Different Evolutionary Histories of the Two Classical Class I Genes BF1 and BF2 Illustrate Drift and Selection within the Stable MHC Haplotypes of Chickens¹

Chickens are assailed with a wide variety of pathogens, some of which also pose a threat to humans, including most recently influenza H5N1. Among the genetic loci potentially involved in the immune response, the chicken MHC can determine decisive resistance and susceptibility to several infectious pathogens, as well as the response to both live and killed commercial vaccines (1, 2, 3, 4). Compared with typical mammals such as humans and mice, the chicken MHC is simpler, smaller, and rearranged (5, 6, 7, 8). Based on these data, we developed the concept of a “minimal essential MHC” to give a molecular basis to the striking functional properties of the chicken MHC (7, 8, 9).

One key feature of this concept is the dominant expression of a single class I gene, which we have shown for three haplotypes, including the B12 haplotype (10). However, we have also shown that there are two classical class I genes in the MHC of the B12 haplotype (7). These two class I genes, now called BF1 and BF2 (11), are in opposite transcriptional orientation with their promoters outside next to the adjacent genes for DMB2 and C4, and with their polyadenylation sites close to the genes they flank, TAP1 and TAP2 (see Fig. 1).

As a hypothesis to explain why only one of the two class I genes is expressed at a high level, we proposed coevolution of the class I genes with other MHC genes (such as the TAP genes) involved in peptide loading (7, 8, 13). Such coevolution requires close genetic linkage (13, 14, 15) and, in fact, no recombination between the serologically defined class I and class II B genes has been observed in thousands of experimental matings (16, 17, 18, 19).

However, many questions remain unanswered. Is the high expression of a single class I gene at the RNA level a general feature of other common chicken MHC haplotypes? Is the genomic organization of other common MHC haplotypes the same, with two class I genes flanking the TAP genes? If there are two genes and only one is expressed at a high level, is it always the same gene? What are the sequence features that lead to one gene being expressed at a high level and the other at a much lower level? Do the two genes have separate evolutionary histories, or do they exchange information (for instance, by double-reciprocal recombination, microrecombination, or gene conversion)? If the two genes have separate evolutionary histories, then what forces shaped these histories?

The presence of a dominantly expressed class I molecule and the close linkage of class I and TAP genes appear to be features of many nonmammalian vertebrates (reviewed in Ref. 13), including at least some birds, frogs, bony fish, and cartilaginous fish. Thus far, multiple class I genes have been found in ducks and quails, with most of the genes expressed at low levels or not at all (20, 21, 22). So, the questions asked above may be relevant for many nonmammalian vertebrates.

In this study, we explore further the generality and the basis for the dominant expression of a single class I gene in the chicken MHC, eventually developing methods to amplify the whole of each gene separately and providing answers to each of the questions above.

The chicken lines H.B2 (MHC haplotype B2), CC (B4), CB (B12), H.B14 (B14), H.B15 (B15), H.B19 (B19), and H.B21 (B21) were bred and maintained at the Obergift Farm of the Basel Institute for Immunology until 2000. The MHC-homozygous lines H.B2, H.B14, H.B15, H.B19, and H.B21 were developed from Scandinavian White Leghorns. The highly inbred MHC-congenic lines CC and CB were developed from the Reaseheath line C at the Institute for Molecular Genetics (Academy of Sciences of the Czech Republic, Prague, Czech Republic) and are still available there and elsewhere. The chicken lines 6₁ (MHC haplotype B2), 7₂ (B2), C-B4 (B4), C-B12 (B12), WL (B14), 15I (B15), P2a (B19), N (B21), and 0 (B21) are bred and maintained under specific pathogen free conditions at the Institute for Animal Health (Compton, U.K.). The inbred MHC-homozygous lines 6₁, 7₂, 15I, N, and 0 were developed by the Regional Poultry Research Laboratory (East Lansing, MI). Lines C-B4 and C-B12 are sublines derived from the Reaseheath line C developed at the Northern Poultry Breeding Station (Reaseheath, U.K.). The Wellcome line (referred to as WL, Wl, or W lines in the literature) was developed by the Wellcome Research Laboratories (Beckenham, U.K.) and is not derived from the W line developed by the Northern Poultry Breeding Station. The P2a line at Compton was acquired from the Institute for Animal Science and Health (Lelystad, The Netherlands), and originated from the line P2a from Cornell University (Ithaca, NY). Some of these histories are reviewed in Refs. 23, 24, 25, 26, 27 . Bleeding chickens was conducted in accordance with Home Office Regulations and Local Ethical Review Committee oversight.

