Fish collection

A total of 233 specimens of Parachondrostoma toxostoma, Chondrostoma nasus and their respective hybrids were collected in Southern France in August 2008, June 2010, September 2010, and June 2011. One allopatric population of P. toxostoma from the Orbieu River (Mediterranean Coastal River) and one allopatric population of C. nasus from the Allier River (Loire drainage) were sampled. In addition, four localities situated in the Durance hybrid zone (Avignon, Manosque, and Pertuis on the Durance River, and Pont de Laragne on the Buech River, a tributary of the Durance River) and two localities situated in the Ardeche hybrid zone (Saint Just and Labeaume) were sampled. The Directions Départementales des Territoires (DDT) from Alpes-de-Haute-Provence, Hautes-Alpes, Vaucluse, (N° 2010-961 and N° 2011-133), Ardèche (N°2010-253-8), Puy-de-Domes (N°2008-01-04) and Aude (N° 2003-3555 and ONEMA11-2008) districts issued the permits for fish catching in France, where fishing was done in collaboration with Officers of the French ONEMA (Office National de l'Eau et des Milieux Aquatiques). A fin sample of each fish specimen was preserved in 96% ethanol. The spleen of each individual was removed, transferred into 1.5 ml tubes with RNAlater™ Storage Solution (Sigma-Aldrich) and stored at −80°C. This study was approved by the Animal Care and Use Committee at the Faculty of Science, Masaryk University in Brno (Czech Republic).

Molecular identification

All individuals were identified using the mitochondrial 5′ part of the cytochrome b gene according to Costedoat et al. [27] and nuclear markers (41 microsatellites distributed in five multiplex PCR kits, [33]) applying the Bayesian clustering method in NewHybrid 1.1 [34], see [32]. For microsatellites, amplification was performed as described in Šimková et al. [32]. Allele sizes were scored against an internal GeneScan-500 LIZ standard and genotypes were obtained using GeneMapper® 3.7 [35].

MHC analyses

RNA extraction and reverse-transcription were performed as described in Ottová et al. [36]. The complete exon 2 of DAB genes (functional genes of MHC class IIB in fish) was analyzed in this study. Exon 2 represents the most variable part of MHC IIB genes including the peptide binding regions (PBR). The primers specific for DAB1-like genes and those specific for DAB3-like genes of P. toxostoma and C. nasus were developed in this study. The exon 2 of DAB1-like genes was amplified using forward primer DAB1_EX2_3F (CTGCTTTCACTGGAGCAGCTA) annealing to position 48 (or alternatively 30) and reverse primer DAB1_EX2_2R (GTCTGCCACCAGCCTGAG) annealing to position 372 (or alternatively 354) of the reference sequence for the DAB1 (or alternatively DAB2) gene in Cyprinus carpio (accession numbers Z49064 and Z49065). The exon 2 of DAB3-like genes was amplified using forward primer DAB3_EX2_1F (GCTTTCACTGGAACAGCTG) annealing to position 33 (or alternatively 40) and reverse primer DAB3_EX2_1R (ACAGCTGGATGATTGCCTT) annealing to position 364 (or alternatively 371) of the reference sequence for the DAB3 (or alternatively DAB4) gene in Cyprinus carpio (accession numbers X95431 and X95434). These pairs of primers amplify the complete exon 2 (276 bp long) and partial exon 1 and exon 3 in both genes (Supporting Information S1).

PCR was performed in a 30 µl reaction volume using 30–50 ng of cDNA. The reaction mixture was comprised of a 1× TaqBuffer with (NH₄)₂SO₄, 1.5 mM MgCl₂, 0.2 mM of each dNTP, 0.3 µM (for DAB1) or 0.35 µM (for DAB3) of each primer of gene specific primer pair, and 1 U (for DAB1) or 0.75 U (for DAB3) of Taq DNA polymerase (Fermentas). The amplification of DAB1 ran in 35 cycles under the following conditions: initial denaturation at 94°C/2 min; cycling at 94°C/30 s, 58°C/40 s, 72°C/1 min; and a final extension at 72°C/10 min. The amplification of DAB3 ran in 30 cycles under the following conditions: initial denaturation at 95°C/4 min; cycling at 95°C/30 s, 56°C/30 s, 72°C/45 s; and a final extension at 72°C/10 min. PCR reactions were performed in a Mastercycler ep gradient thermocycler (Eppendorf). PCR products (342 bp and 350 bp for DAB1-like and DAB3-like genes, respectively) were visualized on 1% GoldView-stained agarose gel and purified using High Pure PCR Product Purification Kit (Roche). For the quality and heterozygosity controls, the purified PCR products were sequenced in both directions using the same primers as in the amplification reaction. Sequencing reactions were performed using the fluorescent chemistry of BigDye Terminator v3.1 Ready reaction Cycle Sequencing Kit (Applied Biosystems). The purified products of the sequencing reaction (BigDye X-Terminator Purification Kit; according to the manufacturer's protocol) were analyzed using an ABI 3130 Genetic Analyzer (Applied Biosystems by Life Technologies, Carlsbad, California) under the appropriate module. Sequences were analyzed using Sequencing Analysis version 5.2 (Applied Biosystems) and Sequencher version 5.0 (Gene Codes Corp., Ann Arbor, MI USA) software.

