Evolution of the V, D, and J gene segments used in the primate γδ T-cell receptor reveals a dichotomy of conservation and diversity

Abstract

γδ T cells are an immunological enigma in that both their function in the immune response and the molecular mechanisms behind their activation remain unclear. These cells predominate in the epithelia and can be rapidly activated to provide an array of responses. However, no homologous γδ T-cell populations have been identified between humans and mice, and our understanding of what these cells recognize as ligands is limited. Here we take an alternative approach to understanding human γδ T-cell ligand recognition by studying the evolutionary forces that have shaped the V, D, and J gene segments that are used during somatic rearrangement to generate the γδ T-cell receptor. We find that distinctly different forces have shaped the γ and δ loci. The Vδ and Jδ genes are highly conserved, some even through to mouse. In contrast, the γ-locus is split: the Vγ9, Vγ10, and Vγ11 genes represent the conserved region of the Vγ gene locus whereas the remaining Vγ genes have been evolving rapidly, such that orthology throughout the primate lineage is unclear. We have also analyzed the coding versus silent substitutions between species within the V and J gene segments and find a preference for coding substitutions in the complementarity determining region loops of many of the V gene segments. Our results provide a different perspective on investigating human γδ T-cell recognition, demonstrating that diversification at particular γδ gene loci has been favored during primate evolution, suggesting adaptation of particular V domains to a changing ligand environment.

Distinct from αβ T cells, γδ T cells represent an alternative lineage of T cells characterized by the expression of a heterodimeric T-cell receptor (TCR) composed of a γ- and a δ-chain. Despite substantial effort, the specific role of γδ T cells in host responses against tumors, viruses, and pathogenic bacteria remains unclear. γδ T cells represent a minority of circulating lymphocytes in the blood, but are the majority of tissue-resident T cells in epithelial tissues such as the gut, lungs, and reproductive tract (1). Defined roles for γδ T cells are few; for example, they have been shown to play an important role in wound healing in the mouse (2), and a particular population of human γδ T cells expressing a Vγ9Vδ2 receptor are known to respond rapidly and potently to small pyrophosphate-containing molecules, a response thought to function in pathogenic bacterial detection and elimination of tumor cells (3). However, the immunological function of most γδ T cells remains unclear.

The enigma of γδ T cells also extends to the molecular details of how they respond to antigens through their TCRs. Detailed analysis of ligand recognition by γδ T cells has primarily focused on the mouse nonclassical class I molecule T22. γδ TCR recognition of T22 is unlike that of conventional αβ TCR recognition of classical MHC molecules presenting peptide fragments, which use an assemblage of their complementarity determining region (CDR) loops to dock on the peptide/MHC composite surface. Instead, T22-reactive γδ T cells recognize T22 through a specific amino acid motif in their CDR3δ loop (W… EGYEL) (4, 5); this loop is necessary and sufficient for recognition of T22 (6). However, the predominant use of CDR3δ in ligand recognition does not seem to be ubiquitous across all γδ T-cell populations, nor is it conserved between mouse and human. Other candidate ligands have been proposed for γδ T cells; most of these are self-molecules such as classical and nonclassical MHC proteins [i.e., CD1c (7), I-E^k (8), and MICA (9)], the aforementioned pyrophosphate-containing small molecules (3), and the F1 subunit of the mitochondrial ATPase (10). However, a foreign protein, the glycoprotein I from HSV-1, has also been characterized as a ligand for murine γδ T cells (11), suggesting that ligand targets are not exclusively self-molecules. Recognition of glycoprotein I was direct, without requirement for processing and presentation by MHC molecules (12). Thus, γδ T-cell ligands represent a structurally and chemically diverse array of molecules, suggesting that antigen recognition by these T cells may not have a conserved architecture like that seen with αβ TCR recognition of peptide MHC (13).

In mice and humans, γδ T cells have a restricted repertoire of V and J gene segments that are used for TCR rearrangement. In humans, the δ-locus is embedded within the α-locus on chromosome 14; only three true Vδ genes exist—Vδ1, Vδ2, and Vδ3—although particular Vα genes are infrequently used in δ-chain rearrangement and thus have a Vδ distinction: Vδ4 (Vα14), Vδ5 (Vα29), Vδ6 (Vα23), Vδ7 (Vα36), and Vδ8 (Vα38) (14). The human Vγ repertoire, located in the γ locus on chromosome 7, is also small, with 12 Vγ genes, only seven of which are known to be functional (Vγ1, Vγ5P, Vγ6, Vγ7, and Vγ10 are pseudogenes). The low diversity is also reflected in the number of J gene segments for both δ and γ (four and five, respectively). This contrasts with significant diversity at the α and β loci [Vα, 47; Vβ, 54; Jα, 58; and Jβ, 14 gene segments (15)] and has led to the suggestion that the apparent low diversity of γδ TCRs reflects their recognition of conserved self-proteins of low variability (16). The additional restriction of certain V segments in different tissues further reduces the V domain diversity within γδ T-cell populations. In the mouse, the use of particular V domains may dictate the location and function of these cells (17).

