Polymorphic HLA-C Receptors Balance the Functional Characteristics of KIR Haplotypes

This article is featured in In This Issue, p.2921

Natural killer cells are cytotoxic lymphocytes that kill virus infected (1) and malignantly transformed cells without prior sensitization (2). They also play an important role in early pregnancy where they control the trophoblast-mediated remodeling of maternal blood vessels during formation of the placenta (3). Controlling the development and function of NK cells are a wide array of activating and inhibitory cell surface receptors (4). A common feature of these germline-encoded receptors is their specificity for MHC class I ligands (in humans the HLA class I molecules). In catarrhine primates (humans, apes, and Old World monkeys) the ubiquitously expressed classical MHC class I molecules are recognized by the killer cell Ig-like receptor (KIR) family (5). Because both KIR and MHC class I are highly polymorphic, these interactions diversify and individualize NK cell responses. The functional importance of such interactions is illustrated by the association of various combinations of KIR and HLA class I with the outcome of viral infection (6–8), susceptibility to autoimmune disease (9–11), survival following bone marrow transplantation (12, 13), and reproductive success (14, 15).

The catarrhine primates share four phylogenetic lineages of KIR that have different structure and specificity for MHC class I molecules (3, 5). Lineage I KIRs recognize HLA-G, which has restricted expression on trophoblast cells and monocytes and is considered to be dedicated to reproductive function (16–18). Lineage II KIRs recognize the A3/A11 and Bw4 epitopes that are each carried by a different subset of HLA-A and HLA-B alleles. Lineage III KIRs, the subject of this study, recognize HLA-C ligands. A ligand for the lineage V KIRs, represented by KIR3DL3 in humans, has yet to be identified. In contrast to HLA-A and HLA-B, every HLA-C allotype forms a ligand for KIRs, and it is these interactions that dominate NK cell responses. Two mutually exclusive HLA-C epitopes are defined by the residue at position 80 in the α₁ domain and are recognized by different KIRs (19). Inhibitory KIR2DL1 and activating KIR2DS1 recognize the C2 epitope (lysine 80) whereas inhibitory KIR2DL3 recognizes the C1 epitope (asparagine 80) and inhibitory KIR2DL2 principally recognizes C1 but also has some cross-reactivity with C2 (20, 21).

The KIR gene family is diversified by both gene content variability and allelic polymorphism and is part of the leukocyte receptor complex on human chromosome 19 (22, 23). The basic organization of the KIR locus is conserved among the catarrhine primates, consisting of four framework genes (common to most haplotypes) and a suite of homologous genes that are variably present between species and individuals (24). A site of recombination at the center of the locus defines two distinct regions: one closer to the chromosomal centromere (termed centromeric), the other closer to the chromosomal telomere (termed telomeric) (25). Since the time of their common ancestor, there has been species-specific reorganization of the centromeric and telomeric regions of the catarrhine primate KIR locus (5). Chimpanzees, orangutans, and gorillas expanded variable KIR genes in the centromeric region of the KIR locus but not in the telomeric region (26). In contrast, macaques and other Old World monkeys expanded variable KIR genes in the telomeric region but not in the centromeric region (27). The human species is unique in having expanded variable KIR genes in both the centromeric and telomeric regions (26). Each of these regions contains different genes and different alleles of shared genes, and it is these differences that are the basis for the division of human KIR haplotypes into two functional groups, A and B (28). KIR A haplotypes encode a fixed suite of largely inhibitory receptors whereas KIR B haplotypes have a variable number of inhibitory receptors and several activating receptors. That these haplotypes are found in all human populations but with different relative frequencies suggests they have different functional properties that are subject to balancing selection (29).

In addition to their accumulation of activating receptors, the KIR B haplotypes have accumulated a particular subset of alleles for the inhibitory receptor genes that are common to KIR A and B haplotypes (28, 30). Furthermore, although a limited number of KIR haplotypes have been studied in detail, there appears to be functional differences between the inhibitory receptors encoded by the KIR A and B haplotypes (21, 31–33). In this study we determined the haplotypic association, phylogeny, and function of every KIR2DL1, KIR2DL2/3, and KIR2DS1 allele defined. In so doing we have shed light on the evolution and function of human HLA-C receptors and their contribution to the distinct functions of the KIR A and B haplotypes.

Sequences encoding the D1, D2, stem, transmembrane, and cytoplasmic tail domains (amino acids 1–328) of 26 KIR2DL1, 7 KIR2DS1, and 36 KIR2DL2/3 alleles were aligned and analyzed by three methods: neighbor joining (using the Tamura–Nei pairwise substitution) (34), maximum likelihood, and parsimony, each with 500 replicates. The bootstrap support for each node is indicated when >50. Representative trees are shown in Figs. 1 and 3 and Supplemental Fig. 1. Evolutionary analyses were conducted in MEGA6 (35).

KIR-Fc fusion proteins were generated from insect cells (provided by Prof. K.C. Garcia, Stanford University) and infected with baculovirus as described (36). The KIR-Fc fusion protein corresponding to each KIR2DL1, KIR2DL2, and KIR2DL3 allotype was tested for binding to a panel of microbeads, in which each bead is coated with 1 of 97 HLA class I allotypes (31 HLA-A, 50 HLA-B, and 16 HLA-C allotypes) (LABScreen single-antigen beads, One Lambda, lot no. 8). To account for differences in the amount of HLA class I protein coating each bead, the binding of each KIR-Fc fusion protein is normalized to the binding of W6/32, a mAb detecting a common epitope of HLA class I. Binding values were calculated using the formula: (specific binding − background bead fluorescence)/(W6/32 binding − background bead fluorescence).

