Natural selection on marine carnivores elaborated a diverse family of classical MHC class I genes exhibiting haplotypic gene content variation and allelic polymorphism

Abstract

Pinnipeds, marine carnivores, diverged from terrestrial carnivores ~45 million years ago, before their adaptation to marine environments. This lifestyle change exposed pinnipeds to different microbiota and pathogens, with probable impact on their MHC class I genes. Investigating this question, genomic sequences were determined for 71 MHC class I variants: 27 from harbor seal and 44 from gray seal. These variants form three MHC class I gene lineages, one comprising a pseudogene. The second, a candidate nonclassical MHC class I gene, comprises a nonpolymorphic transcribed gene related to dog DLA-79 and giant panda Aime-1906. The third is the diversity lineage, which includes 62 of the 71 seal MHC class I variants. All are transcribed, and they minimally represent six harbor and 12 gray seal MHC class I genes. Besides species-specific differences in gene number, seal MHC class I haplotypes exhibit gene content variation and allelic polymorphism. Patterns of sequence variation, and of positions for positively selected sites, indicate the diversity lineage genes are the seals’ classical MHC class I genes. Evidence that expansion of diversity lineage genes began before gray and harbor seals diverged is the presence in both species of two distinctive sublineages of diversity lineage genes. Pointing to further expansion following the divergence are the presence of species-specific genes and greater MHC class I diversity in gray seals than harbor seals. The elaboration of a complex variable family of classical MHC class I genes in pinnipeds contrasts with the single, highly polymorphic classical MHC class I gene of dog and giant panda, terrestrial carnivores.

Introduction

Major histocompatibility complex (MHC) class I molecules are central components of the immune systems in all jawed vertebrates (Flajnik and Kasahara 2010). Their principal function is to bind peptide antigens inside cells and present them on the cell surface to natural killer (NK) and CD8 T cell receptors (Marrack et al. 2008; Neefjes et al. 2011; Parham and Ohta 1996). Such interactions between lymphocyte receptors and MHC class I ligands regulate aspects of both the innate and adaptive immune response to intracellular infections, particularly viral infections (Bashirova et al. 2011; Cheent and Khakoo 2009; Orange and Ballas 2006). The evolution of pathogens to evade, escape, and subvert host immunity imposes strong selection pressures on MHC class I genes. These combine with other evolutionary forces, notably genetic drift, to drive the rapid evolution of MHC class I alleles, genes, and haplotypes. As a consequence,MHC class I diversity, which focuses on exons 2 and 3 that encode sites of peptide-binding and lymphocyte receptor contact, can evolve to become species-specific (Adams and Parham 2001) or population-specific (Kelley et al. 2005).

At the levels of alleles, genes, haplotypes, and populations, the only species in which MHC class I variation has been studied to the depth and breadth where generalizations can be made with confidence is arguably the human species (http://www.ebi.ac.uk/imgt/hla/ (Robinson et al. 2011), http://www.allelefrequencies.net/ (Gonzalez-Galarza et al. 2011)). In the human MHC (the HLA complex), there are six functional MHC class I genes, all of which are fixed in the human genome. Three of these, HLA-A, HLA-B, and HLA-C, are highly polymorphic genes, for which thousands of allelic variants have been defined. Consequently, most individuals are HLA-A, HLA-B, and HLA-C heterozygotes, and they differ from each other in HLA class I type. This diversifies immune systems within human populations, spreading both risks and benefits and increasing the likelihood that any new, emerging pathogen will encounter effective immune defense from some fraction of the population (Trowsdale 2011). Contrasting with HLA-A, HLA-B, and HLA-C, the HLA-E, HLA-F, and HLA-G genes exhibit little variation, being essentially conserved in both structure and function. Because of the qualitative difference in their polymorphism, HLA-A, HLA-B, and HLA-C are collectively referred to as classical MHC class I genes (or MHC class Ia genes), whereas HLA-E, HLA-F, and HLA-G are referred to as nonclassical MHC class I genes (or MHC class Ib genes). In addition to their low polymorphism, nonclassical class I genes tend to be older and more divergent than their classical counterparts; they also embrace a wider range of more specialized functions and can be expressed by fewer types of cells (Hofstetter et al. 2011; Rodgers and Cook 2005). However, these tendencies vary with the species and the gene. For example, human HLA-G is expressed only by extra-villous trophoblast, whereas HLA-E has a ubiquitous tissue distribution like HLA-A, HLA-B, and HLA-C (Moffett and Loke 2006).

Although the classical/nonclassical distinction is clearly defined for the human species, where the functional MHC class I genes are all fixed, that is not the situation in many other mammalian species, where variation in MHC class I gene content makes a significant contribution to the overall genetic diversity. Such species include other primates (Boyson et al. 1996; Doxiadis et al. 2011; Kulski et al. 2004; Saito et al. 2012), rodents (Rada et al. 1990), equids (Tallmadge et al. 2010), swine (Tanaka-Matsuda et al. 2009), and ruminants (Ballingall et al. 2008; Codner et al. 2012). For MHC class I genes that are present on only a subset of MHC haplotypes, the functional effect of the difference between the presence and absence of a gene will likely be greater than that achieved by alternative combinations of alleles of a fixed gene. Thus, gene content variation and allelic polymorphism can be seen to represent alternative, and even complementary, strategies for diversifying MHC class I haplotypes.

Having some similarity to their human counterparts, MHC class I haplotypes of domestic dog (a terrestrial carnivore) have a relatively fixed content of functional MHC class I genes (Kennedy et al. 2012; Wagner 2003; Wagner et al. 2005). A major difference in the dog is a translocation that split the MHC class I region between chromosomes 12 and 18 (Yuhki et al. 2007). The region on chromosome 12 that is syntenic to the HLA region (Breen et al. 1999; Mellersh et al. 2000) contains a highly polymorphic MHC class I gene (DLA-88) and two nonpolymorphic MHC class I genes (DLA-12 and DLA-64) (Burnett et al. 1997). On chromosome 18 is a moderately polymorphic MHC class I gene (DLA-79) (Burnett and Geraghty 1995; Graumann et al. 1998). The only known example of gene content variation is a duplication of the DLA-88 gene, which is estimated to be a feature of ~10 % of DLA haplotypes, but can vary between breeds of dog (Kennedy et al. 2012; Ross et al. 2012). The three functional MHC class I genes of the giant panda (Ailuropoda melanoleuca), a dog-like carnivore, are related to the dog MHC class I genes. Aime-1906 is moderately polymorphic and related to DLA-79. Aime-128 and Aime-152 are related to each other and to DLA-12, DLA-64, and DLA-88. Aime-128 appears highly polymorphic and may be present in two copies on some haplotypes, whereas Aime-152 appears conserved (Pan et al. 2008). By way of contrast, the domestic cat, representing the more distantly related cat-like carnivores, has a distinctive set of 12 MHC class I genes (Yuhki et al. 2008).

Pinnipeds are marine, dog-like carnivores that diverged from their terrestrial relatives ~45 million years ago (mya) in North America. Following the split, pinniped ancestors adapted to living in the marine environment (Arnason et al. 2006; Cao et al. 2000; Nery et al. 2012). Such a radical change in lifestyle involved adaptation to a different microbial universe and its pathogens, with the potential for consequent changes in pinniped MHC class I and II genes. Because MHC class II genes are phylogenetically more conserved, and their regions of variability are limited to a single exon, they have been more tractable to analysis than MHC class I. Indeed, MHC class II genes have been studied in pinniped populations for more than two decades, initially by analysis of restriction fragment length polymorphisms (Slade 1992) and subsequently by sequencing and high-resolution typing (Bowen et al. 2002, 2004, 2006; Cammen et al. 2011; Decker et al. 2002; Hoelzel et al. 1999; Weber et al. 2004).