Procedures for blood cells were exactly as in Ref. 10 . Briefly, chicken B cells, T cells, and thrombocytes were isolated from peripheral blood by slow speed centrifugation, density cushion centrifugation, and FACS. RNA was extracted, cDNA was prepared, and then chicken class I α1 and α2 domain sequences were amplified by PCR using the primer 9447 (CGAGCTCCATACCCTGC), primer 9451 (CTCCTGCCCAGCTCAG), and Taq polymerase (7 s at 94°C, 15 s at 58°C, 2 min at 72°C for 30 cycles). The resulting single bands were cloned into dT-tailed pCRII plasmid vector in OneShot bacteria as per the manufacturer’s instructions (Invitrogen Life Technologies). Minipreps from randomly chosen clones were sequenced using dideoxy fluorescently labeled terminators and an Applied Biosystems 373A DNA sequencer.

RNA preparations from C-B12 spleen cells (unstimulated or stimulated with 5 μg/ml Con A for 4 h), and from C-B12 tissues (liver, thymus, and cecal tonsil), were converted to cDNA using the Superscript III kit (Invitrogen Life Technologies), amplified with primers c71 and c75 (c71 (CGAGCTCCATACCCTGCGGTAC, 60745-60765 and 76881-76301 in AL023516) and c75 (CTCCTGCCCAGCTCAGCCTTC, 61509-61490, 75537-75556)) with conditions as above but using a T3 Thermocycler (Biometra). The cDNA amplicons were cloned into pIST (a variant of pBluescript (Stratagene) adapted for PCR by adding two XcmI sites using a published procedure (28)). DNA minipreps of the clones were prepared, digested with XhoI (cuts both major and minor cDNA sequences), PvuII (cuts minor only), and/or HaeII (cuts major only), and analyzed by agarose gel electrophoresis and ethidium bromide staining.

DNA was isolated from erythrocytes using a salting-out procedure (29). Primer pairs used below are depicted in Fig. 1. Partial gene sequences were amplified from genomic DNA using specific primers in 30 μl using commercial reaction buffer (Proofsprinter; Hybaid) including 1.5 mM MgCl₂, 0.2 mM each dNTP, and 1 U of Taq/Pwo polymerase mixture (Proofsprinter; Hybaid), with amplification conditions of 1 min at 96°C followed by 30 cycles of 1 min at 96°C, 30 s at the reannealing temperature, 2 min at 72°C, followed by a final extension step at 72°C for 10 min on a Hybaid Sprint PCR machine. For sequences from class I exon 2 to exon 3, the details were 100 ng of genomic DNA, 30 pmol of each primer (c71 and c75, described above), and reannealing temperature of 60°C. For sequences from class I to adjacent genes, the details were 10 ng of genomic DNA, 30 pmol of each primer (c75 and c241 from DMB2 exon 6 (AGTGATGGTGTTGGGGCTCAG, 59477-59497); c75 and c350 from C4 exon 2 (AGGAGATGTGAGGTGACATGGGTGACATG, 77823-77795)), 5% DMSO in the reaction mix, and reannealing temperature of 63°C. The DNA fragments were purified following agarose gel electrophoresis using the Qiaex II gel purification kit (Qiagen), and cloned into the pIST vector described above. Transformation into DH5α, isolation of plasmid DNA, and sequencing of three clones were by standard techniques.

