At Least One Class I Gene in Restriction Fragment Pattern-Y (Rfp-Y), the Second MHC Gene Cluster in the Chicken, Is Transcribed, Polymorphic, and Shows Divergent Specialization in Antigen Binding Region^1 2

MHC class I genes are often described as being either classical (class Ia) or nonclassical (class Ib). Class Ia genes are ubiquitously expressed encoding the polymorphic transplantation Ags that are the basis of rapid graft rejection. They include human HLA-A, -B, and -C; murine H2-K, -D, and -L; two loci in the chicken B system; and additional loci in a number of other species. Class Ia molecules present peptide-Ag to CTL. At the cell surface class Ia molecules are tripartite, composed of the polymorphic heavy chain, bound peptide, and β₂-microglobulin (β₂m).⁵ By contrast, the MHC class Ib genes, such as human HLA-E, -F, -G, H, I, and J; murine H2-Q, -T, and -M; mammalian CD1, Hfe, and Xenopus NCI genes, are generally not polymorphic in the manner of class Ia; but, there are exceptions, such as Q2 (1). Nonclassical molecules generally have weak influences in graft rejection. Some class Ib loci lie within or near the MHC, and a portion of these are phylogenetically close to class Ia loci. Other, usually older class Ib loci are located in paralogous regions on entirely different chromosomes. Many class Ib molecules are restricted in expression to particular tissues. Some class Ib molecules have critical roles in regulating NK cell activity (2), and trafficking of these to the cell surface can be dependent upon the binding in the Ag binding region (ABR) of peptide derived from signal sequences of other class I molecules (3). Still others, such as FcRn (4) and Hfe (5), function in the delivery of particular molecules across cellular boundaries via molecular interactions independent of the ABR. Although divergent to various degrees in primary sequences, the class Ib and class Ia heavy chain molecules share many elements of tertiary structure in common. Many, but not all, class Ib molecules bind β₂m.

A characteristic often useful for distinguishing between class Ia and class Ib molecules is a set of 8 aa present in the α1 and α2 domains of the class I heavy chain (6, 7). These amino acids in the ABR are effectively invariant in all class Ia molecules and contrast strikingly with the many polymorphic residues in the region. These residues are essential in sequence-independent anchoring of peptide-Ag. Substitutions are present at one or more of these eight positions in nearly all class Ib molecules, although a few exceptions to this are found among H2-Q and H2-T loci. In some class Ib molecules, particular substitutions are associated with anchoring special forms of Ag (8). For example, N-formylated peptides (9) are bound by the mouse-nonclassical H2-M3 molecules in an ABR made more hydrophobic by the presence of phenylalanine and leucine at two of the eight critical positions. Glycolipids and lipoglycans are presented by transporter associated with Ag processing (TAP)-independent CD1 molecules (10, 11). In the ABR of CD1 there are multiple, generally hydrophobic substitutions occurring at the eight residues. Other substitutions are present in other class Ib molecules and are associated with other modifications of the ABR, such as closing of the groove in FcRn (12).

Restriction fragment pattern-Y (Rfp-Y) and B are two genetically independent clusters of MHC class I and class IIβ genes in the chicken that map to a single microchromosome (chromosome 16) (13, 14, 15, 16). Also mapping to chromosome 16 is the single nucleolar organizer region found within the chicken genome (17) and a single nonpolymorphic classical class IIα locus (18). A chromosomal region supporting highly frequent meiotic recombination, perhaps associated with the nucleolar organizer region, separates B and Rfp-Y such that the two clusters are genetically unlinked even though they are located on the same microchromosome (15, 16). This arrangement is quite different from the arrangement of class I loci in the mouse into H-2K and H-2D where the loci remain linked despite physical separation. Rfp-Y was detected initially when two sets of polymorphic restriction fragments revealed by B system class I and class II probes were found to assort independently of one another in families of fully pedigreed animals (13). Rfp-Y was later found to correspond to the cosmid clusters II/IV and III in the molecular map (19) of chicken MHC genes (14, 15).

At least two class Iα heavy chain genes (YFV and YFVI), three class IIβ genes (YLβIII, YLβIV, and YLβV) (20), a c-type lectin gene (21), and two additional genes (13.1 and 17.8) of unknown identity map to Rfp-Y. The classical B system is a compact gene region that determines rapid allograft rejection. A large portion of the B cluster has been sequenced and found to contain 19 genes within a 92-kb region, virtually all of which have counterparts in the human MHC (22). Included among these are two polymorphic class I heavy chain and two polymorphic class IIβ loci (19, 23, 24, 25).