In contrast, very little is known of pinniped MHC class I. First described was a cDNA sequence encoding anMHC class I polypeptide from a harbor seal of the Eastern Pacific subspecies (Phoca vitulina richardii) (Zhong et al. 1998). More recently, analysis of six Hawaiian monk seals (Monachus schaunslandi) identified six closely related MHC class I variants, each one coming from a different seal (Aldridge et al. 2006). Here we examine MHC class I gene diversity in two seal species from the North Sea: the Eastern Atlantic subspecies of the harbor seal (Phoca vitulina vitulina) and the gray seal (Halichoerus grypus). Monk seals, harbor seals, and the gray seal are all species of phocid pinniped (also known as true seals or earless seals), to be distinguished from the otarid seals (sea lions, also called fur seals, and the walrus). Harbor seal and gray seal are closely related and sympatric species that diverged recently from a common ancestor. Two estimates for the divergence time are ~5–7mya (Arnason et al. 2006) and ~1–4 mya (Higdon et al. 2007).

Our purpose with this study was to perform a systematic analysis of the gray and harbor seal MHC class I gene families. Because each seal was to be investigated in considerable detail, the study was focused on relatively small panels of animals, consisting of eight harbor seals and four gray seals. However, the information obtained should provide sufficient knowledge and insight to facilitate future, broad-based population and epidemiological studies of MHC class I variation in pinniped populations.

Unlike MHC class II, which controls CD4 T cell immune responses, MHC class I is dedicated to immunity mediated by NK cells and CD8 T cells (Neefjes et al. 2011). These killer lymphocytes of innate and adaptive immunity provide crucial defense against viral infections, which are heavily influenced by MHC class I type as illustrated by the striking correlations of HLA class I polymorphism with the progress and outcome of human HIV infections (Bashirova et al. 2011). That recent epidemics of phocine distemper virus (PDV) have severely reduced European harbor seal populations, but not gray seal populations, suggests that investigation of MHC class I variation in seal populations could lead to the identification of genetic factors that provide resistance or susceptibility to PDV (Dietz et al. 1989; Harkonen et al. 2006, 2007; Harris et al. 2008).

Materials and methods

Seal sample collection

Tissue from harbor seals was sampled at two sites: in 2003 at Abertay Sands (AB) on the east coast of Scotland in St. Andrews Bay (latitude 56° 23 00′, longitude −2° 45 00′) and in 2002 at The Wash (WSH) on the east coast of England in East Anglia (latitude 52° 92 00′, longitude 0° 05 00′). Blood samples were obtained from four free-living harbor seals at Abertay Sands, whereas spleen samples were taken from four dead harbor seals discovered at The Wash. Blood samples were obtained from four gray seals, housed temporarily at the Sea Mammal Research Unit (University of St. Andrews, Scotland, UK), following capture at Abertay Sands: two individuals in 2003 and the other two individuals in 2004. All United Kingdom Home Office Procedures performed on the seals were conducted under license no. 60/3303, following the ethical review process at St. Andrews University.

Using microsatellite DNA polymorphisms as markers (Goodman 1998), it was shown that harbor seal populations are outbred and that there is movement of seals between them, resulting in admixture. The seals studied here have not been genotyped for markers other than MHC class I, so we have no independent assessment of their relatedness. For several reasons, including the MHC class I genotypes, it seems unlikely that any of the seals studied were closely related. The WSH harbor seals were randomly discovered animals that died of PDV infection. The AB harbor seals were randomly selected and caught. The gray seals all had different mothers, but we do not know if their parents were in any way related. The populations from which the seals derived are large and outbred; there is no evidence of family structure and the seals were caught or discovered outside of the breeding season, a time when family structure is not maintained.

RNA and DNA extraction

Seals were manually restrained, while blood was withdrawn from the extradural vein into EDTA vacutainers using the Vacutainer system (BD Biosciences, San Jose, CA, USA). Peripheral blood was decanted from the vacutainers, mixed 1:1 with RPMI 1640 medium, and overlaid onto Histopaque-1077 (Sigma-Aldrich, St. Louis, MO, USA) for separation of the mononuclear cells (PBMC) according to the manufacturer’s guidelines. The layer of PBMCs was removed, washed twice in RPMI 1640, and used immediately for RNA extraction. PBMCs were homogenized in TriReagent™ (Sigma-Aldrich) and RNA extracted according to the manufacturer’s guidelines. After assessing the quality of the RNA by agarose gel electrophoresis, cDNAwas synthesized using Superscript III (Invitrogen, Carlsbad, CA, USA).

With removal of the PBMC layer from the Histopaque-1007, the pellet consisting of red blood cells (RBC) and granulocytes (polymorphonuclear leukocytes) was used as the source for gDNA. The pellets were washed twice with 70 ml of RPMI 1640 (700 g for 10 min at room temperature). The pellet was resuspended, by swirling, in 40 ml of phosphate-buffered saline (PBS) containing 1 % saponin for 10 min to lyse RBC. After centrifugation at 2,000×g for 5 min at 4 °C, the granulocyte pellet was washed twice in PBS with centrifugation at 2,000×g for 5 min at 4 °C. The washed granulocytes were then resuspended in 10 ml of 100 mM NaCl, 10 mM Tris pH 7.5, 25 mM EDTA, containing 0.5 % SDS and 0.5 mg/ml proteinase K, and incubated at 50 °C for 6 h. The same solution was used to digest samples of spleen (10 mm²) obtained from the cadaveric, Wash harbor seals, but with the time of digestion being overnight. From the proteinase K digest, gDNA was extracted with phenol/chloroform/isoamyl alcohol in a standard two-step manner. Extracted gDNA was then precipitated with ethanol and suspended in Tris-EDTA buffer.

PCR amplification and sequencing of seal MHC class I genes

From comparison of MHC class I cDNA sequences from a range of mammalian species, conserved regions were identified. PCR primers based on these conserved regions were used to amplify multiple gene fragments from the 5′- and 3′-ends of MHC class I genes from one harbor seal and one gray seal. Sequences determined for 48 of these amplification products were used to identify conserved sequences as priming sites that could potentially amplify all phocid seal MHC class I genes. Of the two primers chosen, one (5′-CTACGTGGACGACACGCAGTTC-3′) corresponded to sequence at the 5′-end of exon 2, encoding the α₁ domain, and the other (5′-GGKCAKGGTGGGCMAVTCTC-3′) corresponded to sequence at the 3′-end of exon 8, specifying the 3′ UTR. When used together in PCR on gDNA, these two primers produced an amplicon of ~2.65 kb. Both primers corresponded to conserved regions of mammalian MHC class I and were shown to amplify all the phocid MHC class I genes recovered from the initial primer screen. The primer pair was also tested for its capacity to amplify cDNA from a second harbor seal and a second gray seal: this PCR produced an amplicon of 1,064 or 1,062 bp (only Phvi-N*04), results confirming the general applicability of these priming sites for amplification from both gDNA and cDNA.

A limitation of the above-described primer pair is placement of the 5′-primer in exon 2, which produces incomplete coverage of that exon and thus lack of sequence for the 5′-end of exon 2, in and beyond the priming site. To rectify this deficiency, we applied 5′-RACE to the genomic DNA of one harbor seal and one gray seal, for the purpose of identifying a conserved sequence for primer design in exon 1 or intron 1. This analysis used the SMART RACE system according to the manufacturer’s guidelines (BD Biosciences). The results enabled the design of a degenerate sense primer (5′-GGCTCCCACTCCMTGARGT-3′) that corresponded to sequence in intron 1 just upstream of exon 2. This primer was used in PCR with an antisense primer (5′-GTGGGGATGGGGGTCGTGAC-3′) derived from a conserved sequence in intron 2. From gDNA, this second pair of primers amplified a segment from intron 1 to intron 2 that included all of exon 2. The 3′-end of this ~300-bp amplification product was designed to overlap the 5′-end of the 2.65-kb amplification product by 200 bp. This overlap allowed sequences obtained from the two amplifications to be matched, thus giving complete coding region sequences (exons 2–8) for mature seal MHC class I proteins.

PCR amplification of the 2.65-kb segment was performed on 200 ng of gDNA, using the Advantage 2 genomic polymerase mix (BD Biosciences). Following initial denaturation of 2 min duration, 28 cycles of amplification (94 °C for 25 s, 61 °C for 25 s, and 68 °C for 3 min) were performed, with a final extension of 15 min at 68 °C. PCR amplification of the 300-bp segment used the same conditions except the initial denaturation time was 3 min and the cycling denaturation time was 40 s.