Contrasting with the low number of V and J gene segments used in γδ TCR rearrangement is the ability of these receptors to incorporate multiple D segments during rearrangement (18). These segments can be incorporated in the forward or reverse frame, and in humans, most of the reading frames do not contain stop codons. This ability to use multiple D segments makes the CDR3δ the most amino acid-diverse loop across all rearranged receptors, including αβ TCRs and preaffinity-matured antibody molecules (19). The contrast between low variability of the V domains (which encode the CDR1 and CDR2 loops) with that of the extensive potential diversity of the CDR3δ loop poses a quandary for predicting the nature of the ligands recognized by these TCRs. If recognition is centered on the CDR3δ, which we have shown previously to be true for T22 recognition (4–6), then are some γδ T cells recognizing highly variable antigens potentially derived from exogenous sources? Or are most γδ T-cell populations recognizing evolutionarily conserved ligands that exhibit low polymorphism? Understanding the nature of γδ T-cell ligands has been a central focus in γδ T-cell biology; however, despite substantial effort, few have been well characterized. We have taken an evolutionary approach to understanding ligand recognition by γδ T cells by focusing on the evolutionary pressures that have shaped the repertoire of V, D, and J gene segments in humans and nonhuman primates. Previous work has revealed important similarities and differences between humans and chimpanzees at specific genes (20), but a comprehensive analysis of these loci across divergent nonhuman primate species has not been performed to our knowledge. Our analysis incorporates genomic location and arrangement, nucleotide and amino acid polymorphisms, and measurement of nonsynonymous versus synonymous substitutions per site of these loci across five species representing time points across approximately 45 million years of primate evolution. We find that diversification varies substantially across the γ- and δ-loci and that some variable gene segments exhibit signatures of strong selection for coding changes at the regions encoding the CDR1 and CDR2 loops. Together, this analysis provides a foundation upon which to build our understanding of the nature of γδ TCR recognition of ligand.

Results

Genomic Organization of V, D, and J Gene Segments.

Gene sequences corresponding to human Vδ, Jδ, Dδ, Vγ, and Jγ homologues in four nonhuman primate species (Pan troglodytes, Pongo pygmaeus, Macaca mulatta, and Callithrix jacchus) were identified based on their homology to human, mapped to their genomic location, and are shown in comparison with human in Fig. 1. Identification of the Vδ5 gene from P. pygmaeus, which was lacking in the genome sequence, was done through amplification and sequencing by using gene-specific primers designed from human Vδ5; therefore, the exact position of this gene is undefined.

Fig. 1.

Genomic organization of the V, D, and J gene segments comprising the primate γδ TCR: (A) Vδ, (B) Jδ and Dδ, (C) Vγ and (D) Jγ from humans, Pan troglodytes (Patr), Pongo pygmaeus (Popy), Macaca mulatta (Mamu), and Callithrix jacchus (Caja). V and J genes are shown as arrows to indicate coding direction. D segments are shown as blocks. Functional genes are shown in black, pseudogenes in gray. Scale for each locus is shown in the upper right of each section. Insertions and deletions are indicated as gray triangles with insertions shown as pluses and deletions as minuses, with distances in kilobases.

In all species examined, the Vδ locus was embedded within the Vα locus, as previously described in humans and mice. Strong conservation of this region across the five species was reflected in the high similarity of gene content and organization. Dot plot analysis, whereby long stretches of genomic sequence can be compared and their similarities and differences represented visually on a dot-based matrix, reflects this high conservation (Fig. 2A) as a mostly straight line with few gaps. Indeed, this conservation extends to the whole Vα locus (Fig. 2B). Three core regions of high homology within the Vδ region were identified: one containing the Vδ4 genes; a central block containing the Vδ1, Vδ5, Vδ6, Vδ7, and Vδ8 gene segments; and a third block that contains the Vδ2 and Vδ3 genes (Fig. 1A); these two genes also flank the Dδ and Jδ segments (Fig. 1B). A second Vδ4 (Vα14) gene, called Vδ4n2, was identified in all the nonhuman primates species examined and thus appears to have been deleted in the lineage leading to humans. Vδ4n2 appears to be a pseudogene in these four species based on the presence of several stop codons throughout the predicted coding sequences. Vδ4 was not found in C. jacchus, perhaps because of a 45.7-kb deletion between Vδ4n2 and Vδ6 relative to the human sequence. Distance differences (represented in Fig. 1 as gray triangles with the indicated length difference in kilobases) between the three blocks of genes was common, representing both insertion and deletion events in these regions. In addition, the Vδ3 gene sequence in C. jacchus has several stop codons, and thus appears to be nonfunctional in this species. However, the organization of the Vδ gene segments relative to humans was overall highly conserved.