Successfully amplified samples were subsequently subjected to clonal 454 amplicon pyrosequencing to distinguish particular alleles (454 Titanium Roche technology; commercial customer service by Beckman Coulter Genomics SAS, France). For template preparation, the non-classical fusion primers consisted of a 454 Titanium amplicon adaptor, a 6 bp sequence tag, and a specific primer sequence, but only tag-modified genotyping primers for PCR amplification were designed; i.e. the tag consisting of 7 nucleotides was added at the 5′ end to each specific PCR primer. The precise conditions for designing the primers' tags were as follows: no homopolymer in the tag, at least 3 substitutions between any pair of tags, and no motif repetition inside the tag. The PERL script used to create and select the tags compatible with the mentioned rules is available on request. In total, 40 unique tags were defined (unpublished), out of which 12 tag-modified forward primers and 8 tag-modified reverse primers for each gene made up the unique primer combination for each sample in a 96-well plate. These 7 bp tags were used to assign 454 sequencing reads to individual samples in later analysis.

Samples were newly amplified, each individual sample employing a unique tagged primer combination. The PCR mixture composition, volume, and cycling conditions were the same as in standard PCR (described above). The PCR amplicons with tagged primers were 356 bp and 364 bp long (for DAB1 and DAB3, respectively; including tagged primers). The quality of the amplified product was checked by electrophoresis on 1% GoldView-stained agarose gel. After purification (High Pure PCR Product Purification Kit, Roche), the products were quantified by a Nanodrop 8000 spectrophotometer (Thermo Scientific) and mixed together using equimolar quantities of PCR products according to the provider's requirements (minimum of 1 µg and 2E⁺¹⁰ molecules of purified material in the final pool). As the amplicons for both genes have similar lengths and sequences, all samples were pooled together in one single tube (up to 96 samples for each gene, i.e. a maximum of 192 amplicons for each run, including positive and non-template controls). The prepared amplicon pool was sent to the service provider Beckman Coulter Genomics, who performed quality control of the template, adaptors ligation, fragment library construction, run reading, and raw data delivery (Supporting Information S2).

To extract, visualize and analyze amplicon sequences, we used the user-friendly SESAME program web application package [37] for analyzing amplicon sequences obtained through NGS (next generation sequencing) technologies. The output of the sequence variants from SESAME was further analyzed in Sequencher 5.0 (Gene Codes Corporation) and compared with the Sanger electrophoretogram of the particular sample. On the basis of the unique tag combination for each sample, the corresponding reads were assigned to each particular sample, and, following the similarity of reads with the gene reference sequences, the reads for particular samples were assigned to two groups of DAB-like alleles (DAB1-like and DAB3-like) (with an eValue corresponding to 1E⁻¹⁰) using SESAME. In addition, SESAME assistant provided the alignment and trimming of sequences, the processing of uncorrected sequences, and the final correction of sequences (pre-set error rate 0.05). After analyzing each run (3 runs were analyzed in this study) the output including the number of analyzed samples, number of raw sequences, number of marker-assigned sequences, number of sample-assigned sequences, number of samples with sequences, and final number of assigned variants for the run was created. For a particular fish sample, a maximum number of 8 DAB-like alleles (because of the hypothetical duplication of the DAB1-like gene and DAB3-like gene) was expected. The number of sequences, number of sequence variants, frequency of a particular sequence variant for each particular fish sample, and frequency of a particular sequence variant in the whole run were obtained. Even though all amplicons were adjusted to the same concentrations before pooling, the reads for different samples varied from hundreds to thousands. The assigned sequence variants for a particular sample matched the Sanger sequencing output.

To distinguish true alleles from artefacts generated at different stages of MHC genotyping, we followed the procedures of Zagalska-Neubauer et al. [38] as applied by Nadachowska-Brzyska et al. [24]. Briefly, we excluded all variants containing indels causing frameshifts or present in one copy in all data sets. We considered all amplicons with sufficient coverage and we calculated the maximum per-amplicon frequency (MPAF) for each sequence variant. Each sequence variant with an MPAF of less than 5% was checked for the potential artefact (i.e. 1 bp substitution or recombination from other sequence variants in a given amplicon). The sequence variants present in ≥50% of the amplicon's reads were not explained by the artefacts and were considered as the true alleles in our study. A large proportion of true alleles were confirmed by two independent PCRs. Nevertheless, as Nadachowska-Brzyska et al. [24] confirmed, in a multiple populations approach, many true alleles are observed as singletons. Because using the method of 454 sequencing avoids the problem of recombination in contrast to Sanger cloning [39], the singletons may represent the true MHC alleles. In our study, all singletons were sequence variants with mean MPAF equal 24%. In addition, the genotyping of all specimens for which the singletons were found was repeated twice.

The specific alleles of P. toxostoma were termed as Pato-DAB alleles and the specific alleles of C. nasus were termed as Chna-DAB alleles following the nomenclature of Klein et al. [40]. If the two species shared the alleles, we applied the nomenclature used by Nadachowska-Brzyska et al. [24] for hybridizing species of newts i.e. the alleles shared in P. toxostoma and C. nasus and the alleles identified solely in their hybrids were termed as Pctn-DAB alleles. Concerning DAB genes in cyprinid fish, Dixon et al. [41] presumed that DAB1 and DAB3 sets of sequences represent two different loci which may or may not contain several genes (i.e. the terminology DAB1-like genes and DAB3-like genes following Rakus et al. [15] and Seifertová and Šimková [42] was therefore applied in our study). The new DAB alleles were deposited in EMBL under Accession numbers from HF969045 to HF969122.

Phylogenetic analyses

Phylogenetic reconstructions were performed in PAUP* 4.0b10 for Microsoft Windows 95/NT [43] applying a distance method using a neighbour-joining (NJ) algorithm. A bootstrap analysis with 1000 replicates was performed to determine the reliability of the branching in the phylogenetic tree. The hierarchical likelihood ratio test (hLRT) implemented in ModelTest v. 3.6 [44] was used to determine the appropriate substitution model of sequence evolution that best fitted the dataset. The maximum likelihood (ML) analyses using a heuristic search with a tree bisection reconnection branch-swapping algorithm were performed separately for P. toxostoma and C. nasus. The resulting ML trees were used in the codon-based analyses of selection performed in PAML 4.3 [45] (see below).