PCR products from the 2.65-kb and 300-bp amplifications were cloned using the TOPO TA cloning kit (Invitrogen). For the 2.65-kb amplicon, 72 clones from each of two or three (when necessary to confirm sequences) independent PCR reactions were sequenced in their entirety from both directions by end-sequencing using vector primers followed by primer-walking. The clones from the 300-bp amplicon were sequenced using the M13F site in the vector. To extend the 5′-end sequence of individual 2.65 kb clones, the sequences of clones from the 300-bp amplicon were compared to the 2.65-kb sequences in the 200-bp region of overlap. Clones containing an identical overlapping sequence with a cloned 2.65-kb sequence were then sequenced from the other direction using the M13R primer to complete and confirm the sequence.

Nucleotide sequencing was performed on a Beckman Coulter CEQ 2000 instrument and the data assembled using the STADEN PACKAGE (Staden et al. 2000). Sequences were only used for further analysis once their validity was confirmed, either by them being identified in two independent experiments, or in two or more seals. All confirmed seal MHC class I sequences are deposited in GenBank, their accession numbers being given in Table 1.

Table 1.

GenBank accession numbers

Allele	Accession number
Harbor seal MHC class I
Phvi-N*01:01:02	JX218867
Phvi-N*01:01:02	JX218871
Phvi-N*02:01:01P	JX218872
Phvi-N*02:01:02P	JX218873
Phvi-N*02:02P	JX218874
Phvi-N*03:01	JX218878
Phvi-N*03:02	JX218879
Phvi-N*03:03	JX218880
Phvi-N*03:04:01	JX218881
Phvi-N*03:04:02	JX218882
Phvi-N*04:01	JX218883
Phvi-N*04:02	JX218884
Phvi-N*04:03	JX218885
Phvi-N*05:01	JX218886
Phvi-N*05:02	JX218887
Phvi-N*06:01:01	JX218894
Phvi-N*06:01:02	JX218895
Phvi-N*07:01	JX218896
Phvi-N*08:01:01	JX218875
Phvi-N*08:01:02	JX218876
Phvi-N*09:01	JX218877
Phvi-N*10:01	JX218888
Phvi-N*11:01:01	JX218889
Phvi-N*11:01:02	JX218890
Phvi-N*11:02	JX218891
Phvi-N*12:01	JX218893
Phvi-N*12:02	JX218892
Gray seal MHC class I
Hagr-N*01:01	JX218870
Hagr-N*01:02	JX218868
Hagr-N*01:03	JX218869
Hagr-N*02:01P	JX218937
Hagr-N*03:01:01	JX218918
Hagr-N*03:01:02	JX218919
Hagr-N*03:01:03	JX218923
Hagr-N*03:02	JX218920
Hagr-N*04:01	JX218917
Hagr-N*05:01	JX218901
Hagr-N*05:02	JX218902
Hagr-N*05:03	JX218903
Hagr-N*05:04	JX218904
Hagr-N*05:05	JX218916
Hagr-N*06:01:01	JX218907
Hagr-N*06:01:02	JX218909
Hagr-N*06:02	JX218910
Hagr-N*06:03	JX218908
Hagr-N*07:01	JX218898
Hagr-N*07:02	JX218899
Hagr-N*07:03	JX218900
Hagr-N*07:04	JX218897
Hagr-N*08:01	JX218922
Hagr-N*08:02	JX218906
Hagr-N*09:01:01	JX218924
Hagr-N*09:01:02	JX218925
Hagr-N*09:01:03	JX218929
Hagr-N*09:02:01	JX218926
Hagr-N*09:02:02	JX218928
Hagr-N*09:03	JX218927
Hagr-N*10:01	JX218933
Hagr-N*11:01:01	JX218930
Hagr-N*11:01:02	JX218931
Hagr-N*11:01:03	JX218932
Hagr-N*12:01	JX218911
Hagr-N*12:02:01	JX218912
Hagr-N*12:02:02	JX218913
Hagr-N*13:01	JX218935
Hagr-N*14:01:01	JX218914
Hagr-N*14:01:02	JX218905
Hagr-N*14:02	JX218915
Hagr-N*15:01	JX218934
Hagr-N*15:02	JX218921
Hagr-N*15:03	JX218936

Recombination and phylogenetic analysis

Sequences were aligned using CLUSTAL X (Thompson et al. 1997) and manually edited, when necessary, using Bioedit version 7.0.5.3 (Hall 1999). Detection of recombinant sequences was performed using the Recombination Detection Program (RDP) 3.41 (Martin et al. 2010). Recombination signals were analyzed in aligned sequence datasets, for each species individually and the two species together, with a maximum accepted probability value of P=0.05. Neighbor-joining (NJ) and maximum likelihood (ML) phylogenetic analyses, the average p distance, and the codon-based Z test of selection were all calculated with MEGA version 4.1 (Tamura et al. 2007). The phylogenetic trees were constructed using either the Tamura–Nei method or the Tamura three-parameter model with pairwise/partial deletion comparison and 1,000 replicates. Construction of the ML trees also used the nearest neighbor interchange optimization method.

In the statistical analyses of the gray and harbor seal MHC class I genes, we made comparisons with MHC class I genes from other carnivore species, for which the sequences were determined by other investigators. These MHC class I sequences are listed here with their GenBank accession numbers, where available, and references to relevant publications: Eastern Pacific harbor seal (PLA-1: U888874) (Zhong et al. 1998); Six Hawaiian monk seal MHC class I sequences (Mosc1.1 through to Mosc1.6) were not available from GenBank, so we copied them from the publication (Aldridge et al. 2006); domestic dog (DLA-12:U55030, DLA-64:U55027, DLA-79:Z25418 and DLA-88:U55028) (Burnett et al. 1997; Burnett and Geraghty 1995; Mellersh et al. 2000; Wagner 2003; Wagner et al. 2005); domestic cat (FLA-A1:U07667, FLA-A10:U07668, FLA-A23:U07669, FLA-A24:U07670, FLA-B2:U0767, FLA-B9:U07672, FLA-X8:U07674, and FLA-X10:U07668) (Yuhki et al. 1989, 2007, 2008; Yuhki and O’Brien 1988, 1994); and giant panda (Aime-126:EU162661, Aime-128:EU162656, and Aime-152:EU162657) (Pan et al. 2008).

Nomenclature for seal MHC class I variants

Because mammalian MHCs can contain multiple polymorphic genes with divergent allelic lineages, it can be difficult to distinguish genes from alleles on the basis of sequence comparisons alone, and such assignments remain uncertain until confirmed, either by segregation patterns, whole-haplotype sequencing, or other independent forms of data. We therefore adopted a provisional nomenclature for the groups of seal MHC class I sequences, following precedent set by the Immuno Polymorphism Database (IPD-MHC; http://www.ebi.ac.uk/ipd/ (Robinson et al. 2010)).

For each seal species, the genomic MHC class I sequences were assigned to different groups according to their sequence similarity: 12 groups were defined for the harbor seal and 15 groups for the gray seal. Sequences assigned to the same group had >99.5 % sequence similarity in the combined intron sequences and >98.5 % sequence similarity in the combined intron and exon sequences. These conservative criteria were designed to prevent alleles from different genes being assigned to the same group. Consistent with the achievement of this goal, no seal had more than two variants in any of the groups. Thus, each group is a candidate for being a discrete MHC class I gene. The harbor seal groups were named Phvi-N*01-N*12 and the gray seal groups Hagr-N*01-N*15, in which the prefix N indicates they have not been officially designated as a locus, but are candidates for that status. For the three harbor seal groups (Phvi-N*01, N*02, and N*03) for which some seals have two variants, this provides corroborating evidence that they do represent discrete genes, likewise for eight gray seal groups (Hagr-N*03, N*05, N*06, N*07, N*08, N*09, N*14, and N*15). In addition, Hagr-N*01 and N*02 are almost certainly discrete genes because of their highly divergent sequences. Because of the conservative criteria used to define the groups, it is possible that for genes with divergent lineages of alleles (like human HLA-B), such lineages would be assigned to different groups. In that case, two or more groups would define a single gene. Groups of seal MHC class I variants for which their status as genes or allelic lineages is uncertain are Phvi-N*04–N*12, Hagr-N*04, and N*10–N*13.

