Sequences, Annotation and Single Nucleotide Polymorphism of the Major Histocompatibility Complex in the Domestic Cat

Two sequences of major histocompatibility complex (MHC) regions in the domestic cat, 2.976 and 0.362 Mbps, which were separated by an ancient chromosome break (55–80 MYA) and followed by a chromosomal inversion were annotated in detail. Gene annotation of this MHC was completed and identified 183 possible coding regions, 147 human homologues, possible functional genes and 36 pseudo/unidentified genes) by GENSCAN and BLASTN, BLASTP RepeatMasker programs. The first region spans 2.976 Mbp sequence, which encodes six classical class II antigens (three DRA and three DRB antigens) lacking the functional DP, DQ regions, nine antigen processing molecules (DOA/DOB, DMA/DMB, TAPASIN, and LMP2/LMP7,TAP1/TAP2), 52 class III genes, nineteen class I genes/gene fragments (FLAI-A to FLAI-S). Three class I genes (FLAI-H, I-K, I-E) may encode functional classical class I antigens based on deduced amino acid sequence and promoter structure. The second region spans 0.362 Mbp sequence encoding no class I genes and 18 cross-species conserved genes, excluding class I, II and their functionally related/associated genes, namely framework genes, including three olfactory receptor genes. One previously identified feline endogenous retrovirus, a baboon retrovirus derived sequence (ECE1) and two new endogenous retrovirus sequences, similar to brown bat endogenous retrovirus (FERVmlu1, FERVmlu2) were found within a 140 Kbp interval in the middle of class I region. MHC SNPs were examined based on comparisons of this BAC sequence and MHC homozygous 1.9× WGS sequences and found that 11,654 SNPs in 2.84 Mbp (0.00411 SNP per bp), which is 2.4 times higher rate than average heterozygous region in the WGS (0.0017 SNP per bp genome), and slightly higher than the SNP rate observed in human MHC (0.00337 SNP per bp).


Introduction
The major histocompatibility complex (MHC) is one of the most extensively analyzed regions in the genome due to the fact that this region encodes the most important molecules in immune function, namely class I and class II antigens, and also other important molecules such as chemical sensing genes (olfactory receptor gene complex), its escort gene, and POU5F1 gene involved in iPS stem cells [1][2][3][4]. Recently, the human MHC, HLA haplotypes were sequenced in the HLA haplotype project [5][6][7][8][9][10]. Eight different HLA -homozygous haplotypes' DNA sequences were determined in order to shed a light on MHC-linked diseases and evolutionary history. These BAC-based sequencings are necessary to examine the details in the regions of the genome, where gene duplications, deletions and selections occurred many times, because the genome project, especially in the human genome, was carried out using a mixture of DNA sources [11]. The same will be true in genome projects in other outbred species. The domestic cat serves excellent animal models to study at least three RNA viruses in humans. Feline leukemia virus (FeLV) shares similarly to human leukemia viruses (HTLV I & II) [12]. Feline immunodeficiency virus is considered to cause similar symptoms to human AIDS in a natural host, the domestic cat [13][14][15][16]. Feline infectious peritonitis virus belongs to the same virus group (corona virus) as human SARS virus [17]. To study host-defense mechanisms, in this animal model, we previously analyzed and reported (i) approximately 750 kbp class II region in feline MHC (FLA) [18], (ii) the unique FLA structure with a single chromosomal split at the TRIM gene family region, and chromosome inversion [19], and (iii) comparison of three MHCs, HLA, DLA, and FLA using human sequence, canine MHC homozygous genomic sequence and feline 3.3 Mbp draft sequence based on BAC shotgun sequences [20]. In this manuscript, much detail of FLA gene contents, promoter structures of predicted functional class I and class II genes, proportional scale comparisons of four mammalian MHCs (domestic cat, human, mouse, dog) and one marsupial MHC (opossum) are presented. SNPs (single nucleotide polymorphisms) between the MHC homozygous sequence of the lightly covered (1.96) domestic cat genome shotgun sequence and this BAC-based MHC sequence were also analyzed to compare the degree and mode of the MHC divergence. In addition, two haplotype BAC-based sequences in functional class II DR region in the domestic cat were analyzed.

