The chicken MHC stands out with an unparalleled simple and compact architecture compared to mammals or other avian groups like the passerines [26, 36]. The tight linkage of MHC genes and reduced recombination rates were suggested to have resulted in a close co-evolution of genes within MHC haplotypes ( but see ), leading to strong disease-associations of specific MHC haplotypes in consequence. In the evolutionary history of the avian MHC it is still not resolved whether this unique MHC structure in the chicken is a domestication artefact or was present before domestication as a galliform feature.
We here present a study on diversity, selection patterns and the co-evolutionary history of two transcribed MHC class IIB genes in a wild galliform bird species, the black grouse. We found evidence for both intra- and interlocus recombination or gene conversion, as well as indication for positive selection on the PBR at both loci. However, we also detected differences in the selection between the two loci, as positive selection was indicated at the BLB2 locus and purifying selection at the BLB1 locus. We were able to relate our findings to the structure of the chicken MHC class II by resolving orthology to the chicken BLB loci.
Similar to the work of  on red jungle fowl, we amplified the two black grouse BLB loci separately in twelve individuals. Eleven of the individuals included in this study were cloned in our previous studies with non locus-specific primers that gave a 125 bp BLB exon 2 product [15, 50]. The resulting sequences showed that individuals carried between two and four BLB alleles, without locus assignment. This is a slight underestimation in number of alleles per individual compared to the here presented study, which revealed between three and four alleles per individual for BLB1 and BLB2 combined. Part of this additional variation was detected through the longer BLB2 fragment, which revealed further variable positions towards the 5′-end of the exon 2. To another part, the underestimation by the cloning/sequencing approach may be a result of too few analysed clones (16–26 per individual). On the other hand, for individual D248 the cloning procedure revealed one allele (BLB*125-04) which could not be confirmed by the single-locus amplification. This additional allele is likely to be a contamination during the cloning process. Considering that cloning is more time-consuming than direct sequencing, can add artefacts derived from mismatch repair or the formation of chimeric sequences  and is more contamination-prone, the here presented single locus sequence-based typing method clearly outmatches the cloning/sequencing approach in both accuracy and effort.
Another finding in this study concerns the expression of the two BLB loci in black grouse. We earlier amplified BLB cDNA sequences for individual D870 , which we could show now derived from both BLB1 and BLB2. Thus, both BLB1 and BLB2 are transcribed loci in black grouse. In domestic chicken BLB2 is considered to be dominantly expressed [44, 45], while BLB1 is less expressed and has even been suggested to be neutral to selection and not involved in peptide binding [30, 46]. This ultimately leads to the question whether the BLB1 and BLB2 loci in black grouse show a similar situation or differ in the degree and mode of selection compared to the chicken. We therefore compared diversity and selection patterns between BLB1 and BLB2 to explore their respective modes of selection.
We observed significant linkage between BLB1 and BLB2, in line with the hypothesis of Kaufman [30, 40] that MHC genes in chicken form a tightly linked cluster of co-evolving genes. Similarly, Agudo et al.  showed in a study on Egyptian vultures (Neophron percnopterus) the existence of linkage groups, containing pairs of MHC alleles in strong linkage disequilibrium. They interpret the presence of linkage groups with similar MHC alleles as indication of concerted evolution acting on the MHC gene duplicates. However, concerted evolution does not necessarily homogenize a gene cluster. Gene conversion can both homogenize and diversify among paralogues, as was shown for the gene family hsp70 in Drosophila.
Intuitively one might think that gene conversion and recombination should tend to break up linkage disequilibrium. But this does not seem to be this simple. The human olfactory receptor (OR) gene cluster is another gene family suggested to evolve under concerted evolution with gene conversion events . Linkage disequilibrium is highly significant in the centromeric part of the gene cluster, whereas no linkage disequilibrium could be observed in the telomeric part of the cluster. These considerable differences suggest the presence of several recombination hotspots within the OR cluster.
In terms of number of different alleles, MHC diversity was similar at the two BLB loci, however, BLB2 revealed higher heterozygosity and was less variable in terms of nucleotide diversity than BLB1. Signals of positive selection were found on both the BLB1 and BLB2 locus. Elevated dN/dS ratios were obtained at the PBR of both loci, though only significant for the longer BLB2 fragment. Maximum likelihood analyses confirmed that models allowing for positive selection fitted our data significantly better than neutral models. With these models, four identical codons were identified to be under significant positive selection in BLB1 and BLB2, and two additional positions were identified for the BLB2 locus. All of these positively selected codon positions were either congruent or directly adjacent to a peptide binding site as identified by Tong et al.  for human MHC class II molecules. This congruence emphasises the structural similarity of codon positions involved in peptide binding throughout evolutionary lineages. However, it has been suggested the methods used here might be prone to overestimate positive selection, particularly when recombination rate is high .