Because the variants within a group are almost certainly alleles of the same gene, they have been distinguished with a nomenclature similar to that used for HLA class I and II alleles (Marsh et al. 2010). For example, the Phvi-N*03:01 and Phvi-N*03:02 alleles differ by nonsynonymous substitution in the coding region, whereas the Phvi-N*03:01:01 and Phvi-N*03:01:02 alleles differ by synonymous substitution in the coding region.

Selection analysis

Estimation of the ratio between the average rate of non-synonymous substitution and the average rate of synonymous substitution (dN/dS or alternatively ω) was performed by maximum likelihood (ML) using the F3 X 4 model of codon frequencies and implemented in PAML, v3.14 (Yang 1997). For the site analysis, the likelihood of tree topology was estimated using four site-specific models that allow the ω ratio to vary among codons, but for which the site-specific patterns are identical across all lineages. These four models, M1a (nearly neutral), M2a (positive selection), M7 (β), and M8 (β and ω) were implemented using the program codeml in the PAML package. To test for positive selection, a likelihood ratio test (LRT) was used to compare a model that does not allow ω>1 with a model that does. LRTs were performed on M1a vs M2a and M7 vs M8. The Bayes empirical Bayes approach was then used to identify the codons with ω>1 (Yang et al. 2005).

Calculation of pairwise differences

Two datasets were assembled: the first consisted of coding region sequences for the harbor and gray seal MHC class I sequences reported here; the second consisted of previously reported coding region sequences of human and chimpanzee MHC class I genes, deposited in curated, specialist databases. HLA-A, HLA-B, HLA-C, HLA-E, HLA-F, HLA-G, and HLA-H were obtained from IMGT-HLA (http://www.ebi.ac.uk/imgt/hla/ (Robinson et al. 2011)). For the highly polymorphic HLA-A, HLA-B, and HLA-C, genes only one allele for each allotype group was included in the dataset, e.g., HLA-A*01:01, HLA-A*02:01, and HLA-A*03:01. Sequences for the chimpanzee MHC class I genes (Patr-A, Patr-AL, Patr-B, Patr-C, Patr-E, Patr-F, Patr-G, and Patr-H) were obtained from IPD-MHC (http://www.ebi.ac.uk/ipd/ (Robinson et al. 2010)).

Pairwise distances for the sequences in each dataset were calculated in MEGA using pairwise deletion and p-dist (Tamura et al. 2007). Histograms were made by sorting the p distance values into bins corresponding to increments of 0.005 substitutions/site. The number of pairwise comparisons in each bin was calculated as a percentage of the total number of pairwise comparisons and plotted as a histogram.

Results

MHC class I diversity was studied in harbor seals and gray seals. The analysis of harbor seal will be described first, followed by analysis of the gray seal and then comparison of the two species.

Eleven groups of transcribed MHC class I variants in the harbor seal

MHC class I genes were isolated and sequenced from the gDNA of eight harbor seals. From individual seals, between 8 and 13 different MHC class I variants were defined. On the basis of nucleotide sequence similarity, each MHC class I variant was assigned to 1 of 12 groups: Phvi-N*01–N*12, each containing variants with >99.5 % sequence similarity in the intron sequences and >98.5 % sequence similarity for the combination of introns and exons (Fig. 1a).

Fig. 1.

The harbor seal has a diverse system of MHC class I genes. a Each column corresponds to 1 of 12 groups (Phvi-N*01–12) of MHC class I variant, predicted from sequence comparisons to represent different genes. Each of the first eight rows corresponds to an individual seal, with the presence of an MHC class I group indicated by gray shading. Variants within a group are distinguished by four digit no-menclature: for example, N*03:01 and N*03:02. To save space, the third and fourth digits are only included when needed to discriminate variants. Rows 9 and 10 give the numbers of alleles and allotypes for each group. Dagger, the two alleles of Phvi-N*01 differ by a single synonymous nucleotide substitution in exon 5. The dashed horizontal line separates the Abertay Sands (AB) seals from the Wash (WSH) seals. b Under “gDNA” is given the number of MHC class I variants detected in the genome of each seal; under “mRNA” is the estimated number of transcribed genes for each seal. Under “Groups” is given the number of MHC class I groups represented in each animal. Under “Site” is given the geographical location (AB or WSH) where the blood or spleen sample from each animal was taken. c Amino acid sequence differences between Phvi-N*03 and Phvi-N*04 allotypes. Presence of a dot denotes identity with the N*03:01 or N*04:01 allotype. Shaded gray are positions predicted to contact bound peptide. Extrapolating from the human system, contacts could be made between residue 62 and TCR, and between residue 83 and the single seal killer cell immunoglobulin-like receptor (Hammond et al. 2009). Allotypic differences not shown in c: Phvi-N*05:02 differs from Phvi-N*05:01 by substitution of proline for alanine at position 185 in the α₃ domain; Phvi-N*11:02 differs from Phvi-N*11:01 by substitution of histidine for tyrosine at position 320 in the cytoplasmic tail; Phvi-N*12:02 differs from Phvi-N*12:01 by substitution of glutamine for glutamate at position 154 in the α₂ domain

Each of these groups represents a different candidate MHC class I gene. Phvi-N*02P was judged to be a pseudogene, because its sequence contains several deletions in exons 2 and 3 and lacks a donor splice site in intron 2 that links these two exons. The sequences of the other 11 candidate genes indicate that they are transcribed and translated to produce functional MHC class I molecules. To assess gene transcription, mRNA was isolated from the cells of Pv149 and Pv150, seals whose genotypes cover all 12 groups (Fig. 1a). Complementary DNA was made from the RNA, subjected to PCR, and then cloned and sequenced. No cDNA clones corresponded to Phvi-N*02P, consistent with it being a nonfunctional gene. In contrast, cDNA clones corresponding to all the other 11 MHC class I genes were obtained. Moreover, for both Pv149 and Pv150, cDNA clones corresponding to each MHC class I gene in the seal’s genotype were identified. Thus, 11 of the 12 groups of harbor seal MHC class I genes were shown to be transcribed. By extrapolation from the patterns of transcription in Pv149 and Pv150, the numbers of transcribed MHC class I genes for the other six harbor seals were estimated (Fig. 1b).

Allelic polymorphism and variable MHC class I gene content in the harbor seal MHC

That individual seals can carry two variants of Phvi-N*01, Phvi-N*03, and Phvi-N*02P demonstrates that each of these three groups corresponds to a different MHC class I gene, and their variants represent different alleles. Because all 12 groups are represented in seal Pv149 (Fig. 1a), the other nine groups must represent a minimum of five and a maximum of nine transcribed MHC class I genes. Thus in total, there is a minimum of seven and a maximum of 11 transcribed harbor seal MHC class I genes, plus the pseudogene. Contrasting with Pv149, seal Pv1912 could have as few as five transcribed class I genes. Six groups (Phvi-N*01, Phvi-N*02P, Phvi-N*03, Phvi-N*04, Phvi-N*07, and Phvi-N*10) are present in all individuals, whereas the other six groups are variably represented. Thus, harbor seal MHC haplotypes are demonstrated to vary in MHC class I gene content.

Eight of the 11 groups of transcribed MHC class I genes exhibited allelic polymorphism, but for three of them, the substitutions are synonymous. Of the remaining five groups, three have two allotypes, one has three allotypes, and one has four allotypes (Fig. 1a). The four Phvi-N*03 allotypes differ by between one and eight substitutions (Fig. 1c). The substitutions are at nine positions: two in the α₁ domain (positions 62 and 83) and seven in α₂ (positions 108, 114, 116, 147, 152, 155, and 156). All but one of these positions are predicted to influence the recognition of MHC class I by T cell and NK cell receptors (Bjorkman et al. 1987; Boyington and Sun 2002; Marrack et al. 2008; Saper et al. 1991; Vivian et al. 2011; Wang et al. 2002). The extent of the allelic differences and their location within the class I molecule are typical of classical MHC class I molecules.