Recombination and selection analyses

GENECONV v.1.81a [46], which employs a substitution model to detect the most likely candidate fragments for recombination/gene conversion events between sequences, was applied. GENECONV defines a fragment as an aligned pair of segments between two sequences that is bounded by either two discordant sites, or else by one discordant site and an end of the alignment. GENECONV looks for inner fragments, which are evidence of a possible gene conversion event between the ancestors of two sequences in the alignment. GENECONV can also look for outer fragments, which are evidence of past gene conversion events that may have originated from outside of the alignment, or else from within the alignment but such that evidence of the source has been destroyed by later mutation or gene conversion. By default, GENECONV finds significant “inner” and “outer” sequence fragments. Global and pairwise statistical tests for recombination events were used with 10,000 permutations of the data to assess significance. In the case of multiple tests, the software has an integrated Bonferroni correction that minimizes the inflation of type I errors in many pairwise tests that are performed in a reasonably large sequence sample.

First, to detect the signatures of positive selection in DAB-like sequences, the relative rate of nonsynonymous (d_N) and synonymous substitutions (d_S) was calculated according to Nei and Gojobori [47], applying the correction of Jukes and Cantor [48] for multiple hits using MEGA 5 [49]. A significant excess of d_N substitution over d_S substitution is considered as convincing evidence of non-neutral evolution. A one-tailed Z test was performed for all codons, using only ABS (antigen-binding sites) codons or only non-ABS codons. The ABS codons were inferred from the structure of the human crystal model of the MHC class 2 DRB1 locus [50].

Next, analyses to detect positive selection on a site-by-site basis were performed using a maximum likelihood approach in the CODEML program implemented in PAML 4.3 [45]. Different codon-based models incorporating selection (M3, M2a and M8) and not incorporating selection (M0, M1a and M7) were used to test for the presence of sites under selection, as recommended by Yang [45]. The models used the nonsynonymous/synonymous rate ratio (ω = d_N/d_S) as an indicator of selective pressure on the protein. According to Yang [45], all three pairs of models (i.e. M0 (one ratio) versus M3 (a discrete model involving three site classes for ω), M1a (nearly neutral) versus M2a (positive selection), and M7 (a β model which uses beta distribution) versus M8 (β and ω ratio estimated from the data)) appear to be particularly useful, forming three likelihood ratio tests of positive selection. A likelihood ratio test (LRT) was used to assess the significance of the differences between models. If the LRT statistic comparing two models indicates that the alternative models (M2a, M3 and M8) fit the data significantly better than the simpler models (M1a, M0 and M7), this suggests the presence of sites with ω>1 and the action of positive selection at specific sites in the protein-coding sequences [51]. Following the recommendation of Yang [45], the Bayes empirical Bayes (BEB) method was used to calculate the posterior probabilities for site classes and to identify sites under selection (the posterior means of ω for positively selected sites (PSS) are >1). BEB was implemented under models M2a and M8 only.

Population genetic analyses

The population genetic structure based on MHC IIB genes was analyzed using the approach of Nadachowska-Brzyska et al. [24]. Binary coding was used to determine the presence/absence of a given DAB-like allele in specimens. Because we were not able to assign the alleles to loci (i.e. our results suggest the duplication of both DAB1 and DAB3 loci, see below), each allele was considered as a separate dominant locus in population genetic analyses. A Bayesian clustering method implemented in STRUCTURE 2.3.3 [52] was used to infer and delineate the most probable number of genetically homogenous groups of sampled specimens (i.e. K genetic clusters) with no a priori assumptions of population structure, and to assign specimens to the inferred groups based on allelic frequencies. Here, a method developed for dominant markers such as AFLP markers [53] was used. A model with correlated allele frequencies and allowing admixture was applied. Five independent runs for K ranging from 1 to 10 were performed, using 4×10⁶ Markov chain Monte Carlo replications with a burn-in period of 4×10⁵ chains. To infer the most appropriate number of genetic clusters in the data set, the ad hoc statistic ΔK method proposed by Evanno et al. [54] was used. The results of ten independent runs using a fixed number of most probable K were subsequently post-processed in CLUMPP 1.1.2 [55]. Pairwise F_st values between populations for MHC were estimated using Arlequin 3.5.1.3 [56] (1) by applying the method developed for dominant markers using binary encoded data and (2) from haplotype frequencies using the allele haplotype and number of individuals with a given allele for each population. 95% confidence intervals (CI) around F_st values were calculated by bootstrapping 20000 times over loci.

For microsatellites, the presence of null alleles was checked in MICRO-CHECKER [57]. FSTAT 2.9.3.2 [58] was used to test for deviations from the Hardy-Weinberg equilibrium (HWE). Significance levels of multiple testing were adjusted using a false discovery rate correction [59] in the QVALUE program [60]. Number of alleles (N_a), allelic richness, and number of private alleles were computed in FSTAT 2.9.3.2 [58] and GenAlEx 6 [61] programs. Arlequin software was also used to calculate the pairwise F_st values between populations with N>8 and their 95% CI (Supporting Information S3 and S4). Population structure using microsatellite allele size data was further studied using a Bayesian clustering approach applying the same model as described above. ANCOVA was used to test the effect of species on the microsatellites (number of alleles and allelic richness) or on the MHC (number of alleles and mean number of alleles per specimen in population); sample size was included as covariate. The Spearman correlation was calculated between MHC allelic richness (measured as the mean number of MHC alleles per specimen) and microsatellite allelic richness in a given population. The Mantel test using the Spearman correlation and 10000 permutations between F_st distances for MHC and those for microsatellites was applied. The Spearman correlation was calculated between the mean individual values of q calculated for microsatellites and those calculated for MHC genes; the mean individual values of q represent the proportion of ancestry in cluster 1 or cluster 2 from the CLUMPP output using the Bayesian clustering approach. All statistical analyses were performed in Statistica 10.0 for Windows, StatSoft Inc, and R software (R version 2.12.2 (20011-02-25), The R Foundation for Statistical Computing using the package by Sarah Goslee and Dean Urban (2012) “Ecodist: Dissimilarity-based functions for ecological analysis”.