Because the small number of chicken class Iα heavy chain and class IIβ chain genes are essentially equally divided between B and Rfp-Y, the two clusters might originate from duplication of an entire chromosomal segment providing duplicate sets of loci with similar functions. Alternatively each may perform specialized, complementary functions as is becoming apparent for mammalian classical and nonclassical regions (8). For example, the Rfp-Y loci might in some instances provide molecules supplementing the less than comprehensive Ag presentation that is so characteristic of the B system (26). To begin to define the basis of the organization of chicken MHC genes into two genetically independent clusters, we subcloned and fully sequenced the YFV and YFVI loci located in the Rfp-Y cosmid cluster map. We analyzed the sequences of these loci with respect to those of known class Ia and class Ib loci and evaluated their polymorphism. We extensively analyzed gene transcription and the capacity of YFV cDNA to produce mature Rfp-Y class I molecules upon transfection.

Clone cβ10 was isolated from a cosmid library made from line CB (B*12, Yw*7.1) (27). (We use the w* notation with all Rfp-Y haplotypes to indicate that assignments are subject to further refinement.) For transcription analysis 12- and 19-day-old line CB embryos were provided by Pierrick Thoraval from stock maintained at Institut National de la Recherche Agronomique (Nouzilly, France). One-year-old male birds from line C were provided by Larry Bacon from stock maintained at the US Department of Agriculture Avian Disease and Oncology Laboratory (ADOL) (East Lansing, MI). Lines C and CB both originate from the Reaseheath line RH-C. Lines CB and C are inbred and homozygous for the same B and Rfp-Y haplotypes. Other Rfp-Y haplotypes analyzed include Yw*1.3, Yw*2.1, Yw*3.1, and Yw*6.1 from Northern Illinois University (13); Yw*4.2 and Yw*5.3 from University of New Hampshire (R.L. Taylor, Jr. and M.M., unpublished data); and Yw*7.2, Yw*8.1, and Yw*9.1, provided by Larry Bacon (28).

The 3.55- and 4.8-kb BglII fragments of cβ10 (27) containing the YFVw*7.1 and YFVIw*7.1 genes, respectively, were subcloned into the BamHI site in Bluescript II KS (Stratagene, La Jolla, CA) to provide pYFVw*7.1 and pYFVIw*7.1. Sequences of the YFVw*7.1 and YFVIw*7.1 genes were obtained through the successive application of two techniques. First, the sequences of exons 2, 3, 4, and 5, and introns 2, 3, and 4 of each gene were defined by sequencing of products obtained by PCR with primers designed from the B-FIV*12 gene sequence (29). Two hundred nanograms of cosmid cβ10 clone DNA, 10 ng of pYFVw*7.1 and pYFVIw*7.1 plasmid DNA were used as templates. PCR amplifications consisted of 35 cycles of 95°C for 1 min, 60°C for 45 s, and 72°C for 45 s using Taq DNA polymerase buffer with 400 nM of each primer, 200 μM of each dNTP (Pharmacia, Piscataway, NJ), and 1 U of Taq DNA polymerase (PE Biosystems, Foster City, CA). A fraction of each reaction product was cloned using the TA cloning vector (Invitrogen, San Diego, CA), and the nucleotide sequence of the insert was fully determined by dideoxy chain termination with a Prism Ready Reaction Dye Terminator Cycle Sequencing Kit and a 370A DNA Sequencer (PE Biosystems). To identify any errors due to misincorporation by Taq polymerase, three to six independent PCR were conducted for each primer set and 4–10 clones per PCR were fully sequenced. The sequences upstream of exon 2 and downstream of exon 5 were obtained by direct sequencing of the pYFVw*7.1 and pYFVIw*7.1 plasmids with annealing primers designed from previously determined sequences.