The three Phvi-N*04 allotypes differ by substitutions at one or two positions. The first is at TCR-contacting position 62 in the α₁ domain and the second at peptide-contacting position 147 in α₂ (Fig. 1b). Despite the presence of three alleles in the panel, each of the eight seals has only one Phvi-N*04 allele (Fig. 1a). This is an unusual phenomenon, for which several explanations could be considered. Given the stringent criteria (>98.5 % sequence similarity) used for assigning variants to the same N* group, it is possible that Phvi-N*04 is one of two divergent allelic lineages of the same gene, which we assigned erroneously to different N groups. Because all eight seals have only one form of Phvi-N*07 and Phvi-N*10, these two groups are both candidates for such a second lineage. Alternatively, the second lineage could have been missed completely by our method for capturing MHC class I genes. A second possible explanation for the phenomenon, is that each of the eight harbor seals has one MHC haplotype containing the Phvi-N*04 gene, combined with a second MHC haplotype that lacks Phvi-N*04. A third possibility is that all eight seals are homozygous for Phvi-N*04 alleles, an improbable occurrence by chance, but a possible consequence of selection. A final and fourth possibility is that the Phvi-N*04 gene was translocated to the X chromosome, so the six male seals in the panel (Fig. 1b) could only express one Phvi-N*04 allele, and for the two females to be homozygous for Phvi-N*04 alleles. Precedent for this latter mechanism is translocation of the mouse killer cell immunoglobulin-like receptors from the leukocyte receptor complex on chromosome 7 to the X chromosome (Welch et al. 2003).

Phvi-N*01 is a recombinant related to a dog and giant panda MHC class I gene

For phylogenetic analysis, the harbor seal MHC class I sequences were split into a 5′-segment that included exons 2 and 3, encoding the peptide-binding site, and a 3′-segment that includes exons encoding the α₃, transmembrane, and cytoplasmic domains (Fig. 2). In trees obtained from the 5′-genomic segment (Fig. 2a), harbor seal MHC class I genes form three branches: the first containing the Phvi-N*02P pseudogene, the second containing the divergent transcribed gene Phvi-N*01, and the third containing the other ten transcribed genes, Phvi-N*03-12. In this 5′-segment, Phvi-N*01 clusters strongly with dog DLA-79 and giant panda Aime-1906. In contrast, this cluster is not reiterated in trees constructed from the 3′-segment sequences (Fig. 2b). Here, DLA-79 and Aime-1906 are at the base of the tree, whereas Phvi-N*01 is situated well within the cluster formed by the transcribed harbor seal MHC class I genes and closest to Phvi-N*08. Similar analysis of coding region sequences, showed that Phvi-N*01 again grouped with DLA-79 and Aime-1906 in the 5′-segment (Fig. 2c), but with the other transcribed harbor seal MHC class I genes in the 3′-segment (Fig. 2d). These results pointed to Phvi-N*01 being a recombinant MHC class I gene, a conclusion supported by investigation of the carnivore MHC class I gene sequences using the recombination detection program RDP3 (Martin et al. 2010). Emerging from this analysis was strong statistical support for only one recombination: the event that formed PhviN*01 (P<0.001) from one gene resembling DLA-79/Aime-1906 and a second resembling Phvi-N*08 (Fig. 2e).

Fig. 2.

Phvi-N*01 is a recombinant: the 5′-segment resembles dog DLA-79 and giant panda Aime-1906, the 3′-segment resembles other transcribed seal MHC class I. a–d Phylogenetic trees constructed from carnivore MHC class I sequences. The recombinant Phvi-N*01 gene taxa are boxed, whereas the likely parental genes are shown with gray shading (Phvi-N*08) or black shading with white lettering (DLA-79/ Aime-1906). a 5′-segment: comprising exon 2, intron 2, and exon 3. b 3′-segment: comprising exons 4–8, introns 4–7, and 3′-untranslated region. Intron 3 was not included in the analysis because it contains the recombination break point. c 5′-segment coding region sequence: comprising exons 2 and 3. d 3′-segment coding region sequence: comprising exons 4–8 and 3′-untranslated region. DLA, FLA, and Aime denote dog (Burnett et al. 1997), cat (Beck et al. 2005), and giant panda (Pan et al. 2008) MHC class I genes, respectively. All other MHC class I taxa are from harbor seal for which the species designation, Phvi, is only given for Phvi-N*01 and Phvi-N*08. “P” denotes pseudogene. NJ and ML analysis gave trees with similar topology. Shown are NJ trees, rooted at the midpoint; for nodes with bootstrap support >50 % the value of the support is shown. e Cartoon illustrating the structure of the Phvi-N*01 recombinant, which has a 5′-segment resembling dog DLA-79 (and Aime-1906) and a 3′-segment resembling Phvi-N*08. Black-filled boxes with white lettering denote exons with greater sequence similarity to their DLA-79 and Aime-1906 counterparts, whereas gray-shaded boxes denote exons with greater sequence similarity to their Phvi-N*08 counterparts. The predicted site of recombination is within intron 3 between exons 3 and 4

With the exception of the 5′-part of Phvi-N*01, the transcribed harbor seal MHC class I genes form a seal-specific gene family that has no single equivalent in dog, giant panda, or cat (Fig. 2). The relationship of the 5′-part of Phvi-N*01 with dog DLA-79 and giant panda Aime-1906 indicates that the last common ancestor of seal, dog, and panda (~45 mya) had an ancestral MHC class I gene with a 5′-segment resembling that of modern Phvi-N*01, DLA-79, and Aime-1906. This gene was passed to the pinniped line, where it recombined with a member of the seal-specific family to produce the recombinant structure seen in Phvi-N*01. That no cat MHC (FLA) class I gene clusters with DLA-79, Aime-1906, and Phvi-N*01 is consistent with the divergence of cat-like carnivores from dog-like carnivores ~52 mya, at a time ~7 mya before the separation of dog, giant panda, and seals (Arnason et al. 2006).

Fourteen groups of transcribed MHC class I variants in the gray seal

MHC class I genes were isolated and sequenced from the genomic DNA of four gray seals. From individual seals, between 11 and 18 different MHC class I variants were obtained. By applying the same criteria used to group harbor seal variants, each gray seal MHC class 1 variant was assigned to 1 of 15 groups: Hagr-N*1-15 (Fig. 3a). Of these candidate MHC class I genes, the Hagr-N*02P pseudogene has the same deleterious exon 2 and 3 mutations as harbor seal Phvi-N*02P, as well as lacking the donor splice site in intron 2.Hagr-N*02P is clearly orthologous to Phvi-N*02P: they have 98.6 % sequence similarity compared to <84% sequence similarity with other seal MHC class I genes.

Fig. 3.

The gray seal has a diverse system of MHC class I genes. Each column corresponds to 1 of 15 groups (Hagr-N*01–15) of MHC class I variant, predicted from sequence comparisons to represent different genes. Each of the first four rows corresponds to an individual seal, with the presence of an MHC class I group indicated by gray shading. Variants within a group are distinguished by four digit nomenclature, for which the third and fourth digits are only included when needed to discriminate variants. Rows 5 and 6 give the numbers of alleles and allotypes for each group. b Under “gDNA” is given the number of genes detected in the genome of each seal; under “mRNA” is the estimated number of transcribed genes for each seal. Under “Groups” is the number of MHC class I groups represented in each seal. Under “Site” is given the geographical location (Abertay Sands) where the seals were captured. c Differences in the amino acid sequences of the N*05, N*06, N*07, and N*15 allotypes. Presence of a dot denotes sequence identity with the :01 allotype. Shaded gray are positions predicted to contact bound peptide. Allotypic differences not shown in c: Hagr-N*03:02 differs from Hagr-N*03:01 by substitution of threonine for isoleucine at position 124 in the α₂ domain; Hagr-N*08:01 differs from Hagr-N*08:02 by eight substitutions: six in α₁ (positions 43, 45, 60, 61, 69, and 70) and two in α₂ (positions 145 and 158); Hagr-N*09:02 differs from Hagr-N*09:01 by eight substitutions: three in α₁ (positions 21, 69, and 70) and five in α₂ (positions 108, 127, 134, 152, and 174); Hagr-N*09:03 differs from Hagr-N*09:01 by substitution of phenylalanine for valine at position 21 of α₁; Hagr-N*14:02 differs from Hagr-N*14:01 by substitution of serine for alanine at position 40 in α₁, and two substitutions in the cytoplasmic domain (glycine for serine at position 328 and threonine for methionine at position 334) that are the consequence of recombination with Hagr-N*04

We were able to isolate and study only the 5′-segment of Hagr-N*01, indicating that its 3′-segment differs from that of Phvi-N*01. In the 5′-segment, Hagr-N*01 exhibits >97.8 % sequence similarity with Phvi-N*01, showing the two 5′-segments are orthologous. Our failure to amplify the 3′-segment of Hagr-N*01 by PCR could have a trivial explanation, such as localized variation within the priming sites. Alternatively, the 3′-segments of Hagr-N*01 and Phvi-N*01 could have substantially divergent sequences. The recombinant nature of Phvi-N*01 (Fig. 2) raises the intriguing possibility that Hagr-N*01 is not a recombinant and that both its 3′- and 5-segments resemble DLA-79/Aime1906. In that case Hagr-N*01 would represent the ancestor of Phvi-N*01.