Whole gene sequences were amplified either using the Taq/Pwo Proofsprinter system (Hybaid) described above, or by using Pfx from Life Sciences (Invitrogen Life Technologies). For the latter, the whole gene or fragments of the gene were amplified from 20 to 500 ng of genomic DNA (depending on primer and allele) using 20 pmol of each primer in 50 μl using commercial reaction buffer including 0.5× commercial enhancer solution, 1 mM MgSO₄, 0.2 mM each dNTP, and 2 U of Pfx (Invitrogen Life Technologies), with amplification conditions of 2 min at 96°C followed by 30 cycles of 30 s at 96°C, 30 s at 66°C, 5.5 min at 68°C, followed by a final extension step at 68°C for 10 min on a T3 Thermocycler (Biometra). For the BF2 locus, the whole major gene was amplified using c69 (GCGGTGCCACTGAGTGCCACCAGGG, 63527-63503, 73558-73579) and c350 (above) or c348 (GCCAGAGTTCATCCTGGACAGCACTTCCAG, 72840-72869) and c350 (above). For the BF1 locus, the whole minor gene was amplified using c477 (GTTACGCCCCGCTTCCCGGTCACAACTAC, 59862-59890) and c480 (GCTCTTTGCCCGCTCACTCCACGCCAAC, 64459-64432), or c178 (TGCACAGGGAGATGTCCAGGCG, 60179-60200) and c73 (TGCACCCTGAGCAGCCAAACTGGG, 62852-62829, 74213-74236). The “coding region” (start codon to stop codon) of either gene was amplified by c395 (ATATAAGCTTTGCGAGGCGATGGGGCCGTGC GGGGCGCTG with the HindIII site underlined and the overhang region in italics, 60559-60585, 76490-76416) and c396 (CTAGTCTAGACACTCAGATGGCGGGGTTGCTCCCT with the XbaI site, 62599-62575, 74466-74490) or by c462 (GCATGTAAGCTTTGCGAGGCGATGGGGCCGTGCGGGGCGC with the HindIII site, 60559-60583, 76490-76463) and c463 (GCATGT TCTAGACACTCAGATGGCGGGGTTGCTCC with the XbaI site, 62599-62577, 74466-74488). PCR product purification and cloning into pIST, pcDNA3.1, or pSTBlue-1, followed by transformation into DH5α, isolation of plasmid DNA, and sequencing of at least three clones, were as above.

Genomic DNA from seven inbred lines of chickens with defined MHC type (line 6₁ (B2), line C-B4 (B4), line C-B12 (B12), line WL (B14), line 15I₅ (B15), line P2a (B19), line N (B21)) were digested with EcoRV and NotI, and Southern blots were performed as described, washing with 0.2× SSC (30, 31). The 550-bp DMB2 probe was produced by cutting a DMB2 genomic clone (derived from line 6₁ DNA by PCR, Ref. 32) with ApaI and HindIII, while the TAP1 probe was produced by cutting a TAP1 ABC genomic clone in pCR2.1 (derived from CBF23 cosmid DNA by PCR; Ref. 33) with EcoRI. Both probes were purified by agarose gel electrophoresis.

Alignments of cDNA sequences were produced using Pileup (GCG10) and trees produced by the neighbor joining method in the PHYLIP (http://evolution.genetics.washington.edu/phylip) software package (34). Alignments for gene and intron sequences were produced using AlignX in Vector NTI Advance 10 (Invitrogen Life Technologies). Human intron sequences were derived from www.anthonynolan.org.uk/HIG/seq/nuc/text/agen_nt.txt.

Comparisons were made using a web site (www.hiv.lanl.gov/content/hiv-db/SNAP) running the Synonymous/Nonsynonymous Analysis Program (SNAP; based on Refs. 35 and 36), which yields ds and dn, parameters related to the number of synonymous and nonsynonymous differences between multiple sequences (35).