A YF-specific clone, 163/164f, was generated by PCR from YFVw*3.2 DNA and corresponds to exons for Cy1, Cy2, Cy3, and portions of surrounding introns of YFV (see Fig. 1). Clone 163/164f was used in turn to isolate a full-length cDNA clone, c36f, from a cDNA library made from the small intestine of a UCD line 330 young adult bird. An additional clone, cos2, was also obtained by 163/164f screening from a SuperCos I cosmid library (Stratagene) produced from a bird (wb3078) heterozygous for two additional Rfp-Y haplotypes designated Yw*1.1 and Yw*5.1. Cos2 was determined to originate from Yw*5.1 by restriction fragment pattern. A YFV subclone, pcr75171–3, was derived from cos2 and sequenced from the clone margins.

RNA was extracted from frozen tissues with RNAzol B (Tel-Test, Friendswood, TX). For 12-day-old whole embryos and 19-day-old embryo tissues RNA was purified on cesium chloride gradients.

Samples containing 10 μg of genomic DNA were digested with restriction endonucleases, electrophoresed in 1% agarose gels, and analyzed by Southern hybridization (13). Probes included 1) a ³²P-labeled oligonucleotide (TGGGGCTGGGGCTGGGGCT) designed from the exon 1 of YFVIw*7.1; 2) a 0.9-kb SacI fragment of pYFVIw*7.1 corresponding to exons 1 to 2; and 3) 163/164f.

Transcription analysis by RT-PCR was performed as follows: 1) first-strand cDNA was synthesized for 15 min at 37°C using 1 μg of total RNA, 40 nM primer, 200 μM dNTPs, 20 U of AMV transcriptase (Life Sciences, St. Petersburg, FL) in reverse transcriptase buffer, and a single oligonucleotide reverse primer RTBYFα2 (CCTCGAGGATGTCACAGCC) corresponding to a site in exon 3 identical in the BF and YF genes; 2) cDNA was tested for purity with PCR performed using primer pairs that span the α1 exon/intron/α2 exon boundary so that any product originating from genomic DNA can be recognized by product length; RTBYFα with BFIVα1-5′ (GGGCAGCCGTGGTTCGTGACT) and with YFα1-5′ (GTGGACGACAAAATCTTCGGTA). Products of the PCR (35 cycles at 95°C for 1 min, at 60°C for 45 s, and at 72°C for 45 s) were analyzed for fragment length on agarose gels; 3) cDNA free of genomic DNA was used as template for PCR performed with primers specific for the two YF loci consisting of YFα1-5′ and YFα1-3′ (TTTGTTGTAGCGTTCCGGCAGCC). For BFI the primer pair was BFIα1-5′ (GGGCTGCCGTGGTTCGTGGAC) and BFIα1-3′ (GTGTTCAAGCTCACTTCCACAC). For BFIV the primer pair was BFIVα1-5′ and BFIVα1-3′ (ATGCCCAGGTTCTCGCGGTCAA); and 4) presence and absence of the BF and YF transcripts were scored on the presence or absence of amplicon in agarose gels. To distinguish the locus of origin for YF amplicons obtained with YFα1-5′ and YFα1-3′ primers were analyzed by SSCP (30). For this, 1–3 μl of the PCR products were denatured in formamide at 80°C for 5 min and electrophoresed for 105 min at 200 V in 10% polyacrylamide, 0.5% TBE (44.5 mM Tris-borate, 44.5 mM boric acid, 1 mM EDTA) gels in a Miniprotean II apparatus (Bio-Rad, Richmond, CA). The gels were fixed and stained with a Silver Stain Plus Kit (Bio-Rad) and dried in gel wrap (BioDesign, New York, NY). The resulting patterns were scored in comparison with those provided by PCR in which line C DNA, cβ10, pYFVw*7.1, and pYFVIw*7.1 served as template.

A FLAG epitope tag sequence was incorporated into the YFV cDNA clone c36f for tagging mature protein at the N-terminal end. The modified clone was transferred into the replication-competent RCASBP(A) vector (31), the viral plasmid was transfected into avian DF1 cells (32), and packaged virus in the DF1 culture supernatant was used to infect avian RP9 cells (33). Extracts were made of intact RP9 cells expressing FLAG-tagged c36F YFV molecule and control cells (uninfected RP9 cells, RP9 cells infected with RCASBP(A) containing FLAGc36f in the nonsense orientation, and RP9 cells expressing FLAGBFIV21; Ref. 34), electrophoresed, blotted, and developed using M2 anti-FLAG mAb and ECL reagents (Amersham Pharmacia Biotech, Piscataway, NJ). Proteins were also immunoprecipitated with M2 anti-FLAG mAb (Sigma, St. Louis, MO) or with anti-chicken β₂m mAb (35), generously provided by Jim Kaufman (Institute for Animal Health, Compton, U.K.). The immunoprecipitates were electrophoresed and blotted as above.

