensable for slender BSF, at least in vitro and in the bloodstream [75], probably in part due to the presence of an alternative type 2 NADH dehydrogenase [76]. Differentiation into stumpy cells has long been known to be associated with a dramatic increase in ‘NAD diaphorase’ activity [9] (an assay for NADH dehydrogenase activity), and we note that both pathways for mitochondrial substrate phosphorylation are dependent on recycling of NADH (Fig 8). Our study shows that kDNA in the sleeping sickness parasite T. brucei is not required for differentiation into the transmissible stumpy stage, but that it is critical for the longevity of this stage and for generation of its ΔΨm. We identified three important differences to slender BSF T. brucei: (i) a L262P mutation in the nuclear-encoded ATPase subunit γ does not enable kDNA-independent generation of ΔΨm, most likely because of considerable mitochondrial ATP production; (ii) loss of ΔΨm does not affect the life span of stumpy T. brucei, presumably because life span is limited by other factors that come into play before loss of ΔΨm-dependent processes can take their toll; and (iii) stumpy form viability depends on kDNA-encoded genes other than FO subunit a. Future studies should, for example, assess the consequences of loss of function of respiratory complex I and the F1FO-ATP synthase on stumpy cell viability with specific genetic mutants and seek to identify the intrinsic factors that limit stumpy cell life span. Culture-adapted pleomorphic T. brucei EATRO 1125 AnTat1. 1 90: 13 parasites [34] were transfected with plasmid pEnT6-γL262P-PURO or pEnT6-γWT-PURO. These plasmids are based on the pEnT6 backbone [77] and contain either F1FO-ATPase subunit γ (systematic TriTrypDB ID Tb927. 10. 180) with the L262P mutation (L262Pγ, [16]) or a wild type version (WTγ); they allow the replacement of one endogenous ATPase γ subunit allele in order to generate cell lines containing a single L262Pγ allele (ATPγ/Δatpγ: : atpγL262P PURO) or, as a control, WTγ (ATPγ/Δatpγ: : atpγWT PURO). The replaced gene is expressed by read-through transcription of the endogenous locus and contains its native 5’ UTR, but the aldolase 3’UTR. For the transfection, the AMAXA Nucleofector II was used with nucleofection solution (90 mM NaH2PO4,5 mM KCl, 0. 15 M CaCl2,50 mM HEPES, pH 7. 3) [78] and program Z-001. T. brucei EATRO 1125 AnTat1. 1 90: 13 clones were selected after 4 days and were maintained in 2. 5 μg/ml G418,5 μg/ml hygromycin and 0. 1 μg/ml puromycin in HMI-9 medium [79] containing 10% (v/v) fetal calf serum (FCS; Gibco). The ATPase γ subunit gene was amplified from genomic DNA via PCR, allowing direct Sanger sequencing of the gel-extracted PCR product to confirm the presence or absence of the L262Pγ mutation using primers 5’-CGG CGG CCG CAT GTC AGG TAA ACT TCG TCT TTA CAA AG-3' (forward) and 5' -ATA GGA TCC CTA CTT GGT TAC TGC CCC TTC CCA G-3' (reverse). WT/L262Pγ cells were treated with 10 nM acriflavine (Sigma) over 3 days; loss of kDNA was assessed by preparing microscope slides and mounting with a cover slip using 50 μl Prolong Gold Antifade with 4’, 6-diamidino-2-phenylindole (DAPI; Life Tech.). To confirm loss of maxicircle genes and of a representative minicircle (type A-like) [16,80] by PCR, total DNA was extracted after expanding the cell culture for a further two days in the absence of acriflavine. The PCR assay was carried out exactly as described in Dean et al. (2013). WT/L262Pγ kDNA0 clone #3 was generated without drug treatment; the cell line lost its kDNA spontaneously after 6 weeks of growth in HMI-9,10% (v/v) FCS. Cells were grown in the presence or absence of 10 nM ethidium bromide (EtBr; Sigma). Cell counts were performed daily using a Beckmann Z2 Coulter counter, and cultures were split to a concentration of 1x105/ml after counting. All animal experiments were carried out in adult MF1 mice after local ethical approval at the University of Edinburgh. All animal experiments were carried out by Caroline Dewar, working under personal license I3997C068 and project licences 60/4373 (Professor Keith Matthews) and 70/8734 (Professor Achim Schnaufer), granted by the UK Home Office under the Animals (Scientific Procedures) Act 1986, section 5. Sex- and age-matched MF1 mice were infected with T. brucei EATRO 1125 AnTat1. 1 90: 13 cells (suspended in 200 μl HMI-9) via intraperitoneal (IP) injection. No immunosuppressant was used. Parasitaemia was monitored by obtaining blood via a tail snip, compressing a drop of blood under a cover slip on a microscope slide, and counting parasites at 400x magnification. Five μl blood was also taken for an immunofluorescence assay and cell cycle analysis. Morphology counts were performed as described [36]. Methanol-fixed blood smear slides were blinded by a colleague with respect to cell line, day and time point to prevent bias. Morphology was scored from these slides independently by two individuals. Parasitaemia was judged by eye, based on the Rapid Matching method [36,81]. This method entails an upper limit of 64 parasites per field of view, correlating to a density of 2. 5x108 cells/ml, above which it becomes difficult to estimate counts accurately. Stumpy form trypanosomes were purified from blood using DEAE-cellulose DE52 (Whatman) anionic exchange columns [82] that were preincubated with PSG (44 mM NaCl, 57 mM Na2HPO4,3 mM KH2PO4,55 mM glucose, pH 7. 8) warmed to 37°C. Western blotting and antibody concentrations were as described previously [35]. Anti-EF1α (Millipore) was used at a dilution of 1/7000. Proteins were detected using Enhanced Chemiluminescence reagents (Amersham) and a SRX-101A X-ray developer (Konica Minolta). A mathematical model described previously [36] was modified to include a parameter for SIF-independent slender to stumpy differentiation. The model was constructed as follows. Let the concentration of non-committed slender cells at time t be L (t). The initial infection is at time t = 0. Non-committed slender cells replicate at rate α (i. e. , a cell-cycle time of ln (2) /α) ). They are cleared by a time-dependent immune response at rate rL (t). They become committed to differentiate at rate βb + βff (t), where f (t) is SIF concentration, βb is the background, SIF independent differentiation rate and βf is the SIF dependent differentiation rate. Therefore, the differential equation that describes the dynamics of non-committed slender forms is: ddtL (t) =[α−βb−βff (t) −rL (t) ]L (t) Let the age of differentiated cells since becoming committed to differentiation be a and let d (a, t) be the age density distribution of differentiated cells at time t. Differentiated cells fall into two classes: i) replicating, committed slender cells, and ii) non-replicating stumpy cells. Committed slender cells replicate at rate α, are assumed to be cleared by the immune system at the same rate as non-committed slender cells (rL (t) ), and develop into stumpy cells at age τC. Stumpy cells do not replicate, they are assumed to be cleared by the immune response at a different rate rS (t), and they die at age τS. Thus, the partial differential equation that describes the dynamics of the age density distribution of committed cells is ∂∂td (a, t) +∂∂ad (a, t) =−d (a, t) ×{rL (t) −αif0≤a<τCrS (t) ifτC≤a<τS The boundary conditions on these equations are determined by differentiation of non-committed slender cells into age a = 0, i. e. , d (0, t) = [βb + βff (t) ]L (t), and stumpy death at age τS, i. e. , d (τS, t) = 0. Let C (t) be the total concentration of committed slender cells, let S (t) be the total concentration of stumpy cells, and let T (t) be the total concentration of all cells. These are given by: C (t) =∫d (a, t) daS (t) =∫d (a, t) daT (t) =L (t) +C (t) +S (t) SIF is produced by both non-committed and committed slender cells. SIF is removed at rate γ. Therefore, the differential equation describing the dynamics of SIF concentration is: ddtf (t) =L (t) +C (t) −γf (t) Note that, because SIF is not measured, its concentration is on a dimensionless scale. The immune response against trypanosomes is multifactorial and highly complex, and only qualitatively understood at best. A detailed mathematical model of the immune response was, therefore, of little use when no data were available to fit to. Instead, we used a simple step function to represent an immune response switching from an inactive to an active state at a time T post infection. The strengths of the immune responses against slender and stumpy cells are assumed to be different. They are given by the equations rL (t) ={0ift<TφLift≥T and rS (t) ={0ift<TφSift≥T where ϕL is the removal rate of slender cells and ϕS is the removal rate of stumpy cells. Naive mice are infected with non-committed slender cells at a concentration L0. Therefore the initial conditions are L (0) = L0, d (a, 0) = 0 for all a and f (0) = 0. These imply C (0) = S (0) = T (0) = 0. All variables and parameters are listed in Table 3. For particular numerical values of the model parameters, the model was solved numerically for each mouse. In order to quantify the fit of the model with these parameters to the data, the log-likelihood of the model solution at each data point by was calculated. Parasite density was estimated by observing a field of cells and estimating the number of parasites in the field. Due to the difficulty of observing many moving parasites in a microscopic field, density estimates were categorised into 0,1, 2,3, 4,5, 6,8, 10,12,16,20,24,32,48,64, and 92 parasites per field. Parameter ρti describes the expected density of parasites at time ti (which is obtained from the model). The volume of blood v, in a microscopic field is v = 25. 6 × 10−8μl. The expected number of parasites per field therefore is λ=vρti. The number of parasites N in a field is Poisson distributed with parameter λ. If N equals 0 to 5 then the likelihood of ρti is equal to λNe−λN! . If N is greater than 5 then it can be assumed that the number of parasites lies within a range. The start of the range is the midpoint between the previous category and the assigned category. For example, if the number of parasites in a field is estimated to be about 48 parasites, then the assigned category is 48, the start of the range is Nl = (32 + 48) /2 = 40 parasites. Similarly, the end of the range is the midpoint between the next category and the assigned category, for example Nu = (48 + 64) /2 = 56. The likelihood of ρti is then equal to ∑i=NlNhλie−λi! which equals Q (Nh, λ) − Q (Nl, λ) where Q is the normalised incomplete Gamma function. The likelihood function also includes the proportion of parasites that are slender forms at a time ti. The number X of parasites that have slender morphology is binomially distributed with parameters M and p, where M is the number of parasites observed and p is the predicted proportion that are slender forms (obtained from the model). Thus, the likelihood of pti is proportional to ptiXti (1−pti) Mti−Xti. The parameter posterior distribution was found by multiplying the likelihood, which is the product of likelihoods at each time point, by the prior distributions, which were taken from [36]. The prior on βb was NT (0. 01,0. 012), a normal distribution truncated at 0. Samples from the posterior were drawn using an adaptive population based Markov chain Monte Carlo algorithm with power posteriors [83,84]. A 5 μl blood sample was pipetted into 100 μl ice cold vPBS (pH 7. 4; 137 mM NaCl, 2. 7 mM KCl, 10 mM Na2HPO4,1. 8 mM KH2PO4,46 mM sucrose, 10 mM glucose), washed and cells were fixed for 10 min by ice cold 3% (w/v) paraformaldehyde (Fisher). 130 μl 0. 2 M glycine was added to allow sample storage. Cells were pelleted, washed in PBS (pH 7. 4; 137 mM NaCl, 2. 7 mM KCl, 10 mM Na2HPO4,1. 8 mM KH2PO4), and resuspended in 500 μl PBS with 100 ng/ml DAPI or 5 μg/ml Hoechst 33342 DNA staining dye (Life Tech.). Cells were analysed by flow cytometry (peak excitation 358 nm, peak emission 461 nm) using a Becton Dickinson LSRII machine with BD FACSDiva software. 2x104 events per sample were measured. Results were analysed with FlowJo software. Cells were harvested from a mouse infection during peak parasitaemia whilst the population was approximately 90% stumpy form. After purification from blood, parasites were washed in PBS-G (PBS, 6 mM glucose) and resuspended in either HMI-9 containing 10% (v/v) FCS or a modified minimal medium (CMM) [52] containing 10% (v/v) FCS and devoid of glucose. Supplements (25 mM glucose, 25 mM α-KG, 50 mM N-acetyl glucosamine, all from Sigma) were added as required. Cells were harvested from culture and washed in sterile warm PBS-G. The pellet was resuspended in PBS-G with 10 μM CFDA-SE (ThermoFisher), and incubated for 15 mins at 37°C. Cells were washed with HMI-9 medium and incubated in HMI-9 for 30 mins at 37°C. Cells were then washed, and fixed with 3. 7% (v/v) formaldehyde for 10 min (a detailed paper on validating CFDA-SE staining as a live/dead assay compatible with fixation of cells will be published elsewhere). Fixative was washed out with PBS-G, and cells were resuspended in PBS-G plus 5 μg/ml Hoechst 33342 (Life Tech.). Samples were analysed with excitation peak of 492 nm and emission peak of 517 nm for CFDA-SE on a BD LSRII instrument. For propidium iodide (PI) staining, 1 μl 500 ng/ml PI was added to the final resuspension before analysis. Samples were analysed at peak excitation at 488 nm and emission at 695 nm. The TMRE Mitochondrial Membrane Potential kit (Abcam) was used. All samples were supplemented with 100 nM TMRE and left at 37°C for 20 min. Cells were preincubated with 20 μM carbonyl cyanide-4- (trifluoromethoxy) phenylhydrazone (FCCP) for 10 min or with 0. 1–2 mM sodium azide for 2 h. Cells were pelleted, and washed in 0. 2% BSA in PBS. Cell pellets were resuspended in 0. 2% BSA in PBS containing 5 μg/ml Hoechst 33342 DNA staining dye before analysis on a BD LSRII instrument, with peak excitation at 549 nm and peak emission at 575 nm for TMRE. Images were captured using a Retiga 2000R Mono Cooled charged-coupled device camera attached to an Axioscope 2 or Axioimager Z2 (Carl Zeiss MicroImaging, Inc.) using either Plan-Apochromat 63x (1. 40 NA) or Plan-Apochromat 100x (1. 40 NA) phase-contrast objectives.

African trypanosomes are single cellular eukaryotes transmitted by tsetse flies that cause important diseases in humans and their livestock. For survival these parasites depend on their mitochondrion, an organelle that has its own genome (‘mtDNA’) and that is traditionally viewed as the ‘power plant’ of cells, but that has other essential roles as well. Interfering with mtDNA is an important part of how some anti-trypanosomatid drugs work. Most mitochondrial research in trypanosomes in the past has focused on forms of the sleeping sickness parasite that proliferate in the mammalian bloodstream or in the insect midgut, as these can be readily cultured in the laboratory. As a consequence, relatively little is known about mitochondrial biology of the so-called ‘stumpy’ form, a non-replicative stage that is critical for transmission to tsetse flies. In this study we use a mouse infection model to show that a certain gene mutation permits formation of stumpy forms that lack mtDNA. However, without mtDNA these parasites have an altered mitochondrial metabolism and a reduced life span, which tells us that stumpy forms depend on additional mtDNA-encoded genes that are not required by the proliferative bloodstream form.

Abstract Introduction Results Discussion Materials and methods

medicine and health sciences chemical compounds parasitic cell cycles parasitic diseases parasitic protozoans cell differentiation parasitology trypanosoma brucei developmental biology trypanosoma brucei gambiense protozoans mitochondria bioenergetics cellular structures and organelles kinetoplasts life cycles chemistry azides biochemistry trypanosoma eukaryota cell biology biology and life sciences physical sciences energy-producing organelles organisms parasitic life cycles

2018

Mitochondrial DNA is critical for longevity and metabolism of transmission stage Trypanosoma brucei

Apicomplexan parasites depend on the invasion of host cells for survival and proliferation. Calcium-dependent signaling pathways appear to be essential for micronemal release and gliding motility, yet the target of activated kinases remains largely unknown. We have characterized calcium-dependent phosphorylation events during Toxoplasma host cell invasion. Stimulation of live tachyzoites with Ca2+-mobilizing drugs leads to phosphorylation of numerous parasite proteins, as shown by differential 2-DE display of 32[P]-labeled protein extracts. Multi-dimensional Protein Identification Technology (MudPIT) identified ∼546 phosphorylation sites on over 300 Toxoplasma proteins, including 10 sites on the actomyosin invasion motor. Using a Stable Isotope of Amino Acids in Culture (SILAC) -based quantitative LC-MS/MS analyses we monitored changes in the abundance and phosphorylation of the invasion motor complex and defined Ca2+-dependent phosphorylation patterns on three of its components - GAP45, MLC1 and MyoA. Furthermore, calcium-dependent phosphorylation of six residues across GAP45, MLC1 and MyoA is correlated with invasion motor activity. By analyzing proteins that appear to associate more strongly with the invasion motor upon calcium stimulation we have also identified a novel 15-kDa Calmodulin-like protein that likely represents the MyoA Essential Light Chain of the Toxoplasma invasion motor. This suggests that invasion motor activity could be regulated not only by phosphorylation but also by the direct binding of calcium ions to this new component. The phylum Apicomplexa is a large group of obligate intracellular parasites of wide medical and agricultural significance. Alone, Toxoplasma gondii infects between 30- 80% of people worldwide and is one the most common infectious agents of humans. Toxoplasma transmission occurs by exposure to the feces of an infected cat, eating undercooked meat or ingestion of contaminated water harboring oocysts [1]. An ocular infection route is also common and is a leading cause of blindness in some countries [2]. Primary exposure or reactivation of tissue cysts in pregnant women can lead to congenital birth defects and spontaneous abortion, while toxoplasmosis is a common secondary infection of AIDS patients and other immuno-compromised individuals and can lead to death if untreated. Invasion of host cells by apicomplexan parasites is an obligatory step for their survival and proliferation. Despite each species having a range of host and cell types that they target the intracellular processes governing invasion appear to be largely conserved. Toxoplasma’s ease of growth in the laboratory and high genetic tractability has provided researchers with an excellent model for investigating the molecular basis of invasion in related species such as Plasmodium spp - the causative agent of malaria. For example, a highly conserved actomyosin-based invasion motor that drives parasite motility and invasion was first identified and has been largely characterized in Toxoplasma. At its core the invasion motor consists of a novel class XIV Myosin, MyoA [3] and Myosin Light Chain 1, MLC1 [4] which is anchored into the outer side of the inner membrane complex (IMC) by the Glideosome-Associated Proteins GAP40, GAP45, GAP50 and GAPM’s [5], [6], [7], [8]. Recently, the architecture of the invasion motor has been mapped and the central role and absolute requirement of GAP45 has been demonstrated [7]. Furthermore, it was shown that GAP45 spans the supra-alveolar space providing cohesion between the IMC and plasma membrane [7]. The current model of invasion suggests that upon host cell contact GAP40/45/50, GAPM’s, MyoA and MLC1 complex and filamentous actin forms [9]. Transmembrane host cell adhesins linked to actin filaments through the glycolytic enzyme aldolase are then pulled rearwards by the action of MyoA, thus driving the parasite forward into the host cell [10]. The release of apical organelles is also required for successful invasion. Micronemes are first to be secreted and contain high affinity host cell ligands that are necessary for strong host cell attachment and invasion [11]. Rhoptries, large club-shaped organelles, rich in lipids are released after micronemes upon contact of the apical end of the parasite with the host cell. Rhoptry release is correlated with host membrane penetration and the creation of the parasitophorous vacuole [12]. The molecular machinery that drives exocytosis of these apical organelles is unknown. Activation of the invasion motor and regulated release of apical organelles occurs after a change in extracellular environment and contact with the host cell [13]. Pharmacologically, calcium-dependent signal transduction pathways have been identified in Toxoplasma and P. falciparum and appear to be essential for micronemal release, motility and invasion [14], [15]. Calcium mobilizing agents potently induce exocytosis of micronemes and gliding motility, whereas membrane permeable calcium chelators such as BAPTA-AM inhibit these processes [11], [15]. In addition, ethanol induces a transient [Ca2+]i increase and micronemal release via a putative signaling pathway involving the activation of phospholipase C (PLC) [16]. Further supporting a role of calcium signaling in apicomplexan parasite invasion is the visualization of intracellular calcium ([Ca2+]i) oscillations that accompany invasion processes in Toxoplasma [14], [17] and Plasmodium [15]. Together, this data indicates that [Ca2+]i signals mediate multiple signaling pathways, including micronemal release and gliding motility and these are essential for regulating the processes involved in host cell invasion. In mammalian cells, spatiotemporally distinct [Ca2+]i signals are translated into enzymatic activity using a variety of serine/threonine-specific protein kinases and phosphatases, which control the function of downstream effectors by promoting local changes in protein conformation, interactions and activity. In Apicomplexa, a conserved family of plant-like Calcium-Dependent Protein Kinases (CDPKs) has been strongly implicated in mediating several critical Ca2+-dependent signal transduction pathways during the complex parasite life cycles [18]. Recent chemical genetic and conditional gene knockout approaches have demonstrated that Toxoplasma CDPK1 is an essential regulator of Ca2+-dependent micronemal exocytosis [19], [20], whereas P. falciparum CDPK1 (PfCDPK1) appears to play a role in regulating red blood cell invasion [21]. Furthermore, PfCDPK1 phosphorylates P. falciparum Myosin A Tail Interacting Protein (PfMTIP) and Glideosome-Associated Protein 45 (PfGAP45) in vitro [21], [22], which suggests a role in regulating invasion motor function during host cell invasion. Together, this data suggests that CDPK’s play crucial roles in regulating invasion processes, yet the in vivo substrates of these kinases remain largely unexplored. To identify substrates of calcium-dependent kinases we have employed proteomics approaches to obtain a global snapshot of the phosphorylation pattern of Toxoplasma proteins following stimulation of Ca2+ signaling pathways. This identified targets of Ca2+-dependent phosphorylation pathways potentially involved in regulating invasion processes. Phosphorylation sites on Toxoplasma GAP45, MLC1 and MyoA were identified and given that their is calcium-dependent phosphorylation deposition on only some sites and the presence of a phospho-tyrosine it appears that the phosphorylation status of the invasion motor is likely regulated by multiple kinases. Further, a novel protein was identified as a new component of the Toxoplasma invasion motor that associates more tightly upon calcium signaling suggesting that the invasion motor may also be directly regulated by Ca2+ binding. To identify Ca2+-dependent phospho-substrates during Toxoplasma invasion we conducted a 2-DE-based screen by labeling parasites with 32[P] orthophosphate and stimulating with either ethanol (thought to activate PLC) or the calcium ionophore ionomycin. Both calcium pathway agonists have previously been shown to induce micronemal secretion and motility after short periods of stimulation [16]. As a control parasite samples were either mock treated (DMSO alone) or pretreated with the membrane permeable calcium chelator BAPTA-AM followed by stimulation with ethanol. Extracts were separated on 2-DE gels and over 50 32[P]-labeled protein spots were identified that were absent or barely detectable in BAPTA-AM or mock-treated negative controls (Figure 1A, C) and exclusively present in samples stimulated with either ethanol or ionomycin (Figure 1B, D – marked with red arrows). False colour 2-DE image overlays of corresponding autoradiographs highlight 32[P]-labeled protein spots that were specifically phosphorylated following stimulation with Ca2+ mobilizing drugs (Figure 1E). Spots that appear yellow likely represent Ca2+-insensitive phosphoproteins or GPI-anchored parasite proteins which are unaffected by stimulation, whereas spots that appear red represent potential substrates of calcium-dependent phosphorylation (Figure 1E). Interestingly, the same proteins appeared to be phosphorylated when stimulated by either calcium ionophore or ethanol (Figure 1B, D), strongly suggesting that both agonists stimulate the same signal transduction pathway (s) in Toxoplasma tachyzoites. A total of 50 2-DE spots that were consistently phosphorylated following Ca2+ pathway stimulation were subjected to LC-MS/MS-based protein identification (Figures 1E, Supplementary Figure S1 and Table S1). In some instances proteins from separate 2-DE spots shared the same identity, as indicated by bounding boxes (Figure 1E, Table S1). Assuming that phosphorylated 2-DE spots are the product of a single gene the present data identified a total of 63 Toxoplasma proteins - some of which represent molecules potentially involved in mediating intracellular signaling cascades, regulating exocytosis of invasion organelles, or controlling parasite motility (Table S1). These included amongst others, a cAMP kinase regulatory subunit (PKA-R, TGME49_042070), a casein kinase subunit II beta subunit (CKIIß, TGME49_072400), a soluble NSF-attachment protein (SNAP, TGME49_018760), an armadillo repeat-containing protein (ARM1, TGME49_061440), a serine/threonine phosphatase 2C (PP2C, TGME49_054770), a Rab GTPase (RAB5B, TGME49_007460), two highly acidic proteins similar to I1PP2A/ANP32 (ANP32; TGME49_071810) or I2PP2A/SET (SET; TGME49_044110), as well as a number of putative uncharacterized proteins (ie. Hypo) (Figure 1E). Importantly, previously identified phosphoproteins were also identified in our analysis including Toxofilin (TGME49_014080), an actin binding protein shown to be phosphorylated by parasitic type 2C phosphatase and CKII activities [23] (Figure 1E, Table S1). Furthermore, we identified 32[P]-labeled spots representing GAP45 and MLC1 (Figure 1E, Table S1), indicating that these components of the invasion motor are also likely targets of calcium-dependent phosphorylation in vivo. A limitation in our protein identification methodology is that highly abundant proteins that co-migrate with our calcium-dependent phosphoproteins will be preferentially identified. We therefore sought to confirm the Ca2+-dependent phosphorylation of some of the interesting putative phosphoproteins that we identified. To do this we tagged candidates with a C-terminal triple HA tag. Upon radiolabelling parasites and immunoprecipitation with HA antibodies we assessed the phosphorylation status of proteins during calcium stimulation, using ethanol alone or in the presence of BAPTA-AM (Figure 1F). This confirmed that HA-tagged Toxoplasma PKA-R, PP2C, ARM1 and MLC1 are in-vivo phospho-substrates (Figure 1F). As a control we compared the signal intensity of immunoreactive bands detected on Western blots probed with anti-HA antibody. The 32[P]-labeling intensity of the corresponding HA-PKA-R PP2C-HA, ARM1-HA or MLC1-HA protein bands was significantly lower in samples treated with BAPTA-AM prior to stimulation with ethanol (Figure 1F). This is consistent with the increased number and intensity of 32[P]-labeled 2-DE spots for native PKA-R, PP2C, ARM1 and MLC1 observed by 2-DE display (Figure 1E) and validates that the phosphorylation of these proteins is dependent on intracellular Ca2+ flux. Focusing in on MLC1 and GAP45 we could see in our 2-DE autoradiographs that both motor components are basally phosphorylated and upon calcium stimulation additional phosphorylated isoforms are detectable (Figure 1E). Multiple 32P-labeled protein bands seen in immunoprecipitates from parasite lines expressing HA-tagged MLC1 are consistent with this protein co-purifying with other phosphorylated components of the native invasion motor complex (Figure 1F). Overall, our 32[P]/2DE analysis has identified a suite of validated and potential phosphoproteins that respond to calcium signaling and further, identified the Toxoplasma invasion motor as a substrate of both calcium-dependent and independent phosphorylation events. Protein phosphorylation often occurs at low stoichiometry and/or involves proteins with low expression levels. Indeed, some of the proteins identified from 32[P]/2DE spots could be more abundant co-migrating proteins that are not responsible for the signal in the autoradiographs. To overcome this issue we employed a Multi-dimensional liquid chromatography Protein Identification Technology (MudPIT) -based strategy for large-scale analysis of phosphorylation sites following in vivo Ca2+ pathway stimulation of Toxoplasma tachyzoites. Proteins were extracted from ethanol-stimulated parasites and digested with trypsin, fractionated via hydrophilic interaction chromatography (HILIC) and partitioned via titanium dioxide (TiO2) affinity chromatography. The resulting unbound or phosphopeptide-enriched (bound) fractions were subjected to MudPIT analyses on a LTQ linear ion trap mass spectrometer. Detailed analyses of TiO2-bound LC-MS/MS spectra using both Sequest and Mascot-based search algorithms resulted in a total of 305 non-redundant Toxoplasma phosphoproteins at 0. 9% false discovery rate (FDR) supported by 496 manually approved unique phosphopeptide identifications (Figure 2, Supplementary Tables S2–S4 and Text S1). This data has been integrated into ToxoDB (www. toxodb. org) and will be available in the September 2011 release. Combining 2-DE analyses of intact 32[P]-labeled proteins with global MudPIT phosphoproteomics enabled us to cross-correlate our analysis for the targeted study of parasite proteins modified by phosphorylation following in vivo stimulation of Ca2+ signaling pathways. This identified 10 major Ca2+-dependent tachyzoite phosphoproteins represented by 21 valid phosphoprotein spots and 86 phosphopeptide spectra (Figure 2A). The table in Figure 2B summarises the 2-DE and MudPIT LC-MS/MS data for this subset of the parasite phosphoproteome, which includes 17 putative Ca2+-dependent phosphorylation sites. In agreement with differential 2-DE display data (Figure 1E), the existing phosphopeptide MS evidence supports the presence of multiple phosphorylation sites on some phosphoproteins in more than one 32[P]-labeled 2-DE spot (Figure 2B and Supplementary Table S1). This includes a cluster of five residues localized on GAP45 (S153, Y158, S163, S167, S184) (Figure S2, A–E) and at least two phosphorylation sites each for PKA-R (S27, S94) (Figure S2, G–H), CKIIß (S22, S309) (Figure S2, I–J), and a hypothetical protein of unknown function TGME49_048700 (S249, S250) (Figure S2, N-O). By contrast, we only detected MS evidence for a single phosphorylation site on MLC1 (S98) (Supplementary Figure S2F), ARM1 (S33) (Supplementary Figure S2M), ANP32 (S187), SET (S103) and SNAP (S6) (Supplementary Figure S2, K–L), most of which were authenticated in multiple 32[P]-labeled 2-DE spots (Figures 1E, 2B). We failed to observe any phosphopeptide evidence for PP2C by LC-MS/MS even though our IP analyses of 32[P]-labeled parasites expressing transgenic HA-tagged suggested that this protein is a bone fide phosphoprotein (Figures 1F, 2B). Complementary 2-DE and MudPIT analyses of whole parasite extracts provided direct evidence that at least 2 of the 5 known myosin motor complex components, namely GAP45 and MLC1, are phosphorylated in vivo in a Ca2+-dependent manner. To further understand this we employed a Stable Isotope Labeling by Amino Acids in Culture (SILAC) -based proteomics approach and quantitatively assess the in vivo phosphorylation of tachyzoite invasion motor complexes following Ca2+ pathway stimulation (Figure 3A). First, we demonstrated that over the labeling period we achieved >95% incorporation of ‘heavy’ stably isotopically labeled arginine and lysine into tachyzoite protein, as detailed in Materials & Methods and Supplementary Figure S3. Then we grew SILAC-labeled bulk cultures and prepared protein lysates of un-stimulated (light) or ethanol-stimulated (heavy) tachyzoites. Samples were quantified using Sypro Ruby protein stain (Figure 3B, lanes 1&2), total protein was mixed at a ratio of 1∶1 and used to immunoaffinity-purify Toxoplasma myosin motor complexes using anti-GAP45 specific antibodies (Figure 3B, lane 3). As expected anti-GAP45 column eluates were specifically enriched in the five major components of the Toxoplasma glideosome complex – MyoA, GAP50, GAP45, GAP40 and MLC1 (Figure 3B) [5], [7]. Intact protein complexes were then digested with trypsin and partitioned using TiO2-based phosphopeptide enrichment [24]. TiO2 eluates and flow through were then run through the high mass accuracy Orbitrap instrument. (Phospho) peptides were then identified and relative abundances calculated using the MaxQuant computational platform [25] (Figure 3A). From this analysis we were able to achieve 70. 8%, 36. 9%, 52. 2%, 3. 7% and 73. 7% sequence coverage across MyoA, GAP50, GAP45, GAP40 and MLC1, respectively (Supplementary Figure S4). A total of 61 Toxoplasma phosphopeptide sequences were identified in digests of immunoaffinity-purified invasion motor complexes (Supplementary Table S5). They were exclusively singly phosphorylated and recovered in the TiO2-bound fraction consistent with a specific enrichment of mono-phosphorylated peptides by TiO2 beads [26]. Computational analyses of Orbitrap LC-MS/MS data identified up to 13 potential phosphorylation sites on the invasion motor complex proteins GAP45, MLC1 and MyoA, but none on GAP50 (Table S5). Evaluation of evidence spectra resulted in manual approval of ten unambiguous phosphorylation sites with highly confident Mascot (>25), ΔMascot (>10) and MaxQuant Phospho S/T/Y site probability scores (>90%) (Table 1 and Supplementary Figure S5, A–J). The phosphorylation of neighboring GAP45 residues S184/S185 was the only uncertainty and was narrowed to either site with 50% probability (Table 1 and Supplementary Figure S5E). The OrbiTrap SILAC data was further analyzed to quantitate the response of calcium signaling on invasion motor phosphorylation. We calculated the ratio of Heavy: Light phosphopeptides normalized for particular components using the MaxQuant algorithm (Table 1, Supplementary Table S5). This showed that phosphorylation of 4 of the identified residues on GAP45 (S153, Y158, S163, S169) did not change upon calcium stimulation, whereas S184/185 and T189 had H: L ratios of 3. 9 and 3. 7 respectively (Table 1), strongly suggesting that phosphorylation on these sites is deposited by a calcium-dependent protein kinase (the quantitation of Ca2+-dependent GAP45 phosphorylation was manually validated as shown in Supplementary Figure S6, and a detailed discussion of results is provided in Supplementary Text S2). For MLC1 we could determine that while one site (S55) did not respond to calcium signaling, T98 and S132 had mild increases in phosphorylation deposition, again suggesting a role of a calcium-dependent kinase in modulating the post-translationally modified state of this protein (the quantitation of Ca2+-dependent MLC1 phosphorylation was manually validated as shown in Supplementary Figure S7 and is discussed in Supplementary Text S3). The only MyoA phosphorylation site that we were able to identify appeared to change negligibly upon calcium stimulation and further work will be needed to identify if S21 phosphorylation truly responds to calcium signaling (Table 1). Given that we identified both serine/threonine and tyrosine phosphorylation and that some residues respond to calcium signaling, while others do not, this suggests that at least three kinases are responsible for the phosphorylation profile of the Toxoplasma invasion motor. Given that the invasion motor must form a complex to promote parasite motility [7] we wanted to see if we could use our SILAC approach to see what effect calcium signaling has on motor assembly. To do this we analyzed the abundance of non-phosphopeptides from motor complex components and in their different labeling states from anti-GAP45 column eluates. Using our LC-MS/MS analyses on the Orbitrap we identified abundant MyoA, GAP50, GAP45, GAP40 and MLC1 peptides, with a total of 186,58,54 or 43 assigned spectra from six independent experiments (including different modifications and/or labeling states) (Table 2 and Supplementary Table S6). Analysis of Heavy: Light ratios of non-phosphopeptides indicated a 2–3 fold increase in the relative abundance of heavy-labeled invasion motor complex components from calcium-stimulated parasites (Table 2 and Supplementary Table S7). By contrast, the average H/L ratio of total Toxoplasma protein in these samples was estimated to be 1. 1 ± 0. 9, as expected given that a 1∶1 mixture of H/L-labeled parasite lysates was used in our pull down experiments (Table 2). This strongly suggests that more invasion motor is available for immunoprecipitation after heavy labeled parasites underwent calcium pathway stimulation. Two possible interpretations of this result are that the invasion motor complex forms more readily upon calcium signaling or the invasion motor could potentially form higher order structures after calcium signaling. This is the first evidence to suggest that changes in the structure or assembly of the invasion motor occurs upon calcium signaling, and until a more in-depth analysis is performed these results should be treated with caution. By interrogating our SILAC data we were able to show that there are potentially changes in the assembly of the invasion motor upon calcium stimulation. We therefore reasoned that we might also be able to identify other unknown components of this complex by this virtue. The most highly represented additional protein was a hypothetical Calmodulin-like Toxoplasma protein with a predicted mass of 15 kDa (TGME49_069440) (Tables 2 and Supplementary Tables S6 and S7), that we have termed Essential Light Chain 1 (ELC1) (see below). To investigate the localization and potential association of this 15 kDa Calmodulin-like protein with the motor complex, we fused this protein with a triple HA epitope tag at the endogenous locus and performed immunofluorescence (IFA) and co-immunoprecipitation/Western blot analyses (CoIP) using anti-HA or anti-GAP45 antibodies (Figure 4). Immunofluorescence microscopy of stably expressing parasites demonstrates this protein resides both at the apical end (likely the conoid) in intracellular parasites and at the parasite periphery as demonstrated by co-localization with GAP45 (Figure 4A). Interestingly, upon egress from host cells extracellular parasites appeared to lose conoid localization and adopt a more typical pattern of an IMC protein (Figure 4A). The significance of this is not yet understood. Clostridium septicum alpha-toxin-induced swelling of extracellular parasites revealed this protein is more firmly anchored to the IMC than the plasma membrane (PM), providing further evidence that this protein could be tightly associated with the motor complex (Figure 4B). CoIP’s performed with anti-HA antibody on tagged and wildtype lines (as a control) followed by Western blot analyses formally established the association of this protein with the invasion motor complex (Figure 4C). In addition, pull-downs using antibodies against MLC1-HA (Figure 4D) or GAP45 (Figure 4E, F) precipitated the endogenous or HA-tagged ELC1. The endogenous ELC1 protein was readily detectable as a 15-kDa Sypro-Ruby stained protein band present in anti-HA IP’s in a MLC1-HA expressing line and proven by LC-MS/MS (Figure 4D and Supplementary Figure S8). The co-purification of ELC1-HA by anti-GAP45 antibodies (Figure 4E and Supplementary Tables S6 and 7) was also shown to be highly specific, as demonstrated by IP/Western blot analysis of immunoprecipitates prepared from ELC1-HA expressing parasite lines using anti-GAP45-specific or non-specific rabbit IgG (Figure 4F). Reproducibly, we noticed that this protein could be seen as two bands (Figure 4C, F) but further work will be needed to determine if this is a degredation product or has physiological significance. Toxoplasma ELC1 displays limited homology with apicomplexan calmodulins, but has significant structural similarity to the essential light chains of more conventional myosins. We found that the sequences of MLC1 and ELC1 to be highly compatible with the structure of the essential light chain of Scallop myosin II complex (PDB 2BL0, chains C and B, respectively) and therefore we constructed a model of MyoA, MLC1 and ELC1 based on this structure (Figure 4G, protein alignments used are presented in Supplementary). We therefore suggest that TGME49_069440 is the Essential Light Chain of MyoA (here named ELC1) and given its predicted calcium-binding ability this new component suggests that Ca2+ binding could also directly regulate invasion motor activity. Invasion of apicomplexan parasites is a complex, multistep process critical for the survival of this group of pathogens. In Toxoplasma calcium signaling is required for the release of apical organelles and the activation of the invasion motor [18], yet the molecular pathways mediating these processes are unclear. Pharmacological evidence suggests that upon host cell recognition by an unknown ligand, PLC is activated to produce soluble Inositol 1,4, 5 triphosphate (IP3), which then promotes the release of Ca2+ from ER stores. An increase in cytoplasmic Ca2+ concentration activates a range of calmodulin and calcium-dependent protein kinases, allowing for phosphorylation of specific targets. This then is thought to change the cellular activity of substrates, thus activating the invasion process [18]. Recent work has identified several kinases that potentially play a role in invasion [18], [19], [21], yet little is known about the substrates that they target. To address this problem we have used global proteomics approaches to identify these substrates and understand the patterns of calcium-dependent phosphorylation upon motility and invasion. We have identified over 50 potential calcium-dependent phosphorylation substrates and detected phosphorylation sites on at least 10 on these proteins. Further, by analyzing the 2-DE pattern of phosphoproteins after stimulation with either ethanol (thought to activate PLC) or the calcium-mobilizing drug Ionomycin we have shown that both these agonists stimulate the phosphorylation of a largely overlapping set of proteins, strongly suggesting that both these agents act to stimulate the same signal transduction cascade. Our analysis therefore provides further evidence that PLC acts in the same pathway as calcium signaling to activate host cell invasion [16], [18]. Using a combination of molecular and biochemical approaches we identified a wide range of Toxoplasma phosphoproteins with different predicted cellular functions and validated at least 10 calcium-dependent protein kinase substrates that are phosphorylated. Our work therefore presents a snapshot of the complexity associated with calcium-mediated signal transduction in apicomplexan parasites. It will now be interesting to understand the function of these individual molecules in mediating host cell invasion. For example, does the phosphorylation of the vesicle fusion components αSNAP and ARM1 (homologous to Vac8 from yeast) activate apical organelle release? The calcium-dependent phosphorylation of the kinase regulatory subunits CKIIß and PKA-R and the protein phosphatase PP2C are also interesting observations as they suggest modulation of the activity of these molecules by calcium signal transduction pathways. Dissecting the role of individual calcium-dependent phosphoproteins is clearly the next important step in revealing how this group of pathogens regulates host cell invasion. We have also shown that upon calcium-pathway stimulation the invasion motor is a major in vivo target of calcium-dependent phosphorylation. This strongly suggests that calcium-dependent phosphorylation events stimulate invasion motor activity. GAP45 is the major target of phosphorylation with seven identified sites. GAP45 is a multifunctional protein essential for both maintaining cohesion between the IMC and the PM and also for recruiting MyoA-MLC1 to the parasite periphery for productive movement [7]. GAP45 is tethered to the cytoplasmic face of the plasma membrane via it’s N-terminal lipid anchor and to extend across the ∼300 Å supra-alveolar space to the outer face of the inner membrane complex, where it interacts with invasion motor components via it’s C-terminal globular domain [7]. All seven GAP45 phosphorylation sites that we identified are clustered together in a semi-conserved region (aa 152-192) between the predicted coiled-coil region and the highly conserved globular domain (Figure 5A, B). Although it has been demonstrated that the coiled-coil and globular domain are both important for parasite motility [7] nothing is yet known about the role of this highly phosphorylated region. Bioinformatic analysis of the pairwise energy content of the GAP45 amino acid sequence using the IUPred algorithm [27] predicts this region to be intrinsically unstructured (Figure 5B, blue line). Furthermore, when the S/T/Y phosphorylation sites are replaced by glutamic acid (mimicking phosphorylated residues), IUPred predicts an additional ∼25% increase in the disorder tendency of this domain (Figure 5B, red line). Interestingly, clusters of phosphorylation are often located in regions predicted to be intrinsically unstructured [28] and are commonly protein-protein interaction domains [29]. This suggests that GAP45’s hyperphosphorylated region may represent a protein-protein interaction domain regulated by phosphorylation. Indeed, recent studies have implicated phosphorylation of S163 and S167, two calcium-independent phosphorylation sites within this region, in modulating an interaction of GAP45 with GAP50 [30], further supporting the notion that this region is involved in motor regulation through protein interaction. It is also apparent that only the calcium-dependent phosphorylation sites on Toxoplasma GAP45 are conserved amongst apicomplexan species (Figure 5C). Of the three Ca2+-dependent residues authenticated in the present study, S184 is replaced by a glutamic acid in sequences of other Plasmodium species, Toxoplasma GAP45 residue S185 maps to in vitro phosphorylation site S149 on PfGAP45 [31], and TgGAP45 residue T189 did not map to any identified PfCDPK1 in vitro phosphorylation site (Figure 5C). Homologous Toxoplasma residues corresponding to other PfGAP45 phosphorylation sites identified as in vitro substrates of PfCDPK1 (S156, S173) [31] were not detected in our present study, putting some doubt on the relevance of the sites identified in vitro (Figure 5C). This data suggests that essential aspects of Ca2+-dependent phosphorylation of GAP45 might be highly conserved amongst apicomplexan species (Figure 5C). In Toxoplasma MLC1 anchors MyoA to GAP45 through a long N-terminal extension (Figure 6A) [7], separate from the degenerate EF hands (Figure 6B). Alignments of MLC1/MTIP’s from other Apicomplexa suggest that this domain can be further broken down into a relatively well-conserved N-terminal portion and a C-terminal region, which, like the hyperphosphorylated segment of GAP45, is also predicted to be intrinsically unstructured (Figure 6A, Supplementary Table S8). The calcium-independent phosphorylation site S55 that we have identified is found within this disordered segment of the N-terminal extension (Figure 6A). Within this region two other phosphorylation sites have been identified in vitro in P. falciparum [21], [22], and like S55 also appear to be unconserved across the phylum. Another post-translational modification has been identified neighboring phospho-S55 and although the nature of this modification is currently unknown it has been implicated in regulating the activity of the invasion motor [32]. This region of MLC1 therefore appears to be a hotspot of posttranslational modifications, suggesting that this could also be a regulatory domain controlling the activity and/or assembly of the invasion motor. Given the role of the N-terminal segment of MLC1 in anchoring MLC1-MyoA to GAP45 [7] it is possible that S55 and other posttranslational modifications found within this region could regulate this interaction. The two calcium-dependent phosphorylation sites that we identified on MLC1 map to the C-terminal region, containing the degenerate EF hands responsible for binding the MyoA tail (Figure 6B). The two Ca2+-dependent MLC1 phosphorylation sites that were located in this work (T98, S132) do not correspond to residues critical for MyoA binding [33], nor are they conserved amongst Apicomplexa (Figure 6B). To understand a potential role of Ca2+-dependent phosphorylation we mapped these sites onto a modeled structure of MLC1-MyoA (Figure 6C). The Calmodulin-like C-terminal region of MLC1 shows significant structural homology to the regulatory domain of scallop (Physarum polycephalum) myosin, which is made up of two highly ordered degenerate EF-hand domains (N- and C-lobe), connected by a central helical ‘hinge’ region (Figure 6B, C) [34]. Interestingly, the calcium-dependent phosphorylation sites that we identified map to the N-lobe and on the surface of one particular side of MLC1’s predicted three-dimensional structure (Figure 6C). This poses the interesting possibility that this particular region could be an interaction face regulating the attachment of MLC1 with another invasion motor component. The only phosphorylation site to be identified on MyoA was at S21. We saw that this site changes mildly in abundance upon calcium signaling. This region is found in the head of the MyoA a region that binds to the actin filament. Although this residue appears conserved across Apicomplexan species (Supplementary Figure S10) we at this stage are unable to speculate on its function. Upon calcium pathway stimulation we also noticed that another protein more strongly associates with the invasion motor and we show that this Calmodulin-like protein is a bona fide component of the Toxoplasma motor complex. This Calmodulin-like Toxoplasma protein is phylogenetically distinct from both canonical calmodulins and myosin-light chains of apicomplexans [35]. Based on the presence of the intact EF hands of this component and the predictions from protein threading we suggest that this new component could represent the Essential Light Chain of MyoA. Myosin activity in muscle tissue is regulated by one of two mechanisms, either by the calcium-dependent phosphorylation of the regulatory light chain (RLC) or by the direct binding of calcium ions to an EF hand within ELC. Both RLC and ELC bind to the lever arm of myosin and either mode of activation is thought to rigidify the myosin motor and allow it to function more efficiently [36]. We found that the sequences of MLC1 and ELC1 to be highly compatible with the structure of the essential light chain of scallop myosin II complex (PDB 2BL0, chains C and B, respectively) and therefore we constructed a model of MyoA, MLC1 and ELC1 based on this structure (Figure 4G and Supplementary Figure S9). The model of Toxoplasma ELC1 includes a calcium-binding site juxtaposed near the region of closest approach between the ELC1 (residues T16-D17) and MLC1 (residues Y177-G178-E179) – homologues to G178 that is highly conserved amongst RLC’s and other calmodulins. Therefore the model predicts that an interaction between Toxoplasma MLC1 and ELC1 is also possible (Figure 4G). This opens the way to investigating the theory that invasion motor activity may be regulated by direct binding of calcium ions as well as calcium-dependent phosphorylation. What kinases are responsible for invasion motor phosphorylation? We hypothesized that enzymes engaged in these events may be associated with the invasion motor complex. We therefore interrogated our proteomic data set to identify possible interacting kinases, but unfortunately we were unable to detect any significant hits which we attribute to be likely due to the transitory nature of kinases with their substrates. Appraisal of the literature and the motif signature of calcium-dependent phosphorylation sites suggest that members of the CPDK family are the most likely candidates (data not shown). Given that PfCDPK1 has been implicated in motor phosphorylation in P. falciparum the most likely candidate in Toxoplasma is the phylogenetically related Toxoplasma kinase CDPK3 [18]. TgCDPK3, like PfCDPK1, has a predicted motif for N-terminal acylation, which could potentially anchor it to the PM, putting this enzyme in the vicinity to modulate invasion motor phosphorylation in this parasite. There are several possible kinases that could modulate the calcium-independent phosphorylation status of the motor; Protein Kinase B (PKB) has been shown to phosphorylate PfGAP45 in vitro, whereas Toxoplasma Protein Kinase G (PKG) is found at the periphery and a specific inhibitor to this enzyme prevents parasite motility and invasion [37]. Another candidate is Casein Kinase II (CKII). The beta regulatory subunit of CKII in Toxoplasma also localizes to the periphery of parasite [38] and two calcium-independent phosphorylation sites on GAP45 conform to a CKII-like substrate motif (data not shown). A tyrosine-based kinase also likely modulates the phosphorylation status of the invasion motor. Tyrosine 158 on GAP45 is phosphorylated in a calcium-independent manner. Given that tyrosine kinases form a unique class of enzyme this suggests a third pathway regulates that invasion motor phosphorylation. Tyrosine kinase and tyrosine phosphorylation have not yet been reported for any apicomplexan parasite, and therefore, it is not possible to speculate on what the identity of this kinase might be. Overall, our data suggests that there are at least three different kinases, and therefore, at least three potential pathways that modulate the phosphorylation status of the invasion motor. Applying specific conditional mutants to our quantitative proteomics approach has the potential to reveal the in vivo kinase for these important phosphorylation events. We have also shown that upon induction of calcium signaling more invasion motor components are found associated with GAP45. This suggests that although the invasion motor can be found pre-formed in resting parasites, calcium signaling induces changes in the association between GAP45 and the other components. Mechanistically such a change could be one of the following: 1. Inducing more invasion motor complexes to form from individual components, 2. The invasion motor complex is rendered more stable after calcium signaling or 3. A higher order structure, such as multimerization is induced. To distinguish from these possibilities and to confirm our finding is physiologically relevant biochemical analyses of isolated invasion motor complexes will need to be performed. Predicted open reading frames of phosphoproteins (as described at www. toxodb. org) were amplified and inserted into Toxoplasma transfection vectors. Toxoplasma MLC1, ARM1 and PP2C were amplified with primers 1&2,3&4,5&6 (Table 3) respectively and inserted into BglII/AflII sites of pCT3H (Tonkin, unpublished), which enabled the tagging of these genes with at the C-terminus with a 3xHA tag. PKA-R was amplified with primers 7&8 and a HA tag introduced by adding this sequence ending this tag onto the 5’ primer (Table 3). PKA-R PCR product was then inserted behind the fkbp-derived destabilization domain (DD) at AvrII/PstI of pCTDD (Tonkin, unpublished). The invasion motor associated Calmodulin-like protein (TGME49_069440) was tagged at the 3’ end of the endogenous gene using ΔKu80 parasite line [39]. A 3’ flank of this gene was amplified with primers 9 and 10 and inserted into pLIC-HA3/HX [40] using the ligation independent cloning strategy. Plasmids were linearized within the gene flank for efficient homologous integration after transfection [39]. Toxoplasma parasites were grown using standard procedures. Briefly, tachyzoites were grown in confluent human foreskin fibroblasts (HFF) or Vero cells maintained in Dulbecco’s Modified Eagle’s Medium (DMEM; GIBCO, Invitrogen) supplemented with 10% cosmic calf serum (CCS) and an additional 2 mM glutamine (Gibco). If HFF were used DMEM/10%CCS was replaced with DMEM supplemented with 1% fetal calf serum (DMEM/1%FCS) with additional 2 mM glutamine (Gibco) upon infection with parasites. Toxoplasma was transfected using standard procedures [41], [42]. Electroporated tachyzoites were inoculated onto confluent HFF cells and selected on 6 ug/ml of chloramphenicol [41] or Mycophenolic Acid (25 µg/ml) /Xanthine (50 µg/ml) for stable transfectants [43]. IFA was performed using standard procedures. Briefly, Parasites, either intracellular or extracellular, were fixed in 4% paraformaldehyde, permeabilized in 0. 1% Triton-X/PBS and blocked in 3% BSA/PBS. Anti-HA mAb (clone 3F10; Roche), anti-GAP45 [5] and anti-SAG1 (a kind gift from L. D Sibley) were decorated onto blocked slides and primary antibodies were detected with antibodies conjugated to AlexaFluor-594 and 488 (Molecular Probes). Extracellular parasites were stuck to slides using 0. 1% polyethylenimine. Parasites were treated with dialyzed C. septicum culture supernatant containing alpha-Toxin for ∼3 hours. Parasites were then imaged on a Zeiss Inverted Axioscope equipped with AxioCam MRM and Axiovision with Deconvolution software. For radioactive or heavy isotope labeling and Ca2+ pathway stimulation studies, specialized cell culture conditions were applied. For radioactive labeling, Vero cells were infected with T. gondii tachyzoites and metabolically labeled in 25 ml of custom sodium phosphate-free DMEM (GIBCO, Invitrogen) supplemented with 10% fetal calf serum, 2 mM glutamine and 20 µCi/ml of 32[P]-labeled monosodium phosphate, overnight. For stable isotope labeling studies, HFF were infected with Toxoplasma and grown in custom DMEM medium devoid of lysine, arginine and leucine. This media was then supplemented with 1% fetal calf serum (dialyzed against sterile PBS), 2 mM glutamine, 0. 802 mM L-leucine, and either 0. 398 mM L-Arginine + 0. 798 mM L-Lysine for the „light“ condition (R0/K0 media), or 0. 398 mM (15N4) L-Arginine + 0. 798 mM (13C6,15N2) L-Lysine (98% isotopic purity; Sigma-Isotec) for the “heavy” condition (R4/K8 media). Heavy label incorporation was applied to parasites for ∼6 days to get maximal stable isotope incorporation before bulking up and subsequent parasite harvest. Analysis demonstrated >95% incorporation of R4 or K8 labels after this time (Supplementary Figure S3). The preparation of free, viable tachyzoites was done as follows; Upon the >80% disruption of HFF cells, parasites were scraped, needle passed and pushed through a 5 µM pore-size membrane filters (Millipore) and then repeatedly washed in ice cold Invasion Medium (DMEM/HEPES/1%FCS) and counted using a hemocytometer. Aliquots of radioactive or heavy-isotope-labeled tachyzoites were resuspended in invasion medium and incubated at 37°C for 10 min prior to Ca2+ pathway stimulation. In vivo Ca2+ pathway stimulation was achieved by adding 1% (171 mM) ethanol or 1 µM ionomycin. As a control, samples of parasites were either mock treated (equal volume DMSO) or were pretreated for 10 minutes with the membrane-permeable calcium chelator BAPTA-AM, followed by stimulation with 1% ethanol. Samples were stimulated for 60 seconds, immediately mixed with 0. 5 ml ice cold 2 × lysis buffer (40 mM HEPES pH 7. 4,300 mM NaCl, 2% Triton X-100,1% NP-40) containing protease and phosphatase inhibitors (HALT protease and phosphatase inhibitor solution, PIERCE) and snap frozen in liquid nitrogen. Parasites were disrupted by freeze thawing and ultrasonication in ice cold lysis buffer (20 mM HEPES pH 7. 4,150 mM NaCl, 1% Triton X-100,0. 5% NP-40) including protease and phosphatase inhibitors (Pierce), and proteins extracted for 30 min on ice with vortexing. Parasite extracts were clarified by ultracentrifugation at 75,000 g for 30 min and the detergent-soluble protein fraction was precipitated using 2-D Clean-Up kit (GE-healthcare). For 2-DE gel electrophoresis, precipitated Toxoplasma protein preparations (∼50 µg protein for analytical gels/∼500 µg protein for preparative gels) were redissolved in 300 µl rehydration/sample buffer (7 M Urea, 2 M Thiourea, 2% ASB-14,1% DTT, 1% ampholytes), loaded onto 13 cm pI 4–7 IPG strips by passive rehydration and focused at a current limit of 50 µA/IPG strip using a fast voltage gradient (8000 V max, 24,000 Vh) at 15°C. The second dimension was carried out on 10% polyacrylamide gels (18 cm×16 cm×1. 5 mm) using a Hoefer SE 600 system (GE Healthcare) at 75 V constant voltage and 10°C overnight. Analytical 2-DE gels were electrophoretically transferred to Immobilon-PSQ PVDF membranes (Millipore). Protein spots on PVDF membranes were visualized using Deep Purple protein stain (GE Healthcare), and protein spots in preparative 2-DE gels were stained with Sypro Ruby Protein stain (Molecular Probes), according to manufacturer’s protocols. Imaging of 32[P]-labeled protein spots was achieved by direct autoradiography (7 day exposure) of dry PVDF membrane blots using FUJIFILM BAS-TR2040 tritium imaging plates. Fluorescent or autoradiographic 2-DE images were digitized on a FLA-3000 laser-scanning detection system (Fuji) and manually matched and annotated using Image Master Platinum v7. 0 2D image analysis software (GE Healthcare). Preparative 2-DE gels were counter-stained using Colloidal Coomassie Brilliant Blue (Sigma) and regions matching 32[P]-labeled protein spots manually excised and subjected to LC-MS/MS analysis. For 1-DE or 2-DE samples, protein spots or bands were manually excised from preparative SDS-PAGE gels and subjected to automated in-gel reduction, alkylation, and tryptic digestion using a MassPREP Station (Micromass, UK). All gel samples were reduced with 10 mM DTT (SIGMA) for 30 min, alkylated for 30 min with 50 mM iodoacetic acid (SIGMA) and digested with 375 ng trypsin (Promega) for 16 hrs at 37°C. The extracted peptide solutions were then acidified (0. 1% formic acid) and concentrated to approximately 10 ul by centrifugal lyophilisation using a SpeedVac AES 1010 (Savant). Briefly, extracted peptides were injected and fractionated by nanoflow reversed-phase liquid chromatography on a nano LC system (1200 series, Agilent) using a nanoAcquity C18 150 mm×0. 15 mm I. D. column (Waters) developed with a linear 60-min gradient with a flow rate of 0. 5 µl/min at 45°C from 100% solvent A (0. 1% Formic acid in Milli-Q water) to 100% solvent B (0. 1% Formic acid, 60% acetonitrile, 40% Milli-Q water). The nano HPLC was coupled on-line to an LTQ-Orbitrap mass spectrometer equipped with a nanoelectrospray ion source (Thermo Fisher Scientific) for automated MS/MS. Up to five most intense ions per cycle were fragmented and analysed in the linear trap, with target ions already selected for MS/MS being dynamically excluded for 3 min. For proteome-wide phosphopeptide analyses using MudPIT, precipitated Toxoplasma protein preparations (2. 5 mg of protein of ionomycin-stimulated tachyzoites) were dissolved in a total volume of 1. 25 ml of 6 M Urea, 100 mM Tris-HCl (pH 8. 0) and sequentially reduced with 5 mM DTT and alkylated with 10 mM iodoacetamide. Samples were diluted 1∶5 (v/v) in digestion buffer (40 mM ammonium bicarbonate, pH 8,10% acetonitrile, 1 mM Ca2Cl) and digested with LysC (Roche) and trypsin (Sigma). Approximately 1 mg tryptic digest was desalted using Sep-Pak C18 6 cc/1 g cartridges (Waters) and fractionated by hydrophilic interaction chromatography on a TSKgel Amide 80 column (TOSOH Biosciences) using an optimized phosphopeptide gradient, as per published protocol [44]. Ten phosphopeptide fractions were collected and lyophilized. Fractions 1-4 were pooled and the resulting 7 samples were resuspended in 100 µl TiO2 loading/wash buffer (1 M glycolic acid, 80% ACN, 5% TFA) and purified on self-packed TiO2 micro-columns using a highly efficient method for the selective enrichment of phosphorylated peptides, as detailed elsewhere [45]. The flow-through and all washings were combined and constituted the phosphopeptide-depleted (TiO2-unbound) fraction. Individual TiO2-bound phosphopeptide fractions were analyzed by multi-dimensional LC-MS/MS on an LTQ linear ion trap mass spectrometer (Thermo Scientific), according to published protocols [46]. MudPIT LC-MS/MS spectra were analyzed with SEQUEST 2. 7 [47] using a non-redundant protein decoy database (Ludwig NR_Q309_con reverse; 9870917 entries for all species). The SEQUEST outputs were analyzed by DTASelect 2. 0. 37 [48]. DTASelect 2. 0 uses a quadratic discriminant analysis to dynamically set XCorr and DeltaCN thresholds for the entire data set to achieve a user-specified false positive rate (1% in this analysis). The false positive rates were estimated by the program from the number and quality of spectral matches to the decoy database [49]. After filtering the results from SEQUEST using DTASelect, MS/MS spectra were analysed using the DeBunker algorithm for automatic validation of phosphopeptide identifications from tandem mass spectra [50]. This software package uses a support vector machine binary classifier to assess the correctness of phosphopeptide/spectrum matches. For protein identification of gel samples and the identification of additional MudPIT phosphopeptide sequences (ie. not included in the LudwigNR database) LC-MS/MS data were searched against a redundant protein decoy database comprising sequences from the latest version of Swiss-Prot (Human, Bovine, Plasmodium, Toxoplasma species), Trembl (Toxoplasma entries), PlasmoDB/ToxoDB (Toxoplasma entries), as well as their reverse sequences (Toxoplasma_decoy; 117496 entries). Mass spectra peak lists were extracted using extract-msn as part of Bioworks 3. 3. 1 (Thermo Fisher Scientific) linked into Mascot Daemon (Matrix Science, UK). The parameters used to generate the peak lists for the LTQ Orbitrap were as follows: minimum mass 400; maximum mass 5000; grouping tolerance 0. 01 Da; intermediate scans 1; minimum group count 1; 10 peaks minimum and total ion current of 100. Peak lists for each nano-LC-MS/MS run were used to search MASCOT v2. 2. 04 search algorithm (Matrix Science, UK) provided by the Australian Proteomics Computational Facility (www. apcf. edu. au). The search parameters consisted of carboxymethylation of cysteine as a fixed modification (+58 Da, for gel samples only), with variable modifications set for NH2-terminal acetylation (+42 Da) and oxidation of methionine (+16 Da), phosphorylation of serine, threonine or tyrosine (+80 Da). A precursor mass tolerance of ±3 Da (LTQ spectra) or 20 ppm (Orbitrap spectra), #13C defined as 1, fragment ion mass tolerance of ±0. 8 Da, and an allowance for up to three missed cleavages for tryptic searches was used. To generate immunoaffinity resin anti-GAP45 IgG was purified using a 1 ml HiTrap Protein-A HP column (GE Healthcare) using an ÄKTA prime FPLC chromatography system (Pharmacia). Purified IgG was desalted using a PD-10 buffer exchange column and chemically cross-linked to CN-Br-activated Sepharose 4B resin (GE Healthcare) in 0. 1 M NaHCO3,0. 5 M NaCl pH 8. 5 for 2 h at room temperature. Un-reacted sites were blocked with 1 M triethanolamine pH 8. 0 for 2 h at room temperature and the resin stringently washed with 50 mM glycine-HCl, 0. 5 M NaCl pH 3. 5 and 50 mM Tris-HCl, 0. 5 M NaCl pH 8. 0 and resuspended in PBS. Large-scale immunoaffinity purification of intact MyoA motor complexes from Toxoplasma parasites was carried out using anti-GAP45 rabbit serum, as previously described [5]. For our SILAC-based quantitative LC-MS/MS analyses, Toxoplasma tachyzoites were grown in confluent HFF’s in “light” R0/K0 media or “heavy” R4/K8 media for 48 h. We found that 4×T150 cm2 flasks each of light- or heavy-labeled Toxoplasma cultures yielded sufficient material to obtain >60% sequence coverage for most of the complex components. Free, viable tachyzoites were prepared and the heavy isotope-labeled parasites were stimulated with 1% ethanol to measure the effect of Ca2+-pathways stimulation, as detailed above. Freshly prepared parasite lysates of un-stimulated (light) or 1% ethanol-stimulated (heavy) tachyzoites were prepared as detailed above, mixed at a protein ratio of 1∶1 in 2 ml lysis buffer with protease and phosphatase inhibitors and incubated with 0. 25 ml of the anti-GAP45 resin for 1 h at 4°C. Unbound protein was removed and the resin washed 5 times with 1 ml lysis buffer and 2 times with 0. 5 ml water by centrifugation at 2000 g for 5 min in 1 ml microcentrifuge spin columns (Pierce). Invasion motor complexes were eluted with 200 µl 0. 1% TFA in water and the eluted material dried in a SpeedVac. Dried samples of anti-GAP45 column eluates were dissolved in a total volume of 250 µl of 6 M Urea, 50 mM ammonium bicarbonate (pH 8. 0) and sequentially reduced with 10 mM DTT and alkylated with 55 mM iodoacetamide. Samples were diluted 1∶5 (v/v) in digestion buffer (40 mM ammonium bicarbonate, pH 8,10% acetonitrile, 1 mM Ca2Cl) and digested with proteomics-grade trypsin (Sigma) at 37°C overnight. The resulting peptide solutions were vacuum-dried, peptides resolved in 0. 2%TFA in water, desalted on MacroSpin C18 columns (The Nest Group Inc.), eluted with 0. 2%TFA/60% ACN in water and again dried in a SpeedVac. Desalted peptide samples were resuspended in 100 µl TiO2 loading buffer (1 M glycolic acid in 80% ACN, 5% TFA) and affinity purified using self-packed TiO2 microcolumns containing ∼10 µg of 5 µm Titansphere TiO2 beads (GL Sciences), using a published method [24]. The flow-through and all washings were combined and constituted the phosphopeptide-depleted (TiO2-unbound) fraction. The TiO2 eluate comprising the phosphopeptide enriched fraction and the phosphopeptide-depleted fraction were vacuum-dried. Both fractions were redissolved in 0. 2%TFA in water, desalted on MacroSpin C18 columns (The Nest Group Inc.) and again dried in a SpeedVac. For LC-MS/MS analyses, dried (phospho) peptide fractions were redissolved in 1 µl 100% formic acid and diluted to 20 µl in MilliQ water. Digested (phospho) peptides were then subjected to nano-LC-MS/MS on an LTQ-Orbitrap instrument (Thermo Scientific), as described above. LTQ-Orbitrap LC-MS/MS data were searched against the Toxoplasma_decoy database using the MASCOT search engine, as detailed above. Search parameters were identical, except that additional variable modifications were set for heavy-isotope labeling of arginine (+4 Da) or lysine (+8 Da) and the precursor mass tolerance was 20 ppm for the SILAC/Orbitrap datasets. Mascot search results were loaded into MaxQuant (v1. 1. 08) [51] for peptide and protein quantification as well as Scaffold (v3. 0) (www. proteomesoftware. com) for manual validation of peptide spectral matches. Manual validation of all peptide spectral matches was done irrespective of peptide scores or expectation values (E-values) to ensure that all major fragment ions were annotated in accordance with known rules of peptide fragmentation. The Xcalibur program (Thermo Scientific) was used for generating XIC' s of selected peptides in order to validate and verify the MaxQuant results. For anti-HA CoIP’s, freshly released tachyzoites expressing wild-type (negative control) or HA-tagged parasite proteins were harvested as above, washed in PBS, and lysed in ice cold lysis buffer (20 mM HEPES pH 7. 4,150 mM NaCl, 1% Triton X-100,0. 5% NP-40) containing protease and phosphatase inhibitors (Pierce). Protein was extracted for 30 min on ice with vortexing and centrifuged at 50,000 g for 30 min at 4°C. Supernatants were mixed with monoclonal anti-HA antibody coupled to paramagnetic MicroBeads (Miltenyi Biotec) and incubated for 1 h at 4°C. Supernatants were then subjected to CoIP, according to a manufacturer’s protocol for the specific isolation of HA-tagged proteins utilizing a MultiMACS separator (Miltenyi Biotech). For anti-GAP45 CoIP’s, 1 ml protein extracted from parasites expressing ELC1-HA were mixed with 100 µl of ProteinG-coupled paramagnetic microbeads (Miltenyi Biotech), mixed with 5 µl of anti-GAP45-specific rabbit IgG or non-specific rabbit IgG (negative control), and processed as above. For Western blot analyses eluted protein samples were separated by SDS-PAGE on precast 10% or 4-12% NuPAGE Bis-Tris gels, transferred onto Nylon membranes using an iBlot blotting system (Invitrogen), and then probed with specific primary or HRP-conjugated secondary antibody as indicated. Protein fold recognition was conducted using the WURST protein-threading web server [52]. Homology models were constructed using the sequence alignments predicted from WURST with the MODELLER (9v7) comparative modeling software [53]. PKA-R, TGME49_042070; CKIIß, TGME49_072400; SNAP, TGME49_018760; ARM1, TGME49_061440; PP2C, TGME49_054770; Rab5B, TGME49_007460; ANP32, TGME49_071810; SET, TGME49_044110; Toxofilin, TGME49_014080; Hypothetical protein, TGME49_048700; Conserved hypothetical protein, TGME49_032440; CaM-like (ELC1), TGME49_069440; MLC1, TGME49_057680; GAP45, TGME49_023940; MyoA, TGME49_035470; GAP50, TGME49_019320; GAP40, TGME49_049850.

Apicomplexan parasites are a group of obligate intracellular pathogens of wide medical and agricultural significance. Included within this phylum is Plasmodium spp, the causative agents to malaria and the ubiquitous parasite Toxoplasma, which inflicts disease burden on AIDS patients, transplant recipients and the unborn fetus. No matter the host cell that they target, all apicomplexan parasites must activate invasion upon host cell contact. Calcium-mediated signal transduction pathways modulate this process, yet the molecular processes are largely unknown. Using a range of proteomics approaches we reveal proteins in Toxoplasma that are phosphorylated upon calcium signaling, and furthermore, identify phosphorylation sites on a range of proteins that may play crucial roles in regulating parasite motility and microneme secretion. By quantitatively monitoring phosphorylation deposition upon calcium signaling we define putative regulatory domains of GAP45 and MLC1 and further show evidence that the invasion motor potentially more strongly associates upon calcium signaling. We also identified that a new Calmodulin-like protein is part of the invasion motor and this suggests that direct Ca2+ binding may also modulate motor activity.