Gene annotation
Sequences were first masked by Repeatmasker program. Gene annotation was made using GENSCAN [26] coding prediction plus BLASTP and BLASTN programs [27], also using megablast for the entire sequences against the latest human Refseq database. Class I, MIC, BAT1, olfactory receptor, MOG, TRIM26, 15, 10 gene annotation was made using human transcripts or FLA class I mRNA sequence (FLAI-A24) [28] by bl2seq [29] and results were parsed using Perl scripts. Repeat sequences were analyzed using Repeatmasker and STR finder programs. These data was graphically presented using Advanced PIPmaker program [30].

SNP analysis
The 1.96feline WGS contigs [32] were aligned with BAC MHC sequence using CROSSMATCH program and SNP was found between sequences selected by reciprocal best matches (.90% sequence identity) and with more than Quality value = 15 [33].

DR haplotypes
BAC clones of 152N13-244j14-16i4 from B2qCen side were sequenced by the shotgun method described above and analyzed by the methods of GENSCAN, Spidey [34] and Genwise [35] for annotation. A sequence assembled from this DR haplotype 2 region was compared with a sequence from DR haplotype 1 region previously published [18].

Comparisons of MHC structures
Sequences of MHC from four species: human, mouse, dog, and opossum which span from KIFC1 gene through UBD plus three olfactory receptor genes were extracted from UCSC Genome Browser. Gene coordinates tables from UCSC site were parsed by Perl script and gene organizations were graphically plotted by R script (http://cran.R-project.org).
Transcription factor binding sites in promoter regions of predicted functional feline class I and class II DR genes Sequences totaling 6 kb (5 kb upstream and 1 kb downstream) from a potential translation start site (ATG) of predicted functional feline class I genes (FLAI-E, I-H, I-K) and class II DR genes (FLA-DRA1, DRA2, DRA3, DRB1, DRB3, DRB4) in addition to human HLA-A, -B, -C class I genes and HLA-DRA, DRB1, DRB3 class II genes were analyzed for the presence of potential transcription factor binding sites using Match TM program with TRANSFAC 7.0 database (http://www.gene-regulation.com/ pub/databases.html). In addition, S-Y-module sequences of human HLA class II and I genes were used to screen above 6 kb sequences with b12 seq [29] with parameters, MATCH = 1, MISMATCH = 21, GAP OPEN 5, GAP EXTENSION 2, X_DROP OFF 0, EXPECT 10.00, WORDSIZE 7.

Results
Sequence 2,975,516 bp and 381,545 bp sequences were assembled for two FLA regions on the pericentromeric and subtelomeric positions of feline chromosome FcaB2. The first sequence covers from KIFC1 gene in the extended class II region through the entire class II, class III and a part of class I regions from the point adjacent to BAT1 gene through HLA-B, -C class I corresponding region, TRIM39 plus HLA-92 (HLA-L) region to alpha satellite-rich region. The second sequence covers from telomeric repeats rich region through TRIM 26/15/10 genes to the third olfactory receptor like gene (GenBank accession Nos. EU153401, EU153402).

