Independent Expansion of Zincin Metalloproteinases in Onygenales Fungi May Be Associated with Their Pathogenicity

To get a comprehensive view of fungal M35 family (deuterolysin) and M36 family (fungalysin) genes, we conducted genome-wide investigations and phylogenetic analyses of genes in these two families from 50 sequenced Ascomycota fungi with different life styles. Large variations in the number of M35 family and M36 family genes were found among different fungal genomes, indicating that these two gene families have been highly dynamic through fungal evolution. Moreover, we found obvious expansions of Meps in two families of Onygenales: Onygenaceae and Arthodermataceae, whereas species in family Ajellomycetace did not show expansion of these genes. The strikingly different gene duplication and loss patterns in Onygenales may be associated with the different pathogenicity of these species. Interestingly, likelihood ratio tests (LRT) of both M35 family and M36 family genes suggested that several branches leading to the duplicated genes in dermatophytic and Coccidioides fungi had signatures of positive selection, indicating that the duplicated Mep genes have likely diverged functionally to play important roles during the evolution of pathogenicity of dermatophytic and Coccidioides fungi. The potentially positively selected residues discovered by our analysis may have contributed to the development of new physiological functions of the duplicated Mep genes in dermatophytic fungi and Coccidioides species. Our study adds to the current knowledge of the evolution of Meps in fungi and also establishes a theoretical foundation for future experimental investigations.


Introduction
Many proteolytic proteases, such as extracellular serine proteases, aspartic proteases, and metalloproteases (Meps) are known to be important virulence factors related to fungal pathogenicity [1,2]. Metalloproteinases, in which zinc is an essential metal ion for the catalytic activity, have been identified to play important roles in the pathogenicity of pathogenic fungi with different life styles, including vertebrate pathogens, entomopathogens and phytopathogens. For example, a 43.5-kDa Mep was identified from the dermatophytic fungus Microsporum canis and it could digest keratin azure during infection of cats and guinea pigs [3]. A thermolysin-like metalloproteinase produced by the entomopathogenic fungus Metarhizium anisopliae has been identified to not only cause insect cell and tissue damage [4], but also facilitate the development of entomopathogenic fungi within the infected insect host [5]. Moreover, a metalloprotease from the rice blast fungus Magnaporthe grisea was shown to function as a gene-forgene-specific avirulence factor by directly binding the plant resistance gene product and triggering a signaling cascade leading to resistance [6]. These studies suggest that Meps have developed distinct functional properties to help fungi adapt to different ecological niches. Their functional diversities in different fungi raise interesting questions regarding the evolution of these important genes.
The Meps identified so far as secreted by pathogenic fungi mainly belong to two different families: the deuterolysin (M35) and the fungalysin (M36) [2]. These two families of proteases share little similarity with each other, with the exception of the HEXXH motif, in which two histidine residues function as the first and the second zinc ligand, respectively [7]. M35 family genes harbor a third zinc ligand known as an Asp that is found in a GTXDXXYG motif at the C-terminal, whereas M36 family genes utilize a Glu, which is present at the EXXXD motif at the C-terminal as the third zinc ligand [7,8]. Previous evidence suggested that members of these two families are actual pathogenic factors in human infectious diseases. For example, Aspergillus fumigatus, an opportunistic human pathogen, can release a M36 family metalloprotease to help it invade mammalian lung [9]. In addition, at least ten Mep genes (designated as Mep1 to Mep10) have been found in Coccidioides posadasii, which is a fungal respiratory pathogen of humans that can cause coccidioidomycosis (Valley fever) in immunocompromised individuals, and most of them were classified into M35 and M36 families. Moreover, the dermatophytic fungi, which cause various dermatophytosis common in animals and humans, can secrete several M36 family proteases with collagenolytic, elastinolytic and keratinolytic activities to help them survive in various niches [10]. Therefore, although the precise roles of these proteases from other phytopathogenic or entomopathogenic fungi have not yet been identified, we believed that the M35 and M36 family genes in different pathogenic fungi may have a wide variety of pathological actions. Thus, characterizing the evolution of M35 family and M36 family genes from different fungi can be of great importance for understanding the diversification of these Mep genes and for inferring their biological significance. Moreover, the increasing availability of completely sequenced fungal genomes also provides a good opportunity for us to characterize the Meps superfamily in different fungal taxa and to infer their evolutionary patterns.