Abstract Introduction Results Discussion Materials and Methods

biochemistry signal transduction molecular cell biology protein interactions biology microbiology proteomics molecular biology parasitology pathogenesis

2011

Quantitative in vivo Analyses Reveal Calcium-dependent Phosphorylation Sites and Identifies a Novel Component of the Toxoplasma Invasion Motor Complex

PIK3C2A is a class II member of the phosphoinositide 3-kinase (PI3K) family that catalyzes the phosphorylation of phosphatidylinositol (PI) into PI (3) P and the phosphorylation of PI (4) P into PI (3,4) P2. At the cellular level, PIK3C2A is critical for the formation of cilia and for receptor mediated endocytosis, among other biological functions. We identified homozygous loss-of-function mutations in PIK3C2A in children from three independent consanguineous families with short stature, coarse facial features, cataracts with secondary glaucoma, multiple skeletal abnormalities, neurological manifestations, among other findings. Cellular studies of patient-derived fibroblasts found that they lacked PIK3C2A protein, had impaired cilia formation and function, and demonstrated reduced proliferative capacity. Collectively, the genetic and molecular data implicate mutations in PIK3C2A in a new Mendelian disorder of PI metabolism, thereby shedding light on the critical role of a class II PI3K in growth, vision, skeletal formation and neurological development. In particular, the considerable phenotypic overlap, yet distinct features, between this syndrome and Lowe’s syndrome, which is caused by mutations in the PI-5-phosphatase OCRL, highlight the key role of PI metabolizing enzymes in specific developmental processes and demonstrate the unique non-redundant functions of each enzyme. This discovery expands what is known about disorders of PI metabolism and helps unravel the role of PIK3C2A and class II PI3Ks in health and disease. Identifying the genetic basis of diseases with Mendelian inheritance provides insight into gene function, susceptibility to disease, and can guide the development of new therapeutics. To date, ~50% of the genes underlying Mendelian phenotypes have yet to be discovered [1]. The disease genes that have been identified thus far have led to a better understanding of the pathophysiological pathways and to the development of medicinal products approved for the clinical treatment of such rare disorders [2]. Furthermore, technological advances in DNA sequencing have facilitated the identification of novel genetic mutations that result in rare Mendelian disorders [3,4]. We have applied these next-generation sequencing technologies to discover mutations in PIK3C2A that cause a newly identified genetic syndrome consisting of dysmorphic features, short stature, cataracts and skeletal abnormalities. PIK3C2A is a class II member of the phosphoinositide 3-kinase (PI3K) family of lipid kinases that catalyzes the phosphorylation of phosphatidylinositol (PI) [5]. PI3Ks are part of a larger regulatory network of kinases and phosphatases that act upon the hydroxyl groups on the inositol ring of PI to add or remove a phosphate group [6]. The combinatorial nature of phosphorylation at the -3, -4, and -5 position of the inositol ring gives rise to seven different PI species, termed polyphosphoinositides. Among these polyphosphoinositides, class II PI3Ks are generally thought to catalyze the phosphorylation of PI and/or PI (4) P to generate PI (3) P and PI (3,4) P2, respectively [7]. PI (3) P, PI (3,4) P2, and the other polyphosphoinositides each account for less than ~1% of the total phospholipid content of a cell [8]. However, despite their relatively low abundance, they play central roles in a broad array of signaling pathways and are central to the pathophysiology underlying cancer, metabolic disease, and host-pathogen interactions [6]. The functions of class II PI3Ks are poorly understood relative to many other kinases and phosphatases that regulate PI metabolism, in part because there was no causal link between any class II PI3K and a monogenic human disease. In contrast, a number of disorders of PI metabolism have previously been described that have provided invaluable insight into the physiological functions of specific PI metabolizing enzymes [9]. These include Charcot-Marie-Tooth type 4J (FIG4) [10,11], Centronuclear X-linked myopathy (MTM1) [12], and primary immunodeficiency (PIK3CD) [13,14], among others. As just one example, detailed studies of FIG4 subsequent to its identification as a cause of Charcot-Marie-Tooth type 4J have revealed both genetic and physiological interactions with VAC14 and PIKFYVE, which together generate PI (3,5) P2 and are required for melanosome homeostasis, oligodendrocyte differentiation, and remyelination [15–18]. Collectively, the array of PI metabolism disorders is striking for its phenotypic diversity, affecting a wide range of organ systems including those described above as well as others that lead to neuromuscular, skeletal, renal, eye, growth, and immune disorders. The diversity of phenotypic manifestations resulting from PI metabolism defects highlights the lack of functional redundancy between genes that regulate nominally the same enzymatic transformation of PIs. PIK3C2A has previously been attributed a wide-range of biological functions including glucose transport, angiogenesis, Akt activation, endosomal trafficking, phagosome maturation, mitotic spindle organization, exocytosis, and autophagy [19–28]. In addition, PIK3C2A is critical for the formation and function of primary cilia [23,26]. However, as mentioned above, there is as yet no link between PIK3C2A or any class II PI3K and a Mendelian disorder. Here, we describe the evidence that homozygous loss-of-function mutations in PIK3C2A cause a novel syndromic disorder involving neurological, visual, skeletal, growth, and occasionally hearing impairments. Five individuals between the ages of 8 and 21 from three unrelated consanguineous families were found by diagnostic analyses to have a similar constellation of clinical features including dysmorphic facial features, short stature, skeletal and neurological abnormalities, and cataracts (Fig 1, Table 1, S1 Table). The dysmorphic facial features included coarse facies, low hairline, epicanthal folds, flat and broad nasal bridges, and retrognathia (S1 Table). Skeletal findings included scoliosis, delayed bone age, diminished ossification of femoral heads, cervical lordosis, shortened fifth digits with mild metaphyseal dysplasia and clinodactyly, as well as dental findings such as broad maxilla incisors, narrow mandible teeth, and enamel defects (Fig 1B and 1C, S1 Table, S1 Fig). Most of the affected individuals exhibited neurological involvement including developmental delay and stroke. This was first seen in individual I-II-2 when she recently started having seizures, with an EEG demonstrating sharp waves in the central areas of the right hemisphere and short sporadic generalized epileptic seizures. Her brain MRI showed a previous stroke in the right corpus striatum (Fig 1E). Hematological studies were normal for hypercoagulability and platelet function (S2 Table). In addition, brain MRI of patient II-II-3 showed multiple small frontal and periventricular lacunar infarcts (S1E Fig). Unclear episodes of syncope also led to neurological investigations including EEG in individual III-II-2, without any signs of epilepsy. Her brain MRI showed symmetrical structures and normal cerebrospinal fluid spaces but pronounced lesions of the white matter (S1E Fig). Other recurrent features included hearing loss, secondary glaucoma, and nephrocalcinosis. In addition to the shared syndromic features described above in all three families, both affected daughters in Family I were diagnosed with congenital adrenal hyperplasia (CAH), due to 17-alpha-hydroxylase deficiency, and were found to have a homozygous familial mutation: NM_000102. 3: c. 286C>T; p. (Arg96Trp) in the CYP17A1 gene (OMIM #202110) [29,30]. The affected individuals in Families II and III do not carry mutations in CYP17A1 or have CAH, suggesting the presence of two independent and unrelated conditions in Family I. The co-occurrence of multiple monogenic disorders is not uncommon among this highly consanguineous population [31]. To identify the genetic basis of this disorder, enzymatic assays related to the mucopolysaccharidosis subtypes MPS I, MPS IVA, MPS IVB, and MPSVI were tested in Families I and II and found to be normal. Enzymatic assays for mucolipidosis II/III were also normal and no pathogenic mutations were found in galactosamine-6-sulfate sulfatase (GALNS) in Family I. Additionally, since some of the features of patient II-II-3 were reminiscent of Noonan syndrome, Hennekam syndrome, and Aarskog-Scott syndrome, individual genes involved in these disorders were analyzed in Family II, but no pathogenic mutation was identified. In patient III-II-2, Williams-Beuren syndrome was excluded in childhood. Additionally, direct molecular testing at presentation in adulthood excluded Leri-Weill syndrome, Alstrom disease, and mutations in FGFR3. Given the negative results of targeted genetic testing, WES and CNV analysis was performed for the affected individuals from all three families. Five homozygous candidate variants were identified in Family I, including the CYP17A1 (p. Arg96Trp) mutation that is the cause of the CAH [29,30], but is not known to cause the other phenotypes. The remaining four variants affected the genes ATF4, DNAH14, PLEKHA7, and PIK3C2A (S3 Table). In Family II, homozygous missense variants were identified in KIAA1549L, METAP1, and PEX2, in addition to a homozygous deletion in PIK3C2A that encompassed exons 1–24 out of 32 total exons (S3 Table). The deletion was limited to PIK3C2A and did not affect the neighboring genes. Sequence analysis of Family III showed a homozygous missense variant in PTH2R, nonsense variant in DPRX, and splice site variant in PIK3C2A (S3 Table). Sequencing analyses revealed that all affected family members in the Families I, II, and III were homozygous for predicted loss-of-function variants in PIK3C2A, and none of the unaffected family members were homozygous for the PIK3C2A variants (Fig 1A and 1G). The initial link between these three families with rare mutations in PIK3C2A was made possible through the sharing of information via the GeneMatcher website [3]. The PIK3C2A deletion in Family II was confirmed by multiplex amplicon quantification (S2A Fig). The single nucleotide PIK3C2A variants in Families I and III were confirmed by Sanger sequencing (S2B and S2C Fig). In Family I, the nonsense mutation in PIK3C2A (p. Tyr195*) truncates 1,492 amino acids from a protein that is 1,686 amino acids. This is predicted to eliminate nearly all functional domains including the catalytic kinase domain, and is expected to trigger nonsense-mediated mRNA decay [25]. Accordingly, levels of PIK3C2A mRNA are significantly decreased in both heterozygous and homozygous individuals carrying the p. Tyr195* variant (Fig 2A). The deletion in Family II eliminates the first 24 exons of the 32-exon PIK3C2A gene and is thus predicted to cause a loss of protein expression. This is consistent with a lack of PIK3C2A mRNA expression (Fig 2B). The variant in PIK3C2A in Family III affects an essential splice site (c. 1640+1G>T) that leads to decreased mRNA levels (Fig 2C). Deep sequencing of the RT-PCR products revealed 4 alternative transcripts in patient-derived lymphocytes (p. [Asn483_Arg547delinsLys, Ala521Thrfs*4, Ala521_Glu568del, and Arg547SerinsTyrIleIle*]) of which the transcript encoding p. Asn483_Arg547delinsLys that skips both exons 5 and 6 was also observed in patient’s fibroblasts (S3 Fig). Although this transcript remains in-frame, no PIK3C2A protein was detected by Western blotting (Fig 2D and 2F). This is consistent with Families I and II, for which Western blotting also failed to detect any full-length PIK3C2A in fibroblasts from the affected homozygous children (Fig 2E and 2F). Thus, all three PIK3C2A variants likely encode loss-of-function alleles. Importantly, among the 141,456 WES and whole genome sequences from control individuals in the Genome Aggregation Database (gnomAD v2. 1) [32], none are homozygous for loss-of-function mutations in PIK3C2A, which is consistent with total PIK3C2A deficiency causing severe early onset disease. To test whether the observed loss-of-function mutations in PIK3C2A cause cellular phenotypes consistent with loss of PIK3C2A function, we examined PI metabolism, cilia formation and function, and cellular proliferation rates. PIK3C2A deficiency in the patient-derived fibroblasts decreased the levels of PI (3,4) P2 throughout the cell (Fig 3A) as well as decreased the levels of PI (3) P at the ciliary base (Figs 3B and S4A). The reduction in PI (3) P at the ciliary base was associated with a reduction in ciliary length (Fig 4A), although the percentage of ciliated cells was not altered (Fig 4B). Additional cilia defects include a reduction in the levels of RAB11 at the ciliary base (Figs 4C and S4B), which functions within a GTPase cascade culminating in the activation of RAB8, which together with ARL13B selectively traffics ciliary proteins to the cilium [33]. Additionally, there was increased accumulation of IFT88 along the length of the cilium (Figs 4D and S4C), which is a component of the intraflagellar transport sub-complex IFT-B, and is essential for the trafficking of ciliary protein cargoes along the axonemal microtubules [34,35]. Together, these findings are suggestive of defective trafficking of ciliary components. Finally, the proliferative capacity of PIK3C2A deficient cells was reduced relative to control cells (Fig 5). As PIK3C2A is a member of the class II PI3K family, we tested whether the expression of the other family members PIK3C2B and PIK3C2G were altered by PIK3C2A deficiency. The expression of PIK3C2G was not detected by qRT-PCR in either patient-derived or control primary fibroblasts. This is consistent with the relatively restricted expression pattern of this gene in the GTEx portal [36], with expression largely limited to stomach, skin, liver, esophagus, mammary tissue, and kidney, but absent in fibroblast cells and most other tissues. In contrast, PIK3C2B expression was detected, with both mRNA and protein levels significantly increased in PIK3C2A deficient cells (Fig 6A–6D). Downregulation of PIK3C2A using an inducible shRNA in HeLa cells also resulted in elevated levels of PIK3C2B (Fig 6E). Together, these data are consistent with increased levels of PIK3C2B serving to partially compensate for PIK3C2A deficiency. Here we describe the identification of three independent families with homozygous loss-of-function mutations in PIK3C2A resulting in a novel syndrome consisting of short stature, cataracts, secondary glaucoma, and skeletal abnormalities among other features. Patient-derived fibroblasts had decreased levels of PI (3,4) P2 and PI (3) P, shortening of the cilia and impaired ciliary protein localization, and reduced proliferation capacity. Thus, based on the loss-of-function mutations in PIK3C2A, the phenotypic overlap between the three independent families, and the patient-derived cellular data consistent with previous studies of PIK3C2A function, we conclude that loss-of-function mutations in PIK3C2A cause this novel syndrome. The identification of PIK3C2A loss-of-function mutations in humans represents the first mutations identified in any class II PI-3-kinase in a disorder with a Mendelian inheritance, and thus sheds light into the biological role of this poorly understood class of PI3Ks [7,37]. This is significant not only for understanding the role of PIK3C2A in rare monogenic disorders, but also the potential contribution of common variants in PIK3C2A in more genetically complex disorders. There are now numerous examples where severe mutations in a gene cause a rare Mendelian disorder, whereas more common variants in the same gene, with a less deleterious effect on protein function, are associated with polygenic human traits and disorders [38–40]. For example, severe mutations in PPARG cause monogenic lipodystrophy, whereas less severe variants are associated with complex polygenic forms of lipodystrophy [41,42]. In the case of PIK3C2A deficiency, the identification of various neurological features including developmental delay, selective mutism, and the brain abnormalities detected by MRI (S1 Table) may provide biological insight into the mechanisms underlying the association between common variants in PIK3C2A and schizophrenia [43–45]. Other monogenic disorders of phosphoinositide metabolism include Lowe’s syndrome and Joubert syndrome, which can be caused by mutations in the inositol polyphosphate 5-phosphatases OCRL and INPP5E, respectively [46]. All three of these disorders of PI metabolism affect some of the same organ systems, namely the brain, eye, and kidney. However, the phenotype associated with mutations in INPP5E is quite distinct, and includes cerebellar vermis hypo-dysplasia, coloboma, hypotonia, ataxia, and neonatal breathing dysregulation [47]. In contrast, the phenotypes associated with Lowe’s syndrome share many of the same features with PIK3C2A deficiency including congenital cataracts, secondary glaucoma, kidney defects, skeletal abnormalities, developmental delay, and short stature [9,48]. The enzyme defective in Lowe’s syndrome, OCRL, is functionally similar to PIK3C2A as well, as it is also required for membrane trafficking and ciliogenesis [49]. The similarities between Lowe’s syndrome and PIK3C2A deficiency suggest that similar defects in phosphatidylinositol metabolism may underlie both disorders. In addition to Lowe’s syndrome, there is partial overlap between PIK3C2A deficiency and yet other Mendelian disorders of PI metabolism such as the early-onset cataracts in patients with INPP5K deficiency [50,51], demonstrating the importance of PI metabolism in lens development. The viability of humans with PIK3C2A deficiency is in stark contrast to mouse Pik3c2a knockout models that result in growth retardation by e8. 5 and embryonic lethality between e10. 5–11. 5 due to vascular defects [20]. One potential explanation for this discrepancy is functional differences between human PIK3C2A and the mouse ortholog. However, the involvement of both human and mouse PIK3C2A in cilia formation, PI metabolism, and cellular proliferation suggests a high degree of functional conservation at the cellular level [26,28]. An alternate possibility is that the species viability differences associated with PIK3C2A deficiency result from altered compensation from other PI metabolizing enzymes. For instance, there are species-specific differences between humans and mice in the transcription and splicing of the OCRL homolog INPP5B that may uniquely contribute to PI metabolism in each species [52]. Alternately, PIK3C2B levels were significantly increased in human PIK3C2A deficient cells, including both patient-derived cells and HeLa cells surviving PIK3C2A deletion, suggesting that this may partially compensate for the lack of PIK3C2A in humans, although it remains to be determined whether a similar compensatory pathway exists in mice. It is intriguing that both PIK3C2A and OCRL have important roles in primary cilia formation [26,53,54]. Primary cilia are evolutionary conserved microtubule-derived cellular organelles that protrude from the surface of most mammalian cell types. Primary cilia formation is initiated by a cascade of processes involving the targeted trafficking and docking of Golgi-derived vesicles near the mother centriole. They play a pivotal role in a number of processes, such as left-right patterning during embryonic development, cell growth, and differentiation. Abnormal phosphatidylinositol metabolism results in ciliary dysfunction [55], including loss of PIK3C2A that impairs ciliogenesis in mouse embryonic fibroblasts, likely due to defective trafficking of ciliary components [26]. The importance of primary cilia in embryonic development and tissue homeostasis has become evident over the two past decades, as a number of proteins which localize to the cilium harbor defects causing syndromic diseases, collectively known as ciliopathies [56,57]. Hallmark features of ciliopathies share many features with PIK3C2A deficiency and include skeletal abnormalities, progressive vision and hearing loss, mild to severe intellectual disabilities, polydactyly, and kidney phenotypes. Many of these disorders, including Bardet-Biedl Syndrome, Meckel Syndrome, and Joubert Syndrome are also associated with decreased cilium length [58], as seen in PIK3C2A deficient cells. Ciliary length is a function of both axoneme elongation and cilium disassembly, and is molecularly regulated by intraflagellar protein transport, including the velocity of transport and cargo loading, as well as soluble tubulin levels and microtubule modifications [59,60]. As defects in intraflagellar protein transport were likely indicated by abnormal IFT88 localization along the length of the cilium in PIK3C2A deficient cells, this may represent a potential mechanism underlying the shortened cilium. Further work and the identification of additional patients with mutations in PIK3C2A will continue to improve our understanding of the genotype-phenotype correlation associated with PIK3C2A deficiency. However, the identification of the first patients with PIK3C2A deficiency establishes a role for PIK3C2A in neurological and skeletal development, as well as vision, and growth and implicates loss-of-function PIK3C2A mutations as a potentially new cause of a cilia-associated disease. The study was approved by the Helsinki Ethics Committees of Rambam Health Care Campus (#0038-14-RMB), the University Hospital Institutional Review Board for Case Western Reserve University (#NHR-15-39), the Ethics Committee of the Friedrich-Alexander University Erlangen-Nürnberg (#164_15 B), and was in accordance with the regulations of the University Medical Center Groningen’s ethical committee. Written informed consent was obtained from all participants. Whole exome sequencing (WES) of two patients from Family I was performed using DNA (1μg) extracted from whole blood and fragmented and enriched using the Truseq DNA PCR Free kit (Illumina). Samples were sequenced on a HiSeq2500 (Illumina) with 2x100bp read length and analyzed as described [61]. Raw fastq files were mapped to the reference human genome GRCh37 using BWA [62] (v. 0. 7. 12). Duplicate reads were removed by Picard (v. 1. 119) and local realignment and base quality score recalibration was performed following the GATK pipeline [63] (v. 3. 3). The average read depth was 98x (I-II-1) and 117x (I-II-2). HaplotypeCaller was used to call SNPs and indels and variants were further annotated with Annovar [64]. Databases used in Annovar were RefSeq [65], Exome Aggregation Consortium (ExAC) [32] (v. exac03), ClinVar [66] (v. clinvar_20150330) and LJB database [67] (v. ljb26_all). Exome variants in Family I were filtered out if they were not homozygous in both affected individuals, had a population allele frequency greater than 0. 1% in either the ExAC database [32] or the Greater Middle East Variome Project [68], and were not predicted to be deleterious by either SIFT [69] or Polyphen2 [70]. Whole exome sequencing was performed on the two affected individuals of Family II and both their parents essentially as previously described [71]. Target regions were enriched using the Agilent SureSelectXT Human All Exon 50Mb Kit. Whole-exome sequencing was performed on the Illumina HiSeq platform (BGI Europe) followed by data processing with BWA [62] (read alignment) and GATK [63] (variant calling) software packages. Variants were annotated using an in-house developed pipeline. Prioritization of variants was done by an in-house designed ‘variant interface’ and manual curation. The DNAs of Family III were enriched using the SureSelect Human All Exon Kit v6 (Agilent) and sequenced on an Illumina HiSeq 2500 (Illumina). Alignment, variant calling, and annotation were performed as described [72]. The average read depth was 95x (III-II-2), 119x (III-I-1) and 113x (III-I-2). Variants were selected that were covered by at least 10% of the average coverage of each exome and for which at least 5 novel alleles were detected from 2 or more callers. All modes of inheritance were analyzed [72]. Variants were prioritized based on a population frequency of 10−3 or below (based on the ExAC database [32] and an in-house variant database), on the evolutionary conservation, and on the mutation severity prediction. All candidate variants in Families I, II, and III were confirmed by Sanger sequencing (primers listed in S4 Table). Microarray analysis for CNV detection in Family I was performed using a HumanOmni5-Quad chip (Illumina). SNP array raw data was mapped to the reference human genome GRCh37 and analyzed using GenomeStudio (v. 2011/1). Signal intensity files with Log R ratio and B-allele frequency were further analyzed with PennCNV [73] (v. 2014/5/7). In Family III the diagnostic chromosomal microarray analysis was performed with an Affymetrix CytoScan HD-Array and analyzed using Affymetrix Chromosome analysis Suite-Software, compared with the Database of Genomic Variants and 820 in house controls. All findings refer to UCSC Genome Browser on Human, February 2009 Assembly (hg19), Human Genome built 37. CNV analysis on the WES data of Families II and III were performed using CoNIFER [74]. Variants were annotated using an in-house developed pipeline. Prioritization of variants was done by an in-house designed ‘variant interface’ and manual curation as described before [75]. Subsequent segregation analysis of the pathogenic CNV in Family II was performed with MAQ by using a targeted primer set with primers in exons 3,10,20 and 24 which are located within the deletion and exons 28,32,34 which are located outside of the deletion (Multiplex Amplicon Quantification (MAQ); Multiplicom). Human dermal fibroblasts were obtained from sterile skin punches cultured in DMEM (Dulbecco' s Modified Eagle' s Medium) supplemented with 10–20% Fetal Calf Serum, 1% Sodium Pyruvate and 1% Penicillin and streptomycin (P/S) in 5% CO2 at 37°C. Control fibroblasts were obtained from healthy age-matched volunteers. Fibroblasts from passages 4–8 were used for the experiments. To measure cell proliferation, cells were detached using trypsin and counted with an Automated Cell Counter (ThermoFisher). Cells (n = 2500) were plated in triplicate in 96-well plates. Viability was measured at day 2,4, 6 and 8. Each measurement was normalized to day 0 (measured the day after plating) and expressed as a fold increase. Viability was assessed by using CellTiter-Glo Luminescent Cell Viability Assay (Promega). Three independent experiments were performed. HeLa cells were infected with lentiviral particles containing pLKO-TET-PI3KC2A-shRNA or pLKO-TET-scramble-shRNA in six-well plates (n = 50,000 cells). After two days, the medium containing lentiviral particles was replaced with DMEM 10% FBS, 1. 5μg/ml puromycin. After 7 days of selection, cells were detached and 100,000 cells were plated in six-well plates in triplicate in the presence of doxycycline (0. 5,1 and 2 μg/ml). Medium containing doxycycline was replaced every 48 hours. After 10 days of doxycycline treatment, cells were lysed and analysed by Western blot. Total RNA was purified from primary fibroblasts using the PureLink RNA purification kit (ThermoFisher) or RNAPure peqGOLD (Peqlab). RNA was reverse transcribed into complementary DNA with random hexamer using a high-Capacity cDNA Reverse Transcription Kit (ThermoFisher). RT-PCR from lymphocytes to detect exon-skipping in family III was performed using primers flanking exon 6. The resulting product was sequenced on an Illumina HiSeq2500 (Illumina) to detect splicing variants with high sensitivity. Gene expression was quantified by SYBR Green real-time PCR using the CFX Connect Real-Time System (BioRad). Primers used are detailed in S4 Table. Expression levels were calculated using the ΔΔCT method relative to GADPH. Protein was extracted from cultured primary fibroblast cells as described [76,77]. Extracts were quantified using the DC protein assay (BioRad) or the BCA method. Equal amounts of protein were separated by SDS-PAGE and electrotransferred onto polyvinylidene difluoride membranes (Millipore). Membranes were blocked with TBST/5% fat-free dried milk and stained with antibodies as detailed in S5 Table. Secondary antibodies were goat anti-rabbit (1: 5,000, ThermoFisher #31460) goat anti-mouse (1: 5,000, ThermoFisher #31430), goat anti-rabbit (1: 2,000, Dako #P0448), and goat anti-mouse (1: 2,000, Dako #P0447). Primary fibroblasts were grown on glass coverslips to approximately 80% - 90% confluency in DMEM + 10% FCS + 1% P/S, at which time the medium was replaced with DMEM without FCS for 48 hours to induce ciliogenesis. Cells were fixed in either methanol for 10 minutes at -20°C or 4% paraformaldehyde for 10 minutes at room temperature (RT). Fixed cells were washed in PBS, and incubated with 10% normal goat serum, 1% bovine serum albumin in PBS for 1 hour at RT. If cells were fixed with paraformaldehyde, blocking solutions contained 0. 5% Triton X-100. Cells were incubated with primary antibody overnight at 4°C, washed in PBS, and incubated with secondary antibody including 4' , 6-diamidino-2-Phenylindole (DAPI) to stain nuclei for 1 hours at RT. Coverslips were mounted on glass slides with fluoromount (Science Services) and imaged on a confocal laser scanning system with a 63x objectives (LSM 710, Carl Zeiss MicroImaging). Primary antibodies are detailed in S5 Table. To induce ciliogenesis, cells were grown in DMEM with 0–0. 2% FCS for 48 hours. Cells were washed in PBS, then fixed and permeabilized in ice-cold methanol for 5 minutes, followed by extensive washing with PBS. After blocking in 5% Bovine Serum Albumin, cells were incubated with primary antibodies for 1. 5 hours at RT and extensively washed in PBS-T. Primary antibodies used for Centrin and ARL13B are detailed in S5 Table. To wash off the primary antibody, cells were extensively washed in PBS-T. Subsequently, cells were incubated with secondary antibodies, Alexa Flour 488 (1: 800, Invitrogen) and Alexa Fluor 568 (1: 800, Invitrogen), for 45 min followed by washing with PBS-T. Finally, cells were shortly rinsed in ddH2O and samples were mounted using Vectashield with DAPI. Images were taken using an Axio Imager Z2 microscope with an Apotome (Zeiss) at 63x magnification. Cilia were measured manually using Fiji software taking the whole length of the cilium based on ARL13B staining. At least 300 cilia were measured per sample. Cilia lengths were pooled for 3 control cell lines and compare to 2 patient-derived samples (II-II-2 and II-II-3). Statistical significance was calculated using a Student t-test. PI (3) P at the ciliary base was detected in randomly chosen cells using the same exposure for each acquisition. A specific anti-PI (3) P antibody (Echelo Z-P003) was used to quantify the PI (3) P by measuring the green fluorescent intensity around the ciliary base in a region with a diameter of 8 μm and a depth of 10 μm as previously illustrated and described [26].

Identifying the genetic basis of rare disorders can provide insight into gene function, susceptibility to disease, guide the development of new therapeutics, improve opportunities for genetic counseling, and help clinicians evaluate and potentially treat complicated clinical presentations. However, it is estimated that the genetic basis of approximately one-half of all rare genetic disorders remains unknown. We describe one such rare disorder based on genetic and clinical evaluations of individuals from 3 unrelated consanguineous families with a similar constellation of features including short stature, coarse facial features, cataracts with secondary glaucoma, multiple skeletal abnormalities, neurological manifestations including stroke, among other findings. We discovered that these features were due to deficiency of the PIK3C2A enzyme. PIK3C2A is a class II member of the phosphoinositide 3-kinase (PI3K) family that catalyzes the phosphorylation of the lipids phosphatidylinositol (PI) into PI (3) P and the phosphorylation of PI (4) P into PI (3,4) P2 that are essential for a variety of cellular processes including cilia formation and vesicle trafficking. This syndrome is the first monogenic disorder caused by mutations in a class II PI3K family member and thus sheds new light on their role in human development.

Abstract Introduction Results Discussion Materials and methods

medicine and health sciences diagnostic radiology lens disorders enzymology fibroblasts magnetic resonance imaging connective tissue cells enzyme metabolism cellular structures and organelles enzyme chemistry research and analysis methods imaging techniques animal cells connective tissue biological tissue metabolic disorders clinical genetics biochemistry radiology and imaging diagnostic medicine cell biology anatomy cilia cataracts gene identification and analysis ophthalmology genetics mutation detection biology and life sciences cellular types

2019

Mutations in PIK3C2A cause syndromic short stature, skeletal abnormalities, and cataracts associated with ciliary dysfunction

Altered DNA methylation patterns in CD4+ T-cells indicate the importance of epigenetic mechanisms in inflammatory diseases. However, the identification of these alterations is complicated by the heterogeneity of most inflammatory diseases. Seasonal allergic rhinitis (SAR) is an optimal disease model for the study of DNA methylation because of its well-defined phenotype and etiology. We generated genome-wide DNA methylation (Npatients = 8, Ncontrols = 8) and gene expression (Npatients = 9, Ncontrols = 10) profiles of CD4+ T-cells from SAR patients and healthy controls using Illumina' s HumanMethylation450 and HT-12 microarrays, respectively. DNA methylation profiles clearly and robustly distinguished SAR patients from controls, during and outside the pollen season. In agreement with previously published studies, gene expression profiles of the same samples failed to separate patients and controls. Separation by methylation (Npatients = 12, Ncontrols = 12), but not by gene expression (Npatients = 21, Ncontrols = 21) was also observed in an in vitro model system in which purified PBMCs from patients and healthy controls were challenged with allergen. We observed changes in the proportions of memory T-cell populations between patients (Npatients = 35) and controls (Ncontrols = 12), which could explain the observed difference in DNA methylation. Our data highlight the potential of epigenomics in the stratification of immune disease and represents the first successful molecular classification of SAR using CD4+ T cells. The modest effects of genetic variants in inflammatory diseases indicate the importance of epigenetic mechanisms like DNA methylation to disease pathology. However, studies of inflammatory diseases have shown conflicting results. In monozygotic twins discordant for multiple sclerosis (MS), no significant differences in DNA methylation profile were found [1]. A more recent study of monozygotic twins discordant for psoriasis identified widespread differences between siblings [2]. Other studies of autoimmune diseases have reported varying findings [3]. Discordant monozygotic twin studies benefit from a constant genetic background on which to identify disease-associated epigenetic changes. However, intrinsically, such studies tend to involve small samples sizes, and may thus lack the power to detect small, rare, disease-associated changes in DNA methylation. The variation between studies may be due to disease heterogeneity, variations in disease course and the confounding effects of treatment, and as the disease causing agent is unknown, it is difficult to experimentally model disease pathogenesis. By contrast, seasonal allergic rhinitis (SAR) occurs at defined time points each year and the disease causing agent, pollen, is known. These unique features of SAR permit analysis of CD4+ T-cells from SAR patients during and after the pollen season in vivo, and by allergen-challenge in vitro [4]. An epigenetic component in SAR is supported by its increasing prevalence in the developing world, failure of genome-wide association studies to identify a consistent genetic component to the disease, and frequent discordance for SAR between monozygotic twins [5], [6]. DNA methylation changes at numerous loci are required for appropriate differentiation of naïve CD4+ T-cells into CD4+ T effector cell subtypes [7]. We generated genome-wide expression (Npatients = 9, Ncontrols = 10) and DNA methylation (Npatients = 8, Ncontrols = 8) profiles for CD4+ T-cells from untreated SAR patients and healthy controls, both during and outside the pollen season. Consistent with previous studies, we found that CD4+ T-cell gene expression profiles were poor classifiers of SAR. However, we observed clear and robust separation of patients and controls by DNA methylation signature, both during and outside the pollen season. Separation by methylation (Npatients = 12, Ncontrols = 12), but not by gene expression (Npatients = 21, Ncontrols = 21) was also observed in an in vitro model system in which purified PBMCs were challenged with allergen in culture. Moreover, we found that these methylation profiles were significantly associated with disease severity in patients during season, and may be due to differing proportions of central memory CD4+ T-cells (Npatients = 35, Ncontrols = 12). This, to our knowledge, is the first successful molecular classification of SAR using CD4+ T-cells, and highlights the potential of genome-wide epigenetic technologies in the stratification of immune disease. Seasonal allergic rhinitis (SAR) is a powerful disease model because, (1) SAR' s clear clinical manifestations make it easy to assess disease severity, (2) the affected cell type, CD4+ T-cells, can be obtained from patients when they are symptom-free (outside the pollen season) and compared to the same cell type in the same individual when symptomatic (during the pollen season). We aimed to test the ability of CD4+ T-cell DNA methylation to separate SAR patients from healthy controls. The cohorts used in the study are outlined in Supplemental Figure S1. In a previous study, we obtained PBMCs from adult SAR patients (N = 21; age = 25. 4 years±7. 8 SD) and healthy controls (N = 21; age = 25. 7 years±10. 1 SD) outside the pollen season, and challenged these cells with either grass pollen extract or diluent (PBS) (Figure 1A). Seven days post-challenge, total CD4+ T-cells were isolated by magnetic-activated cell sorting (MACS, positive-selection), and the mRNA expression profile determined by gene expression microarray (GEO: GSE50223). Consistent with several previous studies, we were unable to separate patients and controls after challenge with allergen by CD4+ T-cell mRNA expression profile [6], [8], [9] (Figure 1B). Here, PBMCs from a new cohort of SAR patients (N = 12; age = 28. 3 years±12. 1 SD) and healthy controls (N = 12; age = 27. 3 years±10. 7 SD) were allergen-challenged, after which purified CD4+ T-cell DNA was analysed using Illumina HumanMethylation27 DNA methylation microarrays. Unsupervised hierarchical clustering of samples by genome-wide DNA methylation profiles resulted in two groups, ‘H’ (healthy controls) and ‘P’ (patients), clearly separating samples by disease-state (Figure 1C, left panel). Consensus clustering, whereby the data are repeatedly re-sampled and re-clustered, found that clusters ‘H’ and ‘P’ were reproducibly and stably identified (P<0. 05). We confirmed this separation in the data using principal component analysis (PCA), which also revealed a clear separation between allergen-challenged CD4+ T-cells from patients versus controls (Figure 1C, right panel). A leave-one out (LOO) cross-validation found that methylation data accurately classified all samples as patient or healthy control (χ2; P<0. 0001). DNA methylation and gene expression signatures did not separate patients and healthy controls after diluent challenge (Figure S2). Though striking, the results obtained from in vitro allergen-challenge of PBMCs may be confounded by cell culture effects. To verify the observations made after in vitro allergen-challenge, the mRNA expression and DNA methylation profiles of in vivo CD4+ T-cells were determined in a new cohort of SAR patients and healthy controls (GEO: GSE50387). In this experiment we used Illumina HumanMethylation450k methylation arrays, which quantitatively assess 450,000 CpG sites across the genome, with over half the probes targeting CpGs outside gene promoters and a quarter of all probes targeting CpGs in non-genic regions. CD4+ T-cells were purified from fresh blood collected from SAR patients and healthy controls by negative magnetic cell sorting both during and outside the pollen season in the same calendar year. Symptom scores for patients were recorded at the time of collection (Table S1). DNA and total RNA were harvested simultaneously from each sample. Subsequently, cDNA and bisulfite converted DNA was applied to Illumina HT12 expression microarrays (Npatients = 9, Ncontrols = 10; both during and outside the pollen season) and HumanMethylation450k microarrays microarrays (Npatients = 8, Ncontrols = 8; both during and outside the pollen season), respectively. Unsupervised hierarchical clustering of samples by mRNA expression profile did not result in separation of samples by disease state either during or outside the pollen season (Figure 2A, left panel). No effect of array batch or sex was evident in the observed clusters (Figure 2A, left panel). These findings are in line with those reported here (Figure 1B) and in numerous previous studies of CD4+ T-cells in allergy and other autoimmune diseases [6], [8], [9]. PCA of the gene expression data also failed to distinguish between SAR patients and healthy controls (Figure 2A, right panel). Differentially expressed genes were identified as those showing differential expression between patients and controls (Mann-Whitney U-test; P<0. 05, unadjusted) both during and outside the pollen season (357 genes identified). The relative changes observed across all genes were small (mean absolute fold change = 0. 39, SEM ± 0. 01). Gene Ontology (GO) term analysis failed to identify any significantly enriched gene annotation clusters after correction for multiple correction. However, four of the top 10 annotations related to lymphocyte activation (Figure S3A). All four clusters consisted of mixtures of the same seven genes, namely; HSPD1, TGFBR2, CD134 (TNFRSF4, OX40), CD45 (PTPRC), SPN (CD43), IL2RA, IL13RA1. This group of genes is highly relevant; CD43−/− mice show a pronounced Th2 phenotype and exhibit increased in inflammation in two allergic mouse models [10] and association studies have identified both IL2RA and TGFBR2 as susceptibility loci for allergy and asthma [11]. CD143 is a regulator of CD4+ T-cell memory and CD45 isoforms are key markers of memory CD4+ T-cells, though it is not possible to determine the exact isoform mis-expressed due to the positioning of the gene expression probe in the 5′ UTR of CD45 [12]. To determine the relationship between the observed changes in gene expression and DNA methylation we determined the average promoter methylation for all genes represented on the 450K methylation array. Changes in gene expression and associated promoter DNA methylation between patients and controls were not significantly negatively correlated (Pearson' s r = 0. 14, P = 0. 07), as would be expected if DNA methylation were acting to silence transcription (Figure S3B). Thus, although it was not possible to separate patients from controls by gene expression profile, a small set-of disease-relevant miss-expressed genes were identified. In contrast, unsupervised hierarchical clustering of all probes by DNA methylation profile resulted in clear separation of the samples by disease state, regardless of season (during or outside pollen season); seven of eight healthy controls grouped in cluster ‘H’, and six of eight patient samples grouped in cluster ‘P’ (Figure 2B, left panel). Consensus clustering identified clusters ‘H’ and ‘P’ with high statistical confidence (Multiscale Bootstrap Resampling; P<0. 05). Interestingly, the paired measurements (during and outside season) of each sample also clustered neatly together; again implicating disease status as the major modifier of DNA methylation profile between patients and controls samples, while also highlighting the reproducibility and accuracy of the data (paired samples were collected and processed at least five months apart) (Figure 2, left panel). PCA resulted in even clearer separation of samples along the main principle component, with only one healthy sample clustering among the patient samples (Figure 2, right panel). LOO analysis correctly classified samples as patient or healthy control in all samples collected outside season (χ2; P<0. 0001), and all but two healthy samples collected during season (χ2; P = 0. 01). PCA1 explained approximately 14% of the variation in the data and 3-times the variation explained by PCA2 (5%). Although our findings indicated that disease status was the major mediator of DNA methylation differences between patients and healthy controls an effect of pollen season was also evident. With the exception of one outlier, healthy samples clustered tightly together along PCA1, whereas patient samples showed greater spread along PCA1 (Figure 2B, right panel). Given this observation, we tested if the position along PCA1 of patient samples was associated with patient symptom score during season. Symptom scores for each patient are listed in Supplemental Table S1. Significantly, we found that symptom score explained 74% of the variation in patient sample variation along PCA1, a strong and highly significant correlation (Spearman' s rho = 0. 86, P = 0. 011) (Figure 2C). To our knowledge, such a strong association between individual or genome-scale markers and symptoms has not been previously been described in allergy, nor in other inflammatory diseases. However, given the small sample size (Npatients = 8) of the study presented here, the use of DNA methylation as a marker of disease severity needs to be tested in a much larger cohort. The association with disease severity may also explain the difficulty in identifying an epigenetic component to other immune-related diseases in which the disease course is more variable and complex to assess. However, as patient symptoms can vary dramatically during the pollen season it is important that the observed correlation between DNA methylation and symptom severity is tested at several time-points during a pollen season to validate the robustness of the preliminary observation reported here. If our findings are applicable to other inflammatory diseases, an important implication is that DNA methylation may help to stratify such diseases. The observed differences in DNA methylation between patients and controls were small (mean absolute change = 1. 2%±2. 3 SD), bi-directional, and genome-wide, 12,000 probes (3. 5% of all probes) were found to have changed significantly (Mann-Whitney U-test; P<0. 01; unadjusted) (Figure S4A). Indeed, the 1,000 most significantly altered probes changed by only ±10% (Figure S4B). Given the small size of the observed methylation differences and the known technical variation associated with 450K methylation arrays [13], we selected five CpG loci for validation by pyrosequencing in 4 SAR patients and 4 healthy controls. We selected three CpGs that showed significant methylation differences between patients and healthy controls (among top 50 altered probes) which were also located in annotated gene promoters (PIEZO1 promoter CpG, RPP21 promoter CpG, HLA-DMA promoter CpG) and 2 control CpG loci (unmethylated CpG; GAPDH promoter & methylated CpG; CD74 promoter). Pyrosequencing primers are listed in Table S2. Array methylation was highly significantly correlated with pyrosequencing methylation for all CpG loci (Spearman' s rho = 0. 98, P<10−6) (Figure S4C), and the absolute difference in methylation measurements between the array and pyrosequencing across all CpGs was small (median difference = 2. 41%±3. 2 MAD; mean difference = 4. 1%±4 SD). The methylated (Figure S4D) and unmethylated (Figure S4E) control CpGs were also validated by pyrosequencing, although agreement between array and pyrosequencing was slightly better for the unmethylated CpG site. Critically, pyrosequencing confirmed the direction and scale of DNA methylation change observed between patients and controls by 450K methylation array at the three disease-associated test loci (Figure S4F–S4H). Our findings agree with a recent quantitative study of genome-wide methylation in CD4+ T-cells from monozygotic twins discordant for the autoimmune disease, psoriasis, in which affected and unaffected siblings were distinguished by numerous, small, bi-directional changes in DNA methylation, with no one CpG exhibiting a significant change in DNA methylation level [2]. Moreover, and very recent study of DNA methylation in CD4+ T-cells from patients with the autoimmune disease Systemic Lupus Erythematosus (SLE) also reported widespread small changes in DNA methylation [14]. Interestingly, we found that the genic location of significantly altered probes was enriched for gene-bodies and non-genic regions (χ2; P<0. 0001) (Figure S5A). This highlights the importance of using unbiased genomic technologies; as assays targeted towards detection of large changes in DNA methylation in promoter regions may miss subtle but informative changes in other genomic compartments. To further dissect the genomic compartmentalization of the observed changes in DNA methylation we focused on regulatory elements whose function is known to be modified by DNA methylation, namely promoters, DNaseI hypersensitive sites (DHS) and enhancer elements. Interestingly, differentially methylated probes appeared enriched in annotated enhancers compared to those located in DHSs and promoters (χ2; P<10−5) (Figure S5B). This enrichment was observed both during and outside the pollen season (Figure S5B). Enhancer probes differentially methylated during and outside the pollen season showed a significant overlap (Fisher' s exact test; P<0. 001) (Figure S5C). Significantly, those enhancer probes differentially methylated both during and outside the pollen season (N = 960) showed a clear tendency towards loss of methylation in patients in contrast to the bi-directional changes observed for all differentially methylated probes (Figure S5D). As the activity of many enhancer elements and consequently their associated genes, is affected by DNA methylation, this finding may reflect the epigenetic re-modeling of enhancer elements in CD4+ T-cells. However, as most enhancers are only active in a small number of tissues we cross-referenced the observed differentially methylated enhancers with a recently published experimentally determined set of Th1-, Th2- and Th0-specific enhancers [15], [16]. Only 1 of the differentially methylated enhancer probes (N = 960) mapped to the Th1/Th2/Th0 enhancers. Thus, is seems unlikely that the observed changes in enhancer methylation have function gene expression consequences in Th2 cells, the key pathogenic cell-type in allergy. However, as Th2 cells constitute <3% of the total CD4+ T-cell population, we cannot exclude large functionally relevant changes in DNA methylation at enhancers in other CD4+ T-cell subsets. Differences in DNA methylation are often assumed to reflect changes in DNA methylation in a given type of cell; however such differences can also result from changes in the proportions of cell types in samples. Indeed, a recent study reported that a significant proportion of the DNA methylation differences observed between breast tumours (N = 248) reflected the number of infiltrating T-cells, and was significantly associated with prognosis [17]. Thus, the observed, small, genome-wide and bi-directional changes in DNA methylation between SAR patients and controls may reflect changes in the proportions of CD4+ T-cell sub-populations between patients and controls. Given that DNA methylation profile clearly separates samples by disease-state both during and outside the pollen season we hypothesized that the observed differences may have been due to differences in memory CD4+ T-cell subsets. Indeed, we have previously observed a reduction in CD4+ central memory T-cells (TCM) in a small cohort of allergic patients [18]. Here we extended this study, using FACS analysis to quantify CD4+ T-cell subsets in an additional 26 subjects (Npatients = 35, Ncontrols = 12) (Figure S6). SAR patients were significantly depleted for CD4+ central memory T-cells (TCM) (11% of total CD4+ T-cells) relative to healthy individuals (18. 7% of total CD4+ T cells), a reduction of 41% in the TCM population in patients relative to controls (Mann-Whitney; P = 0. 001) (Figure 3). The values for TCM cells in healthy individuals reported here are similar to those reported in independent studies [19], affirming the accuracy of our methodology. Altered TCM proportions are being implicated in a growing number of immune-related diseases [19]–[21]. This finding was also consistent with enrichment for the Gene Ontology (GO) term, ‘negative regulation of memory T-cell differentiation’, found when genes containing the 100 most significantly changed probes were analyzed (Figure S5E, right panel). However, given the enrichment for altered probes in non-genic and non-promoter regions of genes, the results of such GO term analyses must be viewed with caution. As the quantification of CD4+ T-cell subsets and DNA methylation profiling were performed on different cohorts of patients and controls, conclusions drawn from these independent observations must be cautious. Determination of DNA methylation profile, CD4+ T-cell subset structure and symptom severity in a large cohort of patients and controls at multiple time-points during and outside the pollen season would allow a more robust analysis of the inter-dependence of these variables. Although our results suggest that differences in T-cell sub-populations may contribute to the observed differences in DNA methylation between SAR patients and controls, they do not exclude a direct role for DNA methylation changes within CD4+ T-cell subsets. Analysis of gene expression and DNA methylation in each memory CD4+ T-cell subset in a large cohort of patients and healthy controls is required to directly test the association between CD4+ T-cell subset changes and observed changes in total CD4+ T-cell methylation. It is technically challenging to obtain sufficient numbers of each subtype of CD4+ T-cell from standard (40 mL) blood samples. An alternate future strategy may be to determine reference genomes for each CD4+ T-cell subtype and use these subtype specific methylomes to estimate sub-type proportions from a total CD4+ T-cell methylome [22]. This approach has recently been successfully applied to the study of rheumatoid arthritis [23]. The ability to separate patients and healthy controls by quantitative DNA methylation array, but not by gene expression array is interesting. Assuming little or no change in DNA methylation or gene expression profiles within CD4+ T-cell subtypes, small changes in the proportions of subtypes should result in very subtle changes in both the DNA methylation and gene expression profiles of total CD4+ T-cells. These subtle changes can be detected by the DNA arrays, given 1) their sensitivity – they are quantitative and accurate, and 2) their power – they assess 450,000 CpG sites. Gene expression arrays are not quantitative and have a restricted dynamic range of values which is often determined by probe design, not gene transcript levels. Moreover, whereas gene expression arrays typically have probes targeting ∼40,000 transcripts, only a small proportion of these are actually expressed (informative) in any given cell type, further reducing the power of the assay. Moreover, DNA methylation profiles are very stable in normal somatic cells, showing less inter-individual variation than RNA levels. Thus, whereas detecting small changes (<5%) in CD4+ T-cell substructure by gene expression microarray is theoretically possible, it would require a very large sample size; much greater than that used in this study (Npatients = 9, Ncontrols = 10; both during and outside the pollen season). Our results support the use of both DNA methylation arrays and gene expression arrays simultaneously, particularly, where cell-type composition may contribute to the molecular signature of the disease. Epigenetic regulation plays a key role in Th differentiation. Pronounced changes in DNA methylation patterns are observed at several key loci during helper T cell differentiation [24]. Up-regulation of Il4, IL5, and IL13 gene expression in Th2 cells is accompanied by a pronounced loss of DNA methylation and gain of permissive histone marks across the locus [25], [26]. Similarly, the promoter of the Th1 gene, Ifng, is unmethylated in Th1 cells, but hypermethylated in Th2 cells, reinforcing helper T-cell identity. DNA demethylation also occurs at regulatory regions of the FOXP3 gene in Treg cells [27], and at the IL17A promoter in Th17 cells [28]. Allergy involves an inappropriate Th2 response to a benign allergen such as pollen [29], and several observations point to a key role for epigenetics in the pathogenesis of SAR. Murine studies have established that a diet rich in methyl donors, such as folic acid enhances allergic airway disease in progeny [30], and knock-out studies of the DNA methyltransferase, Dnmt3a, resulted in dysregulation of important Th2 cytokines, including Il13, and increased inflammation in a mouse model of asthma [31]. In humans, large meta-analysis of GWAS identified few loci associated with SAR [32], and those loci identified did not contain genes encoding Th2 genes or other genes of known relevance for SAR. A recent study identified stable and functional DNA demethylation at the key regulatory gene FOXP3 in patients cured by specific immunotherapy (SIT) [33], directly linking disease reversal with DNA demethylation. Recent studies have reported differences in DNA methylation in airway epithelial cells between asthmatic children and healthy controls [34] and reported that methylation levels at the key cytokine gene, IL2, in cord blood was associated with asthma exacerbations in childhood [35]. However, systematic clinical studies of the genome-wide distribution and role of DNA modifications in Th differentiation in SAR are lacking. Analysis of the transcriptome of CD4+ T-cell subsets have failed to identify clear and reproducible differences between patients and healthy controls in several immune-diseases [6], [8], [9]. The circulating total CD4+ T-cell population is complex, comprised of several subsets that vary markedly between individuals. This complexity renders identification of subtle allergy-specific transcriptional signals challenging with current approaches. Unlike gene expression microarrays which typically assay 20,000–40,000 transcripts and have a small dynamic range, the DNA methylation microarrays employed here assay ∼450,000 individual CpG sites with quantitative accuracy. We suggest that quantitative DNA methylation microarrays can act as a proxy measure of the cell-population structure of samples, and as such, may be powerful analytical tools for diseases in which the proportions of different cell sub-types is likely to be of pathogenic significance such as immune-diseases and cancer [2], [17], [19]. Indeed, a very recent publication of genome-wide methylation in CD4+ T-cells from patients with Systemic Lupus Erythematosus (SLE), identified similar, small widespread changes in DNA methylation and associated these epigenetic changes with changes to CD4+ T-cell populations in SLE patients [14]. This exciting finding is completely consistent with the preliminary results reported here and suggests that alterations in CD4+ T-cell populations may be a general feature of many immune diseases. As CD4+ T-cell population structure and DNA methylation profiles reported here were determined in different cohorts, simultaneous analysis of the DNA methylation and gene expression profiles of CD4+ T-cells in the same cohort of patients and controls is required to directly determine the contribution of changes in subtype structure between patients and controls to the observed difference in total CD4+ T-cell methylation. Our findings highlight the potential to stratify immune diseases with DNA methylation. In vitro and in vivo array studies (cohorts 1–3, Figure S1) were approved by the ethics board of University of Gothenburg and all participants provided written consent for participation. The quantification of CD4+ T-cell subtypes study (cohort 4, Figure S1), was approved by the ethics board of Linkoping University and all participants provided written consent for participation. We recruited patients with SAR and matched healthy controls of Swedish origin at The Queen Silvia Children' s Hospital, Gothenburg (cohorts 1–3, Figure S1). We recruited patients with SAR and matched healthy controls of Swedish origin at Linkoping University Hospital, Linkoping (cohort 4, Figure S1). SAR was defined by a positive seasonal history and a positive skin prick test or by a positive ImmunoCap Rapid (Phadia, Uppsala, Sweden) to birch and/or grass pollen. Patients with perennial symptoms or asthma were not included. The healthy subjects did not have any history for SAR and had negative ImmunoCap Rapid tests. Supplemental Figure S1 provides an overview of the experiments performed on each cohort used in the research presented here. Severity of patient symptoms (itchiness of the eyes, block sinus, running nose) during season were self-assessed using a visual analogue scale (1–10). The score for each symptom was summed to give a single symptom severity score for each patient (Supplemental Table S1). A detailed description of all methods employed in this study can be found in Supplemental Text S1.

T-cells, a type of white blood cell, are an important part of the immune-system in humans. T-cells allow us to adapt our immune-response to the various infectious agents we encounter during life. However, T-cells can also cause disease when they target the body' s own cells, e. g. Psoriasis, or when they react to a harmless particle or ‘antigen’, i. e. allergy. Much evidence supports an environmental, or ‘epigenetic’, component to allergy. Surprisingly, although allergy is viewed as a T-cell disease with an epigenetic component, no studies have identified epigenetic differences between healthy individuals and allergic individuals. Using a state-of-the-art genome-wide approach, we found that we could clearly and robustly separate allergic patients from healthy controls. It is often assumed that these changes reflect changes in DNA methylation in a given type of cell; however such differences can also result from different mixtures of T-cell subtypes in the samples. Indeed, we found that allergic patients had different proportions of T-cell sub-types compared to healthy controls. These changes in T-cell proportions may explain the difference in DNA methylation profile we observed between patients and controls. Our study is the first successful molecular classification of allergy using CD4+ T cells.

Abstract Introduction Results and Discussion Materials and Methods

genomics immune cells immunity gene expression genetics t cells epigenetics molecular genetics biology dna modification immunology allergy and hypersensitivity genetics of disease immune response human genetics

2014

DNA Methylation Changes Separate Allergic Patients from Healthy Controls and May Reflect Altered CD4+ T-Cell Population Structure

Following trauma of the adult brain or spinal cord the injured axons of central neurons fail to regenerate or if intact display only limited anatomical plasticity through sprouting. Adult cortical neurons forming the corticospinal tract (CST) normally have low levels of the neuronal calcium sensor-1 (NCS1) protein. In primary cultured adult cortical neurons, the lentivector-induced overexpression of NCS1 induces neurite sprouting associated with increased phospho-Akt levels. When the PI3K/Akt signalling pathway was pharmacologically inhibited the NCS1-induced neurite sprouting was abolished. The overexpression of NCS1 in uninjured corticospinal neurons exhibited axonal sprouting across the midline into the CST-denervated side of the spinal cord following unilateral pyramidotomy. Improved forelimb function was demonstrated behaviourally and electrophysiologically. In injured corticospinal neurons, overexpression of NCS1 induced axonal sprouting and regeneration and also neuroprotection. These findings demonstrate that increasing the levels of intracellular NCS1 in injured and uninjured central neurons enhances their intrinsic anatomical plasticity within the injured adult central nervous system. Spinal cord injury is a significant clinical problem that produces life long disability, although in a minority of patients some degree of recovery can occur spontaneously without any therapeutic intervention [1], [2]. There are several possible mechanisms that could be responsible for this, one being anatomical plasticity, but such plasticity is very limited [3]–[5]. There is a growing literature suggesting pharmacological interventions can enhance both axonal regeneration [6]–[9] and anatomical plasticity [10]–[14] within the spinal cord, but little is known about the intracellular mechanisms responsible for such plasticity. Recently, we have found that following injury, the lentiviral overexpression of retinoic acid receptor β2 (RARβ2) induces regeneration in sensory and central axons [15], [16]. Microarray analysis of CNS tissue transduced with overexpressing RARβ2 lentivector was carried out to identify the intracellular molecular pathways involved in such regeneration. In unpublished data, this analysis revealed a significant upregulation of neuronal calcium sensor-1 (NCS1) in the transduced tissue as confirmed immunohistochemically and by real-time PCR. NCS1 is highly conserved across species and emerges as a key intracellular calcium signalling component in a number of regulatory pathways in neurons [17], [18]. This small molecule has been implicated in neuronal survival [19], short-term synaptic plasticity [20], and enhanced synapse formation and transmission [21]. Recently, it has also been suggested to regulate neurite outgrowth in pond snails [22], [23] and in primary cultured embryonic chick dorsal root ganglia neurons [24]. Here we show using lentiviral vectors that NCS1 overexpressed in primary cultured adult cortical neurons increases neurite sprouting. Following corticospinal tract (CST) denervation by unilateral pyramidotomy, axons of uninjured corticospinal neurons (CSN) overexpressing NCS1 sprout across the midline to form functional connections in the CST-denervated spinal cord. In axotomized CSN, overexpression of NCS1 induces axonal sprouting and regeneration and also neuroprotection. These studies reveal NCS1 as an important intracellular molecule for promoting anatomical plasticity following CNS injuries in the adult. To transduce adult cortical neurons with NCS1 at high efficiency and to enable visualisation with an extrinsic marker, we constructed a minimal human immunodeficiency virus (HIV) based lentiviral vector expressing NCS1 and GFP under cytomegalovirus (CMV) and spleen focus-forming virus (SFFV) promoters, respectively, termed HIV-GFP-NCS1 (Figure 1A). A control vector termed HIV-GFP was used that expressed only GFP under the CMV promoter (Figure 1B). It has been shown that although CMV is a stronger promoter than SFFV in GFP expression, the percentage of transduced neurons with GFP expression was similar [25]. The level of NCS1 in transduced primary adult rat cortical neurons measured immunocytochemically was more than 5-fold greater than in the control HIV-GFP-transduced neurons (Figure 1C–1I). In control-transduced neurons, few sprouts were observed from cell bodies or neurites (Figures 1C–1E, 1J–1K, and S1). In contrast, neurons transduced with HIV-GFP-NCS1 showed significant increases in the numbers of sprouts from both cell bodies and neurites (Figures 1F–1H, 1J–1K, and S1). The GFP immunostained cortical cells were confirmed as neurons by co-immunopositive staining with the neuronal growth associated marker GAP43 (Figure 1L). Furthermore, although two different promoters to drive GFP were used, the adequacy of GFP immunostaining for neurite distribution in both the NCS1- and GFP-transduced groups was confirmed to be similar to that obtained with phalloidin staining (Figure 1M–1N). The type of neurite that has undergone sprouting with NCS1 overexpression was further investigated using the specific dendritic immunomarker microtubule-associated protein 2 (MAP2). MAP2 has been previously shown strongly and weakly to immunolabel dendrites and axons, respectively [26], [27]. In control HIV-GFP-transduced neurons, few sprouts were observed in both dendrites and axons (Figure 2A–2F). In contrast, a significant increase in the number of sprouts on both dendrites and axons was observed in HIV-GFP-NCS1-transduced neurons (Figure 2G–2N). These data indicate that primary cultured adult cortical neurons overexpressed with NCS1 have significantly more neurites and sprouts from dendrites and axons than the control neurons. NCS1 has been shown to induce neuronal survival via the activation of the PI3K/Akt pathway [19]. We investigated whether this downstream intracellular pathway was also involved in the NCS1-induced neurite sprouting in primary cultures of adult mammalian neurons. The level of phospho-Akt in NCS1-transduced cortical neurons was significantly higher than in the control GFP-transduced group (Figures 3A–3D, 3I, and S2). Blockade of PI3K/Akt pathway with the inhibitor LY294002 caused a significant decrease in levels of phospho-Akt in the NCS1-transduced cortical neurons (Figures 3E–3F, 3I, and S3). This decrease corresponded to a significant 2-fold reduction in number of neurites from cell bodies and a 5-fold reduction in sprouts from neurites (Figure 3A–3K). However, neither phospho-Akt expression nor neurite sprout number was significantly changed in GFP-transduced neurons treated with LY249002 compared to vehicle treatment (Figures 3G–3K and S3). These data indicate the levels of phospho-Akt are elevated in neurons with overexpressed NCS1 and blockade of Akt production reduces neurite sprouting in these neurons. In Western blots, the levels of NCS1 were significantly higher in cortical neurons transduced with HIV-GFP-NCS1 compared with the controls both in vitro and in vivo (Figure 4A–4B, 4E–4F). This increase in NCS1 corresponds with a significant increase in phospho-Akt levels (Figure 4C–4D, 4G–4H). In the presence of LY249002, no significant change in NCS1 level occurred in NCS1-transduced cortical neurons compared to the controls (Figure 4A–4B, 4E–4F). This demonstrates that the reduction in phospho-Akt level was a direct result of LY249002 and not via the NCS1 overexpression itself. These data show that the neurite sprouting induced by NCS1 was indeed via the PI3K/Akt pathway. With the demonstration that NCS1 overexpression induces neurite sprouting in primary cultured adult cortical neurons, it was next determined whether this also occurred in vivo. HIV-GFP-NCS1 or control HIV-GFP lentivector was injected into the forelimb and hindlimb regions of the left sensorimotor cortex of adult rats. High efficiency was achieved of both GFP and NCS1 expression in the CSN at 3 wk with HIV-GFP-NCS1 (Figure 5A–5C). Similar numbers of GFP labelled neurons in layer V of the sensorimotor cortex were observed in both transduced groups as detected by immunohistochemistry (Figure 5D). Within such GFP labelled neurons, the percentage with NCS1 positive immunostaining was significantly higher in NCS1-transduced neurons than in the control GFP-transduced neurons (Figure 5D). Axons from NCS1-transduced CSN were visible in the pyramidal tract with GFP immunostaining (Figure 5E), in the main dorsal component of the CST, and in its collaterals at the cervical cord level (Figure 5F–5I). These data show that GFP allows a precise identification both of co-labelled transduced neurons overexpressing NCS1 and of their axons, thus obviating the need for later labelling with neuronal tracers or by the use of an independent virus expressing LacZ or GFP. Adult Wistar rats received unilateral intracortical injections of either HIV-GFP-NCS1 (NCS1-transduced) or the HIV-GFP (control) lentivector 3 wk before receiving on the contralateral side a unilateral pyramidal tract lesion which, in turn, causes CST-denervation of the contralateral side of the spinal cord (Figures S4A and S5A–D). The lesion site was defined with the astrocytic marker GFAP and the loss of PKCγ immunostaining caudal to the lesion site (Figures S5C–D and S8). At 6 wk post-CST-injury, GFP-immunostaining was performed to define axon collateral sprouting from the intact CST at the cervical and lumbar levels and particularly into the CST-denervated side of the spinal cord. The number of GFP-labelled axons in the CST was not significantly different between the control and NCS1-transduced rats (Figure 6A). In control rats, GFP-labelled collaterals were present in the CST-innervated side of the cord but few in the CST-denervated side at the cervical (Figures 6B–6F, 6M–6N, and S6) and lumbar (Figure 7A–7E) levels. In NCS1-transduced rats, GFP-labelled collaterals were present in the CST-innervated side, with a significant increase in the peak number of GFP positive fibres at the mediolateral region (Figures 6G–6I, 6N, 7F–H, 7L, and S7). More importantly, a significant increase also occurred in the number of GFP positive fibers sprouting across the midline into the CST-denervated cord. At the cervical level, GFP positive fibers in the range of 1–2 fibers per section for the control group compared to 4–5 fibers for the NCS1-transduced rats. At the lumbar level, GFP positive fibers of no more than 1 fiber per section for the control group compared to 5–6 fibers for the NCS1-transduced rats. This significant difference was maintained for up to 850 µm and 350 µm from the midline at the cervical and lumbar level, respectively (Figures 6J–6M, 7I–7K, and S7). The completeness of the tract lesions was confirmed by PKCγ immunostaining in the spinal cord (Figure S8). These data show that overexpression of NCS1 in CSN at the cortical level can induce distal axon collateral sprouting across the midline into the CST-denervated side of the cervical and lumbar spinal cord. It has been shown that a pyramidal tract lesion in adult hamster causes CSN to become atrophied after 2 wk post-injury [31]. To investigate whether NCS1 overexpression can prevent adult axotomized CSN from atrophy, adult Wistar rats received unilateral intracortical injections of either HIV-GFP-NCS1 or the control HIV-GFP lentivector 1 wk before an ipsilateral pyramidal tract lesion at the medullary level. To identify CSN, the retrograde Fast Blue tracer was injected directly into the lesion site immediately after sectioning (Figure S4D). Controls were unlesioned rats with Fast blue injected into the pyramidal tract at the medullary level with no intracortical lentiviral injection. After 2 wk post-lesion, the axotomized CSN in control GFP-transduced rats that have low NCS1 levels exhibited significant cell soma shrinkage compared to the large and healthy CSN in unlesioned rats (Figure 12A–12F and 12J–12K). In contrast, the axotomized CSN in NCS1-transduced rats that have high NCS1 levels did not exhibit significant cell soma shrinkage compared to the CSN in unlesioned rats (Figure 12G–12K). These data suggest NCS1 overexpression in CSN exerts a neuroprotective effect on axotomized CSN. This present study demonstrates that the intracellular levels of NCS1 in adult cortical neurons can be significantly elevated by transduction with a lentiviral vector. In culture, neurons overexpressing NCS1 develop extensive neurite sprouting which by immunocytochemistry and Western blotting was shown to be via Akt phosphorylation. Similarly, analogous experiments conducted in vivo show that CSN overexpressing NCS1 with intact CST axons can undergo distal collateral sprouting and cross the midline into the CST-denervated side of the spinal cord. This anatomical plasticity is also functional as demonstrated by the behavioural and electrophysiological outcomes in NCS1-transduced adult rats. Furthermore, studies on the axotomized CSN show that NCS1 overexpression not only induces axonal sprouting and regeneration at the lesion site but also exerts a neuroprotective affect on injured CSN. To date, several therapies for spinal cord injury models have shown both significant axonal regeneration [6]–[9] and anatomical plasticity [10]–[14] within the spinal cord. However, the intracellular mechanisms for these therapies have been little investigated. Only the purine-sensitive ste20-like protein kinase (Mst3b) has been linked to the anatomical plasticity observed with the purine nucleoside inosine [11], [26]. Interestingly, Mst3b has been shown selectively to induce outgrowth only from axons and not dendrites [26]. Conversely, in the present study, NCS1 overexpression induces sprouting from both axons and dendrites in cultured neurons, suggesting the growth induction process of NCS1 is non-selective. However, despite the existence of morphologically and molecularly distinct differences between dendrites and axons, neurons have been shown to have the capacity to generate axons from dendrites [32]. In addition to observing neurite sprouting in vitro, we demonstrated an NCS1 mediated axon collateral sprouting in vivo following unilateral pyramidotomy. The HIV-GFP-NCS1 lentivector injected into the cortex enabled GFP labelling of neurons overexpressing NCS1. The GFP labelling allowed visualisation and quantification of sprouting of the CST axons without the need of applying an independent tracer, and such labelling of axons and their collaterals can be detected as far distal as the lumbar region. We report for cervical region the number of GFP-positive collateral fibers from uninjured CST axons that have crossed the midline into the CST-denervated side of the spinal cord, as measured per 40 µm section, is 1–2 for the control and 4–5 for the NCS1-transduced rats. It is of interest to consider the possible total number of crossing fibers over the relevant cervical region. The length of the adult rat C5–C8 cervical cord containing the majority of forelimb motoneurons is approximately 11 mm [33], [34]. Extrapolating using these data, the NCS1-transduced rats would have approximately over 800 additional collaterals to account for the functional plasticity demonstrated behaviourally and electrophysiologically in these rats. Furthermore, this recovery was shown to be dependent on the collaterals provided by the intact CST as demonstrated by its loss following the second pyramidotomy as well as the loss of the crossed EMG activity when the intact CST axons were sectioned in the terminal experiment. We have previously induced axonal regeneration by lentivector overexpression of retinoic acid receptor β2 (RARβ2) [15], [16]. In this study, NCS1 overexpression has been demonstrated in axotomized CSN to induce axonal sprouting and regeneration. Recently, we showed that the retinoic acid receptor beta agonist (CD2019) overcomes inhibition of axonal outgrowth via the PI3/Akt pathway in injured adult rat spinal cord [35]. From our unpublished data, lentivector overexpression of RARβ2 also induces an upregulation of NCS1 as initially detected by microarray analysis and confirmed immunohistochemically and with real-time PCR. Thus this present study suggests that upregulation of NCS1 is a major intracellular target linking RARβ2 to the PI3K/Akt pathway in inducing anatomical plasticity and axonal regeneration. A similar explanation may account for GDNF-induced anatomical plasticity as other studies have shown NCS1 is upregulated by GDNF [14], [19], [36]. The successful regenerative responses of CSN after pyramidotomy with delayed post-injury NCS1 overexpression suggests axonal sprouting and regeneration can occur without the need to prime with overexpression of NCS1. These data provide promising support for NCS1 overexpression as a possible therapeutic treatment for CNS injury in a clinical setting. This study also reveals a neuroprotective feature of NCS1 overexpression in reducing cell shrinkage due to retrograde effects of axotomy. Others have shown that neurotrophic factors can prevent atrophy or death of axotomized CSN [31], [37]. Recently, Chondroitinase ABC, which is known to remove the inhibitory scarring at the injury site, has been shown to induce neuroprotection of CSN via a possible retrograde effect mediated at the injured mouse spinal cord [38]. However, our study demonstrates that NCS1 overexpression can increase the intrinsic capacity of CSN to overcome the inhibitory environment and even compensate for the apparent lack of trophic support associated with CNS injury. Our study establishes that NCS1 is an important intracellular component in the regulation of axonal sprouting and regeneration, and neuroprotection in central neurons of an adult mammalian nervous system, as recently shown for the peripheral nervous system in chick embryo studies on dorsal root ganglion (DRG) neurons [24]. Furthermore, the PI3K/Akt pathway mediating these responses in vitro and in vivo is consistent with the experiments on primary cultured adult DRGs and perinatal cortical neurons linking Akt activation with neurite outgrowth [39], [40] and the survival of primary cultured embryonic cortical neurons [19]. The opposite conclusion, that NCS1 contributes to a retardation of neurite growth, may relate to the use of a rat adrenal medullary pheochromocytoma cell line (PC12 cells) and the additional need for NGF to promote differentiation into sympathetic neuron-like cells [41], [42]. In summary, the limited ability of adult CST neurons to undergo functional sprouting may be due to low endogenous levels of NCS1. Thus NCS1 emerges as a potential intracellular target for therapeutic intervention following injury to the central nervous system. The complete cDNA sequence of the rat NCS1 was generated by PCR from adult rat cortex using the following forward primer; 5′-ATGGGGAAATCCAACAGCAAG-3′; and the reverse primer, 5′-CTATACCAGCCCGTCGTAGAG-3′ then cloned into a pCR2. 1-TOPO vector (Invitrogen). The ncs1 gene was inserted under the control of a CMV promoter in a human immunodeficiency virus type 1 (HIV1) vector containing a 5′ central polypurine tract (cPPT) element and a 3′ woodchuck post-transcriptional regulatory element (WPRE) enhancer. To allow for coexpression of NCS1 and the enhanced green fluorescent protein (GFP), the eGFP gene was inserted into the Cla1 site under the control of a SFFV promoter. Viral vector stocks, pseudotyped with the VSV-G envelope glycoprotein, were prepared by triple plasmid transient transfection of HEK293T cells as previously described [43]. The titre of pRRL-CMV-NCS1-SFFV-eGFP (for simplicity termed HIV-GFP-NCS1) was 3. 3–4. 4×108 TU ml−1 and pRRL-CMV-eGFP (for simplicity termed HIV-GFP) was 4. 7–4. 8×108 TU ml−1 determined by transient transfection of the HEK 293T cell line and analysed by flow cytometry. Adult cortical neurons were cultured as previously described [16], [44]. Adult male Wistar rats (220–250 g) were overdosed with sodium pentobarbitone (Lethobarb), transcardially perfused with heparinized saline, and the cortices removed with as little white matter as possible. The cortices were cut into 0. 5 mm longitudinal sections using a McIlwain tissue chopper before digestion in 2 mg/ml papain at 30°C for 30 min. Cortical neurons were mechanically dissociated with a glass Pasteur pipette and separated from debris by centrifugation in four 1 ml steps of Optiprep in HibernateA/B27 medium (7. 5%, 10%, 12. 5%, and 17. 5%) at 600 g for 15 min. Fractions containing neurons were collected, washed, and resuspended in NeurobasalA/B27 medium for plating at a density of 3,000 cells per well on poly-D-lysine (10 µg/ml) pre-coated cover slips. The neurons were allowed to settle onto the cover slips for 1 h, and after washing, HIV-GFP-NCS1 or control HIV-GFP lentivector was added to the media at MOI 10. The neurons were left for a further 3 days in vitro (DIV) before immunocytochemical processing. For phospho-Akt studies, the reversible PI3K/Akt inhibitor LY294002 (hydrochloride, 40 µM dissolved in DMSO, Tocris Biosciences) or DMSO alone at a final concentration of 0. 01% was added to the cultures at 1 DIV and 2 DIV. The neurons were harvested at 3 DIV. After 3 d in culture, neurons were fixed with 4% paraformaldehyde for 20 min and then permeablized with cold methanol for 3 min. The neurons were washed three times for 5 min with 0. 01 M phosphate-buffered saline (PBS) before 2 h incubation with chicken anti-GFP (1∶1000, ab13970, Abcam), rabbit anti-NCS1 (1∶500, NL3750, BioMol International), rabbit anti-phospho-Akt (1∶200, Cell Signalling Technology), or rabbit anti-MAP2 (1∶1000, AB5622, Chemicon). The cover slips were washed 3×5 min with PBS and then incubated with donkey anti-rabbit Alexa Fluor 546 and goat anti-chicken Alexa Fluor 488 secondary antibodies (1∶2000, Molecular Probes) for 45 min. After 3×5 min PBS washes, they were mounted with FluorSave reagent containing 0. 5 µl DAPI (10 µg/ml) to visualise cell nuclei. The staining with phalloidin-TRITC (1∶100, P5282, Sigma) was carried out 1 h before immunostaining with other antibodies was carried out. Image analysis and quantification was made with the observer blinded to the group assignment as previously described [16], [45], [46]. Analyses were restricted only to transduced neurons immunoexpressing GFP. For each experimental group, 50–100 GFP-positive neurons were captured randomly using a Zeiss Axioplan 2 fluorescence microscope. The soma of each neuron was outlined to obtain the fluorescent intensity using the Axiovision V4. 6 software to determine the neuronal levels of NCS1 and phospho-Akt immunoreactivity. To minimise variability between each image, the capture settings were fixed throughout the whole study. The number of neurite sprouts from the cell bodies and of the longest neurites, of length greater than cell body diameter was determined. To differentiate whether a GFP positive neurite was a dendrite or an axon, the specific dendritic marker microtubule-associated protein 2 (MAP2) was used. Neurites with strong and weak MAP2 immunostaining were identified as dendrites and axons, respectively. Western blots were carried out as previously described [47], [48]. After 3 d in culture, primary adult cortical neurons transduced with either HIV-GFP-NCS1 or the control HIV-GFP vector, with or without the PI3K/Akt inhibitor LY294002, media were removed and neurons were harvested in 250 µl ice-cold lysis buffer (20 mM HEPES pH 7. 4,100 nM NaCl, 100 mM NaF, 1 mM Na3VO4,5 mM EDTA, 1% Nonidet P-40 and 1× protease inhibitor cocktail; Roche). To obtain sufficient protein, the same 250 µl lysis buffer was used in three cultured wells and the lysates rotated for 2 h at 4°C. After centrifugation at 13,500 g for 15 min at 4°C, the supernatant was collected and total protein concentration was determined using a bicinchoninic acid protein assay kit (Pierce). Intracortical injections of either HIV-GFP-NCS1 or HIV-GFP lentivector in adult male Wistar rats (n = 4–5 per group) were carried out as described below. To determine the role of Akt activation, the PI3K/Akt inhibitor LY294002 (100 mM) or vehicle (DMSO) was injected intracerebroventricularly via an externalised catheter on every other day of the third post-injection week. At the end of the third week, rats were sacrificed and the injected region of the cortex was freshly and quickly removed and stored at −80°C until further processed. The protein obtained for Western blotting was extracted as described above. Fifteen micrograms of total protein were electrophoresed on 12% acrylamide gel before transfer onto Hybond P membranes (Amersham) and incubated overnight at 4°C with rabbit anti-phospho-Akt (Ser 473,1∶100, #3787S, Cell Signalling Technology), rabbit anti-NCS1 (1∶1000, NL3750, BioMol International), or mouse anti-β III tubulin (1∶1000, G712A, Promega). Visualisation was performed using secondary antibodies, donkey anti-rabbit IRDye-800CW, and goat anti-mouse IRdye-680CW (LI-COR Biosciences). Fluorescent blots were imaged on the Odyssey Infrared Imaging System (LI-COR Biosciences). To allow for visualisation of the total Akt on the same blot as phospho-Akt (both antibodies were raised in the same species), the blot was first stripped with buffer (62. 5 mM Tris-HCl pH 6. 8,2% SDS, 100 mM β-mercaptoethanol) before re-blotting with rabbit anti-Akt (1∶100, #9272, Cell Signalling Technology). Western blotting was carried out with 3–5 independent samples. The surgery was performed aseptically in accordance with UK Home Office regulations as previously described [16]. Briefly, adult male Wistar rats (n = 8–9 per group) were anaesthetized using a combination of ketamine and medetomidine, then fixed in a stereotaxic frame. The skull was exposed and injections were made at a depth of 2 mm dorsoventrally into the sensorimotor cortex region using the injection coordinates as determined from a microstimulation mapping study [28]. These were, with reference to bregma (AP, anterior-posterior; L, lateral); AP: −1. 5 mm, L: 2. 5 mm; AP: −0. 5 mm, L: 3. 5 mm; AP: +0. 5 mm, L: 3. 5 mm; AP: +1. 0 mm, L: 1. 5 mm; AP: +1. 5 mm, L: 2. 5 mm; AP: +2. 0 mm, L: 3. 5 mm. At each site, 1 µl of HIV-GFP-NCS1 or control HIV-GFP lentivector was directly injected at a rate of 0. 2 µl/min using a microinfusion pump via a finely pulled glass micropipette and left in situ for a further 1 min. HIV vector pseudotyped with a VSV-G envelope produced strong expression and anterograde labelling [49]. Three weeks after viral injection, a unilateral pyramidal tract lesion at the level of medulla was performed as described previously [50]. A ventral midline incision was made and the occipital bone exposed by blunt dissection. The ventrocaudal part of the bone was partially removed using fine rongeurs, exposing the right medullary pyramid. The dura was opened and the right pyramidal tract was sectioned approximately 2 mm rostral to the decussation with fine iridectomy scissors using the basilar artery as the midline. Sham operated rats received similar surgery without incision of the tract. In another group of experiments, the left intact pyramidal tract was transected in a second operation. In the delayed lentivector transduction studies, adult Wistar rats (n = 5–6 per group) received intracortical lentiviral injections 2 d after a unilateral pyramidal tract lesion as described above. To study the sprouting effect of delayed lentivector transduction on uninjured and injured CST axons, the lentiviral injections were administered into the sensorimotor cortex corresponding to the unlesioned and lesioned pyramidal tract, respectively. After 4 wk post-surgery, the rats were perfused transcardially with 4% paraformaldehyde and tissue collected for histology. In the neuroprotection study, adult Wistar rats (n = 4–5 per group) received intracortical lentiviral injections 1 wk before receiving on the ipsilateral side a unilateral pyramidal tract lesion as described above. Using a microinfusion pump, Fast Blue tracer (200 nl, 2% wt/vol PBS, EMS-Chemie GmbH) was administered at a rate of 0. 2 µl/min into the lesion site via a finely pulled glass micropipette and left in situ for a further 1 min. Unlesioned rats without intracortical lentiviral injection had Fast Blue tracer (2%, 200 nl) injected into the pyramidal tract at the medullary region. Care was taken to minimise axonal damage by the injection process. After 2 wk post-injection of tracer, the rats were perfused transcardially with 4% paraformaldehyde and tissue collected for histology. The analysis of atrophy in CSN was carried out as previously described [38]. The cell area of CSN co-labelled with Fast Blue tracer and GFP immunostaining were acquired using the AxioVision V4. 6 program by an investigator blinded to the treatment groups. In unlesioned rats, Fast Blue traced CSN from similar coronal levels as for the lentivector transduced rats were analyzed. Size and frequency distributions of CSN were determined for each rat and a mean distribution calculated for each treatment group. At least six transduced sections were analysed and quantified per rat (n = 4–5 per group). A total of over 2,400 neurons were analyzed. Following unilateral pyramidotomy, functional recovery was assessed behaviourally using the staircase reaching and grid exploration tests at 2 d post-lesion and then weekly for 6 wk as described previously [15], [16], [50]. In the staircase reaching test, the rats were trained to reach and grasp the food pellets from a baited double staircase (Campden Instruments) before CST lesion. This test allows assessment of extension and grasping ability independently for each forelimb. On the testing day, rats were placed in the staircase box for 15 min and the number of food pellets removed or displaced was recorded. In the grid exploration test, the rats were allowed to explore the grid freely (40 cm×60 cm containing 5 cm×5 cm mesh, raised 50 cm high) where at least 50 forelimb and 20 hindlimb steps were recorded, typically made within 3 min. The “free” exploration removes any possible learning effect due to training as no pre-training was required and that the rats never move around the grid in the same pattern. The grid exploration captured on video camera was replayed and analysed for limb misplacement on the grid. Analysis involved counting the number of limb misplacement from the first 30 forelimb and 20 hindlimb placements of each rat, to ensure no bias between animals and groups. At the end of the behavioural assessment at 6 wk, the rats were sacrificed and perfused transcardially with 4% paraformaldehyde and tissue collected for histology. Control rats (n = 4) received intracortical injections of HIV-GFP while NCS1-transduced rats (n = 4) received HIV-GFP-NCS1 followed by a unilateral pyramidotomy on the right side 3 wk later as described above. All electrophysiological measurements were performed at least 6 wk post-injury under urethane (1. 25 g/kg body weight, i. p.) anesthesia. Following tracheotomy, the rat was fixed into a frame by ear bars and spinal clamps such that the forelimbs were fully pendent. A pair of hooked stainless steel wires insulated to the tip was inserted into the tricep brachii of both forelimbs approximately 6 mm apart for EMG recording. The area of the sensorimotor cortex where lentiviral vectors were injected was exposed by craniotomy, covered by mineral oil, and stimulated through a flat ended silver wire electrode (0. 5 mm diameter) ensheathed with plastic to its tip to minimize surface spread of the stimulating current. A 2 mm diameter anode was placed on the skull periosteum rostral to the stimulating electrode. The stimuli repeated at 1 Hz consisted of 1 to 4 pulses, 3 ms apart, and 0. 1 ms duration from an isolated stimulator (Digitimer DS2A). The final stimulation site was selected after systematic mapping with varying stimulus parameters until a discrete contralateral (left) forelimb movement was observed with a clear EMG response and a threshold below 25 V for the least number of effective pulses. The EMG was amplified (LF, 30 ms TC; HF, 3 KHz) and digitized using a CED 1401 interface with a sampling rate of 10 kHz. The area of the EMG response (Vs) was measured from 20 averaged responses using Spike 2 V5. 0 software. To check that the EMG response from the CST-denervated forelimb was dependent on contralateral CST input to the spinal cord, the right dorsal CST was transected at the cervical C4 level using a chisel formed by flattening a G25 needle. Data were analyzed using SigmaStat 3. 5 software. Reported values are expressed as mean ± SEM. The in vitro experiments, Western blot analysis, and number of GFP positive axons in the dorsal CST were analyzed with Student' s t test. The GFP immunopositive axon collaterals and sprouts in the cord and brainstem, behavioural tasks, and electrophysiology were analyzed with two-way ANOVA followed by Tukey' s post hoc test. The cell size cumulative frequency distribution of CSN was analyzed with a two-sample Kolmogorov-Smirnov test, performed against a significant threshold of 0. 05 to correct for multiple testing.

Following trauma to the central nervous system (brain or spinal cord), neurons show very little capacity to re-grow their axons, which can lead to a permanent loss of function in those regions. In this study, we show that this failure of axon re-growth is associated with low intracellular levels of a small molecule called neuronal calcium sensor-1 (NCS1). We modified a non-replicating virus in two ways so as to increase the level of NCS1 in neurons while simultaneously labelling them with a green fluorescent protein, which allowed us to track neuronal growth. Using this virus to increase the level of NCS1 in a particular group of neurons that communicate between the brain and spinal cord, we showed that new axonal growth occurred in the spinal cord with or without injury to the neurons. Electrophysiological assessments demonstrated that these new processes formed functional connections in the spinal cord, and behavioural experiments revealed that this recovery also helped the animals move their limbs more effectively. Furthermore, an increase in NCS1 protected these neurons, such that fewer of them died after the injury. These findings demonstrate that increasing the intracellular levels of NCS1 in neurons can aid in the recovery from central nervous system injury, and can help improve behavioural function.