Sequences were assembled with PC Gene and DNAStar. Similarity indices were determined with Wilbur-Lipman (DNA) and Lipman-Pearson (protein) algorithms. Deduced amino acid sequences were aligned using a mutation matrix, together with visual inspection of the modeled structures of YFVw*7.1 and B-FIV*12 and with Megalign (DNAStar). Pileup (GCG) and PaupSearch (maximum parsimony) were used to construct gene trees and to assign bootstrap values.

A homology model of the YFVw*7.1 structure was built using Insight II (Molecular Simulations, San Diego, CA) software. File 2clr.ent (HLA-A*0201) was chosen from the protein database as the template. HLA-A*0201 was one of several class Ia molecules giving essentially equal scores in FASTA alignments with YFVw*7.1 sequence. None of the class Ib molecules scored well when aligned with YFV. Amino acid insertions and deletions were readily placed between secondary structure elements at positions minimizing disruption of the overall fold by searching a high-resolution subset of the Brookhaven database for loops having the required length and similar context. Wherever more than one loop was present, the loop of the highest sequence homology was chosen. Once all coordinates were assigned and several conflicting side chains repositioned, the models were energy minimized with DISCOVER within Insight II using the consistent valence force field and default parameters. Similar steps were followed in modeling the avian β₂m chain on the structure of human β₂m.

BglII subclones containing YFVw*7.1 and YFVIw*7.1 were prepared from the cosmid cβ10 (14, 27) and fully sequenced. The two loci are oriented with 3′ ends opposed and are separated by 11.5 kb. The sequences of both loci are presented in Fig. 1 aligned with the sequence of the BFIV*12 (29), an allele at the classical class I locus that is most strongly expressed in the chicken and with which the Rfp-Y class I genes are highly similar. The exon/intron junctions for the YFVw*7.1 and YFVIw*7.1 were deduced based on the sequence of BFIV*12 and confirmed by sequencing a YFV cDNA (c36f) clone. The intron and exon organization in all three chicken genes is typical of class I genes. Eight exons are present, and their size is generally conserved with variations in the Rfp-Y genes confined to one or two codon differences from BFIV*12. Introns are generally small compared with mammalian class I genes with intron length varying between Rfp-Y and B system genes by 1–29 nucleotides. The Rfp-Y class I genes are C+G rich as has been noted for the B system class I loci (36).

The two YF loci are highly similar (93%) in nucleotide sequence (Table I) except for a large repeat sequence (48 copies of the hexanucleotide GGGCTG) that disrupts exon 1 of YFVIw*7.1 (Fig. 1). This insertion and the absence of a polyadenylation signal sequence downstream of the stop codon indicate that YFVI w*7.1 is most likely unexpressed. As noted below, no transcripts from the YFVI locus were found in any organs examined in RT-PCR assays. To determine whether the hexanucleotide repeat is present in other YFVI alleles in other Rfp-Y haplotypes two probes were prepared for Southern hybridizations. When an oligonucleotide containing three hexanucleotide repeats and a subclone of exon 1 from YFVIw*7.1 were used to probe DNA representing seven additional Rfp-Y haplotypes, only one was found to hybridize (data not shown), indicating that the repeat is not commonly present in other Rfp-Y haplotypes.

There is substantial sequence similarity between the sequence of YFVw*7.1 and other MHC class I genes in extracellular domains (Table I). Overall the nucleotide/predicted amino acid sequence similarities for YFVw*7.1 with BFIV*12 and HLA-A2 are 73%/60% and 49%/38%, respectively, and the similarity is generally uniform over the exons corresponding to extracellular domains. However, YFVw*7.1 and BFIV*12 are relatively less similar to each other in exon 2/α1 domain sequences (66%/49%) than they are in other extracellular exon/domain sequences suggesting that the α1 domains of YFV and BFIV molecules have divergent functional constraints. The exons corresponding to transmembrane and cytoplasmic domains of YFVw*7.1, BFIV*12, and HLA-A2 are mostly dissimilar suggesting further specialization associated with the Rfp-Y locus. As with other class Ib molecules, YFV-encoded molecules lack the phosphorylation motif found in mouse and human MHC class Ia molecules (37).

