Introduction

The genus Rupicapra (chamois) is a group within the subfamily Caprinae of the family Bovidae and is thought to have appeared about 20 MYA. It encompasses the two extant chamois species Rupicapra rupicapra (Alpine chamois) and Rupicapra pyrenaica (Pyrenean chamois) (Lovari 1987; Masini and Lovari 1988). The phylogenetic divergence time between the two chamois species was estimated to range between about 150 kyBP (cf. Masini and Lovari 1988) and 280 kyBP (Hammer et al. 1995). R. rupicapra consists of seven subspecies (rupicapra, tatrica, carpatica, balcanica, cartusiana, caucasica and asiatica), distributed among the Alps, the Tatra massif (Slovak Republic), the Carpathian Mountains (Romania) and various mountain massifs in the Balkans, Asia Minor and the Caucasus. R. pyrenaica consist of three subspecies (pyrenaica, parva and ornata) which are disjunctly distributed in south-west Europe (Pyrenees and Cantabrian Mountains) and the southern Apennine Mountain Range in Italy (Shackleton 1997). Normally, chamois live in small social groups throughout the year, with only older bucks remaining mostly solitary. The population dynamics of R. rupicapra populations in large parts of the Eastern Alps as well as populations in the Iberian peninsular (R. pyrenaica) and of Ibexes (Capra ibex, C. pyrenaica) is influenced by sarcoptic mange epidemics, an infection of Sarcoptes rupicaprae. Commonly, high mortality rates (80%) are noted in local populations in the course of regional epidemics (León-Vizcaíno et al. 1999; Rossi et al. 1995). Given this background chamois provide an interesting system for the study of the dynamics of host–parasite interactions and their effects on the evolution of immune genes in natural populations. Genetic polymorphism in immune-response genes may vary temporally and spatially in host populations, which makes it difficult for pathogens to adapt to all genotypes in a population (‘moving targets’). Genes of the major histocompatibility complex (MHC) are usually highly polymorphic and a striking example of selectively maintained polymorphic genes in vertebrates. The MHC genes encode cell-surface glycoproteins playing a key role in the vertebrate adaptive immune system. The main function of the MHC class II molecules is presenting peptide fragments derived from pathogens on the cell surface to T-helper cells. Subsequently, T-helper cells stimulate B cells to generate and secret antibodies. However, MHC molecules also play an important role in the shaping of the T-cell receptor repertoire during T-cell maturation in the thymus (Janeway et al. 2001). MHC class II molecules are α/β heterodimers and are mainly expressed on specialized antigen-presenting cells such as macrophages, B cells or dendritic cells. The α1 and β1 domains form the peptide-binding region (PBR) in which peptides are bound and recognized by T-helper cell receptors. The β1 domain is encoded by exon 2 of β-genes, and its genetic diversity can be extremely high at the population level, both in terms of the number of alleles present and in the extent of sequence diversity among alleles (Apanius et al. 1997). In humans, for example, 362 MHC class II DRB1 alleles are currently assigned (Robinson et al. 2003). In the PBR the number of non-synonymous substitutions usually exceeds the number of synonymous substitutions, an indication that positive selection is acting at PBR sites within genes, driving the diversification of MHC loci. It is thought primarily that some form of balancing selection (heterozygote advantage or negative frequency dependent selection) maintains the high polymorphism at the MHC (Hughes and Yeager 1998). Balancing selection may act to maintain ancient alleles and amino acid motifs in a trans-specific manner in mammals, that is, allelic lineages are passed from species to species and persist in the populations over long periods of evolutionary time (Klein et al. 1993). MHC loci are becoming increasingly well characterised for a growing number of ungulate species. However, the relative importance of different molecular mechanisms behind the generation of allelic diversity in the MHC remains contentious. Intragenic recombination (or homologous gene conversion) has been suggested as an important evolutionary mechanism for the generation of MHC sequence diversity (Andersson and Mikko 1995; Bergstrom et al. 1998; Gyllensten et al. 1991; Richman et al. 2003a,b). On the other hand, it has also been suggested that steady accumulation of point mutations and occasional convergence evolution cannot be ruled out as an important mode to create new MHC alleles (Klein et al. 1993; Klein and O’hUigin 1995; O’hUigin 1995; Takahata and Satta 1998). The direct measurement of recombination rate by sperm typing or pedigree analyses at a high resolution is technically difficult and time consuming. Moreover for wild mammals it is often impracticable, only really being feasible for laboratory or domestic species. However, the application of recently developed statistical methods to DNA sequences from natural populations now permits indirect, but quantitative, estimation of recombination rate from population–genetic data (Hudson 2001; McVean et al. 2002), allowing quantitative assessments of the relative contributions of different molecular mechanisms in generating allelic diversity at the MHC.