We have previously reported that we found only one classical class I cDNA sequence from birds of the B15 haplotype, and two sequences from the B4 and B12 haplotypes, for which one (major) sequence was detected 10 times more frequently than the other (minor) sequence (10).

To confirm and extend these findings, we performed RT-PCR on RNA from three cell types derived from 10 egg-layer chicken lines (kept at the Basel Institute for Immunology) carrying eight MHC haplotypes and then counted the number of sequenced clones (Fig. 2). Altogether, 293 clones from 32 independent amplifications were analyzed (other details in the legend to Fig. 2). We found only one sequence in two haplotypes, B14 and B15. We found two sequences in the other haplotypes, B2, B4, B6, B12, B19, and B21 (and related recombinant haplotypes R1 and R2), of which one was present as much as 10 times the frequency of the other in most amplifications (with the average proportion of 0.85 for the major sequence in all of the blood cell samples taken together). After this work was done, a separate study also reported that both B19 and B21 express two sequences, of which one was roughly 10-fold more frequent than the other (37).

Similar analysis of B12 spleen cells (from the C-B12 subline kept at the Institute for Animal Health) shows that stimulation with the mitogen Con A results in an even greater proportion of one sequence (Fig. 2). Analysis of three B12 tissues shows that liver and thymus give similar proportions as blood and spleen cells, but that the proportion is less dramatic for cecal tonsil in this experiment (Fig. 2). Given the results with cecal tonsil and some of the thrombocyte samples (see legend to Fig. 2), it is possible that there are changes in the proportion of the two sequences in different tissues or under other kinds of stimulation.

Thus, there is a dominantly expressed class I molecule at the RNA level in many chicken MHC haplotypes, and in several tissues and conditions. As before (10), we called the more and less abundant cDNAs the major and minor sequences, respectively.

We previously reported that there are two classical class I genes present in the B12 haplotype (7) and that the major B12 sequence derives from the BF2 gene and the minor sequence derives from BF1 gene (10).

To confirm and extend these findings, we first performed PCR on genomic DNA using oligonucleotide primers based on conserved regions of the cDNA sequences to amplify exons 2 and 3 (encoding the α1 and α2 domains). We examined genomic DNA from the same birds kept at the Basel Institute for Immunology as were used for the cDNA experiments above. We also examined genomic DNA from birds kept at the Institute for Animal Health (Compton, U.K.), nine lines carrying seven MHC haplotypes that type serologically the same as those from Basel, but with very different histories and genetic backgrounds. Just as we did for the cDNA sequences above, we found only one genomic sequence in the B14 and B15 haplotypes and two genomic sequences in the B2, B4, B12, B19 and B21 haplotypes. Moreover, the appropriate sequences derived from the Basel and Compton birds were identical and differed from the cDNA clones only in having the intervening (229 bp) nearly invariant intron.

We then performed long-distance PCR on genomic DNA derived from the Compton birds (Fig. 3 and data not shown), and sequenced the products (data not shown). One primer was designed to match the conserved sequence at the end of exon 3 of the class I genes and the other primers were designed to match sequences in or near genes that are adjacent to the class I genes in the B12 genomic sequence. In all haplotypes tested, we were able to amplify a product between a class I gene and the C4 gene; in all cases, the genomic sequence corresponded to the major cDNA sequence. In all haplotypes except B14 and B15, we were able to amplify a product between a class I gene and the DMB2 gene; in all these cases, the genomic sequence corresponded to the minor cDNA sequence.

Thus, in a range of lines and haplotypes, there is a dominantly expressed (major) class I molecule at the RNA level that is encoded by the BF2 gene, located between TAP2 and C4. In most haplotypes, there is a poorly expressed (minor) class I molecule at the RNA level that is encoded by the BF1 gene, located between DMB2 and TAP1.