Onygenales, a fungal order associated with a host/substrate shift from plants to animals, includes many important pathogenic fungi that can cause notorious diseases of vertebrates (e.g. Coccidioides immitis, Coccidioides posadasii, Paracoccidioides brasiliensis, Histoplasma capsulatum, Trichophyton rubrum and Microsporum canis) and also contain a number of non-pathogens (e.g. Uncinocarpus reesii) [11]. Previous studies suggested that both the M35 and M36 family genes have separately expanded in Onygenales [11,12]. Our previous study also provided evidence that positive selection has acted on the duplicated M35 family genes in Coccidioides and promoted their novel functional adaptations (e.g., recent acquisition of virulence traits) of these species [13]. Phylogenetic analyses in this paper suggested that M35 and M36 family genes have undergone intriguing evolution in Onygenales fungi. To better understand the evolutionary force that have likely shaped the diversification of these duplicated M35 and M36 family genes in Onygenales species and the implications of such changes, we also investigated the possible selection pressures responsible for the duplicated genes in Onygenales species.

Data sets
In order to understand the evolutionary history and functional divergence of the M35 and M36 family genes in fungi, we performed a comprehensive evolutionary analysis of these two family genes from the sequenced genomes of 50 Ascomycota fungi. These fungi have different life styles and they include saprophytes, opportunistic human pathogens, plant pathogens and entomopathogens. The genome sequences and their corresponding annotated proteins database of the 50 Ascomycota fungi are available from the Fungal Genome Initiative at the BROAD Institute (http://www.broad.mit.edu/annotation/fungi/fgi) or Fungal Genome Research (http://fungalgenomes.org/).
To avoid search bias, we applied two methods to find putative homologs of M35 family sequences and M36 family sequences from the fungal genome databases employed in this study. In the first method, the M35 family genes (Mep2 to Mep8) and M36 family genes (Mep9 and Mep10) from C. posadasii were used as query sequences to do TBLASTN and BLASTP searches with Evalue cutoff of 10 210 against the genome nucleotide database and fungal protein database, respectively. The sequences that met the criteria of both searching procedures and with amino acid sequence identities greater than 30% and covering more than 150 amino acids were considered as candidate genes. In the second method, the HMM profile Peptidase_M35 (PF02102; http://pfam.sanger.ac.uk/family/PF02102#tabview = tab6) and Peptidase_M36 (PF02128; http://pfam.sanger.ac.uk/family/ PF02128#tabview = tab6) were downloaded and used as queries for homologous protein searches using the program HMMSEARCH from the HMMER package (http://hmmer. wustl.edu/). For each search result, hits were considered significant when they matched the Pfam HMM profile with E values ,10 25 . In total, 105 M35 family genes and 58 M36 family genes were identified, including two M35 family genes and four M36 family genes from the Basidiomycete fungus Coprinopsis cinerea which were used as outgroups for these two families, respectively. The taxonomic information and the identified gene number of the fungal species are presented in Table 1.

Phylogenetic analysis
For each of the two families of genes, MUSCLE v3.5 was used to generate protein alignment with default settings [14]. The ambiguous areas of alignments were located and removed by using the program Gblocks 0.91b [15,16] with default parameters. The gap selection criterion ''with half'' was used here.
Phylogenetic analyses of the alignments were performed using PHYML 3.0 [17] for Maximum likelihood (ML) analysis and MrBayes 3.1.2 [18] for Bayesian inference. In the ML analysis, the best-fitting model of sequence evolution were estimated for both M35 family and M36 family genes using program ProtTest [19]. The chosen model WAG+I+G and their parameters (I = 0.03, G = 1.912 for M35 family genes; I = 0.038, G = 1.149 for M36 family genes) were used in the ML analysis. The reliability of the tree topology was evaluated using bootstrap support [20] with 100 for ML analysis. In addition, Bayesian inference was conducted using MrBayes 3.1.2 [18]. The best-fitting model of sequence evolution determined by ProtTest [19] was also used as the priors of Bayesian inference. Bayesian analysis started with randomly generated trees and Metropolis-coupled Markov chain Monte Carlo (MCMC) analyses were run for 2610 6 generations. The run was stopped when the average standard deviation of split frequencies was less than 0.01 in all the cases (MrBayes 3.1.2 manual). To ensure that these analyses were not trapped in local optima, the dataset was run three times independently. We determined the burn-in period by checking for likelihood stability. A 50% majority rule consensus of post burn-in trees was constructed to summarize posterior probabilities (PPs) for each branch.

