Habitat isolation generates barriers to gene flow among populations that often result in loss of genetic diversity through genetic drift and inbreeding . For vertebrates, genetic variation is of special importance in MHC genes due to the significant role they play in immune functions . MHC diversity is presumed to improve parasite resistance, reproductive success, and population viability , and has been studied in species such as western gorilla (Gorilla gorilla) , brown bear (Ursus arctos) , Ethiopian wolf (Canis simensis) , European bison (Bison bonasus) , Bengal tiger (Panthera tigris tigris) , and Namibian leopard (Panthera pardus pardus) . In this study, a relatively high level of MHC variation was found in the golden snub-nosed monkey, with 9 DQA1 and 16 DQB1 alleles found in 64 individuals. However, we found lower MHC variation in the SNJ population, which also showed lower genetic diversity in microsatellites and mitochondrial genes [27, 34]. Small sample size cannot account for the reduced variability, because sample collected across the entire distribution of the SNJ population showed similar patterns of genetic diversity .
For each MHC locus, regardless of how many clones were sequenced from an individual, no more than 2 alleles were observed in an individual, a strong indicator that we amplified single loci in all cases. We assumed that all alleles were from a single functional gene. The assumption that our sequences were from functional genes was supported by three findings: 1) sites inferred to have been exposed to significant selection, most of which were ABS sites, indicating historical selection at functional genes; 2) no reading frame or stop codon disruptions found in any alleles; and 3) the sequences identified from cDNA were parts of or included sequences obtained from DNA.
Historical balancing selection
Golden snub-nosed monkey MHC genes reflect historical balancing selection in that an excess of non-synonymous substitutions was mainly concentrated in the ABS (Table 3, Additional file 4: Table S3, Additional file 5: Table S4). According to neutrality theory , the synonymous nucleotide substitution rate is larger than the non-synonymous substitution rate because a change in amino acid sequence has a greater possibility of being deleterious. The elevated rate of non-synonymous substitutions at the ABS provided clear evidence of positive selection [43, 44] shaping genetic variation . The p-value was not significant at DQA1 ABS may due to a weaker recombination within DQA1 as a lower recombination rate has been shown before. Without higher recombination and stronger selection, some DQA1 alleles might be lost when population size decreased. The sharing of MHC alleles among populations also indicates that MHC alleles may have been conserved by selection . Second, random site models analysis proved the existence of historical selection based on the maximum likelihood method, which revealed that, for MHC genes, the models including selection (M2a, M3, and M8) match MHC alleles better than models without selection (Table 4,and 5). Under the M2a and M8 models, some sites of the 2 MHC loci were under significant selection pressure. Furthermore, trans-species evolution of the MHC alleles revealed historical balancing selection. Under balancing selection, some MHC alleles or allelic lineages are reported in other species, which indicates that they are ancestral alleles .
Genes vary in terms of the level of selection, and both DQA1 and DQB1 revealed different patterns of selection. Each population had unique DQB1 alleles, while not every population had unique DQA1 alleles (Table 2). Population divergence, measured as pairwise F
, was larger in DQB1 than in DQA1 except for that between SG and QL (Figure 3). Similar results were reported in water vole, where balancing selection pressure was different at MHC genes in continuous populations . MHC genes are assumed to be closely linked . In our study, however, linkage disequilibrium between DQA1 and DQB1 was only observed in the SG population. This weak linkage disequilibrium and the different selection pressures on these loci could be a result of recombination, which is common at the MHC genes [49, 50]. The higher recombination rate that was found in DQB1 genes may explain their larger allelic richness compared to DQA1 genes. Recombinants maintained by selection may counteract the linkage of closely linked genes  and play an important adaptive role in DQB1 evolution. In the present study, historical selection was found, but this does not conclusively indicate that balancing selection is acting on current populations. First, an excess of non-synonymous mutations requires a long time to accumulate . Once present, this pattern would take a long time to vanish after the disappearance of selection . Hence, we investigated whether selection continues to play a major role at present.
Patterns of selection and drift
Although selection historically maintained MHC diversity, recent population isolation and fragmentation has increased the role of genetic drift in shaping patterns of MHC variation in snub-nosed monkeys. First, compared with neutral forces, balancing selection is supposed to diminish population differentiation as measured by conventional pairwise F
ST[22, 52]. Thus, the population structure of genes under balancing selection should not be pronounced . However, in the present study, half of the pairwise F
ST values were greater than 0.05, and two F
ST values at DQB were even greater than those at microsatellites (Figure 3). Second, F
ST outlier analysis showed that the structure in the MHC loci was within the neutrality level for all populations and for each population. Considering all populations, one microsatellite (D14S306) showed a F
ST value that was lower than the neutral level, indicating its linkage with other genes under selection . Lastly, a positive correlation was found between allelic richness in MHC and microsatellites. The QL population had the highest allelic richness in microsatellites and MHC, while SNJ had the lowest. This positive correlation indicates that genetic drift plays a significant role in maintaining MHC diversity for snub-nosed monkeys . Maintenance of MHC variation through balancing selection may be hampered in small, isolated populations because of their lower effective recombination rate . In all, our results indicate that even though selection acts on MHC, it is overwhelmed by genetic drift in small, isolated populations.
The positive correlation in the allelic richness of MHC and microsatellites, together with other evidence, indicates that genetic drift has a great influence on the maintenance of MHC variations in small, isolated populations of snub-nosed monkeys [7, 25, 55]. No evidence showed that MHC polymorphism had increased in populations that contained low neutral variation [27, 42]. Under neutral evolution theory , alleles are expected to be neutral when s < 1/2 Ne (s = selection coefficient, Ne = effective population size). Therefore, the smaller Ne becomes, the greater the likelihood of genetic drift . The SNJ population is subject to more genetic drift than the other two populations as found in a previous study . Other animals whose patterns of MHC polymorphism have been contributed to drift over selection include the great crested newt (Triturus cristatus) , black-footed rock-wallaby (Petrogale lateralis lateralis) , tuatara (Sphenodon spp.) , and the Egyptian vulture (Neophron percnopterus) . These results suggest that selection on MHC is not strong enough to counteract drift that results from population fragmentation, isolation and bottleneck.