Molecular profiles of Chondrostoma/Parachondrostoma populations

Chondrostoma/Parachondrostoma specimens from the localities investigated were genetically determined using mtDNA and 41 microsatellite loci. Using the combination of mtDNA and microsatellites, the allopatric status of a population for C. nasus at Allier and for P. toxostoma at Orbieu was confirmed. In the zones where P. toxostoma and C. nasus live in sympatry, our sampling revealed differences in the occurrence of both species between localities (Table 1). The number of hybrids within localities where these two species live in sympatry (i.e. Pont de Laragne, Pertuis, Manosque, Avignon, Saint Just and Labeaume) was variable, ranging from 3% to 20%.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Populations of P. toxostoma (PT) and C. nasus (CN) studied plus sample sizes for hybrids (H) in sympatric zones.

https://doi.org/10.1371/journal.pone.0065883.t001

MHC polymorphism in P. toxostoma and C. nasus: species level

P. toxostoma and C. nasus differ in their levels of MHC polymorphism measured by a spectrum of MHC alleles per species when all “pure” specimens identified using mtDNA and microsatellite markers were taken into account. A total of 55 different DAB-like alleles (13 DAB1-like and 42 DAB3-like) were identified in P. toxostoma. However, only 30 different DAB-like alleles (6 DAB1-like and 24 DAB3-like) were found in C. nasus. Concerning the alleles of the DAB1 lineage, 8 DAB1-like alleles were specific to P. toxostoma (7 of them were found in at least two specimens and one was a singleton), and only one DAB1-like allele was specific to C. nasus. Concerning the alleles of the DAB3 lineage, 34 alleles were specific to P. toxostoma (19 of them were found in at least two specimens and 15 of them were singletons), and 16 alleles were specific to C. nasus (12 of them were found in at least two specimens and 4 of them were singletons). Several alleles were shared by both species, i.e. 5 DAB1-like alleles and 8 DAB3-like alleles were found in the populations of both P. toxostoma and C. nasus. Some shared allelic variants appeared frequently in specimens (from 10% to 34% of specimens expressed a given sequence variant), whilst others were rarely found (from 0.8% to 7% of specimens expressed a given sequence variant). For all shared allelic variants except Pctn-DAB3*34 (this allelic variant was equally distributed between the two species), a high frequency was found in one species and a low frequency in the other (7 allelic variants were frequent in C. nasus and rare in P. toxostoma, whilst only 3 allelic variants were frequent in P. toxostoma and rare in C. nasus). Two shared alleles were also identified in the allopatric population of P. toxostoma and 5 shared alleles were also identified in the allopatric population of C. nasus. Despite the low proportions of hybrids in samples from the localities where both species live in sympatry, a high number of DAB-like alleles were introgressed into the hybrid genomes, i.e. 14 P. toxostoma-specific, 6 C. nasus-specific, and 7 shared DAB-like alleles were found in the hybrid specimens. The majority of shared allelic variants found in hybrids were frequent (see above), whilst species-specific allelic variants were found with variable frequency (0.9 to 12% of specimens expressed a given sequence variant). In addition, 6 other DAB-like alleles were recognized exclusively in the hybrid specimens.

Phylogenetic reconstruction including all DAB-like alleles of P. toxostoma, C. nasus and their hybrids was performed to investigate the ancestry of MHC polymorphism in these fish. DAB-like alleles were clustered in two main lineages, i.e. DAB1-like and DAB3-like. Even though we confirmed the trans-species polymorphism between P. toxostoma and C. nasus, surprisingly, we found that DAB3-like alleles tended to cluster in a species-specific way, i.e. one large clade including a major proportion of P. toxostoma-specific alleles and one large clade including a major proportion of C. nasus-specific alleles were found (Figure 1). The alleles shared by both species were situated randomly within a phylogenetic tree. The position of alleles found exclusively in hybrids indicates that these alleles originated from P. toxostoma.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. The neighbour-joining tree including all DAB-like alleles identified in P. toxostoma and C. nasus.

The alleles found in pure specimens of both species, i.e. P. toxostoma (Pato-DAB alleles) and C. nasus (Chna-DAB alleles), are shown in black. The alleles shared between two species are shown in red, and those identified exclusively in hybrids are shown in green; both of them are termed as Pctn-DAB alleles. All alleles found in hybrids (i.e. species-specific, common or exclusively hybrid) are shown with #. Bootstrap values >50% are shown.

https://doi.org/10.1371/journal.pone.0065883.g001

MHC variability in P. toxostoma, C. nasus and hybrids: population and individual levels

In all populations of C. nasus, this species tended to reach a higher mean number of DAB-like alleles per specimen when compared with P. toxostoma populations (Table 1). C. nasus specimens expressed mainly three DAB-like alleles, whilst the majority of P. toxostoma specimens expressed two DAB-like alleles. The hybrids exhibited a pattern of DAB-like expression intermediate between C. nasus and P. toxostoma, i.e. the majority of them expressed two or three DAB-like alleles (Figure 2). The number of DAB-like alleles expressed in C. nasus was significantly higher than that expressed in P. toxostoma when including all specimens (from allopatric populations and from the sympatric zones), when including only the specimens from sympatric zones, or when analyzing separately the Durance and Ardeche sympatric zones (MW test, p<0.001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Distribution of DAB-like alleles in P. toxostoma (A), C. nasus (B), and hybrid (C) specimens.