Gene duplication and loss analyses
To investigate the gene duplication and loss scenarios of M35 and M36 family genes in Onygenales, we used Notung 2.6 [21,22] to reconcile the species tree and gene tree. The species tree was inferred based on a combined alignment of six genes from each Onygenales species as those used in James et al [23], including 18S rRNA, 28S rRNA, ITS RNA, translation elongation factor 1-a (TEF1a), RNA polymerase II largest subunit (RPB1) and RNA polymerase II second largest subunit (RPB2) (9601-bp in total, data not shown). Phylogenetic analyses of the alignments were performed as described by Li et al [13] and the model TrN+I+G of sequence evolution was optimized using Akaike information criterion [24,25] as implemented in Modeltest version 3.7 [26]. For the gene tree used here, we collapsed those inconsistent nodes produced by different tree-building methods, which were also poorly supported, into polytomies [27].
In addition, we investigated the gene duplication and loss scenarios of M35 genes in Eurotiales fungi. The species tree and gene tree of Eurotiales fungi were inferred by using the same methods mentioned above.

Selective pressure analyses
To investigate the possible selective forces behind M35 and M36 family gene duplications in Onygenales, we conducted likelihood ratio tests (LRT) for these two families respectively. From the phylogeny shown in Figure 1, we can see that multiple gene duplications of the M35 family genes have occurred in Eurotiales, and some of the duplication events occurred after the divergence of Eurotiales and Onygenales. To avoid the interference of within-Eurotiales duplications on our analyses, we used more distantly related Sordariomycetes fungi as outgroup to conduct LRTs analysis of M35 family genes. For M36 family genes, we still used Eurotiales fungi as outgroup. Finally, a 861-bp codon-based alignment for 62 M35 family genes (Fig. S1) and a 1569-bp codon-based alignment for 38 M36 family genes ( Fig. S2) were used to construct phylogenetic trees using the same methods mentioned above except that the best-fitting model of sequence evolution used for ML and Bayesian analyses were estimated using program Modeltest 3.06 [26]. Because the likelihood analysis might be sensitive to tree topology used, we collapsed the nodes that showed inconsistent branching patterns from different treebuilding methods and with poor statistical support into polytomies [27]. The ratio v (dN/dS) is the ratio of the number of nonsynonymous substitutions per non-synonymous site (dN) to the number of synonymous substitutions per synonymous site (dS), which provides an indication of the change in selective pressures [28]. A dN/dS ratio = 1, ,1, and .1 are indicative of neutral evolution, purifying selection, and positive selection on the protein involved, respectively [29,30]. The codon substitution models implemented in the CODEML program in the PAML 4.4b package [31] were used to analyze changes of selective pressure. All models were corrected for transition/transversion rate and codon usage biases (F364). The ambiguous sites were removed in PAML analysis (clean data = yes). Different starting v values were also used to avoid the local optima on the likelihood surface [32].
Two branch-specific models, i.e., ''one-ratio'' (M0) and ''freeratios'', were compared. The M0 model assumes the same v ratio for all branches while the ''free-ratios'' model assumes an independent v ratio for each branch [33]. We constructed LRT to compare the two models. Significant differences between models were evaluated by calculating twice the log-likelihood difference following a x 2 distribution, with the number of degrees of freedom equal to the difference in the numbers of free parameters between models.
Site-specific models, which allow for variable selection patterns among amino acid sites, M1a, M2a, M7, and M8, were used to test for the presence of sites under positive selection. The M2a and M8 models identify positively selected sites. When these 2 positiveselection models fitted the data significantly better than the corresponding null models (M1a and M7), the presence of sites with v.1 would be suggested. The conservative Empirical Bayes approach [34] was then used to calculate the posterior probabilities of a specific codon site and identify those most likely to be under positive selection.
Considering that positive selection may act in very short episodes during the evolution of a protein [35] and affect only a few sites along a few lineages in the phylogeny, the likelihood models accommodating v ratios to vary both among lineages of interest and among amino acid sites, as an improved version of the ''branch-site'' model [36], were also considered here. We used branch-site Model A as a stringency test (test 2) and identified amino acid sites under positive selection by an empirical Bayes approach along the lineages of interest [36,37]. The log-likelihoods for the null and alternative models were used to calculate a likelihood ratio test statistic, which was then compared against the x2 distribution (with a critical value of 3.84 at a 5% significance level) [31]. We used the conservative Bayes Empirical Bayes (BEB) approach, which assigns a prior to the model parameters and integrates over their uncertainties, to calculate the posterior probabilities of a specific codon site and to identify those most likely to be under positive selection. In addition, the Bonferroni correction [38,39] was also applied for correcting multiple testing in the analysis according to the number of tests of significance performed.