Annotation
The entire gene coordinates, and possible functions are listed in Table 1. Gene organization and GC level was depicted in Figure 1. Detailed graphic presentation for exon-intron structure, orientation, repeat sequence, CpG island and sequence identity level to human HLA-6 COX 4.72 Mbp sequence was organized in Figures 2. (Figure 2A-1 was shown in the main text. Please see Figure S1 supporting file).
Extended and classical class II region. Extended class II region spans 230 Kbp from KIFC1 gene to the point adjacent to DPB pseudogene. Fourteen human gene homologues and 2 unknown coding regions were found. Classical class II region spans 884 Kbp. Twenty-five human gene homologues were found in the region defined from DPB pseudogene to the point adjacent region of NOTCH4 gene. Annotation and sequence of a part of this region, (KIFC1, previously assigned as HSET to BTNL2), were described elsewhere [18]. Briefly, a pair of DPA, B pseudogenes, 3 pairs of DRA, B genes were identified with one DRB pseudogene. A set of genes which are involved in antigen processing, including a pair of DOA, DOB, DMA, and B genes, two antigen transporter genes, TAP1, 2, and protease genes, LMP2, 7 were found. In addition, two butylophillin genes, BTNL2, BTL2, and BRD2 (previously assigned as RING3) genes were found.
Class III region. FLA class III region spans 520 Kbp which encodes fifty-one human gene homologues and two unknown coding regions.   Class I region. FLA class I region were classified as three subregions based on chromosomal localization and gene contents. The first class I region, proximal class I region spans 600 Kbp from the first class I gene (FLAI-A) to the last class I gene (FLAI-Q) in this HLA-B, -C corresponding region adjacent to the class III region. This region encodes seventeen class I genes/gene fragments based on sequence alignments with full length feline class I cDNA sequence. Eight BAT1 gene fragments are located in the vicinity of class I genes. Three RPS28 gene fragments, four class I-related (MIC) genes or gene fragments and four unknown coding regions were also identified.
The second class I subregion, a central class I region, spans 600 kb region from POU5F1 (previously assigned as OCT3) gene to the alpha satellite repeat-rich pericentromeric region. There are 32 human gene homologues including two class I gene/gene fragments in HLA-92 (HLA-L) region, three unknown coding regions. The third class I subregion, distal class I region, spans 360 Kbp from 47 telomere repeats of (TTAGGG) through the third olfactory receptor like gene. This region encodes ten human gene homologues and three unknown coding regions. Three TRIM genes, TRIM26, TRIM15, TRIM10 were identified, however, TRIM40, 31 gene homologues were not recognized. PPPR11 and MOG genes are located in 26 Kbp interval, while in human HLA, these two genes are located with 340 kb interval due to the existence of eleven class I genes/gene fragments as HLA-A region. GC contents. GC contents nearly reached at 60% level in the extended class II, class III, and the distal class I regions. The lowest GC content of nearly 40% was found in the classical class II region and sharply increased at the border of class II and class III regions. The proximal/central class I regions kept GC content at 50% level ( Figure 1).

Repeats
Interspersed repeats. Interspersed repeats occupied about thirty-four percentages of MHC region, which is approximately the same level as found in the cat genome (32%), but significantly fewer than human HLA region (48%) or human genome (46%). Simple Tandem Repeats (STRs). Frequency of STRs was calculated in each FLA subregion and was compared with results obtained from human HLA 6COX haplotype sequence. These results were summarized in Table 3. A total of 541 STRs (di-, tri-, tetra-, penta-) with more than 12 and 5 perfect repeats for di-and others, e.g. (CA) 12 and (GGA) 5 , respectively were found in FLA. The frequency of STRs (1 every 6.17 kb) was 50% higher than that in human HLA (1 every 9.93 kb) due to at least 3 times higher frequency of dinucleotide repeats. This trend was more obviously observed in the classical class II region. Approximately 4 times more occurrence of dinucleotide STRs was found in this FLA subregion.
Dotplot analyses of HLA, DLA, FLA. MHC sequences spanning from UBD plus three olfactory receptor genes to KIFC1 in HLA, DLA, FLA were compared in pairwise fashion. These analyses, DLA vs. HLA ( Figure 3A), FLA vs. HLA ( Figure 3B) and FLA vs. DLA ( Figure 3C) revealed mosaic structures of highly conserved regions (solid lines), gene duplication (square with dots), deletions (spaces between solid lines in one axis but not in other) and divergent regions (broken lines). Figure 3A and 3B showed similar patterns between DLA vs. HLA and FLA vs. HLA, indicating conserved class II, III, and central class I regions plus class I gene amplifications, though the level of class I gene amplification was lower in DLA due to the fact that only 3 class I genes exit in HLA-B, -C corresponding region. The observation that DLA and FLA lack HLA-A class I region was also evident in this analysis. Figure 3C also showed that FLA and DLA were highly conserved in gene contents and organization except that the level of class I gene amplification was higher in FLA and sequences around pericentromere and subtelomere had highly divergent sequence due to the numerous and different types of gene translocations from other genome sites, resulting in a large broken solid line in these regions.   Endogenous retrovirus sequences. One of the baboonderived endogenous retroviruses, ECE1 (RD114) which had 99% sequence identity (1631/1633) with GenBank RD114 (ECE1) AF155060 and two new types of endogenous retroviruses FERVmlu1 and 2, which showed high sequence similarity with recently submitted sequences by an NISC Comparative Sequencing Initiative project of brown bat (Myotis lucifugus) BAC clone (95% sequence identity with 83% coverage, and .85% sequence identity with 83% coverage, respectively) were also recognized within 1401kb region (Figure 4). Detailed open reading frame (ORF) analyses showed FERVmlu1 and 2 have 140 and 19 ORFs which sizes range from 102-900 and 102-516, respectively. The largest ORF of FERVmlu1 encodes 324 amino acid residues which have 70% similarity to a part of recombinant mouse-MuLV/RaLV Pol region, half of retroviral aspartyl protease, DNA binding region and a half of putative active site, however, other ORFs have no significant homology to gag, pol, env regions.
The FERVmlu2 have two Pol-like ORFs. The ORF1 is similar to reverse transcriptase like sequence, in which encodes a DNA binding domain and a putative active site. ORF2 has similarity to an integrase core domain. The third ORF showed a gag -p30 superfamily motif. Nine LTR like sequences were recognized in FERVmlu1 by Repeatmasker program, indicating sequence divergences ranging from 13 to 32% in canine, baboon, chimpanzee endogenous retroviral LTRs.
Single nucleotide polymorphism (SNP). The SNP count plot ( Figure 5) in the MHC region from 1.96 cat whole genome shotgun sequence indicated that this region is homozygous. Therefore, MHC BAC sequences were aligned with MHC homozygous 1.96 whole genome shotgun sequence contigs to examine SNP levels in FLA [32]. A total of 2,835,361 bps were aligned with sequence quality value, more than 15 by reciprocal best matches (.90% homology) using the algorithm of Smith-Waterman. This covers more than 85% of the entire FLA region. Distributions of these SNPs and coding SNPs were plotted in Figure 1 and the summary was presented in Table 4. A total of 11,654 SNPs were identified by this method. FLA SNP rate was slightly higher than the rate of two HLA haplotypes (0.00411 vs. 0.00337), and more than 2 times higher than genome wide regions of SNP rate (0.0017) found in regions of heterozygous cat WGS result. Ten to 20 times higher SNP rate than average FLA SNP rate was found in class II DR region, class II/III border region, proximal class I region and pericentromeric and subtelomeric regions. Clustered high coding SNP rates were observed in the proximal class I region.