Abstract Introduction Results Discussion Materials and Methods

2010

Cortical Overexpression of Neuronal Calcium Sensor-1 Induces Functional Plasticity in Spinal Cord Following Unilateral Pyramidal Tract Injury in Rat

Although infection with Toxocara canis or T. catis (commonly referred as toxocariasis) appears to be highly prevalent in (sub) tropical countries, information on its frequency and presentation in returning travelers and migrants is scarce. In this study, we reviewed all cases of asymptomatic and symptomatic toxocariasis diagnosed during post-travel consultations at the reference travel clinic of the Institute of Tropical Medicine, Antwerp, Belgium. Toxocariasis was considered as highly probable if serum Toxocara-antibodies were detected in combination with symptoms of visceral larva migrans if present, elevated eosinophil count in blood or other relevant fluid and reasonable exclusion of alternative diagnosis, or definitive in case of documented seroconversion. From 2000 to 2013,190 travelers showed Toxocara-antibodies, of a total of 3436 for whom the test was requested (5. 5%). Toxocariasis was diagnosed in 28 cases (23 symptomatic and 5 asymptomatic) including 21 highly probable and 7 definitive. All but one patients were adults. Africa and Asia were the place of acquisition for 10 and 9 cases, respectively. Twelve patients (43%) were short-term travelers (< 1 month). Symptoms, when present, developed during travel or within 8 weeks maximum after return, and included abdominal complaints (11/23 symptomatic patients, 48%), respiratory symptoms and skin abnormalities (10 each, 43%) and fever (9,39%), often in combination. Two patients were diagnosed with transverse myelitis. At presentation, the median blood eosinophil count was 1720/μL [range: 510–14160] in the 21 symptomatic cases without neurological complication and 2080/μL [range: 1100–2970] in the 5 asymptomatic individuals. All patients recovered either spontaneously or with an anti-helminthic treatment (mostly a 5-day course of albendazole), except both neurological cases who kept sequelae despite repeated treatments and prolonged corticotherapy. Toxocariasis has to be considered in travelers returning from a (sub) tropical stay with varying clinical manifestations or eosinophilia. Prognosis appears favorable with adequate treatment except in case of neurological involvement. Toxocariasis, caused by intestinal roundworm of dogs (Toxocara canis) or cats (Toxocara catis), is a zoonotic infection with a worldwide distribution [1,2]. Humans can get infected by ingestion of embryonated eggs present on the soil, plants or soil-dwelling invertebrates contaminated by dog or cat feces, and less frequently by ingestion of encapsulated larvae from undercooked paratenic hosts such as chickens, cattle and sheep. Infection is followed by the migration of third-stage larvae through the tissues which is usually asymptomatic but may be associated with a variety of non-specific clinical features [2,3]. Presentation may be acute or sub-acute, with systemic, abdominal or respiratory manifestations, classically described as the syndrome of visceral larva migrans (VLM) sometimes associated with dermatological symptoms as well. Occasionally, the course is complicated by the involvement of the central nervous system (neurological toxocariasis) or the eyes (ocular toxocariasis). Like in other helminthic infections with larval migration, blood eosinophilia is common and/or numbers of eosinophils may be increased in other relevant tissues or fluids; other laboratory abnormalities such as elevated inflammatory parameters or liver enzymes disturbances are often mild and inconstant. Of note, a distinct clinical presentation called “covert” (in children) or “common” (in adults) toxocariasis has been described more recently, with more subtle and chronic symptoms such as cough, abdominal pain or pruritus, associated with mild eosinophilia and Toxocara spp. seropositivity [4–6]. It is however assumed that most cases of human toxocariasis go unrecognized. A definitive diagnosis of human toxocariasis can only be made by histological examination of infected tissue, demonstrating Toxocara spp. larvae within eosinophilic granulomatous lesions. Since sampling of appropriate tissue is rarely justified on clinical grounds, diagnosis of human toxocariasis relies in the daily practice on a constellation of suggestive symptoms (if present), combined with laboratory abnormalities (blood eosinophilia) and the detection of circulating immunoglobulin G (IgG) antibodies to Toxocara excretory–secretory (TES) antigens [7]. In Western countries with low prevalence of helminthic infections, the association of clinical symptoms (although frequently poorly specific) with eosinophilia usually triggers clinicians aware of the disease to perform the specific serological investigations [1]. In the tropics however, where polyparasitism is highly prevalent, the low specificity of eosinophilia, the lack of diagnostic facilities and the cross-reactivity with other helminthic infections often preclude the etiologic diagnosis [2]. While larvae may remain viable for years in the tissues, human toxocariasis is usually a self-limiting disease, although its course can sometimes be invalidating and prolonged (up to several weeks). Treatment firstly aims at controlling the inflammatory reaction when needed, and despite the limited knowledge about its clinical benefit, anti-helminthic therapy is often associated. In such cases, preference usually goes to albendazole for its parasitological efficacy and good clinical tolerance [8]. The global burden of human toxocariasis is poorly quantified [9]. Serological surveys demonstrate that the infection is more frequent in tropical settings and in rural areas, with population-based seroprevalence ranging from 2. 5% in urban Europe up to 85% in rural tropics [9–14]. Visceral larva migrans has been reported in individuals born in tropical countries having migrated to Europe [15]. However, although many susceptible European travelers are assumed to be at increased risk of exposure to Toxocara spp. during a stay in highly endemic developing countries, frequency and presentation of toxocariasis are largely unknown in this population. A recent large multicenter GeoSentinel study has reported 16 cases diagnosed as visceral larva migrans among 42,173 travelers evaluated from 2007 to 2011 (38 cases per 100,000 travelers) but there was no standard diagnostic approach in the 53 participating travel clinics and no clinical description [16]. In the present study, we aimed to assess the frequency, presentation and outcome of toxocariasis diagnosed among travelers and migrants presenting at the travel clinic of the Institute of Tropical Medicine of Antwerp, Belgium. For this retrospective study, a query was first undertaken in the database of the Central Laboratory of Clinical Biology of the Institute of Tropical Medicine, Antwerp (ITMA), to retrieve all results of Toxocara serology requested in patients having attended the travel clinic of the ITMA from 2000 to 2013. The ITMA is the national reference center for tropical medicine in Belgium, with on average about 6500 consultations a year for post-travel care. The medical records of all travelers and migrants found with a positive Toxocara serology during this period were then reviewed. Relevant clinical and laboratory data were extracted, de-identified and entered in a Microsoft Access 2010 database. Variables included: demographic data, month and year of first Toxocara positive serology, most recent travel destination, time of symptom onset after travel, duration of symptoms before consultation, clinical presentation, result of the chest X-rays if requested, absolute blood eosinophil count (and percentage of white blood cell count), Toxocara antibody optical density result, results of parasitological and serological tests prescribed by the physician targeting other helminths according to epidemiological relevance (Ascaris spp. , Echinococcus granulosus, Fasciola spp. , Filaria spp. , Schistosoma spp. , Taenia solium, Strongyloides stercoralis, Trichinella spp. and Anisakis simplex), administered treatment (s), clinical and laboratory evolution and outcome. Next, we categorized all reviewed Toxocara-seropositive cases in four groups according to the clinical reasons for requesting Toxocara serology: 1) presence of symptoms compatible with VLM combined with eosinophilia; 2) asymptomatic eosinophilia; 3) symptoms compatible with VLM without eosinophilia; and 4) possible exposure/no clear reason (Fig. 1). For this clinical study, strict case definitions of symptomatic (Toxocara-associated visceral larva migrans) and asymptomatic toxocariasis were used. Diagnosis of Toxocara-associated VLM was considered as highly probable when the following criteria were all fulfilled: (1) positive Toxocara serology (entry criteria) AND (2) presence of any systemic symptom compatible with toxocariasis (including fever, respiratory signs such as wheezing, dry cough, dyspnea or an infiltrate on the chest X-ray, abdominal symptoms as abdominal pain, vomiting, diarrhea or hepatomegaly, neurological signs such as focal deficit or encephalopathy, and ocular signs such visual disturbances), with or without dermatological symptoms such as pruritus, urticarial rash or angio-edema AND (3) blood eosinophilia (defined as an absolute blood eosinophil count above 500/μL, or > 7% of the white blood cell count at first presentation [17,18]), or presence of eosinophils in another relevant fluid/tissue) AND (4) reasonable exclusion of alternative diagnosis. Diagnosis of VLM was considered definitive when all 4 criteria were present together with unequivocal Toxocara seroconversion documented on paired serum samples. Diagnosis of asymptomatic toxocariasis relied on the presence of blood eosinophilia at presentation in asymptomatic Toxocara-seropositive individuals and no evidence of other infection likely to explain the eosinophilia. To comply with the case definition, we therefore did not further analyze groups 3 and 4 (since there was no eosinophilia), and then carefully assessed for all cases of groups 1 and 2 whether an alternative diagnosis or a co-infection were possible. In accordance with the case definitions, we finally excluded from this study all patients with parasitological or serological evidence of another infection or clinically suspect of alternative diagnosis (such as allergy, scabies, …). From 2000 to 2009, the serological diagnosis of toxocariasis was performed with a commercial anti-TES IgG enzyme-linked immunosorbent assay (ELISA) (Toxocara canis, Bordier Affinity Products SA, Crissier, Switzerland) according to the instructions of the manufacturer. The assay’s threshold for positivity was set as “weak positive” (from 2000 to 2009) and at 1. 0 (measured by optical density, since 2010 onwards) for use in clinical settings [13]. In this context, the assay has been reported to provide a sensitivity and a specificity of 91% and 86% respectively, the latter being evaluated in a population of patients with protozoan and helminthic infections [19]. When toxocariasis is strongly suspected or confirmed, the standard treatment regimen at ITMA is albendazole 2 x 400 mg (in adults) for 5 days, according to international guidelines. Corticosteroids may be added at the physician’s discretion according to severity of symptoms. Diethylcarbamazine (DEC) is restricted as second line therapy for refractory cases [20,21]. Analysis was performed with the SPSS program version 20. 0 (SPSS Inc. , Chicago, IL, USA). Differences were compared using Student’s t-test or Mann-Whitney U-test, when appropriate, for continuous outcome and chi-square and Fisher’s exact tests for categorical outcomes. P values of less than 0. 05 were considered to indicate statistical significance. This was a retrospective analysis of data collected during routine clinical care over a 13-year period. Ethical clearance was obtained from the institutional review board at ITMA. Laboratory queries were obtained in an anonymous way. Clinical data were then retrieved through an encoded link and de-identified for analysis according to the Belgian legislation. No written informed consent was obtained from individual participants because an opt-out strategy is in place at ITMA covering surveillance use of clinical and laboratory data. From January 2000 to August 2013,3436 Toxocara serological tests were ordered for diagnostic purpose in post-travel care by the ITMA physicians. Of these tests, 190 (5. 5%) had positive anti-TES IgG (Fig. 1). Of 187 patients with complete clinical data, 44 had VLM symptoms and eosinophilia (group 1), 35 were found with asymptomatic eosinophilia (group 2) and 54 had symptoms compatible with VLM but no eosinophilia (group 3, excluded from the study); for the remaining 54 individuals with no symptoms and no eosinophilia, the reason for Toxocara testing was considered as unclear (group 4, also excluded from further analysis). In the groups 1 and 2, after exclusion of the cases with alternative diagnosis or possible co-infection, a diagnosis of human toxocariasis was retained in 28 patients, including Toxocara-associated VLM in 23 and asymptomatic toxocariasis in 5 (Fig. 1). Diagnosis of VLM was considered as definitive in 7 cases for whom seroconversion was observed. Histological examination was not performed in any case. Of note alternative diagnoses in the excluded cases of groups 1 and 2 (n = 51) were mainly strongyloidiasis (n = 15), allergic reaction (n = 7), filarial infection (n = 7), Ancylostoma/Ascaris infection (n = 7) and schistosomiasis (n = 6). Suspicion of co-infection was also frequent. The clinical presentation of the 28 cases is detailed in Table 1. Cases were evenly distributed throughout the study period with no cluster phenomenon. All patients were adult travelers born or residing in Europe, except one Ethiopian child evaluated after adoption and one Lebanese adult living in the Democratic Republic of the Congo. Mean age was 46 years (range: 4–68 years) with a male/female ratio of 0. 87. Regions of most recent travels (and presumed exposure) were North and sub-Saharan Africa in 10 patients, Southern and Southeast Asia in 9 and Southern Europe (including Turkey) in 6; for 3 patients (numbers 18,19 and 26), the continent of acquisition could not be traced with certainty because of multiple travel destinations within a short timeframe. Duration of travel was less than 1 month in 12 (43%) patients and more than 3 months in 10 (36%). For the 23 cases presenting with VLM, symptoms started during the stay abroad in 11 (48%) or developed within 3 weeks in average (range: 0–8 weeks) after return. For one patient (number 11), the dates of recent travel were not clearly reported. Clinical manifestations included abdominal complaints in 11 (48%) patients, respiratory symptoms and skin abnormalities in 10 (43%) each and fever in 9 (39%); in most cases, symptoms were combined or developed sequentially. Radiological pneumonia was found in 5 patients and one of them (number 22) had to be admitted elsewhere because of the severity of respiratory symptoms. Two patients (numbers 11 and 16) presented with progressive neurological features of transverse myelitis but no other symptoms. In both cases, the diagnosis was made by the demonstration of an increased eosinophil count and of anti-TES IgG in the cerebrospinal fluid, in the absence of another etiology (Table 1). For the 21 patients without neurological complications, median duration of symptoms before first documented medical evaluation at ITMA or elsewhere was 3 weeks (range 5 days- 4 months), while both patients with neurological toxocariasis were diagnosed several months (2,5 and 7 respectively) after symptom onset. At first presentation, the median blood eosinophil count was 1720/μL (range: 510–14160) in the 21 VLM cases without neurological complication and 2080/μL (range: 1100–2970) in the 5 asymptomatic cases. Of note, the median eosinophil count was significantly lower in the VLM patients presenting more than 4 weeks after symptom onset than in those consulting earlier (700/μL versus 2340/μL, p<0. 001). The blood eosinophil count was within the reference ranges for both patients with neurological toxocariasis. Of the 5 patients with asymptomatic toxocariasis, 3 received the 5-day course of albendazole upon diagnosis, and the eosinophil count normalized uneventfully; the other 2 patients were empirically treated before the complete results were available: one with ivermectin (who required albendazole secondarily because eosinophilia persisted) and the second with ivermectin and praziquantel (with resolution of the eosinophilia). Of the 23 symptomatic cases, 15 received a standard course of albendazole within one or two weeks after presentation (when laboratory results were available): 11 clinically recovered and eosinophil counts normalized (of note 4 had already clinically improved before the start of therapy); one was lost to follow-up; another had to switch to DEC because of immediate anosmia attributed to albendazole; both neurological cases were given steroids concomitantly with albendazole, but since the improvement was slow DEC was also administered (with resolution of the radiological abnormalities and moderate clinical recovery). Four patients were initially treated empirically with praziquantel (n = 1), or ivermectine (n = 3) but 3 of them required albendazole secondarily in the absence of substantial clinical/laboratory improvement. Finally the remaining 4 patients preferred at first not to be treated; 3 of them recovered spontaneously but the 4th patient needed a course of albendazole a few weeks later and persisting symptoms eventually subsided. Of note only one non-neurological patient (number 22) required corticosteroids during hospitalization to control the severe respiratory symptoms. Although toxocariasis is highly prevalent in most tropical areas, this condition has been hardly studied in returning travelers and migrants. We describe here a case series of 28 patients diagnosed with toxocariasis acquired from all over the world. Clinical presentation was extremely varied and resembled that of many other endemic helminth infections. Morbidity was important and complications sometimes serious. This study has many limitations. It was indeed a retrospective single-center study conducted in a reference travel clinic, meaning that collection of data was not systematic and that findings may not be generalizable to all clinical settings. For instance, some cases with longer incubation period may have been directly attended in the primary care setting given the fact that the link between symptoms and travel had become less obvious. Also our observations are not transposable as such to the features observed in autochthonous toxocariasis sporadically seen in Belgium or elsewhere, mainly in children [22]. Other limitations could be related to the restrictive case definition that may have missed several true cases without eosinophilia (as sometimes observed in milder cases of “covert/common” toxocariasis) or with negative Toxocara serology (during the serological window period or because sensitivity does not reach 100%, in particular in low burden infection such as ocular toxocariasis [2]). In the same line, some true toxocariasis cases may have been disregarded just because serological tests against other helminths were also positive, either by cross-reactivity or as part of infection with multiple parasites. Conversely, false positive result was also possible if anti-Toxocara seropositivity was just reflecting remote exposure or cross-reaction while another helminthic infection was missed during the workup. The commercial test we used is widely considered as an adequate screening tool for clinical practice but indeed detects anti-TES IgG that may persist for years; IgM or IgE-based serological tests that could better discriminate recent infection are not routinely available. On the other side, for European travelers with little previous exposure to parasites, cross-reaction is probably not a major issue, since test threshold has been set at value providing good specificity. In addition, the consistent clinical and laboratory diagnostic approach by a stable group of expert physicians throughout the study probably reflected the best accuracy that can be obtained in routine care. Finally, in this series, infection was most likely acquired abroad since symptoms developed during or shortly after travel, but infection in Belgium before travel or after return cannot be fully excluded. The observation of toxocariasis in travelers is not surprising, although poorly studied so far. In a 10-year retrospective study in Spain, Toxocara antibodies were detected in 31 (4. 9%) of 634 Latin American migrants and VLM was diagnosed in 4 of them [15], but this was in a particular segment of the travel population. With 28 highly probable/definitive toxocariasis cases diagnosed in about 85,000 travelers during the 13-year study period (33/100,000 travelers), our findings are in line with the recent multicenter GeoSentinel study (38/100,000 travelers) although the actual frequency was probably somewhat underestimated given the very restrictive case definition. Because of the high prevalence of toxocariasis in tropical countries and the inherent risks related to visiting regions with substandard hygiene (exposure to locally prepared food, incidental contacts with animals [23], it is reasonable to include toxocariasis in the differential diagnosis of most travel-related illnesses. However, surprisingly, the rate of Toxocara seropositivity in the (suspected) travelers for whom the test was requested (5. 5%) was quite similar to that found in suspected autochthonous cases in Denmark (5. 5%) [13] or the Netherlands (5–10%) [12] at similar ages. This observation does not support the idea that exotic travel by itself represents a major risk factor for Toxocara seropositivity, but since the frequency of clinical toxocariasis was not reported in those studies, comparisons with our findings remain inconclusive. Finally, only one study conducted in the Netherlands has investigated the incidence rate of Toxocara infection in travelers by comparing pre- and post-travel serology and found 1. 1 seroconversion per 1000 person-months [24]. We confirm here that, even if infrequent, toxocariasis does occur in travelers and has to be considered after any type or duration of travel and from any destination. Several factors contribute to underdiagnosis of toxocariasis, even in settings with higher resources [25]. Observed symptoms were often little specific and mimicked many other parasitic infections occasionally seen in travelers [26,27]. They were sometimes mild and self-limiting, not always triggering a complete etiological workup. Eosinophilia was rather high within the first weeks after symptom onset but tended to normalize quite rapidly. Laboratory investigations often detected evidence of other helminthic infection, with almost no possibility to discriminate between cross reaction, dual infection or remote exposure. Finally, we observed almost always excellent clinical and laboratory responses to albendazole, which is liberally used as empiric anti-helminthic treatment in travel medicine [28]. Clinicians often tend therefore to consider the specific diagnosis of toxocariasis as somehow difficult and of secondary importance. Morbidity of toxocariasis should however not be underestimated. In the present series, most patients experienced a protracted illness, also because diagnosis was often delayed. Admission was necessary for three patients (10%), including both neurological cases for whom the diagnosis was particularly challenging. The clinical features of transverse myelitis had indeed developed insidiously, with no concomitant systemic symptoms and no blood eosinophilia at diagnosis, but well laboratory findings indicating an eosinophilic meningitis and non-specific alterations at the Magnetic Resonance Imaging of the spine [29–32]. Both cases did not fully recover despite maximal anti-helminthic therapy and prolonged corticosteroids. For all other cases however, administration of a 5-day course of albendazole was, when tolerated, immediately beneficial, suggesting that the current recommended practice is adequate [8]. This was not the case for ivermectin treatment with which clinical failures were observed [33]. Of note, no clinical exacerbation was observed during anti-parasitic treatment in contrast with what often occurs in other helminthic infections with larval invasion such as acute schistosomiasis [26]. Finally, we did not observe any case of ocular toxocariasis, but such cases are often serologically negative (due to low parasite load) and usually diagnosed in specialized ophthalmologic clinics [34]. In conclusion, symptomatic and asymptomatic toxocariasis was sporadically diagnosed in international travelers attending our center and had clinical and laboratory features overlapping those of many other tropical infections. In the present series, morbidity was non negligible and occasionally severe. A standard 5-day course of albendazole provided substantial clinical benefit without evidence of clinical exacerbation. Research is needed to develop antigen-based tests that would better reflect the disease activity both for diagnostic and monitoring purposes in clinical care.

Toxocariasis is a zoonosis of worldwide distribution caused by dog (Toxocara canis) or cat (T. catis) roundworm that can be fully asymptomatic or may cause significant disease such as a the systemic syndrome called visceral larva migrans as well as neurological or eye manifestations. Toxocariasis prevails in tropical areas, but information about this disease in travelers and migrants is scarce. In this study, we describe in detail a case series of 28 international travelers, mostly adults, diagnosed with toxocariasis from 2000 to 2013 at the reference travel clinic of the Institute of Tropical Medicine of Antwerp, Belgium. We found this infection in all types of travelers returning from any part of the world. Clinical symptoms, when present, varied widely and an increase of the blood eosinophil count was almost always present. Morbidity was substantial and 2 patients had severe neurological complications. Diagnosis was difficult in travelers because the illness often resembled other tropical infections. Recovery was, however, complete, either spontaneously or with anti-parasitic drugs, except in both cases with neurological involvement. Toxocariasis is one of the numerous parasitic infections to consider in travelers returning from the tropics with any type of symptoms or with an increased blood eosinophil count.

Abstract Introduction Materials and Methods Results Discussion

2015

Toxocariasis Diagnosed in International Travelers at the Institute of Tropical Medicine, Antwerp, Belgium, from 2000 to 2013

Circadian clocks in eukaryotes rely on transcriptional feedback loops, in which clock genes repress their own transcription resulting in molecular oscillations with a period of ∼24 h. In Drosophila, the clock proteins Period (PER) and Timeless (TIM) operate in such a feedback loop, whereby they first accumulate in the cytoplasm of clock cells as a heterodimer. Nuclear translocation of the complex or the individual PER and TIM proteins is followed by repression of per and tim transcription, whereby PER seems to act as the prime repressor. We found that in addition to PER: TIM complexes, functional PER: PER homodimers exist in flies. Specific disruption of PER homodimers results in drastically impaired behavioral and molecular rhythmicity, pointing the biological importance of this clock protein complex. Analysis of PER subcellular distribution and repressor competence in the PER dimer mutant revealed defects in PER nuclear translocation and a disruption of rhythmic period transcription. The striking similarity of these phenotypes with that of reduced CKII activity suggests that the formation or function of the PER dimer is closely linked to this kinase. Our results confirm a previous structural model for PER and provide strong evidence that PER homodimers are important for circadian clock function. Circadian clocks likely evolved because they provide organisms with the advantage to anticipate changes of environmental conditions. Thanks to such clocks, metabolism, physiology, and behavior can be tuned to occur at advantageous times during the 24-h day [1]. Molecularly, circadian clocks are assembled by transcriptional feedback loops in which several clock gene products regulate their own and other clock genes' transcription [2]. In Drosophila, the first two clock genes identified were period and timeless, which are both required for the maintenance of circadian clock function (e. g. , [3,4]). The Period (PER) and Timeless (TIM) proteins are able to form a heterodimeric complex, which is important for stabilization of both proteins, for nuclear entry, and presumably also for the function of the PER: TIM complex as transcriptional repressor of their own expression [5–12]. More recently, PER has been shown to enter the nucleus and exhibit repressor activity independent of TIM in vitro and in vivo [13–20]. PER and TIM repress their own expression by binding to their transcriptional activators Clock (CLK) and Cycle (CYC), two b-HLH PAS domain-containing transcription factors that bind E-box sequences in the regulatory regions of per, tim, and other clock or clock-controlled genes [10,21]. Binding of PER: TIM to the CLK: CYC complex ultimately results in the release of CLK: CYC from the E-boxes of their target genes thereby deactivating them [10]. The kinase encoded by the double-time (dbt) gene can phosphorylate PER, which results in proteasomal degradation of cytoplasmic and nuclear PER in vivo [22–25]. DBT translocates to the nucleus in a complex with PER (or PER: TIM) [23], and recent work suggests that PER phosphorylation by DBT counteracts PER nuclear entry in vivo [20]. In addition PER may serve as a “bridge” to enable CLK phosphorylation by DBT, which may be a crucial event in inactivating CLK transcriptional activity in cell culture [26–29]. Importantly, the currently available in vivo data do not support this role for DBT. In flies, the lack of PER phosphorylation in a dbt mutant background in the absence of TIM is correlated with strong, DBT-independent PER repressor activity [18,20]. In addition to DBT, CKII has also been implicated in phosphorylating PER thereby enhancing PER nuclear entry and repressor activity in vivo [30–33] and in vitro [14]. De-phosphorylation of PER [34] and CLK [27] by the phosphatase PP2A counteracts DBT and CKII mediated phosphorylation, providing an additional level of nuclear translocation and repressor activity regulation. By an apparently independent mechanism, the bHLH-Orange domain transcription factor Clockwork Orange (CWO) also regulates E-box driven expression of clock genes (including per and tim) by directly binding to CLK target sequences [35–37]. Although it was initially postulated that CWO acts as a repressor, more recent work demonstrates that CWO has also activating properties [38]. In order for PER to exhibit its repressor function, be it direct via altering CLK conformation upon binding, or indirectly by bringing the kinases DBT and CKII into the proximity of CLK, PER needs to be present in the nucleus. Although it had originally been postulated that the PER: TIM interaction is required for nuclear translocation of both proteins, it seems now generally accepted that in flies PER and TIM can enter the nucleus separately [16,39]. These findings were further underscored by the discovery of the timblind mutation, which interferes with TIM but not with PER nuclear localization [40]. A Förster Resonance Energy Transfer (FRET) -based study performed with PER and TIM proteins in an embryonic Drosophila cell line (S2) revealed that PER and TIM form a complex immediately after their synthesis in the cytoplasm, but separate right before nuclear translocation and enter the nucleus independently [17]. Interestingly, cytoplasmic PERL: TIM complex formation is not delayed, but PERL did delay nuclear accumulation [17]. This suggests that events during or after PER: TIM formation are important for the correct timing of nuclear entry. These events most likely involve the reciprocal regulation of the phosphorylation status of PER by DBT and CKII, whereby CKII function seems crucially important for efficient nuclear localization of PER in wild-type flies [20,32,33,41]. In order to enter the nucleus in absence of TIM, PER needs somehow to be protected from DBT-induced degradation. One possible way to stabilize PER in the absence of TIM could be the formation of PER: PER homodimers, which could either form after the PER: TIM complexes dissolve, or co-exist with PER: TIM. The existence of such dimers has long been postulated and even been demonstrated in vitro and in vivo [5,42,43], although they were predicted to exist in very low concentrations [5]. More recently, the crystal structure of a PER fragment (amino acids 232–599) containing the two PAS domains (PAS-A and PAS-B) plus 75 additional C-terminal amino acids has been resolved [44]. It also revealed that this fragment can form a homodimer mediated by several intermolecular interactions between PAS-A, PAS-B, and an α-helix (αF) immediately C-terminal to the PAS domains (Figure 1A–1C). Importantly, one contact is made between Val243 in the PAS-A domain of molecule 1 (the site of the original perL mutation V243D; [45]) and residues Met560 and Met564 in the αF-helix of molecule 2 (Figure 1B and 1C; [44]). Val243 has previously been associated with mediating PER: PER, PER: TIM, as well as intramolecular PER interactions in vitro [42,43,46]. The long circadian period of perL flies was attributed to a delayed nuclear entry of the PERL protein, which can be observed in vivo and in vitro [17,47], suggesting that PER: PER and/or PER: TIM interactions regulate nuclear entry time. Yet, the functional significance of homodimer formation has so far only been tested by analyzing the V243D and M560D PER mutants in vitro [44]. Both amino acid replacements were predicted to weaken the PAS-A: αF interaction by introducing a negative charge into the hydrophobic interface and resulted in increased nuclear translocation and repressor activity of the mutated PER proteins in a cell culture transcription assay [44]. Although this indicated biological relevance for both the PER: PER dimer and the PAS-A: αF interaction, to date no supporting in vivo data exist. Here we show that by weakening the PAS-A: αF interaction via introducing a single amino-acid substitution in αF (M560D) we can drastically reduce PER: PER dimer formation in the fly without compromising the formation of PER: TIM complexes. Moreover, this reduction of homodimer formation coincides with severely impaired behavioral rhythmicity under free running conditions, indicating that the PER: PER dimer is important for clock function. Contrary to the in vitro results described above, our results indicate that PER: PER formation is necessary for efficient nuclear translocation of PER and subsequently for repressing CLK: CYC mediated transcriptional activation. Several reports indicated the existence of a PER: PER homodimer, although its relevance and biological function in vivo has not been revealed [5,44]. On the basis of the crystal structure determined for a PER: PER N-terminal fragment (Figure 1A–1C; [44]), we designed single amino acid substitutions aimed to destroy the major contact points between the two PER molecules in the context of the full length protein. The mutations were introduced in the per cDNA, which was modified to carry either an HA or c-myc tag at the C terminus. Expression of the constructs was facilitated by a 1. 3-kb fragment of per' s 5′-flanking region, a short hsp70 activation sequence, and a 2-kb per 3′-UTR region, known to participate in regulation of normal per expression (Materials and Methods; Figure 1D; [48,49]). All constructs were stably integrated into the genome by P-element transformation and analyzed in flies that carry the loss-of-function mutation per01 (Materials and Methods, [50]). Since per01 flies do not express endogenous PER protein, the only source of PER in the flies we generated, stems from the various transgenes. As controls, we first generated the wild-type versions of the tagged per constructs (Figure 1D) and tested the PER expression levels on western blots at ZT0 and ZT12 (reflecting times of high and low PER expression levels in wild-type flies, respectively). We found robust expression of both HA and c-myc tagged versions when we compared the transgenes to per+ control flies (Figure 2A). We next created a mutant PER version, in which the Trp at position 482 was exchanged to Glu, predicted to disrupt the two (reciprocal) interaction points between PAS-B and the PAS-A domain of the dimerizing molecule (Figure 1B–1D). Western-blot analysis of several independent W482E transgenic lines revealed that they accumulate only very little W482E PER protein compared to the per+ and wild-type transgenic controls (Figure 2A). We obtained the same results when the blots were incubated with anti-HA antibodies, indicating that the low anti-PER signals are not due to altered PER-immunoreactivity of the mutant proteins. Quantitative analysis of mRNA expression driven from the W482E construct indicates that reduced mRNA levels do not account for reduced protein levels (Figure S1). The most likely explanation for reduced W482E levels is therefore reduced protein stability (see Discussion). We next introduced a missense mutation at position Arg345 of PER located in the PAS-A domain. The R345E mutation we introduced is expected to disrupt the salt-bridge observed between Arg345 in PAS-A of molecule 1 and Glu566 in the αF-helix of molecule 2 (Figure 1B–1D); [44]). As in the case of W482E, flies expressing the R345E transgene accumulated only very little R345E PER compared to controls, indicating that this mutant protein is also unstable (Figure 2A and Figure S1, see Discussion). R345E and W482E carry a mutation in the PAS-A, and CLD domain, respectively, which have been described to be involved in binding to the clock protein TIM in cell culture [11,46]. Therefore, instead or in addition to disrupting PER: PER binding, the mutations could interfere with formation of the PER: TIM dimer. Since PER and TIM heterodimerization is thought to stabilize PER [5,7], this could explain the low levels of our mutant PER proteins. We therefore decided to generate a PER mutant that, based on the 3-D structure, should interfere only with the formation of PER: PER dimers and not with the PER: TIM interaction. Residue Met560 is situated in the αF-helix contacting Val243 in the PAS-A domain of the dimerizing molecule (Figure 1B–1D). Although Val243 has been implicated both in PER: PER and PER: TIM interactions (it is the site of the original perL mutation: V243D), no such interactions have been reported for any residues in the αF-helix or C-domain. The M560D mutant we generated indeed exhibited robust levels of mutant PER protein, indicating that stabilizing PER: TIM interactions are not disrupted by the mutant (Figure 2A). In agreement with the results just described, a PER W482E-M560D double mutant expressed a low level of mutated PER protein (Figure 2A and Figure S1). Since the M560D mutant exhibited wild-type levels of PER, we analyzed the temporal expression profile of this mutant PER protein. For this, we performed western blots with extracts prepared from flies at six different time points throughout the day, which were probed with anti-PER antibodies. We compared temporal expression between nontransgenic per+ control flies, PER wild-type transgenics, and the M560D mutants at six different time points throughout the day (Figure 2B). The nontransgenic per+ controls exhibited the typical robust PER oscillations in abundance and mobility (because of temporal regulated phosphorylation, [51]) (Figure 2B, left panels; e. g. , [52]). Similarly, the transgene encoded wild-type PER proteins exhibited daily abundance and mobility oscillations, although the mobility shifts appeared to be reduced (Figure 2B, upper right panel). The reduced mobility changes observed for the transgene-encoded wild-type proteins were apparent in several independent transgenic lines both containing the HA or c-myc tag (unpublished data). Since the transgenic proteins contain the entire PER open reading frame, the altered migration properties are likely caused by the different tags attached to the C termini of PER. Nevertheless the wild-type transgenic PER proteins underwent robust daily oscillations, which were absent or severely diminished from the M560D mutant proteins (Figure 2B and Figure S2A). The M560D protein was almost equally expressed during the 24-h day; and during all time points analyzed, fast and slow migrating mutant PER species were present. Quantification of the band intensities revealed that fast migrating, hypophosphorylated forms of PER are more abundant in the late night in the M560D mutant compared to the slower migrating phosphorylated forms (Figure S2B and Figure 2B; note the distinct fast migrating forms in the mutant at ZT16 and ZT20). Although as discussed above, the tags attached to the proteins likely contribute to this effect, this finding indicates that the M560D protein is less efficiently phosphorylated compared to the wild-type protein. In any case, a significant reduction in the overall PER abundance rhythm could be observed in all M560D transgenic lines analyzed, independent of the insertion site or type of tag (Figure 2B and Figure S2). If the M560D mutant indeed interferes with PER: PER homodimerization, our results imply that dimer formation is necessary for proper PER cycling. We therefore tested whether wild-type polypeptides form homodimers in the living fly, and if their formation may be compromised by the M560D mutation. Homodimer formation in flies was tested by performing co-immunoprecipitation (CoIP) experiments making use of the HA and c-myc tags attached to the wild-type and mutant PER proteins. per01 flies homozygous (on Chromosome 2) for a transgene encoding a HA-tagged PER and homozygous (on Chromosome 3) for a transgene encoding a c-myc-tagged PER were synchronized to a 12-h: 12-h light–dark (LD) cycle and collected at ZT20, a time where homodimer formation was previously reported [5]. Protein head extracts were subjected to CoIP, whereby the extracts were incubated with anti-c-myc coated sepharose beads and the western blots were incubated with anti-HA antibodies (Materials and Methods). Therefore, HA-tagged PER proteins can be detected on the western blot only when they had formed a homodimer with the c-myc-tagged PER molecules. In agreement with the earlier study [5], we were able to detect PER-HA, indicating PER: PER homodimer formation at ZT20 (Figure 3A, lane 3). To test if HA-tagged PER molecules are able to bind to anti-c-myc coated beads unspecifically, we subjected extract from flies expressing only the PER-HA proteins to the same CoIP. The lack of anti-HA signal (Figure 3A, lane 6) clearly shows that PER-HA can only be pulled down in the presence of PER-c-myc. It should be mentioned that we are only able to detect homodimers formed between HA and c-myc tagged molecules and not PER-HA: PER-HA and PER-c-myc: PER-c-myc dimers, which are likely to be formed at similar rates. Therefore, the overall levels of PER: PER homodimers are expected to be two times higher compared to those we detected in our CoIP experiments (there are two possible ways to form PER-HA: PER-c-myc dimers, but only one to form dimers between PER molecules carrying identical tags). The same experiment was then performed using double-homozygous per01 M560D-HA/M560D-c-myc flies. In contrast to the wild-type tagged proteins, M560D mutant PER proteins were not able to support homodimer formation, since little to no PER-HA signals could be detected (Figure 3B, lane 3). But given the fact that we can detect only 50% of the possible PER: PER dimers (see above) we cannot rule out that some homodimers form between the mutant PER M560D proteins. Given that M560D proteins are stable we assumed that they are still able to heterodimerize with the TIM protein (see above). To test this idea, we also incubated the western blots after CoIP with anti-TIM antibodies. Indeed, both the wild-type PER and the mutant M560D proteins strongly bound to TIM (Figure 3A and 3B, lanes 3). This therefore explains the stability of the mutant M560D protein. More important, since M560D-PER: TIM heterodimerization is not affected, the M560D mutant allows for specific functional analysis of PER homodimerization owing to absent or drastically reduced PER: PER formation. To determine if the PER homodimer fulfills biological function we analyzed locomotor activity rhythms of wild-type and mutant PER-encoding constructs in a per01 genetic background. Behavior was analyzed in 12-h: 12-h LD cycles and in constant darkness (DD) to assess effects on synchronization to LD cycles and on the internal clock. The M560D mutation is predicted to disrupt the interaction with Val243, the site of perL mutation, which is defective in temperature compensation [53]; therefore we tested LD and DD behavior at different constant temperatures of 18 °C, 25 °C, and 29 °C. As expected, the wild-type PER-encoding constructs were able to restore robust and largely temperature compensated behavioral rhythms in per01 flies, both in LD and DD conditions (Figure 4 and Table S1). In LD at 25 °C, per+ control and wild-type transgenic flies showed the characteristic anticipation of the D to L and L to D transitions in the mornings and evenings, respectively interspersed by prolonged periods of inactivity during the day (siesta) and night (Figure 4A, cf. [54]). Moreover, at cold (18 °C) and warm (29 °C) temperatures, flies moved their activity peaks to occur mainly in the light and dark portions of the day, respectively (cf. [55]). Next, we analyzed behavior of the M560D mutant flies. Although they behaved similarly to the controls overall, behavioral anticipation of the environmental changes was less pronounced. For example at 25 °C, when wild-type control flies where relatively inactive during the siesta, M560D mutant flies showed increased activity levels (Figure 4A). On the other hand, M560D mutants were still able to move their periods of main activity towards the light or dark phase in cold and warm temperatures, respectively (Figure 4A). In DD and 25 °C about 80% of the per01 flies transformed with the wild-type PER-HA or PER-c-myc encoding constructs exhibited robust circadian rhythms with periods of ∼23 h (Figure 4B and 4C). Moreover, the number of rhythmic animals and the period length did not vary significantly at 18 °C and 29 °C, indicating that the PER-HA and PER-c-myc fusion proteins are able to replace endogenous PER protein (Figure 4B and 4C and Table S1). In contrast, depending on the temperature, only between 30% and 60% (40% at 25 °C) of the M560D mutant flies were rhythmic in DD, indicating that the circadian clock is drastically impaired in these flies (Figure 4C and Table S1). Nevertheless, the period length did not vary much between the different temperatures, indicating that temperature compensation is not affected (Table S1). As was observed for flies from the two control strains, overall rhythmicity for M560D flies was correlated with an increase in temperature (Figure 4C, left panel). At each of the different temperatures tested the M560D mutants showed slightly longer periods compared to transgenic controls (∼1 h longer at each temperature, Figure 4C, right panel; Table S1). This period lengthening could be due to the specific mutation, since in the homodimer M560D is predicted to disrupt the interaction with the perL site Val243. Alternatively, slightly reduced overall PER levels in the M560D transgenics compared to the wild-type transgenics could account for the period lengthening, because per function is dosage sensitive (e. g. , females carrying only one copy of the X-linked per+ allele have 25-h periods, [50]). In any case, together with the biochemical results described, the behavioral defects observed in the M560D mutants clearly point to a crucial function of the PER: PER dimer within the circadian clock. We also analyzed the behavioral rhythms of the other mutants we generated (W482E, R345E, and the double mutant W482E M560D). Neither of the mutant proteins was able to restore rhythmic behavior in per01 in LD or DD conditions in more than 20% of the flies (Figure 4C and Table S1), consistent with the low mutant PER protein levels present in these flies. The few rhythmic flies exhibited widely variable periods, and temperature compensation was severely compromised (Figure 4C and Table S1). Although this effect can be due to many different reasons, it may indicate that PER: TIM interactions are more important for proper temperature compensation compared to PER: PER (temperature compensation in M560D flies is normal, Figure 4C). So far our data indicate that disruption of the PER: PER dimer interferes with circadian clock function and results in abnormal behavioral rhythms (Figures 2B, 3, and 4). But what specific clock process (es) would be carried out by PER: PER dimers? One possibility we tested is that PER' s function as a repressor of CLK/CYC activated per and tim transcription involves the PER: PER dimer. In order to test this we first analyzed if the abundance of PER: PER dimers varies throughout the circadian day. If the PER: PER homodimer is important for PER' s function as a repressor, one would expect to see a higher accumulation of the dimer at times when PER is nuclear, which corresponds roughly with the second half of the night until the early day (ZT18 to ZT4) [16,47]. We therefore performed CoIP experiments with our PER-HA and PER-c-myc expressing per01 flies collected at ZT16 (cytoplasmic) and ZT20 (nuclear). We also included an early morning time point (ZT2), in which PER is in the nucleus and TIM is mostly degraded due to the presence of light [5]. During this time PER has been shown to be the main repressor of CLK/CYC induced transcription [13], and perhaps PER: PER homodimer formation is crucial for this TIM-independent function of PER. Indeed, the temporal analysis of PER: PER abundance revealed that peak amounts are reached at ZT2, suggesting that the PER: PER dimer is an active repressor unit (Figure 5A). Consistent with this idea, PER: PER homodimer formation in the M560D mutant is disrupted at the two nuclear time points (ZT20 and ZT2) investigated (Figures 3B and 5A, lower panel). But also at times when PER is cytoplasmic (ZT16), substantial amounts of PER: PER complexes exist, similar to what is observed when PER is nuclear (ZT20), suggesting a role for PER: PER during the accumulation phase of this clock protein (Figure 5A). These results suggest that the homodimer acts as repressor. In order to further test this hypothesis we wondered if the M560D mutant would decrease PER' s repressive activity. For this, we made use of a period-luciferase (per-luc) transgenic reporter strain that reflects per transcription in vivo [56,57]. In the plo transgene the per promoter is directly fused to the firefly luciferase cDNA and per-luc expression in individual adult flies can be monitored with an automated bioluminescence counter (e. g. , [57]). As expected, the wild-type PER constructs restored robust transcriptional rhythms when introduced into per01 plo flies (Figure 5B and Table S2). Interestingly, when the same reporter flies expressed the M560D PER mutant, transcriptional rhythms were abolished in the majority of the flies, or they were of significantly reduced amplitude (Figure 5B and 5C and Table S2). Although the overall mean levels of plo expression were similar between the wild-type and mutant flies, the latter did not reach the trough levels of expression observed in the wild-type PER transformants; or if they did, only for a very limited amount of time (Figure 5B, arrows). We conclude that repression by the mutant monomeric PER protein is less efficient compared to that of the dimeric wild-type protein, resulting in a breakdown of per transcriptional rhythms. Faulty nuclear translocation could be one possibility why the dimerization defective M560D mutant exhibits reduced repressor activity. Therefore we determined the subcellular distribution of PER within the clock neurons of the adult brain at different times within a circadian cycle. Rhythmic expression and proper cytoplasmic/nuclear shuttling of clock proteins (including PER and TIM) in the lateral clock neurons (LNs) is required for proper clock function and control of rhythmic locomotor activity (e. g. , [58]). We stained brains prepared at ZT16 (PER cytoplasmic), ZT20 and ZT2 (both PER nuclear) with anti-PER and anti-PDF (as a marker for cytoplasmic staining in the LNv' s; the more ventrally located subgroup of the LNs [58]). In nontransgenic per+ control flies (y w) we observed the characteristic cytoplasmic and cytoplasmic plus nuclear staining at ZT16 and predominantly nuclear staining at ZT20 and ZT2 in all clock-neuronal cell types (Figure 6A, left row; Figure 6B). The per01 wild-type transgenics showed a very similar pattern, although the anti-PER signals appeared overall weaker (Figure 6A, middle row; Figure 6B). This is in agreement with the robust behavioral rescue of per01 mediated by these transgenes (Figure 4 and Table S1). When we performed the same stainings in the per01 M560D mutant flies, we detected a significant reduction of the total numbers of LNs expressing PER (Figure 6A and 6B, and Figure S3). In the PER positive cells at ZT16 a significant (p < 0. 05) reduction of cytoplasmic or nuclear/cytoplasmic signals was visible in the s-LNv and l-LNv, respectively (Figure 6A and 6B, and Figure S4). At ZT20 a reduction of cells with nuclear and nuclear/cytoplasmic signals was observed for both cell types (Figure 6A and 6B), but the difference to the wild-type transgenics was significant only for the l-LNv (Figure S4). Importantly, some mutant s-LNv cells exhibited clear nuclear PER signals (Figure 6A and 6B), indicating that the results can not simply be the cause of overall lower PER levels in the mutant. At ZT2, when nuclear PER accumulation is maximal in the nontransgenic and transgenic controls (Figure 6A and 6B), for both cell types we observed significantly fewer cells with nuclear signals, paralleled by an increased number of cells with weak nuclear and cytoplasmic signals (Figure 6A and 6B, and Figure S4). This clearly points to impaired nuclear localization efficiency in the M560D mutant flies in both s-LNv and l-LNv. This abnormal nuclear-cytoplasmic distribution is in good agreement with the poor behavioral rescue and with the reduced repressor activity, mediated by this mutant protein (Figures 4 and 5B, and Table S1). These results further point to the importance for PER: PER homodimer formation for circadian clock function and to a role for the dimer in PER nuclear localization and transcriptional repression. Our in vivo findings are supported by biochemical studies analyzing homodimer-formation of the N-terminal crystallized PER: PER fragment (amino acids 232–599). Although the M560D mutant in the context of this fragment runs as a dimer in gel filtration experiments, the affinity of the dimer is significantly reduced by the mutation (see accompanying report). Furthermore, the V243D mutant (perL) version of this PER fragment and a fragment entirely lacking the αF-helix behave as monomers in gel filtration, demonstrating the importance of the PAS-A-αF interface (including Val243 and Met560) for dimerization in solution [44]. Unlike our in vivo results, when the M560D mutant in the context of the whole PER protein was expressed in S2 cells, it efficiently entered the nucleus and also acted as a potent repressor [44]. This discrepancy is likely linked to the fact that events in S2 cells not necessarily reflect the in vivo situation (for example TIM was not co-expressed in the study just cited) as was observed repeatedly in the past (e. g. , cytoplasmic localization of PERΔ in S2 cells versus nuclear localization in flies [28,29]) and underscores the importance of in vivo studies. Except for the M560D mutation all other mutant proteins we analyzed were unstable in flies. The W482E mutation (Figure 1B and 1C) is predicted to disrupt dimer formation at two symmetrical positions between the Trp482 located at the tip of the βD′-βE′ loop in PAS-B and the hydrophobic pocket formed by the βA, αB, and αC strands and helices of the other PER molecule [44]. As expected from this disruption of a prominent dual interaction point, the W482E weakens the dimer in the context of the PER (232–599) PAS domain fragment (see accompanying Research Article). Moreover, and similar to M560D, in S2 cells a W482A mutation shows enhanced repression compared to the wild-type fragment in the context of full length PER [44]. Since the purified PAS domain fragment carrying the W482E mutation is stable and properly folded (see accompanying Research Article), the instability of the W482E mutation in flies suggests that in vivo additional factors contribute to PER stability. One important feature missing from the stable PAS domain fragments, is a newly discovered N-terminal interaction domain with SLIMB, which is required for efficient degradation of PER [24]. The same applies for the R345E mutant, which is expected to disrupt a salt bridge between Arg345 on the βD strand of molecule 1 and Glu566 located on the αF-helix of molecule 2 (Figure 1B, 1C). Most likely the TIM protein and phosphatase activity supply this additional stability in vivo as discussed above. All mutants analyzed in the current study (except M560D), map to the PAS-A (R345E) or the PAS-B (W482E) domains, which have been implicated in directly mediating the PER: TIM interaction ([11,46], also see accompanying report). The most simple explanation for the observed instability of PER mutants containing amino acid changes in one of the PAS domains is therefore that they also interfere with the PER: TIM interaction (see accompanying Research Article). If true, mutants mapping to the αF helix and predicted to weaken dimer formation (like M560D) should result in stable proteins that are still able to interact with TIM. Such candidates include Met564 and Glu566, predicted to weaken the interaction with Val243 and Arg345, respectively [44]. The perL mutation (V243D) lengthens the free-running period dramatically [50] and also compromises temperature compensation [53]. Molecularly, perL leads to a temperature-sensitive delayed nuclear entry of PERL, which can account for the long behavioral period and loss of temperature compensation in this mutant [47]. Structural and in vitro studies [42–44] suggested an important role for Val243 in PER homodimerization. In agreement with these studies, we revealed that the contact amino acid of Val243 in the αF helix of the partner PER molecule (Met560) is also crucial for homodimerization in the context of the full length PER protein in vivo. But in contrast to the original and in vitro mutagenized perL mutants [42,47,53], our M560D mutant flies only exhibit a slight (1-h) increase in period length and no loss of temperature compensation (Figure 4C and Table S1). Therefore we believe that the period and temperature compensation defects of perL mutants are due to a faulty interaction with TIM as originally proposed by Gekakis et al. [46]. Indeed, we observed normal interactions between M560D PER and TIM, further supporting this hypothesis. It follows, that whereas the Val243 residue of PER is involved in mediating both PER: TIM and PER: PER interactions, the Met560 residue mediates only (or mainly) homodimerization. Further support for the importance of the Val243 residue for PER: TIM interactions came from a genetic screen, in which a “suppressor of perL” mutation (timSL), which ameliorated both phenotypes (period length and temperature compensation), was isolated and found to map to the timeless gene [59,60]. Also, cryb suppresses the temperature compensation phenotype of perL, indicating that the residue mutated in timSL (Thr494) interferes with the CRY: TIM interaction, and that the V243D (perL) induced temperature compensation defect is caused by increased PERL: TIM: CRY interactions [61]. Similar to the model original proposed by Rosbash and colleagues [42], it is therefore possible that the PAS-A: αF interaction between two PER molecules competes with a heterologous PAS-A: TIM interaction, which is likely also influenced by CRY. It has been shown in vitro that, after formation of PER: TIM complexes in the cytoplasm of S2 cells, these dimers dissolve, and both PER and TIM enter the nucleus independently [17]. Also Shafer et al. [16,39] demonstrated advanced nuclear entry of PER (without TIM) in vivo. Therefore, it is possible that PER: PER formation is necessary for, or promotes, nuclear entry after the PER: TIM dimers dissolve. What could be the possible signal for this event? Two kinases have been implicated in nuclear localization of PER. DBT promotes cytoplasmic localization of PER [18,20,41], perhaps via the DBT-dependent phosphorylation of a PER NLS sequence [24]. CKII supports PER nuclear translocation [33], perhaps because PER (or TIM [62]) phosphorylation by CKII serves as a signal for PER: TIM break down, whereupon PER: PER dimers form that could then enter the nucleus. Alternatively, a different signal could promote the PER: TIM break down, and the PER: PER dimer could serve as a prime substrate for CKII, followed by nuclear translocation. This hypothesis is supported by the fact that CKII mutants and M560D PER mutants share similar phenotypes. Both result in period lengthening without compromising temperature compensation, although CKII mutants have a much more drastic effect on clock speed compared to M560D (Figure 4 and Table S1) [30–33]. Also, in both cases PER oscillations (as determined by western blots) are affected and show a reduction of PER phosphorylation (Figure 2B and Figure S2) [30–33]. Finally nuclear translocation is similarly impaired by M560D and CKII mutants (Figure 6 and Figure S4) [30–33]. Probably as a result of this faulty nuclear localization both mutants exhibit reduced repressor activity of PER (Figure 5B, 5C; [33]), which is in agreement with the enhancing effect of CKII on PER repressor activity observed in cell culture [14]. Both CKII [33] and M560D mutants do not completely block PER nuclear entry, indicating that other factors also support PER nuclear translocation, or that to some extent PER monomers or PER: TIM heterodimers are able to enter the nucleus. It is likely that M560D does not completely block homodimer formation, as indicated by the weak band we sometimes see in the M560D CoIP experiments (Figure 3B). We do see nuclear M560D PER, mainly within the small LNv cells, but much less frequently in the large LNv' s. Perhaps this difference is simply due to the smaller cytoplasmic volume in the s-LNv compared to the l-LNv, which could enhance homodimer formation because of a higher local concentration of monomeric proteins. These homodimers would then be able to enter the nucleus to repress CLK. The occasional nuclear staining in the s-LNv cells also explains why not all the M560D mutant flies exhibit arrhythmic behavior in constant darkness. The s-LNv' s are crucial for maintaining sustained locomotor rhythms (e. g. , [63]), and it has been shown that only a few of these cells are sufficient to drive behavioral rhythms if neuritis projecting from them terminate within the dorsal brain [64]. The small fraction of nuclear PER homodimers presumably also explains that some repression is still maintained in the M560D mutants (Figure 5B). Alternatively, homodimer formation may not be required for repression, which could be mediated by PER: TIM heterodimers during the late night [10], followed by PER repression in the early morning [13]. Given that the M560D mutation leads to less efficient transcriptional repression (Figure 5B), and robust levels of wild-type homodimers are present in the early morning (Figure 5A), we favor the idea that PER does form nuclear homodimers to mediate repression at this time of day. In support of this hypothesis, a mutant PER protein lacking a rather large piece of the C-domain (ΔC2, missing amino acids 512–568 and therefore the complete αF-helix; Figure 1A) including M560D, shows drastically reduced repressor activity in vivo [65]. In contrast to M560D PER, the ΔC2 protein is constitutively nuclear (unpublished data in [65]), suggesting that the predicted inability of this protein to form homodimers is the reason for lacking repressor activity. The constant nuclear localization of ΔC2 PER in contrast to M560D is possibly due to a potential nuclear export signal located in the deleted region of ΔC2 (Figure 1A [66,67]). The ΔC2 PER is also stable in constant light [65], indicating that, in addition to PER: PER, the previously documented PER: CRY interaction in yeast [68], is also mediated by this domain in vivo (although, see [69]). Alternatively, the large 56 amino acid deletion in ΔC2 PER could result in a structural change of the PER: TIM complex, which may interfere with the CRY: TIM interaction. DBT promotes PER phosphorylation and turnover, when PER is free from TIM in the cytoplasm and the nucleus [22,23], presumably by creating an optimized binding site for the F-box protein SLIMB [24]. PER can be stabilized either by binding to TIM [7], by preventing progressive DBT-dependent phosphorylation [70] or by phosphatase activity [24,34,71]. Perhaps dimer formation also contributes towards PER stability, although we think this is unlikely because we do observe robust PER levels in the M560D mutants (Figures 2,3B, and 5A). To obtain a definite answer, mutations that completely disrupt dimer formation without compromising other PER protein interactions need to be generated (see “Residues Important for PER: PER and PER: TIM Interactions” in “Discussion” below). In principle it is possible that PER: PER complexes bind to DBT, since an important DBT-binding domain has been mapped to a small region (27 to 54 amino acids, depending on the study) located C-terminal of the PAS and αF interaction surfaces (Figure 1A, [28,29]). The DBT-binding domain also overlaps with the previously identified “CLK CYC Inhibition Domain” (CCID), which presumably explains the reduced in vitro repressor activity of PER molecules lacking the CCID [15]. Therefore DBT could enter the nucleus in a complex with PER: PER and mediate transcriptional repression via phosphorylation of CLK, as recently proposed [26–29]. Importantly, flies expressing a PER protein lacking the DBT-interaction domain (PERΔ) accumulate constantly hypophosphorylated forms of nuclear PER [28,29], reiterating that DBT supports cytoplasmic localization of PER and is not required for nuclear entry in vivo [20,41]. Although it was reported that PERΔ can act as potent repressor in cultured “nonclock” cells [29], this deficient protein shows drastically reduced repressor activity in vivo, supporting a model in which action of DBT within the nucleus contributes to repression of CLK [28]. Our findings suggest that, at least in the early morning, after light-dependent clearance of TIM, the PER-PER homodimer is building the proposed “bridge” for CLK phosphorylation by DBT [26,28]. In this model, disruption of the PER: PER dimer mainly affects nuclear accumulation of PER (perhaps because of an impaired interaction with CKII as discussed above); but since DBT interaction is most likely not disrupted, mutant “escaper” complexes that are able to enter the nucleus can still mediate repression. However, it has been shown that in the absence of TIM and DBT activity, PER exhibits nuclear localization, correlated with strong repression of CLK [18,20]. This indicates that perhaps other kinases (e. g. , CKII) phosphorylate CLK, or that PER-independent mechanisms, in particular repression by CWO, are able to compensate for the lack of DBT function [35–37]. CRY has also been shown to act as a repressor of CLK in peripheral clock cells in vivo [72]. Although this repression was shown to require both PER and CRY, the authors conclude that both proteins regulate distinct parts of the cycle (i. e. , PER still acts as a repressor after CRY has been degraded by light [13]). Although it is tempting to speculate that PER and CRY form a heterodimer to repress CLK, such a complex has so far only been demonstrated to exist in yeast [68], and PER: CRY interactions in vivo are likely mediated via TIM [69]. In mammals PER proteins (mPER1–3) have been shown to interact with Cryptochromes (mCRY1 and mCRY2), and this interaction is thought to mediate nuclear translocation of both proteins [3,73]. In addition, formation of heterodimers between the different mPER proteins in vitro and within the SCN has been demonstrated [73,74], but their biological function remains elusive. Furthermore, the accompanying report indicates the existence of mPER homodimers (i. e. , mPER2: mPER2), and future work will show if they have a similarly important role within the circadian clock, as do the fly PER homodimers described here. We provide strong evidence for an important function of the PER: PER homodimer in the Drosophila circadian clock. The mutant M560D selectively disrupts PER: PER dimer formation in vivo resulting in a significant reduction of molecular and behavioral rhythmicity. The faulty nuclear localization of M560D proteins in circadian clock neurons strongly support a role for the PER: PER homodimer in nuclear translocation, which is probably closely linked to CKII function. Consistent with these findings, M560D shows a reduced ability to repress per transcription in an in vivo transcription assay, resulting in largely abolished per transcriptional rhythms. The per constructs are based on the −1313–34-hs-per, a pP{CaSpeR-4} transfection vector with deleted HindIII, PstI, SalI, XhoI, and HpaI restriction sites containing the complete per cDNA, a −1313–34 5′-flanking region cloned from the per locus, the hsp70 basal promoter, and a 2. 1-kb per downstream sequence [48,49]. For addition of the HA- and the c-myc-tag, a 4. 2-kb XhoI/HindIII fragment (hsp70 promoter and per cDNA) was subcloned into a pBluescript KS+ vector, lacking the BamHI restriction site (further on called pKS-per), followed by addition of an AatII restriction site in front of the tag stop codon, using the QuikChange II XL Site-Directed Mutagenesis Kit (Stratagene) in combination with the oligo nucleotides Aat-S (CCAGACACAGCACGGGGACGTCTAGTAGCCACACCCGC) and Aat-AS (GCGGGTGTGGCTACTAGACGTCCCCGTGCTGTGTCTGG). Annealed and 3′ phosphorylated oligo nucleotides myc-S/myc-AS (TGAGCAGAAGCTGATCAGCGAGGAGGATCTGTACGT, AGAGATCCTCCTCGCTGATCAGCTTCTGCTCAACGT) and HA-S/HA-AS (TTACCCCTACGATGTGCCCGATTACGCCTACGT, AGGCGTAATCGGGCACATCGTAGGGGTAAACGT) were then ligated directly into the AatII restriction site, respectively. After addition of the HindIII/EcoRI fragment of the original −1313–34-hs-per (per downstream sequence) to both of the vectors, the BamHI/XbaI fragments, containing the C-terminal part of the per cDNA and the 3′ sequences, were exchanged in the −1313–34-hs-per, creating the constructs per-HA and per-c-myc. Point mutations leading to the exchanges W482E and R345E in the protein sequence were introduced in pKS-per using W482E-S/W482E-AS (AGCTTCGTCAATCCAGAGTCCCGCAAGCTGG, CCAGCTTGCGGGACTCTGGATTGACGAAGCT) and R345E-S/R345E-AS (CCTGGGGCTCACCTTCGAGGAGGCTCCGGAGGAG, CTCCTCCGGAGCCTCCTCGAAGGTGAGCCCCAGG) oligonucleotides, respectively. The XhoI/BamHI fragments (containing the hsp70 promoter and N-terminal part of the per cDNA) of the two mutated pKS-per vectors were then exchanged with the respective wild-type fragment in per-HA and per-c-myc resulting in per-W482E-HA, per-W482E-c-myc, per-R345E-HA, and per-R345E-c-myc. Similarly, the SanDI/BamHI fragment of pAc5. 1-V5/His-dPer-M560D [44] was introduced in per-HA and per-c-myc to generate per-M560D-HA and per-M560D-c-myc. To clone the double mutant constructs per-W482E-M560D-HA and per-W482E-M560D-c-myc the wild-type XhoI/SanDI fragment in the two per-M560D vectors was replaced with the same fragment carrying the W482E mutation. All constructs were verified by DNA sequencing. As nontransgenic control flies the wild-type strain CantonS, y Df (1) w (y w), and w1118 flies were used [75]. The original per01 allele [50] was also in a y w genetic background to facilitate transgene detection and to increase luciferase signals. P-element transformation was performed using standard techniques (e. g. , [76]) by injecting wild-type and mutant constructs into y w/y w (or y w/Y); KiΔ2–3/+ embryos, whereby the Δ2–3 transposon served as transposase source [77]. Go males were then crossed to y per01 w; Bl/In (2LR) O, Cy (CyO) virgins and the F1 was screened for orange-eyed males. y per01 w; p[w+] males with either Bl or CyO were then backcrossed to y per01 w; Bl/CyO virgins. If the transgene did not map to Chromosome 2, y per01 w; p[w+] males were subsequently crossed to a y per01 w strain containing a dominant marker and balancer chromosome for Chromosome 3. At least five independent transgenic lines were established for each construct, and results were confirmed with at least three independent lines. For per-M560D-c-myc we originally isolated only one homozygous lethal line mapping to Chromosome 3 (10-2-1) and one homozygous viable line mapping to Chromosome 4 (10-2-2) (see Table S1). We therefore mobilized the P-element in line 10-2-2 by backcrossing to the homozygous KiΔ2–3 strain and isolating homozygous viable inserts mapping to Chromosome 2 using standard crossings [77]. One of these lines (per-M560D-c-myc: 10-2-2J5) was used to create a double-homozygous stock with per-M560D-HA: 9–8 for CoIP experiments using standard crosses and balancer chromosomes. To create the double-homozygous flies expressing wild-type per-c-myc and per-HA constructs lines 1-5-2 and 2-2-2 were used. Two- to 3-d-old individual adult male flies were loaded in small glass tube sealed at one end with food (5% sucrose, 2% agar) and closed at the other end by cotton. The locomotor activity is detected by an automated infrared beam monitoring system (Trikinetics) for 4–7 d in a 12-h: 12-h LD cycle and then in DD for another 7 d. Daily average histograms and actograms were plotted using the fly toolbox and MATLab software [78]. The free-running period was calculated using the Autocorrelation function. In this study all flies with an Rhythmicity Statistics (RS) value >1 were considered as rhythmic (see [78,79] for how this cut-off was determined). Anti-PER antibody stainings were performed as previously described [80]. Prior to collection at ZT16, ZT20 and ZT2, the flies from different strains were entrained for at least 2 d under 12-h: 12-h LD conditions. Whole-mounted brains were dissected and collected in Ringer solution before being fixed in 4% paraformaldehyde at 4 °C overnight. After fixation, the samples were washed ten times with 0. 1 M phosphate buffer (pH 7. 4) and three times in PBS with 1% Triton X-100 (PBS-T) at room temperature (RT). The brains were then blocked with 10% goat serum in PBS-T for 2 h in RT and stained with pre-absorbed polyclonal rabbit anti-PER in PBS-T at 1: 1,000 dilution. After washing three times by PBS-T, the samples were incubated at 4 °C overnight with goat-anti-rabbit antibody conjugated with fluorophore, AlexaFluor 488 nm (Molecular Probes) diluted 1: 300 in PBS-T. For double-labeling, samples were washed with PBS-T and incubated with blocking serum and polyclonal rabbit anti-PDF antibody diluted 1: 1,000 in PBS-T at 4 °C overnight. The samples were then treated with goat-anti-rabbit antibody conjugated with fluorophore, AlexaFluor 594 nm (Molecular Probes) diluted 1: 300 in PBS-T at 4 °C overnight. Brains were washed three times in PBS-T and water before being mounted in Vectashield. Samples were stored at 4 °C until examination under a LSM-510 META confocal microscope (Zeiss). PDF signals in the LNvs were used as cytoplasmic marker. Yellow or orange staining of outside the nucleus caused by co-expression of PDF and PER (green) was scored as “cytoplasmic PER” (C). Green signals in the centre of LNvs were scored as “nuclear PER” (N), and neurons with yellow in periphery and green in the centre as “nuclear and cytoplasmic” (N/C). See legends of Figures S3 and S4 for further details. CoIPs were performed as described [69]. Briefly, adult flies from the HA- and c-myc-tagged strains were entrained to a 12-h: 12-h LD cycle for 2 d. At ZT16, ZT20 and ZT2,6 ml of flies were collected in liquid nitrogen and frozen in −80 °C until homogenization. Fly heads were separated by repeated vortexing/cooling in liquid nitrogen. 400 μl of fly heads were then isolated using a 0. 45-mm/0. 14-mm prechilled metal mesh. Heads were then homogenized in 400-μl of Extraction Buffer (20 mM Hepes [pH 7. 5], 100 mM KCl, 1 mM dithiothreitol, 5% glycerol, 0. 05% Nonidet P40,1× Complete Protease Inhibitor [Roche]). 20 μl of head extract were boiled with 5 μl of 5× SDS loading buffer as input control. Protein G Sepharose fast flow beads (Amersham) were coated with anti-MYC antibody (5 μl anti-MYC antibody [Covance Inc. ] + 20 μl beads/sample in 1 ml extraction buffer, 4 °C for 1 h) and incubated with the head extracts for 16 h at 4 °C. Beads were spun down by centrifugation, and 20 μl of supernatant were boiled with 5 μl of 5× SDS loading buffer as supernatant control. The pulled-down beads were washed three times with 750 μl Extraction Buffer before being resuspended in 30 μl 1× SDS loading buffer for western blot. Flies of the indicated genotypes were first kept in LD cycles for at least 3 d and collected on dry ice during the indicated ZT in LD. Preparations of head extracts and protein blots were performed as described [81]. Twenty-five fly heads for each genotype/time point were collected and homogenized with 40 μl of Extraction buffer (20 mM HEPES [pH 7. 5], 100 mM KCl, 5% glycerin, 10 mM EDTA, 0. 1% Triton-X 100,20 mM β-glycerophosphat, 0. 1 mM Na3VO4,1× Complete Protease Inhibitor [Roche]) and centrifuged at 4 °C at 13,000 rpm. The supernatants were transferred and boiled with 1 × SDS loading buffer for 5 min before loading and running on 4. 5%/6. 0% SDS-PAGE overnight at 55–70 V. CoIP samples were separated on the same gels, and proteins were blotted to Nitrocellulose at 500 mA for 1 h with a Semi-dry electro blotting unit (Fisherbrand) according to the manual. After blotting, the nitrocellulose membranes were blocked with 5% nonfat milk in TBS-T at room temperature for 1 h. First antibodies (5% nonfat milk/TBS-T) were diluted and incubated at 4 °C overnight, followed by secondary antibody (5% nonfat milk/TBS-T) incubation at room temperature for 2 h. Dilution of first antibodies were 1: 10,000 for rabbit anti-PER, 1: 2,000 for rat anti-TIM, and 1: 1,000 for mouse anti HA-tag (MMS-101p, Covance). The dilutions of HRP-conjugated secondary antibodies were 1: 100,000 for goat-anti-rabbit IgG, 1: 25,000 for goat anti rat IgG, and 1: 5,000 for goat-anti-mouse IgG. ECL substrates were used, and X-ray films were developed according to manufacturer' s instructions. Bioluminescence assays were performed as previously described [57]. Briefly, 2–3-d-old individual male flies carrying the plo per-luciferase transgene were ether-anesthetized and loaded in a 96-well micro titer plate in which each of the well contains 100 μl of 5% sucrose, 1% agar, and 15 mM luciferin. Flies were measured in a Packard Topcount Multiplate Scintillation Counter at 25 °C and 12-h: 12-h LD cycles for 7 d. Data were plotted using BRASS (biological rhythms analysis software system) Version 2. 1. 3. [82]. Period values and rhythm significance (rel-amp error) were calculated using FFT-NLLS analysis as described in [83]. In the current analysis flies with relative-amplitude errors (rel-amp) <0. 7 and period values of 24 ± 2 h were considered as rhythmic. Rel-amp errors <0. 7 indicate that the rhythms in the bioluminescence data determined by the FFT-NLLS analysis are due to rhythmic gene expression with 95% confidence [57].

The current models of circadian clocks in flies and mammals involve the formation of complexes between clock proteins in the cytoplasm. These complexes are usually heterodimers (that is, made up of two different clock proteins) and appear to enter the nucleus at certain times of the circadian day in order to shut down their own gene expression by deactivating specific transcription factors. After progressive phosphorylation the repressor proteins eventually are degraded so that a new cycle of transcription can begin. Here we present evidence that in addition to heterodimeric complexes, the clock protein PERIOD (PER) also forms homodimers (pairs of identical proteins). Based on a structural model a PER mutant was designed, which is not able to form homodimers but can still bind to its partner TIMELESS (TIM). Flies expressing this mutant PER protein show abnormal clock function in regard to PER nuclear translocation, repressor activity, and behavioral rhythms. The circadian clock model in flies therefore needs to be extended by adding the PER: PER homodimer as a functional unit. Recent structural studies with mammalian PER proteins suggest that homodimers between clock proteins are an important general feature of eukaryotic clocks.