Results
Taxonomic distribution of M35 and M36 gene families As shown in Table 1, large variations of the gene numbers for both M35 and M36 family genes were observed from the analyzed Ascomycota species, suggesting these two gene families have been highly dynamic through fungal evolution.
For the M35 family, a total of 103 genes were identified from the analyzed Ascomycota fungal species. In Family Onygenaceae of the Onygenales Order, 7 genes were identified from the human pathogenic fungi C. posadasii and C. immitis, and 4 genes were predicted from the saprophytic fungus U. reesii. In Family Arthodermataceae of Onygenales, five genes each were identified from the five dermatophytic fungi (M. canis, Microsporum gypseum, Trichophyton equinum, T. rubrum and Trichophyton tonsurans). These fungi can cause dermatosis in humans and animals. In Family Ajellomycetace of Onygenales, only one M35 family gene each was predicted from H. capsulatum and P. brasiliensis, and two genes were identified from Blastomyces dermatitidis. This result suggested that M35 family genes have expanded in two families of Onygenales: Onygenaceae and Arthodermataceae. The genus Aspergillus, which includes species with very different life styles with some very harmful and others beneficial to humans, showed the most variations of gene numbers (range from 0 to 6). For other fungal species, most contained 1 to 4 M35 family genes. However, no M35 family gene was identified from several fungal groups (e.g. Saccharomycotina fungi, Taphrinomycotina fungi, two endophytic fungi and some saprophytic fungi) ( Table 1).
For the M36 family, a total of 54 genes were identified from the analyzed Ascomycota fungal genome sequences. In Onygenales fungi, the dermatophytic fungi (Arthodermataceae) contained the most with 4-5 genes per species, while 2 genes each were identified from Onygenaceae fungi (C. posadasii, C. immitis and U. reesii) and no M36 family gene was identified from Arthodermataceae fungi (H. capsulatum, P. brasiliensis and B. dermatitidis), suggesting that M36 family genes might have expanded in dermatophytic fungi. For other fungal species, 1 or 2 M36 family genes were predicated. However, several fungal groups (including Saccharomycotina fungi, Taphrinomycotina fungi, two endophytic fungi, three Leotiomycetes fungi, M. anisopliae, Neurospora crassa and Podospora anserine) contain no obvious M36 family gene homologs (Table 1). These results suggest that the M36 family genes may have undergone many gene duplication and gene loss events during the evolution of Ascomycota fungi.

Phylogenetic analysis of M35 family
Phylogenetic relationships among the M35 family genes were analyzed based on an alignment consisting of 185 amino acids from 105 M35 family genes (include 103 genes from Ascomycota fungi and 2 outgroup genes from the Basidiomycota fungus C. cinerea; Fig. S3) using maximum-likelihood and Bayesian techniques. As seen in Fig. 1 and Fig. S4, phylogenetic analyses consistently grouped the M35 family genes into 12 distinct clades, which are abbreviated as M35-a to M35-l. Among these clades, M35-a (BS = 99%; PP = 100%) is a dermatophyte-specific lineage, containing exclusively the M35 genes from five dermatophytic fungi with Microsporum spp. branching basally to Trichophyton spp. It seemed that the multiplication of these genes occurred before the divergence of these five species. This is because each subclade within this clade contained one gene from each species, suggesting that these genes had a conserved and essential function in dermatophytes. The genes from Onygenaceae fungi fell into three clades (M35-b, M35-c and M35-d). Clades M35-c (BS and PP = 100%) and M35-d (BS and PP = 100%) were comprised of sequences from C. posadasii, C. immitis and U. reesii, while M35-b (BS = 88%; PP = 98%) was a Coccidioides-specific lineage, consisted of three duplicated genes from each of the two Coccidioides species. The genes from Ajellomycetaceae fell into clades M35-e (BS and PP = 100%) and M35-f (BS and PP = 100%). Clade M35-f, which contained genes from B. dermatitidis and P. brasiliensis, was closely related to Onygenaceae lineage M35-d. In contrast, clade M35-e, which included one gene from H. capsulatum and another from B. dermatitidis, showed close relationships with Eurotiales lineage M35-g (BS = 92%; PP = 100%).