Two DR haplotypes in a single BAC library
We previously constructed a composite nucleotide sequence of the domestic cat MHC class II region spanning 758 kb from HSET to BTLII that included the DR region [18]. As shown in Figure 6A the DR region spans approximately 250 Kb and consists of 3 DRA and 4 DRB genes, both gene families are encoded by 5 exons with the exception of DRB2 which lacks the full complement of exons and thus represents a pseudogene. We determined the sequence of the second haplotype of a domestic cat DR region using BAC clones (152n13-244j14-16i4) from a single individual. The DR haplotype 2 contained three complete DRA genes (DRA1, DRA2 and DRA3), five DRB genes (four complete and one partial) similar to DRB4, DRB1, DRB3 and DRB2 plus new DRB gene, namely DRB5 as well as a BTLII gene all with the same order and orientation as observed in the DR haplotype 1 (except DRB5 adjacent to DRB1 with same orientation) ( Figure 6B). To determine if DRB1 and DRB4 also displayed allelic variation, we aligned the genomic sequences of DRB1, 3 and 4 in the region of exon 2 and flanking introns 1 and 2. To assign these exon 2 sequences to specific DRB alleles we compared them to 71 different domestic cat DRB exon 2 alleles of 238 bp in length present in the NCBI nucleotide sequence database (nr/nt). The results summarized in Figure 6B show that DR haplotype 2 contains DRB3 exon 2 identical to DRB*0504 whereas DR haplotype 1 contained a DRB3 exon 2 that differed by 2 bp from DRB*0204, and thus represents a new domestic cat DRB allele (DRB*0204_new1) but differ from nucleotides of 70 nts between DR hap1 and DR hap2. Similarly, DRB1 exon2 sequences in haplotypes 1 and 2 contain the alleles which were identical to DRB*0511 and DRB*0403, respectively, but differ in 31 nucleotides of 233 nts. In addition, haplotype 1 was also positive for DRB4 which showed 229/233 nucleotide sequence identities with DRB*0107_new1. In contrast, the haplotype 2 DRB4 sequence was identical (233/233) to DRB*0107. The DR haplotype 2 contains additional DRB genes designated DRB5 that was not observed in haplotype 1 that displays identical exon 2 sequences with DRB*0212. In summary, this data show that a single domestic cat (Fca273) contains three or four DRB genes in the order DRB4-DRB1-(DRB5)-DRB3, that the three loci are heterozygous, and resolve into 2 distinct haplotypes consisting of DRB*0107_new1-DRB*0511-DRB*0204_new1 (haplotype 1) and DRB*0107-DRB*0403-DRB*0212-DRB*0504 (haplotype 2). Dotplot analysis confirmed this conclusion, indicating duplicated DRB genes adjacent to DRB1 gene (Figure 7). Deduced amino acid sequences of above alleles were compared in each DRB loci ( Figure 6C). Of these loci, DRB1 alleles were the most polymorphic containing 23 different residues. DRB3 was the second most polymorphic loci, maintaining 17 different residues. In contrast, DRB4 had only one amino acid substitution.
The additional DRB loci found in haplotype 2, named as DRB5 had 16 different residues compared with DRB1 loci in haplotype 2. On antigen recognition site (ARS) defined by X-ray crystallography [36], 22 sites forms ARS. Of these sites, 15 sites were found highly polymorphic in FLA.