https://doi.org/10.1371/journal.pone.0065883.g002

The numbers of different DAB1-like and DAB3-like alleles in the allopatric population of C. nasus was similar. However, the numbers of different DAB3-like alleles in the allopatric population of P. toxostoma and all sympatric populations of both species were higher than the numbers of DAB1-like alleles, and the same trend was observed for hybrids (Table 1). Using all populations or only those situated in sympatric areas, no effect of species on the number of different DAB-like alleles in the population was found (ANCOVA with sample size as covariate, p>0.05). Patterns of expression for the alleles of DAB1 and DAB3 lineages differed between P. toxostoma and C. nasus (Figure 3). Concerning the DAB1 lineage, the proportion of P. toxostoma specimens from the allopatric population and the four populations sampled from sympatric areas that did not express the alleles of the DAB1 lineage was the highest. Only in Pertuis (a locality within the Durance sympatric zone) were the proportions of P. toxostoma specimens expressing one DAB1-like allele, two DAB1-like alleles or no DAB1-like alleles balanced. On the other hand, the proportion of C. nasus specimens expressing one DAB1-like allele was the highest in all populations of this species. Concerning the DAB3-like lineage, the highest proportion of P. toxostoma specimens expressing at least two DAB3-like alleles was found in four populations sampled from sympatric areas, whilst the proportion of P. toxostoma specimens expressing a single DAB3-like allele was the highest in Orbieu (i.e. in the allopatric population) and Pertuis. The highest proportions of C. nasus specimens expressing at least two DAB3-like alleles were found in the samples from sympatric areas; the proportions of specimens without DAB3-like alleles or with at least two DAB3-like alleles in the sample from the allopatric population were low. In hybrids, 50% of specimens did not express the alleles of the DAB1-like lineage and 71% of them expressed at least two DAB3-like alleles.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Frequency of the DAB1 and DAB3 expression patterns in P. toxostoma and C. nasus populations.

https://doi.org/10.1371/journal.pone.0065883.g003

Recombination and selection maintaining MHC polymorphism

GENECONV detected four significant global inner fragments (sim p<0.05) in the DAB-like alignment of P. toxostoma. However, using Bonferroni-corrected KA p-values, no global inner fragment was significant (p>0.05). No significant global outer-sequence fragment was found. No significant inner or outer global fragments were detected in the DAB-like sequence alignment of C. nasus. Five additional inner pairwise fragments were significant (BC KA p-value p = 0.044).

The average rates of nonsynonymous substitutions were higher than those of synonymous ones for all sites and ABS sites for each species (Table 2). However, positive selection was not detected at all sites or at non-ABS sites. The d_N/d_S ratio was significantly elevated only at ABS sites of DAB-like alleles in P. toxostoma and C. nasus, which provides evidence of positive selection acting on the PBS of MHC IIB of both species.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. The average rates of non-synonymous (d_N) and synonymous (d_S) substitutions (mean and standard errors obtained using 1000 bootstrap replicates), Z-test of positive selection with p-value included.

https://doi.org/10.1371/journal.pone.0065883.t002

The LRT statistic comparing the two models indicates that the models that account for the sites under selection (i.e. M2a, M3 and M8) fit the data significantly better (p<0.001) than simpler models that do not allow for selection (i.e. M1a, M0 and M7), which indicates the action of positive selection at specific sites in DAB sequences. Log-likelihood values and parameter estimates under random-site models are included in Table 3. On the basis of BEB analysis, 17 PSS in C. nasus were identified using the M2a and M8 models, whilst only 12 sites using the M2a model and 14 sites using the M8 model were under positive selection in P. toxostoma. 14 out of 17 PSS in C. nasus and 12 out of 14 PSS in P. toxostoma corresponded to human antigen binding sites (ABS) identified using crystallography by Brown et al. (1993). The Bayes identification of sites under positive selection calculated using the M8 model is included in Figure 4. The same 13 PSS were identified in both species. In addition, codons 36, 58, 87 and 92 in C. nasus and codon 75 in P. toxostoma were identified to be under positive selection.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Approximate posterior means of ω under the M8 random sites model.

Approximate posterior means of ω calculated as the weighted average of ω over the 11 site classes and weighted by the posterior probabilities under the M8 random sites model are shown for DAB-like sequence variants for P. toxostoma (A) and for C. nasus (B). Sites inferred to be under positive selection at the 99% level are indicated by large white squares and those at the 95% level are indicated by small white squares. The reference sequences in the graphic presentation are the following: Pato-DAB3*01 for P. toxostoma and Chna-DAB3*02 for C. nasus.

https://doi.org/10.1371/journal.pone.0065883.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Log-likelihood values and parameter estimates under random-site models for P. toxostoma and C. nasus.

https://doi.org/10.1371/journal.pone.0065883.t003

Genetic structure of P. toxostoma and C. nasus populations

Of all 41 loci tested, a total of 16 microsatellite loci (LleA-029, LleA-071, BL1-98, BL2-114, Lsou08, Lsou34, Ppro132, CtoA-256, CtoE-249, CtoF-172, CypG24, IVO4, LCO1, N7K4, CA1 and Z21908) were discarded from the study due to the probable occurrence of null alleles as indicated by MICRO-CHECKER. All the remaining 25 loci that were subsequently used in genetic structure analyses were in HWE for each population.