Fig. 2.

Dot-plot analysis reveals a dichotomy of genomic evolution between the δ- and γ-loci. Dotplot analysis was applied to the genomic regions encoding the Vδ (A), Vα (B), Vγ (C), and Vβ (D) genes between humans, P. troglodytes (Patr), P. pygmaeus (Popy), M. mulatta (Mamu), and C. jacchus (Caja) to determine the relative stability of these regions across primate evolution. The relative positions of the V genes are shown with arrows, with black indicating functional genes and gray pseudogenes. Black dots on the plot correspond to regions with greater than 90% identity. Gray dots indicate greater than 80% identity between the two regions being compared. The position of the Vδ5 gene from P. pygmaeus is uncertain, indicated by a box. The size of the regions compared are shown in base pairs after the species name. In D, pink boxes indicate the nature of the genes located in regions of repetitive sequences.

Organization of the Dδ and Jδ gene segments was also very highly conserved across the species examined, both in location (Fig. 1B) and sequence identity. Examination of the D segment recombinational signal sequence (RSS) flanking regions also shows high identity across species (Fig. S1), suggesting the majority of these segments are functional for V-D-J rearrangement. The exception is the Dδ1 sequences from M. mulatta and C. jacchus, in which an insertion of six nucleotides in the heptamer/nonamer 23 spacer region may result in this segment being unable to be incorporated during V-D-J rearrangement in these species. However, M. mulatta has an additional Dδ3 segment (Dδ3n2), located approximately 1.7 kb upstream from Dδ3. Sequence identity with Dδ3 is high in the D segment region and the RSSs, suggesting this is a fully functional D segment that can be used for rearrangement. Overall, the gene segments used in rearrangement of the δ-chain appear to be remarkably conserved in these primate species.

Contrasting with the evolutionary stability of the δ-locus, investigation of the γ-locus reveals a region that is one part stable, one part highly dynamic. Shown in Fig. 1C is the organization of the Vγ genes in the five species. The Vγ9, Vγ10, and Vγ11 genes [previously designated groups II, III, and IV, respectively (21), and here designated group 2 for simplicity] are present in apes and the Old World monkey representative, M. mulatta. However, Vγ11 was not found in C. jacchus, and a stop codon early in the coding sequence of Vγ11 in M. mulatta suggests this gene is nonfunctional in this species. As noted previously for Pan troglodytes (22), the Vγ10 gene appears to be functional in all these species except for human. The genomic distance separating Vγ9, Vγ10, and Vγ11 varies little across the species, and phylogenetic analysis (discussed here later) establishes the orthology of these three genes across these primate species. Vγ9, Vγ10, and Vγ11 therefore appear to be evolutionarily conserved and stable in relation to the other Vγ genes.

Distinct differences between species at the remaining Vγ genes [previously designated group I (21) and here designated group 1] become evident even in P. troglodytes, in which the Vγ5P gene is missing and Vγ10 is functional (22). Further distinction is clear in P. pygmaeus, in which orthology between gene segments becomes unclear; this is also reflected to a greater degree in the organization of M. mulatta and C. jacchus. Although the region appears to have been evolving dynamically in each species, a gross order to genes is still maintained. Vγ1 maintains a flanking position in each of the species. Although Vγ1 is a pseudogene in humans and P. troglodytes as a result of a deletion in the RSS sequence, this mutation is not present in the remaining primate species, and therefore the Vγ1 gene in these species appears functional. Genes with gross homology to Vγ2, Vγ3, Vγ5, and Vγ6 generally occupy similar positions across the species, but it is clear that gene duplication and/or deletion—and potentially even recombination and gene conversion—frequently occurred in this region, as the orthology between genes is unclear.

