Identification of cDNA encoding Cs-KIR3DL
Sabaeus monkey cDNA sequences encoding a KIR form with three Ig-like domains and a long cytoplasmic tail that has two ITIM were identified (Fig. 3). This family of molecules is designated as Cs-KIR3DL. Four types of Cs-KIR3DL were identified in the PBMC of two unrelated sabaeus monkeys, with no type shared between the two animals (Fig. 1B). The amino acid differences between the various types are not localized to any particular region of the receptor sequences, but greater differences are seen in the Ig-like domains and the cytoplasmic region. The Cs-KIR3DL sequences possess 73–76% amino acid identity to human KIR3DL1 and KIR3DL2 and 86–91% amino acid identity to rhesus monkey KIR3DL types 13, 17. Like Mm-KIR3DL, Cs-KIR3DL sequences have a D0 domain that is three amino acids longer than those of human KIR3DL forms and also a transmembrane domain that is one amino acid shorter than those of the human KIR3DL molecules. On the basis of sequence comparison, none of the sabaeus monkey KIR3DL sequences can be designated as orthologous to any human, ape or rhesus monkey KIR3DL sequences.
Figure 3. Predicted amino acid sequences and structural domains of sabaeus monkey KIR3DL and KIR3DH forms. Tildes (∼), periods (.), inverted triangles (▾), boxed sequences, double bars, asterisk (*), vertical bars, and single horizontal bars indicate the same features as in Fig. 2. Dashes (-) indicating the absence of amino acids are used to aid in alignment. The rhesus monkey KIR3DL and KIR3DH types, Mm-KIR3DL7 and Mm-KIR3DH1 are included for comparison 13. (A) Signal sequence and Ig-like domains. Putative beta-strands and loops of the Ig-like folds are indicated above the consensus sequences of the D1 and D2 domains 22. Loop residues that have been identified to make contact with MHC class I by analysis of human KIR2DL/HLA-C co-crystal structures are underlined in the consensus sequence 20, 21. The three-amino acid deletion in the D0 domain of Cs-KIR3DH4 is highlighted in yellow. The two regions with high rates of non-synonymous substitutions are highlighted in green. (B) Stem, transmembrane, and cytoplasmic domains. A filled circle (•) indicates the cysteine in the stem domain of KIR3DH sequences. The sequence corresponding to exon 8 is missing in the Cs-KIR3DH molecules.
Download figure to PowerPoint
Identification of cDNA encoding Cs-KIR3DH
cDNA sequences were identified encoding a sabaeus monkey KIR form with three Ig-like domains and a short cytoplasmic tail (Fig. 3). These sequences share a number of features with the rhesus monkey KIR3DH form and were therefore designated Cs-KIR3DH. The transmembrane region of the Cs-KIR3DH form has an arginine and an extra amino acid in the transmembrane region not found in KIR3DL. These changes are characteristic of both the KIR2DL4 and Mm-KIR3DH forms. The stem region of the Cs-KIR3DH form contains a cysteine, which is also found in several of the Mm-KIR3DH types, but not in KIR2DL4. This cysteine is not in the same location as in the human KIR3DL2 molecule, but may allow formation of disulfide-linked dimers 18, 19. Like Mm-KIR3DH, Cs-KIR3DH sequences also lack the exon 8 region, which encodes a portion of the cytoplasmic domain. This 53-nucleotide deletion causes a frameshift and early termination of the protein. Consequently, the Cs-KIR3DH molecules are truncated and have no ITIM.
Four distinct types of Cs-KIR3DH were found, with Cs-KIR3DH1 expressed in the PBMC of both animals evaluated (Fig. 1B). The detection of three Cs-KIR3DH sequences in monkey U040 indicates that there is more than one KIR3DH locus in this animal. Cs-KIR3DH1 and Cs-KIR3DH2 have very similar Ig-like domains, but divergent transmembrane domains. Unlike the other Cs-KIR3DH and Cs-KIR3DL sequences, the D0 domain of Cs-KIR3DH4 lacks the same three amino acids as the human KIR3D sequences. The Cs-KIR3DH sequences have only 84–88% amino acid identity to rhesus monkey KIR3DH types and, based on sequence comparison, do not appear to be orthologous to these rhesus monkey KIR3DH sequences.