The predicted, mature product of the YFVw*7.1 gene is a 332-residue protein that has a single potential n-glycosylation site at the same residue found in many class Ia molecules (noted by • in Fig. 2,A). The amino acids critical for folding of class I molecules are generally conserved in the YFVw*7.1 amino acid sequence. The four cysteine residues that form the basis of the highly conserved class I disulfide loops (C98^C101-C161^C164, C199^C203-C255^C259) are present (the superscript denotes the position in HLA-A2). All 18 invariant residues (noted by I in Fig. 2,A) known to form various contacts within and between class I domains that are strictly conserved in the sequences of class I molecules (38) are present in YFVw*7.1. Molecular modeling of YFVw*7.1 provides a structure highly similar to that of HLA-A2 (Fig. 2 B). The structural integrity of the β strands and the α helices forming the ABR is mostly conserved in the YFV protein, even though the YFV ABR is three residues shorter than that of HLA-A2. The positions of the “missing” residues are easily assigned to the margins of the β strand and α helical regions and most likely do not disrupt domain folding. As reflected in the model, a proline substitution at P51^E53 in YFVw*7.1 is likely to disrupt the H1 helical region typical of the α1 domain of classical class I molecules. In addition, the presence of a contiguous pair of flexible glycine residues G67^A69 G68^H70 in YFVw*7.1 indicate that the long H2 helical region of the α1 domain may be broken into two shorter helices. Also, the natural break in the α2 domain α helix is further accentuated in YFV by the insertion of a glycine between positions 150 and 151 (HLA-A2 numbering). In summary, it is likely that YFVw*7.1 will have a structure overall highly similar to HLA-A2 with the ABR displaying a degree of specialization associated with the locus, a feature commonly encountered in comparisons between class Ia and class Ib molecules.

In class Ia molecules eight highly conserved residues define the “left” (Y7, Y59, Y159, and Y171) and “right” (Y84, T143, K146, and W147) pockets of the ABR and secure peptide Ag by bonding with main chain atoms. In YFV two of the left-pocket tyrosine residues are replaced by H57^Y59 and E156^Y159. Further substitutions occur in the right pocket. Position 84 in YFV is polymorphic occupied by R, Q, and C in different Rfp-Y alleles. A further, albeit conserved substitution, R143^K146, is also present in the right pocket. These left and right pocket substitutions make it highly unlikely that YFV class I molecules present Ag in the manner of class Ia molecules. Hence the YFV locus fails to meet a major criterion for inclusion in the class Ia category and, therefore, should be considered a class Ib locus.

The substitutions that are found in the predicted YFV molecules at four of the eight subclass-defining residues in the ABR are extremely rare among class Ib molecules. The E156^Y159 substitution is unique. Substitution of Y159 is generally rare with substitutions of phenylalanine (H2-M9, H2-Q5^k, H2-Mb-1, DLA79, FcRn), glycine (H2-T10), aspartic acid (H2-T9, H2-T22), tryptophan (Mr1), alanine (MICA), and leucine (hCD1c and mCD1) occurring in a limited number of class Ib molecules. The H57^Y59 substitution is shared so far only with Mr1 and some Xenopus class Ib loci. The Y59 is conserved at most class Ib loci with other substitutions such as phenylalanine (H2-M9) and leucine (mCD1) only rarely occurring. The substitution of three different amino acids at tyrosine 84 (R/Q/C82^Y84) in different alleles at any class I locus is unprecedented. Generally alternatives to tyrosine at this position are very rare with isoleucine and glutamic acid found at mouse H2-Mb-1 and CD1. Finally, the R143^K146 substitution occurs occasionally in class Ib molecules (H2-M3, FcRn) and is common among chicken class Ia (BFIV) alleles. Thus no other known class I locus is closely similar to YFV in the substitutions at these four positions suggesting that YFV defines a new type of class Ib locus.