Abstract Introduction Results Discussion Materials and Methods

cell biology neuroscience molecular biology

2009

A Role for the PERIOD:PERIOD Homodimer in the Drosophila Circadian Clock

Two insults often underlie a variety of eye diseases including glaucoma, optic atrophy, and retinal degeneration—defects in mitochondrial function and aberrant Rhodopsin trafficking. Although mitochondrial defects are often associated with oxidative stress, they have not been linked to Rhodopsin trafficking. In an unbiased forward genetic screen designed to isolate mutations that cause photoreceptor degeneration, we identified mutations in a nuclear-encoded mitochondrial gene, ppr, a homolog of human LRPPRC. We found that ppr is required for protection against light-induced degeneration. Its function is essential to maintain membrane depolarization of the photoreceptors upon repetitive light exposure, and an impaired phototransduction cascade in ppr mutants results in excessive Rhodopsin1 endocytosis. Moreover, loss of ppr results in a reduction in mitochondrial RNAs, reduced electron transport chain activity, and reduced ATP levels. Oxidative stress, however, is not induced. We propose that the reduced ATP level in ppr mutants underlies the phototransduction defect, leading to increased Rhodopsin1 endocytosis during light exposure, causing photoreceptor degeneration independent of oxidative stress. This hypothesis is bolstered by characterization of two other genes isolated in the screen, pyruvate dehydrogenase and citrate synthase. Their loss also causes a light-induced degeneration, excessive Rhodopsin1 endocytosis and reduced ATP without concurrent oxidative stress, unlike many other mutations in mitochondrial genes that are associated with elevated oxidative stress and light-independent photoreceptor demise. The causes of progressive dysfunction or death of photoreceptors (PRs) is genetically heterogeneous in humans [1]. PR degeneration is a complex process influenced by numerous genes and environmental factors. Although prolonged exposure to sunlight is one of the major causes of retinal degeneration, more than 200 genes have been associated with retinal diseases in humans [2,3]. Genes associated with retinal diseases affect a variety of cellular processes including phototransduction, cellular stress, metabolism, catabolism, and mitochondrial function [1,3, 4]. PR activity is a highly energy-dependent process [5,6], and mitochondrial dysfunction has been implicated in glaucoma, optic atrophy, Leber hereditary optic neuropathy (LHON), and retinitis pigmentosa [1,7, 8]. A widely accepted view postulates that increased reactive oxygen species (ROS) levels, resulting from mitochondrial dysfunction, is a major cause of retinal degeneration in human and mouse [9]. According to this model, light triggers mitochondrial activity, leading to increased production of ROS and cellular damage. In Drosophila, the function of several mitochondrial genes has been assessed in PRs. These include Succinate dehydrogenase A (SdhA), a subunit of mitochondrial Complex II [10], Sicily, a protein required for Complex I assembly in mitochondria [11], Opa1, a protein required for inner mitochondrial membrane fusion [12], Aats-met, a mitochondrial methionyl-tRNA synthetase [13], and NnaD, a mitochondrial zinc carboxypeptidase [14]. Consistent with previously published data in mammals [9], all of the mutants in which ROS was assessed have been associated with elevated ROS, suggesting that increased oxidative stress promotes PR degeneration [10–13]. Genetic screens in Drosophila have identified mutations in numerous genes that cause PR degeneration and that are also conserved in human. These mutants can be categorized into two broad groups: those that cause light- and activity-dependent PR degeneration and those that cause light- and activity-independent degeneration. The majority of mutations in genes that are primarily implicated in the phototransduction pathway typically cause light-dependent PR degeneration either due to aberrant Rhodopsin1 (Rh1) trafficking or Ca2+-mediated excitotoxicity [15–17]. However, mutations that constitutively activate the phototransduction pathway, leading to excessive Ca2+ influx, cause light-independent PR degeneration, e. g. , loss of function of rdgA [18] or in trpP365, which encodes a constitutively active TRP (Transient Receptor Potential) channel [19]. Light-independent PR degeneration has also been documented for a single fly mitochondrial mutant. The authors showed that the demise of neurons is due to oxidative stress because of the loss of SdhA in mitochondria [10]. Since light dependence has not been tested for the other mutations causing mitochondrial dysfunctions, it is not obvious which mutations cause which type of neurodegeneration, nor what the nature of the insults are that underlie these neurodegenerations. In this study, we show that mutations that impair mitochondrial ATP production without a concurrent increase in oxidative stress exhibit light-dependent PR degeneration. In contrast, mutations that affect ATP production as well as oxidative stress exhibit light-independent PR degeneration that can be exacerbated by light exposure. Furthermore, the observed light-induced PR degeneration in mutants affecting mitochondrial ATP synthesis stems from defects in the phototransduction cascade leading to aberrant endocytosis and delay in the degradation of Rh1. To identify genes required for the maintenance of neurons in the visual system, we performed an unbiased mosaic genetic screen on the X chromosome. We induced large homozygous mutant clones of essential genes in the eyes using the ey-FLP system and screened for age-dependent defects in electroretinograms (ERGs) [20,21]. ERG recordings are induced by light and exhibit “on” and “off” transients (arrow and arrowhead in Fig 1A), indicative of synaptic communication between the PR neurons and postsynaptic cells. They also exhibit a corneal negative response, the amplitude of which signifies the depolarization of PR neurons (dashed line) (Fig 1A). One of the isolated complementation groups, named ppr, pentatricopeptide repeat containing protein (see below), displayed a dramatic reduction in ERG amplitude as well as a loss of “on” and “off” transients in five-wk-old but not 2–3-d-old animals, suggesting a progressive PR degeneration (Fig 1A). The causative mutations of the five alleles of this complementation group were mapped to CG14786 (ppr), an uncharacterized gene in Drosophila (Fig 1B and 1C and S1 Fig). All alleles carry a premature stop codon (Fig 1B and 1C). Two rescue transgenes, a 20 kb P[acman] BAC (P/ΦC31 artificial chromosome for manipulation) CH322-75O21 genomic fragment that contains CG14786 [22] and a 5 kb genomic fragment of CG14786 (Fig 1C), rescue the pupal lethality associated with the loss of ppr. Moreover, ppr mutants (pprA, W150Stop) carrying the genomic rescue transgene (P[acman] BAC CH322-75O21, S1A Fig) show normal ERG amplitudes in aged animals (Fig 1A). The human homolog of ppr is LRPPRC, a mitochondrial protein (Fig 1D) whose loss causes Leigh syndrome [23]. Similar to LRPPRC and other pentatricopeptide proteins, the Ppr protein contains multiple PPR repeats (Fig 1E and S1B Fig; hence ppr). The Ppr protein has a putative amino terminal mitochondrial targeting sequence, as predicted by Mitoprot [24] (Fig 1E). To assess the subcellular localization of the protein, we created transgenic lines carrying a 5 kb genomic rescue transgene in which ppr is tagged at the C-terminus with Green Fluorescent Protein (GFP) (Fig 1C). This construct rescues the lethality of pprA and pprE, is ubiquitously expressed, and the protein colocalizes with a mitochondrial protein, ATP5A (Fig 1F and S1C and S1D Fig). In summary, we identified mutations in a fly homolog of LRPPRC, a protein that is localized to mitochondria and whose loss causes a progressive decline of PR function. To determine whether the progressive age-dependent decay in ERG amplitudes is light-dependent, we raised the flies in constant darkness or a 12 h light/dark cycle for five weeks. The ERG amplitudes of mutant PRs are not affected when the flies are raised in the dark, whereas flies maintained under a 12 h light/dark cycle exhibit severely diminished ERG amplitudes (Fig 2A and 2B). Moreover, the ERG amplitude is dramatically reduced in one-week-old ppr mutant flies if they are maintained under constant light (Fig 2C). Hence, the progressive defect in ERG loss in ppr mutants is induced by light. To assess the morphological features of ppr mutant PRs upon aging and light exposure, we examined cross-sections of the retina by light and Transmission Electron Microscopy (TEM). In the fly eye, PR cells are organized in ~800 ommatidia, and each ommatidium contains eight PR cells (R1–R8). Cross-sections across the retinal PRs reveal the dense microvillar structures of the rhabdomere (Fig 2D arrows), a stack of membranes that are highly enriched in Rh1 and are required for phototransduction [15]. Retina of control, young ppr mutants and ppr mutants reared in the dark for three weeks show very similar morphological features (Fig 2D–2F and S2A–S2F Fig). Moreover, the morphology of PRs of control flies maintained on a 12 h light/dark cycle for three weeks are comparable to young flies (Fig 2G and S2G Fig), whereas the morphology of ppr mutants is highly aberrant (Fig 2H and 2I and S2H and S2I Fig). This phenotype is fully rescued by a genomic rescue transgene (Fig 2J and 2K). Degeneration occurs in PRs R1–R6, which all express Rh1 [25] (blue arrow in Fig 2D), whereas R7 and R8 are spared (red arrows, Fig 2H and 2I; compare to red arrow in 2D). Hence, the residual ERG amplitude in ppr ERG traces may be contributed by R7 and R8. In summary, a light-induced mechanism causes degeneration of PRs in ppr mutants. Although both young and old ppr mutants raised in the dark display normal ERGs, light-dependent PR degeneration typically indicates a defect in the phototransduction cascade [15,16]. To establish if the ppr mutants display defects in the phototransduction cascade, we recorded ERGs upon repetitive pulses of light [26–29]. Flies were kept in the dark for 3–4 min prior to the ERG recordings and stimulated with 10–15 cycles consisting of white light for 1 sec followed by a 1. 5 sec dark period (Fig 3A). In ppr mutant eyes, there is a rapid run-down of the amplitude and “on” and “off” responses, whereas control flies show only a very modest reduction in amplitude (Fig 3A). Since this phenotype is activity-dependent, it prompted us to assess the inactivity period (darkness) needed to recover normal ERG amplitude in ppr mutant eyes. We exposed flies to light for 30 sec to maximally reduce the stimulation response in ppr mutant eyes (Fig 3B). Upon a 5–120 sec rest period, we measured the recovery of the ERG amplitude and observed a full recovery to the light response upon a two minute rest period in the dark (Fig 3B). These results demonstrate that Ppr function is required to maintain PR activity. Phototransduction in Drosophila PRs (Fig 3C) is initiated with the conversion of Rh1 to active meta-Rh1 (MRh1) by blue light (~480 nm) [15,29,35,36]. MRh1 triggers a G-protein cascade that activates Phospholipase C (PLC, encoded by norpA), causing hydrolysis of the membrane phospholipid, phosphatidylinositol 4,5-bisphosphate [PI (4,5) P2] (Fig 3C). Hydrolysis of PI (4,5) P2 activates the light-sensitive TRP channel causing a Ca2+ influx, which is essential to depolarize the PRs [29,37–41]. The observed transient depolarization phenotype upon repetitive stimulation, as observed in ppr mutant PRs (Fig 3A and 3B), could be due to impaired PLC activity [27] and/or the inability to quickly regenerate PI (4,5) P2 [28,42], resulting in diminished TRP activity. However, we did not find any evidence for a loss of PLC activity in ppr mutants. Indeed, PLC loss typically impairs PI (4,5) P2 hydrolysis [43] and causes a delay in repolarization [44], neither of which was observed in ppr mutants (Fig 3A and S3A–S3C Fig). To assess if the expression of other proteins required for the phototransduction pathway are affected in ppr mutant PRs, we performed western blots of many key players in the process [15,16,35]. As shown in Fig 4A, none of the proteins tested display altered expression levels in ppr mutant eyes (2–3-d-old, reared in the dark). Hence, ppr does not seem to affect proteins known to be required for the light transduction pathway. In addition, there are no hints of morphological changes in these PRs prior to testing (Fig 4B and 4C). These data indicate that PLC activity is not impaired and that most known players are present. Upon photoisomerization of Rh1 to MRh1 by a photon of blue light, the latter is quickly inactivated by Arrestin2 (Arr2) binding (Fig 3C and S4E Fig) [45,46]. Subsequently, MRh1 is reisomerized to Rh1 by a photon of orange light (~580 nm), leading to the release of Arr2 [32,45]. The mechanism of Rh1 recycling requires Ca2+ influx through TRP channels [32–34,45,47–49]. A small fraction of Rh1/Arr2 complex is endocytosed and degraded [50]. A reduced Ca2+ influx results in increased levels of the Arr2/Rh1 complex, causing excessive endocytosis of Rh1, which is toxic to cells as it stresses the endolysosomal system [47,50–54]. The inability to maintain a sustained light response in ppr mutant eyes (Fig 3A) suggests an impaired Ca2+ influx in PRs [37–40], which in turn may affect the Rh1 cycle and hence lead to an increased internalization of Arr2-bound Rh1 upon exposure to light. Inducing a constitutive Ca2+ influx, however, severely impairs the function and affects the morphology of ppr mutant PRs, even in newly eclosed flies (S3D and S3E Fig), possibly due to synergizing effects of Ca2+ toxicity and mitochondrial stress. To assess whether Rh1 internalization is affected, we performed whole mount antibody staining for Rh1 [47,55]. As shown in Fig 4D–4G, ppr mutants show no defect in the dark but exhibit Rh1 accumulation when exposed to light. Note that although Rh1 is found throughout the rhabdomeres when sections are performed (S4A–S4D Fig), in whole mount preparations Rh1 is detected on the outer rim as the antibodies cannot penetrate the membrane stack. However, the whole mount protocol reveals internalized Rh1 much better (Fig 4D–4G) than stained sections (S4D Fig) [47,55–57]. In addition, brief exposures to blue light followed by orange light cause very similar accumulations of Rh1 in the cytoplasm, indicating a defect in Rh1 cycling in ppr mutant PRs (S4E Fig). Since increased cytoplasmic Rh1 is known to cause degeneration of PRs [52], our data suggest that Rh1 mediates degeneration of ppr mutant PRs. To determine if increased Rh1 internalization in ppr mutant PRs is associated with a defect in Arr2 dynamics, we tested Arr2 translocation to rhabdomeres upon blue light exposure and its release following orange light exposure. To detect Arr2, we expressed Arr2: : GFP under the control of the Rh1 promoter (Green or Gray, Fig 4H–4J), which is active in R1–R6 PRs [31]. We generated small ppr mutant mitotic clones with ey-FLP in otherwise heterozygous retina. The mutant ppr PRs (dotted circles, Fig 4H–4J) can be distinguished from wild-type PRs by the absence of RFP (shown in red). Upon blue light exposure, Arr2 translocates to the rhabdomere membranes (binding to Rh1). The subsequent exposure to orange light relocates Arr2 to the cytoplasm as it is released from Rh1 [31,45,52]. In wild-type and ppr mutant clones, Arr2: : GFP levels are low in rhabdomeres when flies are kept in the dark (Fig 4H). However, upon exposure to ~1. 5 min of blue light, Arr: : GFP levels are increased in wild-type as well as ppr mutant rhabdomeres (Fig 4I). To assess the release of Arr2: : GFP from rhabdomeres, we kept flies in blue light for 30 min followed by a 60 min exposure to orange light prior to fixation. As shown in Fig 4J and 4K, we observe a higher level of GFP florescence in ppr mutant rhabdomeres than in wild-type rhabdomeres, indicating the slow release of Arr2 in ppr mutant rhabdomeres. Moreover, light-induced internalized Rh1 in ppr mutant PRs colocalizes with Arr: : GFP punctae when compared to wild-type PRs (S4F Fig). Hence, our data indicate that impaired dynamics of Arr2 release from Rh1 in ppr mutant PRs leads to increased internalization of Rh1 and toxicity. To test if excessive Rh1 internalization causes PR degeneration in ppr mutant eyes, we examined whether reducing Rh1 suppresses the light-dependent degeneration of ppr PRs. Maturation of Rh1 requires the binding of the chromophore, 11-cis 3-hydroxyretinal, to the opsin moiety. In the absence of the chromophore, opsin is not exported to the rhabdomere but is instead degraded [17,58,59]. In flies, the major source of the chromophore is derived from dietary β-carotene/vitamin A [59]. Indeed, Rh1 levels can be reduced to less than 3% by raising flies in vitamin A-deficient food, and this reduction has been shown to suppress Rh1-mediated PR degeneration [47,56]. Interestingly, under constant light or dark conditions, the ERG amplitude in flies deprived of β-carotene is comparable to those raised on normal food (Fig 5A and S5A Fig). ppr mutants raised in constant light for seven days on normal food display ERG amplitude that is ~ 20% of control (Fig 5A and 5B), whereas ppr mutants raised on vitamin A-deficient food display an ERG amplitude that is ~60% of control. Hence, removal of most Rh1 in PRs (S5A Fig) strongly suppresses the neurodegenerative phenotypes associated with the loss of ppr. To assess whether depriving flies of vitamin A suppresses the morphological alterations of ppr mutant PRs induced by light exposure, we performed TEM of the retina in flies reared in a 12 h light/dark cycle for three weeks. As previously shown [52], rhabdomeres of flies deprived of vitamin A are small, since Rh1 is an important structural component of rhabdomeres (Fig 5D). When raised on normal food, the morphology of ppr mutant rhabdomeres is severely affected (Fig 5C), but PRs of mutant flies deprived of vitamin A are indistinguishable from controls, albeit reduced in size in both cases (Fig 5D and S5B Fig). Combined with the Arr2 data, these results indicate that increased Rh1 internalization is a major cause of PR degeneration in the absence of ppr. We have previously shown that the retromer complex alleviates endolysosomal stress in PRs by preventing Rh1 from entering the endolysosomal pathway. Hence, overexpression of subunits of the retromer promotes its activity and suppresses Rh1-induced endolysosomal trafficking defects in some mutants [56]. Similarly, overexpression of vps35 in PRs suppresses or delays the neurodegenerative defects in ppr mutants (S5C Fig). These data provide further support that aberrant Rh1 internalization/degradation is a major cause of PR degeneration in the absence of ppr. Since Ppr is a mitochondrial protein, and the phototransduction process consumes a significant amount of ATP [5,60,61], we sought to assess whether loss of ppr compromises ATP production. LRPPRC, the human homolog of ppr, and its homologs are required for polyadenylation and stability of mitochondrial RNA (mtRNA) and translation [62,63]. Mitochondrial DNA is transcribed as two long polycistronic precursor RNAs [64]. The precursor RNAs are then processed to create smaller mtRNAs, which are stabilized by the addition of a polyA tail [64,65]. To assess whether ppr is required for mtRNA stability, we quantified the mtRNA levels for 14 transcripts by RT-qPCR and normalized this data to mitochondrial precursor RNA. As shown in Fig 6A, except for Complex I, all mtRNA levels are significantly reduced in mitochondria of ppr mutant larvae, in agreement with a role of Ppr proteins in mtRNA stability [62,63,66–69]. Moreover, the mtDNA content, normalized to nuclear DNA, in ppr mutants is about four times higher than in control larvae (Fig 6B), suggesting that the loss of mtRNA may induce a compensatory increase in mitochondrial biogenesis. Indeed, when we normalize the mtRNA levels with nuclear RNA (RP49), we found an increase in mitochondrial precursor RNA levels (right of Fig 6C) consistent with an increase in mitochondrial biogenesis (Fig 6B). However, normalization of processed mtRNA with nuclear RNA reveals that the mtRNA of ND5, CoI, and CoII are up-regulated and Cyt-b is down-regulated, whereas others are unchanged (Fig 6C). Hence, the overall mtRNA levels in a cell are not dramatically altered. These data suggest the presence of a compensatory response, which induces mitochondrial biogenesis in ppr mutant and can, in part, counterbalance the reduced mtRNA stability per mitochondrion. Given that mtRNAs encode 13 different proteins that are all components of the mitochondrial electron transport chain (ETC) complex (I, III, IV, and V) [13,64], we sought to determine enzymatic activities of individual ETC components from whole cell lysates (Fig 6D) or isolated mitochondria (S1 Data). We also measured Citrate synthase (CS) activity to normalize ETC complex activity. We observed significant decreases in the activities of Complex I, Complex II, and Complex IV in ppr mutant larvae (Fig 6D). The decreased activity of Complex II is striking, as its subunits are encoded in the nucleus [70]. Nevertheless, these data are consistent with the reduced Complex II activity that was observed in LRPPRC knockout mice [63]. Finally, the defects in ETC activity are rescued by a wild-type genomic copy of ppr, showing that the loss of ppr is indeed responsible for these phenotypes. To assess mitochondrial energy production, we measured the rate of oxygen consumption of intact mitochondria in vitro by polarography. In the presence of the Complex I-specific oxidizable substrates malate and glutamate, ppr mutant mitochondria exhibit a significant defect in state III (ADP-stimulated O2 consumption rate), resulting in a decreased respiratory control ratio (RCR), defined as the ratio of state III to state IV (ADP-limiting O2 consumption rate) (Fig 6E). The observed partial deficiencies of several ETC complexes in ppr mutants, combined with the defective respiration of isolated ppr mutant mitochondria (manifesting as reduced state III rate and RCR), are indicative of a reduced efficiency of oxidative phosphorylation (OXPHOS), or in other words, reduced OXPHOS-dependent ATP production [71]. We therefore measured steady state levels of ATP and observed reduced ATP levels in ppr mutant larvae when compared to control animals (Fig 6F). Together, these results provide compelling evidence that ppr regulates mitochondrial RNA levels and thereby affects OXPHOS and ATP levels. PRs are known to consume up to 10% of total ATP in blowflies [60,72]. ATP consumption increases 5-fold above baseline in Drosophila PR in the presence of light [5,6]. Thus, we tested ATP levels in ppr mutant eyes exposed to light for 1 h (Fig 6G) and found that the ATP deficit in mutant PRs is about twice as high (40%) as in the third instar larvae (20%). Furthermore, we investigated the change in ATP levels following light exposure in control and ppr mutant heads. Interestingly, there is a significant increase (48%) in ATP levels in wild type controls upon a 1 h light exposure but only a subtle increase (13%) in ATP level in ppr mutants (Fig 6H). In summary, there is impaired ATP production in the eye of ppr mutants. It has been shown that mitochondrial activity is triggered by Ca2+ influx in neurons [72–75]. We therefore measured changes in ATP following light exposure in mutants that have an impaired Ca2+ influx (trp [38] and norpA/PLC [37,76]). Indeed, these mutants fail to increase ATP levels following light exposure (Fig 6H), suggesting that a Ca2+ influx is required to activate mitochondrial ATP production. These data indicate the presence of a feedback mechanism required for ATP generation to ensure the continuity of the phototransduction process. Besides ppr, the Drosophila genome contains a single other gene that contains PPR motifs, bicoid stability factor (bsf) [66]. RNAi-mediated knockdown of bsf also affects mtRNA stability [66]. Hence, ppr and bsf may be partially redundant. We identified an allele of bsf (bsfSH1181; Fig 7A) that appears to be a null allele, as no Bsf protein was detected in western blots (Fig 7B). Ubiquitous expression of bsf cDNA rescues the pupal lethality associated with bsfSH118. Finally, Bsf also colocalizes with Ppr: : GFP (Fig 7C). These data permitted us to compare and contrast the mitochondrial phenotype of ppr mutants to bsf mutants. When we assessed mtRNA levels, as shown in Fig 7D and 7E, bsf mutants show a similar phenotypic profile to ppr mutants although typically more severe. In addition, bsf mutants also show defects in the ETC activity (Fig 7F). Similar to ppr mutants, CII activity is reduced in bsf mutants. However, unlike ppr, bsf mutants display a severe reduction in CIII activity. These data suggest that Ppr and Bsf may play partially redundant functions. To test this hypothesis, we created double mutants. As mentioned before, ppr and bsf mutants cause pupal lethality. However, “ppr–bsf” double mutants die as embryos (Fig 7G), suggesting that Ppr and Bsf are partially redundant. Mitochondrial defects have been shown to cause elevated ROS levels and retinal degeneration in mammals and flies [9–13,77]. We therefore tested if ROS levels are elevated in ppr mutants by staining with dihydroethidium (DHE), a dye which detects superoxide radicals [78,79]. As shown in Fig 8A, ppr mutant clones in eye imaginal discs, marked by loss of GFP, do not show differences in fluorescence intensity when compared to neighboring wild-type tissue. We also performed DHE staining in adult eyes exposed to 24 h constant light. As shown in Fig 8A, the level of DHE staining in mutant eye is similar to control eye (Fig 8B–8C). We also assessed ROS levels by assaying mitochondrial aconitase activity. The native activity of this enzyme is extremely sensitive to elevated ROS [80] and a highly reliable readout in Drosophila [11,77]. As shown in Fig 8D, aconitase activity in mutant animals is comparable to control, suggesting that ROS levels are not affected in ppr mutants. Furthermore, we overexpressed human copper-zinc superoxide dismutase (hSOD1), a potent suppressor of neurodegeneration induced by ROS in flies [12,77,81], in ppr mutant PRs. However, we did not observe a suppression of the degenerative phenotype (Fig 8E), again implying that PR degeneration in ppr mutants is not induced by oxidative stress. Based on our findings, loss of ppr causes reduced ATP production but does not alter steady state ROS levels. However, ppr deficiency causes a severe loss of ERG responses and Rh1 accumulation upon repetitive light stimulation as well as a progressive light-induced PR degeneration. In the genetic screen that permitted the isolation of ppr, we identified mutations in numerous genes whose proteins are targeted to mitochondria [20]. To assess if mutations in genes that have been shown to affect ATP production display similar phenotypes, we evaluated an embryonic lethal allele of knockdown (knd16A) [21], which encodes a homolog of CS, and CG7010 (pdha21A, G170E) [20], which encodes the E1 subunit of Pyruvate dehydrogenase. Loss of CS impairs the tricarboxylic acid (TCA) cycle and hence NADH and ATP production [82,83], whereas Pyruvate dehydrogenase converts pyruvate to acetyl-CoA and mediates entry of glycolytic products into the TCA cycle [84]. Mutant clones in the eyes of knd and pdha show normal primary ERG amplitudes in young flies and flies aged in complete darkness, similar to ppr mutant PRs (Fig 9A). However, a seven-day exposure to light nearly abolishes ERG amplitudes in these mutants, whereas wild type control PRs are barely affected. Hence, loss of kdn or pdha causes a severe light-induced degeneration. In addition, both mutants fail to sustain the ERG amplitude upon repetitive stimulation in young animals (Fig 9B), similar to the phenotypes associated with the loss of ppr (Fig 3A). These observations suggest that perturbations of oxidative metabolism leading to loss of ATP production in both mutants underlie these phenotypes. Given that loss of ppr induces Rh1 accumulation upon light exposure, we tested Rh1-localization in knd and pdha mutant PR of flies raised in the dark. Rh1 localization is indistinguishable from controls in 2-d-old flies (Fig 9C, 9E, and 9G). Similar to ppr mutant PRs (Fig 4G), an ~24 h exposure to constant light leads to a substantial increase in cytoplasmic Rh1 in knd and pdha mutants when compared to controls (Fig 9D, 9F, and 9H). Finally, as shown in S6A and S6B Fig, mutant clones of knd and pdha in eye discs do not show any change in DHE staining, suggesting that loss of these enzymes does not affect ROS production. In summary, the key phenotypes associated with the loss of ppr in the eye are very similar to those of knd and pdha, suggesting a common underlying pathology. Since increased ROS is commonly associated with mitochondrial dysfunction and causes retinal degeneration in flies and humans [9–11,13,77], we tested whether mutations that severely affect the ETC and exhibit a severe increase in ROS levels cause both a light-dependent and light-independent degeneration. Mutations in sicily show a severe reduction in Complex I activity, a reduction in ATP levels (S7A Fig), and a significant increase in ROS production and PR degeneration [11,77]. In dark-reared young flies, ERG amplitudes recorded from sicily mutants are comparable to controls (S7B Fig). When raised in the dark for seven days, sicily mutant eyes exhibit a ~50% reduction in ERG amplitude, whereas the ERG amplitudes of sicily mutant eyes is reduced by ~80% when the flies are kept in constant light for seven days (S7B Fig). These findings suggest that both light-independent and light-dependent mechanisms cause degeneration in sicily mutant PRs. As noted in ppr, kdn, and pdha mutants (Figs 3A and 9B), sicily mutants also show a loss of ERG amplitude upon repetitive stimulation in young animals (S7C Fig). Upon exposure to light for 7 d, we observe an increase in Rh1 levels in the cytoplasm of sicily mutant PRs (S7E Fig) when compared to controls (Fig 9D). Finally, Rh1 localization in dark-reared sicily mutant PRs is indistinguishable from controls (Fig 9C and 9D), similar to what we observed in ppr, kdn, and pdha mutant PRs (Figs 4 and 9). These results indicate that in sicily mutants, increased ROS levels [11,77] promote a degeneration that is exacerbated by Rh1 accumulation upon light exposure. In a forward genetic screen designed to identify mutations in essential genes that cause neuronal degeneration, we identified mutations in numerous nuclear genes that encode mitochondrial proteins. One of these genes corresponds to ppr, a homolog of human LRPPRC (Fig 1). Interestingly, ppr mutant PRs do not degenerate in the dark, in contrast to other mitochondrial mutants such as sicily [11] and sdhA [10], suggesting that a different mechanism underlies the degeneration in ppr mutants. Intriguingly, unlike many other mitochondrial mutants [10–13,77], loss of ppr does not affect ROS levels but impairs ATP production (Figs 6G and 6H and 8A–8D), suggesting that a reduced ATP production underlies the light-dependent degeneration. This hypothesis is supported by the identification and characterization of mutations in two other genes encoding Pyruvate dehydrogenase and CS, which play an important role in the TCA cycle. Both are critical to sustain mitochondrial ATP production [82,83,85–87]. These results, however, do not rule out the possibility that the ratios of other metabolites will be altered because of the different mitochondrial defects, and that these alterations contribute to degeneration. Nevertheless, our results indicate that mutations that affect mitochondrial ATP production without altering ROS levels do not cause PR degeneration in the absence of neuronal activity. This is in sharp contrast with other mitochondrial mutations like sdhA that display increased ROS [10]. Hence, reduced neuronal activity in this subgroup of mitochondrial mutants has neuroprotective effects. In a French Canadian population, mutations in human LRPPRC have been associated with Leigh Syndrome, an autosomal recessive neurodegenerative disorder with onset in infancy [23]. LRPPRC is a key regulator of mtRNA polyadenylation and stability as well as translation [62,63,69], and loss of LRPPRC causes a decrease in mtRNA abundance, defects in translation, ETC activity, and mitochondrial ATP production. In agreement with the phenotypes associated with loss of LRPPRC, we observe a reduction in mtRNA stability in mitochondria of ppr mutants (Fig 6A). We also show that ppr and bsf, the two fly homologs of LRPPRC, play partially redundant roles. We find that CIII activity, which is not affected in ppr mutants (Fig 6D), is significantly lower in bsf than in ppr mutants (Fig 7F). In contrast, CIV activity is significantly down-regulated in ppr mutants (Fig 6D) when compared to bsf mutants (Fig 7F). Surprisingly, both ppr and bsf mutants display a decreased activity of the nuclear-encoded CII (Figs 6D and 7F), a phenotype also observed in LRPPRC knockout mice [63]. We do not know the cause for this reduced CII activity. Reduced CII activity in ppr and bsf mutants as well as in LRPPRC knockout mice may be related to the increase in mitochondrial DNA and transcription, as observed in mouse knockouts for Mterf3 and Tfb1m [88,89]. Finally, we show that mitochondria isolated from ppr mutants show reduced ADP-stimulated oxygen consumption (Fig 6E), suggesting a defect in OXPHOS leading to reduced mitochondrial ATP production (Fig 6F–6H). In summary, features associated with the loss of ppr in flies are similar to what has been described in human cell and mouse experiments [62,63,68,69,90]. Phototransduction is a high ATP-consuming process, and eyes have been estimated to consume 10% of total ATP produced in blowflies [5,60,72]. Moreover, neurons primarily rely on mitochondrial OXPHOS for ATP production [72,91,92]. We show that ATP synthesis increases upon exposure to light in controls suggesting the need for a constant energy supply during phototransduction (Fig 6H). Ca2+ has been shown to activate ATP synthesis in mitochondria, and we observe that blocking Ca2+ influx in PRs also inhibits light-induced ATP production (Fig 6H). Hence, the failure to maintain PR activity during repetitive light exposure in young ppr animals (Fig 3A) may result from reduced mitochondrial activity. In ppr mutant PRs, a defect in Rh1 cycling (S4E Fig), due to reduced Ca2+ influx as predicted by reduced ERG amplitude (Fig 3A), induces excessive internalization of Rh1 (Fig 4G and S4E Fig). Excessive Rh1 internalization is known to overload the endolysosomal system, resulting in neurodegeneration upon prolonged light exposure [47]. Indeed, reducing Rh1 by reducing vitamin A uptake strongly suppresses the PR degeneration associated with ppr mutants (Fig 5A–5D). The observation that the overexpression of the retromer complex protein Vps35, which recycles internalized Rh1 and protects PRs from degeneration [56], partially rescues light-induced ERG phenotypes in ppr (S5C Fig) provides further support that excessive Rh1 mediates degeneration of ppr mutant PRs. Mitochondrial dysfunction is one of the leading causes of neurodegeneration [93]. However, mitochondrial disease-associated phenotypes differ significantly depending on the gene that is affected and the nature of the mutations [94]. Comparing the phenotypes observed in previously characterized Drosophila mitochondrial genes allows us to start subdividing them into more discrete phenotypic groups that can be correlated with the observed physiological defects. For example, sdhA mutants exhibit PR degeneration in the dark and an increase in ROS, yet the ATP levels remain normal [10]. In contrast, ppr, kdn, and pydh mutants exhibit reduced ATP levels [82,83,85–87,90], unaltered ROS levels and their PR only degenerate when exposed to light (Figs 2 and 9A). Mutations that cause reduced ATP production and increased ROS levels may show an intermediate phenotype. sicily mutants show a severe CI deficiency, severely increased ROS levels [11,77] and reduced ATP levels (S7A Fig). Indeed, sicily mutants exhibit a light-independent PR degeneration that is accelerated by light exposure (S7B Fig). Consistent with phenotypes observed in ppr mutants, sicily mutant PRs fail to sustain ERG amplitude upon repetitive light exposure (S7C Fig) and accumulate Rh1 when exposed to light (S7D Fig). These observations suggest that reduced mitochondrial ATP production exacerbates the phenotype induced by excessive ROS production through Rh1-mediated toxicity. In summary, our data suggest that the mechanisms that underlie the neurodegenerative phenotypes in a number of mitochondrial mutants are due to differences in key parameters like ATP production and ROS levels. Obviously, other mechanisms are also likely to play a role in mitochondrial dysfunction-associated neurodegeneration. For ERG recordings, flies were immobilized on a glass slide with glue. A sharp glass-recording electrode, filled with 100 mM NaCl was placed on the surface of the eye, and another sharp glass reference electrode was inserted in the thorax. Field potential recordings were performed after three to four minutes of darkness. The PR response was digitized and recorded using AXON-pCLAMP8. 1. To record ERGs from a single stimulation, ~1 sec of light flashed using a halogen lamp (~1,700 Lux). To record ERGs from repeated stimulations, repeated cycles of ~1 sec of light followed by ~1. 5 sec of darkness was used. See [104] for a detailed method of ERG recording in Drosophila. Fly heads were dissected and fixed overnight at 4°C in 4% paraformaldehyde, 2% glutaraldehyde, 0. 1 M sodium cacodylate (pH 7. 2), postfixed in 1% OsO4 for 1 h, dehydrated in ethanol and propylene oxide, and then embedded in Embed-812 resin (Electron Microscopy Sciences). One micron-thick sections were stained with toluidine blue and imaged with a Zeiss microscope (Axio Imager-Z2) equipped with an AxioCam MRm digital camera. Thin sections (~50 nm) were stained in 4% uranyl acetate and 2. 5% lead nitrate, and TEM images were captured using a transmission electron microscope (model 1010, JEOL). Images were processed with ImageJ and Adobe Photoshop. See [105] for detailed methods. For immunostaining of larval tissue and adult testis, tissues were dissected in PBS (pH7. 2), fixed in 3. 7% formaldehyde in PBS for 20 min, and washed in 0. 2% Triton X-100 in PBS (PBT). For whole mount staining of fly eyes, heads were prefixed in 4% formaldehyde in PBS for 30 min after removal of the proboscis. Fly eyes were then dissected from these heads, fixed for another 15 min, and washed in 0. 3% Triton X-100 in PBS. Fixed samples were blocked in 1X PBS containing 5% normal goat serum and 0. 2% Triton X-100 for 1 h (PBTS). Samples were incubated in primary antibody diluted in PBTS overnight at 4°C. For anti- PI (4,5) P2 staining, samples were incubated in primary antibody for two days at 4°C. Samples were washed in PBT, incubated in secondary antibody diluted in PBT for two hours at room temperature, and then washed in PBT prior to mounting. Primary antibodies were used at the following dilutions: Mouse monoclonal anti-Rh1 4C5, DSHB[106] 1: 50, Rabbit anti-GFP (Invitrogen) 1: 500, mouse anti-ATP synthase α subunit (Complex V; MitoSciences) 1: 500, mouse anti- PI (4,5) P2 (Echelon) 1: 100. Secondary antibodies conjugated to Cy3 (Jackson ImmunoResearch Laboratories, Inc.) or Alexa Fluor 488 (Invitrogen) were used at 1: 500. Phalloidin conjugated with Alexa 488 or Alexa 647 (Invitrogen) 1: 250 was added with secondary antibody. Samples were mounted in Vectashield (Vector Laboratories) before imaging with a confocal microscope. Heads from 1–2-d-old flies were used. Samples were processed as described in [57]. Primary antibodies were used at the following dilutions: rabbit Arr2 (1: 2000) [46], rabbit RdgC (1: 2000) [107], rabbit Trp (1: 2000) [108], rabbit Inad (1: 2000) [109], rabbit CalX (1: 2000) [110], mouse Rh1 (1: 2000) DSHB [106], rabbit NinaC (1: 1000) [111], rabbit PKC (1: 1000) [112], mouse Actin (1: 5000) (ICN Biomedicals) and Anti-Bsf (1: 1000) [113]. All secondary antibodies conjugated to HRP (Jackson ImmunoResearch Laboratories, Inc.) were used at 1: 10,000. Total RNA was isolated from control and pprA third instar larvae. Five micrograms of total RNA from each sample were reverse transcribed using Random Hexamer Primers and the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). RT—qPCR analysis of the rp49, mitochondrial precursor and mature mitochondrial transcripts were performed in triplicates using 150ng of cDNA per reaction on a 7900HT Real-Time PCR System using ABI SYBR Green PCR Master Mix (Applied Biosystems). An initial activation step for 10 min at 95°C was followed by 40 cycles of 95°C for 10 s and 60°C for 30 s. The primer sequences used are provided in S1 Table. Data is presented as mean ± SD. Fold change was calculated as previously described [114], and statistical significance was determined using a two-tailed Student’s t test (p < 0. 05). Method adopted from [115]. Drosophila whole DNA (genomic and mitochondrial) was purified from third instar larvae as the template for PCR. Template DNA was mixed with primers and green supermix reagent (iQ SYBR; Bio-Rad Laboratories). PCR was performed in a thermal cycler (iCycler; Bio-Rad Laboratories), and the data were collected and analyzed using the optical module (iQ5; Bio-Rad Laboratories) and related software following the manufacturer’s instructions. The following primer pairs were used to amplify a genomic DNA fragment corresponding to CG9277/β-Tubulin or a mitochondrial DNA fragment corresponding to CG34083/ND5, respectively: β-Tubulin forward, 5′-CCTTCCCACGTCTTCACTTC-3′; and β-Tubulin reverse, 5′-TTCTTGGCATCGAACATCTG-3′; and ND5 forward, 5′-GCAGAAACAGGTGTAGGAGCA-3′; and ND5 reverse, 5′-GCTGCTATAACTAAAAGAGCTCAGA-3′. Dissociation curves for the amplicons were generated after each run to confirm that the fluorescent signals were not attributable to nonspecific signals (primer-dimers). The mtDNA content (mtDNA/β-Tubulin ratio) was calculated using the formula: mtDNA content = 1/2ΔCt, where ΔCt = CtmtDNAΔ—Ctβ-Tubulin. Enzymatic activity assays were performed on larval whole cell extracts or isolated mitochondria from third instar larvae as previously described [13,116]. Polarography was performed on isolated mitochondria from third instar larvae as previously described [13,117]. Aconitase activity assays were performed in isolated mitochondria from third instar larvae as previously described [13]. ATP level for larvae, eyes, and heads were determined by ATP assay kit (Invitrogen) [118,119]. Flies were exposed to light (~1,800 Lux) for 1 h prior to detection of ATP levels in adult eyes and heads. Eyes were dissected in PBS, and heads were frozen on dry ice and separated on a metal plate kept on dry ice. Five third instar larvae, 20 eyes or 5 heads were dissected and homogenized in 50 μl of 100 mM Tris and 4 mM, EDTA, pH 7. 8. These homogenates were snap-frozen in liquid nitrogen and then boiled for 3 min. Samples were then centrifuged, and the supernatant was diluted (1/50 for larvae and 1/2 for heads and eyes) in extraction buffer mixed with luminescent solution. Luminescence was measure on FLUOstar OPTIMA plate reader. DHE staining was performed as described previously [78]. Flies were exposed to 24 h light (1,800 Lux) prior to DHE staining in adult eyes. Percentage protein similarity was determined using BlastP (NCBI). Protein domains were analyzed by PROSITE [120].

Mitochondrial dysfunction is associated with a number of metabolic and neurological diseases such as Leigh syndrome and progressive blindness. Increased oxidative stress, which is often associated with mitochondrial dysfunction, is thought to be a common cause of disease progression. Here, we identified nuclear genes that encode mitochondrial proteins, whose loss causes the demise of photoreceptor neurons. Contrary to the common idea that this degeneration is triggered by elevated levels of oxidative stress, we find no change in the levels of oxidative stress. We show that activating photoreceptor neurons with light significantly increases energy production, and that this process is required to sustain their activity. Mitochondrial dysfunction impairs this capacity and leads to a premature termination of the light response. This in turn impairs the cycling of the light-sensitive receptor Rhodopsin in photoreceptors, and Rhodopsin accumulates in the cell inducing toxicity. This distinct mechanism of degeneration suggests that different mitochondrial diseases may follow different paths of disease progression and would hence respond differently to treatments.

Abstract Introduction Results Discussion Materials and Methods

2015

Impaired Mitochondrial Energy Production Causes Light-Induced Photoreceptor Degeneration Independent of Oxidative Stress

Morphological development of fungi and their combined production of secondary metabolites are both acting in defence and protection. These processes are mainly coordinated by velvet regulators, which contain a yet functionally and structurally uncharacterized velvet domain. Here we demonstrate that the velvet domain of VosA is a novel DNA-binding motif that specifically recognizes an 11-nucleotide consensus sequence consisting of two motifs in the promoters of key developmental regulatory genes. The crystal structure analysis of the VosA velvet domain revealed an unforeseen structural similarity with the Rel homology domain (RHD) of the mammalian transcription factor NF-κB. Based on this structural similarity several conserved amino acid residues present in all velvet domains have been identified and shown to be essential for the DNA binding ability of VosA. The velvet domain is also involved in dimer formation as seen in the solved crystal structures of the VosA homodimer and the VosA-VelB heterodimer. These findings suggest that defence mechanisms of both fungi and animals might be governed by structurally related DNA-binding transcription factors. The fungal and the animal kingdom are related as they both belong to the ophistokonts with a common ancestor existing about 1 billion years ago [1], [2]. Animals have evolved with an elaborate inflammation and immune system for self-defence. Inflammation, the immune system, and animal development are controlled by various mono- and multiprotein assemblies of RHD-containing proteins. Among others, one family, named NF-κB, consists of five members, which respond to external stimuli [3], [4]. In contrast to animals, fungi are normally secured by a thick cell wall and had been misclassified as plants for centuries due to their loss of motility and the establishment of a cell wall. In addition, in response to various abiotic or biotic signals, filamentous fungi produce small signalling and/or defensive bioactive molecules [5], [6]. These secondary metabolites range from antibiotics such as penicillins to mycotoxins such as aflatoxins, affecting everyday life of animals and human beings [7]. Regulation of the secondary metabolism as well as the control of growth and differentiation of the model mold Aspergillus nidulans are coupled by a family of fungal regulators, the velvet proteins (Figure S1) [6], [8]. These velvet regulators are present in most parts of the fungal kingdom from chytrids to basidiomycetes. The velvet proteins share a homologous region comprising about 150 amino acids, which lack significant sequence homology to any other known proteins (Figure S2). In A. nidulans the four velvet proteins VeA, VelB, VelC, and VosA have been identified and characterized. They can interact with each other and also with non-velvet proteins resulting in complexes, which link morphological and chemical development of fungi [9]. The regulation of sexual development and secondary metabolism has been shown to be a light-regulated process coordinated by the heterotrimeric complex, consisting of the velvet proteins VeA, the VeA-like protein B (VelB), and the putative methyltransferase LaeA. The heterotrimeric VelB/VeA/LaeA-complex activates secondary metabolism and sexual development. The ΔlaeA and ΔveA mutant strains are unable to produce hardly any sterigmatocystin, the penultimate precursor of aflatoxins. Similarly ΔveA and ΔvelB strains do not form any sexual fruiting body [9]. Notably, in the dark VeA is predominantly found in the nucleus, whereas it is mostly in the cytoplasm in the light. VeA contains an N-terminally located nuclear localisation signal (NLS) recognized by the nuclear import factor KapA mediating the transport from the cytoplasm into the nucleus [10], once it becomes accessible by a yet unknown factor or mechanism. VosA contains an N-terminally located velvet domain and is required for the transcription of several genes essential for spore viability [11]. Deletion of vosA results in a severe down-regulation of genes associated with trehalose biosynthesis (tpsA, tpsC, and orlA) and the lack of trehalose biogenesis in spores. As a consequence, spores of ΔvosA strains are much less resistant to heat, UV, and other stress conditions and exhibit a strongly reduced survival rate after 10 d. Studies of spore viability of the ΔvelB mutant revealed that the interaction of VosA with VelB is required for proper expression of the trehalose biosynthesis genes in fungal spores [12], [13]. Similar to the ΔvosA strain, the ΔvelB strain produces spores that contain virtually no trehalose, rendering them much more susceptible to desiccation and other stresses. The role of velvet proteins in other fungi has been extensively studied in the past few years. While external stimuli can be different, their regulatory function on secondary metabolism and development seems to be conserved. In the human pathogen Histoplasma capsulatum, the switch from filamentous growth to the pathogenic yeast form is triggered by a temperature increase and requires the VosA and VelB orthologues Ryp2 and Ryp3, respectively [14], [15]. In Fusarium fujikuroi, the deletion of the veA and velB homologues Ffvel1 and Ffvel2 affects the secondary metabolism and virulence on rice [16]. In several cases the veA null mutation could be rescued by complementation of cross-genus veA from other fungi [16]–[18]. Numerous recent studies support a role of velvet proteins in fungal virulence [19]–[22]. Here we report the molecular basis of the velvet-mediated gene regulation. Genome-wide and targeted DNA binding studies of VosA reveal that its velvet domain recognizes an 11-nucleotide sequence present in the promoter regions of many regulatory and structural genes. The crystal structure analysis of the velvet domains of VosA and the heterodimeric VosA-VelB complex demonstrate that the velvet domain is an RHD-like domain related to NF-κB. Besides the velvet domain, VosA contains a predicted C-terminal transcriptional activation domain, implying that it is likely a transcription factor [11]. Taken together, the existence of novel fungus-specific transcription factors possessing a mammalian NF-κB–like DNA-binding domain suggests a common functional origin for the coordination of fungal development with secondary metabolism and the immuno-inflammatory response control in humans. Due to their regulatory roles and nuclear localization, a function as transcription factor was proposed for velvet proteins. However, based on their amino acid sequences, no known DNA-binding domain could be identified. To test for a potential DNA binding activity of velvet proteins, in vivo chromatin immuno-precipitation (ChIP) employing the VosA protein tagged with FLAG followed by A. nidulans tiling microarray analysis (ChIP-chip) was carried out. The results revealed that more than 1,500 genes' promoters were enriched by the VosA-FLAG-ChIP (Table S1). For verification of these results, we further carried out ChIP-PCR and demonstrated that VosA-FLAG-ChIP indeed specifically enriched the promoter regions of the genes associated with asexual development (brlA, wetA, and vosA) and trehalose biosynthesis (tpsA and treA) (Figure 1A; Figure S3). Using VosA-ChIP-chip result, we then performed the consensus motif analysis, which led to several motifs (Table S2). Independent of this computer-based analysis, we carried out a series of electrophoretic mobility shift assays (EMSAs) with full-length or truncated VosA proteins (VosA, VosA_N, VosA_C) using various regions of the brlA promoter as probes. Such a promoter-walking EMSA revealed that full-length and the N-terminal half of VosA (VosA_N, residues 1–216) containing the entire velvet homology region binds to a 35 bp fragment of the brlAβ promoter (−1,395 to −1,361, marked by the arrowhead in Figure 1B). To develop a preliminary motif recognized by VosA, we used 14 sequences from ChIP-chip and Northern blot results (Table S3) and one 35 bp brlAβ fragment from EMSA. These 15 sequences were subject to MEME (Multiple Em for Motif Elicitation) analysis [23] and an 11-nucleotide consensus sequence was found (Figure 1C). This motif (positions 5 through 10) is similar to motif 1 from ChIP-chip results (Table S2) and has a very good palindrome structure (CCGCGG). In addition, the TGG sequence (positions 2 to 4 in a preliminary motif sequence) is included in three motifs from ChIP-chip results. To test whether these sequences (CCGCGG and TGG) in the 35 bp brlAβ fragment are core sequences that are needed for DNA-binding, we designed three probes that specifically deleted these regions (Figure 1C). VosA-DNA binding was decreased when the mutated probes were used (Figure 1D). These results suggested that the VosA velvet domain represents a novel type of DNA-binding domain that recognizes an 11-nucleotide DNA sequence. Previous studies proposed that the VelB-VosA heterodimer is a functional unit of trehalose biosynthesis in spores and of spore maturation [12], [13]. To test whether VelB-ChIP also enriches the promoter regions of the VosA target genes, a VelB-ChIP-PCR analysis was performed. VelB-FLAG-ChIP enriched the same promoter, but not ORF-regions of VosA target genes (Figures 2A and S3). To test the molecular consequences of the lack of VosA or VelB in spores, we then examined the mRNA levels of the high score genes in wild-type, ΔvosA, and ΔvelB strain conidia (Table S3). Deletion of velB, similar to deletion of vosA, caused reduced accumulation of AN8694, nsdD, AN5371, cteA, AN6508, and AN5709 mRNAs, but increased transcript levels of brlA, treA AN8741, and rfeG (Figure 2B). These results suggest that these two velvet regulators play a dual role in activating genes associated with spore maturation and repressing certain development-associated genes. As mentioned above, the velvet domain of VosA is a DNA-binding domain. To test the DNA binding ability of the velvet domains of other velvet proteins, we carried out further EMSAs using VeA, VelB, or a heterodimer composed of VelB and the minimal velvet domain of VosA encompassing amino acid residues 1–190 (VosA1–190), which was designed for crystallization (see below). The VelB protein alone failed to bind to the brlA probe, whereas VosA1–190-VelB and VeA bind to this probe readily (Figures 2C and S4). In addition, VosA1–190-VelB and VeA binding to DNA was decreased when the mutated probes of the brlA-promoter region were used (Figure S5). Overall, these results imply that the homodimers of VosA and VeA may recognize the same promoter regions, while VelB alone at least does not bind the brlA promoter. However, binding sequences and affinities might be different for heterodimers. In order to gain insights into the structural basis of the specific DNA recognition and to increase the chances for crystallization, a minimal velvet domain of VosA encompassing residues 1–190 (VosA1–190) was cloned, expressed, and purified. The DNA binding properties—albeit weaker than for the VosA_N—indicated a properly folded and active entity (Figure 1B), which crystallized readily forming well-diffracting tetragonal crystals. The crystal structure of VosA1–190 was determined de novo by means of single-wavelength anomalous dispersion (SAD). The structure was refined at a resolution of 1. 79 Å (Table 1) and comprises residues 8–185 belonging to one VosA monomer occupying the asymmetric unit. Crystal packing analysis revealed the existence of a homodimer defined by a crystallographic 2-fold symmetry axis, which is consistent with the results from size exclusion chromatography and multi-angle-light-scattering (MALS), demonstrating that VosA1–190 exists in solution as a homodimer (Figures S6 and S7). The VosA velvet-domain folds into a highly twisted β-sandwich containing seven antiparallel β-strands. One side of the β-sandwich is involved in dimer formation, whereas the other one is flanked by several loops of which two fold into an α-helix. These α-helical fragments are located between β-strands 2 and 3 and at the C-terminus (Figure 3A). The 2-fold symmetry of a homodimer results in an antiparallel orientation of β-strands 2,3, 5, and 7 to the same strands of the other subunit (Figure 3B). The surface area covered by the interaction of the subunits is 1,078 Å2 corresponding to 12. 7% of the total molecule surface sufficient for a stable interaction required for dimer formation. Both subunits contribute to a positively charged patch on the homodimer' s surface likely to be involved in binding and recognition of DNA (Figure 3B). A search for proteins structurally homologous to VosA using DALI [24] identified the mammalian transcription factor NF-κB-p50 [25], [26] as the most similar protein structure with a root mean square deviation (r. m. s. d.) of 2. 8 Å for 113 common Cα-atoms (Figure 3C). Given the low amino acid sequence identity of 13. 7%, this structural similarity was quite unexpected (Figure S8). NF-κB itself belongs to the family of Rel-proteins containing the conserved Rel-homology-region, which encompasses 300 residues that fold into two immunoglobulin-like domains [25], . Both of these domains are involved in DNA-binding, but dimerization of Rel-proteins is exclusively mediated by the C-terminally localized, shorter domain (Rel-C). Opposing, the fold of the VosA velvet domain resembles that of the longer N-terminal domain of NF-κB-p50 (Rel-N). The major difference between p50-Rel-N and the velvet domain is an insertion of three additional α-helices between β-strands 7 and 8 of p50-Rel-N (Figures 3C and S8). A C-terminal helix that covers the β-sheet in the region formed by β1, β2, β5, and β4 of VosA is missing in p50-Rel-N. Instead p50-Rel-C interacts with a long loop connecting β1 and β2, extending the protein on that side. The crystallized VosA fragment comprises the 190 N-terminal residues corresponding to the velvet domain. An amino acid sequence analysis of the missing C-terminal part of VosA performed with HHpred [27] revealed a homology to the Sec24 transport protein for a fragment comprising residues 240 to 434 of VosA (sequence identity and similarity of 23% and 35%, respectively). This C-terminally located domain is connected via a 50-residue-long region without any detectable structural homology, which nevertheless could correspond to the C-terminally located Rel-C domain of NF-κB-p50. Superposition of VosA and NF-κB-p50 structures revealed that the flexible loop connecting two domains in NF-κB-p50 could be structurally equivalent to a loop preceding the C-terminal helix (K160-M165) in VosA. The interpretation of the experimental electron density map of that loop was not unambiguous, as it is located close to crystallographic 2-fold axis. Hence the two possible conformations of that loop affect the positioning of the remaining C-terminal fragment that can either cover the β-sheet of the same protein molecule or the β-sheet of the adjacent protein molecule related by the 2-fold symmetry by employing a domain swapping (Figure S9). Utilizing the structural similarity to NF-κB, the mode of DNA binding of the velvet domain may be deduced from the superposition of the VosA monomer to crystal structures of NF-κB-DNA complexes [25], [26]—for example, the NF-κB p50 homodimer bound to DNA (PDB ID code 1SVC). The superposition reveals that the loop connecting the first and second β-strand (loop A) as well as the loop located before the C-terminal helix of VosA (loop B) could be involved in DNA binding by interactions with the major groove (Figure 3C). Indeed, several positively charged residues (Lys, Arg) prone for DNA-binding activity are located within these loops (Figure 3D). To test whether these residues are critical for DNA-binding, K37, K39, R41, and K42 located in loop A and K160 in loop B were individually substituted with alanine. The DNA binding activity of the mutated VosA1–190 proteins was tested by an EMSA with the 35 bp fragment of the brlAβ promoter containing the VosA binding motifs (Figure 4) and additionally with the same brlAβ promoter fragment with deleted VosA binding motifs 1 and 2 (Figure S10), verifying the importance of both motifs in VosA-DNA interaction. Remarkably, severely reduced DNA-binding activity was observed for all mutants located in loop A (Figure 4). In contrast, the loop B K160A mutant retained its DNA-binding capability. This observation suggests a minor, if any, involvement of the second loop in DNA binding and supports its assignment as a flexible loop joining two individual VosA domains. Substituting all four positively charged residues in this loop—namely K37, K39, R41 and K42—simultaneously with alanine completely abolished DNA binding activity. Similar to this quadruple mutant, already the double mutant K37A/K39A cannot bind to DNA anymore (Figure 4). This strongly indicates that this loop of VosA is involved in protein–DNA interactions, which is consistent with structural superposition with the NF-κB-DNA complex (PDB ID code 1SVC) (Figure 3C, D). The function of the double mutated protein K37/39A was also analyzed in vivo in A. nidulans. The introduction of the mutated version of the vosA gene in a vosA deletion strain only partially complemented the vosA deletion phenotype, suggesting that the mutated protein is not working properly due to its reduced DNA binding activity (Figure S11). Adding the second molecule of the VosA homodimer to the model of a VosA-DNA complex reveals that a significant bending of the DNA has to occur if both subunits bind DNA simultaneously. The interacting loop A of the second VosA molecule would be positioned about 10 bases distant from the binding site of the first VosA monomer (Figure S12A, B). Known physiological functions of VosA depend on the concerted action with the velvet protein VelB, which was shown to form a heterodimer with VosA [12], [13]. In order to unveil the molecular basis of the VosA-VelB interaction, we also determined the crystal structure of the VosA1–190-VelB heterodimer at a resolution of 2. 2 Å (Table 1, Figure 5). In contrast to the other velvet-proteins in A. nidulans (VosA, VeA, and VelC), the velvet-domain of VelB is not continuous but is interrupted by an insertion of 99 amino acids (residues 132–231; Figure S2). This insertion is rich in proline, glutamine, glycine, tyrosine, and serine residues, and it is predicted to form an intrinsically disordered region. Even though full-length VelB (369 residues) was used for complex formation and crystallization, the insertion is not present in the crystal structure most likely due to a proteolytic removal (Figure S13). This proteolytic activity might also be the reason why residues 161–190 of VosA in this heterodimer structure are not defined in the electron density map, even though residues 161–185 are defined in the structure of the VosA homodimer. Overall VosA exhibits an almost identical fold as in the homodimer structure with an r. m. s. d. of 0. 80 Å. The VelB velvet domain adopts a fold similar to that of VosA (r. m. s. d. of 1. 01 Å for 136 Cα positions), however it contains one additional N-terminal β-strand. In the heterodimer, VosA and VelB share the same interaction surface with respect to the secondary structure elements involved in the interaction as the monomers of the VosA homodimer (Figure 5). The surface covered by the interaction of both molecules (1,453 Å2) is slightly larger than that in the VosA homodimer (1,078 Å2). A major difference to the VosA homodimer is that the VelB protein in the VosA-VelB heterodimer is oriented differently. With respect to the second VosA-molecule, in the homodimer it is rotated by around 30° and shifted about 3 Å closer toward VosA (Figure S14). Importantly, the lack of the 99-residue-long insertion in the VelB velvet domain does not compromise the binding of VelB to VosA (Figure S5). Superposition of VelB with VosA reveals that DNA-binding loops A and B in VelB differ in arrangement and/or sequence from VosA (Figure S12). Loop B, which is highly similar in the overall structure to VosA, contains two lysines instead of one, both pointing away from the protein molecule. In both VosA and VelB, loop A is made up of eight residues. While loop A of VelB has an irregular conformation, the corresponding loop A of VosA contains a short helical element resulting in an elongated loop. The lysines in loop A of VosA play an important role in DNA binding, however there is only one lysine in loop A of VelB present and K42 is exchanged into R81. Interestingly, R80 (corresponding to R41 in VosA) is not pointing inward as in the VosA homodimer, but is oriented outward and interacts with a sulfate ion and the neighbouring D77 side chain and D79 carbonyl group. Flexibility of the loop regions comprising the positively charged residues might be a prerequisite for these interactions to contribute to both DNA binding and subsequent stabilization of the loop conformation of VelB (Figure S12). All together the differences in the DNA-binding loops are indicative for a different interaction surface, which in turn could cause different DNA sequence specificity of VosA and VelB, respectively. Members of the velvet protein family have been defined by a conserved sequence comprising some 150 amino acids, denoted as velvet domain. Only in VelB the velvet domain contains an insertion of 99 residues of yet unknown function. The crystal structure analysis of VosA and VelB revealed that the velvet domain represents a structural entity. The velvet domain is involved in specific DNA binding as well as in the dimerization of the different velvet proteins, resulting in formation of homo- and heterodimers. The common fold of the VosA and VelB velvet-domain comprises a highly twisted β-sandwich composed of seven antiparallel β-strands, an α-helix inserted in the loop connecting β-strands 2 and 3, and a second, C-terminally located α-helix (Figures 3 and 5). VosA forms a homodimer, as indicated by an interaction surface encompassing 12. 7% of the total surface. This is supported by results from gel-filtration and MALS experiments. Head-to-head homodimer formation buries equivalent outside surfaces formed by strands β4, β3, β6, β7 of each VosA monomer resulting in formation of an intradimer eight-stranded β-sandwich. The newly formed β-sandwich buries several hydrophobic residues located on β3, β6, and β7, in particular three Phe (F72, F136, and F145), positioned in a row perpendicular to β-sandwich axis. Thus a cluster of six Phe forming a large hydrophobic patch could be the major driving force for VosA oligomerization. Interestingly, although using the identical region for interaction in the VosA-VelB heterodimer, the complexes significantly differ in relative orientation of individual monomers. This is mostly due to different length and curvature of the β6 and β7 and interconnecting loop found in VelB β-sandwich, which in order to pack against a β-sandwich of VosA needs to be rotated by 32° and shifted by 3 Å. This results in a better fit of the two subunits and increase of the interaction surface, which similarly to the VosA homodimer buries several hydrophobic side chains in the core of newly formed intradimer eight-stranded β-sandwich (Figure 5). Both proteins reveal a structural similarity to members of the NF-κB-p50-family, especially to a conserved region of about 150 amino acids that resembles an RHD-like fold. In contrast to NF-κB, where the C-terminal domain is responsible for the dimerization and both domains are capable of binding to DNA, the velvet domain harbours both functions. Based on the obtained structural information of VosA and VelB and their specific DNA binding properties (discussed below), we propose that the fungal velvet proteins represent a new class of direct DNA-binding transcription factors sharing a common ancestor (s) with the NF-κB-p50 family. Comparison of VosA and VelB overall structure with NF-κB allowed identification of critical residues for DNA-binding activity located in a loop region within a patch of positively charged surface. According to the superposition, binding would occur to the major groove of the DNA. Substitution of these residues with alanine in VosA clearly abolishes DNA-binding activity. The strongest effect on DNA-binding is observed by replacing K37/K39, suggesting that these two residues together with K42 provide the major interactions with the DNA. The introduction of a mutated vosA K37/39A in the genome of A. nidulans proves that these two residues are important for proper function of the protein. The deletion of vosA results in an up-regulation of the brlA gene and to a loss of viability after 10 d of growth due to the lack of trehalose. Additionally, the deletion influences the production of sexual spores and of pigments released to the agar (Figure S11B). The introduction of the mutated vosA K37/39A allele in a ΔvosA strain fails to fully complement these defects. Only the production of sexual spores was restored, whereas the spore viability was partially restored, but the defected control of pigmentation (Figure S11B) and brlA expression (Figure S11C) was not rescued. This patch of positively charged residues defined by K37, K39, and K42 is well conserved among all velvet proteins, suggesting a common mode of protein-DNA interaction for VosA and VelB. Furthermore, the structural comparison with NF-κB suggests that the VosA homodimer recognizes about 11 base-pairs, which is in agreement with the predicted 11 nucleotides consensus sequence recognized by VosA in our ChIP-chip analysis. However, the arrangement of the VosA homodimer and the superposition and modelling of the DNA reveals that the putative DNA-binding loops of the second molecule is at least ∼13 Å distant to the backbone of a dsDNA with ideal B-form conformation. A simultaneous binding to both VosA molecules would require a kink in the DNA. In contrast to the VosA homodimer, in the putative VosA-VelB-DNA complex model with VosA in close proximity to the DNA, the distance of the velvet domain of VelB to the DNA backbone is reduced and would require less bending of an ideal B-form DNA for tight interaction. Both DNA sequence motifs 1 and 2 play an important role in DNA binding of VosA, however with differing importance. Deletion of motif 1 has a higher influence on binding efficiency than deletion of motif 2. A recent study demonstrated that the DNA-binding sequence of the velvet proteins Ryp2 and Ryp3 in H. capsulatum is highly similar to the motif 1 derived from our ChIP-chip result (Table S2) [28]. The electron density obtained from the VosA1–190 crystals allows an interpretation of two alternate conformations of its C-terminal helix. The missing C-terminal part of VosA might lead to an overall conformation of VosA similar to the one observed for NF-κB and could increase the effect of loop B on DNA binding. The additional residues present in VosA-N might therefore explain its increased DNA binding in comparison to VosA1–190. In summary, we suggest that the modulation of gene transcription is achieved by the use of varying homo- and heterodimers as seen for VosA and VelB, potentially allowing the velvet proteins to specifically recognize different DNA sequences causing differential regulatory outcomes. In A. nidulans there are four velvet proteins, three of which have been studied. VosA-VelB is essential for the regulation of asexual development and spore viability [11]–[13] and VelB-VeA for sexual development [11]. The trimeric VelB-VeA-LaeA complex coordinates differentiation and secondary metabolism in response to external signals [9]. Little is known about the function of the velvet proteins in other clades of fungi outside of the ascomycetes, where much remains to be discovered. Interestingly, the unicellular eukaryote Capsaspora owczarzaki, which is a symbiont in the haemolymph of the tropical freshwater snail Biomphalaria glabrata, carries a gene for both an NF-κB and a VosA-like protein, suggesting the coexistence of both in this organism [29]. Several studies indicate that the velvet proteins are global regulators controlling a diverse set of processes ranging from toxin production and cell wall formation to the development of resting or sexual fruiting structures [6]. It will be interesting to determine what additional common themes and features exist between fungal growth and developmental control by the velvet protein family and the immune, inflammation, and differentiation response of animals by the NF-κB protein family. One candidate for a common denominator of RHD and velvets' functions is the COP9 signalosome, a conserved multiprotein complex controlling the life span of proteins. The COP9 signalosome is required for the control of NF-κB activation [30] as well as the control of fungal development and secondary metabolism [31]. Recently it has been shown that the physical interactions between the COP9 signalosome and an additional developmental regulator protein are conserved between humans and fungi [32]. Similarly to the NF-κB transcription factors, the founding member of the velvet protein family, VeA, resides in the cytoplasm but is prevented from nuclear import in the light by a yet unknown factor or mechanism. It is tempting to speculate that the light signal leads to a posttranscriptional modification—e. g. phosphorylation—of VeA, hindering the interaction with the nuclear import factor KapA. The control of nucleocytoplasmic transport by phosphorylation is known for various proteins like Hxk2, LASP-1, IPMK, and hTERT [33]–[36]. Notably, the regulation on the level of nucleocytoplasmic transport applies also for NF-κB, as the NF-κB inhibitor Iκ-Bα binds to NF-κB, thereby masking the NLS of NF-κB [37]–[39]. However, the similarity of the velvet proteins to NF-κB does not extend to this regulatory mechanism, since no Iκ-B homolog or functional homolog could be identified in fungi yet. Hence, the precise understanding of differences and similarities in the molecular mechanisms of the velvet/Rel family might help to control fungi, which not only cause increasing problems for human health and crop yield/quality, but also play a crucial role as environmental recyclers, fermenters, industrial producers, and agricultural aids. In order to express the truncated VosA (residues 1–190), here denoted as VosA1–190, vosA cDNA was amplified using the oligos OZG479/480 containing the NcoI site. The amplicon was digested with NcoI and inserted into the NcoI site of pETM13 (EMBL, Heidelberg), yielding the plasmid pETM13-VosA190 with a 3′ coding sequence for a Strep-tag. For construction of plasmid pME3815, velB was amplified from A. nidulans cDNA with primers JG45/46 and cloned into plasmid pETM-13 digested with NcoI and XhoI. The VosA1–190 mutations K37A, K39A, K37/39A, R41A, K42A, K160A, and the dead mutation (K37A, K39A, R41A, K42A) were inserted by PCR with mutated primers. For the mutation K37A, the N- and C-terminal VosA fragments were amplified from pETM13-VosA190 with primers OZG479/JG365 and JG366/367. Then, the fragments were fused by PCR with primers OZG479/JG367. The mutations K39A, K37/39A, R41A, K42A, K160A, and the dead mutation were designed in the same way as K37A with the following primers: K39A (OZG479/JG368, JG366/367, OZG479/JG367), K37/39A (OZG479/JG642, JG366/367, OZG479/JG367), R41A (OZG479/JG369, JG366/367, OZG479/JG367), K42A (OZG479/JG370, JG366/367, OZG479/JG367), K160A (OZG479/JG372, JG373/367, OZG479/JG367), and dead (OZG479/JG371, JG366/367, OZG479/JG367). The fused fragments were cloned in NcoI-digested pETM13, resulting in plasmids pME3845 (K37A), pME3846 (K39A), pME3845 (K37/39A), pME3847 (R41A), pME3848 (K42A), pME3850 (K160A), and pME3849 (dead). The plasmid pETM13-VosA190 and the mutant forms were transformed into E. coli Rosetta 2 (DE3). Expression was carried out in ZYM5052 media [40] at 16°C. Cells were harvested by centrifugation for 20 min at 5,300×g and resuspended in lysis buffer (30 mM HEPES pH 7. 4,400 mM NaCl, 30 mM Imidazol). Cell lysis was performed using a Fluidizer (Microfluidics) at 0. 55 MPa. The resulting lysate was cleared by centrifugation at 30,000×g for 30 min at 4°C. The supernatant was applied to a 5 ml StrepTactinHP column (GE Healthcare) equilibrated with lysis buffer. After extensive washing, the protein was eluted with elution buffer S (lysis-buffer +2. 5 mM des-thiobiotin). The eluate was applied to a Superdex 200 16/60 column (GE Healthcare) equilibrated in gel-filtration buffer (10 mM HEPES pH 7. 4,400 mM NaCl). The fractions containing VosA1–190 were pooled, concentrated to 10. 4 mg/ml in centrifugal concentrators (Vivascience), and used for crystallization. VosA mutant forms were expressed and purified as the wild-type, and the gel-filtration step was omitted since the purity was more than 95% as judged by SDS-PAGE. As a final test for integrity, MALS was performed for selected complexes (Figure S7) and the mutant forms of VosA1–190 were compared to the wild-type by means of CD spectroscopy (Figure S15). His-tagged VelB was expressed as described above. Cells expressing VosA1–190 with a C-terminal Strep-tag and full-length VelB-His6 were harvested, mixed, and lysed in lysis-buffer as described before. The cleared supernatant was applied to a 10 ml NiNTA (GE Healthcare), washed, and eluted with elution buffer N (lysis-buffer +400 mM Imidazol). The resulting eluate was directly applied to a 5 ml StrepTactinHP, washed, and eluted with elution buffer S. The complex was further purified using a Superdex 200 16/600 column (GE Healthcare) equilibrated with gel filtration buffer. The fractions containing the monomeric VelB and VosA1–190 proteins were pooled, concentrated to 11. 6 mg/ml, and used for crystallization. To express the GST tagged VosA proteins used for EMSA and the GST-pull-down experiments, cDNA of the full-length vosA, vosA-N (N-terminal region, 1–216 aa), or vosA-C (C-terminal region, 217–430 aa) was cloned between EcoRI and SalI sites (for full-length vosA) in pGEX-5X-1 (Amersham) or cloned between BamHI and SalI sites (for vosA-N and vosA-C) in pGEX-4T-3 (Amersham) to make pNI47,49, and 50, respectively. These plasmids were introduced into E. coli BL21 (DE3). The GST fusion protein expression and purification was performed following the manufacturer' s (GE Healthcare) instructions. To concentrate and buffer exchange, Amicon Ultra Centrifilter Units (Milllipore) were used. BCA Protein Assay Kit (Pierce) was used to estimate the protein concentration. To express the GST tagged VeA protein used for EMSA experiments, cDNA of the full-length veA was amplified with primers OHS723/724 and cloned between SalI and NotI sites in pGEX-5X-1 (Amersham). This plasmid (pHS51) was introduced into E. coli BL21 (DE3). The GST fusion protein was expressed as described for VosA1–190, and purification was performed following the manufacturer' s (GE Healthcare) instructions. VeA1–224 was cloned into the NcoI site of pETM13 (EMBL, Heidelberg), yielding the plasmid pETM13-VeA1–224 with a 3′ coding sequence for a Strep-tag. The protein was expressed and purified as described for VosA1–190. VosA1–190 and VosA1–190-VelB were crystallized by the sitting-drop vapour diffusion method. X-shaped crystals of VosA1–190 grew after 1 d in a condition containing 100 mM MES pH 6. 5,30% PEG 4000 at 20°C. Further optimization led to crystals with the same morphology in a condition containing 100 mM MES pH 6. 5,32% PEG 2000 MME, 150 mM KI, which were used for structure determination. Prior to data collection, crystals were cryo-protected by soaking in reservoir solution supplemented with 12% (v/v) 1,4-butane-diol. X-ray diffraction data were collected at 110 K on a Rigaku MicroMax 007 rotating Cu-anode equipped with a MAR345dtb image-plate detector (Mar Research). The VosA crystals belong to the space group P4122 and have cell dimensions of a = b = 45. 40 Å and c = 189. 43 Å. The VosA1–190-VelB complex was crystallized in 100 mM MES pH 5. 5,150 mM ammonium sulfate, and 25% (w/v) PEG 4000. The plate-like crystals were cryo-protected in crystallization solution +10% (v/v) 1,4-butanediol. Diffraction data were collected at 100 K at the ESRF microfocus beamline ID23-2. The crystals belong to the space group P212121 and have cell dimensions of a = 52. 03, b = 56. 75 Å, and c = 138. 17 Å. X-ray diffraction data from the VosA1–190 crystal were integrated and scaled with the XDS package [41]. Phases were obtained by SAD using SHELXC/D/E [42] navigated through HKL2MAP [43], which found 10 iodine ions. However, for further processing only iodine ions with occupancy >0. 2 were used, which resulted in three sites with occupancies of 1. 0,0. 36, and 0. 24. The initial electron density map was readily interpretable and the majority of residues were built automatically using ARP/wARP [44]. Manual model building of missing residues was done with COOT [45]. Refinement was carried out in PHENIX [46] and Refmac5 [47]. The final model contains one monomer in the asymmetric unit. Diffraction data obtained from the VosA1–190-VelB crystals were integrated and scaled with MOSFLM [48] and SCALA from the CCP4-package [49], respectively. The calculated Matthews coefficient of 1. 59 Å3 Da−1 (corresponding to a solvent content of 22. 9%) excluded the presence of both proteins, VosA and VelB, used for crystallization trials in the asymmetric unit. Assuming that the asymmetric unit comprises two protein molecules with the size of VosA1–190, the calculated Matthews coefficient value is 2. 27 Å3 Da−1 (45. 6% solvent content). At this step of structure solution, the assumption was made to look for two molecules using the available model of VosA1–190 comprising 143 amino acids (Ser 17 to Asp 79 and Ala 86 to Met 165). As the two proteins (VosA and VelB) share about 42% amino acid identity, the VosA1–190 model could also be used to search for the fragments of VelB in case a proteolytic digest had occurred prior to crystal formation. The molecular replacement search was carried out with PHASER [50] using data between 30 and 2. 2 Å resolution. The search yielded two prominent solutions with an overall likelihood gain (LLG) score of 587 (the first molecule RFZ = 10. 7, TFZ = 2 0. 4, LLG = 236; the second molecule RFZ = 4. 4, TFZ = 16. 5, LLG = 501) and R factor of 53. 4%. The quality of initial electron density maps (overall mean FOM of 0. 41) was not good enough to distinguish whether two VosA1–190 or VosA1–190 and truncated VelB molecules were occupying the asymmetric unit. The structure was manually rebuilt in COOT and verified against simulated annealing (SA) omit maps calculated with CNS [51], which was also used during initial refinement steps. Refinement was based on slow-cooling SA (both torsion angle dynamics and Cartesian dynamics) combined with standard minimization and individually restrained B-factor refinement. Careful inspection of the electron density maps indicated differences between the two molecules in the asymmetric unit that were modelled and finally let us identify the second molecule as a proteolytically truncated VelB. The final model contains two molecules in the asymmetric unit. Probes for EMSA were generated by annealing two single-stranded reverse-complementary oligonucleotides. Binding reactions were performed in a 10 µl reaction volume containing 10 mM HEPES/NaOH (pH 7. 4), 150 mM NaCl, ∼53 pmol DNA probe, and appropriate amount of each purified VosA protein. The reactions were incubated at RT for 15 min. The complexes were resolved on a 6% polyacrylamide gel (37. 5∶1 crosslinking) with 0. 5% TBE running buffer at 200 V at RT for 20 min. The gel was stained with ethidium bromide. For sample preparation, 2-d-old conidia (∼5×109 conidia for three ChIP experiments) from the strain TNI10. 34. 1 (ΔvosA; vosA: : FLAG) were crosslinked with fresh 1% formaldehyde at RT for 30 min. Then, 1/20 volume of 2. 5 M glycine solution was added to stop the cross-linking reaction. The conidia were collected by centrifugation and the conidia were washed with cold PBS for three times. About 1. 5 ml FA lysis buffer (50 mM HEPES-KOH [pH 7. 5], 150 mM NaCl, 1 mM EDTA, 1% Triton X-100,0. 1% Na deoxycholate, 0. 1% SDS, 1 protease inhibitor cocktail tablet (Roche) per 50 ml) was added before use. Silica beads (∼300 µl) were added into cross-linked cell lysates and the lysates were broken by a mini-bead beater for three cycles (1. 5 min homogenization with 1. 5 min sitting on ice). Subsequently the samples were sonicated for seven cycles (30 s on, 60 s off) with a sonifier equipped with a microtip at 70% amplitude and level 5 of output control. All steps were carried out on ice. The sonicated cell lysates were cleared of cellular debris by a 2,000×g spin for 3 min five times. The supernatant was collected, and its DNA concentration was checked using a biophotometer (Eppendorf). About 25 µl supernatant was kept as input chromatin control (glycerol was added to 10% final concentration if samples had to be frozen). The rest of supernatant was adjusted to 10 ml with FA lysis buffer, and 50 µl of Anti-Flag M2-agarose from mouse (SigmaAldrich) was added for ChIP with constant mixing overnight at 4°C. The agarose beads were collected by centrifugation and washed with FA lysis buffer three times. About 50 µl elution buffer (10 mM Tris, pH 8. 0,1 mM EDTA, 1% SDS) was added to the beads, and the sample incubated at 65°C for 10 min to elute chromatin samples from the beads. Both the recovered supernatant from the elution and the input chromatin supernatant (saved before ChIP) were adjusted to 170 µl by elution buffer, and incubated at 65°C overnight to reverse the crosslinking reaction. Then, the samples were treated by protease K for 2 h at 37°C, and protein extracted twice with phenol (equilibrated with TE, pH 8. 0) and once with chloroform∶isoamyl alcohol (24∶1). DNA was precipitated by ethanol and resuspended in 30 µl RNase TE (to digest RNA). Finally, the DNA was purified by QIAquick PCR purification kit (Qiagen) and eluted in 30 µl of the elution buffer provided with the kit. To obtain sufficient amounts of DNA for further labelling and hybridization, DNA was amplified using the WGA kit (Sigma, WGA2). DNA labelling and hybridization were performed by Roche Nimblegen. The DNA bound by the VosA-FLAG protein was labelled with Cy5, while the control DNA was labelled with Cy3. We used the A. nidulans whole-genome tiling-oligonucleotide array containing 65,536 oligonucleotides (50∼65 nucleotides at a 75-bp interval, Roche Nimblegen). The array data from hybridization of two independent immune-precipitation experiments for the VosA∶FLAG strain were used for analysis. The data were processed using NimbleScan (Roche Nimblegen). Briefly, NimbleScan detects peaks by searching four or more probes whose signals are above the specified cutoff values using a 500 bp sliding window. The ratio data were then randomized 20 times to calculate the probability for being “false positive. ” Every peak is assigned with a false discovery rate (FDR) score based on the randomization. The lower the FDR score, the higher the possibility that the peak corresponds to a real binding site. When finalizing the candidates, we used the cutoff value “1” (log 2 ratio of experiment to control) as peak score and cutoff value “0. 05” for FDR score. Then we found the overlapping candidates from two chips. To define the DNA motif recognized by VosA, we employed two separate approaches. First, using the VosA-FLAG ChIP-on-chip results, sequences of 400 bp on the midpoint of about 1,500 enriched peaks from the two VosA-ChIP biological replicates were used as target sequences. As a background, 6 kbp sequences on about 6,000 promoter regions of genes without the VosA peaks were used (BIOINFORX Inc. , Madison, WI, USA). The data were then analysed using HOMER (Hypergeometric Optimization of Motif EnRichment; http: //biowhat. ucsd. edu/homer/). The predicted VosA binding motifs are presented in Table S2. Second, the ChIP-on-chip data were imported into RINGO [52], an R/Bioconductor package, for the analysis of ChIP-chip readouts, including the quality assessment, normalization, smoothing, and peak calling. We normalized the probe intensities by variance-stabilizing normalization (VSN) [53]. We only included the peaks that were consistent in two independent VosA-ChIP biological replicates. Then several sequences were selected according to maxLevel, the highest smoothed probe level in the enriched region (Table S3). We then examined mRNA levels of these genes and found that 10 genes' mRNA levels are affected by VosA and/or VelB (Figure 2B). Finally, VosA-ChIP enriched sequences from these 10 genes, four sequences from the known target genes wetA, tpsA, orlA, and treA, and the 35 bp brlAβ promoter fragment shown to be bound by VosA in EMSA were subject to MEME (Multiple Em for Motif Elecitation) analysis, which led to the predicted VosA binding motif CTGGCCaaGGC (Figure 1C). ChIP-PCR analysis was performed according to the manufacturer' s instructions with a minor modification using MAGnify Chromatin Immunoprecipitation System (Invitrogen). Two-day-old conidia (1×109) of WT (FGSC4), ΔvelB (THS16. 1), and ΔvosA (THS15. 1) strains [13] were cross-linked with fresh 1% formaldehyde at RT for 15 min. Then, 1/20 volume of 2. 5 M glycine solution was added to stop the cross-linking reaction. The conidia were washed and broken by a mini-bead beater for two cycles (1 min homogenization with 1 min sitting on ice). The cell lysates were sonicated for four cycles (30 s on, 60 s off) with a sonifier. The sonicated cell lysates were cleared of cellular debris by centrifugation at 13,000×g for 10 min. The diluted chromatin extracts were incubated with 2 µg of anti-FLAG antibody-Dynabeads complex for 2 h at 4°C and then washed three times with the IP buffer. The input control and chromatin sample were eluted from the beads at 55°C for 15 min with reverse crosslinking buffer with proteinase K. DNA was purified by DNA purification Magnetic Beads (Invitrogen). For amplification of precipitated DNA by PCR, the GO Taq DNA polymerase (Promega) was used. The primer sets used for PCR are shown in Table S4. As negative controls, the chromatin extract being incubated with bead only (without anti-FLAG antibody) and the samples of FGSC 4 lacking FLAG-tagged VosA or VelB (Figure S3) were used. Individual input DNA samples before immune-precipitation (IP) were used as positive controls. Two biological replicates have provided essentially identical ChIP-PCR results. Signal intensities of PCR results obtained from ChIP assays were analyzed by the ImageJ software available online (National Institutes of Health; http: //rsbweb. nih. gov/ij/). Total RNA isolation and Northern blot analyses were carried out as previously described [54], [55]. The DNA probes were prepared by PCR-amplification of the coding regions of individual genes with appropriate oligonucleotide pairs using FGSC4 genomic DNA as a template (Table S4). Coordinates and structure factors have been deposited in the PDB, namely VosA as PDB ID code 4N6Q and VosA-VelB as PDB ID code 4N6R.

In many fungi, developmental processes and the synthesis of nonessential chemicals (secondary metabolites) are regulated by various external stimuli, such as light. Although fungi employ them for defensive purposes, secondary metabolites range from useful antibiotics to powerful toxins, so understanding the molecular processes that regulate their synthesis is of particular interest to us. In the mold Aspergillus nidulans the main regulators of these processes are the so-called “velvet” proteins VeA, VelB, and VosA, which share a 150-amino acid region known as the velvet domain. Velvet proteins interact with each other, alone (“homodimers”), in various combinations (“heterodimers”), and also with other proteins, but the molecular mechanism by which these proteins exert their regulatory function has been unclear. In this work we show that velvet proteins form a family of fungus-specific transcription factors that directly bind to target DNA, even though analysis of their amino acid sequence does not reveal any known DNA-binding domains or motifs. We determined the three-dimensional structure of the VosA-VosA homodimer and the VosA-VelB heterodimer and found that the structure of the velvet domain is strongly reminiscent of the N-terminal immunoglobulin-like domain found in the mammalian transcription factor NFκB-p50, despite the very low sequence similarity. We propose that, like NFκB, various homo- or heterodimers of velvet proteins modulate gene expression to drive development and defensive pathways in fungi.