The sequences of the other 13 genes pointed to them being functional. To assess gene transcription, an analysis of MHC class I cDNA was made for gray seals HgQ and HgU1. The genotypes of these two individuals cover almost all the groups, exceptions being Hagr-N*08 and Hagr-N*10. Complementary DNA clones corresponding to Hagr-N*03, Hagr-N*04, Hagr-N*05, Hagr-N*06, Hagr-N*07, Hagr-N*09, Hagr-N*11, Hagr-N*12, Hagr-N*13, Hagr-N*14, and Hagr-N*15 were obtained and were those expected from each seal’s genotype. In the absence of a complete genomic sequence, the functionality of Hagr-N*01 was not tested. No cDNA corresponding to Hagr-N*02P was obtained, consistent with it being nonfunctional. By extrapolating from these results, we estimated the number of transcribed MHC class I genes in the other two gray seals (Fig. 3b).

Allelic polymorphism and variable MHC class I gene content in the gray seal MHC

That individual gray seals can carry two variants of the Hagr-N*03, Hagr-N*05, Hagr-N*06, Hagr-N*07, Hagr-N*08, Hagr-N*09, Hagr-N*14, and Hagr-N*15 genes shows these eight groups represent different MHC class I genes (Fig. 3a). Because Hagr-N*01 and Hagr-N*02P are divergent and orthologous to Phvi-N*01 and Phvi-N*02P, respectively, it is very likely that these two also represent discrete genes. This leaves only Hagr-N*04, and N*10-N*13, as the remaining candidate genes. Minimally, this panel of gray seals has 13 MHC class I genes and pseudogenes, maximally it has 15. Six genes were represented in all four individuals, whereas nine groups exhibited presence/absence polymorphism. Thus, gray seal MHC haplotypes are demonstrated to vary in MHC class I gene content.

Polymorphism is particularly evident for the Hagr-N*05, Hagr-N*06, Hagr-N*07, and Hagr-N*15 genes, each represented by three to five alleles. Contrasting with Phvi-N*03 and Phvi-N*04 (Fig. 1c), the variable residues of these four polymorphic gray seal MHC class I molecules do not concentrate at the peptide-binding site or known receptor-contacting positions of the α₁ and α₁ domains (Fig. 3c). This raises the possibility that the polymorphic MHC class I genes of gray and harbor seals have been subject to different selection pressures. Alternatively, they may affect peptide binding and presentation in an indirect manner.

Gray and harbor seals have the same lineages and sublineages of MHC class I genes

Overall, the systems of MHC class I genes in harbor seal and gray seal are seen to be similar. They comprise the same three types of genes: a diverse seal-specific family of transcribed and likely functional MHC class I genes; a single and divergent transcribed gene related to a gene present in dog and giant panda, but not cat; and a divergent pseudogene (Fig. 4). These three groups represent three lineages of seal MHC class I genes. The lineage corresponding to the seal-specific expansion of transcribed MHC class I genes, we will call the diversity lineage.

Fig. 4.

Lineages and sublineages of seal MHC class I genes. Phylogenetic trees were constructed from nucleotide sequences of the coding regions of MHC class I genes: NJ and ML analysis giving trees with similar topology. Shown is a NJ tree, which is rooted at the midpoint. For nodes with bootstrap support >50 %, the values of the support are shown. Species designations are Phvi (harbor seal); Hagr (gray seal), Mosc (Hawaiian monk seal) (Aldridge et al. 2006), PLA-A1 (Eastern Pacific harbor seal) (Zhong et al. 1998), domestic dog (DLA), giant panda (Aime), and domestic cat (FLA). Excluded from the analysis was exon 1 that encodes the leader peptide, which is not a part of the mature MHC class I molecule. The incomplete sequences included in the analysis had at least 50 % of the coding region sequence. Only the sequence for the 5′-segment of Phvi-N*01 and Hagr-N*01 was used because the 3′-segment of Phvi-N*01 is recombinant (see Fig. 2), and the 3′-segment of Hagr-N*01 eluded our methodology. Pairwise sequence comparison showed Mosc1.2 is in sublineage 1 of the diversity lineage, whereas Mosc1.1, 1.4, 1.5, and 1.6 are in sublineage 2. Mosc1.3 is a recombinant, with exon 2 coming from sublineage 2, and exons 3–8 being like Mosc1.2 of sublineage 1. The recombinant nature of Mosc1.3 contributes ambiguity to the node separating the two groups of Hawaiian monk seal MHC class I genes

Within the diversity lineage, some orthologous relationships between gray and harbor seal MHC class I genes could be discerned (Fig. 4). Phvi-N*07 appears orthologous to Hagr-N*07, Phvi-N*12 to Hagr-N*12, and Phvi-N*06 to Hagr-N*09. Conversely, some genes appear specific to one of the species, harbor seal Phvi-N*08, and gray seal Hagr-N*03, Hagr-N*04, Hagr-N*11, and Hagr-N*15. Such assignments, however, can only be provisional because of the small sample sizes. Forming part of the diversity lineage are the Pacific harbor seal MHC class I variant (PLA-1) (Zhong et al. 1998) and the six Hawaiian monk seal MHC class I variants (Mosc1.1–6) (Aldridge et al. 2006). PLA-1 is a distinctive member of the cluster formed by Phvi-N*07 and Hagr-N*07, suggesting that it is orthologous to these two genes. The monk seal MHC class I sequences add two “new” branches to the tree, consistent with monk seals having diverged from harbor and gray seals 14–18 mya (Higdon et al. 2007).

Gray and harbor seal MHC class I genes of the diversity lineage further divide into two sublineages. Sublineage 1 comprises harbor seal Phvi-N*05, Phvi-N*10, and Phvi-N*11 and gray seal Hagr-N*05 and Hagr-N*14; sublineage 2 comprises all the other MHC class I groups in the diversity lineage (Fig. 4). Distinguishing sublineage 1 from sublineage 2 are substitutions at four positions in the α₁ domain and seven in the α₂ domain (Fig. 5a). These are sites that influence peptide binding (Bjorkman et al. 1987; Saper et al. 1991), TCR interaction (Marrack et al. 2008), and killer cell immunoglobulin-like receptor (KIR) interaction (Boyington and Sun 2002; Vivian et al. 2011), whereas the amino acid substitutions that distinguish the members of sublineage 1 are predominantly not at such sites (Fig. 5b). For sublineage 1, the surface contacting the T cell receptor has more basic and polar residues, compared to more acidic residues for sublineage 2. In addition, the variation at peptide-binding positions is more limited for sublineage 1 than sublineage 2. Monk seal MHC class I sequences also partition between the two sublineages: Mosc1.2 and 1.3 in sublineage 1 and Mosc1.1, 1.4, 1.5, and 1.6 in sublineage 2 (Fig. 4). That both sublineages are represented in gray, harbor, and monk seals shows they existed prior to separation of monk seals, from harbor and gray seals, 14–18 mya (Higdon et al. 2007).

Fig. 5.