To understand the basis for the difference in RNA levels between the BF1 and BF2 genes, we examined the promoter sequences that were amplified by long-distance PCR (Fig. 4). All the promoter sequences of the BF2 genes have nearly identical sequences (13 single nucleotide polymorphisms (SNPs)¹⁰ in 361 bp presented in this figure), and have the transcriptional start sites and transcription factor binding sites originally identified for the B-FIV promoter of the B12 haplotype (38), including enhancer A, IFN regulatory element, and W/S, X, X2, and Y boxes. In contrast, we found two kinds of BF1 gene promoter sequences, both of which appear disabled.

The BF1 promoter sequences of the B12 and B19 haplotypes are identical with each other and are nearly identical with the proximal BF2 promoters, including the IFN regulatory element, W/S, X, X2, and Y boxes, and the major start-points of transcription. However, the sequence identity abruptly ends in the distal promoter region at position −202 (from the first nucleotide of the translation start codon), with a 272-bp deletion between the DMB2 gene and the BF1 gene, so that the enhancer A element is simply not present. The enhancer A element is known to be required for high-level transcription of mammalian class I promoters (40).

The BF1 promoter sequences of the B2, B4, and B21 haplotypes are nearly identical with each other (eight SNPs in 200 bp) and are nearly identical with the BF2 promoters, apart from two important differences. The enhancer A element is diverged, differing by over half of the nucleotides (and including an insertion, depending on where certain small indels are placed). In any case, the changes would be expected to reduce function based on mutagenesis experiments with mammalian class I promoters (41). A second difference is that there are one or more deletions (depending on how the sequences are aligned) in the very proximal promoter, which include the start-points of transcription based on the BF2 gene of the B12 haplotype (−25 and −58, Ref. 38). Upstream of the enhancer A sequence, there may be sequence stretches related to the κB2 site that have been described for human class I promoters (41), but they appear very diverged as well.

The deletions in the BF1 (minor) gene promoters explain the differences in the size of fragments amplified by PCR (Fig. 3) or after digestion with restriction enzymes in the Southern blots described below. The large deletion that removes the enhancer A element from the B12 and B19 minor genes can be explained by simple homologous recombination between two copies of a decamer sequence (GACTCCGTGC). This sequence is found in one copy at one end of the B12 and B19 deletion, but in two copies in the appropriate positions in the B2, B4 and B21 minor genes (both of which are diverged in all of the major genes). The similarity between the upstream regions of the minor and major genes continues at a low level (55–60% identity) all the way to the adjacent genes.

Despite many attempts in different ways, we were unable to amplify an upstream region for the BF1 gene from the B14 or B15 haplotypes. To approach the basis for the defect in these haplotypes, we compared the BF1 genes of all the haplotypes by digestion with a variety of restriction enzymes followed by Southern blots hybridized with probes from the adjacent DMB2 and TAP1 genes. As illustrated for the double digest with EcoRV and NotI (Fig. 5), the B2, B4, and B21 haplotypes share the same patterns of bands. The B12 and B19 haplotypes share patterns of bands that are smaller than those of B2, B4, and B21 by ∼300 bp, as expected from the sequences determined above. The B14 and B15 haplotypes share patterns of bands that are larger than those of B2, B4, and B21 by ∼4 kbp, suggesting the presence of an insertion. Sequencing of TAP and DMB2 genes confirms that the relevant EcoRV and NotI sites are present in the B14 and B15 haplotypes. Thus, it appears that there is some insertion in the BF1 gene of the B14 and B15 haplotypes, but the restriction patterns of all the enzymes taken together indicate a complicated rearrangement as well (data not shown).

To compare the sequences of the whole genes, we developed a number of PCR procedures to amplify separately either the whole BF1 gene or the whole BF2 gene, using primers located outside the genes (details in Materials and Methods). These amplifications were successful for all of the class I genes except for BF1 (minor) genes of B14 and B15. Sequences from representative clones (corrected for the errors discovered by comparison with multiple clones) have been deposited in the nucleotide databases (accession numbers in legend to Fig. 6), from −361 to +2793 from the first nucleotide of the start translation codon (that is, from roughly 120 bp upstream of the enhancer A sequence to roughly 35 bp downstream of the second polyadenylation site).