Abstract Introduction Results Discussion Materials and Methods

biochemistry proteins protein structure biology dna-binding proteins

2013

The Velvet Family of Fungal Regulators Contains a DNA-Binding Domain Structurally Similar to NF-κB

More than 800 published genetic association studies have implicated dozens of potential risk loci in Parkinson' s disease (PD). To facilitate the interpretation of these findings, we have created a dedicated online resource, PDGene, that comprehensively collects and meta-analyzes all published studies in the field. A systematic literature screen of ∼27,000 articles yielded 828 eligible articles from which relevant data were extracted. In addition, individual-level data from three publicly available genome-wide association studies (GWAS) were obtained and subjected to genotype imputation and analysis. Overall, we performed meta-analyses on more than seven million polymorphisms originating either from GWAS datasets and/or from smaller scale PD association studies. Meta-analyses on 147 SNPs were supplemented by unpublished GWAS data from up to 16,452 PD cases and 48,810 controls. Eleven loci showed genome-wide significant (P<5×10−8) association with disease risk: BST1, CCDC62/HIP1R, DGKQ/GAK, GBA, LRRK2, MAPT, MCCC1/LAMP3, PARK16, SNCA, STK39, and SYT11/RAB25. In addition, we identified novel evidence for genome-wide significant association with a polymorphism in ITGA8 (rs7077361, OR 0. 88, P = 1. 3×10−8). All meta-analysis results are freely available on a dedicated online database (www. pdgene. org), which is cross-linked with a customized track on the UCSC Genome Browser. Our study provides an exhaustive and up-to-date summary of the status of PD genetics research that can be readily scaled to include the results of future large-scale genetics projects, including next-generation sequencing studies. Parkinson' s disease (PD) is the second most common neurodegenerative disease with a prevalence of ∼1% over 60 years of age [1]. Approximately 5–10% of the patients show an autosomal dominant or recessive mode of inheritance, and several causative genes have been identified, e. g. SNCA, LRRK2, PARK2, and PINK1 (for review see ref. [2]). Recently, two other novel autosomal dominant PD genes, VPS35 and EIF4G1 [3]–[5], have been identified, the former via application of next-generation sequencing techniques. It can be anticipated that causal mutations in additional genes will emerge within the next years. However, the vast majority of patients suffer from non-Mendelian forms of PD, which are likely caused by the combined effects of genetic and environmental factors. In order to decipher the genetic architecture underlying PD susceptibility, more than 800 genetic association studies have been performed over the past 20 years. While early candidate gene studies and subsequent meta-analyses provided conclusive evidence showing that polymorphisms in SNCA [6] (encoding alpha-synuclein), LRRK2 [7] (leucine-rich repeat kinase 2), MAPT [8] (microtubule-associated protein tau), and GBA [9] (acid beta-glucosidase) significantly impact PD susceptibility, most association studies in the field provided inconclusive or even conflicting results. During the last few years, genome-wide association studies (GWAS) [10]–[19] have postulated additional PD loci. While the early GWAS and a GWAS-meta-analysis [20] were of limited sample sizes and yielded mostly inconsistent results, more recent studies have identified a number of loci that were independently confirmed in follow-up studies (e. g. GAK, BST1, and PARK16, see Table 1 for all proposed GWAS findings across GWAS publications). Very recently, a GWAS meta-analysis [21] implicated several other new putative PD loci which currently await further validation. Despite this progress, approximately 40% or more of the population-attributable risk probably remains unexplained by today' s most promising PD loci [21]. To this end, genetic association studies remain one of the mainstays of PD genetics research. However, GWAS and other large-scale association studies typically only highlight the most promising results and often do not provide data on variants showing suggestive evidence for association, or previously implied variants that could not be confirmed in the GWAS setting. As a result, the cumulative genetic evidence in favor of or against association with certain variants in the PD field is becoming increasingly difficult to follow, evaluate and interpret. To address this problem, we have comprehensively collected, catalogued and systematically meta-analyzed the data from all genetic association studies published in the field of non-Mendelian PD, including GWAS, and made all results publicly available on a regularly updated online database, “PDGene” (http: //www. pdgene. org). The results of this research synopsis are based on a freeze of the PDGene database content on March 31st 2011 (available upon request from the authors). At that time, PDGene included details on 828 individual studies across more than 50 different countries and six continents reporting on 3,382 polymorphisms in 890 genetic loci. Data for more than 2,000 SNPs were supplemented by results derived from up to three publicly available GWAS datasets [10], [12], [13] following extensive quality control and imputation. Ultimately, this procedure yielded a total of 867 polymorphisms across ∼300 genetic loci that met our criteria for meta-analysis (see Methods). Additional independent GWAS data for 147 SNPs yielding P values of ≤0. 1 in these initial meta-analyses were provided by researchers of all remaining currently published Caucasian GWAS datasets [13], [15]–[19], [22]. Following the identification of genome-wide significant association with an intronic SNP (rs7077361) in ITGA8 after addition of these data, we obtained additional data from the same GWAS datasets on ∼1,400 SNPs in the chromosomal region encompassing ITGA8 (chr10: 15346353–15801533, hg18). Finally, independent replication data in Caucasian and Asian populations from the GEO-PD consortium [23] generated for ten recently described PD loci [21] were made available for inclusion. As a result, we were able to substantially increase the sample size (up to 16,452 PD cases and 48,810 controls) for a large number of some of the most promising PD loci. For instance, we were able to add data from up to 48,861 previously not analyzed combined cases and controls to meta-analyses of some of the recently proposed PD loci [21] (median sample size 14,896, see Table 2 and Table S1 for details). In addition to these focused analyses, PDGene displays meta-analysis results for more than seven million additional SNPs originating from up to three publicly available GWAS datasets [10], [12], [13]. The results are available online (e. g. as summarized in http: //www. pdgene. org/largescalemeta. asp), where they are cross-linked to a customized and fully browsable track on the UCSC Genome Browser. The PDGene meta-analyses of the 867 core polymorphisms were based on a median of 7,680 subjects (interquartile range 4,612–16,726). Additional meta-analyses were performed after stratification for Caucasian and Asian ancestry (for details on sample size and included ethnicities for individual meta-analyses see Table S1). In addition, we also performed random-effects meta-analyses across all three publicly available GWAS datasets [10], [12], [13] following genotype imputation using data from the International HapMap Consortium and 1000 Genomes Project. Ultimately this yielded 7,123,920 SNPs that could be meta-analyzed across at least two GWAS datasets (see Figure S1 for a quantile-to-quantile plot of the GWAS-only meta-analyses). All 867 core meta-analysis results are available online on PDGene as forest plots, summarizing the relative contributions of each dataset to the most current summary effect estimate, and in the form of cumulative plots, illustrating how summary ORs evolve over time. All meta-analysis results are plotted in Figure 1 (green dots) alongside the GWAS-only meta-analysis results (black and grey dots). One-hundred-three meta-analyses across 12 genetic loci (BST1, CCDC62/HIP1R, DGKQ/GAK, GBA, ITGA8, LRRK2, MAPT, MCCC1/LAMP3, PARK16, SNCA, STK39, SYT11/RAB25) yielded summary ORs suggesting a genome-wide significant (P≤5×10−8) increase or decrease in PD risk in all ethnicities and/or after stratification for ethnic ancestry (Table 2, Table S1, and Figure S2 [forest plots]). None of these loci contained more than one SNP independently associated at genome-wide significance (as judged by pair-wise linkage disequilibrium assessments using ‘SNAP’ and r2-values of 0. 2 as cut off http: //www. broadinstitute. org/mpg/snap/). The majority of polymorphisms tested in the genome-wide significant loci do not show evidence for publication bias (Table S1). Finally, all genome-wide significant signals were robust against potential undetected sample overlap using a recently proposed procedure [24] (see Table S2 for more details). Combined sample sizes for all 12 loci were substantially larger here as compared to any previously published meta-analysis (Table S1), providing unequivocal evidence for an involvement of these loci in PD susceptibility. While power to detect genome-wide significance was excellent for most of these loci (>80% based on an OR of 1. 15, and a minor allele frequency down to 0. 05 using the Genetic Power Calculator, http: //pngu. mgh. harvard. edu/~purcell/gpc/), power was less for a large number of other meta-analyses due to smaller sample sizes and allele frequencies (see Table S1 for details). Thus, no simple statistic can summarize the overall power of our study. The above list includes an intronic polymorphism in ITGA8 located on chromosome 10p13 for which we identified novel evidence for genome-wide association with PD risk (OR 0. 88, P = 1. 3×10−8, I2 = 0, see Table 2, and Figure 2). This SNP had previously been proposed to be associated with PD risk at sub-genome-wide significance by Simon-Sanchez et al [13]. After obtaining and meta-analyzing GWAS data from ∼1,400 additional SNPs in this region derived from all Caucasians GWAS datasets [10], [12], [13], [15]–[19], [21], [22], rs7077361 remained the most significantly associated SNP in this region (Figure S3). In addition to using random-effects models, we also performed exploratory fixed-effect meta-analyses on all eligible polymorphisms. These analyses did not reveal genome-wide significant effect sizes for any additional locus, except ACMSD/TMEM163 (most significant SNP rs6723108, OR 0. 91, P = 1. 3×10−9, I2 = 46% [95% CI 0–73%], Figure S4, panel 1) and HLA (most significant SNP chr6: 32609909, OR 0. 78, P = 8. 8×10−15, I2 = 84% [95% CI 70–91%], Figure S4, panel 2), both of which were reported to be associated with PD risk at genome-wide significance in previous work [16], [21]. In both instances, the lack of genome-wide significance in the random-effects models (Table S1) was due to relatively pronounced heterogeneity of effect estimates across studies. However, the heterogeneity across the 11 datasets in the ACMSD/TMEM163 meta-analysis is almost entirely due to variance of effect size estimates in the same direction (see Figure S4, panel 1), making it likely that ACMSD/TMEM163 represents a genuine PD risk locus. For the SNP tested in the HLA locus (chr6: 32609909, Figure S4, panel 2), heterogeneity is more pronounced and more complex owing to ORs on either side of 1. This could be due to a number of reasons, e. g. subtle and uncorrected population substructure and/or different LD patterns between the analyzed SNP and the actual functional variant (s) [16]. Thus, although the evidence is currently not as conclusive as for ACMSD/TMEM163 it still appears quite possible that there is one or more PD association signals in the HLA region. Regardless of these considerations, additional data are needed to more firmly assess the role of both loci in contributing to PD susceptibility. SNCA, LRRK2, BST1, and PARK16 show evidence for genome-wide significance in meta-analyses restricted to Caucasian and Asian populations (Table 2). Furthermore, data obtained from the GEO-PD consortium [23] suggest that the effect estimates for some of the recently discovered PD loci (i. e. CCDC62/HIP1R, MCC1, and STK39) [21] may be comparable in Caucasian and Asian populations (Table S1), although additional datasets are needed to establish genome-wide significance in populations of Asian-descent for these loci. Conversely, only insufficient data are currently available to assess the effect sizes of GAK and SYT11/RAB25 on PD risk in Asians: GAK rs6599388 violated Hardy-Weinberg equilibrium in Asian datasets from the GEO-PD consortium and was thus excluded from further analyses on that ethnic group [23]. SYT11/RAB25 chr1: 154105678 was excluded from all analyses due to technical reasons in the study by the GEO-PD consortium [23]. Moreover, none of the reported SYT11/RAB25 and GAK SNPs from the recent GWAS meta-analysis [21] were captured directly or by proxy (with an r2≥0. 8) in the Japanese GWAS dataset [14], [23]. Finally, Asian-descent populations cannot be appropriately assessed for PD association with the MAPT-H1/H2 haplotype, rs10928513 in ACMSD, and rs7077361 in ITGA8 owing to monomorphicity at these sites [14], [23]. To estimate the epidemiologic credibility of associations with polymorphisms showing sub-genome-wide significant association with PD (P>5×10−8), we applied two “credibility” measures for each such result. First, we calculated Bayes factors (BF, expressed here as log10-values, “logBF”) assuming an average non-null odds ratio of 1. 15, as approximation of a typical “complex disease effect size”, and a spike and smear prior distribution of effects [25]. Our second assessment was based on the Human Genome Epidemiology Network' s (HuGENet) interim criteria for the assessment of cumulative epidemiologic evidence in genetic association studies [26], [27]. The results of these analyses are summarized in Table S1. There was strong epidemiologic support in both assessments for all loci showing genome-wide significant association. This included several additional polymorphisms in these same loci that only showed sub-genome-wide significant association. However, there was no additional sub-genome-wide significantly associated locus that received unequivocally strong support from both credibility assessments (Table S1). In this list, the strongest support was assigned to SNP chr6: 32588205 in the HLA locus receiving the best possible grade in the HuGENet criteria (grade A), but more moderate support in the Bayesian analyses (logBF = 4. 4). However, the relevance of this assessment needs to be evaluated as the underlying analysis was only based on four GWAS datasets. The PDGene database represents a comprehensive, regularly updated and freely available online research synopsis of genetic association studies in PD. Detailed summaries of the most compelling findings are provided within an easy-to-use, dedicated online framework, displaying forest plots, cumulative meta-analyses, and an up-to-date ranking of “Top Results”. To allow comparison of PDGene results with association findings from other complex diseases and to facilitate their interpretation with respect to functional genetics data, all meta-analysis results have been ported as a customized track onto the UCSC Genome Browser. This will also allow for a integration and visualization [28] of association results from large-scale resequencing data (e. g. from whole-exome or whole-genome studies) into PDGene once these become available. To the best of our knowledge, our study represents the most comprehensive research synopsis in the field of PD genetics. In addition, it represents the first disease-specific genetic database that allows a systematic and exhaustive inclusion of GWAS data, and may serve as a model for similar databases in other complex genetic diseases. Owing to our multi-pronged data retrieval and analysis protocol we were able to perform meta-analyses on the vast majority of PD risk-gene candidates, including those “featured” as top association results in all published GWAS. In particular, this includes the five novel loci recently featured in the recent GWAS meta-analysis [21]. Through collaboration with other PD genetics laboratories we obtained independent summary data for these and 142 additional SNPs, substantially extending the hitherto available evidence. Taken together, our analyses provide unequivocal evidence that BST1, CCDC62/HIP1R, DGKQ/GAK, GBA, ITGA8, LRRK2, MAPT, MCCC1/LAMP3, PARK16, SNCA, STK39, SYT11/RAB25 represent genuine PD risk loci, while the role of several other loci (e. g. ACMSD/TMEM163, and the HLA locus) remains to be determined. The unpublished data aggregated here from various PD genetics groups for selected candidate genes represents the first step towards a systematic meta-analysis across the full GWAS datasets from the same populations. Once completed, the results of this “mega” meta-analysis will be posted on the PDGene database, allowing users to browse the complete results via the customized genome browser track already in place. Of particular interest are loci with unusually large effect sizes. While most loci in PDGene have only small effects on PD risk (with ORs ranging from 1. 10 to 1. 35, which are typical for complex diseases), for some loci much larger ORs were estimated (i. e. GBA [OR 3. 51 in Caucasians], LRRK2 [OR 2. 23 in Asians], and SYT11/RAB25 [OR 1. 73 in Caucasians], see Table 2). The risk-allele frequencies at these polymorphisms are typically rather small (i. e. below 0. 05), resulting in low population attributable risks for these loci (for the above mentioned loci individually less than 2%). Interestingly, the meta-analysis results of GBA N370S as well as the LRRK2 rs34778348 are solely based on candidate-gene approaches since these SNPs are not on any of the current GWAS arrays or imputation reference panels. Thus, even in the “GWAS era” smaller-scale, non-GWAS but “focused” genetic studies, will likely continue to play an important role. This is also true when it comes to providing independent replication of proposed disease associations and/or when validating imputation-derived results by direct genotyping in sufficiently sized datasets. PDGene systematically concatenates all these different types of data into one database framework, vastly facilitating an assessment of the overall evidence for any given SNP or locus. The strength of our approach is further exemplified by the identification of genome-wide significant association between disease risk and a SNP in ITGA8, which was not featured as a relevant PD gene in any previous study. ITGA8 (encoding integrin alpha 8, a type-I transmembrane protein) is functionally interesting as it is expressed in brain [29], mediates cell-cell interactions and regulates neurite outgrowth of sensory and motor neurons [30]. Additional studies are needed to further assess the potential role of this gene in PD pathogenesis. Furthermore, PDGene shows that two additional loci, not highlighted by the recent GWAS meta-analysis [21], yield genome-wide signficiant results in the PDGene meta-analyses, i. e. PARK16, originally implicated as a PD susceptibility locus in an Asian GWAS [14] but not highlighted in the recent GWAS meta-analysis on Caucasian samples [21] and GBA, a gene that was found soley by candidate-gene approaches. Another strength of our study is that it combines genetic data from currently more than 50 different countries allowing a systematic assessment of genetic associations across populations of different ethnic descent. For instance, these analyses suggest that variants in BST1, LRRK2, the PARK16 locus, and SNCA show genome-wide significant association with PD risk in both Caucasian and Asian-descent samples. Furthermore, the recently described Caucasian GWAS loci CCDC62/HIP1R, MCC1, and STK39 [21] also show similar effect size estimates in populations of Asian-descent [23]. PD association data originating from other ethnic groups are still relatively scarce. However, they could easily be added to the already existing data on the respective polymorphisms available on PDGene. In summary, we have created a continuously updated online resource for genetic association studies in the field of PD. Synthesizing essentially all available data in the field led to the identification of ITGA8 as a novel potential PD risk locus. Our quantitative approach to data integration across a multitude of different study designs can be readily scaled to include large-scale resequencing efforts that will emerge over the coming years, making the complex field of PD genetics accessible to a broad range of investigators. After completion of all data-management and analysis steps, all study-specific variables, genotype data (except for GWAS), and meta-analysis plots are posted on a dedicated, publicly available, online adaptation of the PDGene database using the same software and code as our databases for Alzheimer' s disease [33] and schizophrenia [34]. The online database is hosted by the “Alzheimer Research Forum” and can be accessed via its own designated URL (http: //www. pdgene. org). The database software can easily be ported to other genetically complex diseases and will be made available on a collaborative basis to interested researchers upon request.

The genetic basis of Parkinson' s disease is complex, i. e. it is determined by a number of different disease-causing and disease-predisposing genes. Especially the latter have proven difficult to find, evidenced by more than 800 published genetic association studies, typically showing discrepant results. To facilitate the interpretation of this large and continuously increasing body of data, we have created a freely available online database (“PDGene”: http: //www. pdgene. org) which provides an exhaustive account of all published genetic association studies in PD. One particularly useful feature is the calculation and display of up-to-date summary statistics of published data for overlapping DNA sequence variants (polymorphisms). These meta-analyses revealed eleven gene loci that showed a statistically very significant (P<5×10−8; a. k. a. genome-wide significance) association with risk for PD: BST1, CCDC62/HIP1R, DGKQ/GAK, GBA, LRRK2, MAPT, MCCC1/LAMP3, PARK16, SNCA, STK39, SYT11/RAB25. In addition and purely by data-mining, we identified one novel PD susceptibility locus in a gene called ITGA8 (rs7077361, P = 1. 3×10−8). We note that our continuously updated database represents the most comprehensive research synopsis of genetic association studies in PD to date. In addition to vastly facilitating the work of other PD geneticists, our approach may serve as a valuable example for other complex diseases.

Abstract Introduction Results Discussion Methods

medicine public health and epidemiology epidemiology neurology neurological disorders

2012

Comprehensive Research Synopsis and Systematic Meta-Analyses in Parkinson's Disease Genetics: The PDGene Database

APOBEC3 (A3) family proteins are DNA cytosine deaminases recognized for contributing to HIV-1 restriction and mutation. Prior studies have demonstrated that A3D, A3F, and A3G enzymes elicit a robust anti-HIV-1 effect in cell cultures and in humanized mouse models. Human A3H is polymorphic and can be categorized into three phenotypes: stable, intermediate, and unstable. However, the anti-viral effect of endogenous A3H in vivo has yet to be examined. Here we utilize a hematopoietic stem cell-transplanted humanized mouse model and demonstrate that stable A3H robustly affects HIV-1 fitness in vivo. In contrast, the selection pressure mediated by intermediate A3H is relaxed. Intriguingly, viral genomic RNA sequencing reveled that HIV-1 frequently adapts to better counteract stable A3H during replication in humanized mice. Molecular phylogenetic analyses and mathematical modeling suggest that stable A3H may be a critical factor in human-to-human viral transmission. Taken together, this study provides evidence that stable variants of A3H impose selective pressure on HIV-1. Apolipoprotein B mRNA editing enzyme catalytic polypeptide-like 3 (APOBEC3; A3) enzymes are cellular single-stranded DNA cytosine deaminases that are specifically encoded in mammals [1,2]. Rodents including mice (Mus musculus) have a single A3 gene, while primates including humans (Homo sapiens), chimpanzees (Pan troglodytes) and Old World monkeys have seven A3 paralogous genes (A3A, A3B, A3C, A3D, A3F, A3G and A3H). Gene duplication is a hallmark of the genes that are under evolutionary selective pressures [3], and indeed, the seven primate A3 genes have been positively selected during evolution [4], These observations suggest that primate A3 proteins play crucial roles in primates including humans. Human A3G was discovered first and was shown to be capable of restricting the replication of human immunodeficiency virus type 1 (HIV-1) in an in vitro cell culture system [5]. Subsequent investigations revealed that several human A3 family proteins exhibit the ability to reduce HIV-1 infectivity [2,6–8]. Moreover, previous studies including ours have demonstrated that A3D, A3F, and A3G, which are endogenously expressed in human CD4+ T cells, are restriction factors potently controlling HIV-1 replication in human CD34+ hematopoietic stem cell (HSC) -transplanted humanized mouse models [9–12]. To antagonize the anti-viral effect of A3 proteins, HIV-1 encodes a protein named viral infectivity factor (Vif). Vif orchestrates cellular ubiquitin ligase complex and degrades anti-viral A3 proteins via ubiquitin/proteasome-dependent pathway in infected cells [2,13]. In addition to A3D, A3F and A3G, human A3H is known as a potent restriction factor against HIV-1. Human A3H is polymorphic and has seven haplotypes [14,15]. Three of them, called haplotypes II, V, and VII, produce stably expressed enzymes that exhibit anti-HIV-1 activity in model cell culture experiments as well as primary T lymphocytes ex vivo [14–16]. In contrast, the other three haplotypes (III, IV, and VI) do not exhibit detectable protein expression [14–16]. Additionally, our recent study has demonstrated that A3H haplotype I (A3H-I) has intermediate stability and clear enzymatic activity [17] (Fig 1A). Importantly, the frequency of each haplotype differs among human population, with a higher frequency of stable A3H in the African-descendant population [14,15]. Furthermore, it is more intriguing that the Vif proteins of certain HIV-1 strains are unable to counteract stable A3H haplotypes, and the ability of Vif to antagonize stable A3H is determined by at least two residues at positions 39 and 48 (Fig 1B) [18–20]. These observations suggest that both the A3H-mediated anti-viral effect and the antagonistic ability of Vif against A3H are co-mingled in the human population, in contrast to the functional relationships between Vif and A3D, A3F and A3G, which appear much less variable. However, the robustness of the effects of stable/intermediate A3H haplotypes on viral replication at an individual scale and a population scale remains unclear, and the dynamics by which HIV-1 may circumvent and/or counteract the anti-viral effect of stable A3H is yet to be addressed. Here we use an HSC-transplanted humanized mouse model to demonstrate that stable A3H, but not intermediate A3H, which is endogenously expressed in human CD4+ T cells, is a bona fide restriction factor capable of controlling HIV-1 replication in vivo. In addition, we reveal that HIV-1 Vif readily acquires the ability to counteract stable A3H during viral expansion in vivo. Additionally, we use molecular phylogenetic analysis and mathematical modeling to further address the impact of stable A3H on HIV-1 epidemics. Our analyses suggest that stable A3H may control HIV-1 dissemination in both intra- and inter-individual scales. To address the impact of endogenous A3H haplotypes (Fig 1A) on HIV-1 replication in vivo, two derivatives of the replication-competent CCR5-tropic HIV-1 strain NLCSFV3 [21] were made with differing A3H haplotype II (A3H-II) neutralization capabilities [20]. One virus encodes a Vif protein that is able to counteract stable A3H (" hyper Vif" ), while the other encodes a Vif protein that does not (" hypo Vif" ) (Fig 1B) [20]. Importantly, previous reports demonstrated that the Vif' s ability to degrade stable A3H is determined by the two amino acid residues at positions 39 and 48 (Fig 1B) [18–20]. Consistent with a prior study [20], hyper HIV-1 fully counteracted the anti-viral activity mediated by A3H-II, whereas hypo Vif was not able to counteract A3H-II (Fig 1B). In the absence of A3H-II, the infectivity of both of these HIV-1 molecular clones is similar (Fig 1B). To investigate the impact of endogenous A3H on HIV-1 replication in vivo, a series of hyper versus hypo Vif competition experiments was conducted in humanized mice reconstituted with stable A3H-expressing HSCs. The first experiment used eight humanized mice, which were heterozygous for stable A3H haplotypes (S1 Table): two out of the eight had blood cell compartments reconstituted with haplotypes III and V cells, and the other six mice expressed haplotypes I and II (S1 Table). Next, these eight mice were intraperitoneally co-inoculated with equal amounts of hyper and hypo viruses (1,500 TCID50 each; Fig 1C), and the amount of viral RNA in the plasma and the level of human CD4+ T cells in the peripheral blood (PB) were routinely analyzed for 6 weeks post-infection (wpi). HIV-1 efficiently expanded in the humanized mice, as observed in our previous studies [11,12,22–24], and the level of peripheral CD4+ T cells was significantly reduced compared to mock-infected mice (Fig 1D). At 6 wpi, viral RNA was extracted from the plasma of infected mice and the sequences of the vif gene were analyzed. As anticipated, hyper vif and its derivatives were able to outcompete hypo vif virus in mice expressing stable A3H (73. 1% ± 7. 7% [286/391] in Fig 1E; see also S2 Table). However, hypo vif-related sequences were the majority in infected mouse no. 5 and were still present at significant levels in all animals (no. 5,67. 3% [35/52] to no. 1,7. 9% [3/38]; Fig 1E & S2 Table). These results raised the possibility that these hypo vif viruses may have adapted in vivo and gained a better ability to counteract stable A3H during viral replication. To address this idea, we subcloned the 13 hypo vif-related open reading frames (ORFs) into the expression plasmid and evaluated their anti-stable A3H activity using in vitro cell culture system. As shown in Fig 1F, adapted Vif proteins with V39F/N48H (26 clones from 4 mice), V39F/N48H/D113N (1 clone from 1 mouse) and V39F/N48H/L148S (1 clone from 1 mouse) mutations, degraded A3H-II and impaired the A3H-II packaging into the released viral particles. Additionally, the HIV-1 restriction capacity of A3H-II was significantly counteracted by these 3 adapted hypo Vif derivatives as evidenced by hyper Vif levels of infectivity (Fig 1G). We verified that these 3 hypo Vif derivatives as well as parental hypo Vif were active in counteracting other HIV-1 relevant A3s such as A3D, A3F and A3G (S1 Fig). Together, the sequencing results and the tests of the functionality of the adapted hypo Vif proteins indicated that 80. 3% ± 4. 8% (314/391) of the vif sequences in the plasma of infected stable A3H mice are able to counteract stable A3H (Fig 1E–1G & S2 Table). These findings indicate that the ability to antagonize stable A3H is required for efficient HIV-1 replication in humanized mice. We next investigated whether HIV-1 undergoes selection as a result of pressure from A3H-I in vivo. Six humanized mice were reconstituted with HSCs from three individual donors. Five out of the six mice were homozygotes for A3H-I and one mouse was heterozygous for A3H haplotypes I and VI (S1 Table). These six mice were co-inoculated with hyper and hypo HIV-1s (Fig 2A). All mice exhibited a high level of viremia and a declined level of peripheral CD4+ T cells (Fig 2B). We then analyzed the vif sequences in the plasma of these six infected mice at 6 wpi. In contrast to the observations from animals with at least one copy of stable A3H (hap II or V; Fig 1E), the proportion of hyper/hypo vif sequences varied in each infected mouse, and no obvious replication biases were observed (Fig 2C & S3 Table). On average, the percentage of hyper and hypo vif-derived sequence were similar, 46. 5% ± 13. 7% (127/273) and, 53. 5% ± 13. 7% (146/273), respectively (S4 Table). To assess whether A3H genotype affects the expression level of other HIV-1 relevant A3 genes, we analyzed the expression levels of A3D, A3F and A3G in the splenic CD4+ T cells of humanized mice. Consistent with our previous studies with primary CD4+ T cells ex vivo and with humanized mice [24,25], the mRNA expression levels of these A3 genes in HIV-1-infected mice were significantly higher than those in mock-infected mice (S2A Fig). The expression levels of these A3 genes were comparable between the humanized mice expressing stable A3H and intermediate A3H (S2B Fig), indicating that the A3H genotype is not associated with the expression levels of other HIV-1 relevant A3 genes. In addition to A3H, single nucleotide variants (SNVs) in human A3G [26–28], A3F [29,30] and A3D [31] have been reported. The variants of A3G [28] and A3D [31] are degraded efficiently by HIV-1 Vif and are therefore unlikely to play significant roles in vivo. In contrast, An et al. have recently reported that an SNV of A3F, V231I, confers partial resistance to Vif-mediated degradation by certain strains of HIV-1 [30]. To address the possibility that A3F V231I mutant affects viral growth and the sensitivity to hyper/hypo HIV-1, we assessed the genomic sequences of A3F. However, this A3F SNV was not detected in the human cells used in our studies (data not shown). Altogether, these findings suggest that no specific selective pressure is elicited against either hyper or hypo HIV-1 in intermediate A3H humanized mice and that virus expansion is occurring in a stochastic manner. Co-infection studies revealed that hyper Vif HIV-1 dominates over hypo Vif virus in animals humanized with stable A3H expressing cells (Fig 1). These observations suggest that the stable A3H protein, which is expressed endogenously in human CD4+ T cells, exhibits a robust anti-viral effect and impairs the expansion of the viruses without full A3H counteraction abilities (i. e. , V39 hypo Vif). We used three HIV-1 strains, NLCSFV3, JRCSF and AD8 to demonstrate that the Vif proteins of these viruses are unable to antagonize A3H-II (S3 Fig). We then inoculated these viruses into 50 humanized mice reconstructed from 16 HSC donors (Fig 3A). The genotyping PCR revealed that 13 out of the 16 HSC donors encode A3H-I, and 3 donors possessed one stable A3H allele (S5 Table). Based on A3H haplotypes, these infected mice were classified into two groups, intermediate A3H (n = 37) and stable A3H (n = 13), and the level of peak viral load in each group was compared. As shown in Fig 3B, surprisingly, the peak viral load was comparable between intermediate A3H mice and stable A3H mice (P = 0. 92 by Mann-Whitney U test). Because certain hypo vif derivatives acquired anti-stable A3H activity in the stable A3H mice co-inoculated with hyper/hypo HIV-1 (Fig 1E–1G), we hypothesized that these viruses acquired de novo resistance to stable A3H in vivo (strains NLCSFV3, JRCSF, or AD8; Fig 3B). To test this hypothesis, we analyzed the vif sequences in the plasma of the 4 stable A3H mice infected with HIV-1 (strain NLCSFV3) at 6 wpi. Notably, some Vif sequences were commonly detected in the 4 stable A3H mice infected with NLCSFV3 (Fig 3C; the raw data and mutation matrix are shown in S4 Fig). To investigate whether these mutant variants acquired anti-stable A3H activity de novo, we prepared the expression plasmids of these Vif derivatives and conducted in vitro experiments using our cell culture system. As shown in Fig 3D & 3E, we detected 2 Vif variants, N48H and V13I/N48H/GDAK60-63EKGE, that are able to antagonize A3H-II at the level observed for hyper Vif. In summary, 59. 7% ± 5. 1% (145/243) of the vif sequences in the plasma acquired the ability to counteract stable A3H (S6 Table), but such mutants were not detected in the intermediate A3H mice infected, solely, with NLCSFV3 (S5 Fig & S7 Table). Taken together, these findings suggest that the ability of HIV-1 Vif to antagonize stable A3H is acquired de novo during viral expansion in vivo. We previously reported that endogenous A3H mRNA expression levels in primary human CD4+ T cells are significantly lower than those of anti-viral A3 genes such as A3D, A3F, and A3G, and that activation and/or infection stimuli induces higher A3H expression [25,32]. In agreement with these prior works, the activation stimuli driven by anti-CD3/CD28 antibodies induced the expression of CD25, a marker of activated human CD4+ T cells (S6A Fig), and also A3H mRNA expression levels (P = 0. 010 by paired t test, S6B Fig). However, it should be noted that human CD4+ T cells in humanized mice [12] and human PB [33] are less activated (S6C Fig) and in a quiescent state. In this regard, previous studies reported that HIV-1 infection induces the activation of CD4+ T cells of infected individuals [34,35]. Therefore, we hypothesized that HIV-1 infection induced CD4+ T-cell activation and augmented A3H expression in humanized mice, and this resulted in robust anti-viral effect by endogenous A3H (Figs 1 & 3). To investigate the immune activation status in detail, we performed RNA sequencing (RNA-seq). Human mononuclear cells (MNCs) were isolated from the spleen of 4 HIV-1-infected mice and 4 mock-infected mice at 6 wpi, and RNA-seq analyses were conducted. As shown in Fig 4A, 93 genes were significantly up-regulated by HIV-1 infection, whereas 16 genes were down-regulated. Parametric gene set enrichment analysis (GSEA) revealed that the genes associated with T-cell/lymphocyte activation, inflammatory response, and positive regulation of T cell activation were significantly up-regulated in the human MNCs of HIV-1-infected mice (Fig 4B; the GESA result is listed in S8 Table). In addition, various interferon-stimulated genes such as RSAD2 (encoding Viperin), DDX58 (encoding RIG-I), EIF2AK2 (encoding PKR), MX1, ISG15, MOV10 and BST2 (encoding tetherin) were up-regulated in HIV-1-infected mice (Fig 4A). As observed in infected patients [34,35], our findings suggest that HIV-1 infection triggers immune activation in humanized mouse model. We then addressed the possibility that the immune activation caused by HIV-1 infection (Fig 4A & 4B) leads to the up-regulation of A3H in humanized mice. As shown in Fig 4C, the proportion of the splenic CD4+ T cells (CD3+ CD8− cells) of infected mice was significantly lower than that of uninfected mice, and particularly, CD25+ activated CD4+ T cells were severely depleted by HIV-1 infection (P = 0. 0039 versus mock infection). Consistent with our previous findings [12], HIV-1 infection led to the severe depletion of activated CD4+ T cells in humanized mice. Next, we sorted the fractions of non-activated CD4+ T cells (CD45+ CD3+ CD8− CD25− cells) and activated CD4+ T cells (CD45+ CD3+ CD8− CD25+ cells) of mock-infected mice (S7 Fig) and analyzed the mRNA expression level of A3H in each population by real-time RT-PCR. In mock-infected mice, A3H expression in the activated CD4+ T cells was significantly higher than that in non-activated cells (P = 0. 0090 by Mann-Whitney U test; Fig 4D). This finding further suggests that the CD4+ T-cell activation augments A3H expression, as observed in in vitro experiments (S6B Fig) and in our previous reports [25,32]. Because CD25+ CD4+ T cells were severely depleted in infected mice (Fig 4C), we sorted only the fraction of CD25− CD4+ T cells (CD45+ CD3+ CD8− CD25− cells) of HIV-1-infected mice for real-time RT-PCR. Interestingly, the A3H expression level in the CD25− CD4+ T cells of HIV-1-infected humanized mice was significantly higher than that of CD25− CD4+ T cells of uninfected mice (P = 0. 0062 by Mann-Whitney U test; Fig 4D). Altogether, these findings suggest that the immune activation triggered by HIV-1 infection augments A3H expression in CD4+ T cells of infected humanized mice. Finally, we addressed how hyper and hypo HIV-1 sequences circulate in the human population. The HIV-1 Vif sequences were obtained from the Los Alamos National Laboratory HIV-1 sequence database (https: //www. hiv. lanl. gov/components/sequence/HIV/search/search. html). Fig 5A shows a phylogenetic tree of Vif sequences of HIV-1 group M (n = 2,976), which is a pandemic strain worldwide. The phylogenetic tree indicates that Vif sequences cluster based on subtype (Fig 5A). Interestingly, the sequences of hyper Vif (here we defined" hyper Vif" as a sequence that possesses F or Y at position 39 and H at position 48) scattered in this tree and did not form a unique cluster (Fig 5A). Additionally, the percentage of hyper Vif varied in each subtype (Fig 5B), suggesting that hyper and hypo Vif mutually swap in human population. To evaluate the counteracting ability of HIV-1 clinical isolates (group M) against stable A3H, we used 15 infectious molecular clones (IMCs): 8 subtype B and 7 subtype C; 10 transmitted/founder (TF) viruses and 5 chronic control (CC) viruses. As shown in Fig 5C, the infectivity of the 3 IMCs (strains AD17, RHPA and MCST) was significantly suppressed by A3H-II with statistical differences, suggesting that these viruses have established new infection as TF viruses in individuals without anti-stable A3H activity. In contrast, other 12 IMCs overcame A3H-II-mediated restriction (Fig 5C), suggesting that these viruses exist in human population as hyper HIV-1. Importantly, the anti-stable A3H ability of these IMCs corresponded well to the amino acid residues positioned at 39 and 48 (S9 Table). These findings suggest that anti-stable A3H ability is not a necessary requirement for certain viruses circulating within individuals. We then assessed the anti-stable A3H ability of non-pandemic HIV-1 groups N, O, and P. As shown in Fig 5B, the proportion of hyper Vif sequences varied in each group. In particular, HIV-1 group O strains (n = 51) did not encode a hyper vif sequences. However, the cell-based experiments demonstrated that the IMCs of group O (strains BCF183 and RBF206) overcame A3H-II-mediated anti-viral effect (Fig 5C). The IMC of HIV-1 group P (strain RBF168) also counteracted A3H-II, while that of HIV-1 group N (strain DJO0131) did not (Fig 5C). Interestingly, in contrast to the results of HIV-1 group M, the anti-stable A3H ability of HIV-1 group O Vif was not governed by the two residues at positions 39 and 48 (S9 Table) These findings suggest that other residues than those positioned at 39 and 48 determine the ability of Vif proteins of HIV-1 group O to counteract stable A3H. Furthermore, we assessed the correlation between the frequency of hyper HIV-1 and the proportion of the individuals harboring stable A3H haplotype worldwide. The HIV-1 Vif sequences were obtained from HIV-1 sequence database (S10 Table), and the information of A3H haplotype was obtained from the 1000 Genomes Project (http: //www. internationalgenome. org) [36] (S11 Table). As shown in Fig 5A & 5B, the Vif sequences are highly diversified and the logoplot (S8 Fig) further indicated that the amino acids at position 39 and 48 were not highly conserved when compared to the YRHHY motif, which is essential for A3G degradation [24,37]. Additionally, consistent with previous reports [14,15,20], both the percentage of hyper Vif and the proportion of stable A3H haplotype were highest in Africa, particularly in Nigeria (Fig 5D & S12 Table), and these two parameters were correlated each other with statistical significance (Spearman' s r = 0. 720, P = 0. 017 by Spearman rank correlation test; S9 Fig). To further investigate the relationship between hyper HIV-1 and stable A3H haplotype, we conducted a mathematical simulation. As shown in Fig 5E, the frequency of hyper HIV-1 increased dependent on the proportion of the people harboring stable A3H haplotype. Taken together, our analyses at a human population level suggest that stable A3H elicits a selective pressure against HIV-1, and that HIV-1 overcomes stable A3H-mediated anti-viral immunity by acquiring the ability to counteract stable A3H. In this study, we used a humanized mouse model to show that HIV-1 infection induces immune activation and augments the expression of endogenous A3H in human CD4+ T cells (Fig 4). We also showed that the ability of HIV-1 Vif to counteract stable A3H-mediated anti-viral effect is crucial for efficient viral expansion in vivo when endogenous A3H is expressed stably (Figs 1 & 3). In contrast, the ability of HIV-1 Vif to counteract stable A3H is dispensable when stable A3H is absent in vivo (Fig 2). Furthermore, we addressed the significance of the stable A3H-mediated anti-viral effect on HIV-1 dissemination in human populations using molecular phylogenetic analysis and mathematical modeling. The occurrence of hyper Vif variants and stable A3H haplotypes correlates worldwide, suggesting that the ability of Vif to antagonize stable A3H was acquired during viral spread throughout the human population (Fig 5). These findings suggest that the A3H polymorphism influences HIV-1 dissemination at individual and population levels. In the stable A3H humanized mice co-inoculated with hyper and hypo HIV-1 infectious clones, several hypo Vif viruses acquired V39F and N48H changes, which resulted in the gain-of-function to counteract A3H-II (Fig 1E–1G). Given that these two amino acids are identical to those in hyper Vif, the emergence of the hypo Vif derivatives, which can potently antagonize A3H-II in the hyper/hypo HIV-1 co-inoculated stable A3H mice (Fig 1), may be due to the recombination between hypo and hyper Vif sequences. However, the results shown in Fig 3 argue against this possibility. We demonstrated that some Vif sequences with the ability to antagonize stable A3H emerge during viral replication in the stable A3H mice within only 6 weeks. In contrast, in the intermediate A3H mice co-inoculated with hyper and hypo HIV-1 clones, the hyper or hypo Vif viruses expanded randomly with no evidence of selection on vif (Fig 2). These findings suggest that the stable A3H, which is endogenously expressed in CD4+ T cells, has a robust anti-viral activity in vivo and that it is feasible for Vif to acquire the counteracting ability against stable A3H de novo. We favor a model in which the starting hypo Vif virus is constrained evolutionarily, likely by needing to counteract A3D, A3F, and A3G, and that de novo (not recombination mediated) amino acid substitutions at positions 39 and 48 provide the most efficient route to optimize anti-A3H activity. Moreover, it is important to note that all the stable A3H humanized mice used in this study were heterozygous for A3H stability (S1 & S5 Tables). It appears that an allele of stable A3H is sufficient to induce a robust selective pressure against HIV-1. In sharp contrast to the findings in the stable A3H mice co-inoculated with hyper and hypo HIV-1s (Fig 1), hyper HIV-1 was not commonly selected in the intermediate A3H mice co-inoculated with hyper and hypo HIV-1 clones (Fig 2). Also, de novo emergence of hyper Vif was not detected in the intermediate A3H mice infected with NLCSFV3 (S5 Fig). On the other hand, we recently showed that the intermediate A3H (A3H-I) is enzymatically active and contributes to breast and lung cancer mutagenesis despite being expressed at lower levels compared to its stable A3H counterpart [17]. These findings suggest that A3H-I, which is endogenously expressed in human CD4+ T cells, is not sufficient to impose selective pressure on HIV-1 replication in vivo. Here we detected the emergence of Vif sequences that acquired the ability to antagonize stable A3H (Figs 1 & 3). In contrast, in the humanized mice infected with a vif-mutated HIV-1 (designated 4A HIV-1), which is sensitive to A3D and A3F, we have previously demonstrated that Vif sequences with the ability to antagonize A3D and A3F do not emerge [24]. We confirmed the absence of Vif revertants in the plasma of two 4A-HIV-1 infected mice infected at 6 wpi (S10 Fig). These observations suggest that HIV-1 is able to overcome the restriction mediated by stable A3H but not by A3D and A3F during viral replication in vivo. Two nonexclusive models may explain the observed differences. One possibility is that it might be more feasible for Vif to overcome stable A3H-mediated restriction than A3D/A3F because the anti-viral activity of endogenous stable A3H is lower than those of endogenous A3D and A3F. However, at least four previous studies have demonstrated that the anti-HIV-1 activity of stable A3H (haplotype II) is similar to that of A3F and is higher than that of A3D [18,19,25,38] and argue against this possibility. In addition, it should be noted that the endogenous expression levels of the respective A3 genes in primary human CD4+ T cells are different from each other. Indeed, Refsland et al. have revealed that endogenous expression levels of A3D and A3F mRNAs are higher than that of A3H in primary CD4+ T cells [32]. Another possibility is the number of amino acids responsible for A3 counteraction: only two amino acids at positions 39 and 48 are responsible for counteracting stable A3H [18–20], while there are four that are responsible for counteracting A3D and A3F (known as DRMR motif at position 14–17) [37,39]. The emergence of Vif revertants harboring the ability to counteract stable A3H is reminiscent of the observations that the sub-optimal drug concentrations facilitate the emergence of drug-resistant viruses in infected patients [40,41]. In fact, it appears difficult for Vif to acquire the ability to counteract A3F and A3G de novo during viral replication in cell cultures [42,43] and a humanized mouse model [11,24]. In contrast, previous studies have successfully selected the viruses that acquired the ability to counteract stable A3H in the in vitro culture infection experiments using the human CD4+ T cell lines such as MT-4 cells [18] and SupT11 cells [20] that ectopically express A3H-II. Here we demonstrated that HIV-1 infection induces immune activation in humanized mice, as observed in infected individuals [34,35], and augments the expression of endogenous A3H in the human CD4+ T cells of infected mice (Fig 4). But still, the anti-HIV-1 activity of endogenous stable A3H is not sufficient to control viral expansion in vivo, and therefore, Vif may easily acquire the ability to counteract the restrictive activity of endogenous A3H. Our findings in infected humanized mice revealed that hyper HIV-1 is predominantly selected in the mice expressing stable A3H (80. 3% ± 4. 8%; S2 Table), while the viruses replicated in the mice with intermediate A3H were selected stochastically (46. 5% ± 13. 7%; S4 Table). We also demonstrated the de novo emergence of hyper HIV-1 in the stable A3H mice infected with NLCSFV3 (59. 7% ± 5. 1%; S6 Table). Based on these findings and numerical parameters, we investigated the dynamic effect of A3H haplotypes on HIV-1 epidemic in the human population through molecular phylogenetic and mathematical modeling and revealed that the occurrence of hyper Vif and stable A3H variants are correlated positively in the human population (Fig 5). This suggests that stable A3H may not just provide an intrinsic immunity at the level of individual patients, as elaborated here in humanized mice, but it may also function to control the dissemination of hypo HIV-1 isolates in the human population [44,45]. All procedures including animal studies were conducted following the guidelines for the Care and Use of Laboratory Animals of the Ministry of Education, Culture, Sports, Science and Technology, Japan. The authors received approval from the Institutional Animal Care and Use Committees (IACUC) /ethics committee of the institutional review board of Kyoto University (protocol number D15-08). All protocols involving human subjects were reviewed and approved by the Kyoto University institutional review board. All human subjects were provided written informed consent from adults. NOG mice (NOD/SCID/Il2r KO mice) [46] were obtained from the Central Institute for Experimental Animals (Kawasaki, Kanagawa, Japan). The mice were maintained under specific-pathogen-free conditions and were handled in accordance with the regulation of the IACUC/ethics committee of Kyoto University. Human CD34+ hematopoietic stem cells (HSCs) were isolated from human fetal liver as previously described [47]. The humanized mouse model (NOG-hCD34 mouse) was constructed as previously described [11,22,23,48–50]. In the experiments shown in Figs 1 & 2,14 newborn (aged 0 to 2 days) NOG mice from 7 litters were irradiated with X-ray (10 cGy per mouse) using an RX-650 X-ray cabinet system (Faxitron X-ray Corporation) and were then intrahepatically injected with the human fetal liver-derived CD34+ cells (1. 0 × 105 to 2. 3 × 105 cells; 5 donors). A list of the humanized mice used in this study is summarized in S1 Table. In the experiments shown in Fig 3, the 35 NOG-hCD34 mice infected with HIV-1 were used in our previous studies [12,23,24] (Fig 3) and the 15 NOG-hCD34 mice were newly infected with HIV-1. These humanized mice were constructed using 16 independent HSC donors with 29 NOG litters (summarized in S5 Table). HEK293T cells (a human embryonic kidney 293 T cell line; ATCC CRL-3216) and TZM-bl cells (obtained through the NIH AIDS Research and Reference Reagent Program) [51] were maintained in Dulbecco’s modified Eagle' s medium (Sigma) containing FCS and antibiotics. Human peripheral CD4+ T cells were isolated human CD4+ T cell isolation kit (Miltenyi) according to the manufacturer’s protocol. These cells were activated with anti-CD3/anti-CD28 dynabeads (Thermo Fisher Scientific) and maintained in RPMI1640 (Sigma) containing FCS and antibiotics with human interleukin-2 (100 U/ml) as previously described [23]. To construct the IMCs of hyper HIV-1 and hypo HIV-1 derivatives (based on a CCR5-tropic strain NLCSFV3 [21]), the hyper and hypo Vif variants of the HIVIIIB A200C proviral constructs [20] were digested with AgeI and EcoRI, then the resultant DNA fragment was inserted into the AgeI-EcoRI site of pNLCSFV3 [21]. The IMCs of HIV-1 strains JRCSF [52] and AD8 [53] were also used. The two vif-mutated derivatives based on pNLCSFV3, vif-deleted virus (pNLCSFV3Δvif) and DRMR/AAAA-mutated virus (4A HIV-1), are constructed in our previous study [11,24]. The IMCs of transmitted/founder (TF) and chronic control (CC) viruses as well as those of HIV-1 groups N (strain DJO0131), O (strains BCF183 and RBF206) and P (strain RBF168) (Fig 5C) were obtained kindly provided by Drs. Beatrice H. Hahn (University of Pennsylvania, USA) and Frank Kirchhoff (Ulm University Medical Center, Germany). To prepare virus solutions of hyper and hypo HIV-1s, 30 μg of each IMC was transfected into HEK293T cells according to calcium-phosphate method as previously described [11,12,23,24]. At 48 h posttransfection, the culture supernatant was harvested, centrifuged, and then filtered through a 0. 45-μm filter (Millipore) to obtain the virus solution. The amount of viral particles was quantified using an HIV-1 p24 (Gag) antigen ELISA kit (Zeptometrix). Virus solutions of hyper and hypo HIV-1 clones (containing 2. 5 ng of Gag antigen each) were intraperitoneally co-inoculated into NOG-hCD34 mice. RPMI 1640 was used for mock infection. PB and plasma were routinely collected as previously described [11,12,22–24]. The mice were euthanized at 6 wpi with anesthesia and the spleen was crushed, rubbed, and suspended as previously described [11,12,22–24]. To obtain splenic human MNCs, the splenic cell suspension was separated using Ficoll-Paque (Pharmacia) as previously described [11,12,22–24]. The amount of HIV-1 RNA in 50 μl plasma was quantified by Bio Medical Laboratories, Inc. (the detection limit of HIV-1 RNA is 800 copies/ml). In the experiments shown in Figs 1 & 2, genomic DNA was extracted from the PB of NOG-hCD34 mice using a DNeasy Blood & Tissue kit (Qiagen) as previously described [24]. In the experiments shown in Fig 3, genomic DNA was extracted from the splenic MNCs of NOG-hCD34 mice in the same procedure. Genotyping PCR of A3H was performed using PfuUltra High Fidelity DNA polymerase (Agilent) according to the manufacturer’s protocol, and the following primers were used: Exon2_Fwd, 5' -GAA ACA CGA TGG CTC TGT TAA CAG CC-3' ; Exon3_Rev, 5' -CGG GGG TTT GCA CTC TTA T-3' ; Exon4_Fwd, 5' -AGG AAG GAA GGA TTG TGG CTC A-3' ; Exon4_Rev, 5' -GAG TCC TCA TGC TCA GCA CA-3' (see also Fig 1A). For genotyping PCR of A3F, the following primers were used: A3F_exon5_8822_Fwd. 5' -GGT CTC TGC ATT GGG GTT TC-3' ; A3F_exon5_9069_Rev: 5' -TGC ATT CCT AGC TGC TTA GC-3' . The resulting DNA fragments were directly sequenced, and, if needed, were cloned using a zero blunt TOPO PCR cloning kit (Thermo Fisher Scientific). The sequence was analyzed with Sequencher v5. 1 software (Gene Codes Corporation). Flow cytometry was performed with FACS Calibur (BD Biosciences) and FACSJazz (BD Biosciences) as previously described [11,12,22–24], and the obtained data were analysed with Cell Quest software (BD Biosciences) and FlowJo software (Tree Star, Inc.). For flow cytometry analysis, anti-CD45-PE (HI30; Biolegend), anti-CD3-APC-Cy7 (HIT3a; Biolegend), anti-CD4-APC (RPA-T4; Biolegend), anti-CD25-APC (BC96; eBioscience), and anti-Ki67-PE (B56; BD Biosciences) antibodies were used. Hematometry was performed with a Celltac α MEK-6450 (Nihon Kohden Co.) as previously described [11,12,23,24,49]. Live cell sorting was performed using FACSJazz (BD Biosciences) according to the manufacture' s procedure. The purity of each population was >94% (see also S7 Fig). Transfection, the TZM-bl assay and Western blotting were performed as previously described [11,12,23,24]. Briefly, in the experiments shown in Fig 1B & S3 Fig, HEK293 cells were cotransfected with an expression plasmid for flag-tagged A3H-II (0,25,50 and 100 ng) and the indicated IMCs (1 μg). In the experiments shown in Figs 1F, 1G, 3D & 3E, HEK293 cells were co- cotransfected with an expression plasmid for flag-tagged A3H-II (10 ng), pNLCSFV3Δvif (500 ng) and an expression plasmid for the indicated Vif tagged with HA (500 ng). In the experiments shown in Fig 5C, HEK293 cells were cotransfected with an expression plasmid for flag-tagged A3H-II (50 ng) and the indicated IMCs (1 μg). In the experiments shown in S1 Fig, HEK293 cells were cotransfected with an expression plasmid for flag-tagged A3D (50 ng), A3F (10 ng) or A3G (10 ng), pNLCSFV3Δvif (500 ng) and an expression plasmid for the indicated Vif tagged with HA (500 ng). For Western blotting, anti-Flag antibody (M2; Sigma), anti-HA antibody (3F10; Roche), anti-p24 antiserum (ViroStat), and anti-α-tubulin (TUBA) antibody (DM1A; Sigma) were used. RT-PCR was performed as previously described [24]. Briefly, viral RNA was extracted from the plasma of infected mice at 6 wpi using a QIAamp viral RNA mini kit (Qiagen), and cDNA was prepared as previously described [24]. RT-PCR was performed using PrimeSTAR GXL DNA polymerase according to the manufacturer’s protocol, and the following primers used are used: Vif-Fwd, 5' -GTT TGG AAA GGA CCA GCA AA-3' ; Vif-Rev, 5' -GCC CAA GTA TCC CCG TAA GT-3' . The resulting DNA fragments were cloned using a zero blunt TOPO PCR cloning kit (Thermo Fisher Scientific), and the sequence was analyzed with Sequencher v5. 1 software (Gene Codes Corporation). The vif ORF sequences (Figs 1E, 2C & 3C) were aligned by using MUSCLE [54] implemented in MEGA 6 software [55]. ML phylogenetic trees were constructed using MEGA 5. 1 software [55]. The Vif sequences (Fig 5A and 5B & S10 Table; one sequence per patient) were extracted from Los Alamos National Laboratory HIV-1 sequence database (https: //www. hiv. lanl. gov/components/sequence/HIV/search/search. html). These sequences were aligned and the phylogenetic tree was constructed as described above. A series of HA-tagged Vif expression plasmids are based on pDON-AI (Takara) and are constructed in our previous study [24]. To prepare the expression plasmids of Vif derivatives (Figs 1F, 1G, 3D & 3E), the pCRII-blunt-TOPO containing vif ORFs were digested with EcoRI and blunted. The resultant DNA fragments containing vif ORF were subcloned into the HpaI site of pDON-AI (Takara). Human MNCs were isolated from the spleen of humanized mice as described above and RNA was extracted using QIAamp RNA Blood Mini kit (Qiagen) as described above [11,23,24]. RNA-seq analysis was conducted in Medical & Biological Laboratories, co (Nagoya, Japan). The obtained raw sequence data (. fastq files) were mapped to the human reference genome (NCBI hg19) by Bowtie2 version 2. 2. 5 [56], followed by spliced junction detection by Tophat2 version 2. 1. 0 [57]. Several R (versions 3. 1. 1) and Bioconductor packages were used to further process the gene expression data. Read count data for each sample were extracted by package ‘Rsubread’ [58]. The obtained raw read count data were then normalized by applying repeated edgeR normalization defined in package ‘TCC’ [59]. The normalized read count data were classified into two groups according to infection status (HIV-1 infected, or uninfected as control). The expression data were analyzed to detect differentially expression genes by package edgeR [60]. Top-ranked genes were selected as differentially expressed genes (DEGs) with the following threshold values: False Discovery Rate (FDR) less than 0. 001 calculated by the Benjamini-Hochberg method [61], and more than twice up-regulated or less than half down-regulated normalized gene expressions compared with the control (see Fig 4B & S8 Table). DEGs were then used to obtain enriched biological functions by a parametric gene set enrichment analysis by using package ‘gage’ [62]. The method defined in ‘gage’ enabled to extract gene ontology terms associated with up-regulated DEGs. Finally, a distance matrix was calculated from the expression data for DEGs based on the correlation distance [63], and the distance matrix was converted by the Z-transformation defined in package ‘gplots’ to visualize the result with a heatmap (Fig 4A). Real-time RT-PCR was performed as previously described [20,24] using CFX connect real-time system (Biorad) and the following primers: A3H-Fwd (RSH2757), 5' - AGC TGT GGC CAG AAG CAC-3' and A3H-Rev (RSH2758), 5' -CGG AAT GTT TCG GCT GTT-3' . A3D, A3F, A3G were amplified by using the primers reported previously [32], and the primers for GAPDH were purchased from Thermo Fisher Scientific. The information of A3H haplotypes of individuals was extracted from the 1000 Genomes Project (http: //www. internationalgenome. org) [36]. We obtained the Phase 1 VCF (variant call format) data of 1092 individuals from all available human populations. From this phased variant dataset we extracted the information of 5 A3H SNPs 15,18,105,121, and 178 and estimated the frequencies of A3H haplotypes for each population. The following simple model describes the HIV-1 transmission among human population: dS (t) dt=b−dS (t) −βS (t) I (t) N (t), dI (t) dt=βS (t) I (t) N (t) −μI (t), where S (t) and I (t) represent the number of susceptible and infected individuals, respectively [64]. N (t) is the total population size at time t, and N (0) = b/d is the initial size. Susceptible individuals are born at rate b and removed at rate d, and infected individuals transmit HIV-1 at a rate β during their infectious period of 1/μ. To describe the dissemination of hyper HIV-1 in the human population, we modified the above model as follows: dSU (t) dt=bU−dSU (t) −βSU (t) N (t) {IrU (t) +IoU (t) +IrS (t) +IoS (t) }, dIrU (t) dt=βSU (t) N (t) {IrU (t) +IrS (t) }−μIrU (t), dIoU (t) dt=βSU (t) N (t) {IoU (t) +IoS (t) }−μIoU (t), dSS (t) dt=bS−dSS (t) −βSS (t) N (t) {IrU (t) +IoU (t) +IrS (t) +IoS (t) } dIrS (t) dt=βSS (t) N (t) {IrU (t) +IrS (t) +fIoS (t) }−μIrS (t), dIoS (t) dt=βSS (t) N (t) {IoU (t) + (1−f) IoS (t) }−μIoS (t). The variable SU (t) is the number of susceptible individuals harboring unstable A3H haplotype, and IrU (t) and IoU (t) are the number of infected individuals with hyper and hypo HIV-1s, respectively. On the other hand, the variable SS (t) is the number of susceptible individuals harboring stable A3H haplotype, and IrS (t) and IoS (t) are the number of infected individuals with hyper and hypo HIV-1s, respectively. We assumed that the susceptible individuals harboring unstable and stable A3H haplotype are born at the rates bU and bS = b − bU, respectively. Furthermore, we considered that the fraction, f, of newly infected individuals harboring stable A3H haplotype with hypo HIV-1 become infected individuals with hyper HIV-1 because of adaptive evolution of hyper HIV-1 from hypo HIV-1 in vivo, as we observed in the stable A3H mice infected with NLCSFV3 (Fig 3 & S6 Table). To investigate how the frequency of hyper HIV-1 at 100 years after the initial infection (i. e. , (IrU (100) +IrS (100) ) / (IrU (100) +IrS (100) +IoU (100) +IoS (100) ) ) is determined depend on the proportion of the people harboring stable A3H haplotype (i. e. , SS (0) /N (0) = (bS/d) / (b/d) = bs/b), we simulated the transmission dynamics of hyper and hypo HIV-1s among 1 million individuals for 0 < bs/b < 1 based on the above modified mathematical model. Here we simply fixed 1/d = 35 years (i. e. , adults aged 15–49 years), which implies b = dN (0) = 2. 86 × 104 per year. As previously estimated in [65,66], we assumed that β = 4. 53 per year, and 1/μ = 35 years corresponding to HIV-1-infected individuals with the mean set-point viral load of 3. 2 × 104 RNA copies/ml. The fraction, f, is fixed to be 0. 60 in our simulations based on our findings in the stable A3H humanized mice infected with NLCSFV3 (Fig 3 & S6 Table). Our simulations well reproduced that the prevalence of hyper HIV-1 in the human population with different stable A3H proportion (Fig 5E). The data are presented as averages ± SDs or SEMs. Statistically significant differences were determined by Student' s t test, Paired t test, and Mann-Whitney U test. To determine statistically significant correlations (S9 Fig), the Spearman rank correlation test was applied to the data. An accession number for the data generated in this study is as follows: the RNA-seq data of the splenic MNCs of HIV-1-infected (n = 4) and mock-infected (n = 4) humanized mice (GEO: GSE92262).

Human APOBEC3 family proteins are known as intrinsic defenses against HIV-1, whereas HIV-1 Vif counteracts APOBEC3-mediated anti-viral action. Using a hematopoietic stem cell-transplanted humanized mouse model, we demonstrated that endogenous APOBEC3D, APOBEC3F and APOBEC3G play pivotal roles in restricting HIV-1 replication in vivo. In addition to these three APOBEC3 family proteins, certain haplotypes of APOBEC3H have the ability to control HIV-1 replication in cell culture studies. However, the anti-viral effect of APOBEC3H polymorphism in vivo and in human population is yet to be addressed. Here we use a humanized mouse model to show that acquiring resistance to anti-viral APOBEC3H is necessary for HIV-1 replication. Together with phylogenetic analyses and mathematical modeling, we conclude that APOBEC3H is a critical determinant of HIV-1 replication within infected individuals and we propose that it may also be a factor in human-to-human HIV-1 transmission.

Abstract Introduction Results Discussion Materials and methods

blood cells taxonomy medicine and health sciences immune cells pathology and laboratory medicine evolutionary biology pathogens immunology microbiology genetic mapping plasmid construction retroviruses viruses immunodeficiency viruses animal models phylogenetics data management model organisms rna viruses phylogenetic analysis experimental organism systems dna construction molecular biology techniques research and analysis methods sequence analysis computer and information sciences white blood cells animal cells bioinformatics medical microbiology hiv gene expression microbial pathogens t cells biological databases mouse models hiv-1 evolutionary systematics molecular biology haplotypes sequence databases cell biology heredity viral pathogens database and informatics methods genetics biology and life sciences cellular types lentivirus organisms

2017

HIV-1 competition experiments in humanized mice show that APOBEC3H imposes selective pressure and promotes virus adaptation