The evolutionary dichotomy of this region is strongly supported by dot-plot analysis of the Vγ-locus (Fig. 2C). Comparisons of the region encoding group 2—Vγ9, Vγ10, and Vγ11—are represented by a single diagonal straight line, indicating a linear orthology between species, whereas the region encoding the group 1 Vγ genes has a hatched appearance on the plot, consistent with repeated homologous segments produced by duplication events. For comparison, a representative dot-plot analysis of the β-locus of human and P. troglodytes is shown in Fig. 2D. There are small regions that have a similar hatched pattern (the duplicated block of Vβ6, Vβ7, and Vβ5 genes and a family of trypsinogen genes, TRY and PRSS), suggesting that parts of the Vβ gene locus are evolving rapidly through a similar process to that of the Vγ locus. Overall, our genomic mapping analysis has revealed distinct differences between the genomic and evolutionary forces that have shaped the gene repertoire at the δ- and γ-loci.

Evolutionary Relationships as Revealed by Gene Phylogenies.

To understand the evolutionary relationships between the V and J segments characterized from the δ- and γ-loci, we generated neighbor-joining phylogenetic trees by using the Jukes–Cantor method and performed bootstrap analysis (500 replications) to test for confidence of groupings. Phylogenies for Vδ, Vγ, Jδ, and Jγ are shown in Fig. 3 A–D, respectively. Included in the phylogenies are the homologous mouse genes (Mumu) to identify whether gene conservation has extended to the rodent lineage.

Fig. 3.

Phylogenetic relationships of the primate V, D, and J gene segments. Shown are neighbor-joining trees of Vδ (A), Vγ (B), Jδ (C), and Jγ (D). Bootstrap confidence values are shown for most branches. Well supported groupings are indicated by colored shading. Identical sequences are enclosed in boxes. A scale for distance is shown in the lower right of each tree.

Phylogenetic analysis of the Vδ genes (Fig. 3A) reflects the conservation seen in genomic organization. There are clear groupings of the species at each of the Vδ genes (Fig. 3A, shaded in blue), and the bootstrap confidence values for these groupings are generally very high (98–100%). In addition, as noted previously (23), there are also mouse Vδ genes that clearly group within primate Vδ clades. For example, the four mouse DV6 genes [International Immunogenetics (IMGT) project nomenclature] group with 99% bootstrap confidence with the primate Vδ1, mouse DV4 groups with 100% confidence with primate Vδ2, and mouse DV5 groups with 100% confidence with primate Vδ3. Clear mouse homologues also exist for primate Vδ4 (mouse DV11), Vδ6 (mouse DV8), and Vδ8 (mouse DV2-1 and -2). Mouse DV1, DV7, and DV10 group with primate Vδ7, but the confidence for this grouping is weaker (66%). The Jδ gene segments group similarly, forming four distinct clades with high bootstrap confidence values (83–100%). Within the Jδ1 and Jδ3 clades group the TRDJ1 and TRDJ2 genes from mouse, respectively. Remarkably, the general organization of the mouse Vδ and Jδ gene fragments is similar to that of primates, suggesting these genes may be true orthologues.

Neighbor-joining trees of the primate Vγ genes show a distinctly different morphology than those of the Vδ and Jδ gene fragments but again reflect the bipolar nature of the Vγ locus (Fig. 3B). The group 2 Vγ genes—Vγ9, Vγ10, and Vγ11, which are the three anchor genes similar in location across all primate taxa—form distinct clades with long branch lengths supported by strong bootstrap confidence. The group 1 Vγ genes, however, form a bush-like structure, with short branch lengths and low bootstrap confidence values. Within the group 1 Vγ, the Vγ1, Vγ6, Vγ7, and Vγ8 form well supported subgroupings within this region. The Vγ2 and Vγ4 genes group together, as do the Vγ3 and Vγ5 genes, but no clear subgroupings within these clades are evident, suggesting these two represent sets of paralogous genes, generated through a recent block duplication of a segment containing the Vγ2 and Vγ3 ancestors, or genes that have frequently exchanged sequences through recombination or gene conversion. The hashed pattern corresponding to this region identified by dot-plot analysis supports this prediction. Rapid duplication also appears to have shaped the Jγ gene segments as well; sequence analysis and phylogeny construction reveal that Jγ1 and Jγ2, despite being separate genes, are identical at the nucleotide level, and JPγ1 and JPγ2 are highly homologous. The organization of these genes, combined with their phylogenic signature, is consistent with a block duplication of a Jγ1/JPγ1 ancestor. Overall, our analysis of the γ-locus suggests it has been evolving dynamically, with gene duplication (and presumably deletion) generating new repertoires of genes even between closely related species, such as the primates studied here. Supporting this hypothesis is the lack of clear homologues in mouse for any of the group 1 Vγ genes, or for the Jγ genes. All the mouse Vγ cluster outside of this group, with only mouse TRGV1, TRGV2, TRGV3, and TRGV4 clearing grouping with one primate gene, Vγ11, homology noted previously based on amino acid sequences (24).