From our phylogenetic analysis, we found that except for Arthodermataceae and Onygenaceae, the M35 family genes from other Ascomycota fungi within the same categories were consistently clustered into two clades respectively. For example, the genes from Ajellomycetaceae fell into M35-e and M35-f clades and those from Eurotiales were clustered into M35-g and M35-h clades (BS = 89%; PP = 86%). Clades M35-i (BS and PP = 100%) and M35-l (BS = 67%; PP = 70%) were comprised of genes from Sordariomycetes while clades M35-j (BS and PP = 100%) and M3k (BS and PP = 100%) mainly consisted of those from Leotiomycetes fungi. Noticeably, one gene from Neurospora crassa showed a close relationship with clade M35-j. Interestingly, although the genes from the same taxonomic groups of fungi were consistently clustered into two clades, gene duplication and loss events might have also occurred during their evolution, leading to the diversification of M35 family genes in different fungi. For instance, though clades M35-i and M35-l both contained genes from Sordariomycetes, clade M35-l contained more genes than clade M35-i.

Phylogenetic analysis of M36 family
Results from our phylogenetic analyses based on an alignment consisting of 538 amino acids from 58 M36 family sequences (include 54 genes from Ascomycota fungi and 4 outgroup genes from the Basidiomycota fungus C. cinerea, Fig. S5) were largely consistent with the taxonomical relationships among the major fungal groups. The results suggested that the Dothideomycetes showed close relationships with Sordariomycetes and that the Eurotiales species were closely related to Onygenales fungi. Phylogenetic analyses based on the M36 family genes consistently showed eight strongly-supported monophyletic clades which were designated as M36-1 to M36-8 in the trees (see Figure 2 and

Gene duplication and loss analysis in Onygenales and Eurotiales
Based on the above results, we found that M35 and M36 family genes have undergone expansion in Onygenales fungi. To infer the gene duplication and loss scenarios of M35 and M36 family genes in Onygenales, We used Notung 2.6 [21,22] to reconcile the species tree (Fig. S7) and the gene tree of each family. Using various tree-building methods, our inferred species tree consistently showed three strongly supported clades (Fig. S3), consistent with previous reports, which suggested that the fungal family Onygenaceae was more closely related to Arthodermataceae than to Ajellomycetace [13,[40][41][42]. The inferences of gene duplication and loss events in Onygenales species are shown in Figure 3.
For the M35 family genes (Figure 3a), there were at least 10 gene duplication and 5 gene loss events in the Onygenales species. An initial duplication event occurred before the divergence of Onygenales fungi. Subsequently, four and five gene duplications occurred independently in Arthodermataceae and Onygenaceae, respectively. In Onygenaceae, two gene duplications occurred before the divergence of Coccidioides and U. reesii, while three occurred before the divergence of C. posadasii and C. immitis (Figure 3a). For comparison, one M35 gene might have been lost in each of U. reesii, P. brasiliensis and H. capsulatum. Moreover, the lineages of Arthodermataceae might have also suffered two gene loss events after the duplication but before the divergence of the five dermatophytic fungi (Figure 3a).
For the M36 family genes (Figure 3b), there were four gene duplication and four gene loss events in Onygenales species. Before the divergence of Onygenales, there was one initial duplication event. Subsequently, three duplication events occurred in Arthodermataceae. There was one gene loss event in each of the two lineages of Onygenaceae and Arthodermataceae respectively and one gene was also lost in each of T. equinum and T. rubrum.
In our study, we also found large variations of M35 family gene number in Eurotiales, ranging from 0 to 6 ( Table 1). To infer the gene duplication and loss scenarios of M35 genes in this group, we also used Notung 2.6 [21,22] to reconcile the species tree (Fig. S8) and the gene tree of this family. The inferences of gene duplication and loss events in Eurotiales species are shown in Figure 4. We found at least 7 gene duplication and 14 gene loss events in the Eurotiales species. While most of the duplication events occurred before the divergence of Aspergillus spp, the gene loss events in Eurotiales fungi were more dynamic. There were three gene loss events occurred in Aspergillus nidulans and Aspergillus terreus each, and one gene might have been lost in Aspergillus oryzae, Aspergillus clavatus and Penicillium chrysogenum each. Moreover, four gene loss events occurred in the lineage including three species (A. fumigatus, Neosartorya fischeri and A. clavatus) and one gene might have been lost before the divergence of the four species including A. nidulans, A. terreus, A. oryza and Aspergillus flavus. The tree was performed using PHYML 3.0 [17]. The best-fitting model WAG+I+G and their parameters (I = 0.03, G = 1.912) which were estimated by program ProtTest [50] were used in the ML analysis. The reliability of the tree topology was evaluated using bootstrap support [20] with 100. doi:10.1371/journal.pone.0090225.g001 Table 2 shows the evidence of positive selection analyses of M35 family genes in Onygenales. In the branch-specific model analyses [43], the free-ratio model, M1, revealed a significantly better fit to the data than did the one-ratio model, M0 (p,0.001, Table 3), suggesting that these M35 family genes have been the subjects of Figure 2. ML tree based on amino acid sequences of 58 M36 family genes. The tree was performed using PHYML 3.0 [17]. The best-fitting model WAG+I+G and their parameters (I = 