The possibility of emergence of praziquantel-resistant Schistosoma parasites and the lack of other effective drugs demand the discovery of new schistosomicidal agents. In this context the study of compounds that target histone-modifying enzymes is extremely promising. Our aim was to investigate the effect of inhibition of EZH2, a histone methyltransferase that is involved in chromatin remodeling processes and gene expression control; we tested different developmental forms of Schistosoma mansoni using GKS343, a selective inhibitor of EZH2 in human cells. Adult male and female worms and schistosomula were treated with different concentrations of GSK343 for up to two days in vitro. Western blotting showed a decrease in the H3K27me3 histone mark in all three developmental forms. Motility, mortality, pairing and egg laying were employed as schistosomicidal parameters for adult worms. Schistosomula viability was evaluated with propidium iodide staining and ATP quantification. Adult worms showed decreased motility when exposed to GSK343. Also, an approximate 40% reduction of egg laying by GSK343-treated females was observed when compared with controls (0. 1% DMSO). Scanning electron microscopy showed the formation of bulges and bubbles throughout the dorsal region of GSK343-treated adult worms. In schistosomula the body was extremely contracted with the presence of numerous folds, and growth was markedly slowed. RNA-seq was applied to identify the metabolic pathways affected by GSK343 sublethal doses. GSK343-treated adult worms showed significantly altered expression of genes related to transmembrane transport, cellular homeostasis and egg development. In females, genes related to DNA replication and noncoding RNA metabolism processes were downregulated. Schistosomula showed altered expression of genes related to cell adhesion and membrane synthesis pathways. The results indicated that GSK343 presents in vitro activities against S. mansoni, and the characterization of EZH2 as a new potential molecular target establishes EZH2 inhibitors as part of a promising new group of compounds that could be used for the development of schistosomicidal agents. Schistosomiasis is a chronic and debilitating disease caused by trematodes of the genus Schistosoma and is one of the most prevalent and neglected diseases of tropical and subtropical regions, affecting more than 250 million people in 78 countries [1,2]. Despite efforts to control schistosomiasis in Brazil, it remains the South American country with the highest number of registered cases [3,4]. The severity of the disease and the organic deficit it produces make schistosomiasis the second most important neglected tropical disease in terms of death and morbidity, behind only malaria [4,5]. Current schistosomiasis treatment is based on the use of praziquantel (PZQ), which is effective against all Schistosoma species infecting humans [6]. Despite the advantages of PZQ as an anthelmintic drug of choice, which include tolerability, safety, efficacy, and low cost, it does not protect individuals against reinfection [7]. A major disadvantage of PZQ is its low efficacy against immature forms [8] and female worms [9]. In addition, the appearance of resistance of some strains of Schistosoma to the drug is a constant concern for the public health authorities [7,10,11]. Thus, there is a need for identifying new targets [12] and for planning, development and research of new compounds as potential schistosomicides [13]. Schistosoma, like other parasites, has some physiological characteristics similar to malignant tissues, such as intense and out-of-control cell division (for egg production), and a high level of metabolic activity similar to tumors [14]. Such metabolic similarities have generated interest in testing histone deacetylase inhibitor anti-cancer drugs such as valproic acid (VPA), suberoylanilide hydroxamic acid (SAHA) and Trichostatin A (TSA) as schistosomicidal compounds [14]. In particular, TSA has been shown to cause mortality of schistosomula and adults, an increase in apoptosis and increase in caspase 3/7 activity [14]. In our group, we have recently identified a set of genes differentially expressed in schistosomula exposed to TSA in culture [15], which are genes associated with DNA replication, recombination and repair, cell cycle and cell death. Interestingly, a network of genes with altered expression related to cell death was identified [15], among which the Embryonic Ectoderm Development (EED) gene was present; EED expression was strongly suppressed by TSA treatment [15]. EED encodes a regulatory protein that is required for the catalytic activity of the histone methyltransferase EZH2 enzyme, the subunit of polycomb repressor complex 2 (PRC2) that is responsible for trimethylation of histone H3 lysine 27 (H3K27me3) and inhibition of gene transcription [16]. Inhibition of EED expression by TSA led us to test the hypothesis that the decrease in the activity of PRC2 would be related to the death of the parasites [15]. We therefore targeted the S. mansoni PRC2 with GSK343 [15], a compound that is a known inhibitor of the human EZH2 enzyme [17,18] that was predicted to bind to SmEZH2 and act as a competitive inhibitor of its cofactor S-adenosyl methionine (SAM) [15]; prediction was done by homology modeling [15] of the SmEZH2 protein (Smp_078900) to the crystal structure of the human hEZH2 [19], followed by GSK343 docking to the SAM cofactor pocket of the SmEZH2 protein [15]. In that work, we showed that GSK343 has a synergistic effect along with TSA in promoting S. mansoni in vitro death [15]. In the present work we investigated in detail the effect of exposing S. mansoni in vitro to GSK343 alone. GSK343 was found to kill three different developmental forms of S. mansoni tested, namely schistosomula, adult males and females, and western blotting confirmed a decrease in the H3K27me3 histone mark in all three forms. Scanning electron microscopy and phenotypic parameters such as motility, viability, oviposition and mating were analyzed to document the changes upon GSK343 exposure that accompanied parasite death. Analysis of the effect of GSK343 on the large-scale gene expression of treated worms was performed with a high-throughput sequencing strategy (RNA-seq). In addition, the effect of GSK343 on the relative abundance of proteins was determined with quantitative mass spectrometry. Profiles of altered gene expression and of altered proteins abundance were identified. The experimental protocols were in accordance with the Ethical Principles in Animal Research adopted by the Brazilian College of Animal Experimentation (COBEA) and the protocol/experiments have been approved by the Ethics Committee for Animal Experimentation of Instituto Butantan (CEUAIB n° 1777050816). The BH strain of Schistosoma mansoni (Belo Horizonte, Brazil) was maintained in the intermediate snail host Biomphalaria glabrata and as definitive host the golden hamster (Mesocricetus auratus). Female hamsters aged 3–4 weeks, freshly weaned, weighing 50–60 g, were housed in cages (30x20x13cm) containing a sterile bed of wood shavings. A standard diet (Nuvilab CR-1 Irradiada, Quimtia S/A, Paraná, Brazil) and water were made available ad libitum. The room temperature was kept at 22 ± 2°C and a 12: 12 hour light–dark cycle was maintained. Hamsters were infected by exposure to a S. mansoni cercariae suspension containing approximately 200–250 cercariae using the ring technique [20]. After 49 days of infection, the S. mansoni adult worms were recovered by perfusion of the hepatic portal system [21]. Alternatively, after 21 days of infection juvenile worms [22,23] were collected by perfusion [21]. Cercariae were released from infected snails and mechanically transformed to obtain schistosomula in vitro [24]. Schistosomula, juvenile worms and adult worms were treated with different final concentrations of GSK343 in culture medium specific to each stage as indicated in the Results (from a stock solution of 20 mM GSK343 in DMSO), and with the equivalent amount of DMSO in the control assays. Newly transformed schistosomula (NTS) were maintained for 16 h in M169 (Vitrocell) medium supplemented with 2% fetal bovine serum (FBS) (Vitrocell), 1 μM serotonin, 0. 5 μM hypoxanthine, 1 μM hydrocortisone, 0. 2 μM triiodothyronine, penicillin/streptomycin, amphotericin, gentamicin (Vitrocell) at 37°C and 5% CO2 [25]. Only after 16h incubation with the culture medium was the GSK343 treatment initiated. Juvenile worms and paired adult worms were maintained in RPMI medium (Gibco) supplemented with 10% fetal bovine serum (FBS) (Vitrocell), penicillin/streptomycin, amphotericin, gentamicin (Vitrocell) at 37°C and 5% CO2. An inverted microscope was used to evaluate the general condition of adult worms, including motility and mortality rate. Parasites were observed after 3,6, 12,24 and 48 h of treatment with 20 μM GSK343 or vehicle dimethyl sulfoxide (DMSO 0. 1%). The motility and survival of worms were assessed according to the criteria scored in a viability scale of 0–3 [26]. The scoring system was as follows: 3, complete body movement; 1. 5, partial body movement or immobile but alive; and 0, dead. Treatment was considered lethal whenever no worm movement was detected when observed for more than 2 min. In addition, pairing status and oviposition were recorded. The viability of schistosomula and S. mansoni adult worms after treatment was determined by a cytotoxicity assay based on the CellTiter-Glo Luminescent Cell Viability Assay (G7570, Promega, Madison, Wisconsin, EUA) [27]. The assay determines the amount of ATP present in freshly lysed adults or in intact schistosomula; the assay signals the presence of metabolically active cells. In addition, viability of schistosomula was evaluated by the presence or absence of dead parasites, through staining with propidium iodide (PI) [28] and fluorescein diacetate (FDA) [29], as follows. Schistosomula were equally distributed in 96-well microtiter plates, incubated with the concentration of GSK343 indicated in the figure or the corresponding DMSO vehicle (control), and 2 μg/mL propidium iodide (PI) (Sigma-Aldrich) plus 0. 5 μg/mL fluorescein diacetate (FDA) (Life Technologies) were added at the time points indicated in the legend to the figure. The parasites were immediately observed with light microscopy at 10 x magnification using a Nikon Eclipse fluorescent inverted microscope. Under light microscopy, viable parasites were scored by preserved mobility and lack of opacity. Under fluorescence microscopy, schistosomula death was scored by a red fluorescence signal (572 nm emission microscope filter); in living schistosomula cells FDA is converted into charged fluorescein by parasite esterase activity, staining the schistosomula with a green fluorescence signal (492 nm emission microscope filter) [28,29]. For each time point a new set of wells was used, because the staining procedure was lethal to the parasites. The number of biological replicates that were assayed, as well as the number of parasites that were counted per replicate, is stated in the legends to the figures. Caspase 3/7 activity was measured using the Caspase-Glo 3/7 assay kit (Promega) following manufacturer’s instructions. Briefly, 2,000 schistosomula per well were cultivated in a 24 well plate with complete medium (M169 medium supplemented as described above) plus 20 μM GSK343 or vehicle (DMSO 0. 1%). At 48 h incubation the medium was replaced, the reporter lysis buffer was added, followed by 3 h incubation with the Caspase-Glo reagent, in the dark with agitation. A negative control was carried out without the Caspase-Glo reagent. Luminescence was measured in a white-walled 96-well plate in a Wallac Victor2 1420 multilabel counter (PerkinElmer). Detection of DNA strand breaks in GSK343-treated schistosomula was performed using the In situ Cell Death Detection kit (Roche) [30]. Briefly, 2,000 schistosomula per well were cultivated in a 24-well plate with complete medium (M169 medium supplemented as described above) plus 20 μM GSK343 or vehicle (DMSO). At 72 h incubation, the schistosomula were fixed for 60 min, made permeable prior to labeling as recommended by the manufacturer and mounted in a glass slide using Prolong with DAPI (Invitrogen, USA) for nuclear visualization. Images were taken on a Zeiss Axio Observer Z1 inverted microscope equipped with a 40× objective lens and an AxioCam MRm camera, in the ApoTome mode. Adult and juvenile worms collected by perfusion were immediately transferred to supplemented RPMI medium. The parasites were distributed in 24-well plates (adults: one paired worm couple per well; juveniles: five worms per well) with the same medium. The schistosomula were distributed in 6-well plates (2,000 per well) containing the supplemented M169 medium. The worms were kept in culture (oven at 37°C and 5% CO2) for 2 h for adaptation, and then GSK343 was added. Ultrastructural analysis was performed with scanning electron microscopy. Adult worms, juvenile worms and schistosomula incubated in 20 μM GSK343 or DMSO vehicle for 24 h and 48 h were fixed with modified Karnovsky reagent (1% paraformaldehyde, 2. 5% glutaraldehyde, 1 mM calcium chloride in 1 M sodium cacodylate buffer, pH 7. 4) and after the fixing stage the material was washed with sodium cacodylate buffer (0. 1 mol / L, pH 7. 2) and post-fixed with 1% osmium tetroxide (w / v) for 1 h. The same protocol was used for SEM of eggs released from the couples incubated with GSK343 or the vehicle. The samples were dehydrated with increasing concentrations of ethanol and then dried with liquid CO2 in a critical-point dryer machine (model Leica EM CPD030, Leica Microsystems, Illinois, USA). Treated specimens were mounted on aluminum microscopy stubs and coated with gold particles using a ion-sputtering apparatus (model Leica EM SCD050, Leica Microsystems, Illinois, USA) [31]. Specimens were then observed and photographed using an electron microscope (FEI QUANTA 250, Thermo Fisher Scientific, Oregon, USA). In the preparation of histone acid extracts of adult parasites, fifty pairs of worms (either treated or control) were soft lysed with 500 μL lysis buffer (PBS containing 0. 5% Triton X-10,0. 02% NaN3 and Mini Protease Inhibitor Cocktail-Complete from Roche) in a glass Potter homogenizer. The samples were centrifuged (10 min, 2000 g at 4°C) and pellets containing the nuclear material were washed once in 200 μL lysis buffer then centrifuged again [14]. The pellet containing the nuclear fraction with the histones was resuspended in 400 μL 0. 25 M HCl with protease inhibitor and the solution was incubated overnight at 4°C in order to precipitate acid proteins [32]. The samples were centrifuged (30 min, 16000 g at 4°C) and the histone proteins contained in the supernatant were concentrated with trichloroacetic acid 33% [33]. The final pellet with histones was taken up in MilliQ water with protease inhibitors. For schistosomula nuclear histones extraction, treated or control samples containing approximately 10,000 schistosomula each were used. Each sample was sonicated on ice in an Eppendorf tube for 10 min with an Active Motif sonicator model Q120AM (5 s ON, 5 s OFF, 20% amplitude) in lysis buffer (PBS containing 0. 5% Triton X-100,2 mM PMSF, 0. 02% NaN3 and Mini Protease Inhibitor Cocktail—Complete from Roche). After the sonication, total lysis of the schistosomula was checked under an optical microscope, the samples were centrifuged 20 min at 13,000 g, 4°C, and the supernatant containing the nuclear protein extract was stored in a clean tube at -80°C. The nuclear protein concentration of extracts of adult worms or schistosomula were determined with the Micro BCA Protein Assay kit (Pierce Biotechnology). From each sample, 10 μg of extracts was loaded on 15% SDS-Polyacrylamide gels, and after electrophoretic protein separation, the proteins were electro-transferred (TE 77 Amersham Biosciences) from the gel to a nitrocellulose membrane (Amersham) for 1. 5 h at 100 mA. Briefly, membranes were blocked with Tris-buffered saline (TBS) containing 0. 1% Tween 20 and 2% bovine serum albumin (TBST/2% BSA), and then probed overnight with specific antibodies in TBS/2% BSA. Membranes were washed with TBST and incubated for 1 h in TBST/2% BSA with secondary antibody conjugated with IRDye (IRDye 800CW goat anti-rabbit or IRDye 700CW goat anti-mouse from Licor Biosciences). After washing the membranes in TBST, the bands were visualized and images were recorded with the Odyssey Infrared Imaging System (Licor Biosciences), and quantified with Image J [34]. Acetylation and methylation of histones was measured with specific monoclonal antibodies to the following lysine modifications: Histone H3 acetyl K9 antibody C5B11 (Cell Signaling) (1: 1000), Histone H3 acetyl K27 antibody ab45173 (Abcam) (1: 500), Histone H3 tri-methyl K4 antibody 04–745 (Millipore) (1: 2000), Histone H3 tri-methyl K27 antibody ab6002 (Abcam) (1: 1000), Histone H3 mono-methyl K27 antibody 61015 (Active Motif) (1: 250). For normalization of the signals across the samples anti-Histone H3 antibody ab24834 (Abcam) (1: 1000) was used. Adult worms or schistosomula were maintained in the stage-specific culture media described above (at 37°C and 5% CO2) and were treated for 48 h with either 20 μM GSK343 or the DMSO (0. 1%) vehicle. Total RNA extraction from parasites was performed using the RNeasy Mini Kit (QIAGEN) for adults (in three biological replicates) and the RNeasy Micro Kit (QIAGEN) for schistosomula (four biological replicates), following the instructions provided by the manufacturer, including the optional DNase treatment during extraction of total RNA to ensure the removal of eventual contaminating genomic DNA present in RNA. RNA purity was evaluated with NanoDrop (ND-100 Spectrophotometer, Thermo Scientific) and it was adequate for RNA-seq library construction (ratio A260 / A280 = 2. 2, A260 / A230 = 2. 0). RNA integrity was assessed by electrophoresis with the 2100 Bioanalyzer (Agilent). The Agilent RNA 6000 Nano or Pico Assay protocols were followed, depending on the amount of RNA in the samples. Only the band corresponding to 18S ribosomal RNA and not the 28S band was present, as expected [35], plus a smear above 500 nt, which indicated no RNA degradation. The RNA concentration was measured with Qubit 2. 0 fluorimeter (Invitrogen by Life Technologies), and the total RNA obtained was in sufficient quantity (> 1 μg per sample of adult worms) for library construction with the KAPA Stranded mRNA-Seq Kit (KK8421) for adult worm samples, and with the SMART-Seq v4 Ultra Low Input RNA Kit (634891) for schistosomula samples (> 10 ng per schistosomula sample). Both kits have steps where polyA+ RNA is purified prior to cDNA generation. Library construction and RNA-Seq were performed in biological triplicates for adult worms and in quadruplicates for schistosomula at the Duke Center for Genomic and Computational Biology facility (Duke University Durham, NC, USA). Sequencing was on the Illumina HiSeq 2500 platform. The Illumina sequencing data (raw reads) were downloaded from the FTP server at the Duke University facility and verified for files integrity with MD5 checksum. A total of 22 to 32 million reads was obtained per each replicate sample of adult worms and 11 to 13 million reads per each replicate sample of schistosomula. Quality check of reads was with FastQC (www. bioinformatics. babraham. ac. uk/projects/fastqc/), and Trimmomatic [36] was used to trim Illumina adapters and to filter out reads with low quality values using the following parameters: MAXINFO with parameters 50: 0. 05, HEADCROP of 28 nucleotides, a LEADING and TRAILING value of 5, and minimum read length of 30 nucleotides. Mapping of reads to the genome and gene differential expression analysis were performed with three different pipelines as described below. The different tools handle the problem of mapping reads to alternatively spliced genes in different ways; the intersection of results of differentially expressed genes (DEGs) from the three analyses was used as the core, most significant DEGs. Version 5. 2 of the S. mansoni genome [37] was used for all analyses. The first pipeline used Tophat2 [38] that is part of the Tuxedo Tools, along with a custom reference transcriptome gff file, which was based on the Sanger transcriptome [37] and includes corrections for inconsistencies in some coordinates of start and end of UTRs; this custom reference transcriptome gff file is available at http: //verjolab. usp. br. Read counting was done with cuffquant using the Tophat2 BAM files as input, and a metagenes transcriptome [39] as reference; this metagenes transcriptome reference gtf file is available at http: //verjolab. usp. br. Statistical significance analysis of differential expression was obtained by using cuffdiff [40] with cuffquant counting as input and the default parameters, which imply the pooled dispersion method and the geometric library normalization method. A q-value ≤ 0. 05 was used as significance threshold. The metagenes transcriptome reference takes into account the evidence from public RNA-seq data either that two or more genomic neighbor Smp_nnnnnn predicted genes do encode fragments of the same protein and should be merged, or that in a given genomic locus a novel transcript was assembled from the public RNA-seq data [39]. In the two cases the corresponding transcript was named “SmVG_xxxxxx”, where xxxxxx stands for the XLOC_xxxxxx identifier number generated for that locus during the assembly [39]. In the case that one or more Smp genes comprise the SmVG_xxxxxx metagene, it was further annotated as SmVG_xxxxxx: Smp_nnnnnn to indicate the corresponding Smp_nnnnnn genes that comprise that SmVG_xxxxxx; the expression value of that metagene was assigned to all SmVG_xxxxxx: Smp_nnnnnn genes belonging to the same SmVG_xxxxxx metagene. Sequence and genomic coordinates information for all SmVG_xxxxxx genes is available at the metagenes transcriptome reference gtf file mentioned above. The second pipeline employed Kallisto [41], which uses a pseudo-alignment approach to map the reads against the genome. The metagenes transcriptome was again used as reference. Sleuth [41] statistical package was used to find statistically significant DEGs with a q-value ≤ 0. 05. Finally, the third pipeline employed Sailfish [42], which uses a quasi-mapping approach to map the reads against the genome. EdgeR [43] was used for the statistical analyses and the exact test proposed by Robinson and Smyth [44] was selected. A q-value ≤ 0. 05 was used as significance threshold. As mentioned, we retained for further processing only the consensus DEGs found by all three procedures described above. Detailed expression values for all replicates and differential expression statistical significance for these consensus genes are available at http: //verjolab. usp. br. Simplified lists with gene IDs, q-values, GO terms and product description are given in the tables indicated in the Results section below. Gene enrichment analysis was performed with BINGO [45] using the S. mansoni annotated Gene Ontology (GO) terms (www. geneontology. org) and the list of DEGs for each tested condition. An enrichment cutoff value of corrected p-value ≤ 0. 05 was used as significance threshold. For quantitative RT-PCR, complementary DNAs were obtained by reverse transcription (RT) of 100 ng schistosomula or adult worms total RNA using SuperScript III Reverse Transcriptase (Invitrogen) and random hexamer primers in a 20 μL volume, according to the manufacturer’s recommendations. The resulting cDNA was diluted 8-fold in water and qPCR amplification was done with 2. 5 μL of diluted cDNA in a total volume of 10 μL using SYBR Green Master Mix (Life Technologies) and specific primer pairs (S1 Table) designed for S. mansoni genes by Primer3 online software. The Light cycle 480 II (Roche) qPCR was used. The results were analyzed by comparative Ct method and the statistical significance was calculated with the t-test. To find adequate normalizer genes for qPCR, we looked for evidence of genes with non-detectable changes in expression upon GSK343 treatment in the RNA-Seq data. For this, genes with normalized TPM values > 1 in the control samples, as computed both by Sailfish and by Kallisto (see above) were selected. Further filtering retained the genes that were not in the list of DEGs, and had an average TPM ratio treated/control in both Sailfish and Kallisto analyses within the interval 0. 9–1. 1 (maximum allowed difference between control and treated expression is 10%); additionally, the selected genes in the adult females and males analysis were pooled and ranked by their coefficient of variation, from smaller to larger, and the top 5 genes were selected for testing by qPCR; for schistosomula, the top 4 genes in a similar analysis were selected for testing by qPCR. The final set of normalizers was confirmed by qPCR as having their expression ratio within the interval 0. 9–1. 1. For adult worms, the normalizers were granulin (Smp_170550), ATP synthase subunit beta (Smp_029390) and serine threonine protein phosphatase 2A (Smp_166290). For schistosomula, the normalizers were mitochondrial ATP synthase B subunit (Smp_154530) and casein (Smp_025010). In all cases, the average expression of normalizers was used for calculating the expression of the genes of interest. Schistosome parasites were lysed in lysis buffer containing urea (8 M), Mini Protease Inhibitor Cocktail-Complete from Roche and 50 mM ammonium bicarbonate (Ambic). Extracted proteins were quantified using Micro BCA Protein Assay kit (Pierce Biotechnology), reduced with dithiothreitol (10 mM) for 30 min at room temperature and subsequently alkylated with iodoacetamide for 30 min in the dark. Following incubation, the samples were diluted in 20 mM Ambic, pH 7. 5, and digested with trypsin (Promega V5111) 1: 50 (w/w) overnight at room temperature. Samples were acidified with trifluoroacetic acid (TFA) (1% final concentration) and centrifuged at 14,000 g for 10 min to stop trypsin digestion. Supernatant was collected and dried prior to desalting. Three biological replicates for treated and untreated conditions were processed and analyzed. Samples were resuspended in 0. 1% TFA and desalted using homemade micro-columns comprised of a C18 plug taken from a C18 disk (Sigma-Aldrich) with the constricted end of a P200 tip inserted into the disk. The acidified samples were loaded onto the micro-column by applying a gentle air pressure with the syringe and washed three times with 0. 1% TFA. Peptides were eluted with 50% acetonitrile (ACN) / 0. 1% TFA, followed by 70% ACN/ 0. 1% TFA. Peptide samples were dried in the Speed-Vac, resuspended in 0. 1% formic acid (FA) and analyzed using an EASY-nLC liquid chromatography system (Thermo Scientific) coupled to LTQ-Orbitrap Velos mass spectrometer (Thermo Scientific). The peptides were loaded on a Reprosil-Pur C18-AQ (3 μm) column and separated in an organic solvent gradient from 100% phase A (0. 1% FA) to 30% phase B (0. 1% FA, 95% ACN) during 80 min for a total gradient of 105 min at a constant flow rate of 300 nL/min. The LTQ-Orbitrap Velos was operated in positive ion mode with data-dependent acquisition. The full scan was acquired in the Orbitrap with an automatic gain control (AGC) target value of 10e6 ions and a maximum fill time of 500 ms. Peptides were fragmented by collision-induced dissociation. Ions selected for tandem mass spectrometry (MS/MS) were dynamically excluded for a duration of 30 s. Each MS scan was acquired at a resolution of 60,000 FWHM followed by 20 MS/MS scans of the most intense ions. All raw data were accessed with the Xcalibur software (Thermo Scientific). Raw data were processed using MaxQuant software version 1. 5. 2. 8 and the embedded database search engine Andromeda [46]. A custom reference Proteins Database was used (http: //verjolab. usp. br); this fasta file contains the >Smp_nnnnnn protein sequences [37] from Ensembl (http: //metazoa. ensembl. org/Schistosoma_mansoni/Info/Index) plus the 212 novel protein sequences (>cnnnn-gn-in) described by Anderson et al. [15] and the micro-exon genes (>MEGnnnnnn and >GUnnnnnn sequences) described by DeMarco et al. [47]. The MS/MS spectra were searched against this custom reference database with the addition of common contaminants, with an MS accuracy of 4. 5 ppm and 0. 5 Da for MS/MS. Cysteine carbamidomethylation (57. 021 Da) was set as the fixed modification, and two missed cleavages for trypsin. Methionine oxidation (15. 994 Da) and protein N-terminal acetylation (42. 010 Da) were set as variable modifications. Proteins and peptides were accepted at FDR less than 1%. All raw data have been submitted to the PRIDE archive [48,49]. Label-free quantitation (LFQ) was performed using the MaxQuant software with the “match between runs” feature activated. Protein LFQ intensity of each of the three biological replicates of treated and untreated samples were used for further statistical analyses. Proteins with valid values in at least two replicates of each biological condition were kept in the analyses. Exclusive proteins were identified in at least two biological replicates of one condition and in none of the other. Significantly regulated proteins were determined using the Welch’s t-test using the Perseus v. 1. 5. 1. 6 software [50], and a p ≤ 0. 05 was used as significance threshold. The S. mansoni adult worms exhibited decreased motility when exposed to GSK343, and this was dependent upon concentration and incubation time (Fig 1A). After 48 h the adult worms lost approximately 40% of their motility at the sublethal dose of 20 μM GSK343 (Fig 1A), there was impairment of the peristaltic movement and the ability of their suckers to adhere to the bottom of the culture plates. When exposed to 50 μM GSK343, the motility of the worms was altered after 12 h of incubation (Fig 1A). At 24h the motility was further decreased to 50%, and some dead worms were already observed. At 48 h with 50 μM GSK343, all worms were classified as dead (Fig 1A). Pairing of the couples was reduced to approximately 50% in relation to the control with 20 μM GSK343 after 48 h of exposure (Fig 1B). With 50 μM GSK343 pairing was drastically reduced to nearly zero already at 3 h of exposure (Fig 1B). Viability measured by ATP quantitation in adult worms after 48 h exposure to the compound (Fig 1C) was reduced to 50% in parasites treated with 20 μM GSK343 compared with controls (DMSO), and it was further reduced to less than 20% with 50 μM GSK343. Throughout the observation intervals (3,6, 12,24 and 48 h), S. mansoni adult worms exposed to vehicle (DMSO 0. 1%) exhibited normal peristaltic movements and characteristic waves along the body axis, suckers in constant movement, showing occasional adherence to the bottom of the culture plate through the ventral sucker; Pairing was maintained in the controls (Fig 1A and 1B). With scanning electron microscopy (SEM), it was possible to detect the phenotypic differences between control and treated male adult worms. The tegument of control adult worms showed no morphological and structural changes (Fig 1D, panels 1 to 5). With 20 μM GSK343, after 48 h exposure, the presence of blebs in the dorsal region was observed, causing the tegument to appear distended (Fig 1D, panel 8) in some parts, possibly due to ballooning of the subtegumental region, and this was a constant characteristic present in all adult worm males analyzed with SEM. Areas of lesion near the tubercles were sometimes visualized, with fissures in the tegument surface and the presence of deep chasms between the dorsal tubercles (Fig 1D, panel 9). At the anterior and posterior regions of the parasite, no alterations were detected (Fig 1D, panels 6,7 and 10). In females treated with 20 μM GSK343 for 24 or 48 h of incubation, severe damage to the tegument is visualized (Fig 2) exhibiting extensive sloughing that exposes subtegumental tissues. The presence of peeling (Fig 2M) and blisters can be seen throughout the extension of the dorsal region of the female (Fig 2E, 2F and 2N). In Fig 2G and 2O, the larger magnification permits visualization of the rupture of blisters where possibly there was extravasation of cytoplasm. The anterior region of the females presented damage to the tegument only after 48 h of incubation with the compound (Fig 2P). In addition, after 24 or 48 h of incubation with 20 μM GSK343 there was a decrease of about 40 to 60% in the number of eggs laid by these females in relation to the control (Fig 3C). It is worth pointing out that the percentage reduction in egg laying remains fairly constant from 24 to 48 h treatment, in contrast to the observed increase in lesions on the worms. SEM of eggs released by females incubated with GSK343 was performed (Fig 3A, panels 5 to 8). These eggs were phenotypically different from control eggs (Fig 3A, panels 1 to 4). They showed fissures in the shell (Fig 3A, panels 6 and 8) and a 30–40% reduction in area (Fig 3B). S. mansoni juvenile forms were treated with GSK343 and observed by SEM (Fig 4). In control (untreated) parasites the tegument appeared intact, the anterior region presented no structural alterations (Fig 4, panels 1 to 3) and at high magnification, the surface had a spongy appearance without morphological changes (Fig 4, panels 4 and 5). In juvenile worms treated with GSK343 (20 and 50 μM) the dorsal surface ridges were severally damaged, they presented blisters in much of the dorsal tegument, with sloughing and erosion (Fig 4, panels 9,10,14 and 15). Anterior region damage was seen only in worms treated with 50 μM GSK343, with the oral sucker partially destroyed (Fig 4, panels 11 to 13). The viability of juvenile worms was measured by the quantitation of total ATP in the parasites. At 24 h an approximate 20 to 30% decrease in viability of parasites was observed when treated with 20 μM and 50 μM GSK343 (Fig 4B, black bars). At 48 h exposure of juvenile parasites to GSK343 an approximate 20 to 30% decrease in viability was already observed at 10 and 20 μM GSK343 (Fig 4B, gray bars). At 50 μM GSK343 a 70% decrease in viability of juvenile worms was observed (Fig 4B, gray bar). Schistosomula mechanically transformed from cercariae were incubated for 16 h in culture medium and micrographs of typical schistosomula were obtained (Fig 5A, panels 1 and 6). At this time vehicle (0. 1% DMSO) was added to control schistosomula, parasites were further incubated and normal development was documented at 24 h (Fig 5A, panels 2 and 3) and at 48 h (Fig 5A, panels 4 and 5). In contrast, when GSK343 was present in the medium, no apparent development or change in size was observed (Fig 5A, panels 7 and 9) compared with schistosomula pre-treatment (Fig 5A, panel 6). Deterioration of the tegument was observed at higher magnification (Fig 5A, panels 8 and 10), with the presence of many surface blisters, intense contraction and loss of numerous spines that covered the whole body of the parasite. Fig 5B shows that while the average length of control schistosomula significantly increased 1. 4-fold in the period from 24 to 48 h on incubation with 0. 1% DMSO (110 μm / 79 μm = 1. 4), the average length of schistosomula treated with 20 μM GSK343 significantly increased only 1. 1-fold (74 μm / 67 μm = 1. 1) during the same period of time (Fig 5B), resulting in a significantly 4-fold lower rate of growth (p < 0. 001) and in significantly smaller treated schistosomula at 48 h incubation (average length 74 μm) compared with controls (average length 110 μm) (p < 0. 001). Exposure of schistosomula for 48 h to 20 μM GSK343 significantly increased caspase 3/7 activity by 17-fold, when compared with control parasites (Fig 5C). Caspases 3 and 7 are activated in both the extrinsic and intrinsic apoptosis pathways [51], and their activity was also increased after treatment of schistosomes with the HDAC inhibitor TSA [12]. Additionally, the capacity of GSK343 to induce apoptosis in schistosomula was measured using a TUNEL assay to detect DNA double strand breaks in schistosomula treated with 20 μM GSK343 for 72 h. The results indicate that this inhibitor induced fragmentation of DNA, with specific green staining clearly observed along the entire worm (Fig 5D, panel 4). No auto fluorescence was detected, as can be seen by the absence of green coloring in the control (0. 1% DMSO) worms observed under UV light (Fig 5D, panel 2). Fluorescein diacetate (FDA) [29] and propidium iodide (PI) [28] were used to differentially quantitate viability of schistosomula at 24,48 and 72 h incubation (Fig 6A) in the presence of vehicle or of 10 to 50 μM GSK343. Under fluorescence microscopy, death of schistosomula in the presence of GSK343 was detected by a red fluorescence signal from PI (536 nm emission microscope filter). Viability was quantified in schistosomula living cells, where FDA is converted into charged fluorescein by the parasite esterase activity, staining the schistosomula with a green fluorescence signal (492 nm emission microscope filter) (Fig 6A). Living green-stained schistosomula were counted at each GSK343 concentration and time. Percentage viability was determined with respect to the total number of parasites present in the well, as checked and counted using light optical microscopy (see Fig 6A, gray images), and plotted as shown in Fig 6B. Incubation with vehicle (0. 1% DMSO) for up to 72 h showed no reduction of parasite viability (Fig 6A, upper panels and 6B), whereas 20 μM GSK343 proved to be a sub-lethal concentration because there was less than 20% decrease in schistosomula viability after 2 days of incubation with GSK343 (Fig 6B). On the other hand, ATP quantitation (Fig 6C) in treated schistosomula showed that the parasites were stressed by the compound, and a reduction of 40% in total ATP content was observed after 2 days of incubation at 20 μM GSK343. It was observed that at longer incubation times for up to 5 days, all tested concentrations of GSK343 caused a 70–100% decrease in viability as measured by ATP content. To detect and generate the profile of changes in histone marks triggered by GSK343 in S. mansoni, the nuclear extracts of adult worms and schistosomula treated with this inhibitor of EZH2 histone H3K27 methyl-transferase were analyzed on western blots developed with a monoclonal antibody specific for the H3K27me3 post-translational modification (PTM), which is related to chromatin condensation and repression of transcription. Four other histone marks, not directly related to EZH2 histone activity, were assayed in order to document the profile of histone marks in the treated parasites. A significant decrease in the level of the H3K27me3 mark was quantitated in adult worms (Fig 7A) and in schistosomula (Fig 7F), treated with GSK343 for 48 h relative to the control (vehicle). In adult worms the acetylated histone marks H3K9ac and H3K27ac (Fig 7B and 7C) were concomitantly and significantly increased in relation to the control samples (vehicle) in three biological replicates. In schistosomula, on the other hand, these marks showed a slight tendency to increase, although these changes were not statistically significant (Fig 7G and 7H). Two more histone marks were analyzed, namely H3K4me3 and H3K27me1; in adult worms there was a significant increase only in H3K4me3 (Fig 7D and 7E), while in schistosomula there was a significant increase only in H3K27me1 (Fig 7I and 7J). With the above panel of histone PTMs, it is possible to conclude that the decrease in the H3K27me3 mark caused by GSK343 treatment probably affected the chromatin architecture [52], thus triggering a compensatory reprogramming and fine-tuning of the parasite gene expression that possibly resulted from a change in the balance between H3K27me3 and other histone marks such as H3K9ac, H3K27ac, H3K4me3 and H3K27me1 [52–54], which are known to be related to activation of gene transcription. To explore the effect of GSK343 on gene expression, adult worms and schistosomula were exposed in vitro to the compound or to vehicle (DMSO), and large-scale gene expression changes were assessed by RNA-Seq. It is known that using different RNA-Seq statistical analyses tools can affect both the number of differentially expressed genes (DEGs) and the altered processes or functions that are detected [55,56]. Therefore, we used three different pipelines, namely Kallisto + Sleuth [41], SailFish + EdgeR [43] and Tuxedo Tools (including Cuffdiff) [40], and looked for a core set of consensus DEGs among them, in order to increase the accuracy of the list of DEGs identified. An example is given in S1 Text, Fig A (i), left panel, which shows the Venn diagram with 585 consensus DEGs (S2 Table) that were detected as downregulated in adult female parasites in the intersection of these three statistical analyses pipelines. It is apparent that between 56 and 245 genes were detected by only one pipeline, and these genes were not further analyzed. In all other analyses of females, males, and schistosomula a similar pattern was obtained, with a core set of consensus genes being considerably larger than the genes exclusively detected by only one pipeline (S1 Text, Figs A (i), A (ii) and A (iii) ). The lists of significantly downregulated (S2 Table) and upregulated (S3 Table) consensus genes in treated females, as well as the significantly downregulated (S4 Table) and upregulated (S5 Table) consensus genes in treated males are given in the Supplementary Information. Interestingly, the effect of GSK343 on adult worms was sex-specific (S1 Text, Figs A (i) and A (ii) ): only a small number of genes was detected in common as downregulated between the two sexes (45 genes, corresponding to 7. 7% of all downregulated genes in females and 7. 9% of all downregulated genes in males). Similarly, only 82 genes were detected in common as upregulated between females and males, corresponding to 18. 3% of all upregulated genes in females and 15. 1% of all upregulated genes in males (S1 Text, Figs A (i) and A (ii) ). It is noteworthy that in schistosomula we detected a smaller number of consensus DEGs when compared with adult worms, namely 143 downregulated (S6 Table) and only 29 (S7 Table) upregulated genes (S1 Text, Fig A (iii) ). The consensus set of DEGs can be better visualized in the heatmaps of Figs 8and 9, where a marked difference in the gene expression levels can be seen between control and treated samples, in females (Fig 8A), males (Fig 8B) and schistosomula (Fig 9), for both upregulated (red) and downregulated (blue) genes. With the lists of female and male differentially expressed genes at hand, a gene ontology (GO) analysis was performed, and Figs 10 and 11 show the top 20 GO enriched terms for downregulated genes in female (Fig 10A) and male worms (Fig 10B) as well as the enriched terms for upregulated genes in males (Fig 11). No significantly enriched GO terms (FDR ≤ 0. 05) were detected among the upregulated genes in females. GSK343 caused alterations in genes related to various metabolic pathways, which are different between females and males, as expected from the minimal overlap of DEGs (S1 Text, Figs A (i) and A (ii) ). In female worms, the GSK343-induced downregulated genes that belong to enriched Molecular Function GO terms are related to nucleic acid and nucleoside binding (Fig 10A). Downregulated genes belonging to enriched Biological Process GO terms are essentially related to nucleic acid metabolic processes, and more specifically the DNA replication pathway and the ncRNA metabolic process (Fig 10A). A heatmap representing all DNA replication genes that were detected as downregulated in GSK343 treated females (32 genes) is shown in Fig 12A; the list of these genes with Smp ID numbers and gene descriptions is given in S8 Table. Interestingly, a principal component analysis (PCA) using the expression levels of this set of 32 genes in females, males and schistosomula (Fig 12B) revealed that females treated with GSK343 have an expression profile of the replication machinery genes quite different from that of untreated females and very similar to those of treated and untreated male worms, suggesting that the active DNA replication machinery in control females is involved with embryonic division and oviposition. The schistosomula expression profile of DNA replication genes is very different from that of adult worms (Fig 12B). The heatmap of ncRNA metabolic process genes that were detected as downregulated in GSK343 treated females (Fig 13A, 51 genes) shows that these genes essentially encode rRNA proteins and tRNA processing proteins as well as proteins involved in translation; the list of these genes with Smp ID numbers and gene descriptions is given in S8 Table. Again, a PCA using the expression levels of these 51 genes involved in ncRNA metabolic processing (Fig 13B) shows that females treated with GSK343 have an expression profile of ncRNA metabolic process genes very similar to that of treated and untreated male worms, possibly indicating that, as is the case for the active DNA replication machinery, the active ncRNA metabolism in control females is related to protein synthesis involved in embryonic division and oviposition. Considering that the oviposition pathways in S. mansoni are not annotated in the GO database, and that correct eggshell formation is essential for embryonic cell differentiation and development, we searched for genes described in the literature as related to egg formation [15,57]. Fig 13C shows the expression levels of genes encoding four eggshell proteins, two tyrosinases, one serine/threonine kinase, all involved in egg production, and nanos-1 and -2, related to egg production and cell differentiation [58,59]; the list of these genes with Smp ID numbers and gene descriptions is given in S8 Table. All genes except NEK7 serine/threonine kinase were downregulated by GSK343 treatment (Fig 13C). Accordingly, we have searched for genes described in the literature as related to the tegument and gut [60–62] and we found 19 gut genes downregulated in males exposed to GSK343, including nine genes from the microexon gene (MEG) family (MEG-1, two MEG-4, MEG-6, two MEG-8, MEG-9, MEG-11, MEG-14), one annexin, one tetraspanin family member, two phospholipase genes, two genes encoding saposin domain containing proteins, two cathepsin peptidases, the phosphatidylcholine sterol acyltransferase gene and a gene encoding a 25 kDa integral membrane protein (S9 Table). In females, 5 gut genes were downregulated including one from the MEG-4 family, one encoding the venom allergen-like 7 protein (VAL 7), two tRNA synthetases and one phospholipase gene (S9 Table). Schistosomula early gut genes expression was as well extensively affected by GSK343 exposure, with 19 downregulated genes, including genes encoding MEGs, cathepsins, saposins, VAL 7, tetraspanins, phospholipase and palmitoyl protein thioesterase (S9 Table). Interestingly, only 2 gut genes were upregulated in GSK343-exposed males, namely one annexin and one zinc finger DHHC type gene, whereas in females only 6 gut genes were upregulated, namely three tetraspanins, one annexin and two palmitoyl protein thioesterase genes; no gut genes were upregulated in schistosomula (S9 Table). An additional 16 genes related to the tegument [61,62] were found downregulated by GSK343 exposure of male worms, namely the genes encoding aquaporin-9, dysferlin1, sodium coupled neutral amino acid transporter, Na/K ATPase beta subunit, calcium transporting ATPase, cation transporting ATPase, MDR transporter, Sm200 surface glycoprotein, seven different tegument-allergen-like proteins including Sm13 and a hypothetical protein (Smp_081920) (S9 Table). In females, 10 genes related to the tegument were found downregulated, including the genes encoding aquaporin-9, Na/K ATPase alpha and beta subunits, two cation-transporting ATPases, three MDR transporters and two tegument-allergen-like proteins including Sm13 (S9 Table). In schistosomula, 4 genes associated with the tegument were found downregulated, encoding a sodium/chloride dependent transporter, a tetraspanin, the Sm200 surface glycoprotein and a tegument-allergen-like protein (S9 Table). Again, much fewer genes associated with the tegument were found upregulated in GSK343-exposed parasites compared with the downregulated, namely 4 genes in males, encoding a transient receptor potential cation channel, two MDR transporters and acetylcholinesterase; in females, 5 genes were found, encoding acetylcholinesterase, a RAB18 RAS oncogene family member, carbonic anhydrase, major tegumental antigen Sm15 and a hypothetical protein (Smp_105220) (S9 Table). No upregulated genes associated with the tegument were found in GSK343-exposed schistosomula. In male worms, GSK343 treatment caused downregulation of genes belonging to enriched Molecular Function GO terms related to the membrane constituents and transport of water, glycerol and urea (Fig 10B). Genes related to the structural constituents of the extracellular matrix were downregulated and belong to a number of enriched Cellular Component and Molecular Function GO terms (Fig 10B). Upregulated genes in males belong to enriched Biological Process GO terms related to regulation of RNA metabolic process and regulation of transcription (Fig 11). In schistosomula treated with GSK343, the set of downregulated genes that belong to enriched Molecular Function GO terms are related to peptidase activity and those belonging to Biological Process GO terms are related to proteolysis (Fig 14A). Also downregulated were genes belonging to enriched Cellular Component GO terms related to membrane constituents (Fig 14A). Of note, very few genes were upregulated in schistosomula treated with GSK343 (29 genes) (Fig 9) and only one gene was present in each of the enriched GO terms that have been detected (Fig 14B), which include a number of Biological Process GO terms related to DNA repair and DNA replication proof-reading (Fig 14B). A set of sixteen differentially expressed genes with statistically significant change of expression was selected for RT-qPCR validation of the RNA-Seq results (S1 Text, Fig B). The selection was based on the following criteria: genes downregulated in treated females (S1 Text, Fig B (i) ) or schistosomula (S1 Text, Fig B (ii) ) that are involved in egg formation, and participate in cell differentiation, signaling pathways, membrane constituents, PRC2 complex and interaction between the mating pair. Tested genes involved in egg biosynthesis were all confirmed by qPCR as downregulated in treated females (S1 Text, Fig B (i) ); they are p48, p14, ESP (egg shell protein), Tyrosinases 1 and 2. The other genes confirmed by qPCR as downregulated in GSK343 treated-females were the nanos gene, which is involved in vitelline cell differentiation [58], the EED gene, which encodes a PRC2 component that is required for regulation of the catalytic activity of the EZH2 enzyme, and EEFB1 (eukaryotic translation elongation factor 1 beta) (S1 Text, Fig B (i) ). In schistosomula, genes that were validated by qPCR as downregulated (S1 Text, Fig B (ii) ) included MEG-4 and Val7, present in the digestive tract of the different developmental forms of S. mansoni [47,63]. Other validated genes encode LDL receptor, FA binding protein and Glucose transport protein (Smp_105410), components of the tegument membrane. Genes encoding the gynecophoral canal protein, involved in mating, the Dnase II protein, involved in degradation of exogenous DNA for cell homeostasis, and the FGFR substrate 3, a ligand of the FGFR that stimulates the Ras pathway, were also validated by qPCR as significantly downregulated in treated schistosomula (S1 Text, Fig B (ii) ). Given the fact that RNA-seq analyses suggested that GSK343 treatment of parasites caused a marked decrease in the expression of 51 genes belonging to the ncRNA GO term and involved with protein synthesis pathways (see above), we decided to perform label-free quantitative mass spectrometry-based proteomic analyses to identify and quantitate proteins differentially abundant between treated and untreated parasites. A total of 1136 proteins (S10 Table) were identified in total protein extracts of adult worms treated with GSK343 or of controls (0. 1% DMSO). Among this set of proteins, 58 were significantly differentially abundant (S1 Text, Fig C), of which 24 of them (41%) were detected as less abundant in treated parasites compared with controls (S1 Text, Fig C). The list of proteins with the Smp ID number, gene product description and the LFQ values is given in S11 Table. Interestingly, three proteins related to the protein synthesis pathways were detected as less abundant in treated worms compared with controls, namely Ribosomal protein S12 (Smp_060090), ATP dependent RNA helicase Ddx1 (Smp_163110) and Ribosomal protein l5 (Smp_200920) (S1 Text, Fig C). Two proteins that were found more abundant in treated parasites, namely calnexin (Smp_043150) and calreticulin (Smp_030370) (S1 Text, Fig C), specifically act to retain unfolded or unassembled N-linked glycoproteins in the endoplasmic reticulum [64]. Three additional proteins detected by LFQ MS as less abundant in treated worms (S1 Text, Fig C) caught our attention because they were downregulated as well in the RNA-Seq of schistosomula and adult worms. The first was epidermal growth factor receptor (Smp_035260), an important key mediator of cell communication during animal development and homeostasis; the second was fatty acid binding protein (Smp_046800), which is related to the trafficking of molecules and the formation of membranes; the third was tegument-allergen-like protein (Smp_086470) that may play a role in parasite–host interactions such as nutrient transport, environmental signal transduction, and evasion of host’s immune system [65,66]. Of note, an additional twenty proteins were uniquely detected (S12 Table) in the control samples (16 proteins) or in samples of treated adult worms (4 proteins). Finally, staphylococcal nuclease domain containing-1 (SND1) (Smp_081570) was detected as less abundant in treated parasites (S1 Text, Fig C); SND1 is a very conserved multifaceted protein that participates in various intracellular processes, such as RNA interference, acting as a nuclease in the RNA-induced silencing complex (RISC), mRNA splicing and stability, and regulation of transcription acting as a co-activator [67]. In the present work we show that GSK343, a known inhibitor of human EZH2 [17], is a schistosomicidal compound, effectively acting in vitro against female and male adult worms. We have previously detected that GSK343 acted synergistically with TSA, a histone deacetylase (HDAC) inhibitor, to enhance the death of schistosomula in vitro [15]. Recently, Roquis et al. [68] have used GSK343 to show that histone H3K27 trimethylation is required for life cycle progression from miracidium to sporocyst. The LD50 for GSK343 was previously determined to be 24. 5 μM [15], and here we confirmed that exposing adults and schistosomula for 24 h to 20 μM GSK343, a sub-lethal dose below the LD50, caused only a 10% decrease both in motility and in adult worm pairing. Notably, under this condition the compound already produced extensive morphological changes in the tegument of adults, and a significant reduction in egg size and oviposition. Also, a marked slow-down in schistosomula growth and development was observed. Similarly, juvenile schistosomula, which are resistant to praziquantel [23] and become sensitive when PZQ is combined with other compounds [22,29], had their tegument significantly affected by GSK343 after 48 h of exposure, although exhibiting a lesser decrease in viability as measured by ATP quantitation, when compared with adult worms. This is consistent with the finding that juvenile worms express higher levels of multidrug resistance-associated protein 1 (SmMRP1) compared with adult worms [69]. Interestingly, RNA-seq large-scale gene expression analyses of GSK343-treated adult females showed 32 genes with significantly downregulated expression that are related to DNA replication. It is noteworthy that TSA inhibition of histone deacetylases (HDACs) caused an opposite effect, namely a marked increase in gene expression of 21 genes involved in DNA replication [15]; it is apparent that the DNA replication program in S. mansoni is submitted to an extensive control by both histone trimethylation and deacetylation, which may be one factor that contributes to the previously reported synergy between TSA and GSK343 [15]. In addition, a pronounced downregulation in the expression of 51 genes related to ncRNA metabolism, such as genes encoding rRNA proteins and tRNA processing enzymes, was observed here, implying a possible reduction of the protein synthesis pathways in GSK343-treated females. In future studies it would be interesting to look at the GSK343 effect on the expression levels of noncoding genes such as tRNAs, by preparing RNA-seq libraries from ribosomal-RNA-depleted samples instead of from polyA+ RNA samples. It is interesting to note that label-free quantitative proteomic analyses showed that three ribosomal proteins are less abundant in total protein extracts from adult worms treated with GSK343, compared with controls. An interesting characteristic of GSK343-treated females was that the overall gene expression profiles of DNA replication genes and of ncRNA metabolism genes were more correlated to the respective expression profiles of control and treated males, and quite different from control females. Recently, unpaired females, in which the gonads are not developed, were also shown to have an overall gene expression profile more correlated to that of males [70]. Because mature paired female schistosomes lay approximately 350 eggs per day [71], they must be metabolically more active than males in order to support oogenesis. The fact that DNA replication and ncRNA metabolism gene profiles of GSK343-treated paired females are similar to those of males suggests that GSK343 predominately inhibits DNA replication and protein synthesis of actively dividing embryos in the female egg-producing gonads. In this respect, we found that nanos, a gene that is highly transcribed in the vitellarium [58], was downregulated in GSK343-treated females. Because nanos is also a stem cell gene marker in S. mansoni somatic stem cells [72] and larvae [59] it should be interesting in a future work to test the possible effects of GKS343 treatment on stem cells, using the EdU labeling assay [72]. GSK343 produced a decrease in egg laying in the adult female worms, and eggs were deformed. It is unlikely that deformed eggs would hatch, and detailed future studies on the possible blocking effect of GSK343 on egg development are warranted. It is noteworthy that in the parasites that have been exposed to GSK343 a number of downregulated genes known to be related to the tegument do correlate with the phenotypic alterations detected by electron microscopy. Thus, dysferlin, which belongs to a family of genes similar to C. elegans ferlin and plays an important role in muscle fiber membrane repair [73] was found downregulated in males and could correlate with tegument surface damage. Also, sodium coupled neutral amino acid transporter, found downregulated in males, is known to influence the cell content of most amino acids, thus determining the overall size and the composition of the intracellular amino acid pool [74]. As amino acids represent a large fraction of cell organic osmolytes, changes of sodium coupled neutral amino acid transporter activity are followed by modifications in both cell amino acids and cell volume [74]; a decrease in the expression of this gene caused by GSK343 could be related to the appearance of bubbles seen in the male tegument. In addition, the expression downregulation of aquaporin 9 and of a number of ion transporters, observed in the GSK343-exposed parasites, probably resulted in a dysregulation of fluid, small solutes and ions homeostasis and may lie behind the distended appearance of the subtegumental region and the presence of bubbles in the tegument. Treatment of adult worms and schistosomula with GSK343 was accompanied by a reduction in the H3K27me3 repressive mark, and by the upregulation of expression of hundreds of genes, as expected from a direct effect of an inhibitor of the EZH2 histone modifying enzyme (the enzyme that places the H3K27me3 repressive mark). In this respect, it should be noted that in females, males and schistosomula, some expression variability was observed among the control sample replicates as well as among the treated sample replicates, and that this is not unusual when dealing with non-clonal whole organisms. Nevertheless, a closer inspection of the expression heatmaps of Figs 8 and 9 shows that even when one replicate is more discordant from the others, most of those discordant genes have shown a change in expression (compared with their corresponding controls) that goes in the same direction as the change detected in the other replicates (albeit with a lesser magnitude of change). Interestingly, downregulation in GSK343-treated females of tens of genes involved in DNA replication and ncRNA metabolism was observed, indicating that this was an indirect effect of epigenetic marks reprograming eventually related to an increase in expression of a transcriptional co-repressor [75] induced by the treatment. Recent studies have shown that, as in vertebrates, S. mansoni also has a vast network of epigenetic marks [68] that could act in a stage-specific way on transcriptional control, resulting in different profiles of gene expression throughout its life cycle [15,76–78]. The present results suggest that DNA replication and ncRNA metabolism are two important processes related to egg-laying whose gene expression undergoes significant control by histone modifications. This is in addition to the known metabolic processes that control schistosome egg production such as fatty acid oxidation and the oxidative phosphorylation pathway [79], which may be required for fueling the energy required for replication and protein synthesis. Other egg-laying epigenetic regulators have recently been described, such as genomic DNA cytosine methylation [77], and two histone acetyltransferase modifying enzymes, SmGCN5 and SmCBP1 [57]. At this point we cannot identify the detailed mechanisms involved in expression downregulation of genes involved in DNA replication and ncRNA metabolism, which should be the subject of future studies; we speculate that such regulators could involve sex-biased microRNAs [78,80] and microRNAs involved in the regulation of ovary development [81]. The histone modifying enzymes are the central axes of the epigenetic regulation in eukaryotes, being involved in different signaling pathways. Our work points to S. mansoni EZH2 as a potential target for development of new schistosomicidal drugs. One concern is the ability to develop inhibitors that are active on the parasite and not on the human enzyme. In this respect, the parasite may be uniquely susceptible to a particular inhibitor, because schistosomes have only one H3K27 methyltransferase [12] whereas humans have two, namely EZH1 and EZH2 [82]. This difference suggests that the human enzymes may have undergone some particular evolutionary changes with respect to the single EZH2 in the parasite. Further work is required to screen existing compound libraries for selective inhibitors and/or to develop selective inhibitors based on structural analyses.

Schistosomiasis is a chronic and debilitating disease caused by a trematode of the genus Schistosoma. The current strategy for the control of the disease involves treatment with praziquantel, the only available drug. The development of new drugs is therefore a top priority. Drugs that inhibit histone modifying enzymes have been used in cancer, altering gene expression, replication, repair and DNA recombination. Schistosoma parasites have some characteristics similar to malignant tumors, such as intense cell division and high levels of metabolic activity. Here we evaluate in Schistosoma mansoni the effect of GSK343, an inhibitor of the histone methyltransferase EZH2 that had been shown to arrest or reduce the growth of human cancer cells. We show that GSK343 causes damage to the parasite tegument and reduces egg laying in vitro, concomitant with a decrease in levels of H3K27me3, the histone mark put in place by EZH2. RNA-seq and proteomic analyses of treated parasites showed changes in the expression of hundreds of genes involved in important metabolic processes. In females, a marked decrease was observed in the expression of genes related to processes such as DNA replication and noncoding RNA metabolism. In conclusion, the histone methyltransferase EZH2 seems to be a promising novel drug target against schistosomiasis.

Abstract Introduction Methods Results Discussion

schistosoma invertebrates schistosoma mansoni helminths metabolic processes dna-binding proteins light microscopy animals dna replication microscopy genome analysis dna research and analysis methods genomics proteins fluorescence microscopy gene expression scanning electron microscopy histones gene ontologies biochemistry eukaryota nucleic acids genetics electron microscopy biology and life sciences computational biology metabolism organisms

2018

Inhibition of histone methyltransferase EZH2 in Schistosoma mansoni in vitro by GSK343 reduces egg laying and decreases the expression of genes implicated in DNA replication and noncoding RNA metabolism

Tamoxifen is one of the most commonly employed endocrine therapies for patients with estrogen receptor α (ERα) -positive breast cancer. Unfortunately the clinical benefit is limited due to intrinsic and acquired drug resistance. We previously reported a genome-wide association study that identified common SNPs near the CTSO gene and in ZNF423 associated with development of breast cancer during tamoxifen therapy in the NSABP P-1 and P-2 breast cancer prevention trials. Here, we have investigated their roles in ERα-positive breast cancer growth and tamoxifen response, focusing on the mechanism of CTSO. We performed in vitro studies including luciferase assays, cell proliferation, and mass spectrometry-based assays using ERα-positive breast cancer cells and a panel of genomic data-rich lymphoblastoid cell lines. We report that CTSO reduces the protein levels of BRCA1 and ZNF423 through cysteine proteinase-mediated degradation. We also have identified a series of transcription factors of BRCA1 that are regulated by CTSO at the protein level. Importantly, the variant CTSO SNP genotypes are associated with increased CTSO and decreased BRCA1 protein levels that confer resistance to tamoxifen. Characterization of the effect of both CTSO SNPs and ZNF423 SNPs on tamoxifen response revealed that cells with different combinations of CTSO and ZNF423 genotypes respond differently to Tamoxifen, PARP inhibitors or the combination of the two drugs due to SNP dependent differential regulation of BRCA1 levels. Therefore, these genotypes might be biomarkers for selection of individual drug to achieve the best efficacy. Approximately 80% of breast tumors express estrogen receptor α (ER) [1–3], a receptor that binds and mediates many of the effects of estrogens. Estrogen signaling is known to modulate several processes relevant to breast cancer cell proliferation, predominately as a result of the activity of ER as a transcription factor [4]. Therefore, selective estrogen receptor modulators (SERMs) such as tamoxifen have been widely used clinically in endocrine therapies for patients with ERα-positive (ERα+) breast cancer [5–7]. Tamoxifen is not only effective in the treatment of ERα+ breast cancer, but it is also effective in the chemoprevention of breast cancer [8,9]. However, resistance to tamoxifen therapy also occurs in that 22. 7% of patients treated in the adjuvant setting had recurrence of breast cancer by 10 years in a meta-analysis, and in the prevention setting [10] tamoxifen reduces risk by 49%, but the number needed to treat to prevent one case of breast cancer is in excess of 50 [8]. Several mechanisms have been associated with resistance to tamoxifen [11,12]. Of particular importance are the effects of estrogen/ER on BRCA1. The BRCA1 protein directly interacts with ERα and inhibits ERα transactivation and downstream signaling [13]. Decreased BRCA1 expression has been shown to be present in 30–40% of sporadic breast cancers [14]. BRCA1 deficiency is known to play a role in breast cancer development. Furthermore, decreased BRCA1 expression results in tamoxifen resistance by altering ERα co-regulator association in breast cancer cells [15]. These findings suggest that BRCA1 may regulate the response of ERα to its canonical ligand E2 and to tamoxifen, a compound known to exert either agonistic or antagonistic activity toward ERα in different cellular and tissue contexts [16]. In addition, BRCA1 is also known to play a major role in the DNA double-strand break (DSB) repair during the S and G2 phases by mediating homologous recombination (HR) to maintain replication fidelity and genome integrity [17]. Studies have demonstrated that BRCA1 dysfunction results in the lack of HR and markedly sensitizes cells to the inhibition of PARP enzymatic activity, which seemed to be attributable to the persistence of DNA lesions that are normally repaired by homologous recombination [18,19]. Therefore, genetic factors that might contribute to BRCA1 regulation could significantly affect response to drugs like SERMs and PARP inhibitors. Our previous case-control genome-wide association study (GWAS) performed with samples from the NSABP P-1 and P-2 breast cancer SERM chemoprevention trials identified two SNP signals that were associated with breast cancer risk, including one in which the variant SNP genotype near the CTSO gene was associated with increased risk for the development of breast cancer and a second signal for which the variant SNP genotype in the ZNF423 gene was associated with decreased risk for the development of breast cancer in women treated with tamoxifen or raloxifene [20]. ZNF423 appeared to be a transcription factor that regulated BRCA1 expression in an estrogen-dependent fashion, while CTSO also showed weak estrogen-dependent induction of BRCA1 mRNA expression in a CTSO SNP-dependent fashion [20]. In a separate study, it was also shown that the variant GG genotype for the CTSO rs10030044 SNP was an independent factor indicating a poor prognosis in ER+ breast cancer patients receiving adjuvant tamoxifen therapy [21], which suggested the involvement of this genetic locus in tamoxifen response. CTSO, cathepsin O, is a member of the cysteine protease family that is involved in cellular protein degradation and turnover. Another member, cathepsin D, has been associated with poor prognosis for breast cancer as a result of stimulation of breast cancer cell proliferation, fibroblast outgrowth, angiogenesis, breast tumor growth and metastasis formation [22]. Even though in our previous study, we have observed a correlation between CTSO and BRCA1 in an estrogen and SNP dependent fashion, how CTSO regulates BRCA1 remains unclear. In the current study, based on our prior findings [20], we investigated the possible role of CTSO in drug response and breast cancer risk as a result of the regulation of ZNF423 and BRCA1. Finally, we also explored the role of both ZNF423 and CTSO SNP genotypes to help selection of tamoxifen and PARP inhibitors. Our previous GWAS involved 592 cases and 1171 matched controls selected from the 33,000 participants enrolled in the NSABP P-1 and P-2 breast cancer prevention trials identified two SNPs on chromosome 4 (rs10030044 and rs4256192) that were associated with breast cancer risk, with odds ratios of 1. 42 and 1. 44 respectively [20]. To gain a comprehensive understanding of the contribution of genetic variants in that region, together with the two top genotyped SNPs, based on our previous imputation results [20], we chose additional six imputed SNPs associated with increased risk for the development of breast cancer (OR 1. 42–1. 45) with adjusted p-values < 5. 00E-6 (rs6835859, rs4550865, rs10030044, rs62328155, rs11737651, rs6810983, rs4256192, rs11724342). All eight SNPs were located at 5′ of the CTSO gene. These variant SNP genotypes are common with MAFs ranging from 0. 39 to 0. 45. We then performed linkage disequilibrium (LD) analysis and the analysis showed that all 8 SNPs were in significant linkage with each other. The top two genotyped SNPs, rs10030044 and rs4256192 were in strong LD (r2 = 0. 78). The SNP rs10030044 was also in strong LD with the three imputed SNPs: rs6835859 (r2 = 1), rs4550865 (r2 = 1), and rs6810983 (r2 = 1), while the rs4256192 SNP was in strong LD with the other three imputed SNPs: rs11724342 (r2 = 1), rs62328155 (r2 = 1) and rs11737651 (r2 = 1). Because of the importance of understanding breast cancer risk and because P-1 and P-2 are the largest breast cancer chemoprevention trials ever performed, we pursued the possible functional implications of these SNP signals. We began by analyzing the top 8 SNPs for their associations with expression levels of all genes including CTSO within 1 Mb up- and downstream of the SNPs of interest using the Genotype-Tissue Expression (GTEx) database. Although, we did not find eQTL relationships between these SNPs and CTSO in normal breast tissue in GTEx, significant eQTL associations between the SNPs and CTSO were present in stomach, skin, pancreas, and testis. The variate SNP was associated with higher CTSO expression (p = 0. 0077–4. 3E-7). We did not observe eQTL relationships between these SNPs and CTSO at baseline in our panel of LCLs for which we had genome-wide genotype data and mRNA expression data [23]. Because 94. 2% of the participants on P-1 and P-2 were Caucasian, our GWAS was restricted to only Caucasian subjects [20]. Therefore, we randomly selected LCLs from Caucasians that were either homozygous wild type (WT) or variant for the SNPs 5’ of CTSO to validate the eQTL relationships in a setting mimicking the estrogenic environment in patients. These LCLs were grown in medium containing charcoal-treated serum to deplete the levels of endogenous steroids and supplemented with physiological concentrations of E2. CTSO mRNA and protein were higher in LCLs homozygous for the variant genotype as compared with LCLs homozygous for the WT genotype (p<0. 05; Fig 1A). However, the induction of CTSO mRNA was more significant in the WT than variant cells, consistent with our previous finding [20], even though the variant cells had higher baseline level of CTSO (S1 Fig). We next determined which of the SNPs 5’ of CTSO might influence expression. Our previous study suggested that the expression of CTSO was estrogen-dependent, and only the rs6810983 SNP disrupted an estrogen response element (ERE) for the variant SNP genotype [20]. We decided to directly determine the possible role of these eight SNPs in transcription regulation using luciferase reporter gene assays performed in ZR75-1 breast cancer cells. Specifically, we cloned a 200 bp DNA sequence that included either WT or variant sequence for each of the eight SNPs, together with the CTSO promoter, into the pGL3 basic reporter plasmid. We then transfected these constructs into the ER+ cell line, ZR75-1 cells in a normal medium with 10% FBS. Cells transfected with constructs with variant genotypes for rs10030044 and rs6810983 SNPs displayed 2–3 fold greater luciferase activity than did those transfected with constructs with WT SNP sequences, indicating increased transcriptional activity (Fig 1B) —compatible with the results in LCLs. We then determined the possible functional effect of CTSO on BRCA1 based on our previous finding [20]. We genotyped the ZNF423 SNP and CTSO SNP in a panel of breast cancer cell lines and chose T47D, CAMA-1, and ZR75-1 cell lines carrying homozygous genotypes for ZNF423 and CTSO SNPs (S1 Table) for further functional study. When CTSO was overexpressed significantly in T47D, CAMA-1, and ZR75-1 cells, there was a striking decrease of BRCA1 protein levels as well as protein levels for the BRCA1 transcription factor, ZNF423, in all the cell lines tested (Fig 2A, left panel). To determine how generalizable this phenomenon might be, we also measured the level of BRCA1 protein in triple negative MDA-MB-231 breast cancer cells. In agreement with ER+ breast cancer cell line data, BRCA1 protein was significantly decreased after overexpressing CTSO in triple negative breast cancer cells (Fig 2A, left panel). Quantitative RT-PCR revealed excellent transfection efficiency of CTSO in all of the cell lines, with modest but statistically significant decreases in BRCA1 transcript levels (Fig 2A, right panel), while ZNF423 mRNA remained unchanged after CTSO overexpression (Fig 2A, right panel). Next, we asked whether CTSO might influence BRCA1 and ZNF423 protein stability through its cysteine proteases activity. Overexpression of CTSO decreased ZNF423 and BRCA1 protein levels in CAMA-1 and ZR75-1 cells, while treatment with the cathepsin inhibitor E-64 resulted in increased levels of BRCA1 and ZNF423 protein (Fig 2B). Previous work has largely focused on CTSO SNP-dependent estrogen induction of CTSO and BRCA1 mRNA in LCLs. Consistent with our previous finding [20], both CTSO and BRCA1 mRNA was moderately induced by E2 in LCLs with WT CTSO SNP genotype (S1 Fig). However, in this study, we further demonstrated that, more importantly, CTSO can also directly regulate BRCA1 protein turnover in breast cancer cells. Since CTSO is able to stimulate BRCA1 and ZNF423 protein degradation, we determined the possible interaction between CTSO and BRCA1or ZNF423. Immunoprecipitation using CTSO antibody showed endogenous interaction of CTSO with BRCA1 and ZNF423 (Fig 2C). These results indicated that CTSO regulates BRCA1 and ZNF423 protein stability through a cysteine protease- mediated degradation pathway—at least in part. We next examined possible mechanisms by which CTSO might influence BRCA1 transcription. We first confirmed that knockdown of CTSO resulted in increased BRCA1 expression, both at the mRNA and protein levels in both CAMA-1 and ZR75-1 cells (Fig 3A). Our previous GWAS study had reported that ZNF423 binds to the 5′-flanking region of BRCA1 and regulates BRCA1 transcription [20]. We also showed in the present study that CTSO interacts with ZNF423, leading to ZNF423 degradation (Fig 2), suggesting that CTSO may regulate BRCA1 transcription partially through its effect on ZNF423. In order to identify additional factors involved in the CTSO-dependent regulation of BRCA1 transcription, we performed mass spectrometry screening of a pool of proteins that co-precipitated with CTSO. During this process, we identified 130 proteins that interacted with CTSO (S2 Table). We then interrogated the Cancer Genome Atlas (TCGA) breast cancer data [24] for possible relationships between the expression of BRCA1 and these 130 genes, and identified 20 genes that were associated with BRCA1 with p< 1E-05 (S3 Table). We then knocked down these 20 genes to determine the effect on BRCA1 levels (S2 Fig), and found that knockdown of 4 out of the 20 genes, MTDH, PABPC4L, LMNA, and EEF1A1, resulted in striking decreases of BRCA1 mRNA expression level (Fig 3B), consistent with the TCGA data that showed positive correlations between these 4 genes and BRCA1. Furthermore, in CAMA-1 and ZR75-1 cells, overexpression of CTSO decreased expression of all four genes (Fig 3C), which could explain the down-regulation of BRCA1 mRNA level when overexpressing CTSO (Fig 2A). In summary, these results indicate that the up-regulation of CTSO could reduce BRCA1 levels by promoting the cysteine protease—mediated degradation of MTDH, PABPC4L, LMNA, and EEF1A1 protein levels in addition to the effect on ZNF423 that we had already identified, all of which regulate BRCA1 transcription. Thus, it appears that tumor expression of CTSO may play a role in the regulation of BRCA1 transcription in addition to having an effect on BRCA1 protein degradation. We hypothesized that, because CTSO regulates BRCA1 stability, it may play a role in endocrine resistance. Previous studies demonstrated that BRCA1over-expression can inhibit cell proliferation by activating p21WAF1/CIP1 [25,26]. We had demonstrated that CTSO regulates the stability of BRCA1 (Fig 2). Therefore, we next determined whether the down-regulation of CTSO inhibited cell proliferation in breast cancer cells due to the up-regulation of BRCA1. BRCA1 protein increased after CTSO knockdown in CAMA-1 and ZR75-1 cells (Fig 4A, lower panel). Depletion of CTSO inhibited cell growth compared with negative siRNA transfected control cells (Fig 4A, upper panel). To further confirm that the CTSO effect on cell proliferation was mediated through the regulation of BRCA1, we knocked down BRCA1 in cells with down-regulation of CTSO. Knockdown of BRCA1 in CTSO-depleted cells resulted in the abrogation of decreased proliferation due to CTSO depletion in both cell lines (Fig 4A, upper panel). We next tested the effect of CTSO on tamoxifen treatment based on the observations from our previous study [20] and others. In the presence of 100 nM 4OH-tamoxifen (4OH-TAM), CTSO-deficient cells exhibited increased sensitivity to 4OH-TAM compared with negative siRNA-transfected control cells (Fig 4B), and BRCA1 might be responsible for the increased sensitivity since BRCA1 depletion in siCTSO cells significantly decreased 4OH-TAM sensitivity (Fig 4B). These results demonstrated that depletion or inhibition of CTSO can increase BRCA1 levels with potential therapeutic effects, resulting in growth arrest. Since our previous study had identified ZNF423 and CTSO SNPs that were associated with breast cancer risk [20], both of which appeared to regulate BRCA1, we examined their joint effect on cell proliferation in the presence of tamoxifen or E2 treatment. We utilized a model system consisting of 300 individual human LCLs (100 European-American, 100 African-American and 100 Han Chinese-American subjects). The “Human Variation Panel” that had been SNP genotyped previously and has repeatedly demonstrated its value as a platform to study genetic variants [20,27,28]. Specifically, we selected 4 groups of LCLs to perform 4OH-TAM treatment: Notably, in the presence of 4OH-TAM, the growth of CTSO WT/ZNF423 WT and CTSO V/ZNF423 V cells decreased significantly (Fig 5A and 5B, and Table 1) suggesting that the therapeutic effects of tamoxifen are seen mainly in the CTSO WT/ZNF423 WT and CTSO V/ZNF423 V groups, not the CTSO WT/ZNF423 V and CTSO V/ZNF423 WT groups (Fig 5C and 5D, and Table 1). We also measured BRCA1, CTSO and ZNF423 protein levels in cells with different ZNF423 SNP and CTSO SNP combinations (Fig 6). The estradiol-, 4OH-TAM -dependent and SNP-dependent regulation of BRCA1 protein level was more pronounced against the background of homozygous variant for the CTSO SNP. BRCA1 protein level in the CTSO V / ZNF423 WT group was significantly upregulated in the presence of E2 and then decreased upon addition of 4OH-TAM treatment (Fig 6). The opposite effects on BRCA1 protein level upon treatment of E2 or E2 plus 4OH-TAM were observed in CTSO V / ZNF423 V group compared with CTSO V / ZNF423 WT group (Fig 6B). The higher BRCA1 level in CTSO V / ZNF423 V group compared to the CTSO V / ZNF423 WT group in the presence of 4OH-TAM could explain the tamoxifen response seen in CTSO V / ZNF423 V group, but not in CTSO V / ZNF423 WT group (Figs 4 and 5B and 5D). In the presence of TAM, cells with CTSO W / ZNF423 W genotype were also showed relatively higher BRCA1 levels, even though with this genetic background the baseline BRCA1 was higher compared with other genotype groups (Fig 6B). Therefore, cells with CTSO W / ZNF423 W also benefit from TAM treatment (Fig 5A). We also measured ER level in these four groups of LCLs upon different treatment to account for its potential impact, and did not observe difference in ER level among the four genotype combination groups, furthermore, E2 and TAM treatment did not change the level of ER compared to vehicle treatment for each genotype combination (Fig 6). Therefore, the ZNF423 and CTSO SNPs-dependent effects on TAM response were not due to ER expression level. When compared the cell proliferation in the presence of different treatments among different genotypes, cells with CTSO V/ZNF423 W showed the fastest growth rate, regardless of whether they received no treatment, estradiol (E2) alone, 4OH-TAM alone, or the combination of E2 plus 4OH-TAM (S3 Fig), while cells with CTSO WT/ZNF423 V grew slowest among all genotype combination groups in all treatment groups (S3 Fig). This was consistent with our previous finding that the odds ratios for CTSO V/ZNF423 W (OR = 5. 71) was the highest, and that for CTSO WT/ZNF423 V (OR = 1. 00) was the lowest for breast cancer risk in the P-1, P-2 trials [20]. Loss of BRCA1 function leads to defects in the HR DNA repair pathway, which renders cells more sensitive to PARP inhibitors [29–32]. In BRCA1/2 mutated cells, the DSBs at the replication fork caused by PARP inhibitor treatment cannot be repaired, resulting in synthetic lethality and cell death. We have shown that the LCL CTSO WT/ZNF423 WT (Fig 5A) and CTSO V/ZNF423 V (Fig 5B) groups respond to 4OH-TAM treatment but not the CTSO WT/ZNF423 V (Fig 5C) and CTSO V/ZNF423 WT (Fig 5D) groups (Table 1). In addition, comparing the two 4OH-TAM-resistant groups, CTSO WT/ZNF423 V cells showed higher BRCA1 level upon 4OH-TAM treatment than CTSO V/ZNF423 WT cells (Fig 6B). As a result, we hypothesized that the combination of a PARP inhibitor and 4OH-TAM might achieve better therapeutic outcomes in the CTSO V/ZNF423 WT group that displayed lower levels of BRCA1. To determine the effect of a PARP inhibitor in this setting, we treated 4OH-TAM-responsive CTSO WT/ZNF423 WT and CTSO V/ZNF423 V LCLs as well as 4OH-TAM-resistant CTSO WT/ZNF423 V and CTSO V/ZNF423 WT LCLs with either 4OH-TAM alone or 4OH-TAM plus the PARP inhibitor, olaparib. Olaparib did not increase 4OH-TAM sensitivity in the two 4OH-TAM-responsive CTSO WT/ZNF423 WT and CTSO V/ZNF423 V groups (Fig 7A, upper panel, and Table 1). However, olaparib significantly sensitized the 4OH-TAM-resistant CTSO V/ZNF423 WT cells to tamoxifen treatment, but not the CTSO WT/ZNF423 V cells (Fig 7A, lower panel, and Table 1). The differential effects of olaparib in the two 4OH-TAM-resistant groups can be explained, at least partially, by the differences in BRCA1 levels (Fig 6B). Upon 4OH-TAM treatment, the 4OH-TAM-resistant CTSO V/ZNF423 WT cells had lower BRCA1 levels compared with the CTSO WT/ZNF423 V cells, resulting in sensitization by combining olaparib with 4OH-TAM. The 4OH-TAM-resistant CTSO WT/ZNF423 V cells had high level of BRCA1, consistent with olaparib having little effect. We also confirmed the therapeutic effect of the combination of olaparib and 4OH-TAM in ER+ breast cancer cells, CAMA-1 and ZR75-1 that had WT BRCA1 and were resistant to olaparib (Fig 7B). Knock down of CTSO resulted in striking increases of BRCA1 protein level (Fig 3A), therefore, the addition of olaparib did not increase 4OH-TAM sensitivity (Fig 7B). However, olaparib significantly increased 4OH-TAM sensitivity in cells transfected with negative control siRNA due to lower baseline BRCA1 level comparing with CTSO knockdown cells (Fig 7B, p<0. 05). 4OH-TAM showed the 50% inhibitory concentration (IC50) of 11. 22 μM for CAMA-1, and 10. 17 μM for ZR75-1 cells transfected with negative control siRNA respectively. The IC50 of 4OH-TAM decreased significantly when co-treated with olaparib in negative control siRNA transfected CAMA-1 and ZR75-1 cells (CAMA-1: IC50 = 5. 10±0. 26μM; ZR75-1: IC50 = 4. 70±0. 18 μM) (Fig 7B, p<0. 05). In summary, these results indicated that the down-regulation of CTSO could increase BRCA1 levels, resulting in decreased cell growth and potential therapeutic effects. In conclusion, we present evidence in the present study that CTSO is a new factor of importance for tamoxifen efficacy as a chemopreventive agent in women at high risk of developing breast cancer as well as evidence for a potential mechanism by which this effect involves BRCA1. The underlying mechanisms identified require validation and further refinement but they also provide pharmacogenomic insights into tamoxifen as a preventative agent. We have demonstrated that a PARP inhibitor, which can effectively restore tamoxifen sensitivity in tamoxifen—resistant ER+ breast cancer cells, might be a potentially promising addition to tamoxifen as a combination regimen for patients carrying the CTSO V/ZNF423 W SNP genotype. As a result, our study has revealed a new potential biomarker signature involving CTSO and ZNF423-related SNPs for the therapeutic stratification of patients at high risk for the development of breast cancer. Dulbecco' s minimum essential medium (DMEM), glutamine and penicillin/streptomycin/glutamine stock mix were purchased from Life Technologies, Inc. (Carlsbad, CA, USA). Fetal bovine serum (FBS) and charcoal-stripped FBS were from Invitrogen (Carlsbad, CA, USA). L-trans-Epoxysuccinyl-leucylamido (4-guanidino) butane (E-64) was from Sigma-Aldrich (St. Louis, MO USA). CTSO, MTDH, PABPC4L, LMNA, EEFiA1and control small interfering RNAs (siRNA) were purchased from Dharmacon (Thermo Scientific Dharmacon, Inc.). CTSO plasmid was purchased from OriGene (Rockville, MD, USA). Affinity purified rabbit and mouse antibodies against human BRCA1 and CTSO were from Santa Cruz Biotechnologies (Santa Cruz, CA, USA). ZNF423 antibody was purchased from Abcam (Cambridge, MA, USA). Actin, MTDH, PABPC4L, LMNA, and EEFiA1 antibodies were from cell signaling (Danvers, MA, USA). For standard PCR, HotStart Taq Plus DNA Polymerase was used (Qiagen, Germantown, MD, USA). Reagents and primers for real time PCR were purchased from Qiagen (Valencia, CA, USA). The protease inhibitor cocktail kit was obtained from Pierce Biotechnology (Rockford, IL, USA). 17β-estradiol (E2) and 4-hydroxytamoxifen (OH-TAM) were purchased from Sigma Aldrich (Saint Louis, MO USA). Olaparib was from Selleckchem (Houston, TX, USA). Lymphoblastoid cell lines (LCLs) with known genotypes for the chromosome (chr) 4 CTSO SNPs were cultured in RPMI 1640 media containing 15% (vol/vol) FBS (Invitrogen, San Diego, CA). T47D, ZR75-1, CAMA-1, MDA-MB-231 cell lines were obtained from American Type Culture Collection (ATCC) (Manassus, VA). T47D and ZR75-1 were cultured in RPMI-1640 (Grand Island, NY) containing 10% fetal bovine serum (FBS). CAMA-1 cells were cultured in Eagle' s Minimum Essential Medium containing FBS to a final concentration of 10%. MDA-MB-231 cells were cultured in Leibovitz' s L-15 Medium containing 10% FBS at 37°C without CO2. Luciferase reporter gene constructs containing various SNP genotypes were generated by PCR based mutagenesis. Specifically, a 1924 bp segment of the CTSO promoter containing ERE was PCR amplified with the primers: 5’- TAAGCAGATATCACTGACATCATGCCACACCT’ and 5- ACGATGCTGAGATTGACCCTAAGCTTTAAGCA -3’ and was cloned into the EcoRV and HindIII sites of pGL3 basic plasmid to make the pGL3-CTSO construct. A 150–250 bp DNA segment that included the rs10030044, rs6810983, rs6835859, rs4550865, rs62328155, rs11737651, and rs4256192 SNPs respectively was also PCR amplified using primers as described in S1 File. These fragments were cloned into the KpnI and NheI sites upstream of the CTSO promoter sequence to make the plasmids pGL3-WT-CTSO or pGL3-V-CTSO. The WT SNP sequence was amplified with LCL genomic DNA as a template that was homozygous for this WT SNP genotype. This variant SNP sequence was amplified using LCL genomic DNA shown to be homozygous for the variant genotype as template. These 150 -250bp amplicons contained the rs10030044, rs6810983, rs6835859, rs4550865, rs62328155, rs11737651, and rs4256192 SNPs respectively. T47D and ZR75-1 cells were then seeded in triplicate in 12-well cell culture plates at a concentration of 105 cells / well. After 24 h, the cells were transfected using Lipofectamine 2000 (Invitrogen) with 4 μg of the pGL3-WT-CTSO or pGL-V-CTSO constructs and 2 μg pRL-CMV encoding a CMV-driven renilla luciferase vector (Promega), together with the carrier DNA (pGL3 basic). Luciferase assays were performed 48 h after transfection using a luciferase reporter assay system (Promega). The renilla luciferase activity was used to correct for the transfection efficiency. The human variation panel model system consists of LCLs from 300 healthy subjects (100 European-Americans, 100 African-Americans, and 100 Han Chinese-Americans). This panel was generated by the Coriell Institute (Camden, New Jersey). We genotyped all 300 cell lines for genome-wide SNPs using Illumina 550K and 510S SNP BeadChips (Illumina), and the Coriell Institute obtained Affymetrix SNP array 6. 0 (Affymetrix) data for the same cell lines. These combined SNP genotype data (~1. 3 million genotyped SNPs) were used to impute a total of approximately 7 million SNPs per cell line. This LCL model system has been used repeatedly to generate and/or test pharmacogenomic hypotheses arising from clinical GWAS [3,12,17–19,53]. The application of these cell lines made it possible to evaluate the function of CTSO and ZNF423 SNP genotypes. To study the effect of the SNP on CTSO expression, LCLs were cultured in base media containing 5% charcoal-stripped FBS for 24 hours and were subsequently cultured in FBS-free base media containing 0. 1 nM E2 for another 48 hours. Cell lysates were used to perform Western blot analysis, and total RNA was isolated for qRT-PCR. Breast cancer cells were cultured in specific base media, as described above, supplemented with 10% FBS. 5000 cells were seeded in triplicate in 96-well plates, and were cultured in base media containing 5% (vol/vol) charcoal-stripped FBS for 24 hours and were subsequently cultured in FBS-free base media for another 24 hours. Cells were then transfected with either control siRNA or siRNA targeting CTSO. Twenty-four hours after transfection the media was replaced with fresh FBS-free base media and the cells were treated with 0. 1 nM E2 for 24 hours, and then treated with 100 nM 4-OH- tamoxifen. Cell growth was measured at different time points (0,24,48, and 72 hours) post tamoxifen treatment using the BrdU Cell Proliferation Assay kit (Cell Signaling, Danvers, MA) at intervals of 24 h following the manufacturer' s instructions. The plates were measured in a Safire2 microplate reader (Tecan AG, Switzerland). LCLs selected based on ZNF423 and CTSO genotypes were cultured in RPMI 1640 media (Cellgro) supplemented with 15% FBS. Cells were cultured in RPMI 1640 media containing 5% (vol/vol) charcoal-stripped FBS for 24 hours and were subsequently seeded in triplicate in 96-well plates and cultured in FBS-free RPMI 1640 media for another 24 hours before treatment. Cells were treated with 0. 1 nM E2,50nM tamoxifen, or the combination of both 0. 1 nM E2 and 50nM tamoxifen. Cell growth was measured at different time points (0,24,48,72, and 96 hours) post treatment using the CYQUANT Direct Cell Proliferation Assay (#C35012, Invitrogen) following the manufacturer’s instructions at intervals of 24 h. The plates were measured in a Safire2 microplate reader (Tecan AG, Switzerland). Cells were plated at 70% confluence in culture medium supplemented with 10% FBS, and were transfected with empty vector or CTSO plasmid (OriGene) using lipofectamine 2000 (Invitrogen, Carlsbad, CA) according to the vendor' s protocol. Cells were collected for protein analysis 48 hours after transfection. In some experiments, 24 hours after transfection, cells were treated with 10 μM E-64, a cysteine proteases inhibitor, for additional 24 hours. Cells were then collected for protein analysis. Specific siGENOME siRNA SMARTpool reagents against a given gene as well as a negative control, siGENOME Non-Targeting siRNA, were purchased from Dharmacon Inc. (Lafayette, CO, USA). Cells were transfected with control siRNA, and specific siRNAs (10nM) in 96-well plates or 12-well plates using lipofectamine RNAiMAX (Invitrogen, Carlsbad, CA) according to the vendor' s protocol. For the purpose of cell growth assay, cells were plated in base medium supplemented with 5% charcoal stripped FBS for 24 hours, and then cultured in FBS-free RPMI 1640 media for another 24 hours before transfection. Different treatments were started 24 hours after transfection. For the purpose of testing gene expression level, cells were transfected with control siRNA and specific siRNAs (10nM) in 12-well plates using lipofectamine RNAiMAX for 48 hours. Breast cancer cells were harvested by trypsinization, lysed in SDS buffer. Cell lysates were heated to 95°C for five minutes. Protein samples (10 to 20 μg) were resolved by electrophoresis on 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) gels and electrophoretically transferred to PVDF membranes (Millipore Corporation, Bedford, MA, USA). The blots were probed with the appropriate primary antibody and the appropriate horseradish peroxidase conjugated secondary antibody. The protein bands detected with the Pierce enhanced chemiluminescence Western blotting substrate (Thermo Scientific, Rockford, IL, USA) and were visualized using Geldoc (Bio-Rad Laboratories). LCLs selected based on ZNF423 and CTSO genotypes were cultured in RPMI 1640 media containing 5% (vol/vol) charcoal-stripped FBS for 24 hours and were subsequently seeded in 6-well plates and cultured in FBS-free RPMI 1640 media for another 24 hours before treatment. Cells were treated with 0. 1 nM E2,50nM tamoxifen, or combination of both 0. 1 nM E2 and 50nM tamoxifen for 48 hours and lysed in RIPA buffer supplemented with protease and phosphatase inhibitors. Cell lysates were used to perform Western blot analysis. Quantification of the blots was analyzed using Image J. Cells were lysed in NETN buffer (20 mM Tris-HCl, pH 8. 0,100 mM NaCl, 1 mM EDTA, 0. 5% Nonidet P-40) supplemented with protease and phosphatase inhibitors. Lysates were clarified by centrifugation (13,000 r. p. m. , 20 min, 4°C) and 500 μg–1mg proteins were used per immunoprecipitation. Proteins were captured with 2 μg CTSO antibody and protein G-sepharose Fast-Flow (Sigma). Immunoprecipitation with mouse serum was used as negative controls. The immuno-complexes were then washed with NETN buffer three times followed by separation on SDS-PAGE. Proteins were resolved by SDS–PAGE, transferred onto PVDF membranes and probed using the appropriate primary and secondary antibodies coupled to horse-radish peroxidase. Total RNA was isolated from cultured cells with the QIAGEN RNeasy kit (QIAGEN Inc. , Valencia, CA, USA), followed by qRT-PCR performed with the one-step Brilliant SYBR Green qRT-PCR master mix kit (Stratagene, La Jolla, CA, USA). Specifically, primers purchased from QIAGEN were used to perform qRT-PCR with the Stratagene Mx3005P real-time PCR detection system (Stratagene). All experiments were performed in triplicate with GAPDH as an internal control. Reverse-transcribed Universal Human Reference RNA (Stratagene) was used to generate a standard curve. Control reactions lacked the RNA template. The 2-δδcycle threshold method was used for statistical data analysis. Drugs were dissolved in DMSO, and aliquots of stock solutions were frozen at −80°C. Cytotoxicity assays were performed in triplicate at each drug concentration. Specifically, 4000 breast cancer cells were seeded in 96-well plates and were cultured in base media containing 5% (vol/vol) charcoal-stripped FBS for 24 hours and were subsequently cultured in FBS-free base media for another 24 hours. Cells were then transfected with either control siRNA or siRNA targeting CTSO. Twenty-four hours after transfection the media was replaced with fresh FBS-free base media and the cells were treated with 10 μL of tamoxifen at final concentrations of 0,0. 5,1, 2,4, 6,8, 12,24, and 48 μM with or without 10 μM olaparib. After incubation for an additional 72 hours, cytotoxicity was determined by quantification of DNA content using CYQUANT assay (#C35012, Invitrogen) following the manufacturer’s instructions. 100μL of CyQUANT assay solution was added, and plates were incubated at 37°C for one hour, and then read in a Safire2 plate reader with filters appropriate for 480 nm excitation and 520 nm emission. LCLs selected based on ZNF423 and CTSO genotypes were cultured in RPMI 1640 media containing 5% charcoal-stripped FBS for 24 hours and 5x104 cells were subsequently seeded in triplicate in 96-well plates and cultured in FBS-free RPMI 1640 media for another 24 hours before treatment. Cells were treated with 10 μL of tamoxifen at final concentrations of 0,0. 5,1, 2,4, 6,8, 12,24, and 48 μM with or without 5 μM olaparib. After incubation for an additional 72 hours, cytotoxicity was determined by quantification of DNA content using CYQUANT assay. ZR75-1 cells were transfected with CTSO plasmid. After 72 hr, cells were lysed by NETN buffer. Cell lysates were incubated with control IgG or CTSO antibody at 4°C for 4 hr, and then incubated with protein G-sepharose Fast-Flow for 2 hr. After washing with NETN buffer three times, bound proteins were eluted, and size fractionated by 10% SDS-PAGE. Coomassie-stained gel slices covering the entire molecular weight range were processed for analysis by mass spectrometer following a standard protocol at the Harvard Medical School Taplin Mass Spectrometry Facility. All data were presented as mean ± SD of at least three independent experiments. Statistical analysis was performed using SPSS22. 0 and Prism 5 (GraphPad Software Inc. , San Diego, CA, USA). Single-factor analysis of the variance test was used for comparisons among multiple groups, and a t-test was used for comparisons between two groups; P <0. 05 was considered statistically significant.

Many studies have demonstrated that germline genetic variation can contribute to both breast cancer disease risk and treatment response. However, the underlying mechanisms associated with these biomarkers often remains understudied. As part of functional genomic studies following up a case-control genome-wide association study (GWAS) performed with the large and influential National Surgical Adjuvant Breast and Bowel Project P-1 and P-2 SERM breast cancer prevention trials, we investigated the top GWAS SNPs in CTSO gene on chromosome 4 and mechanisms of CTSO involvement in the regulation of BRCA1 and response to therapy. We showed that, based on individual’s genotype, CTSO contributes differentially to tamoxifen response in ERα-positive (ER+) breast cancer cells by regulating ZNF423 and BRCA1levels and that PARP inhibitors can effectively restore tamoxifen sensitivity in subjects with unfavorable genotypes of CTSO and ZNF423 associated with tamoxifen resistance. Our work highlights the potential value of a new biomarker signature involving CTSO and ZNF423-related SNPs for selection of tamoxifen or PARP inhibitors.