Variability in the Ig-like domains of Cs-KIR3D forms
The amino acid substitutions in the three Ig-like domains of Cs-KIR3DL and Cs-KIR3DH sequences suggest that the extracellular regions of these different sequences may recognize distinct MHC class I alleles. The primary sequences of the Cs-KIR3D forms can be overlaid into the crystal structure of the homologous D1 and D2 domains of the inhibitory human KIR2DL1 and KIR2DL2 molecules in complex with their HLA-C ligands 20–22. In these crystal structures, six loops linking the beta-strands in the Ig-like domains contain the residues that interact with the HLA-C ligands. The variable residues of the Cs-KIR3D sequences frequently appear in regions homologous to these loops. In particular, the sequences homologous to loops 3, 5 and 6 contain a number of non-conservative residue substitutions, at positions where KIR2DL1 and/or KIR2DL2 have been shown to contact HLA-C (underlined in Fig. 3A).
When an analysis of nucleotide substitutions was performed on the sabaeus monkey KIR3D sequences, two regions containing a concentration of codons with high rates of non-silent substitutions were identified (Fig. 4A). Sharp local increases in the number of non-synonymous changes were detected in the loop 3 region of the D1 domain and in the loop 5 region of the D2 domain. In the first region, seven of the amino acids located in positions 186–196 are under strong positive selection (Figs. 3A, 4A). Residue 186 is frequently an asparagine, forming the typical N-linked glycosylation signal NXS/T, and may be glycosylated. Fewer non-synonymous mutations were observed in the loop 3 region of rhesus monkey KIR3D sequences, suggesting that the contribution of the interaction of loop 3 with the ligand may differ in these two species. The other region under strong selective pressure is located at positions 250–252 in loop 5 of the D2 domain (Figs. 3A, 4A). By tracking substitutions through the phylogenetic trees, non-synonymous substitutions within a primate species in KIR genes could be found to be propagated either through recombination or through base substitution (data not shown, see results below).
Figure 4. Location of the polymorphic positions in the Ig-like domains of the Cs-KIR3D forms. (A) Analysis of non-synonymous and synonymous substitutions in the three Ig-like domains of the sabaeus monkey, human, and rhesus monkey KIR3D sequences, with the substitutions plotted cumulatively for each codon, moving from left to right along the protein. This plot allows one to visualize regional mutation hot spots within a set of protein sequences (very steep regions) and highly conserved regions (very flat regions). The seven amino acids shown here in capital letters using the single amino acid code, NStTsDlAGtY (positions 186–196), of the D1 domain and the amino acids QNS (positions 250–252) of the D2 domain are under strong positive selection. There is a potential for approximately 25% of all substitutions to be silent; therefore, if there were no constraints on mutations and if they occurred at random (a dN/dS ratio of 1), non-synonymous substitutions would accumulate three times faster than synonymous ones. For this reason, the slope for non-synonymous substitutions is steeper on the plots. Thirteen rhesus monkey sequences, Mm-KIR3DL 1–11, 17, and 18, and Mm-KIR3DH1–4 and the three human reference sequences for KIR3DL1 (NM_013289.1), KIR3DL2 (NM_006737.2), and KIR3DS1 (NM_014514.1) were used in this analysis 13, 17. (B) Ribbon diagram of the D1 and D2 domains of human KIR2DL1 indicating (in purple) the amino acid positions of Cs-KIR3DL3 that vary from the consensus sequence of Fig. 3A. The side chains shown are those of the human KIR2DL1 sequence 22. (C) Ribbon diagram indicating the amino acid positions in the D1 and D2 domains of Cs-KIR3DH3 that vary from the consensus sequence of Fig. 3A.
Download figure to PowerPoint
Comparisons of representative KIR3D sequences from humans, common chimpanzees, pygmy chimpanzees, orangutans, rhesus monkeys and sabaeus monkeys revealed that the longest stretches of amino acids that are perfectly preserved across all KIR3D sequences are two regions of the D2 domain (data not shown) 2, 6, 10, 11, 13, 17. These two regions of five amino acids, positions 245–249 (LSCSS) and positions 294–296 (YRCFG), are found in the B and F beta-strands of the D2 domain, respectively. These regions include the cysteine residues that form the intradomain disulfide bridges and therefore may be structurally constrained 22.