A different pattern of selection was found, however, regarding neutrality tests as Tajima’s D. We found a negative Tajima’s D value for BLB1, indicative of purifying selection, and a positive Tajima’s D for BLB2 indicating positive or balancing selection, although not significant in both cases. In a sliding window analysis of Tajima’s D comparing BLB1 and BLB2 it became more apparent that these different selection footprints are due to differences at some specific nucleotide sites only, while at other positions positive selection between the two loci coincides. These opposite but vague signals of differential selection at BLB1 and BLB2 made us explore this further. Worley et al.  found likewise a negative Tajima’s D for BLB1 and positive value for BLB2 (both n.s.) in a captive population of red jungle fowl but the authors points out that a negative Tajima’s D could also be a result of a population bottleneck. As a further comparison, we obtained domestic chicken BLB1 and BLB2 sequences from GenBank and calculated Tajima’s D the same way as we did for the black grouse sequences. This analysis repeated the observation of positive Tajima’s D for BLB2 and negative for BLB1 (data not shown). We also observed that theta k and π were higher in BLB1 than BLB2 in the domestic chicken, which was likewise observed for black grouse. In conclusion, we interpret this as a repeated pattern of differential selection on BLB1 and BLB2 in galliform birds.
Detecting recombination or gene conversion events using statistical methods can be highly problematic, in particular when small fragments are transferred or the gene conversion rate is too high. Hence, we have to keep in mind that a lack of evidence is no evidence for absence. To minimize the risk of missing the footprint of gene conversion, it is recommended to use multiple statistical methods for detection . Following this recommendation we applied a set of seven statistical methods, which have been evaluated to be powerful and accurate for different conditions and scenarios . While two of the methods (RDP and GENECONV) failed to detect most recombination signals, the other methods detected recombination or gene conversion both within and 'between the two BLB loci. As an additional hint for the occurrence of gene conversion, we detected significantly higher d
values at the PBR sites compared to the non-PBR sites at both loci. This situation is likely to be created by a combination of positive selection and gene conversion [78, 83]. Gene conversion events transferring advantageous non-synonymous substitutions at the PBR are positively selected, and in doing so synonymous substitutions are carried along. This way, synonymous substitutions appear more often at the PBR than expected by point mutations under neutrality. We conclude that both intra- and interlocus genetic exchange play an important role in shaping the black grouse MHC class II.
The 3′UTRs between BLB1 versus BLB2 differ in domestic chicken as well as between the corresponding DAB1 and DAB2 loci in pheasants in both length and nucleotide composition, so that the sequences cluster together as orthologous genes rather than according to species . We have observed similar differences in the 3′UTRs between black grouse BLB1 and BLB2 and could prove orthology of the BLB1 gene between the black grouse, chicken and pheasant, and orthology of the BLB2 gene between the three species. The origin of the pheasant like birds Phasianoidea is estimated to approximately 40 million years ago  and chicken and black grouse/turkey diverged approximately 30 million years ago . It seems that the duplication of BLB1 and BLB2 is a case of pre-speciation duplication that has arisen in the ancestral species before the split into chicken, pheasant and black grouse. The observed differences in the length of the 3′UTRs between the two loci may also reflect a difference in function .
In all other phylogenetic reconstructions based on the exon 2, the third codon positions of exon 2 and the exon 3, sequences clustered species-specific and not locus-specific. This is indication for frequent interlocus genetic exchange homogenizing sequences between the paralogous loci, as suggested by Wittzell et al. . The fact that the whole exon 2 sequences and third codon positions of the exon 2 networks present similar phylogenetic relations contains more indication for an early duplication in the evolutionary history of BLB1 and BLB2. Under convergent selection on the two loci, the two phylogenies would reveal deviating patterns . A phylogenetic tree based on the codon sequences should reflect functional similarities, whereas the third codon positions are expected to mirror the neutral gene history. Under the early duplication hypothesis, in contrast, a phylogeny based on the third codon positions will match the phylogenetic relation based on the whole codon sequences, as in our case. Note that in this analysis all third codon positions are considered and not only the synonymous positions at the PBR, which are possibly under a selective sweep of the positions under selection, as discussed earlier.
In summary, we infer that the BLB gene duplicated before the species divergence into chicken, black grouse and pheasant and thus is a case of pre speciation duplication. Further, we conclude that BLB1 and BLB2 in black grouse are subjected to homogenizing concerted evolution due to inter-genetic exchange between loci after species divergence. Different selection patterns indicated for BLB1 and BLB2 may be a sign of different immunogenetic functions. Both BLB1 and BLB2 have been under balancing selection during their evolutionary history. Nevertheless, it is likewise possible that, at present, balancing selection may be operating directly only on BLB2. Because of the tight linkage between the loci, BLB1 would be hitchhiking with BLB2.