The α3 domain sequences of the Rfp-Y class Ib loci were aligned using PileUp (Genetics Computer Group, Madison, WI) with the corresponding sequences from class Ia and class Ib molecules from several vertebrate species. The alignment was analyzed with phylogenetic analysis using parsimony to generate the gene tree and bootstrap values presented in Fig. 3 (class Ib are underlined). The two Rfp-Y class Ib α3 sequences are closely similar to those of chicken and quail class Ia molecules (see large bracket). This group forms a clade restricted to gallinaceous birds indicating that the Rfp-Y loci may be relatively young genes sharing recent ancestors with class Ia genes in gallinaceous birds. Similar relationships occur in other taxa between class Ib loci and class Ia loci as can be seen in the other bracketed regions of the tree where Xenopus, humans, and mice also form clades in which class Ib and class Ia share recent ancestry. Most of the class Ib molecules within these clades are known to require TAP for processing of specialized Ag indicating that these relatively recently derived class Ib share Ag processing pathways with class Ia (8). These molecules contrast with more distantly derived class Ib molecules known to be TAP independent in the presentation of specialized Ag.

Because polymorphism or the lack thereof is another characteristic that is often used to separate class Ia from class Ib loci, we examined Rfp-Y class I genes for evidence of haplotypic and allelic polymorphism. We first examined Rfp-Y class I haplotypic variability using a Rfp-Y class I specific probe, 163/164f, in Southern blots. Nine different TaqI restriction fragment patterns were obtained from nine previously defined Rfp-Y haplotypes (Fig. 4 A). Surprisingly, the number of restriction fragments varies among the haplotypes from only two in Yw*1.3 and Yw*7.2 to at least 10 in Yw*4.2 and Yw*6.1. Similar differences were found in PstI and BglI restriction fragment patterns (data not shown). It is likely that the number of class I loci varies among Rfp-Y haplotypes. Similar variation in gene number occurs in the class Ib region in H-2 haplotypes (39).

Predicted amino acid sequences for YFVw*7.1, YFVw*3.2, and YFVw*5.1 are aligned in Fig. 2 A. It is likely that these three sequences all originate from the YFV locus. YFVw*7.1 and YFVw*3.2 are identical in cytoplasmic domains and differ by only one amino acid in the transmembrane domain (TM). In contrast, YFVw*3.2 differs from YFVIw*7.1 by 15 of 34 aa over the same regions making it highly unlikely that YFVw*3.2 originates from the YFVI locus. The third clone, YFVw*5.1, was obtained by PCR using primers designed for YFV specificity and differs from YFVw*7.1 by only a single amino acid in transmembrane domain sequence.

The three clones display considerable sequence variability (Fig. 2, A and B). The α1 domain variability among the three sequences (21%, 18 of 87 aa) is nearly as great as the variability that is present among BFIV alleles (27%, 24 of 88 aa in 11 alleles) (40). This variability is almost entirely confined to the helical region of the α1 domain and to the floor of the ABR with a small remainder of variation present in loop regions (Fig. 2 B). Amino acid replacements are most often nonconservative. For example, charged residues are interchanged with neutral, nonpolar (70R/L, 59D/A, 67D/G, 85K/I, 87K/G) or with neutral, polar residues (37D/N, 55Q/R, 66Q/R, 73D/C, 82R/Q/C). In other instances neutral, polar residues are interchanged with neutral, nonpolar residues (32N/I, 35T/I, 60T/A, 78W/G, 75N/F/L, 77N/G). Conservative substitutions (47V/A and 71D/E) in these regions are more rare.

The α2 domain is less variable (10%, 9 of 92 aa). Substitutions are mostly confined to two β sheet strands in the floor of the ABR, as they are in the classical BFIV alleles (40). The residues are often hydrophobic, and substitutions are often conservative. One nonpolar residue often substitutes for another (92L/M, 95 M/I, 96I/F, 120L/I), or charged residues are interchanged (117R/K). In other instances the substitutions are nonconservative (91T/M, 94 M/R, 119F/H/Y, and 178R/T). Similar to the chicken class Ia molecules, some variability is also present in the α3 domain (Fig. 2 A). Whether this reflects functional specialization among the YFV alleles is not yet understood.

Nonsynonymous-to-synonymous substitution ratios vary across α helical and β sheet regions of the α1 and α2 domains of the three YFV sequences (data not shown). Briefly, the α helical portion of the α1 domain has a high ratio compared with the β sheet region suggesting that, as in class Ia loci (40, 41), this region may be under selection for interactions with diverse Ag or variability in a counterreceptor. In contrast, in the α2 domain the values for the nonsynonymous-to-synonymous substitution ratios are reversed. The YFV α2 α helical region has an extremely low ratio indicating that diversification of this region of the molecule is restricted, as apparently it is in BFIV alleles (40). So although YFV molecules are nonclassical and unlikely to bind typical peptide Ag, they are polymorphic with the distribution of sequence variability among alleles not unlike the classical class Ia molecules of the chicken.