Pairwise Comparison of Amino Acid Sequences of Human Vγ and Vδ Reveal a High Degree of Diversity.

Because γδ TCRs have few V genes from which to rearrange their receptors, we sought to assess the functional diversity between the protein sequences of the human V domains and to compare them with amino acid sequences of human V domains from αβ TCRs and antibody molecules. Shown in Fig. 4 are graphs of these data, binned for every five differences for simplicity. Shown to the right of the chart legend are the numbers of sequences included in this analysis; all allelic forms of these genes, obtained from the IMGT database, were included in this analysis. We find that the distribution of pairwise differences is much more extensive for the Vδ and Vγ domains than for Vα, Vβ, V_H, or Vκ, suggesting that, although the number of gene segments may be few for γδ V domains, they are functionally very divergent from each other. The bimodal distribution in particular for the Vγ domains is likely a result of comparisons of within group (i.e., group 1 vs. group 1) and between groups (i.e., group 1 vs. group 2). The Vδ domain pairwise differences demonstrate the divergence of every Vδ domain from each other. The maintenance of these genes throughout the primate lineage suggests these distinctions likely have important functional consequences.

Fig. 4.

Vδ and Vγ domains exhibit high levels of amino acid diversity. Shown are histograms of pairwise amino acid differences: Vδ (blue) and Vγ (green; Top), Vα (purple) and Vβ (yellow; Middle), and IgK (pink) and IgH (orange; Bottom). The numbers of sequences used in this analysis are shown to the right of the legend labels. All sequences available (including allelic forms) were included in this analysis.

Coding Substitutions Outweigh Silent Substitutions in Key Regions of the V Genes.

To understand the evolutionary forces that have generated diversity at the V and J genes of the γ- and δ-loci, we have calculated the number of substitutions per site that result in coding (i.e., nonsynonymous; d_n) versus silent (i.e.,synonymous; d_s) mutations. This analysis, often represented as a ratio of d_n to d_s (i.e., d_n/d_s) has been very informative when applied to rapidly evolving genes such as the classical MHC genes, demonstrating that diversifying selection (i.e., d_n/d_s ratio > 1) has been acting on regions of the MHC genes encoding the peptide binding regions (25). Genomic regions under no selective pressure are considered to be neutrally evolving (i.e., d_n/d_s ratio of 1), whereas those under purifying selection typically have a d_n/d_s ratio lower than 1. Although caveats exist for this analysis, it can be generally informative for how different regions of a gene have been evolving in relation to each other. For our analysis, we have compared across species within each Vδ, Jδ, Vγ, and Jγ gene classification. For the V gene analysis, we have further divided the V gene sequences into four categories: the regions encoding (i) the CDR1 loop, (ii) the CDR2 loop, (iii) the framework region (excluding the CDR regions), and (iv) the entire V gene sequence. Where clear orthologues exist, such as at the Vδ and Jδ gene loci, comparisons were made within these genes. For the Vγ loci in which orthology was ambiguous, such as the Vγ2/Vγ4 and Vγ3/Vγ5 gene clusters, these genes were analyzed as a group and are thus represented as Vγ2/4 and Vγ3/5. Because Caja Vγ1, Vγ3, and Vγ7 formed their own clade in our phylogenetic analysis, they have been grouped together for this analysis. Only functional genes (i.e., those capable of encoding a protein product) were included.