0.038, G = 1.149) which were estimated by program ProtTest [50] were used in the ML analysis. The reliability of the tree topology was evaluated using bootstrap support [20] with 100. doi:10.1371/journal.pone.0090225.g002 Independent Expansion of Meps in Onygenales PLOS ONE | www.plosone.org different selective pressures. In the site-specific model analyses, although the LRT of M2a/M1a did not achieve statistical significance (P = 1.000), M8, another positive-selection model provided a significantly better fit to the data than did the neutral model (M7) (P,0.001), suggesting the possibility of positive selection acting on the M35 family genes in Onygenales examined here. When we performed the branch-site model tests for those branches resulted from gene duplications of M35 family in Onygenales (19 branches in total, a-t as indicated in Figure 5), we found that there were five branches (branches a, d, e, f and j) showing signs of positive selection ( Figure 5). After Bonferroni correction for multiple testing, we found that LRT tests were still significant in four of the branches (branch a, e, f and j; p,0.002632) (Table 2, Figure 5). Remarkably, several positively selected residues (A150R/Q, K166A/R/G and M203R/H for branch a, V177Q and Y255Q for branch e, E57T, G79K/Q, E85S/T/L and A99K for branch f, Y63V/I and A84R for branch j) were also identified for these branches with high posterior probabilities (Table 2 and Figure 5). Among these eleven positive selected sites, ten of them (A150R/Q, K166A/R/G, M203R/H, V177Q, E57T, G79K/Q, E85S/T/L, A99K, Y63V/I and A84R) were involved in amino acids substitutions with different physicochemical properties (Table. S1). . Duplication and loss events of M35 family genes and M36 family genes in Onygenales fungi. The reconciliation between species tree and gene tree of Onygenales fungi along with the confirmation of the gene loss/duplication scenario were determined by using Notung 2.6 [21]. The species tree of Onygenales fungi is shown as Fig. S7. Putative duplication events are indicated with solid cycles, while loss events are indicated with thick branches. a, duplication and loss events of M35 family genes. b, duplication and loss events of M36 family genes. doi:10.1371/journal.pone.0090225.g003  Table 3 shows the results of positive selection analyses of M36 family genes in Onygenales. In the branch-specific model analyses [43], the free-ratio model, M1, revealed a significantly better fit to the data than did the one-ratio model, M0 (p,0.001, Table 3), suggesting that these M36 family genes have been the subjects of different selective pressures. Similar as the results in M35 family, only the LRT of M8/M7 achieved statistical significance (P,0.001) in the site-specific model analyses, also suggesting the possibility of positive selection acting on the M36 family genes in Onygenales examined here. Intriguingly, the LRT tests based on the branch-site models for those branches resulted from gene duplications of M36 family in Onygenales (12 branches in total, a-l as indicated in Figure 6) suggested that there was significant evidence along several lineages leading to the duplicated genes in dermatophytic fungi (branch a, b and g). After performing Bonferroni correction for multiple testing, the LRT tests still showed significant results in these branches (p,0.005). In addition, several residues (G183P, Y267V and K401H for branch a, T40E for branch b, A139H, I221A/P, Y232W, L300C and W510M for branch g) were predicted as positively selected sites along these branches (Table 3 and Figure 6). Among these nine positive selected sites, five of them (G183P, Y267V, T40E, A139H, Y232W) were involved in amino acid substitutions with different physico-chemical properties (Table. S1)

Discussions
To get a comprehensive view of fungal M35 family and M36 family genes, we conducted genome-wide investigations and phylogenetic analyses of these two family genes from 50 sequenced Ascomycota fungal genomes with different life styles. Our results provided valuable insights for understanding the evolution of these two gene families in various fungi. In Sordariomycetes, the saprophytic fungi had 0 to 2 Mep genes, whereas the phytopathogenic fungi contained 2 to 6 genes within each species. Phylogenetic analyses suggested that the M35 family genes of Sordariomycetes fungi fell into two clades: M35-i and M35-l. Clade M35-l contained the genes from Sordariomycetes that included saprophytic, phytopathogenic and entomopathogenic fungi, while clade M35-I comprised six duplicated genes from six phytopathogenic fungi each (Figure 1). In addition, the M36 family genes in Sordariomycetes fungi also grouped into two clades: M36-7 and M36-8, with M36-8 containing the duplicated genes from the two phytopathogenic fungi V. albo-atrum and V. dahlia (Figure 2). Similarly, M35 family genes from the three Leotomycetes fungi (Botryotinia fuckeliana, Botrytis cinerea and Sclerotinia sclerotiorum) were also clustered into two clades: M35-j and M35-k, indicating that the genes have duplicated before the divergence of these three phytopathogenic fungi. Previous studies suggested that protease gene family expansion appeared to be an important evolutionary step among the phytopathogenic fungi and likely served important roles in plant infection [44]. We speculate that the duplication of Mep genes occurred in the phytopathogenic fungi belonging to Sordariomycetes and Leotomycetes may be associated with their pathogenicity. However, we did not identify any Mep gene from the two endophytic fungi, Chaetomium globosum and Epichloë festucae, and no M36 family gene was identified from the three Leotomycetes fungi, suggesting that these genes may be lost during the evolution of these fungi.