Mutation provides a constant influx of new MHC variants into the population, but a high recombination rate would speed up this influx. Therefore, high recombination rate in the MHC enables the host to keep up with the usually faster evolving parasites and might have beneficial fitness effects for the host. Therefore, we aimed to provide insights into the scale of the population recombination rate in chamois as compared to point mutations. After the divergence of the genus Rupicapra into Pyrenean chamois and Alpine chamois, both species might have experienced different intensity of adaptive evolution at the DRB locus. A further main objective was therefore to test for the intensity of diversifying selection in the PBR in both chamois species and to identify those sites that are probably under strong positive selection.

Materials and methods

Samples, DNA isolation and sequence data from GenBank

This study is based on the combined analysis of new data on 18 Pyrenean chamois (R. pyrenaica) from two Spanish populations in the Pyrenees (Rupicapra p. pyrenaica) and the Cantabrian Mountains (R. p. ornata) as well as one Italian population from the Abruzzo Mountains (R. p. parva) (for sample overview see Table 1) and the following sequences of MHC class II second exon alleles obtained from GenBank: 19 alleles from Alpine chamois (Rupicapra r. rupicapra, accession nos. AF324840–AF324861, Schaschl et al. 2004) and nine alleles from Pyrenean chamois (R. pyrenaica, accessions nos. AY212149–AY212157, Alvarez-Busto and Jugo, unpublished). Genomic DNA from the Pyrenean chamois samples was extracted with a DNA extraction kit (Deasy Tissue Kit, Qiagen) according to the manufacturer’s protocol.

Table 1 List of Rupicapra pyrenaica (Pyrenean chamois) subspecies used in the present study; observed Pyrenean chamois DRB alleles (Rupy-DRB) are given
Full size table

PCR amplification, cloning and sequencing

PCR amplification of the DRB exon 2 was achieved by following the protocol of Schaschl et al. (2004). PCR products were cloned into the pCR 2.1 TOPO plasmid (Invitrogen). Between five and eight clones per individual were sequenced. Sequences were determined on both DNA strands using BigDey Terminator Cycle Sequencing Kit v3.1 (Applied Biosystem) and an ABI 3100 DNA Sequencer. Samples with poor cloning efficiency were discarded, and sequences were only considered when they were found in two or more clones. Preliminary sequence processing and analysis was performed with BioEdit (Hall 1999).

Sequence analysis and nomenclature

All nucleotide and amino acid sequences were aligned with the program ClustalX (Thompson et al. 1997). Phylogenetic and molecular evolutionary analyses were conducted using MEGA, version 2.1 (Kumar et al. 2001). The relative frequencies of non-synonymous (dN) and synonymous (dS) substitution in the exon 2 were calculated according to Nei and Gojobori (1986) and applying Jukes and Cantor’s correction for multiple substitutions (Jukes and Cantor 1969). The significance of the difference between these rates was tested with a Z-test of selection at the 5% level, whereby the P-values are the probability of rejecting the null hypothesis of neutrality (dN=dS; Nei and Kumar 2000). In accordance with the proposed nomenclature for MHC in nonhuman species (Klein et al. 1990), we designated the exon 2 alleles Rupy-DRB for Pyrenean chamois (R. pyrenaica) with serial numbers attached.

Recombination/gene conversion analysis

Intragenic recombination involves the exchange of sets of DNA segments from the same gene, generating DNA blocks of compatible sites that are incompatible with contiguous DNA blocks.

The programme Geneconv, version 1.81 (Sawyer 1989; Sawyer 1999; available at http://www.math.wustl.edu/~sawyer/mbprogs/), was employed to find the most likely candidate alleles for intragenic recombination/gene conversion events in the genealogical history the genus Rupcapra. This method uses pairwise comparison of sequences in the alignment to find blocks of sequence pairs that are more similar than would be expected by chance. Genconv finds and ranks the highest-scoring fragments globally for the entire alignment. Global permutation test P-values of <0.05 (derived from BLAST-like global scores using 10,000 replicates) were considered as evidence of intragenic recombination. These global permutation test P-values have an intrinsic multiple-comparison correction for all sequence pairs in the alignment. The underlying method in Geneconv has a high statistical power for detecting recombination when recombination is present or likely to be present, while the risk of obtaining false positive results is low (Posada 2002).