Results of our d_n and d_s calculations are shown in Fig. 5. The highest d_n and d_s values in the variable genes were found in the regions encoding the CDR loops, the highest being a d_s of 0.439 for the Vδ6 CDR1 region (Fig. 5A). Averages for the framework regions ranged from 0.019 (Vδ8 d_n) to 0.103 (Vδ4 d_s). The d_n/d_s ratios calculated for the framework and V domain regions were almost always significantly lower than 1, consistent with the action of purifying selection (Fig. 5B). Although the short sequence length and close evolutionary distance complicated d_n/d_s analysis of the CDR regions, several trends were noteworthy. In many cases, d_s was 0, which made calculation of an informative d_n/d_s ratio impossible; however, in a disproportionate number of these cases, d_n was greater than 0 despite the short sequence. In these cases, the ratio is shown graphically with a maximum of 4, with an additional plus sign to highlight that only coding changes exist in these particular regions. In some cases, such as the CDR2 region of Vδ1 and CDR1 region of Vδ3, the average d_n substitution per site are among some of the highest averages noted (Fig. 5A). Overall, there appears to be a bias for d_n/d_s ratios greater than 1 in the regions encoding the CDR1 and CDR2 loops for several of the V domains, suggesting that evolution has favored coding changes in these regions throughout the evolution of the primate lineage. In contrast, d_n/d_s analysis at the J gene segments revealed all ratios significantly lower than 1. The important role that the J segment performs in encoding a region of the rearranged TCR that is involved in chain heterodimerization may explain why coding changes have been selected against in these gene segments. Because of the small coding length of the D segments (Dδ1, 8 nt; Dδ2, 9 nt; and Dδ3, 14–15 nt) and their ability to be read in multiple frames, d_n/d_s analysis was not performed on these sequences. Instead, they were aligned (with the flanking RSS regions as an aid; Fig. S1), and the D segment regions were translated in all three frames, both forward and reverse. These are shown in Fig. 6; the human D sequences are shown as consensus. The three D segments exhibit different levels of variation; all functional Dδ1 sequences are identical, even at the nucleotide level (Fig. S1), whereas Dδ2 and Dδ3 differ between species to varying degrees. Identity at Dδ2 is preserved in the great apes, with differences apparent in M. mulatta and C. jacchus. However, differences at Dδ3 are evident even in P. pygmaeus (an insertion and 2-nt differences) and extend to the more distantly related species. This suggests that the D segments, despite the small genomic distance between them, are evolving at different rates. The average nucleotide pairwise differences for the functional Dδ1, Dδ2, and Dδ3 are 0.000, 0.167, and 0.148 substitutions per site, respectively.

Fig. 5.

Coding (d_n) versus silent (d_s) mutations are unequally distributed across the Vδ and Vγ genes. (A) Average d_n (green) and d_s (yellow) substitutions per site across the primate species are shown for the Vδ, Vγ, Jγ, and Jδ genes. Average substitutions per site were calculated for the regions encoding the CDR1 loop (1), CDR2 loop (2), framework region (F), and the entire coding region (V) for the variable genes and for the entire coding region of the J gene segments. Error bars represent SE of these averages. (B) Ratio of d_n to d_s substitutions for each region calculated in A. A ratio of 1, indicated as a dashed purple line in each plot, is consistent with neutral selection. Significant values greater than this are consistent with diversifying selection. Those lower than this are consistent with purifying selection. CDR1 and -2 loop regions are colored in red, framework in orange, and V domain in purple. For those values at which d_s is 0, ratios are shown extending to the maximum value of 4 with a plus sign. (Significant at *P ≤ 0.05, **P ≤ 0.005.)

Fig. 6.

Diversity of the Dδ segments across primate species. Translations of the Dδ segments are shown relative to the human consensus. Reading frames are shown at the top in the forward or reverse orientation. Dashes indicate identity with the human consensus, and differences are shown as single amino acid abbreviation. (*Stop codons; ^†Dδ segments that have insertions or deletions relative to human.) Segments deemed potentially nonfunctional are shaded lavender.

Discussion

Our analysis of the V, D, and J gene segments used in primate γδ TCR rearrangement through use of primate genome sequences provides a comprehensive perspective on the evolutionary forces that have shaped these regions. We have found that distinctly different evolutionary mechanisms have acted at the primate δ- and γ-loci, with the δ-gene segments exhibiting a grossly conserved nature across species (some extending to mouse) whereas many of the γ-gene segments have undergone substantial diversification within the primate lineage. Vγ9, Vγ10, and Vγ11 represent the exceptions to this trend, and appear to be conserved within the primate species examined here. Vγ11, although absent in the C. jacchus sequence, even appears to have orthologues in mouse (Fig. 3) (23). Dot-plot comparisons have provided broad insight into the evolution of these regions, supporting the conclusions derived from the gene maps. Repetitive elements (visualized as hashmarks) are apparent in the group 1 Vγ region, suggestive of recent, repeated duplications as the source of the gene content of this region. Patterns such as these are also apparent in regions of the Vβ locus (particularly for the Vβ6, Vβ7, and Vβ5 gene families) but do not appear to have played a substantial role in the recent evolution of the Vα or Vδ loci. The rapid evolution of the group 1 Vγ genes, which have very high amino acid identity with each other and differ mainly in the sequences of their CDR loops, suggests that they may be evolving to keep up with a selective force, perhaps particular ligands that may also be rapidly evolving (assuming these γ-chains are used in ligand recognition). This analysis is also informative for understanding the clear differences between the primate species and the various model systems that have been used to study the functions of γδ T cells. Our analysis supports the suggestion that human γδ T cells, at least those that use the group 1 Vγ domains, are not orthologous to those in mouse (23), and may explain in part the lack of clear functional γδ T-cell homologues between the two species. Furthermore, primate species used as models for human research, such as M. mulatta, have different repertoires of the group 1 Vγ genes, and therefore caution should be used in making direct comparisons of cells expressing these receptors within these species.