A comparison between these 12 sequences and their consensus shows that the differences are not distributed evenly along the genes (Fig. 6). Most striking is the nearly invariant intron 2 located between the highly variable exons 2 and 3 that encode the peptide-binding domains. The regions upstream of exon 2 and downstream of intron 3 are also relatively conserved. However, the gene is relatively variable downstream of exon 6, although the protein coding regions (exons 6, 7, and part of 8) are nearly invariant (parenthetically, the first polyadenylation site is diverged in the BF1 gene of the B2 haplotype as well as the BF2 genes of the B14 and B15 haplotypes).

Overall, the variability in the whole BF1 (minor) genes follows the relationships in the cDNA sequences (and the promoter sequences): B12 is nearly identical with B19 (differing in the positions of five SNPs and one indel), B4 is nearly identical with B21 (differing in the positions of four SNPs) and related to B2 (differing in the positions of many SNPs and several indels). Overall, the variability over the whole BF2 (major) genes groups differently (and not so closely), but follows the relationships in the cDNA sequences: B2, B4, and B12 form one group (indeed, B2 and B12 are nearly identical over much of the sequence), B14, B15, and B19 form another group, and B21 is most different throughout the sequence.

We previously reported that the minor cDNA sequences from B4 and B21 are identical, that the minor cDNA sequences from B12 and B19 differ by one nucleotide, and that we were unable to amplify cDNAs for a second class I gene from the B14 and B15 haplotypes (10). These patterns are nearly identical with the groupings based on the BF1 promoter/gene defects and on the variability for the whole BF1 genes: B4 and B21 (and B2) vs B12 and B19 vs B14 and B15, as described above. This prompted us to quantify these relationships by phylogenetic analysis (Fig. 7).

There is no similarity between the trees for the BF2 (major) and BF1 (minor) class I sequences. However, there is a striking similarity between the BF1 (minor) class I tree and the BLB1 (minor) class II β-chain tree, based on sequences from the same lines of chickens (42). In contrast, there are no similarities between the trees for the BF2 and BLB2 (major) sequences or between the trees for either of the major sequences and the trees for the minor sequences. Thus, BF1 and BLB1 (minor) genes appear to have evolved together.

One explanation for this surprising finding is that the minor class I gene and minor class II β-chain gene in each haplotype are coevolving, but thus far there is no precedent or mechanistic reason why this might be so. A simpler explanation is that both these minor genes, being poorly expressed, are not under much if any selection, and the allelic variation is mainly the result of neutral changes accumulated over time. In this view, the trees for the minor genes represent the history of the haplotypes through descent from a common ancestor, while trees for the BF2 and BLB2 (major) genes represent predominantly the effects of selection (as well as other forces). In agreement with this view, the trees for the minor genes are similar to other genes in the chicken MHC which may not be under much selection (D. A. Marston, B. A. Walker and J. Kaufman, unpublished). Moreover, the ratio of synonymous (silent) to nonsynonymous (replacement) differences for the minor genes is what would be expected for genes that are not under selection (for BF1, ds/dn is 1.28 for all differences and 1.79 for presumed peptide contacts).