The population recombination rate (ρ=4Ner) in chamois was estimated using the programme LDhat (see for details McVean et al. 2002). This programme implements Hudson’s (2001) composite–likelihood estimate approach to estimate the population recombination rate conditioned on the estimate of mutation rate per site (θ=4Neμ) from an approximate finite-sites version of the Watterson estimate. This method has been extended by McVean et al. (2002) to take into account high rates of recurrent mutations in sequences. The estimate of 4Ner is taken as the value that has the highest composite likelihood estimate (McVean et al. 2002). We used the implemented likelihood permutation test to test the null hypothesis of no recombination (ρ=0). Extensive computer simulation, carried out by Richman et al. (2003b), revealed that the LDhat estimates recombination rates fairly accurately with regard to sequences evolving under symmetric balancing selection. Furthermore, we calculated the ratio ρ/θ as an estimate of the relative amount of recombination compared to mutation, which is robust against several violations of the underlying coalescent model (Fearnhead and Donnelly 2001).

Test for positive selection using maximum-likelihood analysis

We used the programme CODEML of the PAML, version 3.14, package (Yang 1997) to test for the presence of codon sites affected by positive selection and to identify those sites. Positive selection is indicated by ω=dN/dS>1. The models considered in this study were M7 (beta), and M8 (beta and ω) (Yang et al. 2000). Under the model M7 (beta) the ω ratio various according to the beta distribution and does not allow for positive selected sites (0<ω<1) and thus serves as the null model by comparing with model M8 (beta and ω). Model M8 adds an additional site class to the beta model to account for sites under positive selection (ω>1). The models M7 and M8 can be compared in pairs using the likelihood-ratio test (LRT) (Nielsen and Yang 1998). The LRT statistics calculates twice the log-likelihood difference compared with a χ2 distribution with degrees of freedom equal to the difference in the number of parameters between the two compared models. The best tree for both species by maximum-likelihood search was in accordance with the one-ratio model (M0) used to provide phylogenetic information. A Bayesian approach implemented in CODEML was used to identify residues under positive selection in the MHC class II DRB sequences.

Results

Allelic polymorphism at the MHC class II DRB locus

We obtained seven MHC class II DRB exon 2 sequences (Rupy-DRB) from R. pyrenaica (Pyrenean chamois) samples (Fig. 1). A GenBank search revealed that three of our observed sequences were identical to three of the nine previously published Pyrenean chamois DRB exon 2 sequences (Rupy-DRB01, Rupy-DRB02, and Rupy-DRB04; see “Materials and methods” for details and accession numbers). The four novel DRB exon 2 sequences in our study were named in accordance with the published sequences as Rupy-DRB10 to Rupy-DRB13 (see Table 1), and were submitted to GenBank (GenBank accession numbers AY898752–AY898755). We included in our subsequent analyses the published R. pyrenaica sequences as well as published DRB exon 2 sequences from R. rupicapra (Alpine chamois). Nucleotide sequence variation among all pairwise comparisons of Rupy-DRB sequences, corrected for multiple substitutions, ranged from 0.9% to 8.6%, with a mean of 4.8±1.0% (±standard deviation). Mean nucleotide sequence variation among all pairwise comparisons of Alpine chamois DRB alleles is 4.4±0.9% and thus similar to Pyrenean chamois variation. The nucleotide divergence between Pyrenean chamois and Alpine chamois is 4.7±0.9%. In both chamois species, dN occurred significantly more frequently than dS at the PBR sites (Rupy-DRB: dN=0.188±0.051, dS=0.026±0.017, dN/dS=7.2, P=0.001; Ruru-DRB: dN=0.142±0.044, dS=0.010±0.007, dN/dS=14.2, P=0.0003). Among all the sequences in the current study, two Pyrenean chamois sequences were identical with two Alpine chamois sequences (Rupy-DRB02 and Ruru-DRB01 (also identical at the nucleotide level), and Rupy-DRB04 and Ruru-DRB13 (different in one dS at the codon position 84). In both chamois species the dS was far below the dS found in most other studied ungulates (Gutierrez-Espeleta et al. 2001; Jugo and Vicario 2000; Mikko et al. 1999; Van Den Bussche et al. 1999). This finding is probably a consequence of a young phylogenetic age of the alleles (that is, chamois DRB alleles have evolved more recently than that of other ungulates so far studied). It potentially reflects a complex demographic history of chamois populations (see “Discussion”).