To determine whether the YFVw*7.1 and YFVIw*7.1 are transcriptionally active and to learn whether transcription is confined to particular organs as often is found for class Ib genes, we performed RT-PCR SSCP assays using a YF gene-specific primer set. With the exception of small regions immediately upstream of the start site, the regulatory and promoter sequence elements (see GenBank sequences) of YFVw*7.1 (and YFVIw*7.1) genes are similar to those of class Ia genes suggesting that YF could be generally active in many tissues. RNA free of genomic DNA was obtained from embryos and from a number of organs of young adult line C birds (Fig. 5 and Table II). The YF-specific primers provided a means of specifically detecting transcripts from the Rfp-Y class I loci, and SSCP provided a means of distinguishing between YFVw*7.1 and YFVIw*7.1 transcripts. RT-PCR assays for BFI and BFIV served as positive controls. YFVw*7.1 transcripts were detected in all organs tested, except for three (brain, heart, and pancreas). No evidence was found for transcriptional activity of YFVIw*7.1 locus. Results of the full analysis are summarized in Table II and are consistent with a limited number of RNase protection assays demonstrating the presence of YFVw*7.1 transcripts in several tissues (42). We conclude that YFVw*7.1 is constitutively transcriptionally active in many organs much like MHC class Ia genes. It remains to be determined whether YFV transcription is inherent in all or many of the tissues in these organs or whether transcription is limited to a particular cellular subset with perhaps only limited quantities of YFV reaching the surface of these cells.

To determine whether mature protein can be produced from expression of YFV cDNA, the clone c36f was FLAG-epitope tagged, inserted into the RCASBP(A) retroviral vector, and expressed in the chicken B cell lymphoma line RP9. FLAG-tagged protein was found by flow cytometry to be at the surface of cells expressing FLAGc36f inserted into the vector in the sense orientation but not in the nonsense orientation (data not shown). In immunoprecipitations with the anti-FLAG mAb M2, RP9 cells expressing FLAGc36f and the positive control FLAGBFIV*21 were both found to produce FLAG-tagged protein with masses (∼48 kDa) typical of class I molecules (Fig. 6,A). The FLAGc36f encoded protein could also be immunoprecipitated with an anti-chicken β₂m mAb (Fig. 6 B) providing evidence for association between YFV protein and β₂m. Hence protein similar to typical class I molecules in molecular mass and in β₂m association can be obtained by expressing the YFV cDNA c36f as a transgene.

In these experiments we have shown that the YFV locus shares many qualities with MHC class Ia genes. As summarized in the structural model in Fig. 2,B, the YFV molecule is likely to be structurally similar to class Ia, as are several other class Ib molecules. The unusual substitutions in the putative ABR of the YFV molecules are the strongest characteristic separating YFV from class Ia loci. Alleles at the YFV locus differ from one another by multiple changes in predicted amino acid sequence. Many of the polymorphic residues surround the ABR. Over 20 aa differences in the ABR separate the three YFV alleles in this study from each other. Such variability is on par with the differences between alleles at class Ia loci and between alleles at the one previously identified highly polymorphic class Ib locus, H2-Q2 (1). In addition, there may be many YFV alleles. SSCP analyses of the Rfp-Y class I exon 2 sequences in a variety of genetic lines provide indirect evidence for many more YFV alleles than the three presented here (M.M., unpublished data). As with class Ia loci, polymorphism and the relatively high ratios of nonsynonymous-to-synonymous substitutions over portions of the YFV ABR indicate that this region may be under selection for diversity. YFVw*7.1 and YFVIw*7.1 occupy positions in the evolutionary tree in Fig. 3 that further emphasize their close relationship with class Ia genes in gallinaceous birds. The gene tree in Fig. 3 indicates that the YFVw*7.1 and YFVIw*7.1 were derived relatively recently with their appearance likely predating only the separation of gallinaceous taxa (Fig. 3). Perhaps YFVw*7.1 and YFVIw*7.1 are intermediates in the evolutionary tide postulated by Shawar and colleagues (37) to flow between loci encoding molecules with less selective (class Ia) and highly selective (class Ib) ABRs.