Structure analysis using the Bayesian clustering approach revealed two genetic clusters when using microsatellites or MHC data without reflecting the pattern of geographical distribution or the origin of populations (i.e. the greater distances between allopatric and sympatric populations, the close proximity of sympatric populations, the different origins of the Durance hybrid zone and the Ardeche hybrid zone) (Figure 5). The analysis performed using microsatellites revealed the partial admixture between P. toxostoma and C. nasus in the areas where both species live in sympatry, confirming the hybridization process between these two species. Using MHC data, a more evident admixture between genetic clusters of P. toxostoma and C. nasus was observed when compared with microsatellite data. In the areas where both species live in sympatry, MHC data revealed bidirectional gene flow, as also indicated by microsatellites. The evident admixture of MHC alleles from P. toxostoma to C. nasus was observed even in the allopatric population of C. nasus. A higher admixture of MHC alleles from C. nasus to P. toxostoma than from P. toxostoma to C. nasus was observed in sympatric zones for specimens identified as hybrids based on the identification using microsatellites (especially evidenced in the locality Pertuis).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Genetic structure of P. toxostoma and C. nasus populations in MHC IIB genes and microsatellites.

Genetic structure of P. toxostoma and C. nasus populations in MHC IIB genes (A, B) and microsatellites (C) inferred from Structure analysis for K = 2. The squares, circles and triangles correspond to C. nasus, P. toxostoma and hybrids identification based on the cyt b and microsatellites. Assignment of populations and hybrid samples to P. toxostoma and C. nasus clusters using MHC IIB genes is shown on the map (A).

https://doi.org/10.1371/journal.pone.0065883.g005

Pairwise F_st distances inferred from microsatellites revealed stronger genetic differentiation between C. nasus and P. toxostoma populations than that indicated by pairwise F_st distances inferred from MHC data. Based on microsatellites, both allopatric and sympatric populations of C. nasus were well differentiated from those of P. toxostoma (average pairwise F_st ranging from 0.415 to 0.524). Within species, higher F_st values were observed between the allopatric population and sympatric populations than between the pairs of sympatric populations (this trend was more evident in P. toxostoma) (see Supporting Information S3). Based on F_st calculated for binary encoded MHC data (Supporting Information S4), the allopatric population of P. toxostoma was much more different from its sympatric populations when compared to the allopatric and sympatric populations of C. nasus. The populations of P. toxostoma from sympatric zones were more similar than the allopatric and sympatric populations of this species. Such differences were less in evidence for C. nasus. The genetic differentiation based on MHC data calculated between sympatric populations of P. toxostoma and C. nasus was variable and independent of the origin of the populations (i.e. Durance versus Ardeche hybrid zone). Based on F_st calculated for haplotype MHC data (Supporting Information S4), the genetic differentiation between the P. toxostoma population from Pertuis and other P. toxostoma populations from Durance or Ardeche zones was higher than the genetic differentiation between other pairs of P. toxostoma populations from sympatric areas.

At population level, no effect of species on microsatellite diversity expressed as the number of different alleles or microsatellite allelic richness was found (ANCOVA with sample size as covariate, p>0.05). However, a significant effect of species on the mean number of MHC alleles per specimen in the population was found (ANCOVA with sample size as covariate, total F_2,8 = 15.38, p = 0.002, sample size F = 1.00, p = 0.35, species F = 30.58, p<0.001) with a significantly higher MHC number in C. nasus populations when compared to P. toxostoma populations. At the population level, a positive correlation was found between microsatellite allelic richness (see Table 4) and mean number of MHC alleles per specimen in a given population (N = 11, R = 0.772, p = 0.005). A significant correlation between F_st distances calculated between pairs of populations for microsatellites and haplotype MHC data (N = 55, 0.507, p = 0.002) or between microsatellites and binary encoding MHC data (N = 55, R = 0.738, p<0.001) was found. The assigning of probabilities based on the individual values of q inferred from Structure using MHC and microsatellites is shown in Figure 6. At the individual level, a positive correlation was found between MHC and microsatellites based on the individual values of q inferred from structure analyses (N = 232, R = 0.831, p<0.001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 6. Probabilities assignation of specimens inferred from Structure using MHC and microsatellites.

The red squares, green circles and yellow triangles correspond to C. nasus, P. toxostoma and hybrids identified using cyt b and microsatellites.

https://doi.org/10.1371/journal.pone.0065883.g006

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Microsatellite diversity for P. toxostoma populations (PT), C. nasus populations (CN) and hybrid sample (H) including number of different alleles (N_a), number of private alleles, allelic richness per sample with sample size >8 (i.e. per population or hybrid sample).

https://doi.org/10.1371/journal.pone.0065883.t004

MHC diversity in P. toxostoma and C. nasus

Our analyses revealed different levels of MHC polymorphism in P. toxostoma and C. nasus. Native endemic P. toxostoma showed an extensive degree of MHC polymorphism measured by the number of different allelic variants per species when compared to C. nasus, probably resulting from the different origins and dispersion histories of both species. Concerning the populations of different endangered species, low genetic diversity of MHC especially due to a bottleneck effect has been shown (e.g. [3], [62]), suggesting that such species may be exposed to increased vulnerability to novel pathogens. However, P. toxostoma, a fish with protected status in southern France, expressed similar total numbers of MHC IIB allelic variants at the population level when compared to the populations of invasive C. nasus. This finding may suggest that similar evolutionary mechanisms operate to generate MHC diversity in both species (see below). However, at the individual level, a higher number of allelic variants were found for C. nasus specimens when compared to P. toxostoma specimens. Šimková et al. [32] demonstrated that C. nasus is a species more parasitized by monogenean species (i.e. this fish species is especially parasitized by the wider range of Dactylogyrus species that achieve a high intensity of infection) when compared to P. toxostoma. MHC genes are considered as possible candidates for co-adaptation genes evolving in interactions with parasite genotypes, especially those concerning host-specific parasites [63], [64]. The presence of a higher number of MHC allelic variants in C. nasus specimens compared to the number of MHC allelic variants in P. toxostoma specimens could result from co-evolutionary interactions between host and parasite genotypes and should explain the presence and high abundance of some monogenean parasite species in C. nasus and the low intensity of infection (or absence of some monogenean species) in P. toxostoma. However, further analyses are necessary to test such an evolutionary scenario.