Transcription factor (TF) binding sites in predicted classical genes
We have analyzed transcription factor binding sites a total of 6 kbp (5 kb upstream and 1 kb downstream of ATG putative translation start site) of predicted feline classical class II genes (DRA1, 2, 3, DRB1, 3, 4: Figure 8A, Figure 8B) and classical class I gene candidate genes (I-E, I-H, I-K: Figure 8C), plus human classical class II and I genes (DRA, DRB1, DRB3, HLA-A, -B, -C). Figure 8A depicts the result of DRA genes. All three feline DRA genes have CCAAT-box. The DRA1 and DRA2 genes have striking similarity with TF binding sites up to about 4 kb upstream of ATG site and at least NF-Y binding site, indicating recent gene duplication. In contrast, the DRA3 gene has distinct TF binding sites from the other two genes and is relatively similar to those of the human DRA gene, (e.g., NFY-RFX1-RFX1 sites, Oct-1 sites, sox-9 sites). It may be suggested that the expression pattern is different in these two groups of DRA genes.   Figure 8B depicts the difference between TF binding sites in feline DRB1, -3, -4 genes, and human DRB1 and -3. All of these genes lack the CCAAT-box site. The NF-Y site was found in FLA-DRB1, DRB4, and HLA-DRB3. However, no apparent similar TF binding site patterns were found. In class I genes, all predicted feline classical class I genes have a CCAAT-box site plus a unit of AP1-HNF4-Pax-4 sites adjacent to the CCAAT-box ( Figure 8C). FLA I-H, I-K had relatively similar TF binding sites (e.g., Pax-4-Pax4, Evi-1, FOXD3-COMP1-Hand1/E47, Nkx2-5). Human HLA-B, -C had relatively similar TB binding sites (e.g., CCAATbox, Oct1, Evi-1-FOXD3, Evi-1) however, the HLA-A gene had no CCAAT-box and was quite different in TF site pattern from HLA-B, -C genes. No apparent similar TF binding patterns were found in the FLA and HLA classical class I genes.
MHC class I and class II gene promoter structures were well documented and intensely analyzed by many molecular biological methods [37,38]. MHC class II genes are regulated by a complex system containing two gene-specific transcription factors, regulatory factor X complex (RFX) and CIITA, and maintain an approximately 67 bp sequence, a strictly conserved regulatory module (S-X1-X2-Y) immediately upstream of the promoters [37].
Similar conserved regulatory modules were identified in most of FLA class I and II genes analyzed here and summarized in Figure 8D.