Fig. 1
figure 1

Alignment of the putative amino acid sequences for MHC class II DRB exon 2 from R. pyrenaica (Rupy-DRB) (Pyrenean chamois) and R. rupicapra (Ruru-DRB) (Alpine chamois). Sequences were arranged to display similar DRB alleles together. Dots indicate identity in the amino acid sequence of the Rupy-DRB01, and a cross indicates codons involved in peptide binding regions (PBRs) in human (Brown et al. 1993). Thin arrows indicate PBRs under strong positive selection in R. pyrenaica, and thick arrows indicate PBRs under strong positive selection in R. pyrenaica and in R. rupicapra (>0.95 posterior probabilities)

Full size image

Level of intragenic recombination

The Geneconv analysis shows that intragenic recombination (or homologous gene conversion) events at the DRB locus have occurred in both species. In fact, intragenic recombination events were not only detected within segmental variants of each species but also between alleles of the two different species. Table 2 shows the Rupy-DRB and Ruru-DRB alleles that may have been involved in intragenic recombination events as well as the relative position of the DNA segments from 83 bp to 100 bp in length, which were involved in recombination. In total the analysis revealed that five Rupy-DRB and nine Ruru-DRB alleles were found to be involved in recombination events. Apparently, some sequence blocks (for example, DNA block 98–196 and DNA block 147–234, see Table 2) were repeatedly involved in recombination events and may have served as recombination hot spots. Further evidence for recombination in the chamois DRB sequences comes from the r2 correlation test for recombination (Awadalla et al. 1999) implemented in the LDhat programme. This test detected a significant (P<0.05) decay of linkage disequilibrium with pairwise distance, suggesting recombination in the analysed sequences. The population recombination rate (ρ) was estimated as ρ=78 in Pyrenean chamois and ρ=37 in Alpine chamois (Table 3). In both cases these values are very high, being an order of magnitude greater than the corresponding population mutation estimates (θ), and indicate a large contribution from recombination in the history of these sequences. The likelihood permutation test showed that the ρ estimates were significantly different from those expected under the null hypothesis of no recombination (P<0.001).

Table 2 Statistical test for recombination for chamois MHC class II DRB exon 2 alleles as assessed by the programme Geneconv, version 1.81
Full size table
Table 3 Statistic and P-values for population mutation (Watterson’s θ=4Nμ) and population recombination rate (ρ=4Nr, McVean et al. 2002)
Full size table

Detecting positive selection at sites using maximum-likelihood analysis

The LRT statistic comparing the two models indicates that M8 fitted the data significantly (P<0.01) better than M7. The estimates from M8 suggested that about 23% of the sites were under strong positive selection in the Pyrenean chamois sequences (ω=20.49) and 13% in the Alpine chamois sequences (ω=15.23) (Table 4). Bayes identification of sites under positive selection is listed in Table 4. Those sites that were identified as positively selected sites are mostly in accordance with the human PBR sites (HLA-DRB1 gene) (Brown et al. 1993; for PBR sites see Fig. 1).

Table 4 Log-likelihood values and parameter estimates for the MHC class II DRB alleles of chamois
Full size table

Discussion

In this study we found extensive sharing of amino acid motifs between the DRB alleles of the two extant chamois species. Further, two Alpine chamois DRB alleles are identical to two Pyrenean chamois alleles. If these two alleles do not result from convergent evolution, they could present shared ancestral alleles. Alternatively, relatively recent extensive hybridisation and introgression between the two chamois species (e.g. events of reticulate evolution) could explain these findings. However, the two species are geographically isolated by considerable distance and unsuitable habitat. There is no evidence of recent gene flow from neutral DNA marker studies (Hammer et al. 1995; Pérez et al. 2002), but some 20,000 years ago, during the late glacial maximum, chamois roamed over wide areas in central Europe (Sägesser and Krapp 1986). The late Pleistocene distribution of both species could have resulted in temporary contact or overlap of ranges and might have thus enabled introgression. Preliminary mtDNA and nuclear gene sequence data (Hammer et al., unpublished data) suggest such episodes of reticulate evolution in chamois. One of the two pairs of shared alleles among the Pyrenean chamois and Alpine chamois alleles is Rupy-DRB02 and Ruru-DRB01. The allele Ruru-DRB01 is the most common Alpine chamois DRB allele, with an overall frequency of 0.297 (Schaschl et al. 2004), while the Rupy-DRB02 is one of the most common Pyrenean chamois allele with a frequency of 0.222. Among the Pyrenean chamois subspecies, R. p. pyrenaica and R. p. parva, five DRB alleles were identified in this study plus six alleles previously. In contrast in the Pyrenean chamois subspecies R. p. ornata from the Italian population, only two DRB (Rupy-DRB12 and Rupy-DRB13) alleles have been detected, which were not found in the Pyrenean chamois samples from Spain. Hence the DRB alleles from the Apennine subspecies may represent novel alleles not present in the ancestral population.