Our d_n and d_s analysis reveals that the average substitutions per site (both coding and silent) at many of the V gene CDR regions are higher than the average calculated in the framework regions. Furthermore, in several of the V gene CDR regions, substitutions causing amino acid mutations appear to be favored over those that are silent. Indeed, many of these regions, as a result of their short sequence and relatively short evolutionary separation between species, had a d_s of 0, making assigning significance to these d_n/d_s ratios impossible. However, this trend toward coding level diversification of these CDR loop sequences suggests that these regions may be coevolving with a similarly rapidly evolving ligand. This insight may prove useful in the assessment of candidate γδ T-cell ligands to ascertain their conservation between primate species. Our observation that the silent substitution rate was elevated in several CDR loop regions raises interesting questions regarding the mechanisms behind the evolution of these DNA stretches. Factors such as localized increased mutation rates, introduction of genetic variation through gene conversion and/or codon use bias could be at play in this region; however, testing for their effects may be difficult or impossible, so this remains speculative.

Application of our results to known human γδ TCR chain pairings such as the Vγ9Vδ2 γδ T-cell population found in the blood reveals that three of the four CDR loops have a d_n/d_s ratio greater than 1 (Fig. 5B). This population of γδ T cells is known to be stimulated in the presence of small phosphoantigens; however, it is speculated that there is an antigen-presenting molecule required on the target cell for this stimulation to occur (26), and that all CDR loops appear to be required for this recognition (27). Although certain residues within the CDR loops found to be critical for phosphoantigen-mediated stimulation were conserved between humans and macaques, our analysis suggests that positive selection has favored alterations in these CDR loops, suggesting that selection is subtly adapting the Vγ9Vδ2 population to a changing stimulus. This adaptation has not completely abrogated cross-species reactivity, as chimeric TCRs between macaque and human maintain phosphoantigen reactivity (28), but not to the degree of WT, suggesting that differences between macaque and human coding regions can modulate binding. Positive selection to keep up with evolving ligands may provide a reason for why this population has not been found outside of the primate lineage.

The Vδ1 chain is commonly found in γδ T cells resident in the intestine, paired with various Vγ chains. Some of these intestinal Vδ1 positive cells recognize the MICA molecule, a nonclassical MHC class I molecule that exhibits moderate polymorphism in the human population, both through the natural killer cell-activating receptor NKG2D and through the γδ TCR (9). In our analysis, the CDR2 of Vδ1 exhibits a high nonsynonymous substitution rate (and no synonymous substitutions) whereas the CDR1 loop appears to be evolving neutrally (Fig. 5). It is yet unclear which loops are involved in MICA recognition, and it has not been explored whether nonhuman primates also have a MICA reactive population; however, a recent structure of a human MICA-reactive Vδ1 γδ TCR suggests that the CDR3δ may not play a major role, and affinity measurements and competition assays suggest that the γδ TCR may be binding to a polymorphic region of MICA (29). Although Vδ1 chains are likely also involved in recognition of other non–MICA-related antigens in other tissues, our observation of a rapidly evolving CDR2 loop is consistent with recognition of a polymorphic site on MICA. However, further mutational analysis will be necessary to test the significance of the CDR2δ loop on MICA recognition.

The function and ligand recognition capacity of human γδ T cells still remains much of a mystery, and unfortunately there are no clear model systems in the mouse to aid our understanding. Our investigation of the evolutionary forces that shape primate γδ TCRs is focused toward gaining a better understanding of how they interact with their environment through this receptor. Our work suggests that these “innate-like” lymphocytes are not stationary in evolutionary time. Indeed, the Vγ locus has been rapidly evolving even within the primate lineage, such that several of the human Vγ genes do not have clear orthologues even in P. pygmaeus. Within many of the V genes, the regions encoding the CDR loops appear to be diversifying, implicating their involvement in recognition of ligands that may also be under pressure to change. Conversely, loops that appear conserved in other V domains may implicate these domains in recognition of invariant ligands. These cells reside in the epithelial tissues and are implicated as a first line of defense (16) against invading pathogens and cancers. Selection on these cells is therefore likely strong; therefore, the patterns of diversification, both positive and negative, that we reveal in this analysis can be used to better understand the nature of the ligands that mediate tissue-specific and stimulus-specific reactivity.