Discussion
We report here annotation and SNP analysis of cat MHC (FLA). This study revealed one hundred forty-seven human gene homologues with mostly conserved gene order in five subregions, extended class II, class III, proximal class I, central class I, and distal class I regions. Extensive rearrangement events were obvious in classical class II and class I regions by dotplot analyses of three mammalian MHC, human HLA, canine DLA, and feline FLA ( Figure 3). Especially, deletion of HLA-A and -E regions in both DLA and FLA, and expansion of the regions in FLA corresponding to HLA-B, -C were clearly observed ( Figure 3A, B). A dotplot between DLA and FLA ( Figure 3C) suggests that these two MHC PIPmaker dotplot ( ) was used. + and 2 represent full-length and partial length, respectively compared with full length FLAIA24 cDNA. b. GENSCAN was used to predict coding region for only full-length class I genes. + and 2 symbols represent right and wrong prediction of exon and intron boundaries in each gene. c.
Spidey was used to examine sequence alignment of genomic cDNA class I sequences and splicing donor/acceptor sites. ++, +, and 2 symbols represent typical class I exon/intron structures reported in human class I genes with all correct splicing donor/acceptor sites, with one or two missing splicing donor/acceptor sites, and atypical exon/intron structures, respectively. d.
Class I cDNA sequences from MHC homozygous feline fibroblast cells were compared with all class I genomic sequences by Megablast Search ( ). + symbol represents .99% sequence identity. e.
Thirty-one highly conserved amino acid residues found in a1 and a2 domains of human and cat class I antigens were examined. ++, +, 2 numbers represent all 31 conserved residues, one substitution and more than one substitutions, respectively. systems are more syntenic than those to HLA. However, the manner of class I rearrangement was unique in each DLA and FLA. Each DLA and FLA also had unique sequences near the heterochromatin regions (near telomere and centromere) in canine chromosome cfa35ter/cfa12cen and feline chromosome B2pter/ qcen regions. Among mammalian and MHC class I regions reported so far, only mammals which belong to the group Euarchontoglires (Primates and Rodentia) have class I E and A subgroups, plus the evidence of the recombinant origin of the class I E gene between class I A and B/C [39] suggests that the formation of these two class I subgroups (A, E) occurred after the split of two major mammalian groups, Euarchontoglires and Laurasiatheria [Carnivora (dog and cat), Perissodactyla (horse), Certartiodactyla (pig and cattle)].

Class II genes in FLA
Unlike all other mammalian MHCs which have a single DRA gene, FLA maintains three possible functional DRA genes due to two possible duplication events and one inversion [18]. The deduced amino acid sequences coding a mature DRA peptide are identical in these three DRA genes. However, significant levels of difference in amino acid sequences in the signal peptide region, which may suggest distinct roles in this region. In addition, distinct TF binding sites in DRA1/2 and DRA3 may suggest distinct expression patterns. All three DRB genes, common in two haplotypes examined had significant levels of polymorphism in exon 2 sequence which encodes peptides forming antigen binding and T cell receptor recognition sites. The well documented S-X1-X2-Y promoter module sequences were found in all DRA and DRB genes immediate upstream of CCAAT-box site , except that DRB4, which maintains this module sequence 7 kb upstream from ATG site and 5.5 kb upstream of CCAAT-box. FLA is also unique among mammalian MHC due to the fact that the entire DQ region is deleted. Since canine MHC (DLA) maintains a pair of A and B genes in its DQ region, this deletion event may occur after the split of canids and felids (55MYA).

Class I genes in FLA
Class I gene amplicon of a combination of class I and BAT1 gene fragments are found here in FLA-specific manner, though the human HLA-A region has two BAT1 gene fragments, suggesting that relatively new origins of multiple class I genes than classical class II families (DP, DQ, DR), which were estimated more than 80 MYA [40]. Gene structure of 19 FLA class I genes was characterized and summarized in Table 5. Eleven class I genes maintained full-length exons by dotplot, when compared with FLA class I cDNA sequence and their coding sequences were predicted by GENSCAN. Of those, six class I genes had intact splicing donor/acceptor sites. Three genes (FLA I-E, I-H, I-K) had 31-32 highly conserved amino acid residues in a1 and a2 domains which were reported in deduced amino acid sequences of FLA class I transcripts from fibroblast cell lines [41]. Analysis of FLA class I transcripts of a fibroblast cell line from MHC homozygous Abyssinian cat used for cat genome project indicated that these transcripts are derived from FLA I-H and I-K. In addition, all three of these class I genes maintain the conserved enh.A-ISRE-W/S-X1/ X2-Y/enh.B promoter motif immediately upstream of CCAATbox. Together, we tentatively assigned FLA I-E, I-H, I-K as classical class I genes and nine other genes as nonclassical class I genes.
This promoter analysis also revealed potentially distinct gene regulation of other FLA class I genes. For example, FLA I-S and I-O genes had an intact conserved promoter motif. However, I-S class I gene did not have intact coding region nor expression in fibroblast (Table 5). Also I-O class I gene did not maintain 5 highly conserved deduced amino acid residues in its peptide binding groove. Other class I genes, FLA I-A, I-Q lacked NFkB1, 2 and IFN-c binding sites, and FLA I-J, I-L lacked Y/enh.B site.