Abstract Introduction Results Discussion Materials and methods

genome-wide association studies medicine and health sciences breast tumors gene regulation cancer risk factors variant genotypes cancers and neoplasms genetic mapping oncology genome analysis molecular genetics small interfering rnas transcriptional control gene expression breast cancer molecular biology genetic causes of cancer biochemistry rna nucleic acids heredity genetics biology and life sciences genomics non-coding rna computational biology human genetics

2017

SNPs near the cysteine proteinase cathepsin O gene (CTSO) determine tamoxifen sensitivity in ERα-positive breast cancer through regulation of BRCA1

Postzygotic reproductive barriers such as sterility and lethality of hybrids are important for establishing and maintaining reproductive isolation between species. Identifying the causal loci and discerning how they interfere with the development of hybrids is essential for understanding how hybrid incompatibilities (HIs) evolve, but little is known about the mechanisms of how HI genes cause hybrid dysfunctions. A previously discovered Drosophila melanogaster locus called Zhr causes lethality in F1 daughters from crosses between Drosophila simulans females and D. melanogaster males. Zhr maps to a heterochromatic region of the D. melanogaster X that contains 359-bp satellite repeats, suggesting either that Zhr is a rare protein-coding gene embedded within heterochromatin, or is a locus consisting of the noncoding repetitive DNA that forms heterochromatin. The latter possibility raises the question of how heterochromatic DNA can induce lethality in hybrids. Here we show that hybrid females die because of widespread mitotic defects induced by lagging chromatin at the time during early embryogenesis when heterochromatin is first established. The lagging chromatin is confined solely to the paternally inherited D. melanogaster X chromatids, and consists predominantly of DNA from the 359-bp satellite block. We further found that a rearranged X chromosome carrying a deletion of the entire 359-bp satellite block segregated normally, while a translocation of the 359-bp satellite block to the Y chromosome resulted in defective Y segregation in males, strongly suggesting that the 359-bp satellite block specifically and directly inhibits chromatid separation. In hybrids produced from wild-type parents, the 359-bp satellite block was highly stretched and abnormally enriched with Topoisomerase II throughout mitosis. The 359-bp satellite block is not present in D. simulans, suggesting that lethality is caused by the absence or divergence of factors in the D. simulans maternal cytoplasm that are required for heterochromatin formation of this species-specific satellite block. These findings demonstrate how divergence of noncoding repetitive sequences between species can directly cause reproductive isolation by altering chromosome segregation. A critical stage of speciation is the development of reproductive isolating mechanisms that prevent gene exchange between diverging populations. Hybrid sterility and lethality are major components of reproductive isolation. A key to understanding how these hybrid incompatibilities (HIs) evolve is discovering the causal genes and determining how they inhibit or perturb normal development. A number of HI genes have been identified, all of which are protein-coding. These genes are characterized by two distinct modes of evolution: either high rates of coding-sequence divergence that are consistent with adaptive evolution in many [1]–[3] but not all [4] cases, or structural changes such as in gene location [5] or gene silencing and loss following duplication [6], [7]. These cases suggest that rapid evolution of either protein-coding gene sequence or structure is a general principle underlying the evolution of HIs. Are rapidly evolving protein-coding genes the only cause of HI? Noncoding repetitive sequences, including transposable elements (TEs) and satellite repeats, are major contributors to genome evolution in higher eukaryotes. These sequences comprise heterochromatin, chromosomal regions found primarily around the centromeres and telomeres that remain more condensed than gene-containing euchromatin through the cell cycle. Pericentric heterochromatin is known to play important roles in mitotic and meiotic chromosome segregation [8]–[10]. Heterochromatin may also be important for the transcriptional regulation of flanking sequences such as ribosomal DNA (rDNA) loci, since rDNA genes are often found in heterochromatic regions [11], [12]. Paradoxically, however, despite these apparently conserved functions in higher eukaryotes, heterochromatin can vary greatly in abundance and sequence composition even between closely related species [13]–[16]. These observations have led to speculation that divergence of repetitive noncoding sequences may also directly cause reproductive isolation between nascent species [13], [17]. However, to our knowledge no examples have been clearly demonstrated. One hint that heterochromatin divergence may contribute to HI came from the discovery that the protein encoded by the Drosophila hybrid lethality gene Lhr localizes to pericentric heterochromatin. Lhr itself shows strong evidence of having diverged under the force of adaptive evolution, leading to the hypothesis that it may be co-evolving with heterochromatic sequences [18]. An additional possible link between HI and heterochromatin comes from the identification of the gene Prdm9 as causing hybrid male sterility between subspecies of mice, because the heterochromatic meiotic sex body is defective in both sterile hybrids and in Prdm9-mutant pure-species mice [19]. The sibling species D. melanogaster and D. simulans exhibit large differences in heterochromatin content [15] and strong reproductive isolation [20]. F1 hybrid females produced from D. simulans mothers and D. melanogaster fathers die as embryos [21]. This female-specific lethality is intriguing for several reasons. First, this lethality appears to have a different genetic basis than the F1 male lethality that occurs in the reciprocal cross [22]. While two major-effect genes causing this male lethality have been cloned [18], [23], nothing is known about the molecular basis of the female lethality. Second, this female-specific lethality is an exception to Haldane' s rule, the observation that unisexual hybrid sterility or lethality typically affects the heterogametic (XY or ZW) sex rather than the homogametic sex (XX or ZZ) [24]. Third, a link between hybrid female lethality and heterochromatin was strongly suggested by studies of Sawamura and colleagues of the D. melanogaster Zygotic hybrid rescue (Zhr1) mutation, which suppresses lethality of these otherwise lethal hybrid females. Zhr1 was discovered on an X-Y translocation chromosome that is deleted for much of the X chromosome pericentric heterochromatin [25]. The deleted region is thought to consist primarily of satellite DNA composed of a tandemly repeated 359-bp long monomer [25]. We refer henceforth to the monomer unit as the 359-bp repeat, and the heterochromatic region of the D. melanogaster X chromosome as the 359-bp satellite block, and revisit in the Discussion the question of what specific DNA sequences within this block cause hybrid lethality. In the wild type this satellite DNA (also known as the 1. 688 g/cm3 satellite) is estimated to form a multi-mega-bp block of heterochromatin [26]. Experiments showing that hybrid viability is sensitive to the dosage of a mini-chromosome containing part of the 359-bp satellite block led to the suggestion that repetitive sequences within the 359-bp satellite block are responsible for the hybrid lethal effect [27]. However, the mapping studies are consistent with the alternative possibility that the Zhr locus is a protein-coding gene embedded within this heterochromatic region. This is a plausible alternative, as an unexpected number of protein-coding genes have recently been found on Drosophila Y chromosomes, which otherwise contain mega-bp amounts of heterochromatic repeats [28]. If Zhr is not a protein-coding locus, then the possibility that an HI locus consists of noncoding, repetitive DNA raises important questions regarding how such sequences could kill hybrids. One possibility is that heterochromatic sequences such as those comprising the X-linked Zhr locus cause hybrid lethality by inducing in trans a global effect on chromatin structure or gene expression. Alternatively, the Zhr locus might operate in cis by affecting other adjacent, X-linked sequences such as the rDNA genes or the centromere. A third alternative is that the lethal effects are confined to this heterochromatic locus itself, such that an aberration in its structure somehow directly disrupts embryonic development. Given the difficulties involved in the genetic manipulation of heterochromatic sequences, we addressed these questions by combining genetic and cytological approaches to determine when hybrid females die during development, to identify the cellular basis of the lethality, to investigate whether possible heterochromatic defects occur genome-wide or are confined to the Zhr locus, and to test whether such defects are suppressed in hybrid females carrying the Zhr1 rescue mutation and induced in hybrid males carrying a Zhr duplication. Our results strongly suggest that the Zhr locus directly causes hybrid lethality by inducing mitotic failure in early precellularized embryos, and that the underlying defect is a failure of the 359-bp satellite block to form or maintain a proper heterochromatic state. These results provide compelling evidence that noncoding heterochromatic DNA can directly cause HI and thus contribute to speciation. To address the timing and nature of hybrid female lethality, we examined young (0–3 h) hybrid embryos produced from several different wild-type parental strains (Table 1). Normal embryonic development in Drosophila begins with a single diploid nucleus that gives rise to several thousand nuclei through 14 synchronous mitotic divisions in the large, single-celled blastula. During the first nine divisions, the nuclei migrate from the interior of the embryo to the cortex as they expand in number. Four additional nuclear divisions occur at the cortex before the formation of membrane furrows that transform the syncytial blastoderm into the cellular blastoderm. This process, termed cellularization, is followed by gastrulation (for a detailed review of early Drosophila embryogenesis see [29]). As expected, hybrid male embryos, which survive to adulthood [20], underwent normal nuclear divisions during the blastula stage and progressed into the gastrula stage (Figure 1A). Hybrid female embryos also had normal nuclear spacing and synchrony during the first nine mitotic divisions (Figure 1A). However, between mitotic divisions 10–13, hybrid female embryos exhibited large areas near the cortex devoid of nuclei and abnormal amounts of nuclei remained deep within the cytoplasm, indicating a high level of failed nuclear divisions (Figure 1A). The nuclei at the cortex were irregularly shaped and spaced (Figure 1A) and stained unevenly for the mitotic marker phospho-Histone-3 (PH3) (Figure 1B), demonstrating that these nuclei have asynchronous cell cycles. We also observed lagging chromatin between the dividing chromosome sets during anaphase and telophase in hybrid female embryos (Figure 1C and 1D). Lagging chromatin was observed in all analyzed hybrid female embryos (n = 16), ranging from 40% (13/32) to 100% (11/11) aberrant anaphase spindles per embryo, which is consistent with the high hybrid female lethality (∼87%–100%) produced from these crosses (Table 1). It is likely that the lagging chromatin is the direct cause of the mitotic asynchrony and other nuclear defects in hybrid female embryos, an idea supported by studies showing that mutations causing chromosome bridges lead to similar mitotic defects in D. melanogaster embryos [30]–[32]. To determine whether the lagging chromatin in hybrid female embryos results from a general defect in chromosome segregation or is instead chromosome-specific, we performed fluorescent in situ hybridization (FISH) with probes that recognize distinct satellite sequences in the pericentric regions of different D. melanogaster and D. simulans chromosomes (Figure 2A). Probe signals for sequences on D. melanogaster Chromosomes 2 and 3 and the D. simulans X chromosome were found in condensed regions near the spindle poles and never within the lagging chromatin (n = 75/75 spindles from 13 embryos; Figure 2B), indicating normal segregation of these chromosomes. We also analyzed the segregation of the D. melanogaster X chromosome in hybrid female embryos by using a probe for the 359-bp repeat. The 359-bp repeat probe labeled two abnormally stretched strands leading outward from the lagging chromatin toward opposite spindle poles (n = 56/100 spindles from nine embryos; Figure 2B). Stretched 359-bp repeat DNA was also observed in anaphase spindles in which there was no lagging chromatin (n = 37/100 spindles; Figure S1). Our mapping of the 359-bp repeat probe on chromosome spreads from larval brain tissue confirmed the presence of the major block of 359-bp satellite located on the D. melanogaster X, as well as several minor blocks of related satellites (353-bp, 356-bp, and 361-bp repeats) on the left arm of D. melanogaster Chromosome 3 (also see Figure S2) [33]. These smaller regions appeared unstretched and segregated normally in hybrid female embryos (Figure S3). A variant of the 359-bp repeat is also present in a small satellite block in the pericentric region of the D. simulans X chromosome (Figure S2) but does not cross-hybridize with the 359-bp repeat probe under our experimental conditions (Figures 2B and S2), presumably because of its high level of sequence divergence from the D. melanogaster repeats [34]. The lagging chromatin in hybrid female embryos, therefore, is derived solely from the D. melanogaster X chromosome. Moreover, this stretching effect likely results from partial or complete failure of the sister D. melanogaster X chromatids to separate during anaphase rather than from defective X chromatin condensation because the 359-bp satellite block appeared properly condensed during metaphase (Figure 2C). We used FISH with additional probes to determine whether separation failure of the D. melanogaster X chromatids is confined to the 359-bp satellite block or occurs in other regions of this chromosome. Probe signals from a euchromatic region located at the distal end (cytogenetic location 1C3-4) of the major left arm and from the tandemly repeated rDNA genes (bobbed+ locus) in the distal pericentric heterochromatin appeared as unstretched and condensed foci (for euchromatic region, n = 28/28 spindles from six embryos; for rDNA locus, n = 13/13 spindles from eight embryos; Figure 3A). A discrete signal of the simple-repeat satellite AATAT, which spans a portion of the minor right arm and part of the centromere immediately adjacent to the large 359-bp satellite block [35], was present at each end of the stretched 359-bp satellite block near the spindle poles in a pattern similar to the centromeric regions of the other chromosomes (Figure 3A). Therefore, the centromeres of the sister D. melanogaster X chromatids are active and separate at anaphase. However, we also observed small amounts of AATAT DNA stretched across the spindle and in the lagging chromatin, similar to the 359-bp satellite block (n = 20/34 spindles from three embryos; Figures 3A and S4). These results demonstrate that the stretched DNA is confined to the proximal X pericentric heterochromatin containing 359-bp and AATAT satellites, suggesting that sequences in this region are responsible for separation failure of the D. melanogaster X chromatids. To determine the particular causal region of the pericentric heterochromatin, we examined the segregation of the Zhr1 compound-XY chromosome (Figure 2A) in hybrid female embryos. Consistent with previous results [25], crosses between wild-type D. simulans females and D. melanogaster Zhr1 males resulted in full viability of F1 hybrid female adults (Table 1). Our analysis of larval brain chromosome spreads from the Zhr1 strain revealed that the compound-XY chromosome is completely devoid of the 359-bp satellite block but contains Y-derived AATAT repeats (Figure S2). In hybrid female embryos the Zhr1 compound-XY chromosome segregated normally, as indicated by the complete absence of lagging chromatin during anaphase (n = 68/68 spindles from six embryos; Figure 3B). Furthermore, these embryos advanced properly through subsequent developmental stages into adulthood. We also analyzed hybrid male embryos whose Y chromosome carries a translocation of approximately half of the X-linked 359-bp satellite block to the Y long arm (see Figure 2A) [36]. This Zhr+ chromosome resulted in hybrid male lethality that was less severe than hybrid female lethality induced by the wild-type X chromosome (Table 1). A subset of these hybrid male embryos (4/14) exhibited mitotic asynchrony and lagging chromatin during anaphase and telophase (n = 34/45 spindles; Figure 3C), similar to but not as common as the defects described above in hybrid females. Moreover, FISH analysis showed that the chromatin bridges were comprised of Y-derived 359-bp repeat DNA in these hybrid males (Figure 3C). These results, together with our analyses of the Zhr1 chromosome, strongly suggest that sequences contained specifically within the 359-bp satellite block induce chromosomal segregation failure in hybrid embryos. Segregation failure of the D. melanogaster X chromatids in hybrid females occurs between nuclear cycles 10–13, a period when embryonic development is primarily under control of maternally contributed RNA and proteins [37]. Our findings, therefore, suggest that the D. simulans maternal cytoplasm lacks factors that are compatible with and necessary for proper segregation of the D. melanogaster X-linked 359-bp satellite block. This hypothesis is consistent with the fact that hybrid females produced from the reciprocal cross, carrying the 359-bp satellite block and the D. melanogaster maternal cytotype, are fully viable [20]. We therefore investigated the localization patterns of D1 and Topoisomerase II (TopoII), two proteins known to associate with the 359-bp satellite block in D. melanogaster [38]–[40]. Previous studies showed that the protein D1 localizes to AT-rich heterochromatin, including the 359-bp and AATAT satellites, in larval mitotic tissues [39], [40]. Additionally, D1 was found to influence the localization of heterochromatin protein 1 (HP1) to the 359-bp satellite block [40]. On the basis of these results, it was suggested that D1 may be a structural heterochromatin component of these satellites. To determine if D1 plays a role in the defective structure of the 359-bp satellite block in hybrids, we analyzed the localization of D1 in wild-type D. melanogaster and hybrid embryos with an antibody raised against D. melanogaster D1 [39]. In Western blots, this antibody recognized a single band of approximately 60 kDa, the predicted size of D1 in both D. melanogaster and D. simulans (Figure S5). In D. melanogaster and hybrid embryos, D1 was present during anaphase at numerous sites near the spindle poles, which are likely the AT-rich satellites in the centric and pericentric regions (Figure 4A). However, in hybrid female embryos, we observed no D1 localized to the lagging chromatin containing the 359-bp DNA (Figure 4A). These observations suggested the possibility that D. simulans D1 fails to bind these sequences in hybrids. To test this hypothesis, we expressed D. melanogaster and D. simulans D1 in D. melanogaster embryos using the GAL4-UAS system (see Materials and Methods). Transgenic D1 localized to pericentric regions that completely overlapped with endogenous D1 (Figure 4B). We performed immuno-FISH experiments to simultaneously visualize D. melanogaster or D. simulans D1 with several satellite sequences. Both orthologs exhibited identical binding patterns in young embryos (Figure 4C–4F). Contrary to the prominent localization of D1 to 359-bp DNA in larval mitotic cells (also see Figure S2) [40], we observed barely detectable levels of D1 on this satellite block (Figure 4C–4F). Instead, D1 localized primarily to AATAT satellite DNA (Figure 4G). We propose that the major foci of D1 detected in embryos in earlier studies [39] and presumed to correspond with the 359-bp satellite block actually represent the large regions of AATAT on Chromosome 4. Our results demonstrate that unlike in larval brain cells, D1 is not a major component of the 359-bp satellite block during early embryogenesis, and likely does not play a role in the 359-bp structural defects observed in hybrid female embryos. We also analyzed the localization pattern of TopoII in hybrid female embryos. TopoII is the primary enzyme in Drosophila that decatenates newly replicated DNA strands and is also believed to be a structural component of condensed chromatin [41], [42]. In control D. melanogaster embryos, TopoII localized to 359-bp DNA during interphase and became more evenly distributed across the chromosomes through mitosis, with an occasional, slight enrichment on the 359-bp block at anaphase (Figures 5 and S6). However, in hybrid female embryos TopoII localized to the 359-bp satellite block during interphase but remained highly and consistently localized to this DNA through mitosis (Figures 5 and S6). We observed no TopoII foci during anaphase in hybrid male or D. simulans male or female embryos (Figure S7), in which the 359-bp satellite block is absent, further supporting the conclusion that abnormal TopoII persistence in hybrid female embryos occurs specifically on the 359-bp satellite block. This finding and the observed stretched and lagging 359-bp DNA together indicate the presence of a structural defect in this heterochromatin block that prevents chromatid separation. We have shown that hybrid females produced from D. simulans mothers and D. melanogaster fathers die during early embryogenesis because of widespread mitotic defects induced by separation failure of the 359-bp satellite block on the paternal X chromatids. Elegant genetic experiments by Sawamura and colleagues first suggested that hybrid female lethality is caused by a D. melanogaster heterochromatic locus Zhr [25], [36]. Genetic mapping localized Zhr to a pericentric region of the X chromosome containing the 359-bp satellite block. Because it is otherwise unprecedented for a heterochromatic locus to cause HI, this finding raised the key question of how Zhr kills wild-type female hybrids. We suggest that our results strongly support the conclusion that the 359-bp satellite block directly and specifically causes hybrid lethality, as opposed to alternative possibilities outlined in the Introduction, including indirect effects on other genomic regions. First, we found that hybrid female embryos exhibit large chromatin bridges during anaphase and telophase of mitotic cycles 10–13 that are almost exclusively comprised of DNA from the 359-bp satellite block on the D. melanogaster X chromosome. While these bridges also included some flanking AATAT satellite, a large amount of this satellite is present on the Zhr1 chromosome, which segregates normally, arguing against the AATAT satellite being causal for lethality. The small amount of lagging AATAT DNA detected in hybrid female embryos may result from over-catenation and tangling of AATAT DNA with the 359-bp DNA due to mis-localized TopoII (see below) when the chromatin is uncondensed, and is thus likely a secondary effect. Second, the entire multi-mega-bp satellite block appears to be stretched across the metaphase plate, suggesting that hybrids suffer from a structural defect in this block. Third, concomitant with these chromatin bridges we observed mitotic asynchrony and other aberrations that have been found in D. melanogaster mutants that have chromatin bridges [30]–[32]. In these cases, the lagging chromatin prevents complete separation of the daughter chromosome sets, thus inhibiting further mitotic divisions. Fourth, we found that all of these mitotic defects are suppressed in the Zhr1 mutant, which lacks the 359-bp satellite block, and are induced on a Y chromosome that contains a translocation of the 359-bp satellite block and causes hybrid lethality in males, albeit with incomplete penetrance. An important clue comes from our finding that TopoII localizes abnormally to the 359-bp satellite block during mitosis in hybrid female embryos. Both the DNA-decatenating and structural roles of TopoII are believed to be essential for normal chromatid separation [42]. These observations suggest several possible explanations for the hybrid phenotype. One possibility is that X chromatid separation failure results directly from incompatibility between D. simulans TopoII and the D. melanogaster 359-bp satellite block. TopoII is well conserved in the melanogaster subgroup (D. melanogaster and D. sechellia TopoII proteins are 95. 6% identical based on analysis of the full-length D. melanogaster TopoII and the ∼98% of TopoII sequence available for D. sechellia; only ∼78% of D. simulans TopoII sequence has been assembled), arguing that TopoII is not a primary incompatibility factor. Nevertheless, future transgenic experiments will be important for testing this idea. Alternatively, the persistence of TopoII may reflect a response to incomplete replication of the 359-bp satellite block as a result of incompatibilities with the D. simulans replication machinery. Extensive and unresolved tangling of daughter DNA strands would prevent separation of the D. melanogaster X chromatids at anaphase. Our observations suggest that the centromeres of the X chromatids are active and pulled toward the spindle poles, thus creating tension that results in stretching of the 359-bp satellite block. However, it is unlikely that an incompatibility with the D. simulans replication machinery is the primary cause because the first nine mitotic divisions occur normally, suggesting that replication during these divisions is normal. A third possibility is that abnormal TopoII persistence may result from improper heterochromatin formation of the 359-bp satellite block. Chromatid separation failure in hybrid females occurs during mitotic cycles 10–13 when heterochromatin initially forms. This process involves visible changes in chromatin condensation and localization of HP1 to pericentric and telomeric regions, and precedes the major transition from maternal to zygotic gene expression [43], [44]. Our data thus argue that chromatin bridges and lethality result from a failure of heterochromatin formation at the 359-bp satellite block. Defective heterochromatin formation may lead to other effects such as improper replication and tangling of daughter DNA strands, ultimately causing failure of chromatid separation. What DNA sequences are responsible for these Zhr lethal effects? Our data argue strongly against the possibility that Zhr corresponds to an unknown protein-coding gene embedded within the 359-bp satellite block. Such a hypothetical gene would have to have the highly unusual property of causing mis-segregation of the entire satellite block in which it happens to be located. Furthermore, there is unlikely to be sufficient time to transcribe such a gene to cause lethality since the mitotic defects occur during the early stages of embryogenesis when zygotic transcription is minimal [45]. Previous genetic studies by Sawamura and colleagues led them to propose that the Zhr hybrid lethal effect is caused by repetitive elements in the pericentric region of the D. melanogaster X [27], [36], [46]. By assaying a series of X pericentric deletions and duplications of different sizes they further concluded that the lethality is quantitative, and correlates with the amount of pericentric heterochromatin present. Several Zhr− stocks contained less 359-bp repeat DNA than a wild-type Zhr+ stock [46], a finding consistent with the possibility that a dosage threshold of the 359-bp repeat causes hybrid lethality. However, they excluded the 359-bp repeat (referred to as the 1. 688 g/cm3 satellite) as causing hybrid lethality because two copies of two different mini-chromosomes containing 359-bp repeats did not induce hybrid lethality [46]. The authors inferred that the double dosage of these mini-chromosomes would contain more 359-bp repeats than a single dose of another mini-chromosome that did reduce viability, thus concluding that dosage of the 359-bp repeat does not correlate with hybrid lethality. We suggest two caveats to this conclusion. First, while Southern blots suggested that differences in the abundance of 359-bp repeats are present in the mini-chromosome stocks, quantitative methods were not used to estimate the abundance of 359-bp repeats that are present specifically on the mini-chromosomes. Second, increased dosage of 359-bp repeats may induce lethality only when present as a single block on a single chromosome, and not when dispersed over multiple chromosomes. Subsequent experiments, however, showed that a different mini-chromosome can induce lethality when in two doses [27]. Although the cause of the discrepancy between the two studies remains unclear, they were later interpreted to indicate that either the 359-bp repeat or other repetitive elements are causing hybrid lethality [47]. Our experiments do not allow us to rule out the possibility that other repetitive elements present in the 359-bp satellite block and also unique to the D. melanogaster X chromosome contribute to hybrid lethality. Various TEs are known to be interspersed within the 359-bp satellite block [48]–[50], however none are specific to the X chromosome and thus cannot account for the X chromosome-specific segregation defects we observed. In contrast, several lines of evidence argue that the 359-bp repeat is the primary contributor to the Zhr hybrid lethal effect. First, the 359-bp repeat is among the most highly abundant satellite repeats in the D. melanogaster genome [15]. And while there are scattered repeats along the D. melanogaster X chromosome [51], the vast majority are found in the proximal pericentric heterochromatin where Zhr maps. Second, the 359-bp satellite is essentially species-specific, being ∼50-fold more abundant in D. melanogaster than in D. simulans and highly diverged in primary sequence of its monomers between these species [15], [34]. This species-specificity makes it an attractive candidate in evolutionary models that can account for the nonreciprocal nature of the F1 female lethality in D. melanogaster/D. simulans hybrids (see below). Third, the entire 359-bp satellite block becomes stretched during mitosis in hybrids. If another unidentified repetitive element is causing this effect, it must be distributed evenly across the entire 359-bp satellite block and not on other chromosomes. Our experiments are consistent with the idea that large amounts of the 359-bp repeat present in one block are required to induce chromosome segregation defects. First, the related 353-bp, 356-bp, and 361-bp repeats, located in much smaller amounts on D. melanogaster Chromosome 3, do not induce any observable mis-segregation in hybrids. This observation could mean that only the 359-bp monomer is capable of disrupting chromosome segregation, or, alternatively, that large amounts of this satellite class are required to cause lethality. Second, translocation of approximately half of the X-linked 359-bp satellite block to the Y chromosome resulted in lagging Y chromatin and hybrid male lethality that are proportionally less penetrant than the effects induced by the full-length 359-bp satellite block in hybrid females. The multi-mega-bp size of the 359-bp satellite block precludes definitive genetic tests using transgenic methods. We suggest, however, that the available evidence strongly supports the hypothesis that the 359-bp repeat is the sequence element within the 359-bp satellite block that is the cause of the Zhr hybrid lethal effect. The fact that hybrid females are lethal when produced from D. simulans mothers and D. melanogaster fathers but viable when produced from the reciprocal cross clearly demonstrates the involvement of a maternal effect in this incompatibility. Our results can explain this maternal effect as follows. First, we suggest that the 359-bp satellite block requires maternal factor (s) in order to be packaged as heterochromatin during normal embryonic development in D. melanogaster. Second, D. simulans does not require such factors because it does not contain the 359-bp satellite block. These factors are therefore diverged in or absent from D. simulans. Third, in F1 hybrids from D. simulans mothers, the paternally inherited D. melanogaster 359-bp block fails to be packaged properly as heterochromatin because the requisite maternal factors are missing or functionally diverged. Our proposal that the heterochromatin structure of the 359-bp satellite block is defective in hybrid females provides several promising hypotheses to explain the molecular nature of this incompatibility and the underlying maternal component (s). Satellites and other repetitive DNA elements are normally packaged into heterochromatin with general heterochromatin factors such as HP1 [52], [53], and, in some cases, with repeat-class-specific proteins like D1 [40], GAGA [54], and Prod [55]. These findings suggest a model in which high divergence in both the primary sequence and the abundance of repeat elements leads to incompatibilities with DNA-binding proteins expressed in the hetero-specific maternal cytoplasm. We tested D1 as a candidate maternal incompatibility factor because of its specific association in larval tissues with AT-rich satellite DNA, including the 359-bp repeat, but found that D1 does not localize to the 359-bp satellite block during early embryogenesis. Additional studies will be required to identify new candidate proteins that associate with the 359-bp satellite block in embryos in order to further test this model. Alternatively, hybrid female lethality may be due to a mechanism involving small RNAs. In the yeast Schizosaccharomyces pombe and other organisms, RNA interference pathways and small RNAs are required for heterochromatin formation [56]. Recent studies have identified 359-bp satellite-derived small RNAs in the maternal cytoplasm of D. melanogaster [57], [58], raising the possibility that they may be required for initial heterochromatin formation and epigenetic silencing of the 359-bp satellite block during early embryogenesis. We have proposed that hybrid female lethality occurs owing to the absence of 359-bp–derived small RNAs in the D. simulans maternal cytoplasm [59]. According to this model, hybrid females from the reciprocal cross are viable because these small RNAs are present in the D. melanogaster maternal cytoplasm. Regardless of the mechanistic basis of the maternal effect, it remains interesting that only the 359-bp satellite block is aberrant in hybrids, even though other satellite DNAs show significant differences in abundance and location between D. melanogaster and D. simulans [15]. Larger, more complex satellite repeats such as the 359-bp repeat may be more prone to cause HI than simple-repeat satellites, which are also known to vary in abundance between Drosophila species [15], because of their greater repeat sequence variation and perhaps, more complex heterochromatic structure. The incompatibility between the D. melanogaster X-linked 359-bp satellite block and the D. simulans maternal cytoplasm likely explains why the lethality from this cross violates Haldane' s rule [24]. In this cross only hybrid females are lethal because they inherit the paternal (D. melanogaster) X chromosome carrying the 359-bp block, while viable hybrid males inherit the paternal Y. Although Haldane' s rule is observed in many taxa, it is frequently violated in other Drosophila hybridizations that produce unisexual lethality, and in several of these cases, hybrids die during embryogenesis from a paternal X-linked locus [60]. We propose that paternally inherited X-linked heterochromatic repeats are strong candidates for causing hybrid female lethality in these interspecific crosses. Much of repetitive DNA evolution is likely governed by neutral evolutionary processes [61]. However, variation in satellite DNAs can also be driven by their ability to mediate genetic conflicts such as segregation distortion [62]. In the D. melanogaster segregation distorter (SD) system, sperm bearing high-copy alleles of the 240-bp Responder (Rsp) satellite are targeted for destruction while sperm with low-copy alleles are immune to this effect [63], thus selecting strongly against high-copy alleles. Variation in satellite DNA abundance may also be influenced by meiotic drive in female meiosis [64]–[66]. Female meiosis is particularly prone to meiotic drive because only one of the four meiotic products becomes the maternal pronucleus of the egg. This situation creates an opportunity for competition among chromatids to gain access to the egg, with variation in centromeres being a prime candidate for mediating such antagonism among chromosomes. A recent example of this phenomenon was found in Mimulus, in which distinct centric or pericentric repeat alleles appear to confer a substantial chromosomal transmission advantage during female meiosis in conspecific crosses and a more extreme advantage in interspecific hybrids [67]. Meiotic drive and other types of genetic conflict may therefore be important for causing rapid evolution of repetitive sequences within species and fixed differences between closely related species. Our data demonstrate that as these interspecific differences accumulate, repetitive sequences can inhibit chromosome segregation in hybrids and thus directly cause reproductive isolation. Strains used were: wild-type D. simulans C167. 4 and NC48S [68], and white501 (SA32) (made by introgressing the white501 allele into an isofemale South African D. simulans strain that mates well with D. melanogaster, provided by C. Aquadro), and wild-type D. melanogaster Oregon R and Canton S. The hybrid rescuing Zhr1 chromosome (full genotype is XYS. YL. Df (1) Zhr) is described in [25] and the Zhr+ Y chromosome (full genotype of strain is Ts (1Lt; Ylt) Zhr/Dp (1; Y) y+) is described in [36]; the structures of both chromosomes are shown in Figure 2A. Crosses were conducted by combining 40–50 0–8-h-old virgin D. simulans females and 60–80 12–24-h-old D. melanogaster males. Flies were allowed to mate for 48 h in a 25°C incubator with a 12-h light/12-h dark cycle prior to embryo collection. Embryos were collected on grape juice agar plates [69] over a 3-h period and dechorionated in 50% bleach. Immediately afterward, embryos were fixed for 10 min in 4% EM-grade paraformaldehyde (Electron Microscopy Sciences) and heptane (Sigma-Aldrich) and then devitellinized in 100% methanol (Sigma-Aldrich). Fixed embryos were hydrated by using a series of methanol∶1×PBTA buffer solutions (9∶1,5∶5,1∶9 by volume) and treated with RNaseA (Sigma-Aldrich) at 37°C for 2 h before carrying out FISH or immuno-staining. The following sequences were used for FISH probes: (TTT-TCC-AAA-TTT-CGG-TCA-TCA-AAT-AAT-CAT) recognizing the 359-bp satellite block on the D. melanogaster X as well as minor variants on D. melanogaster Chromosome 3 [34]; (AAT-AC) 6 recognizing a small block of this sequence on the D. melanogaster Y [70]; (AAT-AT) 6 recognizing large amounts of this sequence on D. melanogaster and D. simulans Chromosome 4 as well as a small region on the D. melanogaster X [15], [70]; (AAG-AG) 6 recognizing primarily a large block of this sequence on the D. melanogaster 2 and a small block on the D. simulans X [15]; (AAT-AAC-ATA-G) 3 recognizing a single block of this sequence on D. melanogaster Chromosomes 2 and 3 [70]; and (AAT-AAA-C) 4 recognizing a single region on the D. simulans Y (S. Maheshwari, personal communication). These sequences were chemically synthesized (MWG Biotech) and modified at the 5′ terminus with either fluorescein, Cy3, or Cy5 for fluorescent detection. The euchromatic FISH probe was made by random priming of BAC DNA (BACR19J01 from CHORI BACPAC Resources) using the BioPrime DNA Labeling System (Invitrogen), which incorporates amino-allyl-dUTP into amplified DNA products (protocol from A. Minoda). These products were sonicated into 100–150-bp fragments, cleaned with MiniElute columns (Qiagen), and conjugated with an Alexa546 fluorophore reactant group (Invitrogen). The labeled probe was ethanol-precipitated twice before hybridization. The D. simulans 360-bp probe was made by end-labeling a 360-bp PCR product with poly-amino-allyl-dUTP and Terminal Transferase (Roche). This product was amplified from D. simulans C167. 4 genomic DNA using the following primers that recognize the D. simulans 360-family monomer repeat [34]: forward ACT-CCT-TCT-TGC-TCT-CTG-ACC-A and reverse CAT-TTT-GTA-CTC-CTT-ACA-ACC-AAT-ACT-A. The rDNA probe was made by using a 1,200-bp fragment of five tandem repeats of the rDNA IGS region cloned into pBluescript (construct was a gift from G. Bosco). This fragment was digested out of the plasmid using the restriction enzymes KpnI and SacI, digested into ∼150-bp fragments with AluI and MseI, and purified using the Qiaex II gel extraction kit (Qiagen). End-labeling was conducted as described above. FISH was performed as described [70] with minor modifications. Overnight incubation of fixed tissues with probe was conducted in a thermocycler with a denaturation temperature of 92°C for 3 min and a hybridization temperature of 32°C overnight. This lower hybridization temperature (standard is 37°C) was important for detecting the 359-bp variant sequences on D. melanogaster Chromosome III and did not result in excessive nonspecific hybridization. Three additional 10-min washes in 50% formamide/2×SSCT were performed to maximize the removal of any nonspecifically bound probe. For immuno-FISH experiments, fixed embryos were hybridized with primary and secondary antibodies as described [69]. Embryos were subsequently fixed in 4% paraformaldehyde for 30 min, and washed 3× with PBTX and 3× with 2×SSCT. FISH was then conducted as described above. For immuno-cytological or immuno-FISH experiments, Rat anti-HA (Roche 3F10) and rabbit anti-D1 antibodies (a gift from E. Käs) [39] were used at 1∶100 and 1∶1,000, respectively. Rabbit anti-TopoII (a gift from T. Hsieh) [71] and mouse anti-PH3 (Santa Cruz Biotechnology) were used at 1∶1,000. Alexa555 anti-rabbit, Alexa555 anti-rat, and Alexa633 anti-mouse secondary antibodies were used at 1∶300 (Molecular Probes). Tissue preparation, FISH, and D1 immuno-staining of larval brain chromosomes were performed as described [39]. DNA was stained by using either OliGreen or TO-PRO-3 iodide (Molecular Probes) for embryos and Vectashield containing DAPI (Vector Laboratories) for brain tissues. All imaging was conducted at the Cornell Microscopy and Imaging Facility, using either a Leica DM IRB confocal microscope or an Olympus BX50 epifluorescent microscope. Confocal images were generated by using sequential collection of each wavelength to eliminate bleed-through of fluorophores and generated as maximum projections of multiple scans. Images were processed using Photoshop (Adobe, version 7. 0). Contrast and brightness changes, when used, were applied globally across the image. Images shown in the figures were taken from either hybrid or pure species embryos produced from C. 167. 4 and/or Canton S strains, unless otherwise specified. However, cytological analyses were also performed on hybrid embryos produced from other parental lines shown in Table 1 for verification of the observed phenotypes. 0–3-h embryos from C167. 4 and Canton S flies were collected as described above and washed in 1× PBS buffer. 50 µl embryos from each strain were lysed in an equal volume of 2× SDS Sample Buffer [72] and boiled for 5 min. Five µl of protein extracts were separated on a 10% polyacrylamide gel for 1 h at room temperature and 100 V. Proteins were transferred to a nitrocellulose membrane overnight at 4°C and 20 V, blocked with 5% powdered milk, and then blotted overnight at 4°C with anti-D1 serum (1/10,000 dilution). Membranes were then blotted with goat anti-rabbit HRP antibodies (1/5,000 dilution; Jackson) for 1 h at room temperature. HRP was detected using ECL Western blotting substrate (Pierce). The complete D. melanogaster and D. simulans D1 coding sequences were PCR amplified from Canton S and C167. 4 adult cDNA, respectively, by using the following primers: for D. melanogaster, forward CAC-CAT-GGA-GGA-AGT-TGC-GGT-AAA-GAA-G and reverse TTA-GGC-AGC-TAC-CGA-TTC-GG; for D. simulans, forward CAC-CAT-GGA-AGA-AGT-TGC-GGT-AAA-GAA-G and reverse TTA-GGC-AGC-TAC-CGA-TTC-GG. The resulting fragments were cloned into the pENTR/D-TOPO vector (Invitrogen). Positive clones were fully sequenced to confirm the absence of any errors. Each sequence was recombined into the pPHW plasmid downstream of the UAS transcriptional activation sequence and in frame with an N-terminal 3× HA peptide (Murphy collection; described at http: //www. ciwemb. edu/labs/murphy/Gateway%20vectors. html). Additionally, the attB sequence was subcloned into this plasmid for site-specific integration into the D. melanogaster strain y1 w67c23; P{CaryP}attP2 [73]. The resulting transformants were crossed with the strain w; P{matα4-GAL-VP16}V37 (Bloomington Stock Center) for expression of the HA-tagged D1 in the early embryo.

Speciation is most commonly understood to occur when two species can no longer reproduce with each other, and sterility and lethality of hybrids formed between different species are widely observed causes of such reproductive isolation. Several protein-coding genes have been previously discovered to cause hybrid sterility and lethality. We show here that first generation hybrid females in Drosophila die during early embryogenesis because of a failure in mitosis. However, we have discovered that this is not a general failure in mitosis, because only the paternally inherited X chromosome fails to segregate properly. Our analyses further demonstrate that this mitotic failure is caused by a large heterochromatic region of DNA (millions of base pairs) that contains many repetitive copies of short noncoding sequences that are normally transcriptionally quiescent. Interestingly, this block of heterochromatin is only found in the paternal species. We suggest that a failure of the maternal species to package this paternally inherited DNA region into heterochromatin leads to mitotic failure and hybrid lethality. If this is a general phenomenon it may explain other examples of hybrid lethality in which F1 females die but F1 males survive.

Abstract Introduction Results Discussion Materials and Methods

evolutionary biology/developmental molecular mechanisms developmental biology/developmental evolution evolutionary biology/nuclear structure and function

2009

Species-Specific Heterochromatin Prevents Mitotic Chromosome Segregation to Cause Hybrid Lethality in Drosophila

Adaptation in spatially extended populations entails the propagation of evolutionary novelties across habitat ranges. Driven by natural selection, beneficial mutations sweep through the population in a “wave of advance”. The standard model for these traveling waves, due to R. Fisher and A. Kolmogorov, plays an important role in many scientific areas besides evolution, such as ecology, epidemiology, chemical kinetics, and recently even in particle physics. Here, we extend the Fisher–Kolmogorov model to account for mutations that confer an increase in the density of the population, for instance as a result of an improved metabolic efficiency. We show that these mutations invade by the action of random genetic drift, even if the mutations are slightly deleterious. The ensuing class of noise-driven waves are characterized by a wave speed that decreases with increasing population sizes, contrary to conventional Fisher–Kolmogorov waves. When a trade-off exists between density and growth rate, an evolutionary optimal population density can be predicted. Our simulations and analytical results show that genetic drift in conjunction with spatial structure promotes the economical use of limited resources. The simplicity of our model, which lacks any complex interactions between individuals, suggests that noise-induced pattern formation may arise in many complex biological systems including evolution. Our computer model, illustrated in figure 1, 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 (1) (2) 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 [4]. 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. 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 [23]. For instance, yeasts can switch their metabolism from respiration to fermentation plus respiration [24], [25]. 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 [22]. 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 [21], [22], [24]. This case of a trade-off [21] 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. 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 (left). The importance of sampling noise can be verified in purely deterministic simulations that neglect genetic drift, see Fig. 1c (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, 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 suggest that the wave speed in both regimes can be summarized as (3) 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 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 (4) 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). This is made possible by a nonlinear variable transformation due to E. Hopf and J. D. Cole [26], [27], 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 [19], [28], [29]. 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 [21], [22], [24]. To account for this trade-off between growth rate and yield [21], we have studied our model for a selective disadvantage of the mutants. We find both in simulations (Fig. 3) and theory (Methods) 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 (5) which is the carrying capacity for which mutations with non-zero are unable to invade. In the frame work of evolutionary game theory [30], the condition in Eq. (5) is called an evolutionary stable strategy towards which populations are expected to evolve on long evolutionary time scales. 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 [31]. 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 [22], [24]. 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 [22]. The emergence of this basic form of cooperation in spatially extended habitats has been observed in individual-based simulations [21], [22], [32], 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 [33], [34]. 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” [35], which is a special case of group selection [36], 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 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). 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 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 [36]. Barton and Clark in Ref. [36] 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 [36].) 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 [37], 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 [14], or vary among species, which can lead to Turing patterns [38]. Thus, pattern formation by genetic drift may be an important mechanism in many complex systems including biological evolution. 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. 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 [33], [34] 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, (6) where the local growth rate depends on the ratio between total population density and local carrying capacity, and reads (7) 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 (8) 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), (9) 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 (10) We seek a solution of equation (6) and equation (10) for a step function initial condition, . In the Supporting Text S1, we show that the density closely follows the carrying capacity, provided that the migration rate is small, (11) 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, (12) 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 [26], [27]. Thereto, we introduce the new dynamical field (13) which represents the fraction of mutants to leading order in, . In terms of this new field, equation (12) transforms into a simple diffusion equation (14) 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 (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 [4], cause a noise term in the equations (10,12), which reads [39], [40] (15) where is the variance in offspring number and the stochastic forcing term has white noise correlations, (16) 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 (17) 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 [41]. 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, (18) favoring the growth of the mutants. The noise term takes the form (19) 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 (20) (21) With these substitutions, the stochastic equation of motion takes the form (22) 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 [7], [8]. 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 now follows from known results [7], [8], [19], [28], [29] 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. As mentioned earlier, there are general reasons to posit a trade-off between growth rates and densities, at least in heterotrophic organisms [21]. 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 (23) 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.

Mutations that increase an organism' s fitness are the fuel for biological evolution. When such beneficial mutations enter a spatially extended population, they spread through the population in a “wave of advance”, first described by R. Fisher and A. Kolmogorov. The force driving these traveling waves is Darwinian selection, which favors individuals with higher fitness. Here, we describe a new type of traveling mutant wave that is driven by non-selective forces instead-- namely by random genetic drift, which refers to the randomness in the reproduction process. These noise-driven waves promote the economical use of a limited resource because they occur whenever a mutation increases the growth yield, which refers to the biomass produced per unit of resource. Since a change in growth yield and growth rate often occur together and with opposite signs, we argue that both types of mechanisms will jointly decide over the fate of a novel mutation. We predict that the population evolves towards an evolutionary optimal carrying capacity, at which selective and non-selective forces just balance.

Abstract Introduction Results Discussion Methods

physics statistical mechanics theoretical biology evolutionary biology population modeling evolutionary modeling population genetics biology computational biology population biology genetics and genomics

2011

Noise Driven Evolutionary Waves

In gregarious species, social interactions maintain group cohesion and the associated adaptive values of group living. The understanding of mechanisms leading to group cohesion is essential for understanding the collective dynamics of groups and the spatio-temporal distribution of organisms in environment. In this view, social aggregation in terrestrial isopods represents an interesting model due to its recurrence both in the field and in the laboratory. In this study, and under a perturbation context, we experimentally tested the stability of groups of woodlice according to group size and time spent in group. Our results indicate that the response to the disturbance of groups decreases with increases in these two variables. Models neglecting social effects cannot reproduce experimental data, attesting that cohesion of aggregation in terrestrial isopods is partly governed by a social effect. In particular, models involving calmed and excited individuals and a social transition between these two behavioural states more accurately reproduced our experimental data. Therefore, we concluded that group cohesion (and collective response to stimulus) in terrestrial isopods is governed by a transitory resting state under the influence of density of conspecifics and time spent in group. Lastly, we discuss the nature of direct or indirect interactions possibly implicated. Aggregation is one of the most common social phenomena. Although it may be a response to the heterogeneities of the environment of individuals sharing the same needs, aggregation is frequently the joint response of individuals to the presence of conspecifics and the environment [1], [2]. The ultimate and proximate causes leading to animal aggregations are topics of ongoing research and are broadly known for many taxa. Response to predation risk is probably an important factor for understand groupings [3], [4] as well as feeding efficiency (e. g. , [5]) or desiccation limitation (e. g. , [6] for woodlice). From a mechanistic viewpoint, the use of aggregation pheromones is demonstrated in many invertebrate species [7], but many other vectors such as thigmotaxis [8] are reported. The stability of formed groups can considerably vary both in time and space according to environmental pressure and collective or individual decision making [9], [10]. Several stimuli such as predation attacks can disturb the cohesion of aggregates and lead to more or less coordinated group effects. In this context, the interactions between individuals and particularly between naive and informed (those who perceived the risk or perturbation) individuals faced with perturbation may modulate collective reaction through an information cascade within the group. In cockroaches, the individual probability of fleeing a light perturbation decreases with the number of close and immobile individuals [11]. In many species such as gregarious harvestmen [12] or whirligig beetles [13], larger groups react faster than smaller groups. In these examples, the density-dependent dispersion of groups results from signal amplification (pheromone, collision) during signal propagation among individuals. On the other hand, in many invertebrates, individuals chronically rest in aggregation sites and are characterised by relative phases of behavioural immobility (e. g. , ants [14]; beetles [15]; solitary bees [16]; butterflies and caterpillars [17], [18]; Opiliones [19]). Individual resting time may be regulated by a basic circadian rhythm [20] but also by other biotic factors as prior individual activity [21] or presence and behaviour of conspecifics [20]. In Drosophila, social context increases the total amount of sleep in individuals and its effectiveness [22]. In addition, despite a strong difference in their social organisation, the individual probability of leaving the resting phase decreases with increasing numbers of individuals in the group both in the cockroach Blattella germanica [23] and in the ant Lasius niger [24]. Thus, the departure from a group can have several origins, involving external stimuli to the group or a spontaneous departure, involving social factors or not, and may act in an antagonistic manner. Group size is certainly one of the most important variables in understanding the collective mechanisms of group maintenance and how social facilitation may act [25–30]. However, essential for understanding the collective dynamics of groups and spatio-temporal distribution of organisms in environment, the social influence on group stability during the resting phase, and thereby on the ability to change state when faced with a stimulus, is poorly known and seldom discussed in the literature. Gregarious woodlice (Crustacea: Isopoda: Oniscidea) present long aggregative phases (more often in daytime) to limit desiccation risk and short dispersal phases (often nocturnal) for solitary forage [31–35]. Aggregation in these organisms is a particularly recurrent, fast and stable process [36–40]. Aggregate initiation and formed patterns result from a dynamic trade-off between inter-attraction between conspecifics and environmental heterogeneities such as shelter [39]. In other words, aggregation in woodlice is a social phenomenon [39], [41], making it a good model for the study of gregariousness in non-eusocial arthropods (see [17]) and especially non-insect arthropods. The collective response to a perturbation of a group composed of individuals may be a complex phenomenon to study and model. The aim of this study is a better understanding of the dynamics and mechanisms—especially social interactions—governing the group stability in gregarious species. Accordingly, we investigated group cohesion in the anthropophilic woodlouse Porcellio scaber, one of the most common terrestrial isopods [42]. In this context, we hypothesize that group cohesion and group dispersion may be governed by a behavioural contagion between group members. The collective response of woodlice is tested under two simple parameters: the number of individuals engaged in aggregate and the time spent in the aggregate before the perturbation. Our experimental and theoretical results show that some individuals respond faster to the perturbation and flee more quickly than others. Individuals do not interact while fleeing, but only during the aggregation period. During this period, the individuals adopt one of two behavioural states: calm or excited. Interestingly, it appears that the proportion of calm and excited individuals depends on social context and the time spent in the aggregate before the perturbation. Individuals of the common rough woodlouse Porcellio scaber Latreille were trapped in the gardens of the Catholic University of Lille (Northern France) and placed in culture. The culture consisted of glass boxes (410×240×225 mm) containing a humid plaster layer and soil litter (humidity≈ 80%) and were placed at 21°C under a natural photoperiod of the region. The experimental set-up consisted of a circular arena with a diameter of 193 mm containing at its centre a small and removable arena with a diameter of 65 mm (see S1 Fig). A sheet of white paper covered the bottom of the set-up. This sheet was changed between each experiment to eliminate potential traces of chemical deposit (see [37]). The set-up was lighted by a 40 W lamp placed 80 cm above the set-up, equivalent to a brightness of 156 lux. The small central arena (Ø 65 mm) represented a retention area in which woodlice were first introduced and kept enclosed (see the experimental conditions below), before release into the large arena (Ø 193 mm). The removal of the inner retention arena has been performed by hand in a quick movement that is perpendicular to the support of the set-up (S1 Fig). Each experiment was filmed with a Sony camera CCD FireWire—DMK 31BF03 from the release of individuals until the last individual left the retention area. Three experimental conditions were performed: Each experiment can be divided in three steps: retention, perturbation and departure. Figures and regression analyses were obtained using GraphPad Prism 5. 01 software (GraphPad Software Inc.). Statistical tests were performed using GraphPad InStat 3. 06 (GraphPad Software Inc.). First, circular statistics show that in almost 75% of experiments, individuals did not present particular orientation during the dispersion phase (S1 Table; Rayleigh test, p> 0. 05). In the other cases where a particular trajectory is observed, the orientation (mean angle) is identical to the circular distribution of individuals in the retention arena at t = 0s (S1 Table; Watson-Williams test, p< 0. 05). No experiment with a random initial distribution of individuals has given an orientated pattern during dispersion. There is no inter-experimental bias in the mean angle of dispersion (Rayleigh test, R = 0. 1941, p = 0. 29067). Whatever the experimental conditions, the dispersal dynamics from the retention area are qualitatively similar: a rapid fall in the number of individuals in the first seconds of the experiments followed by more gradual departures (Figs 1A and 2A). However, the dynamics of dispersion are quantitatively affected by the number of individuals initially introduced (Fig 1A). The higher the number of individuals initially introduced, the slower the dynamics of dispersion (Fig 1A and 1B). The average dispersion time of the isolated individuals kept enclosed during 300 s is low (Table 1) and similar to the condition with n = 10 individuals (Fig 1B). There is a significant difference between the half-life time of an aggregate according to the number of conspecifics introduced (Fig 1B; Kruskal-Wallis test, KW = 57. 526, p< 0. 0001). The half-life time of an aggregate of 10 individuals is significantly lower than those of an aggregate of 80 woodlice (Dunn test, p<0. 001) and 120 woodlice (Dunn test, p<0. 001). Additionally, the half-life time of an aggregate of 40 woodlice is significantly lower than with 80 woodlice (Dunn test, p<0. 05) or 120 woodlice (Dunn test, p<0. 001). See similar results for the ¼ life time and the ¾ life time in the supplementary material (S2A Table). In addition, the variance of the half-life time of aggregates increases with group size (Fig 1B; ANOVA test, F = 10. 46, p = 0. 0001). The group of 120 woodlice dispersed with a greater variability than the groups of 10 (Bonferroni test, t = 5. 510, p<0. 001), 40 (Bonferroni test, t = 5. 069, p<0. 001) or 80 woodlice (Bonferroni test, t = 3. 917, p<0. 01). Other conditions did not differ between treatments (Bonferroni test, p>0. 05). There is a significant correlation between the coefficient of variation and the group size (Spearman test, r = 1, p = 0. 0058). The coefficient of variation increases as a function of the group size according to a power law 0. 49N0. 15 (R² = 0. 93). In experiments with isolated individuals, increasing retention time significantly affects the departure time of individuals (Table 1; Kruskal-Wallis test, KW = 8. 555, p = 0. 0139). The woodlice held for 30 seconds left the retention area faster than the woodlice held for 5 min (Table 1; Dunn test, p<0. 05) and for 10 min (Table 1; Dunn test, p<0. 05). Note that one individual held for 600 s remained in the retention area for 2417 s. This outlier was excluded from the analysis (Grubbs test, Z = 6. 05, p<0. 05). The data distribution is given in the supplementary material (S2 Fig). In addition, in group experiments, the dynamics of group dispersion are also affected by initial retention time (Fig 2A). The longer the initial retention time of individuals, the slower the dispersion after release (Fig 2A and 2B). There is a significant difference between the half-life time of an aggregate according to the retention time of individuals (Fig 2B; Kruskal-Wallis test, KW = 40. 33, p< 0. 0001). The half-life time of an aggregate held for 30 s was significantly lower than those of aggregates held for 300 s (Dunn test, p<0. 01) or 600 s (Dunn test, p<0. 001). Additionally, the half-life time of aggregates held for 60 s or 120 s is significantly lower than in the case of aggregates held for 600 s (Dunn test, p<0. 001). Other conditions did not differ from each other (Dunn test, p>0. 05). See similar results for the ¼ life time and the ¾ life time in the supplementary material (S2B Table). In addition, the variance of the half-life time of aggregates increases with the retention time of individuals (Fig 2B; ANOVA test, F = 10. 37, p< 0. 0001). Groups held for 600 s disperse with a greater variability than groups held for 30 s (Bonferroni test, t = 5. 495, p = 0. 001), 60 s (Bonferroni test, t = 5. 260, p = 0. 001), 120 s (Bonferroni test, t = 4. 940, p = 0. 001) and 300 s (Bonferroni test, t = 3. 650, p = 0. 01). Other conditions did not differ from each other (Bonferroni test, p>0. 05). There is a significant correlation between the coefficient of variation and the retention time (Spearman test, r = 0. 9, p = 0. 0417). The coefficient of variation increases as a function of the retention time according to 0. 21t0. 25 (R² = 0. 87). To explain our results of dispersion, several assumptions can be made on whether a social effect was or not involved in the conformation of the groups during the retention time, the releasing of individuals (i. e. , perturbation) and during the departure of the individuals. Hypotheses involving social effects during departure, different behavioural responses to the perturbation or a probability to become tired dependent on population size and retention time are not consistent with our experimental results. For this reason, they are detailed respectively in S1 Text, S2 Text and S3 Text. In brief, the retention effect of individuals during departure (i. e. , an increasing departure rate) is not supported by the decreasing departure rate calculated from our experimental results (see S1 Text). In addition, the perturbation hypothesis predicting that the dynamic of dispersion is the by-product of a disturbance event and spatial conformation of individuals in the retention arena was not supported by the data of spatio-temporal distribution of individuals in the set-up (see S2 Text). Finally, the fatigue hypothesis of a fraction of individuals in the retention area, increasing with the retention time and the number of individuals, cannot reproduce the experimental results quantitatively or qualitatively (see S3 Text). First, we assume that individuals, before and during the perturbation, are in a slow or fast behavioural state with varying proportions depending on the experimental conditions. We add all the experiences assuming that the probability of departure for the slow and fast individuals are independent of the situation and that only the proportions changes with the experimental conditions. The average survival curve of all experiments is well fitted by the two phase exponential (Eq (4), Mat and Meth), where Fs represents the fraction of slow individuals, Ff represents the fraction of fast individuals (i. e. , Ff = 1-Fs), ks represents the inverse of mean time of departure of slow individuals, kf represents the inverse of mean time of departure of fast individuals and t represents the time (s). Fs = 0. 3481 (95% CI: 0. 3457 to 0. 3505); Ff = 0. 6519; ks = 0. 003404 (95% CI: 0. 003376 to 0. 003431); kf = 0. 0687 (95% CI: 0. 06772 to 0. 06968); df = 4381; R² = 0. 9903. With the constants ks and kf obtained previously, we calculated the mean fraction of slow individuals Fs and the mean fraction of fast individuals Ff using Eq (4) for each conditions of Figs 1A and 2A. The variation of Fs (and by extension Ff) according to the number of initially introduced (P0) and the retention time of groups (R) are given in Fig 3A and 3B, respectively. The fraction of slow individuals (Fs) increases with group size (P0) according to a power law Fs=5. 2210−4P01. 5 (R² = 0. 9930). The values given here are valid for P0< 120. Additionally, the fraction of slow individuals (Fs) increases with the retention time of groups (R) according to a similar power law Fs = 3. 12 10−5 R1. 4 (R² = 0. 9964). The values given here are valid for t< 600 s. The experimental distribution of Fs (see S5 Fig) was obtained by fitting each experiment of each condition one by one with Eq (4) and the constants ks and kf obtained previously. During the retention phase, woodlice reduced interindividual distances in both experiments increasing number of individuals introduced (S2 Text) and retention time (S3 Fig). Based on previous studies showing the role of the interactions between conspecifics in cluster formation and mimetic behaviour [43], [44], we hypothesise that increasing spatial tightening between individuals promotes social interactions and therefore behavioural changes. To test this hypothesis, we developed and analysed a 2-state behavioural contagion model. We assume that during the retention phase, individuals can be in two states: excited and calm; these two states lead/correspond to the slow and fast states observed during the departure phase. The model describes the dynamic process – during the retention period – in terms of individuals adopting or leaving the slow or fast state. A schematic illustration of the model is given in S4C Fig. Individuals in the fast (slow) state can spontaneously shift in the slow (fast) state. Moreover, the individuals affect each other: the interactions between a slow and fast animal enhance the probability for the fast one to become slow or the slow one to become fast. We assume a linear relationship between the probability of transition between the state fast –> slow (slow –> fast) and the number of slow (fast) individuals: Φ (fast→slow) =α+βSΦ (slow→fast) =μ+ωF where α (μ) is the spontaneous probability that a fast (slow) individual adopts the slow (fast) state. The coefficient β (ω) is the coefficient of imitation: the greater the number of slow individuals (S), the greater the probability of transition between the fast and slow states; the greater the number of fast individuals (F), the greater the probability of transition between the slow and fast states. Without social interaction β = ω = 0, the fraction of individuals in a slow (or fast) state is independent of the total population. The differential equation describing the time evolution of the mean number of S is dSdt=αF−μS+βSF−ωSF=αF−μS+ (β−ω) SF (6,1) In this study we neglected the parameter ω, and as F+S = N (total number of individuals): dSdt= (α+βS) (N−S) −μS (6,2) In this paper, we worked with the stochastic version of the model (master equation). The master equation describes the time evolution of the probability Θ (S) of the system to occupy each of the discrete sets of states (S = 0,1, …. , N). At the beginning of the retention, all individuals are fast S = 0, F = N: dΘ (S) dt= (α+β (S−1) ) (F+1) Θ (S−1) − (α+βS) FΘ (S) +μ (S+1) Θ (S+1) −μSΘ (S) (7,1) The mean values of S and of the fraction of slow individuals are <S>=∑S=0NSΘ (S); FS=<S>N=∑S=0NSΘ (S) N (7,2) We assumed that α, β and μ are constant and independent of the conditions (total number of individuals or retention time). We performed a numerical resolution of this master equation, and at the end of the retention period, the model predicts the probability Θ (S) of having a population of S individuals in the slow state, and a mean value of <S> or the mean theoretical fraction of slow individual (Fs). For each condition, we searched for parameter values of α, β and μ for which the theoretical mean values of Fs is the closest to experimental mean values of Fs. The results give α = 10–4 s-1; β = 2,5. 10–4 s-1; and μ = 9. 10–4 s-1. For all conditions, the mean fraction of slow individuals Fs observed is particularly close to the Fs obtained with the model that is described here (Fig 4). Moreover, for each condition, we used the Kolmogorov-Smirnov test to determine whether the distribution of experiments as a function of the slow individuals is different from the theoretical distribution Θ (S) for the values of the parameters (α, β, μ) giving the best fit to the mean fraction of slow individuals. The distributions are not different (S5 Fig; S3 Table, p> 0. 05), except for N = 80 and t = 600 s. In the case of N = 80, the model overestimates the proportion of experiments with a small number of slow individuals, and in the case of t = 600 s, the model underestimates the proportion of slow individuals compared to the experimental results (S5 Fig and S3 Table). Additionally, the distribution of experiments with t = 120 s (although consistent with theoretical data; S5 Fig) does not pass the statistical test (S3 Table), but this result is only due to only one experience exhibiting a particularly long plateau phase of approximately 20 individuals (3/4 life time of 317 s when the other experiments in this condition oscillate between 3 s and 47 s). The theoretical distribution Θ (S) (7,1) is also used to calculate the dynamics of dispersion (mean, variance) and mean half-life time of the dispersion (Figs 5 and 6). Eq (5,1), where Ψ (S, F) gives the probability that S and F individuals (and therefore the total population) are still in the area at time t after the release of individuals, is solved with the combinations S = i, F = N- i; i = 1, …, N at the end of the retention phase. The dispersion dynamics for each initial condition weighted by Θ (S) gives the theoretical distribution of the experiments according to the total population that is still in the area at time t. The theoretical results (mean half-life time, variance and dynamics of dispersion) are close to the experimental results regardless of the conditions (Figs 5 and 6; see also S5 Fig and S3 Table). Fundamental rules leading to the initiation of aggregates or collective choices for resources (food sources, shelters) are now well known in many group-living invertebrates [1], [45]. In contrast, the collective mechanisms governing the stability and dispersion of group-living gregarious arthropods are poorly known. These mechanisms are nevertheless important for understanding how social influence may modulate the individual decision to stay in a place or react when faced with perturbation. In this study, we examine the collective response of groups of terrestrial isopods (Crustacea) faced with an environmental disturbance according to the number of individuals engaged in aggregate and the time spent by individuals in aggregates. It is well known that woodlice present a strong thigmotaxis [36], [46], [47]. Contact stimuli might come from abiotic (e. g. , wall of the arena) or biotic elements (e. g. , conspecific). In the field, search contacts, either in micro-shelters and/or in aggregates, should be adaptive in decreasing desiccation risks [6], [48], [49]. Similarly, woodlice are strongly photonegative [36], [47], [50]. In the field, this behaviour should be adaptive to limit desiccation risks in avoiding sunlight. Here, the overall populations in the arena were exposed to a stressful condition due to light, and a part of population was exposed to a mechanical perturbation during release due to the loss of contact with the edge of the arena. Therefore, a relatively quick fleeing behaviour is not surprising. However, the dynamics of group dispersion slows with increasing initial retention time and the number of individuals initially introduced in the set-up. This slowdown during dispersion cannot be explained by a positive or negative feedback at work between individuals during this dispersion. They leave independently of each other. Contrariwise, we show the existence of two behavioural states of individuals or two behavioural populations: calm (slow individuals during departure) and excited (fast individuals during departure). These fractions of slow or fast individuals depend on the size of the group and the duration of retention. This dichotomy between slow and fast individuals, though it may seem excessive (it is probable that there is a greater range of excitability), is a useful simplification for a minimalist fitting. The group effect is based on the modulation of the fraction of slow and fast individuals. At the individual level, an increase in retention time slowed the departure rate of isolated individuals probably because the stress level decreases with time spent in the retention area and increases the probability that an excited individual spontaneously shifts into a calm state. A similar phenomenon could be involved at the group level but appears insufficient to explain the group size effect. Increased group size and time spent in the group increases the potential amount of interactions between individuals [51–53] and increases the probability that individuals initiate aggregation and shift to the calm state. We assume that aggregated individuals are in the calm state. The theoretical model based on these social interactions assumes that individuals in the excited state can spontaneously shift to the calm state (and conversely) during the retention phase and that the probability of switching activity depended on the number of conspecifics and their behavioural states. This model well reproduced the experimental results. In particular, the results show that the fraction of slow individuals (Fs) increases with the number of initially introduced individuals and the retention time of groups. This increase in the fraction of slow individuals leads to maintaining group stability and the resting state. Moreover, the model is able to reproduce the high intra-condition variability between experiments. This high variability strongly suggests that positive feedbacks are at work between calm and excited individuals [1], [23]. Note that the existence of a social effect during the departure phase (i. e. , retention mechanisms) or non-social mechanisms (i. e. mechanical disturbance effect, increasing individual fatigue during retention, etc.) are not exclusive to the scenario proposed here during the retention phase; however, such complementary mechanisms should be of minor importance in contrast with the social transition between excited and calm individuals during the retention phase. In this way, this study proposes a new understanding of aggregation in arthropods in involving the amount of time spent in the group, an often underestimated parameter in collective decision-making in view of the highly emphasised number of individuals. For example, in cockroaches, the probability of leaving a resting site decreased with the individual time spent in the site (isolated individual experiments [54]) or with the presence of conspecifics in the site (collective experiments [23], [43]), Thus, our study argues in favour of a complementarity of these two mechanisms: the individual decision to stay in the area relies on the interplay between individual resting time and the amount of interactions with settled conspecifics and their behavioural states. Our experimental and theoretical results strongly suggest that the aggregation process during the retention phase may promote a collective entry into a behavioural quiescence or sleep-like state (see [20], [55–57]). Entrance into a sleep-like state could explain why woodlice can be observed for several dozen minutes in apparently unfavourable places (exposed at light and without cover) when isolated individuals quickly run away in similar conditions. Our results differ from the classical vision under which larger groups react faster to perturbation than smaller groups [3] due to social facilitation and amplification of the alarm signal (e. g. , mechanical [13], [58] or chemical [12] signal). In this study, woodlice in larger groups respond proportionally slower than individuals in smaller groups. This observation could highlight the particular importance of aggregation in woodlice for maintaining collective resting phases and the associated benefits (e. g. , for reducing water loss [6]), in contrast with an anti-predation strategy as in many models. Woodlice, due to their cryptic way of life [59], should rarely be faced with hungry predators consuming many individuals in a short time scale in micro-habitats. Furthermore, mechanisms leading to maintaining cohesive groups in resting to reduce chronic water loss should be more adaptive than mechanisms leading to quick dispersion during rare predatory events. However, the nature of direct or indirect interactions between woodlice in groups remains poorly understood and deserves further study. The suggested aggregation pheromone in woodlice faeces [37] probably cannot explain the rapidity of the grouping but could be involved in its stability and the transition from an excited to a calm state. The humidity generated by transpiration of a group of woodlice (see [60]) could also indirectly be involved in the collective resting process. Lastly, direct contacts between conspecifics could also be involved in these strongly thigmotactic animals [46]. In many cases, aggregation is the interplay between individual responses to environmental stimuli and to groups [1], [61–64], including in woodlice [39]. In the field, the factors inducing dispersion in woodlice are fairly well known. Woodlice in temperate environments present a (more or less endogenous) circadian rhythm of activity [32–35], [65], [66]. Foraging appears as the main activity during their short nocturnal dispersal phases [33], [34]. Thus, the combined influence of environmental factors (such as the alternation of day/night) and individual conditions (such as physiology, satiety state) probably disturbs the self-amplification process of group retention and allows for group explosion. For example, the dispersal rate increases with increasing set-up brightness (unpublished data). It would be interesting to discern which steps are affected by the brightness. We know nothing about the effect of predation on group dispersion in terrestrial isopods. In this experimental study, we forced the formation of groups on a relatively short time scale. An important next step would be to investigate freely formed aggregates and test their stability over periods of time with a true biological significance. Additionally, characterising the behavioural state of the individuals in the aggregate would give finer information about the group effect on the resting state at the individual level.

Terrestrial isopods, commonly named woodlice or pill bugs, are commonly distributed soil-dwelling arthropods, particularly important in soils as macro-decomposers of leaf litter. Many species of woodlice are synanthropic and, for this reason, are easily observable in gardens, urban parks or composts. Harmless organisms and easy to raise, the woodlice represent an excellent pedagogical model in many schools, so that children may perform on these organisms various behavioral tests such as light escape or introduction to social behaviors. Indeed, woodlice are gregarious species and exhibit long phases of aggregation. Here, we propose a model based on simple rules involving calmed and excited individuals and a social transition between these two behavioural states to explain group cohesion in woodlice. This contagion model well reproduces our experimental results. Our approach provides important clues for the understanding of how social group effects and collective mechanisms may govern the stability and dispersion of aggregates in gregarious arthropods.