Sequence motifs at functional sites in the α₁ and α₂ domains distinguish two sublineages within the diversity lineage of seal MHC class I molecules. a Shown are the amino acid residues present at 11 positions distinguishing the two sublineages of diversity lineage MHC class I proteins. For sublineage 2, the relative frequencies of alternative amino acids at a position are given by the size of the colored rectangles. b Shown is the diversity of amino acid residues at positions that distinguish the members of sublineage 1. Colors denote the types of amino acid: aliphatic (I, L, V): olive green; basic (K, R): dark blue; acidic (D, E): red; polar (N, Q, S, T): gray; small hydrophobic (A): light blue; small polar (G): gold; bulky hydrophobic (W): green; hydrophobic (M): light green; aromatic hydrophobic (Y): mid blue; aromatic charged (H): pink; proline (P): purple. Under “Function,” P denotes contact with peptides, T denotes contact with T cell receptors, and K denotes contact with killer cell immunoglobulin-like receptors (KIR), as deduced from structural and functional studies on other species, mainly human and mouse (Bjorkman et al. 1987; Boyington and Sun 2002; Marrack et al. 2008;Saper et al. 1991; Vivian et al. 2011; Wang et al. 2002)

That harbor seals Pv149 and Pv150 have three members of sublineage 1 (Phvi-N*05, Phvi-N*10, and Phvi-N*11) shows that at least two different MHC class I genes contributed to the evolution of sublineage 1 in this species. At least two MHC class I genes also contributed to the evolution of sublineage 1 in gray seals because Hagr-N*05 and Hagr-N*14 represent different genes. The presence of this sublineage 1 in three seal species, with its preservation of a distinctive site for binding peptides and interacting with T cell receptors, points to sublineage 1 MHC class I molecules having a shared, distinctive, and beneficial function.

Application of the RDP3 program to the gray seal MHC class I sequences detected only one well-supported (P<0.01) recombination event, from which the product was Hagr-N*14:01. In exons 1–6 and introns 1–5, Hagr-N*14:01 is almost identical to Hagr-N*14:02, but their sequences diverge at the 3′-end of the gene, in exon 7, intron 7, and exon 8. Here, Hagr-N*14:01 is identical in sequence to Hagr-N*04. Thus, Hagr-N*14:01 was formed by recombination between Hagr-N*14 of sublineage 1 and Hagr-N*04 of sublineage 2 (Fig. 4).

Positive selection on seal MHC class I has diversified peptide presentation

To investigate the effects of natural selection on the diversity lineage of seal MHC class I, we used pairwise comparisons to estimate the synonymous (dS) and nonsynonymous (dN) substitution rates. Both harbor and gray seal MHC class I have dN/dS ratios>1, evidence for positive diversifying selection. Comparison of the 5′-segment, encoding α₁ and α₂ domains, with the 3′-segment, encoding α₃ and the transmembrane domain, showed selection had focused on α₁ and α₂ (Fig. 6a). Correspondingly, a codon-based Z test significantly rejected (P<0.05) the null hypothesis H0 (dN = dS) in favor of H1 (dN>dS) for full-length harbor seal MHC class I sequences. Significance in this analysis was due to the α₁ and α₂ domains (P<0.05) because selection was not detected in the α₃ and transmembrane domains. For gray seal MHC class I, the Z test approaches significance (P=0.065), but only when analyzing the α₁ and α₂ domains in isolation, a difference that could reflect the smaller gray seal dataset (Fig. 6a).

Fig. 6.

Harbor and gray seal MHC class I genes of the diversity lineage have been subject to positive diversifying selection. a Comparison of mean dN and dS for exons 2–8 encoding the mature MHC class I protein, for exons 2 and 3 encoding the peptide binding site (α₁ and α₂ domains) and for exon 4 encoding the α₃ domain. The null hypothesis (H0) was tested against the alternative hypothesis (H1), using a one-tailed Z test. The significance level (α) is indicated when H0 was significantly rejected. b Shown are the results of likelihood ratio tests of positive selection, comparing site models without (M1a and M7) and with (M2a and M8) positive selection. Comparisons were performed for the harbor seal diversity lineage, Phvi-N*03 alone, and the gray seal diversity lineage. Shown are the likelihood ratio test statistic (2Δl) and the significance level (α) indicating the statistical support for positive selection. c Shown are positively selected codons identified by the Bayes empirical Bayes approach. Black boxes denote positively selected residues detected by models M8 and M2a and light gray boxes denote positively selected residues identified with model M8 only. Confidence levels are indicated by asterisks: *P<0.05 and **P<0.01. Below each residue are the contacts it makes with peptides (P), αβ T cell receptors (T), killer cell immunoglobulin-like NK cell receptors (K), and Ly49 NK cell receptors (Bjorkman et al. 1987; Boyington and Sun 2002; Marrack et al. 2008; Saper et al. 1991; Vivian et al. 2011; Wang et al. 2002). d The ribbon diagrams show positions of positive selection in the α₁ and α₂ domains of harbor seal (left) and gray seal (right) MHC class I molecules of the diversity lineage. Yellow color denotes a significance level of P<0.05, and red color denotes a significance level of P<0.01. The bound peptide is colored blue. The ribbon diagrams were derived from the crystal structure of HLA-C*03:04 (Boyington et al. 2000)

To identify sites under positive selection, ML methods using a substitution model were used (Yang 1997). The two models that allow ω>1 (M2a and M8) were shown to be significantly more likely than the corresponding models that do not, confirming the results from the Z test that positive diversifying selection has driven the variation in harbor seal and gray seal MHC class I (Fig. 6b). The selection tests, models 8 and 2a, identified 24 sites in the harbor seal and 28 sites in the gray seal to have been subject to positive selection (P<0.05) (Fig. 6c, d). Within the α₁ and α₂ domains, almost all the residues identified are predicted to contact bound peptides, the αβ T cell receptors, the KIR of NK cells, or the Ly49 NK cell receptors (Fig. 6c, d).

The four Phvi-N*03 variants with nonsynonymous substitutions have a dN/dS ratio of 1.4, and ML methods showed that the models permitting ω>1 are significantly more likely than their equivalents that do not (α=0.001) (Fig. 6b). Of the eight positions that distinguish the Phvi-N*03 variants, positions 114 and 152 have been subject to positive selection (P>0.95), and both are predicted to contact peptide (Fig. 6c). In contrast, no significant evidence for selection was found for the three Phvi-N*04 variants.

The diversity lineage of seal MHC class I genes appears equivalent to classical MHC class I genes of humans and chimpanzees

The distribution of pairwise differences between MHC class I genes is very similar for gray seal (Fig. 7a) harbor seal (Fig. 7b) and combination of the two (Fig. 7c). The four components to this distribution, with increasing distance, are comparison between alleles, orthologs, and within sublineage 1 of the diversity lineage; other comparisons within the diversity lineage; comparisons with N*01; and comparisons with N*02P. We then compared the seal distributions to those of MHC class I gene sequences from human and chimpanzee (Fig. 7d), hominid species that have been extensively characterized for MHC class I (IMGT-HLA; http://www.ebi.ac.uk/imgt/hla/ (Robinson et al. 2011) and IPD-MHC; http://www.ebi.ac.uk/ipd/ (Robinson et al. 2010)), and which have a divergence time of 7–10 mya (Chatterjee et al. 2009) comparable to the higher estimate for that for gray and harbor seals (Arnason et al. 2006).

Fig. 7.