In this study, we explore further the generality and the basis for the expression of a single class I gene in the chicken MHC and report the following findings. Seven common chicken MHC haplotypes have the same genomic organization of two class I genes flanking the two TAP genes. For each haplotype, it is the BF2 gene that is dominantly expressed at the level of RNA, which correlates with an intact and nearly invariant promoter. We found three different ways in which expression of the BF1 gene was crippled, two involving independent deletions in the promoter region and one involving an insertion/rearrangement leading to a pseudogene. The patterns of nucleotide polymorphisms in the whole BF2 gene follow those of the peptide-binding domains, and are completely different from the BF1 gene. The patterns of nucleotide polymorphisms in the whole BF1 gene match the three independent sequence features which lead to poor expression and match the dendrograms of the poorly expressed class II B gene (BLB1). These data show that, despite their many sequence similarities, the dominantly expressed BF2 locus and the poorly expressed BF1 locus have separate evolutionary histories, with the BF1 gene illustrating the descent of stable haplotypes, a prerequisite for coevolution of genes leading to a single dominantly expressed class I gene, which is a cornerstone of the minimal essential MHC hypothesis.

Below, we discuss three points that arise from our findings: the function (if any) of the minor gene, the importance of the conserved regions (particularly intron 2) within the gene, and, finally, the importance of haplotypes to coevolution.

Both classical class I genes in the BF/BL region are polymorphic and diverse, with all the sequence features expected for functional class I Ag-presenting molecules. In particular, the BF2 gene is well-expressed from an intact and nonpolymorphic promoter, is highly polymorphic with a different sequence for every MHC haplotype, and has dendrograms that are not similar to those for apparently unselected DNA. In contrast, several observations indicate that the BF1 (minor) class I gene is under less selective pressure for function compared with the BF2 (major) class I gene. First, the BF1 gene is expressed at a level much lower than the BF2 gene, as though there is less selective pressure for sufficient numbers of cell surface molecules to be recognized by T cells. Second, there are three apparently independent events leading to different molecular bases for the lower level, as though there was significant pressure to reduce the number of functional class I genes. Third, there are fewer alleles of the BF1 gene (some of which are frank pseudogenes) than the BF2 gene, as though there is less selective pressure driven by pathogens on BF1 (consistent with the lower level of expression). Fourth, the ratio of nonsynonymous (replacement) to synonymous (silent) substitutions in the coding region of the BF1 gene is consistent with a low level of selection. Fifth, the phylogenetic relationships of the BF1 alleles (but not the BF2 alleles) are the same as the BLB1 (minor) class II β-chain alleles (and other chicken MHC genes apparently under little or no selection).

This view is consistent with two previous studies on chicken class I sequences. Hunt and Fulton (43) analyzed cDNAs presumed to be from the BF2 (then called B-FIV) gene of 11 egg-layer lines and concluded that the BF2 gene is under strong selection. Livant et al. (44) analyzed exon 2 to exon 3 genomic sequences from 16 MHC haplotypes from broiler lines and conclude that there are two groups of sequences, one of which has fewer alleles, lower diversity, and much less selection than the other. They propose that the poorly selected sequences are derived from the BF1 gene, based on our previous work (7, 8), and suggest that this gene is involved in recognition by NK cells, based on a motif in the α helix of the α1 domain which in mammals is implicated in recognition by killer Ig receptors. Chickens have recently been shown to have a large family of KIR genes in the leukocyte receptor complex (45, 46). The low level of allelic polymorphism and the very low level of expression that we have found for the BF1 gene are not inconsistent with such a function, but it cannot be an essential function, given the presence of pseudogenes in at least two haplotypes.

One possibility that we considered early on was that the BF1 gene is a reservoir for diversity transferred to the BF2 gene by recombination or gene conversion (as has been suggested for mammalian class I genes, Refs. 47 and 48), or for concerted evolution (as found for blackbird class II B genes, Ref. 49). The fact that the genes were in opposite transcriptional orientation (facilitating intrachromosomal crossing over without loss of either gene) and that the most polymorphic exons were flanked by conserved regions (facilitating gene conversion or double-reciprocal microrecombination) made these possibilities attractive. However, both possibilities now seem highly unlikely, given the fact that the trees for the two class I genes (either the peptide-binding exons or the whole gene including introns) are completely different.