Abstract Introduction Materials and Methods Results Discussion

2015

Behavioural Contagion Explains Group Cohesion in a Social Crustacean

Genome sequences of several economically important phytopathogenic oomycetes have revealed the presence of large families of so-called RXLR effectors. Functional screens have identified RXLR effector repertoires that either compromise or induce plant defense responses. However, limited information is available about the molecular mechanisms underlying the modes of action of these effectors in planta. The perception of highly conserved pathogen- or microbe-associated molecular patterns (PAMPs/MAMPs), such as flg22, triggers converging signaling pathways recruiting MAP kinase cascades and inducing transcriptional re-programming, yielding a generic anti-microbial response. We used a highly synchronizable, pathogen-free protoplast-based assay to identify a set of RXLR effectors from Phytophthora infestans (PiRXLRs), the causal agent of potato and tomato light blight that manipulate early stages of flg22-triggered signaling. Of thirty-three tested PiRXLR effector candidates, eight, called Suppressor of early Flg22-induced Immune response (SFI), significantly suppressed flg22-dependent activation of a reporter gene under control of a typical MAMP-inducible promoter (pFRK1-Luc) in tomato protoplasts. We extended our analysis to Arabidopsis thaliana, a non-host plant species of P. infestans. From the aforementioned eight SFI effectors, three appeared to share similar functions in both Arabidopsis and tomato by suppressing transcriptional activation of flg22-induced marker genes downstream of post-translational MAP kinase activation. A further three effectors interfere with MAMP signaling at, or upstream of, the MAP kinase cascade in tomato, but not in Arabidopsis. Transient expression of the SFI effectors in Nicotiana benthamiana enhances susceptibility to P. infestans and, for the most potent effector, SFI1, nuclear localization is required for both suppression of MAMP signaling and virulence function. The present study provides a framework to decipher the molecular mechanisms underlying the manipulation of host MAMP-triggered immunity (MTI) by P. infestans and to understand the basis of host versus non-host resistance in plants towards P. infestans. Plants possess innate defense mechanisms to resist microbial infection [1], [2]. Efficient plant disease resistance is based on two evolutionarily linked layers of innate immunity. One layer involves cell surface transmembrane receptors that recognize invariant microbial structures termed pathogen- or microbe-associated molecular patterns (PAMPs/MAMPs), hereafter referred to as MAMPs [3]–[5]. MAMPs are not only shared by particular pathogen races, but are broad signatures of a given class of microorganisms. They constitute evolutionarily conserved structures that are unique to microorganisms and have important roles in microbial physiology. Typical MAMPs include lipopolysaccharides (LPS) of Gram-negative bacteria, bacterial flagellin and fungal cell wall-derived carbohydrates or proteins, some of which were shown to trigger plant defense in a non-cultivar-specific manner [3], [6]. The best-studied MAMP receptor in plants is FLAGELLIN-SENSITIVE 2 (FLS2) from Arabidopsis, a receptor-like kinase (RLK) with extracellular leucine-rich repeat domains [7]. The 22 amino acid peptide (flg22) corresponding to the highly conserved amino-terminus of flagellin is sufficient to trigger immune responses in Arabidopsis, tomato, tobacco and barley but not in rice [8]–[12]. Although different MAMPs are perceived by different receptors, convergent early-signaling events, including MAP kinase activation and specific defense-gene induction, have been observed in Arabidopsis plants and protoplasts [13]–[15]. Suppression of flg22-induced defenses by bacterial virulence effectors suggests that manipulation of MAMP-triggered immunity (MTI) in plants is a key strategy for successful pathogens to grow and multiply (reviewed in [16]–[19]). A major target of bacterial effectors is the plant MAP kinase cascade, probably because of the central role of MAP kinase signaling in MTI. The Pseudomonas syringae effector HopAI1 displays phosphothreonine lyase activity and inactivates MPK3, MPK6, and MPK4 in Arabidopsis by dephosphorylating them [20]. P. syringae effector HopF2 blocks MAMP-induced signaling by targeting MKK5, a MAP kinase activating MPK3/MPK6, through a different mechanism of action i. e. ADP-ribosylation [21]. Bacterial effectors can also suppress MAP kinase signaling by targeting the pattern recognition receptor complex as illustrated by the P. syringae effectors AvrPto and AvrPtoB that block FLS2-mediated signal transduction in Arabidopsis and tomato [22]–[24]. Other effectors appear to act downstream of the activation of the MAPK cascade by blocking the expression of defense-associated genes in the nucleus. Such an effector is XopD from Xanthomonas campestris that inhibits the activity of the transcription factor MYB30, resulting in suppression of basal immune responses and promotion of pathogen growth [25], [26]. Unlike bacterial effectors, little is known about the molecular functions of effectors from eukaryotic plant pathogens. It remains to be demonstrated whether these pathogens have evolved effectors that subvert early-induced MTI signaling above, at, or immediately downstream of MAP kinase cascades. Oomycetes, including downy mildews and Phytophthora species, establish intimate association with host plant cells through structures such as appressoria, infection vesicles and haustoria, which are believed to facilitate the delivery of effectors into the host cytoplasm [27]. The genome sequences of Phytophthora sojae, P. ramorum, P. infestans and Hyaloperonospora arabidopsidis are published [28]–[30]. Each genome encodes several hundred putative RXLR effectors. Most oomycete Avirulence (Avr) proteins characterized so far carry a signal peptide followed by a conserved motif centered on the consensus RXLR- (EER) sequence, where X is any amino acid [31]. It has been shown that the RXLR peptide motif acts as a host-targeting signal for translocation into plant cells [31], [32]. Amongst the best-characterized oomycete RXLR effectors are AVR3a, AVRblb2 and PITG_03192 from P. infestans, AVR1b and AVR3b from the soybean pathogen P. sojae and ATR1 and ATR13 from H. arabidopsidis [33]–[47]. P. infestans Avr3a alleles encode secreted proteins of 147 amino acids that differ in two residues which determine recognition; only the isoform AVR3aKI is recognized by the potato resistance protein R3a, whereas AVR3aEM evades detection by R3a. When expressed in Nicotiana benthamiana cells, AVR3a suppresses host cell death induced by the elicitin INF1, a typical MAMP [35], [37]. It has since been shown to suppress cell death elicited by perception of a range of pathogen molecules by direct interaction with, and stabilization of, the plant E3 ligase CMPG1 [36], [42]. The Avrblb2 gene family is highly polymorphic and different forms/alleles are present in different P. infestans isolates. Sequence alignment of the deduced amino acid sequences of the Avrblb2 family members showed that the C-terminal effector domain undergoes positive selection, which is strong evidence for co-evolution with host resistance and/or target proteins [44]. The amino acid residue at position 69 was shown to be crucial for recognition by the cognate resistance protein Rpi-blb2 [44]. AVRblb2 was shown to block the secretion of a C14 cysteine protease that is involved in plant resistance against P. infestans [38]. Recently, the RXLR effector PITG_03192 has been shown to enhance P. infestans colonization of N. benthamiana by its interaction with NAC DNA binding proteins at the host endoplasmic reticulum, preventing their re-localization into the nucleus following pathogen perception [43]. Suppression of MTI has also been reported for ATR1 and ATR13 in Arabidopsis [47]. Nevertheless, for the majority of RXLR effectors, their biological functions and potential host targets are unknown. Transient expression in protoplasts has proven fast and reliable for studying the function of bacterial type III effectors that suppress early MAMP signaling [48], [49]. Moreover, the assay allows the measurement of synchronized responses and it does not require the use of bacteria for protein or DNA transfer into the host cell. In addition, the protoplast system offers the possibility to test large sets of effectors in a medium-high throughput manner. In this study, we have used tomato mesophyll protoplasts to screen a library of 33 P. infestans RXLR effector candidates (PiRXLRs) for their ability to suppress flg22-triggered defense signaling. Our additional aim was to test whether PiRXLRs that suppress early MTI signaling in the host plant tomato retain that ability in the distantly-related non-host plant Arabidopsis. For the experimental read-out we measured the abilities of these effectors to suppress: i) flg22-induced promoterFLG22-INDUCED RECEPTOR-LIKE KINASE 1 - LUCIFERASE (pFRK1-Luc) reporter gene activity; ii) flg22-induced post-translational MAP kinase activation; and iii) flg22-induced gene expression. In addition, we performed sub-cellular localization studies of fluorescent protein-tagged PiRXLR effectors by confocal microscopy. Finally, we tested the potential of the PiRXLR effectors suppressing early MTI signaling to enhance N. benthamiana susceptibility to P. infestans. Unraveling the mode-of-action of PiRXLR effectors within plant cells will help to gain insight into the specific mechanisms that coordinate different signaling and metabolic pathways to ensure proper plant development and response to environmental changes or stresses. A prerequisite to performing a screen that would allow us to identify PiRXLR effector candidates suppressing early events of MAMP signaling pathways in both a host (tomato) and a non-host (Arabidopsis) of P. infestans was to develop comparative bio-assays. Several components of the flg22-triggered signaling pathway are conserved in Arabidopsis and tomato. SlFLS2, the ortholog of AtFLS2, binds flg22 [50]. The MAP kinase orthologs of AtMPK3 and 6 in tomato are SlMPK3 and 1, respectively [51]. We adapted most of the techniques and materials that were generated for the identification and functional characterization of the P. syringae type III effector AvrPto, a well-studied suppressor of early MAMP signaling in both Arabidopsis [48], [49] and tomato [52]. Figure S1 shows that we could reproduce the AvrPto-mediated suppression of early MTI signaling observed in Arabidopsis protoplasts [48]. Moreover, we were able to extend this assay to tomato, and the induction of luciferase activity under control of the flg22-responsive promoter of FRK1 (pFRK1-Luc) was strongly impaired in Arabidopsis and tomato protoplasts expressing AvrPto with a C-terminal Green Fluorescent Protein (GFP) fusion (Figure S1A, B). An inactive AvrPto in which the Gly residue in position 2 is replaced by an Ala (AvrPto G2A-GFP), preventing the myristoylation and membrane localization of the effector protein [53], could not suppress pFRK1-Luc activation by flg22 (Figure S1A, B). Furthermore, we confirmed that AvrPto-GFP but not the AvrPto G2A-GFP mutant blocks post-translational activation of flg22-responsive MAP kinases in both protoplast systems (Figure S1 C, D). We searched for PiRXLR effectors interfering with flg22-induced early immune responses in protoplasts of tomato, a host for P. infestans. Thirty-three PiRXLR effector genes, most of which were selected on the basis of their up-regulation during the biotrophic phase of infection [28], [32], [44], were cloned without the native secretion signal peptide into pDONR Gateway vectors (Table S1). We sub-cloned these sequences into Gateway destination vectors of the p2GW7 series to allow transient expression with/without an N-terminally fused GFP tag. For the initial read-out, we measured pFRK1-Luc activity upon flg22 treatment. Of the 33 PiRXLR effectors screened, 8 (PITG_04097, PITG_04145, PITG_06087, PITG_09585, PITG_13628, PITG_13959, PITG_18215 and PITG_20303) reduced consistently and reproducibly flg22-induced pFRK1-Luc activation in tomato protoplasts, when compared to control protoplasts expressing only GFP (p-value<0. 05 - Figure 1: S. lycopersicum). We named these effectors Suppressor of early Flg22-induced Immune response (SFI) 1 to 8, respectively. Protoplast staining with vital dyes, 24 h after plasmid transformation, showed that the percentage of dead cells is, with the exception of a higher (but non-significant) value for SFI6, similar for each of the tested PiRXLR effectors and the GFP control (Figure S2). Therefore, the suppression of reporter gene activity is not the consequence of a toxic or a programmed cell death process in transformed protoplasts. Five PiRXLR effectors (SFI1 and SFI5-8) reduced pFRK1-Luc activation by flg22 with an efficiency comparable to the bacterial effector AvrPto (+flg22/−flg22≅1). Among PiRXLR effectors with a reported avirulence function in potato, only AVRblb2 (SFI8) [44] was able to suppress flg22-induced pFRK1-Luc activity. SFI8 is a representative member of a large family of AVRblb2-related proteins but it bears a Phe residue at position 69 in its sequence and, therefore, is predicted not to be recognized by Rpi-blb2 [44]. Thus, we extended our analysis to three more AVRblb2 family members with either an Ala (PITG_20300 and PITG_04090) or Ile (PITG_04085) at position 69 and crucial for Rpi-blb2-mediated HR (Table S2). Both predicted Rpi-blb2-recognized and -unrecognized isoforms of AVRblb2 equally suppressed reporter gene activation (Figure S3A). Other PiRXLR effectors identified as avirulence proteins such as AVR1 [28], AVR3a [34], AVR4 [54] and AVRblb1/IPI-O1 or IPI-O4 [55], [56] did not interfere with early flg22-induced responses in our assay (Figure 1: S. lycopersicum). In the case of AVR3a, both R3a-recognized AVR3aKI and R3a-unrecognized AVR3aEM had no effect on flg22-induced pFRK1-Luc activity (Figure 1: S. lycopersicum). Using quantitative real-time PCR (qRT-PCR) we monitored the expression levels of the eight PiRXLR effector genes that suppressed pFRK1-Luc activation in tomato protoplasts at different stages of potato infection, relative to their expression in sporangia. Previous expression analyses of P. infestans RXLR effector genes showed that, when detected by either qRT-PCR [24] or in microarray experiments [28], [57], they are up-regulated in the first 48–72 hours of infection, i. e. during biotrophy. Transcripts of SFI1-8 accumulated during the first 48 hours post-inoculation (Figure S4), consistent with a potential role in effector-triggered susceptibility. We extended our analyses to determine whether PiRXLR effectors that suppress pFRK1-Luc activity in the host tomato are able to also suppress such responses in the non-host plant Arabidopsis. The pFRK1-Luc reporter gene assay, which turned out to be more sensitive in Arabidopsis than in tomato, showed that four effectors (SFI1, SFI2, SFI5 and SFI8/AVRblb2) were also able to attenuate activation in Arabidopsis (p-value<0. 05 - Figure 1: A. thaliana). As observed in tomato, each tested AVRblb2 isoform suppressed reporter gene activation by flg22 in Arabidopsis protoplasts (Figure S3B), whereas AVR3a had no effect (Figure 1: A. thaliana). We found a further four effectors (PITG_00821, PITG_05750, PITG_16737 and AVRblb1/PITG_21388) that attenuated the flg22-dependent pFRK1-Luc activation only in Arabidopsis (p-value<0. 05 - Figure 1: A. thaliana). Like in tomato, transient expression of PiRXLR effectors in Arabidopsis protoplasts did not cause significant cell death (Figure S5). One effector, PITG_18670, significantly induced a stronger flg22-dependent pFRK1-Luc activity than did the GFP control (p-value<0. 05 – Figure 1: A. thaliana), but did not do so in the host plant tomato (Figure 1: S. lycopersicum). This effector was not pursued further in this work. The observation that 4 PiRXLR effectors suppress flg22-mediated pFRK1-Luc induction in the non-host plant Arabidopsis, but not in the host plant tomato, was unexpected. This prompted us to test whether all 8 PiRXLR effectors that suppress pFRK1-Luc induction in Arabidopsis also inhibit the endogenous expression of early MAMP-regulated genes. First, we measured the level of endogenous FRK1 in Arabidopsis following flg22 treatment. This experiment confirmed the data obtained in the reporter gene assay with 3 PiRXLR effectors (SFI1, SFI2 and SFI8/AVRblb2) attenuating the up-regulation of FRK1 expression by flg22 (Figure 2A). In contrast, SFI5, as well as PITG_00821, PITG_05750, PITG_16737 and AVRblb1/PITG_21388, failed to suppress flg22-induced FRK1 expression (Figure 2A). We extended our analysis to an additional MAMP-induced gene, WRKY DNA-BINDING PROTEIN 17 (WRKY17), and observed that its up-regulation was also notably diminished by SFI1, SFI2 and SFI8/AVRblb2 (Figure 2B), whereas SFI5, PITG_00821, PITG_05750, PITG_16737 and AVRblb1/PITG_21388 again had no effect. The expression of the housekeeping gene ELONGATION FACTOR 1A (EF1α) was generally not altered. Only with SFI2 did we observe a 2–3 fold decrease of the EF1α transcript level, possibly as a consequence of reduced cellular fitness due to effector expression (Figure 2C). Indeed, the expression of all genes tested was barely detectable in the presence of this effector. Together, our initial results revealed a set of 8 PiRXLR effectors that are candidate suppressors of early flg22-mediated MTI signaling in tomato, and assigned a novel function to the previously described AVRblb2 effector family. Moreover, our data predict that 3 of these PiRXLR effectors target processes contributing to MTI that are conserved in Arabidopsis and tomato. We proceeded to study all 8 effectors that suppress flg22-inducible reporter gene activation in tomato in more detail. From the initial screen for MTI signaling suppression we hypothesized that the function of 3 PiRXLR effectors (SFI1, SFI2 and SFI8/AVRblb2) may be conserved in both tomato and Arabidopsis while 5 effectors (SFI3, SFI4, SFI5, SFI6 and SFI7) may function specifically in tomato. We expected that the sub-cellular distribution of PiRXLR effectors might provide additional important information about their function in the cell. Therefore, these PiRXLR effectors, N-terminally fused to GFP, were transiently expressed in tomato (all SFI effectors) and Arabidopsis (only SFI1, SFI2 and SFI8/AVRblb2) protoplasts, and in N. benthamiana leaves for comparison, and visualized by confocal microscopy (Figure 3). We performed immunoblot analysis to confirm protein expression and stability of intact GFP-fusion proteins (Figure S6), and verified that GFP-tagged PiRXLR effectors were still functional and effectively suppressed pFRK1-Luc activity in protoplasts (Figure S7). Most of the GFP-tagged PiRXLR effectors were as active as the un-tagged proteins. Notably, GFP-SFI8/AVRblb2 functioned only weakly or not at all in Arabidopsis, but retained its function in tomato (Figure S7). SFI8/AVRblb2, C-terminally fused to GFP (SFI8-GFP) was also unable to suppress pFRK1-Luc activity in Arabidopsis protoplasts (Figure S8). The sub-cellular localizations of the 3 PiRXLR effectors (SFI1, SFI2 and SFI8/AVRblb2) affecting pFRK1-Luc/MAMP gene activation in both tomato and Arabidopsis are similar in each plant species (Figure 3A). GFP-SFI8/AVRblb2 showed nuclear-cytoplasmic localization whereas GFP-SFI1 and GFP-SFI2 localized predominantly in the nucleus, and were also apparent in the nucleolus (Figure 3A). In the case of GFP-SFI1, additional fluorescence signal was observed in the cytoplasm (and possibly at the plasma membrane [PM]) (Figure 3A). The 5 PiRXLR effectors (GFP-SFI3, -SFI4, -SFI5, -SFI6 and -SFI7) with a tomato-specific effect showed different subcellular localizations. GFP-SFI3 was enriched in the nucleus/nucleolus, GFP-SFI4 showed nuclear-cytoplasmic localization, and GFP-SFI5, -SFI6 and -SFI7 showed differing degrees of cytoplasmic localization and association with the PM (Figure 3B), with GFP-SFI5 almost exclusively localized to the PM. Additional sub-cellular localization studies performed upon Agrobacterium-mediated expression in N. benthamiana leaves confirmed the results obtained in protoplasts, suggesting that protoplasts are accurate in reflecting sub-cellular localizations of these effectors in planta. Confocal microscopy revealed distinct sub-nuclear localization patterns for the 3 PiRXLR effectors (GFP-SFI1-3) that were predominant in this compartment. GFP-SFI1 appears to localize in the nucleolus, GFP-SFI3 forms a ring around the nucleolus, whereas GFP-SFI2 showed a range of sub-nuclear localizations (Figure S9). The obvious differences in sub-cellular localization between effectors imply that different steps and/or pathways may be targeted by individual effectors that have in common the suppression of flg22-triggered pFRK1-Luc activity. We performed an epistatic analysis to find out which step of the flg22-triggered signaling pathway in tomato or Arabidopsis is affected by the PiRXLR effectors that suppressed pFRK1-Luc/MAMP responsive gene activation. We conducted immunoblot assays using the p44/42 antibody, raised against phosphorylated MAP kinases, to assess the impact of our effectors on the activation by flg22 of endogenous SlMPK1/3 and AtMPK3/6 in tomato and Arabidopsis protoplasts, respectively. AvrPto was used as a positive control, as it is known to block MTI signaling upstream of the MAP kinase cascade at the FLS2/BAK1 receptor complex [23], [24], [48]. In tomato, 3 effectors (SFI5-SFI7) consistently suppressed flg22-dependent post-translational MAP kinase activation (Figure 4A). We confirmed this result by performing transient expression of HA-tagged SlMPK1 and SlMPK3 in protoplasts followed by immunoprecipitation and in vitro MAP kinase assay (Figure 4B). In contrast, none of the 8 SFI effectors attenuated flg22-dependent post-translational MAP kinase activation in Arabidopsis (Figure 4C). This suggests that the effectors (SFI1, SFI2 and SFI8/AVRblb2) that were shown to attenuate flg22-induced gene activation in both tomato and Arabidopsis are most likely doing so downstream of MAP kinase activation. In the case of SFI5, the demonstration that it attenuates MAP kinase activation only in tomato (Figure 4A, 4C) is consistent with the observation that, although this effector suppressed pFRK1-Luc activation in Arabidopsis, it failed to suppress flg22-mediated up-regulation of endogenous FRK1 in that plant. To further elucidate the molecular mechanism (s) underlying the mode of action of SFI5-SFI7 in suppressing flg22-induced post-translational MAP kinase activation in tomato, we performed gain-of-function experiments using components that activate the MAP kinases SlMPK1 and SlMPK3 in the absence of flg22 signal. The ectopic expression of known key players of MAMP-signaling pathways, such as MAPK kinases and MAPKK kinases [48], [58] have helped to elucidate the steps at which bacterial effectors such as AvrPto interfere with MTI in Arabidopsis [48], [59]. In tomato and other solanaceous plants, MAP kinase signaling cascades are best studied in the context of programmed cell death (PCD) associated with effector-triggered immunity [51], [60], [61]. In N. benthamiana, PCD triggered by perception of the P. infestans MAMP INF1 requires NbMKK1 and its interaction with SIPK (salicylic acid-induced protein kinase; an ortholog of SlMPK1) [62]. The role of MAPKK kinases in tomato immunity is only documented for SlMAP3Kα and SlMAP3Kε [60], [61] and the best characterized MAPK kinases are SlMEK1 and SlMEK2 [60]. Whether these kinases contribute to flg22-triggered signaling in tomato is unknown. As shown in Figure S10, transient expression in tomato protoplasts of a constitutively active SlMEK2 (SlMEK2-DD), or the kinase domain of SlMAP3Kα (SlMAP3Kα-KD), led to post-translational activation of SlMPK1 and SlMPK3 in the absence of flg22. The constitutively active SlMEK1 (SlMEK1-DD) and kinase domain of SlMAP3Kε (SlMAP3Kε-KD) did not activate SlMPK1 and SlMPK3. The expression of the constitutively active SlMEK2 (SlMEK2-DD) and the kinase domain of SlMAP3Kα (SlMAP3Kα-KD) overrode the suppression of flg22-dependent activation of SlMPK1 and SlMPK3 by SFI5-SFI7 (Figure 5A, 5B). These results indicate that the three effectors suppress the signaling cascade very early; either upstream of MAPKK kinase activation, or specifically at the MAPK- and/or MAPKK kinase (s) involved in flg22 signaling. This is consistent with association of these effectors with the plant plasma membrane, where they may interfere with the earliest components of MAMP perception or signal transduction. Since in N. benthamiana PCD triggered by perception of the MAMP INF1 [62], or perception of Cladosporium fulvum effectors Avr4/9 by tomato Cf-4/9 receptors [60], [61], involves MAP kinase cascades, we tested whether SFI5-SFI7 were able to suppress either PCD event. In contrast to AVR3a, which is known to suppress PCD triggered by INF1 or by co-expression of Cf-4/Avr4 ([42]); Figure 6a, 6b – p-value<0. 01), GFP-SFI5 and GFP-SFI6 did not attenuate PCD triggered by either recognition event (Figure 6A, 6B). However, whereas GFP-SFI7 also failed to suppress Cf-4/Avr4-mediated PCD (Figure 6B), this effector significantly attenuated INF1-mediated PCD, albeit less efficiently than AVR3a (Figure 6B – p-value<0. 01). Our results indicate that SFI5 and SFI6 display functional specificity by targeting the flg22/FLS2 MAP kinase cascade, but not suppressing MAP kinase cascades leading to Cf-4- or INF1-mediated PCD, whereas SFI7 has a broader suppressive effect which includes INF1- but not Cf-4-mediated PCD. The 8 PiRXLR effectors suppressing early MTI signaling in tomato are assumed to contribute significantly to the virulence of P. infestans. N. benthamiana was further used to explore the role of the 8 selected PiRXLR effectors in host colonization. Agrobacterium tumefaciens strains containing the PiRXLR effector construct were infiltrated into leaves of 2–3 week-old N. benthamiana plants. Leaves were challenged with P. infestans 1 day after agro-infiltration and lesion size (Figure 7A), as well as disease symptoms (Figure 7B), were recorded after an additional 7 days. Except SFI2, whose overexpression in N. benthamiana leaves caused cell death that interfered with the pathogen assay, we found that the remaining 7 PiRXLR effectors enhanced colonization of N. benthamiana by P. infestans (Figure 7A, 7B). Compared to the empty vector control, the expression of the PiRXLR effectors caused a two- to five-fold increase of the lesion size (p<0. 001) due to enhanced P. infestans colonisation. The strongest effect was observed when expressing GFP-SFI1. Interestingly, this is one of the effectors that localizes predominantly to the nucleus/nucleolus and suppresses flg22-mediated induction of MTI response genes in both Arabidopsis and tomato, but does not suppress MAP kinase activation, suggesting that it may act downstream of this step. We were thus prompted to look further at the significance of the nuclear/nucleolar localization of SFI1 on its virulence function. We attempted to address the importance of the nuclear localization for the function of SFI1 and hypothesized that mis-targeting of the effector away from the nucleus could impact its virulence function. We generated a construct introducing a myristoylation site at the N-terminus of GFP-SFI1. Transient expression of myr-GFP-SFI1 in planta showed that the myristoylation site was functional in targeting SFI1 to the plasma membrane in Arabidopsis protoplasts (Figure 8A) and N. benthamiana leaves (Figure 8B). Both GFP-SFI1 and myr-GFP-SFI1 fusion proteins were stable and intact in planta (Figure 8F). Strikingly, whereas the flg22-dependent induction of pFRK1-Luc activity was suppressed by GFP-SFI1 in Arabidopsis protoplasts, no such suppression was observed in the presence of myr-GFP-SFI1 (p-value<0. 05 - Figure 8C). Notably, myr-GFP-SFI1 lost its ability to enhance P. infestans colonization of N. benthamiana (Figure 8D, 8E), providing strong evidence that suppression of MAMP-induced immune responses by this effector in both host and non-host plants requires its localization to the nucleus/nucleolus. In this study, we used a protoplast-based system to assess the potential for RXLR effectors from P. infestans (PiRXLRs) to manipulate MAMP-triggered early signaling in both a host and non-host plant species. Of 33 PiRXLR effector candidates, selected on the basis of up-regulation during the biotrophic phase of late blight infection, 8 (SFI1-SFI8) were able to suppress flg22-mediated induction of pFRK1-Luc activity in protoplasts of the host plant tomato (summarized in Table 1). Of these, three (SFI5-SFI7) were shown to suppress flg22-dependent MAP kinase activation at - or upstream of - the step of MAPK- and/or MAPKK kinase activation, indicating that they target the earliest stages of MTI signal transduction in tomato (Table 1). As P. infestans does not possess flagellin, the ability of these effectors to attenuate flg22-mediated MAP kinase activation and early defense gene expression indicates that these events are likely stimulated following perception of as yet undefined oomycete MAMPs. We confirmed that 7 of the 8 PiXRLR effectors that suppress early MTI signaling in tomato also enhance colonization by P. infestans in the host plant N. benthamiana (Table 1). We found that 3 PiRXLR effectors (SFI1, SFI2 and SFI8/AVRblb2) suppress flg22-mediated induction of pFRK1-Luc activity in protoplasts of both the host plant tomato and the non-host plant Arabidopsis. We confirmed that suppression by all 3 effectors attenuates transcriptional activation of endogenous MAMP-induced marker genes in Arabidopsis (Table 1), indicating that some effectors may function efficiently across diverse (host and non-host) plant species. Interestingly, we found another set of 4 PiRXLR effectors that suppressed pFRK1-Luc activation only in the non-host Arabidopsis. This was a surprise, albeit the assay is potentially less sensitive in the host plant tomato. However, none of these effectors were able to prevent the activation of endogenous (Arabidopsis) MAMP-induced marker genes (Table 1). Therefore, additional experiments are necessary to determine to what extent suppression of flg22-induced post-transcriptional or translational processes may account for the activity of these effectors on the pFRK1-Luc reporter system in this plant. While 3 PiRXLR effectors (SFI5-SFI7) suppressed MAMP-dependent MAP kinase activation in tomato, no PiRXLR effector had a similar effect in Arabidopsis (Table 1). This is an important finding, consistent with the hypothesis of Schulze-Lefert and Panstruga [63] that non-host resistance in plants (in this case Arabidopsis), which are distantly related to the host of P. infestans, is likely to include failures in effector-triggered susceptibility, due to effectors that are not sufficiently adapted to adequately manipulate plant immunity. Each of these observations will be discussed below. The large number of RXLR effector gene candidates in Phytophthora genomes complicates their functional analysis by reverse and forward genetics. Thus, the development of a medium/high- throughput system to explore their function in plants is strongly desired. Other large-scale effector functional screens have been conducted recently. A study by Fabro et al. [64] identified 39 out of 64 RXLR effectors from Hyaloperonospora arabidopsidis that enhance P. syringae growth in Arabidopsis Col-0 when delivered via the type III secretion system (T3SS). A majority of these effectors was additionally able to suppress callose deposition in response to bacterial MAMP perception. Thirteen of the H. arabidopsidis RXLR effectors promoted bacterial growth in turnip (Brassica rapa), a member of the Brassicaceae that is a non-host of H. arabidopsidis, indicating that they likely retain their virulence function in this closely related plant. Although the authors did not provide molecular evidence of the influence of these RXLR effectors on MTI in turnip, their results are in line with our conclusions, in that the activity of some RXLR effectors is not restricted to the pathogen' s host (s). Nevertheless, a number of H. arabidopsidis RXLR effectors that promoted P. syringae growth in Arabidopsis either failed to do so (44 effectors) in turnip, suggesting that they fail to function in the non-host plant, or reduced P. syringae growth (7 effectors), suggesting that they had activated ETI. Whereas we have identified a set of PiRXLRs that suppress early MTI signaling in tomato but not in Arabidopsis protoplasts, none of the tested PiRXLRs in our study significantly promoted cell death in Arabidopsis protoplasts. In apparent contradiction to the molecular evolutionary concept of non-host resistance [63] we have also identified three PiRXLR effectors that potentially attenuate early flg22-mediated MTI signaling events in Arabidopsis. In order to demonstrate whether failure to suppress MTI has the potential to contribute to non-host resistance to P. infestans in Arabidopsis, it would be necessary to extend the analysis to all PiRXLR effectors and provide an in-depth study of their precise function in both host and non-host plant. Our primary goal was to identify and ascribe functions to PiRXLR effector proteins that interfere with early plant defense responses. Interestingly, AVRblb2 family members (such as SFI8), but not AVR3a, were among effectors suppressing flg22-induced pFRK1-Luc activity. This apparently contrasts with the results obtained from the screen for suppression of cell death mediated by the MAMP INF1 in N. benthamiana, in which AVR3a but not AVRblb2 family members acted as a suppressor [35], [37], [44]. Similarly, PITG_14736/PexRD8 also suppressed INF1-mediated PCD [44] whilst failing to attenuate flg22-mediated responses in this study, and SFI5/PexRD27 suppressed flg22-mediated MAP kinase activation here, whilst failing to suppress INF1-mediated PCD ([44]; Figure 6). Possible explanations would be that AVR3a and PexRD8 disable components located downstream of the MAMP signal transduction and early transcriptional changes studied here, or that these effectors act specifically on alternative signal transduction events related to INF1-mediated cell death, but not the FLS2/flg22 pathway. The opposite may be true for SFI8/AVRblb2 and SFI5. Moreover, SFI7 suppresses flg22/FLS2-mediated signal transduction and attenuates INF1-mediated PCD, but not Cf-4-mediated PCD, whereas AVR3a attenuates both INF1-mediated and Cf-4-mediated PCD. Evidence is thus emerging of effectors with overlapping functions, at the phenotypic level, that are likely mediated by distinct modes of action at the mechanistic level. The epistatic analysis of the MAP kinase signaling cascade showed that SFI5-SFI7 presumably act upstream of the activation of the SlMPK1/SIPK and SlMPK3/WIPK MAP kinases in tomato protoplasts following flg22 perception. These effectors potentially function at the FLS2 receptor complex, or upon MAPKKK or MAPKK activity, or upon alternative regulators associated with this signal transduction pathway. As P. infestans does not possess flg22, and is thus unlikely to activate FLS2, the activity of any effectors upon the receptor complex must involve targets that are associated with bacterial and oomycete MAMP detection. Nevertheless, the absence of any suppressive activity of these effectors against CF4-mediated cell death and the modest suppression of INF1-mediated PCD only by SFI7 – two defense pathways that utilize alternative MAPKK kinases - imply specificity in the signal transduction pathways targeted by these effectors. It is important to note that all three effectors, to differing degree, associate with the plasma membrane, consistent with a potential action at the level of signal perception. Mukhtar et al. , [65], postulated that an overlapping subset of host proteins, so-called hubs, are targeted by oomycete (H. arabidopsidis) and bacterial (P. syringae) effectors that have arisen independently through convergent evolution. Therefore, future work will focus on identification of host proteins with which SFI5-SFI7 interact to better elucidate the molecular mechanisms underlying the action of these effectors. An effect of AVRblb2 on early MAMP signaling in solanaceous plant species has not been reported before, but it is has been shown that AVRblb2 affects plant immunity by inhibiting the secretion of C14, an apoplastic papain-like cysteine protease [38]. It is worth noting that in that study, AVRblb2 was exclusively localized at the plasma membrane, whereas in our experiments SFI8/AVRblb2 appeared mainly in the nucleus and cytosol. Yet, as AVRblb2 forms a large family and it is not clear which AVRblb2 isoform was exactly tested for the inhibition of C14 secretion [38], any apparent discrepancies in our results raise the possibility that different members of the AVRblb2 family have distinct or multiple cellular activities. Nevertheless, in our study all tested members of the AVRblb2 family were able to significantly suppress flg22-mediated induction of pFRK1-Luc activity in protoplasts of the host plant tomato. As the effectors SFI1, SFI2 and SFI8/AVRblb2 interfere with transcriptional up-regulation of MAMP-responsive genes in both host and non-host plants, we presume that they target conserved processes upstream of the earliest transcriptional responses. None of these effectors prevented MAP kinase activation, suggesting that they act downstream of such signal transduction. The nuclear localization of SFI1 and SFI2 in Arabidopsis, tomato and N. benthamiana may indicate that they directly manipulate regulatory processes leading to transcriptional up-regulation. For SFI1 we showed that its mislocalization to the plasma membrane, via addition of a myristoylation signal, prevented both its ability to suppress flg22-mediated MTI gene activation in Arabidopsis and its ability to enhance P. infestans colonization of the host plant N. benthamiana. This strongly implicates the nucleus as the site of effector activity for SFI1. It also indicates the importance of determining subcellular localization of effectors, as mis-targeting them provides a strategy for investigating their virulence function. The fact that SFI1 activity is apparently conserved in the non-host plant Arabidopsis indicates that we may draw on the wealth of genetic resources available in the model plant to further dissect the functions of this effector. Future work will employ additional mis-targeting approaches, for example nuclear export (NES) and nuclear localization (NLS) signals, to better elucidate the potential contributions of SFI1-SFI8 activities, either within or outside the nucleus, to suppress early MTI signaling. Three of the PiRXLR effectors, SFI5-SFI7, suppressed flg22-mediated post-translational MAP kinase activation in tomato but not in the non-host Arabidopsis. A further two effectors, SFI3 and SFI4, were shown to suppress specifically pFRK1-Luc activation in tomato, although we need to confirm their inhibitory effect on the expression of endogenous MAMP-responsive genes. Nevertheless, each enhanced P. infestans colonization when transiently expressed in N. benthamiana, consistent with a role in MTI suppression. Functional characterization of all these effectors is thus better pursued in host plants within the Solanaceae. The availability of genome sequences for potato [66], tomato [67] and N. benthamiana [68], the genetic tractability of the diploid tomato [67], and the range of functional analyses that can be performed in N. benthamiana [52], considerably broaden opportunities to do this. Moreover, the adaptation of the Arabidopsis protoplast-based system [48], [49], [58] to investigate the earliest stages of MTI in tomato, presented here, further enhances capabilities to study the functions of effectors from pathogens that infect solanaceous hosts. Future work will employ transgenic host and nonhost plants expressing the effectors revealed here, and additional RXLR effectors from P. infestans, to more specifically investigate their precise mechanistic action. Such studies will also reveal those effectors which may act downstream of the earliest signaling events in order to suppress MAMP-triggered immunity. Ectopic expression in N. benthamiana of 7 of the 8 SFI effectors selected through the protoplast-based screen enhanced plant susceptibility toward P. infestans infection. This result suggests that the suppression of early signaling events triggering basal immunity is an important step toward successful host colonization by this pathogen. P. infestans itself offers the possibility to further study functional aspects of PiRXLR effectors, and gain- and loss-of function experiments may confirm the importance of our candidate effectors for virulence. However, it should be noted that the functional redundancy of the PiRXLR effectors studied here in suppressing early FLS2/flg22 MTI signaling suggests that silencing of these effector genes in P. infestans may not lead to clear virulence phenotypes, as has been shown by deletion studies with type III effectors in P. syringae [69]. Nevertheless, silencing of single PiRXLR effector genes Avr3a [36], or PITG_03192 [43] compromised P. infestans pathogenicity, indicating that (at least some of) the functions of these effectors are not redundant. In conclusion, the tomato protoplast system provides a new medium/high-throughput tool to identify effectors that modulate the earliest stages of MTI signal transduction. We have identified 8 PiRXLR effectors that suppress early flg22-mediated MTI in tomato. Three of these reveal association with the plant plasma membrane and act at, or upstream of, MAPKK activation specifically related to flg22-mediated MTI signal transduction. Two of these effectors, SFI5 and SFI6, apparently do not act on other MAP kinase-mediated signal transduction events studied in this investigation. In addition, five of the effectors act downstream of the MAP kinase cascade, 3 of which also clearly suppress early flg22-mediated gene induction in Arabidopsis. This demonstrates that the effector complement of P. infestans contains functional redundancy in the context of suppressing early MTI signal transduction and gene activation. It remains to be established why such functional redundancy is necessary, or is selected for, and it is consistent with studies of bacteria such as P. syringae [69] that plant pathogens evolve multiple means of confounding the host immune system. Solanum lycopersicum cv. Moneymaker was kept in a greenhouse under controlled growth conditions: 16 h light at 24°C/8 h dark at 22°C, 40%–45% humidity, ∼200 µE m−2 s−1 light intensity. They were grown on soil containing a 4. 6∶4. 6∶1 mixture of type P soil, type T soil (Patzer, Germany) and sand. Leaves from 3 to 4 week-old plants were used for experiments. Arabidopsis thaliana plants of the Col-0 ecotype were cultivated in a phytochamber under stable climate conditions: 8 h light at 22–24°C/16 h dark at 20°C, 40%–60% humidity, ∼120 µE m−2 s−1 light intensity. They were grown on soil composed of a 3. 5∶1 mixture of GS/90 (Patzer, Germany) and vermiculite. Leaves from 4 to 5 week-old plants were used for protoplast preparation. Nicotiana benthamiana was grown as described previously [36]. Phytophthora infestans putative RXLR effector genes (PiRXLR) were amplified minus the signal peptide from gDNA of the sequenced isolate T30-4 in a two-step nested PCR reaction in order to add flanking attB sites to the RXLR coding sequence. The cloning primers are shown in Table S1 and Table S2. The PCR products were recombined into pDONR201 or pDONR221 vectors (Invitrogen) to generate entry clones, which were further recombined into the vectors p2GW7, p2FGW7 (N-terminal GFP fusion), pB7WGF2 (N-terminal GFP fusion), p2GWF7 (C-terminal GFP fusion) or p2HAGW7 (N-terminal hemagglutinin-tag; derived from p2GW7) (VIB, Ghent University, Belgium) using the Gateway recombination cloning technology (Invitrogen). The myristoylation signal sequence MGCSVSK was added to the amino-termini of the GFP-PiRXLR fusions using PCR with modifying primers and restriction cloning into pENTR1a (Invitrogen) before recombination into p2GW7 or pB2GW7 (VIB, Ghent University, Belgium). The Gateway destination vectors used are designed for transient 35S promoter-driven gene expression in protoplasts or, in the case of pB7WGF2 and pB2GW7, in N. benthamiana plants. To generate the constructs used for epistasic analysis of the MAP kinase signaling cascade, four primer combinations: SlMEK1-attB1/SlMEK1-attB2, SlMEK2-attB1/SlMEK2-attB2, SlMAP3Kα-attB1/SlMAP3Kα-attB2 and SlMAP3Kε-attB1/SlMAP3Kε-attB2 (listed in Table S3) were used to amplify by PCR SlMEK1-DD, SlMEK2-DD, SlMAP3Kα-KD and SlMAP3Kε-KD from pER8 plasmid constructs, respectively. Subsequently, Gateway attB linkers were added via PCR using primers attB1-adapter and attB2-adapter. The obtained PCR products were introduced into pDONR201 to generate entry clones using the Gateway recombination cloning technology (Invitrogen). The genes were further recombined into the vector p2GWF7 (C-terminal GFP fusion – VIB, Ghent University, Belgium). The resulting plasmid constructs, p35S-SlMEK1-DD-GFP, p35S-SlMEK2-DD-GFP, p35S-SlMAP3Kα-KD-GFP and p35S-SlMAP3Kε-KD-GFP were used for protoplast transfection as described below. Plasmid DNA was isolated from E. coli DH5α liquid cultures by column purification using the PureYield Plasmid Midi-prep system (Promega) following the manufacturer' s instructions. For selected candidate gene, control genes and reporter gene constructs, higher amount of the corresponding plasmids were purified on a cesium chloride density gradient. S. lycopersicum mesophyll protoplasts were prepared as described by Nguyen et al. , [52] with slight modifications. The lower epidermis of fully expended leaflets was gently rubbed with grated quartz, rinsed with sterile water and leaf strips were floated on enzyme solution containing 2% cellulose ‘Onozuka’ R10 (Yakult Pharmaceutical Industry), 0. 4% pectinase (Sigma) and 0. 4 M sucrose in K3 medium. After 30 min vacuum-infiltration and 3 h incubation at 30°C in the dark, the enzyme-protoplast mixture was filtered through a 45–100 µm nylon mesh. Viable protoplasts were collected by sucrose gradient centrifugation and washed in W5 buffer. After recovery on ice for 2 h, protoplasts were harvested by centrifugation and suspended to a density of 6*105 cells/ml in MMG buffer prior polyethylene glycol-mediated transfection. 100 µg plasmid DNA/ml protoplast suspension was used during transfection. Protoplasts samples were then incubated in W1 buffer at 20°C in the dark for 12 to 16 h allowing plasmid gene expression. The preparation of A. thaliana mesophyll protoplasts was performed according to the protocol from Yoo et al. , [70] with minor changes. Briefly, thin leaf stripes were dipped into 1. 5% cellulose ‘Onozuka’ R10 – 0. 4% macerozyme R10 solution (Yakult Pharmaceutical Industry), vacuum-infiltrated for 30 min and digested for 3 h at 20°C in the dark. After two subsequent washing steps with W5 buffer Arabidopsis protoplasts were suspended in MMG buffer to a concentration of 2*105 cells/ml. Arabidopsis protoplast transfection was performed as for tomato. Luciferase and GUS reporter gene assays were conducted to screen for immunity-suppressing effector genes. For this, A. thaliana or S. lycopersicum protoplasts were co-transfected with pFRK1-Luc, pUBQ10-GUS and an effector gene construct (or empty p2FGW7 serving as GFP control). For the luciferase assay, luciferin was added to 600 µl transfected protoplast solution to a final concentration of 200 µM. Protoplasts were transferred to an opaque 96-well plate (100 µl per well). For each sample, flg22 was added to 3 wells to a final concentration of 500 nM (+flg22). The remaining 3 replicates were left untreated (−flg22). The luminescence reflecting the luciferase activity was measured at different time-points using a Berthold Mithras LB 940 luminometer. For the GUS assay, 50 µl transfected protoplast solution of each sample was treated with 500 nM flg22 (+flg22) and 50 µl were left untreated (−flg22). Protoplast pellets were collected 3 or 6 h after flg22 elicitation. The cells were lysed in 100 µl CCLR solution (cell culture lysis reagent, Promega). For each sample, 3 technical replicates of 10 µl were incubated with 90 µl MUG substrate (1 mM 4-methyl-umbelliferyl-β-D-glucuronide, 100 mM Tris-HCl pH 8. 0,2 mM MgCl2) for 30 min at 37°C. The reaction was stopped with 900 µl 0. 2 M Na2CO3. The fluorescence was monitored in an opaque plate using a MWG 96-well plate reader with λex = 360 nm and λem = 460 nm. Raw data of Luciferase and GUS assays were processed using Microsoft Excel. First the mean value of the +flg22 and the −flg22 triplicates was calculated for each sample in both assays of an experiment. Next, the +flg22/−flg22 ratio was calculated using the values from the 3 or 6 h time-point of the Luciferase assay and divided by the corresponding +flg22/−flg22 ratio of the GUS assay for normalization. Statistical analysis was performed using one-way ANOVA followed by Dunnett' s multiple comparison test. Total RNA from 400 µl A. thaliana protoplasts was extracted with TRI reagent (Ambion) and treated with DNase I (Machery-Nagel) following the suppliers' protocols. Poly A-tailed RNA (1 µg) was converted to cDNA using the RevertAid reverse transcriptase (Fermentas) and oligo-dT primers. qRT-PCR reactions were performed in triplicates with Maxtra SYBR Green Master Mix (Fermentas) and were run on a Biorad iCycler according to the manufacturers' instructions. Relative gene expression was determined with a serial cDNA dilution standard curve. The Actin transcript was used as an internal control in all experiments. Data was processed with the iQ software (Biorad). qRT-PCR to measure PiRXLR gene expression was carried out on a time-course of potato leaves (cv Desiree) infected with P. infestans isolate 88069. Total RNA from infected leaf discs was extracted with RNeasy Plant mini kit (Qiagen) and treated with DNase I (Qiagen) following the suppliers' protocols. Poly A-tailed RNA (1 µg) was converted to cDNA using the Superscript II reverse transcriptase (Invitrogen) and oligo-dT primers. qRT-PCR reactions were performed in triplicate with Power SYBR Green Master Mix (ABgene) and run on a Biorad Chromo4 cycler according to the manufacturer' s instructions. Relative gene expression was determined using the ΔΔCT method, and P. infestans ActA gene was used as an internal control in all experiments, as described in Whisson et al [32]. Data was processed with Opticon monitor software (Biorad). Primers used in qRT-PCR reactions are listed in Table S3. To monitor the activation of MAP kinase, transfected protoplasts were challenged with 500 nM flg22. Pellets from 100 µl protoplast solution were collected 0,15 and 30 min after treatment and denatured in protein loading buffer. Proteins were loaded onto a 13. 5% SDS-polyacrylamid gel and separated by electrophoresis (SDS-PAGE) using the Biorad MiniProtean equipment according to the manufacturer' s instructions. PageRuler Prestained protein ladder (Fermentas) was used as a molecular weight marker. Proteins were blotted onto nitrocellulose membranes (Hybond–ECL, Amersham) and stained with 0. 1% Ponceau S to visualize equal sample loading. The membranes were blocked for 1 h at room temperature in 5% skimmed milk in TBS-T buffer (50 mM Tris-HCl pH 7. 5,150 mM NaCl, 0. 1% Tween 20), incubated overnight at 4°C in primary antibody solution (anti-phospho-p44/42 MAPK antibody, dilution 1/1000 in 5% BSA in TBS-T, Cell Signaling Technology) and finally incubated for 1 h at room temperature in secondary antibody solution (alkaline phosphatase-coupled anti-rabbit IgG antibody, dilution 1/3000 in TBS-T, Sigma). The immunoblot was revealed in NBT/BCIP detection solution. The expression of GFP-tagged PiRXLR effectors was assessed by immunoblotting using polyclonal anti-GFP antibody produced in rabbit or in goat (Acris Antibodies) at a 1/3000 dilution in 5% BSA in TBS-T. For this, protoplast samples were collected 12 (for S. lycopersicum) or 24 h (for A. thaliana) after transfection and SDS-PAGE and immunoblotting were carried out as described above. The MAP kinase in vitro kinase assay was carried out as described by He et al. , [48]. Briefly, 1 ml transfected protoplasts were lysed in 1 ml of immunoprecipitation (IP) buffer (150 mM NaCl, 50 mM HEPES pH 7. 4,1 mM EDTA, 1 mM DTT, 0,1% Triton X-100,1× phosphatase inhibitor cocktail [PhosphoSTOP, Roche Applied Science] and 1× protease inhibitor cocktail [Complete EDTA-free, Roche Applied Science]). HA-tagged SlMPK1 and SlMPK3 kinases [52] were immunoprecipitated from lysates after adding 20 µl anti-H antibody-coupled beads (Roche Applied Science) and incubated for 3 h at 4°C with gentle shaking. After centrifugation at 500 g for 1 min, the immunoprecipitated material was washed with IP buffer followed by a wash with kinase buffer (20 mM Tris-HCl pH 7. 5,20 mM MgCl2,5 mM EDTA and 1 mM DTT). The kinase reaction was performed by adding 25 µl of kinase buffer (0. 25 mg/ml MBP, 100 µM ATP and 5 µCi [γ-32P] ATP) for 30 min at RT. The reaction was stopped with 4× SDS-PAGE loading buffer. The 32P-labeled MBP was separated by SDS/PAGE (15%) gel and visualized by autoradiography. To determine the cell death rate after transfection (percentage of dead cells/total number of cells), 100 µl protoplast samples were incubated for 24 h and subsequently stained with 1 µg propidium iodide. Stained protoplasts were counted using a Nikon Eclipse 80i epifluorescence microscope with the following filter: TRITC EX 540/40, DM 565, BA 605/55. For sub-cellular localization studies protoplasts were monitored 12 h post-transfection and N. benthamiana leaves at 2 days post-infiltration. Imaging was performed using Leica TCS SP2 AOBS confocal microscopes with HCX PL APO lbd. BL 63×1. 20 W, L 40×0. 8 and L 20×0. 5 water immersion objectives. Samples were excited by an argon laser and fluorescence emission was detected at 496–552 nm for GFP and 620–726 nm for chloroplasts. The pinhole was set to 1. 5 airy units for protoplasts and 1 airy unit for leaf cells. Single optical section images were acquired from protoplasts and z-stacks were collected from leaf cells and projected and processed using the Leica LCS software and Adobe Photoshop CS3. A. tumefaciens transformed with pB7WG2 or pB7WGF2 vector constructs were grown overnight, pelleted, re-suspended in infiltration buffer (10 mM MES pH 5. 6,10 mM MgCl2 and 200 µM acetosyringone) and adjusted to the required OD600 before infiltration into N. benthamiana leaves. A. tumefaciens cultures were grown as above and subsequently mixed together to a final optical density at 600 nm (OD600) of 0. 3 for each construct except Cf4, which was used at 0. 6, N. benthamiana plants were infiltrated using a 1 ml needleless syringe through the lower leaf surface. Three leaves on six plants were used for each biological replicate. Cell death was scored at 7 d post-infiltration (dpi). An individual inoculation was counted as positive if >50% of the inoculated area developed clear cell death. The mean percentage of total inoculations per plant developing cell death of combined data from at least two biological replicates was calculated. One-way ANOVA was performed to identify statistically significant differences. A. tumefaciens Transient Assays (ATTA) in combination with P. infestans infection were carried out as described [36]. Briefly, Agrobacterium cultures were re-suspended in infiltration buffer at a final concentration of OD600 = 0. 1 and infiltrated in N. benthamiana with the bacteria harboring the vector control on one side of the leaf midrib and the bacteria harboring the PiRXLR effector constructs to be tested on the other. P. infestans strain 88069 cultured on Rye Agar at 19°C for 2 weeks was used for plant infection. Plates were flooded with 5 ml cold H2O and scraped with a glass rod to release sporangia. The resulting solution was collected and sporangia numbers were counted using a haemocytometer and adjusted to 30,000 sporangia/ml. After 1 day, each agro-infiltration site was inoculated with 10 µl droplets of sporangia. Three leaves per plant for 4–6 intact plants were used for each biological replicate. Lesions were measured and photographed at 7 days post-infection and data of at least two biological replicates were combined. Statistically significant differences in lesion size were identified by one-way ANOVA with pairwise comparisons performed using the Holm-Sidak method.

Phytophthora species are among the most devastating crop pathogens worldwide. P. infestans is a pathogen of tomato and potato plants. The genome of P. infestans has been sequenced, revealing the presence of a large number of host-targeting RXLR effector proteins that are thought to manipulate cellular activities to the benefit of the pathogen. One step toward disease management comprises understanding the molecular basis of host susceptibility. In this paper, we used a protoplast-based system to analyze a subset of P. infestans RXLR (PiRXLR) effectors that interfere with plant immunity initiated by the recognition of microbial patterns (MAMP-triggered immunity - MTI). We identified PiRXLR effectors that suppress different stages early in the signaling cascade leading to MTI in tomato. By conducting a comparative functional analysis, we found that some of these effectors attenuate early MTI signaling in Arabidopsis, a plant that is not colonized by P. infestans. The PiRXLR effectors localize to different sub-cellular compartments, consistent with their ability to suppress different steps of the MTI signaling pathway. We conclude that the effector complement of P. infestans contains functional redundancy in the context of suppressing early signal transduction and gene activation associated with plant immunity.

Abstract Introduction Results Discussion Materials and Methods

biochemistry plant biochemistry innate immune system plant science medicine and health sciences plant microbiology pathology and laboratory medicine immunity medical microbiology host-pathogen interactions microbial pathogens plant pathology biology and life sciences immunology microbiology pathogenesis immune system

2014

Functionally Redundant RXLR Effectors from Phytophthora infestans Act at Different Steps to Suppress Early flg22-Triggered Immunity

Brucellosis is a neglected tropical zoonosis allegedly reemerging in Middle Eastern countries. Infected ruminants are the primary source of human infection; consequently, estimates of the frequency of ruminant brucellosis are useful elements for building effective control strategies. Unfortunately, these estimates are lacking in most Middle East countries including Egypt. Our objectives are to estimate the frequency of ruminant brucellosis and to describe its spatial distribution in Kafr El Sheikh Governorate, Nile Delta, Egypt. We conducted a cross-sectional study in which 791 sheep, 383 goats, 188 cattle milk tanks and 173 buffalo milk tanks were randomly selected in 40 villages and tested for the presence of antibodies against Brucella spp. The seroprevalence among different species was estimated and visualized using choropleth maps. A spatial scanning method was used to identify areas with significantly higher proportions of seropositive flocks and milk tanks. We estimated that 12. 2% of sheep and 11. 3% of goats in the study area were seropositive against Brucella spp. and that 12. 2% and 12% of cattle and buffalo milk tanks had antibodies against Brucella spp. The southern part of the governorate had the highest seroprevalence with significant spatial clustering of seropositive flocks in the proximity of its capital and around the main animal markets. Our study revealed that brucellosis is endemic at high levels in all ruminant species in the study area and questions the efficacy of the control measures in place. The high intensity of infection transmission among ruminants combined with high livestock and human density and widespread marketing of unpasteurized milk and dairy products may explain why Egypt has one of the highest rates of human brucellosis worldwide. An effective integrated human-animal brucellosis control strategy is urgently needed. If resources are not sufficient for nationwide implementation, high-risk areas could be prioritized. Brucellosis is one of the most common zoonotic diseases worldwide, and as such poses a major threat to human health and animal production [1]–[2]. It is considered a neglected zoonosis by the World Health Organization (WHO), and has been identified as having the highest public health burden across all sections of the community; livestock keepers, consumers of livestock products and general population [3]. Several Middle Eastern and central Asian countries have recently reported an increase in the incidence of human brucellosis and the appearance of new foci [4]. Among the Middle East countries, Syria, Saudi Arabia, Iraq, Iran and Turkey have reported the highest annual incidence rates of human brucellosis worldwide with the exception of Central and Inner Asian countries; 160,21,28,24 and 26 cases/100,000 persons-years at risk, respectively [4]. In Egypt, brucellosis is endemic among humans and domestic ruminants [5], and it has recently been found that catfish in the Nile Delta region can be naturally infected with Brucella melitensis [6]. There is a lack of information on the frequency of human brucellosis at the national level in Egypt, with few available figures obtained mainly from small scale surveys and hospital-based studies [4]. In the Nile delta region, the incidence was estimated at 18 cases/100,000 population in 2000 [7] and the seroprevalence within a village in the Gharbia governorate was estimated at 1. 7% in 2003 [8]. To try to address the lack of reliable information, Jennings et al. [9] used population-based surveillance data to estimate the frequency of human brucellosis in one of the Upper Egypt governorates (Al Fayoum). They reported an incidence of 64 and 70 cases /100,000 population in 2002 and 2003 respectively, and found that hospital based surveillance identified less than 6% of human brucellosis cases. Reliable estimates of the frequency of brucellosis among ruminants in Egypt are also lacking despite an official control policy based on annual serological testing of all ruminant species over 6 months of age. Failure to test all eligible animals every year as per official guidelines, and non-random selection of herds/flocks or animals to be tested, are the reasons why accurate estimates of the seroprevalence of ruminant brucellosis in the country are not available [10]. The largest survey conducted so far across all governorates was carried out from 1994 through 1997, when 40% of the total ruminant population in the country was serologically tested against Brucella spp. as part of a national brucellosis surveillance and control project funded by United States Agency for International Development (USAID). The seroprevalence of brucellosis was estimated then at 0. 9%, 0. 3%, 1. 8% and 8. 2% of the cattle, buffalo, sheep and goat population, respectively [5], [11]. A recent study of 126 herds found 17. 2%, 26. 6% and 18. 9% of the cattle farms, sheep flocks and goat flocks tested to be seropositive [12], but no information is given about the selection of herds/flocks which seem to have been conveniently or purposively selected. Ruminant species infected with Brucella spp. are known to be the primary source of human infection in Egypt and other endemic countries [5], [13]. In Egypt, the close contact between farmers and their animals due to the predominance of small scale farms, occupational exposure of farmers, veterinarians and butchers to infected animals and consumption of unpasteurized milk and dairy products are considered to be the major risk factors for human infection with Brucella spp. [8], [9], [14]. This suggests that measures aimed at reducing the occurrence of brucellosis in animals are the most effective means of reducing human infection [15]. In order to undertake any control program, good quality data regarding the seroprevalence of infection among animals is highly desirable. As previous experiences in different countries have demonstrated, the more appropriate combination of specific measures for the control of ruminant brucellosis depends on the baseline frequency of infection; this is reflected in guidelines issued by international organizations such as the Food and Agriculture Organization (FAO) and the World Organization for Animal Health (OIE) [16]. The objectives of the present study are therefore to estimate the seroprevalence of ruminant brucellosis and to describe its geographic distribution in one of the largest governorates of the Nile Delta region, the Kafr El Sheikh governorate. Up to 85% of the cows and buffaloes in Egypt are reared as household animals in small herds typically of less than five animals. They have frequent contact with sheep and goats, which are sometimes also kept as household animals in the farmers' houses [17]. A typical village in the study area would have several milk tanks (usually between five and 15 for cow' s milk and the same number for buffalo' s milk), one milk collector is usually responsible to manage one to three tanks for each species, to which farmers take twice a day the milk surplus that they want to sell. Milk collectors have three main channels to sell the milk they collect. First, they can sell milk directly to local consumers in the same village. Second, they can sell milk to food shops in nearby villages which sell it to consumers as fresh unpasteurized milk. Third, there are several small and a few large dairy processing plants which buy milk from collectors and either sell it as fresh milk, cream or butter without heat treatment or as pasteurized milk and milk products [18]. Not all farmers sell all their milk surplus to milk collectors, some sell milk and dairy products directly in the local markets and this milk is typically sold without heat treatment. The majority of small ruminant flocks in the villages were kept as small sheep flocks, goat flocks, or mixed flocks of both species managed by sheepherders [17], [19]–[20]. One sheepherder would often keep sheep from a number of different owners; as a result animals from different households are part of the same flock for grazing and breeding during most of the year. A multistage random sampling strategy was used to select cattle milk tanks and individual sheep and goats within the governorate. The first level sampling units in this study were the villages, the second level sampling units were the cattle milk tanks and the individual sheep/goat. The sampling frame consisted of the 206 villages within the governorate. In each district (stratum), the number of villages to be sampled was proportional to the size (total number of villages) of the district (sampling proportional to size). Within each selected village, sample frames of milk collectors and of sheep/goat flocks managed by individual sheepherders were constructed with the help of the village veterinarians and some farmers. Milk collectors were selected using simple random sampling and for each of them a milk sample for each species was taken from the milk tank. If the collector managed more than one tank for either species, one tank for each species was selected by the investigator by pointing at one of the tanks without applying any defined rule (haphazard selection). All the sheep and goats reared in the village were considered as belonging to a single flock: the “village flock”. However, the management of this “village flock” is typically the responsibility of a small number of sheepherders, among which the village flock is divided for purpose of management. The number of sheep and goats to be sampled within one village was equally divided between the existing sheepherders and individual animals were selected when passing through an opening with a flock-size specific sampling interval, or, when this was not possible, the investigator pointed at individual animals for sampling without a specific rule. One liter of whole milk was collected from each selected bulk milk tank and kept at room temperature for three to six hours until transported to the laboratory. Fifteen ml of milk was placed in a sealed McCartney bottle and preserved at −20°C until tested. Whole blood samples were collected from all selected individual sheep and goats using centrifuge tubes and transported directly to the laboratory where the sera were separated after centrifugation and preserved at −20°C until tested. Milk samples were tested using an indirect enzyme linked immunosorbent essay (iELISA) for the presence of Brucella spp. antibodies. Serum samples were tested using Rose Bengal Plate test (RBPT). Only serum samples that were seropositive by RBPT were sent to the Animal Health Research Institute (AHRI) in Cairo for confirmation using Complement Fixation test (CFT). Serum samples which gave positive results in both tests were considered seropositive, while negative samples were those which gave negative results to either RBPT or CFT. All serological kits and reagents used were obtained from the OIE Reference Centre and an FAO/WHO Collaborating Centre for Brucellosis at the Veterinary Laboratories Agency, Weybridge, United Kingdom. All techniques were done according to the instructions of the manufacturer. A range of likely values of sensitivity (Se) and specificity (Sp) of the RBPT and CFT tests when applied at the individual animal level and of the iELISA test when applied to bulk milk samples were obtained from the literature: RBPT (0. 72≤Se≤1; 0. 8≤Sp≤1); CFT (0. 81≤Se≤1; 0. 8≤Sp≤1); iELISA (0. 95≤Se≤1; 0. 92≤Sp≤1) [21]–[25]. For purpose of sample size calculation fixed values of Se and Sp were used for each test: For the series combination of RBPT and CFT we used Se = 0. 9 and Sp = 0. 9 and for the iELISA we used Se = 0. 95 and Sp = 0. 92. For purpose of seroprevalence estimation the likely values of combined sensitivity (CSe) and specificity (CSp) of the series interpretation of RBPT and CFT were calculated as CSe (0. 78) and CSp (0. 99), respectively in another study by the authors (Y. Hegazy, unpublished. data). In this study, most likely values of CSe and CSp were obtained using simulation. The values reported in the literature for the Se and Sp of individual tests and mentioned above were used as input probability distributions in the simulation. For estimation of the true seroprevalence of milk tanks we used values sensitivity (SeELISA) = 0. 98 and specificity (SpELISA) = 0. 98. The number of milk tanks to be sampled was calculated in order to estimate the proportion of seropositive tanks with 95% confidence and 6% absolute error (d), for an expected proportion of seropositive tanks of 50%. The necessary sample size (N) was calculated as in [26] as following: The resulted number of samples needed was multiplied by a design effect to consider the multistage level clustering of the sampling design. The design effect was calculated as: Where m is the number of animals per cluster and ICC is the intracluster (intravillage) correlation coefficient. In the absence of suitable estimates of ICC for brucellosis under local husbandry systems, we used ICC = 0. 1, calculated from what we believed was a plausible scenario for the within and between village distribution of positive tanks. We calculated that 35 villages in total and 5 milk tanks for each species per village would be sufficient to estimate the prevalence of seropositive tanks in the governorate with the desired absolute error. We decided to study 40 villages. The number of sheep and goats to be sampled was calculated in order to estimate the proportion of seropositive individual animals against Brucella spp. with 95% confidence and 6% absolute error, for an expected seroprevalence amongst sheep and goats of 15%. The same equations as for the calculation of the number of milk tanks were used. Using plausible scenarios of within and between sheep and goat seroprevalence, we calculated ICC values of 0. 1 for sheep and 0. 05 for goats. The low value for goats reflects our expectation that due to the relatively low density of goats the impact of the presence of a positive goat within a cluster (village) would be smaller than for sheep. We calculated that if 40 villages were to be sampled, 20 sheep and 10 goats from each village flock would be sufficient to estimate the seroprevalence among small ruminants with the desired absolute error. Latitude and longitude of each milk tank and small ruminant flock sampled were obtained using a Global Positioning System (GPS). An electronic map of Egypt was provided by the General Organization of Veterinary Services (GOVS) in Egypt and the locations of the main markets in the study area identified. Results of serological testing of serum samples of small ruminants and milk tank samples of cattle and buffalo against Brucella spp. are shown in Table 1. A total of 82 (10. 4%) sheep and 37 (9. 7%) goats were classified as seropositive against Brucella spp with true seroprevalence among sheep and goats calculated as 12. 2% and 11. 3% respectively. The VFCSe and VFCSp were estimated at 0. 93 and 0. 76 for sheep and as 0. 87 and 0. 89 for goats, respectively. The true seroprevalence of villages with at least one seropositive sheep or goat was estimated at 41. 3% and 32. 2% respectively (Table 1). The true seroprevalence of villages with at least one seropositive small ruminant animal – either sheep or goat- was 60. 5% (95% CI: 45. 4%, 75. 7%). The distribution of VFTP is shown in figure 1. The distribution of true sheep+goat brucellosis seroprevalence estimates by district is shown in figure 2. A. A total of 188 cattle milk tanks and 173 buffalo milk tanks were sampled in the 40 villages. Of them, 22 (11. 7%) cattle milk tanks and 20 (11. 6%) buffalo milk tanks were classified as seropositive against Brucella spp and the true seroprevalences were calculated as 12. 2% and 12. 0%, among cattle and buffalo milk tanks, respectively (Table 1). The VTSeELISA and VTSpELISA were calculated as 0. 98 and 0. 88 respectively. The true seroprevalence of villages where at least one seropositive tank was found was 38. 4% (95% CI: 19. 6%, 49. 1%). The true seroprevalence of villages with at least one seropositive cattle milk tank was 15. 1%, and the same value was obtained for the true seroprevalence of villages with at least one positive buffalo milk tank (Table 1). When considering cattle and buffalo milk tanks together, we estimated that 22 (55%) of villages had no seropositive tanks and 18 (45%) had at least one seropositive tank. Of those villages with seropositive milk tanks, 11 (27. 5%) had less than 25% seropositive tanks, four (10%) of the villages had between 25% and 50% of tanks seropositive and in three (7. 5%) of the villages more than half of the tanks were seropositive against Brucella spp. The distribution of the true proportion of seropositive milk tanks against brucellosis by district is shown in figure 2. B. Intracluster correlation coefficients for sheep and goat flocks were estimated at 0. 21 and 0. 38 respectively. The southern districts of the governorate, near its capital, had the highest seroprevalence of small ruminant brucellosis (figure 2. A). Significant clustering of seropositive small ruminant flocks was identified within a 3. 3 km radius area in the proximity of the capital of the governorate (P<0. 001; figure 3). Flocks within this cluster were 3. 4 times more likely to be seropositive than flocks outside the cluster. When focused scanning was conducted around major animal markets, there was also evidence of clustering of seropositive flocks around three animal markets, one near the capital of the governorate (radius 2 km, relative risk 3. 4, P<0. 001) and two in the neighboring district of Byala (radius 17 km, relative risk 3 and radius 13 km, relative risk 3; P<0. 001) (figure 3). Although the seroprevalence of seropositive tanks appeared to be higher in southern districts (Figure 2. B), we did not find any significant clustering of seropositive tanks across the study area. The Nile Delta region has one of the highest human and ruminant densities in the world; with more than 125 person per km2 and more than 196 ruminant/km2 [30]–[31]. Most households in the region raise small numbers of cattle, buffaloes, sheep or goats which are kept in close contact with household members [14]. These animals are a source of meat and dairy products that are consumed within the same household or sold in local markets or to middlemen [14]. In the study area milk is mostly sold unpasteurized, either directly by the producers or indirectly by milk collectors or food shops. Cream and butter made by the farmers or by local dairy processing plants are also often sold without heat treatment. The potential for human exposure to zoonotic pathogens such as Brucella spp. is amplified by these demographics, husbandry practices and dairy production and marketing systems, which closely tie the incidence of brucellosis in the livestock and human populations [13]. To our knowledge, this is the first formal survey with probabilistic sampling carried out with the objective of estimating the seroprevalence of ruminant brucellosis in one governorate of Egypt. The results show that brucellosis is widely spread in the study area where seroprevalence values are very high among all ruminant species, suggesting a very intense transmission within the livestock population. In fact, considering all the sheep in one village as a single flock – which, given the production system, seems appropriate – the proportion of seropositive flocks in the area (60. 5%) is among the highest reported in the scientific literature for a small ruminant population [32]–[33]. Our estimates in the ruminant population are in accord with reports that identify Egypt as having one of the highest rates of human infection worldwide [9]. The coexistence with a heavily infected domestic ruminant population managed under husbandry systems such as those in place in Egypt and widespread marketing of unpasteurized milk and dairy products inevitably results in a high level of exposure of the human population. In ruminants, Brucella spp. is transmitted either in-utero or by direct contact between infected and susceptible animals, therefore, a high seroprevalence is necessarily indicative of a high frequency of contacts between infected and susceptible animals. It is likely that, in the study area, a high density of ruminants with free movement of small ruminant flocks results in frequent contact between animals from different households and villages. In the absence of vaccination and other sanitary measures, this contact structure creates the necessary conditions for sustaining Brucella spp. infection at higher seroprevalence levels than in other regions [5], [12], [33]. Our estimates for the intravillage correlation, especially among goats, are higher than those reported in Mexico and Ireland [34]–[35]. This may suggest a high within-villages transmission of brucellosis in the study area. These estimates could be used for study designs in future surveys to insure a proper sample size and better prevalence estimates. Although our study does not differentiate between Brucella strains, in Egypt, the main isolate in different animal species and humans is Brucella melitensis [5]. Given the high seroprevalence in small ruminants, it is likely that cattle act as spill-over hosts of Brucella melitensis. [36]. The recent isolation of Brucella melitensis from Nile Catfish in different regions of the Nile Delta points out the potential extent of Brucella melitensis infection pressure currently in the area [6]. Comparisons of our estimates with the results of the 1994–1997 national control campaign have to be made with great caution, since that nationwide study was not designed to generate unbiased prevalence estimates for the governorates. However, if the 1994–1997 estimates did not heavily underestimate the existing seroprevalence of infection at the time (an assumption that seems reasonable to us), the seroprevalence of ruminant brucellosis in the study area has increased considerably in the last 10 years. The establishment of infection as endemic at such high levels across the different species is also indicative of the ineffectiveness of the control program that has been in place since 1981. Recent reports have shown the inability of the test and slaughter element of the program to test more than 7% of the total ruminant population each year in this governorate as well as the noncompliance with the official vaccination and quarantine policies [9]–[10]. The need for a better implement the existing official strategy or the consideration of other control measures that are better suited to the high frequency of infection across all species, the available resources and the structure of the production systems are highlighted by our results [10], [16], [37]. Across this study, taking into account the imperfect performance of the serological tests, the calculated true village flock prevalence was lower than the apparent prevalence and vice versa for the animal prevalence. In addition, we estimated the positive and negative predictive values at the flock level at 72% and 94. 2% respectively (data not shown). Therefore, ignoring the imperfect performance of the serological tests would result in an overestimation of the proportion of infected flocks and an underestimation of the proportion of infected animals. Control programs for brucellosis that are based on the apparent prevalence estimates will result in considering many non infected villages as false positives. In the light of the local dairy processing and marketing practices outlined above, the finding of 38. 4% of milk tanks seropositive against Brucella spp. , suggests that unpasteurized milk and dairy products may be a major source of exposure of the general population to Brucella spp, including people not keeping livestock in their households. These findings should be considered by public health authorities in the study area and highlight the need for coordinated action between public health and veterinary services. Interventions that would effectively reduce the prevalence of ruminant brucellosis in the Nile Delta would benefit not just livestock keepers but the general population. Therefore, a combined strategy for the control of brucellosis designed and implemented in collaboration by veterinary and public health authorities would be justified and could result in a better allocation of resources [38]. Finally, this study shows that the distribution of brucellosis among different ruminant species within the Kafr El Sheikh governorate is spatially heterogeneous, with clustering of the infection around the capital of the governorate and the main animal markets. The finding of higher seroprevalence towards the south of the study area may be associated to higher livestock density compared to the northern part of the governorate (more dependent on fishing) and to the proximity to the largest animal market in the Nile Delta region in the Gharbia governorate. The spatial clustering of infection suggests that there may be potential for the prioritization of control activities in certain areas. By applying different control measures at specific locations it may be possible to maximize public health benefits and to minimize spread of the infection to areas with lower seroprevalence [39]. A recent FAO/WHO report on Brucella melitensis in Eurasia and the Middle East proposes zoning/compartmentalization within a country as one of the generic disease control measures that could be applicable to the control of Brucella melitensis [37]. Such a control strategy was one of the elements of the program successfully applied for the eradication of brucellosis in Chile [40]. For compartmentalization to be effective it has to be accompanied by a biosecurity border that could be difficult to implement in Egypt given the intensity of unregulated animal movements [5]. However, consideration should be given to this approach and others that may be more realistic than achieving elimination by testing a limited fraction of the population with slaughtering of seropositive reactors in the absence of vaccination, which is the strategy currently in place in the area [5], [10]. The results here presented are highly compatible with an intensity of infection transmission within livestock higher than in any other ruminant population studied in Egypt and nearby Middle Eastern countries. Our reference population was restricted to only one of the five governorates of the Nile Delta, mainly because of the availability of relatively detailed information concerning the implementation of brucellosis control activities in this specific governorate in previous years. However, husbandry practices are similar across the entire Nile delta region and thus the situation in neighboring governorates is not likely to differ considerably. Similar surveys in other parts of the country or a survey with nationwide coverage could be a worthwhile investment to provide the basis for the redesign and implementation of control strategies that are more appropriate to the baseline level of infection, structure of the production systems and availability of resources. The sampling strategy presented in this paper and some of our results including seroprevalence estimates by species, test performance indicators and values and intracluster correlation may prove useful in the design of such surveys. Our experience here presented suggests that even relatively small surveys based on inexpensive diagnostic strategies such as bulk tank milk testing for antibodies may provide enough evidence to justify changes in the existing control strategies. In the light of the results here reported and other concordant published evidence, we recommend that serious consideration should be given to an integrated human-animal brucellosis control program in the Nile delta region and that surveys aimed at estimating the frequency of ruminant brucellosis are carried out in other parts of the country such as Upper Egypt and the dessert governorates.

Brucellosis is a zoonosis of mammals caused by bacteria of the genus Brucella. It is responsible for a vast global burden imposed on human health through disability and on animal productivity. In humans brucellosis causes a range of flu-like symptoms and chronic debilitating illness. In livestock brucellosis causes economic losses as a result of abortion, infertility and decreased milk production. The main routes for human infection are consumption of contaminated dairy products and contact with infected ruminants. The control of brucellosis in humans depends on its control in ruminants, for which accurate estimates of the frequency of infection are very useful, especially in areas with no previous frequency estimates. We studied the seroprevalence of brucellosis and its geographic distribution among domestic ruminants in one governorate of the Nile Delta region, Egypt. In the study area, the seroprevalence of ruminant brucellosis is very high and has probably increased considerably since the early 1990s. The disease is widespread but more concentrated around major animal markets. These findings question the efficacy of the control strategy in place and highlight the high infection risk for the animal and human populations of the area and the urgent need for an improved control strategy.

Abstract Introduction Materials and Methods Results Discussion

infectious diseases/epidemiology and control of infectious diseases

2011

Ruminant Brucellosis in the Kafr El Sheikh Governorate of the Nile Delta, Egypt: Prevalence of a Neglected Zoonosis

Understanding the prevalence of sexual reproduction in eukaryotes is a hard problem. At least two aspects still defy a fully satisfactory explanation, the functional significance of genetic recombination and the great variation among taxa in the relative lengths of the haploid and diploid phases in the sexual cycle. We have performed an experimental study to explore the specific advantages of haploidy or diploidy in the fungus Aspergillus nidulans. Comparing the rate of adaptation to a novel environment between haploid and isogenic diploid strains over 3,000 mitotic generations, we demonstrate that diploid strains, which during the experiment have reverted to haploidy following parasexual recombination, reach the highest fitness. This is due to the accumulation of recessive deleterious mutations in diploid nuclei, some of which show their combined beneficial effect in haploid recombinants. Our findings show the adaptive significance of mitotic recombination combined with flexibility in the timing of ploidy level transition if sign epistasis is an important determinant of fitness. Sexual cycles involve an alternation between a haploid and a diploid phase. The relative duration of both ploidy phases may differ strikingly among taxa. Extremes are, on the one hand, diploid multicellular animals and plants with haploidy restricted to the gamete stage, and on the other hand, haploid algae and fungi with diploidy restricted to the zygote stage. The great variation in predominance of the diploid state in eukaryote life cycles is not fully understood, despite a long history of theoretical [1–4] and experimental [5–9] investigation. Possible evolutionary advantages of diploidy include the masking of deleterious recessive mutations [2,4] and faster adaptation [3,8], but a higher rate of adaptation in haploids has also been found in several studies with yeast [6,9]. The mycelial fungus and genetic model organism A. nidulans allows for facile comparisons between the advantages of haploidy and diploidy. Along with the sexual cycle it has, like many other fungi, a “parasexual cycle” [10] (Figure 1). In growing mycelia, haploid nuclei may fuse with a probability of 10−6 to form relatively stable vegetative diploid nuclei. Diploid nuclei can spontaneously produce diploid recombinants by mitotic crossing-over and haploid recombinants by repeated loss of whole chromosomes with a probability of 10−3 per mitosis [11–13]. Therefore, the ploidy level of the nuclei is polymorphic in mycelia of intermediate-to-large size. At any point in time a small fraction of its nuclei will exist in the diploid state and act as a source of recombinant haploid nuclei. If a diploid nucleus is heterozygous at some loci (because of mutations or when the mycelium is a heterokaryon), a haploid recombinant may contain novel allelic combinations. If adaptation involves “sign-epistasis”—defined as an interaction between mutations that are individually neutral or deleterious but advantageous when combined—for which there is growing empirical support [14], the vegetative diploid stage may act as an accumulator of recessive mutations of this type. Specific advantageous combinations of mutations may then appear in haploid nuclei produced by parasexual recombination. Such successful haploid segregants may then get fixed in the mycelium as the sector containing them outgrows the rest of the colony. We hypothesize that the flexibility of switching between diploidy and haploidy within the vegetative organism, as provided by the parasexual cycle, may allow particularly fast adaptation if sign epistasis is involved. We tested this hypothesis in an experimental study of adaptation over 3,000 mitotic generations by comparing adaptation in 15 haploid and 20 homozygous diploid strains of A. nidulans. At the start of the experiment, all strains were genetically identical except for ploidy and three neutral genetic markers. Propagation in our experiment was exclusively vegetative by inoculating a small part of the growing front of a colony with the highest mycelial growth rate (MGR) six days after incubation onto fresh medium. Fitness is therefore defined as the MGR of a colony on the surface of a solid medium [15,16]. The strains used carry a resistance mutation to the fungicide fludioxonil causing a 50% reduction in MGR in the absence of the fungicide when compared to wild-type strains. We studied adaptive recovery during growth on fungicide-free medium. Figure 2A and 2B show the fitness trajectories of all evolving strains per ploidy level. The rate of adaptation of individual populations is estimated by the slope of the fitness trajectory; the mean of these fitness trajectories is shown in Figure 2C. Fitness improvement was caused by compensatory mutations rather than by reversion to fungicide sensitivity, since all strains retained their resistance during the evolution experiment. Haploid strains show a fast response; most variation among the 15 evolving strains is reached within the first 1,500 mitotic generations. Change in fitness occurs gradually and, when compared to the diploid strains, in relatively small steps. By using benomyl and by measuring spore diameter (see Materials and Methods), we found that all strains that began as a haploid remained haploid during the 3,000 mitotic generations. Diploid strains show a delayed response; most variation among the 20 evolving strains begins to arise after 1,500 mitotic generations. Change in fitness occurs in relatively large steps as compared to the haploid strains. Of the 20 initially diploid strains, four reverted to haploidy in the course of the experiment. Three of these (marked A, B, and C in Figure 2B) showed an instantaneous and dramatic increase in fitness upon haploidization. In one case, fitness improvement was more gradual and continued after haploidization. In all groups of strains the mean mycelial growth rate increased over the course of the experiment (t-tests, haploids: t14 = 2. 35, p = 0. 034; diploids that remained diploid: t15 = 2. 94, p = 0. 010; haploidized diploids: t3 = 28. 7, p < 0. 0001). The four haploidized strains had a significantly higher mean rate of adaptation (Figure 2C) than the strains remaining haploid or diploid. Strains that started and remained haploid did not differ in mean rate of adaptation from strains that remained diploid (ANOVA, F2,32 = 19. 6, p < 0. 0001 and post–hoc testing using the Tukey-Kramer method, α = 0. 05). We further analyzed the four haploid strains with a diploid history in order to understand their remarkably fast adaptation. First, diploids were constructed combining evolved and ancestral haploid strains [12]. These heterozygous diploids expressed the non-adapted phenotype, demonstrating that the adaptive mutations are recessive. Second, the three haploid recombinants that made the largest jump in MGR upon haploidization (marked A, B, and C in Figure 2B) were crossed with a haploid ancestor. From each cross, the MGR on fungicide-free medium of 40 progeny was measured (Figure 3). The large genetic variance in MGR among the progeny and the absence of a clear segregation pattern into a limited number of phenotypic classes points to the involvement of multiple mutations in the successful spontaneous haploid revertants. The occurrence of progeny with a fitness lower than that of the nonevolved ancestor suggests that some of the mutations that have accumulated in the diploid phase during the experiment are deleterious by themselves. Third, the three diploids that showed a dramatic fitness increase upon haploidization were retrieved from the −80 °C stock 240 generations prior to their spontaneous haploidization, and haploidization was induced [12]. For these diploids, the MGR of five analyzed haploid segregants varied in between that of the diploid strain before spontaneous haploidization and the spontaneous haploid segregant (see Figure 4). This indicates the presence of recessive mutations that were adaptive either alone or in combination, but also indicates that the unique combinations of mutations later found in the haploid recombinants were not present yet. It also confirms our earlier conclusion that the large fitness increase of the haploid recombinants was due to a beneficial combination of multiple mutations. Among the 16 diploid strains that remained diploid during the experiment, we found one strain in which the MGR had increased by a factor of 2. 5 after 600 mitotic generations, reaching about the same level as that of the haploid recombinants (see Figure 2B). After induced haploidization of this evolved diploid strain, all 20 analyzed haploid segregants showed the same high MGR as their diploid progenitor. We performed a dominance test by combining one of these haploid segregants and a nonevolved fungicide resistant haploid strain (WG631). The resulting diploid did not show the same elevated MGR. These observations indicate that this particular diploid strain had become homozygous for a part of the genome (loss of heterozygosity), most likely by mitotic crossing-over or nondisjunction, that carried a recessive adaptive mutation with large effect, or several linked recessive mutations. The diploid strains that show a fitness increase towards the end of the evolution experiment could in a similar way have partially homozygous genomes due to mitotic recombination in the parasexual cycle, allowing expression of (combinations of) recessive adaptive mutations. Our experimental results show that the shuttling between ploidy levels during vegetative growth enhances adaptation. We have demonstrated that evolved populations that started as diploid but reverted to haploidy have a higher rate of adaptation than populations that began and remained haploid or diploid. The haploids with diploid history first benefited from the diploid state, during which recessive mutations could accumulate irrespective of their fitness effect in haploid state. During haploidization, high-fitness recombinants were produced, in all likelihood containing combinations of interacting recessive mutations—otherwise the haploid-at-all-times strains would have adapted equally well or better. The data suggest an important role for sign epistasis [14], because some of the offspring from the crosses between the evolved haploid recombinants and the haploid ancestor had lower fitness than both parents. The occurrence of deleterious mutations is further indicated by the fact that some haploid strains evolved to a lower MGR. We think this is due to genetic drift caused by our transfer regime. In several cases there were no visible superior sectors from which the strain could be transferred, so occasionally mycelium containing a mutation with a deleterious effect on growth may have been transferred. The findings of this study shed new light on the evolutionary role of the parasexual cycle in fungi (Figure 1). After its discovery in A. nidulans by Pontecorvo [17], the usefulness of the parasexual cycle for genetic analysis was quickly recognized, and its role was initially seen as an alternative for sexual recombination [18]. Traditionally, the starting point of the parasexual cycle is considered to be the formation by anastomosis of a heterokaryon from two different haploid mycelia. Clearly, when two different nuclear genotypes occur in a common cytoplasm, recombination by a parasexual process can generate substantial genetic variation and can serve as an alternative for sexual recombination [17]. However, heterokaryons are rare in nature due to the widespread occurrence of somatic incompatibility [19]; this has led to skepticism about the evolutionary role of the parasexual processes [20,21]. We believe that our results justify a resurrection of the view that the parasexual cycle has an important evolutionary role in fungi [18], because we show that in initially homozygous diploids sufficient genetic variation is generated by mutation to make genetic recombination effective. The important evolutionary significance of parasexual (mitotic) recombination is that it allows the organism to combine specific advantages of both ploidy levels at the somatic level. Diploid nuclei may accumulate recessive mutations that can be recombined and tested in haploid nuclei. In this way the vegetative organism may undergo genetic adaptation, resulting in the production of better-adapted spores. Finally, we believe that the relevance of our findings is not restricted to ascomycetous fungi, but applies more generally to genetic systems that are characterized by alternation of extended haploid and diploid somatic growth, such as in basidiomycetes, including yeast, and in many algae and mosses. The parasexual cycle occurs naturally in fungi but also in distantly related oomycetes [22] and there are indications of very similar processes in human pathogens such as Cryptococcus neoformans [23] and Candida albicans [24]. In all these systems recessive mutations may accumulate in the diploid phase followed by the segregation and selection of successful haploid recombinants that may clonally spread. Even more generally, parasexual recombination at the somatic level appears to be one of the mechanisms by which competition between cells within a multicellular individual may have evolutionary significance. Within an individual soma, cells (or nuclei) may differ at the genetic level due to the switching between haploidy and diploidy or to mitotic recombination causing loss of heterozygosity [25] or at the phenotypic level due to, for example, epigenetic modification or cell-cycle position [26]. In all these cases, within-individual competition between variants may result in differential clonal outgrowth, affecting the fitness of the organism as a whole and so indirectly its reproductive success. The A. nidulans strains used in this study were isogenic and derived from the original Glasgow strain collection [11]. From strain WG562 (lysB5), a spontaneous mutant resistant to the fungicide fludioxonil (Novartis; 0. 2 ppm) was isolated (WG561; lysB5; fldA1: resistance to fludioxonil) [16]. Neutral genetic markers were introduced into the resistant mutant WG561 to construct WG615 (wA3; fldA1; pyroA4); WG561 and WG615 were used to construct the fungicide-resistant diploid strain WG561//615. WG631 (yA2; proA2; fldA1) was used in a cross to analyze the number of adaptive mutations and to assess dominance of adaptive mutations. Strains were cultured on solid Minimal Medium (MM) [11] supplemented with lysine (2. 0 mmol/l), pyridoxin (0. 1 mg/l) and proline (2. 0 mmol/l) where needed. Strains were always incubated at 37 °C. Whether the evolved populations retained their resistance to fludioxonil was assayed by comparing the MGR of evolved and nonevolved populations on MM with fludioxonil (Novartis; 0. 2 ppm). Haploidization of diploids was induced [12] using 1. 7 ppm of benomyl in Complete Medium [11]. Diploids (before and after adaptation) were distinguished from haploids and the moment at which spontaneous haploidization had occurred was assessed by using Complete Medium with benomyl and by using a Coulter counter to measure the diameter of asexual spores, diploids having larger spores than haploids [11]. Crosses were performed as described by Pontecorvo et al. [10]. We defined fitness as the MGR of fungal mycelium [12,15]. After 6 d of growth, we determined the MGR by averaging the colony diameters as measured in two randomly chosen perpendicular directions. The MGR was expressed in mm/d or made relative to the MGR of the nonevolved haploid (WG561 and WG651) or diploid ancestor (WG561//651) that founded the evolution experiment. Due to physiological differences between haploids and diploids, the fitness in terms of MGR of isogenic haploid and diploid counterparts is not identical; the ancestral haploid strain grows 33 mm in 6 d, the diploid counterpart 27 mm in the same amount of time. To check whether the differences found between adaptation in haploid and diploid strains are affected by these physiological differences, we also scaled the rates of adaptation during the evolution experiment by expressing the MGR relative to the initial value of the ancestral haploid or diploid strain (unpublished data). The MGR of diploid strains was expressed relative to the haploid ancestor after the haploidization event. In this comparison, the difference found between the four haploidized strains and the haploid-at-all-times and diploid-at-all-times strains remains highly significant (t-tests; p < 0. 0001). The MGR of WG561 and WG615 is not different (t-test, t8 = 0. 367; p = 0. 72). For this experiment, eight haploid strains were founded from WG561, seven haploid strains from WG651, and twenty diploid strains from WG561//615. All strains used carry a resistance to the fungicide fludioxonil (fldA1), resulting in a lowered fitness when compared to wild-type lab strains growing on funigcide-free medium, due to costs of around 50% associated with the resistance [16,27]. Adaptive recovery on solid medium without fungicide is measured by monitoring the MGR on the surface of a Petri dish. All strains evolved independently. After every 6 d of incubation at 37 °C and 1 d at 4 °C the part of the growing front with the highest MGR was identified. From here, a small piece of mycelium containing between 10,000 and 40,000 (nearly) genetically identical nuclei was transferred to fresh medium. The total experiment comprised 25 transfers, with about 120 mitotic generations between each transfer (estimation based both on nuclear division time and on the position of nuclei in the mycelium combined with growth characteristics of the fungus) [16,28]. At every transfer, samples of all populations were stored in a nonevolving state (at −80 °C). After 25 transfers the MGR was measured under standardized conditions of samples from the frozen stocks from every fifth transfer of all strains with three replicates for each time point. The rate of adaptation was computed as the slope of the fitness trajectory [9]. Every five transfers, the ploidy of all diploid-derived strains was assessed.

Sexual reproduction involves an alternation of ploidy. Haploid gametes, carrying a single set of chromosomes, fuse to form a diploid zygote with a double set of chromosomes. The gametes are formed from diploid progenitor cells by meiosis, which involves genetic recombination—the key evolutionary aspect of sexual reproduction. In this paper we show that in the fungus A. nidulans, during somatic growth, mitotic recombination occurs at a sufficiently high rate to allow an acceleration of the adaptation to novel environmental conditions. Because fungi (unlike animals) lack a clear soma-germline distinction, nuclei with a novel recombinant genotype in the somatic tissue (the mycelium) can give rise to progeny in the form of asexual spores. The results show that recombination at the somatic level (so-called parasexual recombination) appears to be of evolutionary relevance. This finding recalls a suggestion that was made 50 years ago by Pontecorvo, but was discredited soon afterwards.

Abstract Introduction Results Discussion Materials and Methods

oncology ecology yeast and fungi microbiology evolutionary biology genetics and genomics

2007

Mitotic Recombination Accelerates Adaptation in the Fungus Aspergillus nidulans