Comparison of MHC class I diversity in closely related species of phocid seal and hominid. For various groups of MHC class I coding region sequences, the complete set of pairwise differences was calculated and the distribution of values plotted as a histogram. a Comparison of all gray seal MHC class I variants. b Comparison of all harbor seal MHC class I variants. c Comparison of the combined sets of gray and harbor seal MHC class I variants. The horizontal lines above the histogram denote the range of values for particular types of comparison: I between alleles, orthologs, and between diversity lineage genes of sublineage 1; II other comparisons between diversity lineage genes; III between N*01 and other genes, 5′-segment only; and IV between N*02P and other genes. d Comparison of the combined set of functional MHC class I variants from two hominid species: humans, Homo sapiens (Hosa) and chimpanzee, Pan troglodytes (Patr). The range values are for the following comparisons: V between alleles of the same gene, VI between alleles from different classical class I genes, VII between classical and nonclassical class I genes, and VIII between alleles from different nonclassical class I genes. e Comparison of differences between alleles of the three classical MHC class I genes of humans and chimpanzees. All comparisons are between alleles of the same gene, including interspecies comparisons. The results from the three genes are combined in the histogram. f Comparison of the differences between pairs of alleles from different hominid classical MHC class I genes. g Comparison of the differences between pairs of hominid MHC class I alleles, in which one allele comes from a classical MHC class I gene and the other comes from a nonclassical MHC class I gene. h Comparison of the differences between pairs of alleles from hominid nonclassical MHC class I genes. This includes pairwise differences between alleles from the same gene as well as those between alleles of different genes

The comparison of alleles within the same hominid classical class I gene (A, B, or C) gives a distribution (Fig. 7e) with comparable range (0.01–07) to that observed for genes in the seal diversity lineage (Fig. 7c). In contrast, comparison of alleles from different hominid classical class I genes gives a distribution of differences (Fig. 7f) with a range (0.07–0.125) that is greater, and non-overlapping, than that for seal diversity lineage genes. The distribution of comparisons between a hominid classical gene and a hominid nonclassical gene (Fig. 7g) also has a greater (0.65–0.16), and non-overlapping, range compared to that for the seal diversity lineage. Within this range are the comparisons with seal N*01, pointing to this nonpolymorphic gene (Figs. 1 and 3) being a nonclassical MHC class I gene. The distribution of comparisons with the seal N*02P pseudogene has a range (0.185–0.21) that is higher, and non-overlapping, than that for all functional hominid MHC class I genes (Fig. 7d). This result shows that seal MHC class I genes are not inherently constrained to be closer in sequence similarity than hominid MHC class I genes.

From these comparisons, we see that the diversity generated by MHC class I genes of the seal diversity lineage is comparable to that of hominid classical MHC class I genes. That the seal diversity lineage consists of classical MHC class I genes is supported by the distribution of positively selected sites (Fig. 6), which strongly argues for the variation having been selected to change the binding and presentation of peptides to T cell and NK cell receptors. In seals, the evolution of a family of MHC class I genes that exhibits gene sequence and gene content variation appears to have achieved a similar functional endpoint as the smaller number of fixed and highly polymorphic MHC class I genes in humans and chimpanzees.

Discussion

We find the MHC class I genes of harbor and gray seals to have much in common, consistent with their recent divergence ~5–7 mya. Both species have the same three lineages of MHC class I genes. One lineage comprises a pseudogene, and the other two lineages comprise expressed, and likely functional, genes. Divergent N*01, which shares ancestry with dog DLA-79 and giant panda Aime-1906, is the only representative of the first lineage of transcribed genes. In contrast, the second lineage of transcribed genes, the diversity lineage of seal MHC class I genes, is a gene family that has no counterpart in terrestrial carnivores (dog, giant panda, and cat) and consists of at least six genes in the harbor seal and 12 genes in the gray seal.

The divergent sequence of N*01 and its lack of functional polymorphism are consistent with it being a nonclassical MHC class I gene. In contrast, the collective properties of the diversity lineage genes are characteristically the ones associated with classical MHC class I genes. They contribute to MHC haplotype variability in two ways: gene content variation and allelic polymorphism. This variation is founded on two sublineages of MHC class I genes, present in both harbor and gray seals, and distinguished by functional motifs in residues of the α₁ and α₂ domains that determine the presentation of peptide antigens to the antigen receptors of CD8 T cells and KIR NK cell receptors. The evolutionary process of positive selection that drove this bifurcation occurred prior to the divergence of harbor and gray seals and has continued following the separation of harbor and gray seal ancestors. As a consequence, individual seals vary in the number of MHC class I variants they express: six to ten variants among the harbor seals we studied and 9–16 variants among four gray seals studied. Such use of a combination of gene content variation and allelic polymorphism to generate MHC diversity appears to be a common feature of mammalian species (Ballingall et al. 2008; Boyson et al. 1996; Codner et al. 2012; Doxiadis et al. 2011; Kulski et al. 2004; Rada et al. 1990; Saito et al. 2012; Tallmadge et al. 2010; Tanaka-Matsuda et al. 2009), but is not the case in the human species, where gene content variability is absent and diversity arises purely from the allelic polymorphism of three fixed genes.

An obvious limitation to this exploratory investigation of MHC class I gene diversity in seals is the small number of seals studied. Consequently, all allele and gene frequency estimates can only be considered provisional, and they will undoubtedly be modified by future studies of larger populations. In addition, there is uncertainty for some of the seal MHC class I groups, whether they uniquely define a gene or just one of two constituent sets of alleles. Despite these limitations, we have obtained unambiguous evidence that allelic polymorphism is a mechanism that contributes to the diversity of seal MHC class I haplotypes. Thus, two or more alleles, encoding different protein variants, were characterized for three harbor seal and ten gray seal MHC class I groups in the diversity lineage. The allelic polymorphism of Phvi-N*03 and Phvi-N*04 has characteristics associated with selection for changes in the specificity of peptide presentation to CD8 T cells and in the interaction of MHC class I molecules with KIR NK cell receptors, as does the extended motif that distinguishes sublineages 1 and 2 of the diversity lineage.

Other aspects to the polymorphism, particularly for gray seal MHC class I, do not conform to this paradigm. Such differences may have arisen from other types of selection upon the NK cell response. For example, residue 134, which exhibits positive selection in the gray seal, is known to be a contact site for mouse Ly49 receptors (Wang et al. 2002). Three types of NK cell receptor (CD94:NKG2, Ly49, and KIR) recognize MHC class I and have alternatively been diversified in different mammalian species (Parham et al. 2010). In gray and harbor seals, Ly49 and KIR are each represented by a single functional gene that is highly conserved (Hammond et al. 2009). This raises the possibility that in these species, variation in MHC class I is exclusively used to diversify the interactions with these conserved NK cell receptors. Such a situation could produce patterns of MHC class I polymorphism that differ from those in species like humans and mice, where both the NK cell receptors and their MHC class I ligands are highly variable (Carlyle et al. 2008; Rajalingam 2012).

There are hints in our data for MHC class I differences between the Abertay Sands and Wash populations of harbor seals (Fig. 1a, b). Groups Phvi-N*05 and Phvi-N*06 MHC class I were present in all Abertay Sands seals, but absent from the Wash seals. In contrast, the Abertay Sands seals lacked Phvi-N*03 polymorphism, this group being represented only by Phvi-N*03:01, whereas the Wash seals lacked Phvi-N*03:01, but had Phvi-N*03:02, Phvi-N*03:03, and Phvi-N*03:04.

The 1988 outbreak of PDV is estimated to have killed ~50 % of European harbor seals (Dietz et al. 1989). Characterizing this epidemic was a lower mortality around Scotland (~50 %) compared to eastern and southern England (~60 %), which could have a genetic basis (Goodman 1998). In 2002, a further epidemic of PDV in 2002 also killed ~50 % of the European harbor seals (Harkonen et al. 2006). Thus in recent time, European harbor seal populations have undergone strong selection from PDV. Survivors of the 2002 epidemic were the Abertay Sands seals we studied, whereas the Wash seals did not survive. It is therefore possible that the Phvi-N*03, Phvi-N*05, and Phvi-N*06 differences between these two groups of harbor seals are a reflection of their relative resistance and susceptibility to PDV.

Gray seal populations were relatively unaffected by the PDV epidemics. Although being susceptible to PDV infection, gray seals remain asymptomatic and clear the virus rapidly. Circumstantial evidence suggests that healthy, PDV-infected gray seals actually contributed to the spread of PDV between distant populations of harbor seal (Hall et al. 2006; Hammond et al. 2005; Harkonen et al. 2006). That gray seals have a greater number of MHC class I genes, and more polymorphic MHC class I genes than harbor seals, could in general contribute to the resilience of their immune defense against PDV infection. In particular, Hagr-N*09, which is polymorphic and present in all four gray seals, appears orthologous to Phvi-N*06, a characteristic of the four harbor seals in our study that survived the 2002 PDV epidemic. These speculations provide testable hypotheses for future population and epidemiological studies.