The allospecificity of human KIR2DL is determined by dimorphisms of residue 44 in loop 2 of the D1 domain of these molecules, which directly interacts with residue 80 of HLA-C 23, 24. The conservation at the analogous position to residue 44 (position 163) in the D1 domain of all four Cs-KIR3DL sequences and three of the Cs-KIR3DH sequences indicates that allospecific recognition may not occur at this location in Cs-KIR3D forms. However, the Cs-KIR3D sequences do show variation in additional loops predicted from the human KIR2DL crystal structures that are not involved in the interaction with the HLA-C ligands. Amino acids of Cs-KIR3DL3 that vary from the consensus sequence shown in Fig. 3 include three amino acids in the loop located in the D2 domain between the beta-strands C’ and E. These amino acid changes are indicated on the crystal structure of human KIR2DL1 (Fig. 4B). In addition to amino acid differences in the putative ligand contact loops (loops 2, 4–6), Cs-KIR3DH3 varies from the consensus in loops formed between the B and C beta-strands and the F and G beta-strands of the D1 domain, and the C’-E loop of the D2 domain (Fig. 4C). These additional loops may also contribute to ligand interactions of Cs-KIR3D molecules. The C’-E loop is particularly intriguing because variability of this loop is also seen in human KIR sequences 25. Moreover, a significant rotation of this loop was found in the crystal structure of the activating receptor KIR2DS2, despite sequence conservation 26.
Additionally, a substantial number of residue substitutions were observed in the D0 domains of the Cs-KIR3D sequences, which may contribute to ligand recognition. In fact, studies of deletion mutants of human KIR3DL1 have demonstrated that all three intact Ig-like domains are required for binding to its HLA-B ligand 27. Of the Cs-KIR3D forms, the D0 domain of Cs-KIR3DH4 is unique, and, like the human KIR3D forms, it has three fewer amino acids in the region predicted to form the C beta-strand. This deletion of three amino acids results in the loss of two predicted N-linked glycosylation sites in this domain (Fig. 3A). A two-amino acid deletion in the predicted C' beta-strand was previously shown to enhance interaction of Pan troglodytes (Pt)-KIR3DL1/2 with HLA-B ligands 28. The insertion/deletion of the three amino acids into/from the D0 domain and the alteration in glycosylation of Cs-KIR3D molecules may also affect ligand interaction.
Identification of cDNA encoding a Cs-KIR2DL5 splice variant and the novel form Cs-KIR3DH/vL5
cDNA sequences were identified that encode two unique KIR forms, Cs-KIR2DL5sv1 and Cs-KIR3DH/vL5 (Fig. 5). These forms possess a D0 domain that is strikingly divergent from Cs-KIR3DL, Cs-KIR2DL4, and Cs-KIR3DH, but which has 98–100% identity to the D0 domain of rhesus monkey KIR2DL5 (Mm-KIR2DL5) types. Like Cs-KIR2DL4, this D0 domain is three amino acids shorter than that of Cs-KIR3DL and most Cs-KIR3DH types. These forms also possess an extra amino acid in the transmembrane domain that is found in Cs-KIR2DL4 and Cs-KIR3DH, but not in Cs-KIR3DL.
Figure 5. Predicted amino acid sequences and structural domains of sabaeus monkey KIR2DL5sv1 and KIR3DH/svL5. Tildes (∼), periods (.), inverted triangles (▾), dashes (-), boxed sequences, double bars, putative beta-strands, loops of the Ig-like folds, MHC class I contact residues from human KIR2DL/HLA-C co-crystal structures, filled circle (•), asterisk (*), vertical bars, and single horizontal bars are indicated as in Fig. 3. The rhesus monkey sequence Mm-KIR2DL5.2 and sabaeus monkey sequences Cs-KIR3DH3 and Cs-KIR3DL2 are included for comparison 13.
Download figure to PowerPoint
The first form is designated Cs-KIR2DL5sv1 because of the similarity of its sequence to that of Mm-KIR2DL5 and its apparent lack of both D1 and D2 domains. It may represent a splice variant of an as yet unidentified full-length Cs-KIR2DL5 sequence encoding a D0–D2 KIR form. In fact, a human splice variant of KIR2DL5 without a D2 domain has been previously reported 29. The transmembrane domain of Cs-KIR2DL5sv1 is identical to Mm-KIR2DL5.2, one of the two Mm-KIR2DL5 types identified in rhesus monkeys. However, Cs-KIR2DL5sv1 has a cytoplasmic region with greater identity to the sabaeus monkey KIR3DL sequences than Mm-KIR2DL5 sequences. It appears that a recombination event occurred at the second cytoplasmic exon (exon 9) with a Cs-KIR3DL sequence. The cytoplasmic tail of the Cs-KIR2DL5sv1 molecule is of the same length as that of Cs-KIR3DL and is 31 amino acids shorter than that of Mm-KIR2DL5.