Concerning the presence of DAB-like alleles of different allelic lineages, i.e. DAB1-like and DAB3-like previously recognized in different cyprinid fish species (e.g. [36], [41], [65], [66]), we confirmed the expression of alleles of DAB1 and DAB3 lineages in both P. toxostoma and C. nasus. In accordance with previous studies [66], [67], the specimens expressed the alleles of one lineage (i.e. DAB1 or DAB3) or expressed the alleles of both. However, different patterns of expression were found when comparing native P. toxostoma and invasive C. nasus. Whilst the majority of C. nasus specimens expressed the alleles of both lineages, P. toxostoma specimens mainly expressed only the alleles of the DAB3 lineage. Ottová et al. [67] showed that the expression of alleles of different DAB lineages is advantageous for fish males of common bream (Abramis brama), i.e. it is related to high fish condition and more intensive sexual coloration. Further immunoecological studies investigating whether the expression of the DAB-like alleles of the two different lineages represents some advantages potentially linked to body condition, immunity, survival or the dispersion of invasive C. nasus could be helpful in predicting the evolution of hybrid zones over time and, from a more general point of view, could also be helpful in clarifying the function of DAB genes in cyprinid fish species.

MHC variability in hybrids

Our study indicated that the number of MHC variants found in P. toxostoma x C. nasus hybrid specimens is intermediate between the two parental species. Such a finding does not support the hypothesis of the super-optimal variability of MHC in hybrids, which was proposed as a mechanism selecting against interspecific hybrids by Eizaguirre et al. [68]. They hypothesized that if largely divergent sets of MHC alleles are combined from two mates of different species, an increased number of MHC variants in hybrid specimens may not only bind more foreign, but also a greater variety of self-derived peptides that are recognized by T cell clones. Thus, the T-cell repertoire available for parasite recognition is reduced. Therefore, hybrids should suffer from a higher parasite load, which was not, however, the case for P. toxostoma x C. nasus hybrids in our study. As shown by Šimková et al. [32], hybrids and P. toxostoma have similar intensities of ectoparasite infection, whilst C. nasus is more infected than P. toxostoma and the hybrids. This may suggest that the P. toxostoma-specific MHC alleles introgressed into the hybrid genomes may represent an advantage, i.e. such alleles probably generate low hybrid susceptibility to ectoparasites widely infecting invasive C. nasus.

Concerning the DAB-like alleles expressed in hybrids in our study, a higher number of alleles specific to P. toxostoma (which was the more diverse species concerning total MHC allelic diversity) were found in hybrids compared to the number of alleles specific to C. nasus; moreover, the phylogenetic reconstruction indicates that the alleles identified solely in hybrids were of P. toxostoma origin, suggesting that the hybrids are more introgressed by P. toxostoma than by C. nasus. This may indicate that the higher degree of MHC variability in hybrids originated from P. toxostoma. Nevertheless, many alleles shared between P. toxostoma and C. nasus were common in C. nasus and rare in P. toxostoma, and such alleles were also identified in hybrids, suggesting bidirectional MHC introgression. Bidirectional introgression, which is not common in hybridizing species, was previously demonstrated in these cyprinid fish using morphological traits, allozymes, and mtDNA [26], [27]. Following Bayesian cluster analysis using MHC data, P. toxostoma populations seem to be introgressed more often by C. nasus than C. nasus by P. toxostoma in sympatry. Surprisingly, a high level of introgression of MHC alleles from C. nasus was observed in the pure P. toxostoma population from Pertuis, i.e. in the specimens identified as P. toxostoma determined on the basis of mtDNA and microsatellites. In our study, all such specimens expressed three or four DAB-like alleles, which is an unusual situation for pure P. toxostoma. This may indicate that these specimens were the hybrids that we were not able to determine using 41 microsatellitte and mitochondrial markers. This result clearly demonstrates that different evolutionary forces influence the genomic structure of hybrid specimens. Indeed, a majority of microsatellite loci showed P. toxostoma alleles, and the DAB loci (MHC IIB genes) displayed C. nasus alleles. This result clearly shows how hybrid genotypes can serve as intermediaries for the transfer of adaptive genetic variation between parental populations. Such a pattern is well known in plants [69] but rarely demonstrated in vertebrates.

MHC diversity: trans-species polymorphism, interspecific hybridization, or other factors

As already described, high MHC variability is a result of the action of several evolutionary mechanisms. Similarly to many previous studies performed on fish species (e.g. [36], [70]–[71]), we confirmed the trans-species evolution of MHC IIB genetic polymorphism in P. toxostoma and C. nasus, i.e. one DAB3-like allele (Pctn-DAB3*34) was shared between two species sampled in allopatric populations (however, only one allopatric population was sampled per species because of the protected status of P. toxostoma in southern France). In addition, the phylogenetic analyses revealed that some C. nasus-specific alleles were clustered within the groups of P. toxostoma-specific alleles, or that P. toxostoma-specific alleles were clustered within the groups of C. nasus-specific alleles. The other 12 alleles were found in both P. toxostoma and C. nasus in sympatric areas, which indicates that intergeneric hybridization plays an important role in sharing MHC alleles between P. toxostoma and C. nasus living in sympatry.