Given the close relationship the YF loci have with avian class Ia genes and their sequence polymorphism, it seems likely that YFV molecules bind peptide Ag. The unique substitutions of glutamic acid and histidine in the left pocket of the ABR may provide a means for selecting a particular form of Ag. Perhaps the charged residues form salt bridges with Ag in the ABR providing a means for selection of a particular subset of Ags. Or alternatively, perhaps the left end of the putative ABR of YFV is actually closed by interactions between these and other residues surrounding this region of the groove. In this instance, shorter forms of peptide might fill the remaining open portion of the groove through a selective interaction based on another characteristic of antigenic peptide, such as hydrophobicity of amino acid side chains. Because YFV molecules are apparently close relatives of the MHC class Ia molecules, it seems likely that peptide Ag load into the ABR through a TAP-dependent pathway. It would make sense that the YFV Ag is peptide, but this remains to be determined. How YFV molecules bind Ag and what form the Ag has will be the subject of additional experiments, as will be the consequence of YFV expression in cellular interactions with cytotoxic T and NK cells.

The YF loci have other features that define them as class Ib. Most class Ib loci have little or only weak influences in graft rejection. The structural identification of YFV as a class Ib gene is consistent with conclusions drawn by Pharr and colleagues (28) on the influence of Rfp-Y incompatibility on transplantation immunity. In experiments conducted with carefully defined genetic stock, Pharr et al. found skin graft rejections attributable to Rfp-Y incompatibilities occurred at moderate rates. They were clearly slower than B incompatibilities but significantly faster than Rfp-Y-compatible grafts. These authors suggested that one reason for the observed intermediate rate of graft rejection with Rfp-Y incompatibility might be the presence of class I-like (nonclassical) loci within the Rfp-Y gene region.

The YFV locus may share another feature with class Ib genes. There may be little YFV normally on the surface of cells despite the presence of YFV transcripts. Other investigators have found no evidence that Y-FV molecules are immunoprecipitated by anti-chicken β₂m (43) and so YFV molecules may be normally less abundant at the cell surface than, for example, chicken class Ia molecules derived from the BFIV locus. It will be interesting to learn whether there are conditions under which surface expression of YFV becomes abundant. Because the Ag for YFV is likely to be atypical, it may be that it is not normally abundant and that trafficking of YFV to the cell surface is limited by Ag availability. This would be particularly interesting to explore given the evidence that in some but not all instances Rfp-Y haplotype has been found to influence resistance to virally induced tumors in chickens (44, 45, 46). If YFV is dependent on TAP molecules encoded in the B system for Ag processing, it could also be that interaction with chicken TAP affects YFV surface expression. Because chicken TAP genes are themselves polymorphic (47) and the YFV and TAP loci are unlinked, it might be that in particular combinations of TAP and YFV alleles there is either less or more YFV at the cell surface even in the presence of ample YFV Ag.

Finally, the organization of chicken class I genes into two genetic units composed of class Ia and class Ib loci is not unique. The class Ia and class Ib loci in Xenopus are also located in two independent genetic units and, just as in chickens, the two regions map to the same chromosome (48) separated by a region supporting highly frequent recombination. Considering the evolutionary relationship that exists between class Ia and class Ib genes in these two species, as illustrated in Fig. 3 A, it is likely that this manner of organizing class I genes has been arrived at independently in these two species. Genetic separation of the two class I subclasses could be a means by which the integrity of class Ia and class Ib loci are maintained. Perhaps the class Ia loci isolated by this arrangement evolve in concert with adjacent Ag processing loci as has been suggested by others (47, 48), whereas the class Ib loci are free to evolve in a different manner. In isolation the class Ib loci may be able to change in number, allelic variation, and ABR specificity through a variety of recombination events in a system for selective Ag presentation that evolves rapidly in response to disease challenge.

We thank Larry Bacon for providing C line birds and DNA from Cornell line N and P; Henry Hunt for providing RP9 cells expressing FLAGBFIV*21; Pierrick Thoraval for providing tissue from line CB embryos; and Jim Kaufman for providing anti-β₂m mAb. Elwood Briles and Robert Taylor, Jr. generously provided blood samples from Rfp-Y-typed birds. We thank Larry Bacon, Pamela Bjorkman, Louis DuPasquier, Henry Hunt, and Iwona Stroynowski for helpful discussions.