The nucleotide variation in Pyrenean chamois is slightly higher than in Alpine chamois, a pattern that is repeated for other types of markers. In allozyme surveys Pyrenean chamois have marginally higher values of allozyme diversity than Alpine chamois from the Eastern Alps (Nascetti et al. 1985; Schaschl et al. 2003). A striking feature of the pattern of sequence variation in both chamois species is the low level of silent variation. This suggests a disproportionate loss of silent variation in species that may have been caused by one ore more bottlenecks, sometime during their population history. Similar observations and conclusions have been recorded for European and North American moose (Alces alces) populations by Mikko and Andersson (1995), for Madagascan lemur species (Go et al. 2002), and also for deer mouse (Peromyscus) species (Richman et al. 2003a). Chamois likely have experienced a complex demographic history, probably governed by multiple processes acting over ancient and contemporary time scales, generating changes in population size. This includes population expansions and contractions associated with Pleistocene glacial cycles, contemporary habitat fragmentation and reduction in population size due to human hunting pressure, and disease epizootics.

Population bottlenecks that reduce levels of allelic diversity at the MHC do not appear to be uncommon (Hedrick et al. 2000; Mikko and Andersson 1995; Mikko et al. 1999; Schmulder et al. 2003). Following such population bottlenecks, intragenic recombination provides a mechanism that could regenerate allelic diversity rapidly (Andersson and Mikko 1995).

In has been shown for the human genome that recombination rate occurs in a higher frequency in non-coding region than within genes (McVean et al. 2004). McVean et al. (2004) found also that 80% of recombination occurs in less than 10% of the human MHC sequence. Consequently, the human MHC is thought to be a recombination hotspot. Thus, density and intensity of recombination rate might be optimised for different sections in the genome. This study revealed incompatible sequence blocks in the sequences and a significant (P<0.05) decay in linkage disequilibrium with distance in several chamois DRB alleles, suggesting recombination in the MHC class II DRB gene. In both chamois species the estimated population recombination rate (ρ) differs significantly (P<0.001) from that expected under the null hypothesis of no recombination. In addition, the estimated recombination rate exceeds the estimated mutation rate (θ) by an order of magnitude. This indicates that the accumulation of new recombinant alleles greatly exceeds that of alleles derived from new point mutations and that intragenic recombination may have an adaptive significance in the evolution of the MHC class II DRB gene. Apparently, within exon 2, two DNA segments (positions 98–196 and positions 147–234) may have acted as hot spots for recombination. These DNA blocks were found in several alleles in either chamois species. Putative intragenic recombination events have taken place also between the two sets of species–specific alleles, suggesting frequent segmental sequence exchanges among the DRB alleles in the common history of the genus Rupicapra. The very high population recombination rate (ρ) in either chamois species indicates also that in their short intraspecific evolutionary time a rapid accumulation of novel alleles generated by recombination has taken place and the origin of some of these alleles might be relative recent. This fact also has consequences for phylogenetic inferences made from MHC sequence data, because the occurrence of recombination means that different parts of the alleles have different phylogenies. Schierup and Hein (2000) showed that with recombination the length of terminal branches, and the total branch lengths are larger and the time to the most recent common ancestor is smaller than for a phylogenetic tree reconstructed with no recombination. Therefore, conclusions based on MHC class II gene phylogenies should be considered with caution (e.g. trans-species polymorphism).

Finally, positive selection was determined for the exon 2 in both chamois species. In the Pyrenean chamois sequences about 23% of the sites were identified to be under strong positive selection (w=20.42), whereas in the Alpine chamois sequences, only 13% of sites were found to be under strong positive selection (w=15.23). We used only the M7 model, which acts as the null model versus the alternative model M8 (Yang et al. 2000). It has been shown that these two models are much more robust against the occurrence of recombination in the sequences than the other implemented models in CODEML (Anisimova et al. 2003). Furthermore, the Bayes’ prediction of sites under positive selection appears to be robust to recombination effects (Anisimova et al. 2003). The sites that were identified as positively selected sites were mostly in accordance with the putative human PBR sites (Brown et al. 1993). In the Pyrenean chamois sequences 14 sites were identified as positively selected sites out of 22 putative PBR sites. From those 14 sites only two (amino acid position 41 and 57) are not considered as putative PBR sites. In Alpine chamois only seven PBR sites were found to be under strong diversifying selection (Fig. 1; Table 4). However, the positively selected PBR sites shared between both species are identical.

In summary, the study revealed that the contribution of intragenic recombination for generating sequence polymorphism in the chamois DRB gene is about ten times high than point mutations. Thus, intragenic recombination coupled with strong positive selection are the main forces in generating sequence diversity in the MHC class II DRB gene.