Nadachowska-Brzyska et al. [24], investigating two sister species of newts, showed that MHC variability in their parapatric ranges is increased due to interspecific hybridization. Nevertheless, Wegner and Eizaguirre [23] pointed out that it is difficult to distinguish between ancestral trans-species polymorphism and MHC introgression in species living and hybridizing in sympatry. In our study, MHC diversity seems to be shaped by both these processes. On the basis of Bayesian clustering using microsatellites, admixture of P. toxostoma and C. nasus genetic clusters was observed solely in the areas where both species live in sympatry, supporting the presence of intergeneric hybridization between the two cyprinid species. On the other hand, using MHC data, admixture was observed also in allopatric populations (in this case, the admixture of P. toxostoma into C. nasus was more evident), confirming that MHC diversity is shaped by trans-species polymorphism. However, a higher level of admixture was observed in areas where both species lived in sympatry, suggesting that intergeneric hybridization also contributes to MHC diversity in P. toxostoma and C. nasus populations. The role of intergeneric hybridization in increasing MHC variability in areas where both species live in sympatry is also supported by the fact that some hybrids carried alleles that were not identified in parental species. Thus, it could be suggested that intergeneric hybridization generates new MHC allelic variants by means of the crossover between non-sister chromatids during meiosis. New genomic combinations of P. toxostoma and C. nasus hybrids were also identified previously using nuclear data [27]. However, the presence of hybrid-specific alleles may also result from the limited sampling size in our study.

Nevertheless, in areas where both species live in sympatry, admixture was also observed when considering only the specimens of pure species (identified on the basis of mtDNA and 41 polymorphic microsatellite loci), which is also in line with the idea of the trans-species evolution of MHC polymorphism. However, a higher level of MHC polymorphism in sympatric populations when compared to MHC polymorphism in allopatric populations could also result from other factors such as habitat fragmentation and human activity linked with environmental pollution (e.g. Cohen [72]) or parasite-mediated selection (in our study, both species came into contact, which especially facilitated the transmission of ectoparasites). The introduction of a new parasite species following the introduction of its host fish could induce selection on the host immunity of native fish species in order to increase the number of MHC allelic variants within a population.

Genetic differentiation of P. toxostoma and C. nasus populations

Using both the Bayesian clustering method and F_st distances, it was found that all populations of P. toxostoma were well differentiated from those of C. nasus on the basis of microsatellites. However, no genetic differentiation based on genetic clustering was evidenced between the allopatric population and all sympatric populations within each species. The allopatric population of P. toxostoma was well differentiated from all sympatric populations of this species based on F_st distances using microsatellites, whilst such genetic differentiation for C. nasus was little evident, which is in line with the recent history of the dispersion of invasive species (the Durance sympatric zone is about 100–150 years old, following Costedoat et al. [28], and hybridization in the Ardeche sympatric zone is of more recent origin, unpublished data). However, our results using binary encoded MHC data indicate stronger genetic differentiation between allopatric and sympatric populations of P. toxostoma than indicated by microsatellites, which may suggest that MHC in the allopatric population of native P. toxostoma investigated in our study is under different selective pressure than MHC in its sympatric populations. One of the possible explanations is the existence of different selective pressures from parasites. Šimková et al. [32] showed that the composition of parasite communities in the allopatric population of P. toxostoma from Orbieu is very different from that found in populations of P. toxostoma living in sympatry with C. nasus. Concerning invasive C. nasus, low genetic differentiation between the allopatric population and sympatric populations of this species was found: this differentiation was even comparable with the genetic differentiation between sympatric populations of P. toxostoma and C. nasus. Šimková et al. [32] revealed the highly similar composition of parasite communities between allopatric and sympatric populations of C. nasus, and the similar composition of parasite communities (based only on parasite presence data) in P. toxostoma and C. nasus populations living in sympatric areas. Nevertheless, to confirm the pattern of genetic differentiations between allopatric and sympatric populations of native P. toxostoma in further studies, it would be useful to include a wider range of allopatric populations for this species, which was not done in our study because of its threatened status.

Positive versus neutral selection acting on MHC genes

Using a combination of specific functional loci (such as MHC genes) and genome-wide random markers (such as microsatellite loci) is considered to be a good way to assess the threat posed by reduced genetic diversity in species [5]. Such an approach combining microsatellites and MHC analyses was performed for several endangered species and revealed reduced microsatellite and MHC diversity and/or a positive correlation between MHC variation and the variation at neutral loci across populations, indicating that genetic drift significantly affects MHC variability on a short timescale and that even the genetic drift in some small bottlenecked and fragmented populations may overshadow the role of balancing selection [6]–[7], [73]. However, several studies identifying the effect of genetic drift on MHC diversity also demonstrated the significant role of balancing selection in shaping MHC polymorphism on a long timescale (see the review study on MHC diversity by Radwan et al. [20]). In our study, we showed a positive correlation between microsatellite allelic richness and mean number of MHC alleles per individual in a population, suggesting the effect of neutral selection on MHC variability in C. nasus and P. toxostoma populations. In addition, a positive correlation between measurements of microsatellite variability and MHC diversity was found at the individual level. Nevertheless, our analyses also indicated strong positive selection acting on PBS in both species. The maximum likelihood codon-based analyses, which were performed to detect whether MHC IIB in each species was under selection and to identify the PSS, revealed a similar ω-ratio in the DAB-like genes of both species, indicating a similar strength of positive selection acting on MHC in native and invasive species. The majority of sites identified under positive selection were the same in both species and corresponded well to ABS identified for humans by Brown et al. [50], even if the total number of PSS was slightly higher in invasive C. nasus when compared with endemic P. toxostoma. This invasive species was also found to be more parasitized by Chondrostoma-specific parasites than P. toxostoma and recombinant genotypes of C. nasus and P. toxostoma hybrids [32].