Single Nucleotide Polymorphism (SNP)s
Overall, the SNP rate found in FLA (BAC sequence versus MHC homozygous 1.96 WGS contigs) was at least twice as much higher than the SNP rate in average heterozygous region in the WGS cat genome, (0.00411 versus 0.0017) and slightly higher but nearly equivalent to the SNP rate found in two human HLA haplotypes (6COX and 6QBL) ( Table 4). A total number of coding SNP (CDS SNP) is higher than human HLA (732 versus 341). A total of 193 CDS SNPs were found in class II and class I genes. Of these, both class II DRB4 and DRB1 genes had a higher number of nonsynonymous CDS SNPs than synonymous ones, and two class I genes (FLA-I, -F, -H) had similar tendencies (Table 6). These data suggest that those genes are under positive selection.

New Endogenous Retrovirus Sequences
Phylogenetic analysis of three FLA endogenous retrovirus sequences using the neighbor-joining method ( Figure 4) suggested that in addition to previously described baboon-derived RD114 retrovirus (or ECE1) [42][43][44] the other two sequences showed equidistance to FeLV derived [45,46] and RD114 endogenous sequences, but more similar to the sequence recently submitted to GenBank as comparative genome initiative research derived from brown bat (Myotis Lucifugus) BAC sequence. Because all sequences described above maintained retrovirus POL region, newly identified feline retrovirus sequences was assigned as FERVmlu1 (previously FERV1) and FERVmlu2.

MHC Class I Related Genes
Of four MHC class I-related genes (MIC) which encodes clectin type NK receptor ligands in HLA, none of them had full length exon sequences when compared with human MICA transcripts (data not shown). Interestingly, neither cat nor dog genomes maintain multigene families of KIR and Ly49 found in primates and rodents genomes, respectively. These evidences may suggest distinct control systems for NK cells in cats and dogs.
Comparison of genomic structures in cat, dog, human, mouse, and opossum MHC genes A proportionally scaled MHC genomic structure was presented for four mammalian genomes (cat, human, mouse, and dog) and one marsupial genome (opossum) (Figure 9). The MHC region spanning from KIFC (except mouse H2 which has a translocation in this region, so that H2 here compared from Rps 28) to UBD plus 3 olfactory receptor genes was compared in these MHCs. The result depicts striking similarity in gene contents and order of framework genes from marsupial through mammalian evolution. Three MHC (opossum, human, and mouse) have one contiguous gene content, suggesting depiction of an ancestral form of MHC, while two MHC (cat and dog) have a same split form of MHC at TRIM31 and TRIM26 in the class I region as compared with human HLA. In dog MHC, these two pieces were located on two chromosomes (cfa12qcen, cfa35qter), while in cat MHC, these were located on a single chromosome by an inversion (FcaB2 qcen, FcaB2pter) as previously described [19]. Further, two class I genes in dog MHC were located with two additional chromosomes (cfa7, cfa18). The size variation of MHC from about 3.3 Mbp (cat and dog excluding percentromeric, subtelomeric regions) to about 5 Mbp in opossum was also seen in this analysis. The difference in size observed here is mainly due to the magnitude of class I gene amplification and size of class II/III border regions. Cat MHC consists of 650 Kbp class I gene region, spanning from BAT1 to POU5F1 maintaining 17 class I genes/gene fragments, while human and dog MHC have only 2-3 class I genes in this region. Mouse H2 has 7 class I genes and there are no class I genes in opossum MHC in this region. Accordingly, class I gene amplification seemed to have occurred in a species-specific fashion. Additional evidences that e.g. opossum MHC, class I genes were amplified in the class II region, human HLA have at least 11 class I genes in the HLA-A region between the ZNRD1 and MOG genes and in mouse H2, at least 15 class I genes were found between Abcf1 and Trim26 genes, all support adaptive evolution of this importance immune system. Interestingly, the sizes of class II/III border regions vary in each MHC. Cat and dog MHC have approximately 400 Kbp in these regions. In cat MHC, this region was occupied with LINE repeats however, in dog and opossum there are multiple BTNL genes. These evidences reaffirmed the dynamic nature of evolution and maintenance of genome organizations in MHC.   Figure 9. Comparisons of MHC genomic structures in cat, human, mouse, dog, and opossum. Framework genes in class II, III, I regions were shown as green, blue, red boxes, respectively. Forward and reverse orientations of each gene were shown above and below line, respectively. Classical class II antigen coding genes/gene fragments were shown in orange and classical and non-classical class I genes were shown in black. doi:10.1371/journal.pone.0002674.g009