Thus, it seems likely that most of the polymorphisms in the BF1 gene (including the peptide-binding regions) are a matter of drift rather than selection. However, many of the BF1 alleles examined in this report have a distinctive constellation of residues in the peptide-binding site, similar to those found in the BF2 molecule of the B21 haplotype (10). These residues might be under selection to bind a particular set of peptides, constitutively to bring the BF1 molecule to the surface as a ligand for NK cells, or inducibly under certain conditions of disease. However, again this cannot be an essential function, given the presence of pseudogenes in some haplotypes.

The second point concerns the striking finding that the sequence diversity of the introns (and 3′ untranslated region (UTR)) varies across the genes. The highly variable exons 2 and 3 are flanked by much less variable regions (intron 1 on one side and exon 4 to intron 4 to exon 5 on the other) and separated by the nearly invariant intron 2. In contrast, there are highly variable introns (and 3′UTR) between highly conserved exons 5–8. The level of diversity in different exons is easily explained in terms of selection (for diversity in exons 2 and 3, and against diversity in other exons). However, there is no obvious reason why the introns are not all equally diverse, following either descent (for BF1) or hitchhiking (for BF2), and why the conserved regions have the same sequences for both genes.

There were no obvious clues to functional or structural constraints on the nearly invariant intron 2 found by using web-based analyses for open-reading frames, structural RNAs, matrix attachment sites, or transcription factor binding sites (data not shown). The possibility that the conserved introns promote recombination between the BF1 and BF2 loci was also not tenable, as discussed above. However, intron 2 has been found to be important in regulating transcription in response to IFN and during embryonic development (50, 51, 52). Certain transcription factor binding sites have been implicated in this regulation, only some of which are known to be important in immune system function (53, 54, 55). Moreover, in human classical class I genes, the regions flanking and separating exon 2 and exon 3 are reported to be much less polymorphic than expected (47). Intron 2 is nearly the same size in humans, mice, and chickens (∼220–250 bp), and aligns with over 50% nucleotide identity, but few transcription factor binding sites are in common between humans and mice and none with chickens (data not shown). So, if the lack of diversity in certain introns is an evolutionarily stable feature, then different transcription factors are involved in each species, consistent with developmental differences between species.

The third point concerns the important finding that particular alleles of BF1 and BLB1 (and other genes in between; D. A. Marston, B. Walker, and J. Kaufman, unpublished data) travel together through evolution as MHC haplotypes, based on dendrograms and on sequence features such as deletions in the promoters. In fact, no recombinants between the chicken BF (class I) and BLB (class II) genes have been found in several experiments involving thousands of matings (16, 17, 18, 19). This situation is clearly different from humans (and all mammals examined), in which the MHC is huge (4 Mbp) and broken by relatively frequent recombination (2–4 cM across the whole MHC) (56, 57). Given the low level of recombination observed, the chicken MHC haplotypes may be very old, which is consistent with the evidence for B21 haplotypes being shared between domesticated chickens and Red Jungle fowl (37). These data are also consistent with the evidence that some genes (such as BF2, TAP1 and TAP2, tapasin) are coevolving for function (Refs. 7 and 8 ; A. van Hateren, A. Williams, J. Jacob, T. Elliott, and J. Kaufman, submitted for publication; B. Walker, A. van Hateren, and J. Kaufman, unpublished observations), giving further support for the concept of a “minimal essential MHC” of chickens. However, if this view is not correct, then there may be some unexpected connection between the evolution of the poorly expressed class I (BF1) and class II B (BLB1) genes.

We thank Mark Dessing and Brit Johansson for excellent technical assistance, Drs. Sally Rogers and Tuang Yeoh Poh for the kind gift of RNA preparations, Mick Gill for invaluable help with figures, Dr. Steven G. E. Marsh for assistance with human genomic sequences, Sally Rogers, Shirley Ellis, John Young, and Gillian Griffiths for critical reading of the manuscript, and two reviewers whose comments greatly improved the paper.

The authors have no financial conflict of interest.