Based on the homologous search results, large variations of gene number, ranging from 1 to 8 were also observed in Eurotiales (Table 1). Phylogenetic analyses and gene duplication/loss analyses suggested that the Mep genes, especially the M35 family genes have been highly dynamic through the evolution of Eurotiales fungi (Figure 1, 2 and 4). In our study, A. oryzae, which is a saprophytic fungus and plays a key role in the fermentation process of several traditional Japanese beverages and sauces [45], contains 7 Meps in total (5 in M35 family and 2 in M36 family), whereas A. fumigatus, an important opportunistic pathogen of vertebrates with compromised immune systems [46], has only 4 Mep genes in total (3 in M35 family and 1 in M36 family). However, although A. fumigatus contains fewer Mep genes, it has been reported that this fungus can utilize the M36 family metalloprotease to help it invade mammalian lung and degrade its host's proteinaceous structural barriers [9]. Moreover, other metalloproteases, such as Aspf2 from A. fumigatus and Aspnd1 from A. nidulans which are known cell-surface antigens during fungal infections and can bind specific ligands within mammalian hosts, were also previously identified from Aspergillus spp. [47]. Therefore, we believed that the opportunistic human pathogens in Eurotiales might not only use the M35 or M36 family genes as virulence factors, but also choose other families of metalloproteases as pathogenic factors to carry out the infection of immunocompromised individuals. Surprisingly, neither M35 gene nor M36 gene was found in Saccharomycotina and Taphrinomycotina fungi, suggesting that both the two family genes have been lost during the evolution of the yeast-like fungi. However, although no M35 or M36 family proteins have been identified in yeast-like fungi, other types of metalloproteases that serve important roles in the development and pathogenicity of yeast-like fungi have been identified before [48]. For example, a Zps1p metalloproteases which involved in zinc-responsive transcriptional regulation was observed in Saccharomyces cerevisiae [49]. Moreover, a Zps1p-like metalloproteases named Pra1 which function as virulence factors in mediating critical host-parasite interactions were identified from the opportunistic human pathogen Candida albicans that can cause diseases of vertebrates [48]. Sequence alignments suggested that some critical residues such as the cysteine residues involved in disulfide bond formation and a zinc-binding domain were conserved in both Zps1p and M35 family metalloproteases [48]. These results suggest that the yeast-like fungi may utilize other metalloproteases but not M35 or M36 family proteins to carry out their development and pathogenicity.
The most striking result from our study is the significant varieties of Meps in Onygenales fungi. Phylogenetic reconstruction and gene duplication/loss analyses suggest that the first duplication event of both M35 and M36 family genes occurred in the common ancestor of the Onygenales species (Figure 1, 2 and 3).
Subsequently, additional gene duplication events mainly occurred after the divergence of Arthrlodermataceae (dermatophyte) and Onygenaceae species (Figure 1, 2 and 3), leading to a large number of Mep genes shared between Arthrlodermataceae (dermatophyte) and Onygenaceae species. However, the Ajellomycetace species (B. dermatitidis, H. capsulatum and P. brasiliensis) did not show evidence of expansion of these genes (only containing 1 or 2 M35 family genes and no M36 family gene was identified from them). Although the species in Onygenales showed a host/ substrate shift from plants to animals during evolution, their pathogenicity factors on animals might be different. Dermatophytic fungi cause the majority of superficial dermatophytosis in humans and animals [12], while Coccidioides spp. can cause lifethreatening respiratory disease known as coccidioidomycosis (Valley fever) in immunocompromised individuals [50,51]. However, Ajellomycetace species cause geographically widespread systemic mycosis of humans and other mammals [52]. For example, Paracoccidioides can cause paracoccidioidomycosis [40], B. dermatitidis is the causative agent of blastomycosis. However, H. capsulatum is the most common cause of histoplasmosis in the world [53]. Therefore, the distinct expansion of Meps in Arthrlodermataceae and Onygenaceae species suggest that the Meps may have played different roles during the evolution of dermatophytic and Onygenaceae species, providing flexibility for these species to occupy different ecological niches.