The second novel KIR form is designated Cs-KIR3DH/vL5 because of the similarity of its sequence to that of Mm-KIR2DL5 and Cs-KIR2DL5sv1 in the D0 region, and because of its homology to Cs-KIR3DH in the D2, stem, transmembrane and cytoplasmic regions. Numerous amino acid changes not shared by any other sabaeus monkey KIR molecule are found in the D1 domain of Cs-KIR3DH/vL5, most notably in the loop 2 and loop 3 regions. In particular, residue 160, which is analogous to the loop 2 residue 44 of the human KIR2DL forms, is the nonpolar amino acid isoleucine, not a polar amino acid like that found in all Cs-KIR3DL and Cs-KIR3DH molecules. Like in Cs-KIR3DH sequences, there is a cysteine in the stem region and an arginine in the transmembrane region of Cs-KIR3DH/vL5. Also as seen in Cs-KIR3DH sequences, Cs-KIR3DH/vL5 lacks the sequence corresponding to exon 8 and, accordingly, should terminate early without a substantial cytoplasmic domain.
Phylogenetic analysis of KIR3D sequences
To examine the evolutionary history of the sabaeus monkey KIR3D molecules, Cs-KIR3DL and Cs-KIR3DH sequences were compared to KIR3D sequences previously obtained from humans and four other primate species: common chimpanzees, pygmy chimpanzees, orangutans, and rhesus monkeys 2, 6, 10, 11, 13, 17. Initially, extensive analysis using Simplot and bootscanning was performed 30. Multiple crossover points were apparent between the different exons of the Ig-like domains, suggesting recombination was occurring between these exons (Fig. 6). Because the genomic sequence of the KIR3D genes is not available for any of the nonhuman primate species, the precise boundary of the recombination breakpoints within the introns was not determined. While phylogenetic analyses of primate KIR3D forms have previously been performed, they have generally been done using the full coding sequences, or with all three of the Ig-like domains analyzed together 10, 11, 31. In order to examine more precisely the evolutionary relationships of the primate KIR3D sequences, phylogenetic trees were constructed using the maximum-likelihood method with bootstrapping, treating each of the three Ig-like domains as a distinct entity (Fig. 7). Branching patterns consistent with those shown in Fig. 7 were also observed using neighbor-joining trees (data not shown).
Figure 6. Bootscan analysis provides evidence of recombination events. (A) Bootscan analysis of the Mm-KIR3DL5 sequence as the query against Mm-KIR3DL2, Mm-KIR3DL18, and Pt-KIR3DL6 sequences was performed with a sliding-window size of 200 nucleotides, a step size of 20 base pairs, and 100 bootstrap replicates. (B) Bootscan analysis using the Cs-KIR3DH3 sequence as the query against Cs-KIR3DL3, Mm-KIR3DL7, and Mm-KIR3DL2 sequences.
Download figure to PowerPoint
Figure 7. Phylogenetic relationships between the three Ig-like domains of the KIR3D forms. Maximum-likelihood trees were constructed for each of the KIR3D Ig-like domains. Bootstrap values were determined for 100 replications of the maximum-likelihood tree, and values >50% are shown. The scale bar represents 10 mutations per 100 nucleotide sequence positions. The KIR3D sequences were examined from the following species: human (no prefix), common chimpanzee (Pt-), pygmy chimpanzee (Pp-), orangutan (Popy-), rhesus monkey (Mm-), and sabaeus monkey (Cs-) 6, 10, 11, 13, 17. The human KIR reference sequences for KIR3DL1 (NM_013289.1), KIR3DL2 (NM_006737.2), and KIR3DS1 (NM_014514.1) were used. The outgroup was a sabaeus and rhesus monkey KIR3D consensus sequence. Examples of rearranging associations of Ig-like domains between KIR3D sequences from the rhesus or sabaeus monkey sequences are highlighted in color to assist in tracking. Trees are drawn vertically to better illustrate potential recombination events.