Materials and Methods

Genomic Sequence Analysis.

Genome sequences for P. troglodytes, P. pygmaeus, M. mulatta, and C. jacchus were accessed through the University of California, Santa Cruz, Genome Browser Web site (30). Human γδ TCR gene sequences obtained from IMGT (15) were used to retrieve sequence matches for P. troglodytes (Chimpanzee Sequencing and Analysis Consortium), P. pygmaeus (Genome Sequencing Center at Washington University), M. mulatta (Baylor College of Medicine Genome Sequencing Center), and C. jacchus (Genome Sequencing Center at Washington University) by using the Genome Browser's BLAST-Like Alignment Tool search algorithm. Sequences showing homology to the human Vδ (TRAV/DV), Vγ (TRGV), Jδ (TRDJ), Dδ (TRDD), and Jγ (TRGJ) genes were mapped according to their chromosomal localization within each species by using a prerelease version of the SnapGene molecular biology software (GSL Biotech), establishing their location, orientation, and genomic organization relative to each other. Duplicate/overlapping hits and sequences on irrelevant chromosomes were eliminated. All nonhuman primate genes retrieved were named according to homology with the human gene sequence with which they were identified. Where available, bacterial artificial chromosome sequences and coding sequences derived from deposited cDNA and expressed sequence tag databases were used to validate the genome sequences used in this analysis. Species names were abbreviated according to that proposed for MHC nomenclature (31) and are used this way throughout the figures: Homo sapiens, “Hosa,” P. troglodytes, “Patr,” P. pygmaeus, “Popy,” M. mulatta, “Mamu,” C. jacchus, “Caja,” and Mus musculus, “Mumu.” All gene names used are according to IMGT nomenclature (15). Dot plots were generated using the Genome Shovel server (http://www-btls.jst.go.jp/cgi-bin/Tools/dotBLAST/index.cgi).

Phylogenetic Analysis.

Sequence alignments were created using ClustalW alignment software, version 2 (32). Neighbor-joining, nucleotide phylogenetic trees were constructed using Molecular Evolutionary Genetics Analysis (MEGA) software, version 5.02 (33), with direct p-distance measurements of distance. The confidence of the phylogenetic arrangement was confirmed by using bootstrapping analysis with 500 replicates; these values are shown on the most significant branches. Gene sequences from M. musculus were obtained from IMGT for inclusion in the phylogenetic analysis.

PCR Amplification of Genes of Interest.

Locus-specific PCR primers (Integrated DNA Technologies) were designed from alignments of human γδ TCR gene fragments. Genomic DNA samples from P. troglodytes and P. pygmaeus were used as template for locus specific amplification. PCR products were analyzed via agarose gel electrophoresis, gel-purified (Qiagen), and cloned by using the TOPO TA cloning kit (Invitrogen), following 3′-adenosine addition using Taq polymerase. Clones containing the proper size insert were subcultured and DNA-purified with a commercial miniprep kit (Qiagen). Sequencing of these clones was conducted by the University of Chicago Cancer Research Center DNA Sequencing Facility.

Pairwise Difference and d_n/d_s Analysis.

Pairwise amino acid differences for each gene between species were calculated by MEGA software (33). Sequences included those retrieved from the genomic database and those isolated through molecular cloning using locus specific primers. The number of differences between each pair of sequences was normalized over the total number of sites compared.

MEGA software was additionally used to conduct d_n/d_s substitution per site analysis. Counts of substitutions per site were performed using the Nei–Gojobori method with Jukes–Cantor correction, generating d_n and d_s. The mean d_n and d_s values and the SE for each gene segment across species were calculated by using Excel. To assess the significance of the d_n/d_s ratios calculated, a two-tailed paired t test was conducted on all of the pairwise d_n and d_s values calculated within each gene, with the null hypothesis was that d_n/d_s ratio was equal to 1. All d_n/d_s ratios with a P value less than 0.05 were considered significant. In cases of a d_s of 0, statistical analysis was not performed; ratios in which the average d_s was 0 are represented at the maximum (ratio, 4) with a plus sign.