Interestingly, LRT analysis in both M35 family and M36 family suggested that the duplicated Meps genes most likely underwent positive selection during the evolution in dermatophytic and Coccidioides species. For dermatophytic fungi, significantly different selective pressures have acted on several lineages leading to both duplicated M35 family (branches a, e and f) ( Figure 5) and M36 family genes (branches a, b and g) ( Figure 6). The results suggested that functional divergences of both M35 family and M36 family genes in dermatophytic fungi most likely occurred after gene duplication and before the speciation of these five dermatophytic species.  [3]. However, the expression of genes encoding M36 fungalysins in A. benhamiae assayed by cDNA microarray analysis suggested that the M36 family genes not only help A. benhamiae digesting keratin, but also facilitate the cutaneous infection of guinea pigs [54]. Both the M35 family and M36 family genes have the capacity to degrade keratin, suggesting that the cooperation between the two family genes may have facilitated the fungi to adapt to various environmental niches. For Onygenaceae fungi, the positive selection pressure might have acted on the ancestral lineage (branch j) ( Figure 5) leading to the three additional duplicated genes specifically in Coccidioides species, consistent with the results of our earlier study which suggested that significant selective pressure on the ancestral lineage of three additional duplicated M35 family genes from Coccidioides species may be associated with recent pathogenesis acquisition of M35 family genes in the Coccidioides species [13]. Noticeably, Figure 6. Phylogenetic tree of 38 M36 family genes used for codon-based maximum likelihood analysis in PAML. Phylogenetic trees were collapsed with inconsistent nodes from different tree-building methods and poor statistical supports into polytomy. Branches a-j indicated putative duplication events in Onygenales fungi. The branches with significant evidence of positive selection are indicated as a thick branch. The putative positively selected residues along these branches were shaded in grey. doi:10.1371/journal.pone.0090225.g006 although the M36 family genes from Coccidioides species and U. reesii were duplicated before the split of Arthrlodermataceae and Onygenaceae species, positive selection did not seem to have acted on the duplicated M36 family genes in Onygenaceae species. These results suggest that the M36 family genes in Coccidioides species and U. reesii are unlikely important functional genes during the evolution of these fungi. Of course, we can't exclude the possibility that these M36 family genes in Coccidioides and U. reesii may serve other yet unknown physiological functions.
In our study, eleven and nine amino acid substitutions were predicted as positively selected sites respectively for the M35 and M36 family genes in Onygenales fungi (Table 2 and 3, Figure 5 and 6). Most of the positively selected sites had substitutions of amino acids with different physical-chemically properties, such as differences in polarity, charge and/or hydrophobicity (Table. S1). It has been recognized that substitutions with chemically very different amino acids often have significant effects on the overall flexibility, structure stability and/or activity of proteins [55,56]. Therefore, although the explicit functional changes of the Meps brought by these positive selected residues identified in our study cannot be predicted at present, we propose that these residues may have allowed the development of new physiological functions of the duplicated Mep genes, providing valuable information for functional adaption of the duplicated Meps in dermatophytic fungi and Coccidioides species. Experimental investigations are needed in order test the functional effects of the potentially adaptive amino acid replacements discovered by our analysis.

Conclusions
In summary, our study analyzed the phylogenetic relationships among genes in both M35 and M36 families from Ascomycota fungi. Our results increase the current knowledge of the evolution of Meps in fungi and support the possibility that crucial physiological functions of both duplicated M35 family and M36 family genes may have been developed during the evolution of fungi. Furthermore, our detection of significant positive selections in several lineages of dermatophytes and Coccidioides species indicates that rapid functional divergence of duplicated Meps have most likely occurred after gene duplications in these fungi. These species-specific adaptations could be important in host immune system interaction or in their survival in the environment. The hypotheses proposed in our analysis will provide valuable information for functional analysis of these genes in future studies.