Download figure to PowerPoint
In each of the trees, the human, common chimpanzee, pygmy chimpanzee, and orangutan KIR3D sequences form one distinct lineage, and the rhesus and sabaeus monkey sequences form another. Within these lineages, there was not a clear intraspecies association of sequences, nor was there an association of KIR3D forms based on potential receptor signaling. Most strikingly, many of the KIR3D sequences form relationships with different sequences in each tree, indicating that each Ig-like domain has a different phylogenetic history. Like the simplots and bootscans, the shifting of associations between KIR3D sequences determined for Ig-like domains suggests that recombination between Ig-like domains occurred during the evolution of these KIR sequences. For many of the sabaeus and rhesus monkey KIR3D sequences, only low bootstrap support was obtained for the phylogenetic relationships. In contrast, the relationship between Cs-KIR3DH1 and Cs-KIR3DH2 shows at least 90% support in all three Ig-like domains, consistent with their high amino acid identity.
We were able to detect examples of rearrangement between Ig-like domains of KIR3D sequences from an individual primate species for both the rhesus and sabaeus monkey. The rhesus monkey KIR3DL5 (Mm-KIR3DL5) sequence clusters with Mm-KIR3DL18 in the D0 domain with a bootstrap support of 82%. Yet, in the D1 and D2 Ig-like domains, the Mm-KIR3DL5 sequence forms a cluster with the Mm-KIR3DL1, 2, 3, and 4 sequences. These associations are supported by bootstrap values of 94% and 100% in the D1 and D2 domains, respectively. These alterations in relationship were confirmed using bootscanning, and the crossover point was detected in the region of the exon boundary between the D1 and D2 domains (Fig. 6A). Importantly, the Mm-KIR3DL5 sequence was identified from a different rhesus monkey than Mm-KIR3DL1–4 and Mm-KIR3DL18 sequences, indicating that these findings do not represent a PCR-generated artifact. Although the relationships do not have high bootstrap support, a similar example is seen with the sabaeus monkey Cs-KIR3DL2 and Cs-KIR3DL3 sequences. These sabaeus KIR3DL sequences associate in both the D0 and D2 domains, but not the D1 domain, indicating that the D1 domains of these sequences are more divergent (Fig. 7).
In an example of rearrangement of individual Ig-like domains that illustrates recombination between KIRDH and KIRDL genes, the Mm-KIR3DLl0 sequence associates with Mm-KIR3DL9 in the D2 domain with 100% bootstrap support, and a trend for a relationship between these molecules is shown in the D1 domain tree. However, in the D0 Ig-like domain, Mm-KIR3DL10 is most closely related to a cluster of rhesus monkey KIR sequences that includes Mm-KIR3DL (5, 7, and 18) and Mm-KIR3DH (1 and 2). This D0 cluster containing both KIR3DH and KIR3DL sequences has a bootstrap value of 77%. The differing relationships of these KIR3DL and KIR3DH sequences indicates that shuffling of Ig-like domains has occurred during the evolution of these two forms of immune receptors. These recombinations between Ig-like domains occurred in addition to the distinct recombination events at the transmembrane exon that most likely generated the KIR3DH form.
An additional example of changing associations between KIR3DL and KIR3DH forms involves sequences found in the two different monkey species (Figs. 6B, 7). The sabaeus monkey KIR3DH3 sequence (Cs-KIR3DH3) is associated with Cs-KIR3DL3 in the D0 domain, but forms a cluster with the rhesus monkey sequences Mm-KIR3DL7 and Mm-KIR3DL18 in the D1 domain. Although these associations did not have high bootstrap values using maximum-likelihood analysis, there is 88% bootstrap support for the clustering of these sabaeus and rhesus monkey sequences in the D1 domain using the neighbor-joining method (data not shown). The change in association of Cs-KIR3DH3 with Cs-KIR3DL3 in the D0 domain to Mm-KIR3DL7 in the D1 domain was confirmed by bootscanning (Fig. 6B). Interspecies phylogenetic clustering suggests that an Ig-like domain that existed prior to speciation gave rise to the Ig-like domains currently identified in these two primate lineages. While the Ig-like domain was then maintained in both species, further recombination events occurred to alter the relationship with the other two Ig-like domains, as well as the type of KIR3D form in which this Ig-like domain is found.