We examined the cell surface expression of four natural KIR2DL1 allotypes and nine KIR2DL1*003 mutants when transfected transiently into HeLa cells. N-terminal 3× FLAG-tags were attached to each KIR2DL1 construct, so that the expression of each KIR2DL1 variant could be measured using the same anti-FLAG Ab binding to the identical FLAG epitope. Recombinant cDNA encoding the extracellular, stem, transmembrane, and cytoplasmic domains (amino acids 1–328) of KIR2DL1*003 with an N-terminal 3× FLAG-tag (amino acid sequence, DYKDHDGDYKDHDIDYKDDDDK) was manufactured by GenScript (Piscataway, NJ) and subcloned into the pcDNA3.1⁺ expression vector. Site-directed mutagenesis was performed with the QuikChange kit (Stratagene), according to the manufacturer’s instructions, to generate three other natural KIR2DL1 alleles and nine mutants containing termination codons at different positions in the sequence encoding the stem, transmembrane, and cytoplasmic tail domains. HeLa cells were cultured in DMEM (Life Technologies) supplemented with 10% FBS, 100 μg/ml streptomycin, 100 U/ml penicillin, and 2 mM l-glutamine (DMEMc). Cells were plated in 15.6-mm wells at 5 × 10⁴ cells/well in 500 μl DMEMc for 24 h and then transfected with a pcDNA3.1⁺ vector encoding FLAG-tagged KIR2DL1 allotypes using the FuGENE transfection reagent (Promega). Thirty-six hours after transfection, adherent cells were dissociated from the wells using 200 μl 0.05% trypsin-EDTA solution and washed with flow cytometry buffer (Dulbecco’s PBS [DPBS] containing 2% EDTA and 1% BSA at 4°C). Cells were then stained with 25 μl FITC-conjugated mouse polyclonal anti-FLAG Ab (Sigma-Aldrich) at a final concentration of 3 μg/ml. Following a further wash, cells were suspended in 50 μl 50× propidium iodide (BD Biosciences) and fixed in 50 μl 2% paraformaldehyde. Cells expressing FLAG-tagged KIR2DL1 were detected by flow cytometry (Accuri C6 cytometer, BD Biosciences). Expression levels of each allotype or mutant were determined from the median fluorescence intensity of FITC-conjugated anti-FLAG Ab bound to each positive staining cell. Three independent transfections with at least 50,000 cells each were performed for each allotype tested.

HeLa cells were plated at 5 × 10⁵ cells/well in 500 μl DMEMc on a 12-mm round no. 1 German glass poly-d-lysine–coated glass coverslip (BD Biosciences) and transfected with FLAG-tagged KIR2DL1 allotypes. Forty-eight hours after transfection the cells were fixed in 4% paraformaldehyde (Electron Microscopy Sciences) diluted in DPBS with magnesium and calcium (Life Technologies). DPBS (275 μl) containing glycine (25 mM) was added to quench reactive aldehydes. Cells were permeabilized with DPBS containing 0.04% saponin and blocked in blocking buffer (BB; DPBS containing 2% heat-inactivated goat serum, 1% BSA, 0.1% cold fish skin gelatin, 0.02% SDS, 0.1% Nonidet P-40, and 0.05% sodium azide [pH 7.2]). Rabbit polyclonal IgG anti-FLAG primary Ab (Sigma-Aldrich) was applied at 5 μg/ml in BB and incubated overnight at 4°C. After washing with BB, cells were incubated for 20 min at room temperature with 1 U Alexa Fluor 555 phalloidin solution diluted in BB. Cells were then washed with BB and incubated for 1 h at room temperature with Alexa Fluor 488 goat anti-rabbit IgG (Life Technologies) secondary Ab suspended in BB. Cells were washed in BB and DPBS and coverslips were mounted for microscopy in ProLong Gold antifade reagent (Life Technologies). Cells were analyzed by confocal laser-scanning microscopy using an upright system (DM6000, SP5; Leica) with an oil immersion objective (×63, 1.3 numerical aperture; HCX Plan-Apochromat; Leica) and argon (488-nm) and HeNe (543- and 633-nm) lasers. Images were acquired using LAS AF SP5 software (Leica) in sequential scan mode with a 400-Hz scan rate, line averages of two, and a 512 × 512-pixel resolution. Z-stacks were collected at 0.3-μm intervals. Raw images were processed using Volocity (PerkinElmer). To improve image quality, raw images were processed using a fine filter before any analysis was performed. The same settings were maintained for all samples within an experiment.

Quantitative colocalization analysis in three dimensions was performed between the phalloidin and anti-FLAG channels. Colocalized voxels were identified by performing colocalization analysis in Volocity (PerkinElmer) and generating channels that display the product of the differences of the mean, which were generated by calculating the product of the difference from the mean for each voxel intensity from the two channels analyzed. This gave a clear visual display of areas of positive (and negative) correlation. The total volume of colocalized voxels per cell was calculated in Volocity using eight cells for each of the six transfections performed.

KIRs that recognize HLA-C divide into two groups: the inhibitory KIR2DL2/3 that recognize the C1 epitope of HLA-C, and the inhibitory KIR2DL1 and activating KIR2DS1 that recognize the C2 epitope of HLA-C. KIR2DL1 and KIR2DL2/3 are highly polymorphic, whereas KIR2DS1 is relatively conserved (37) (http://www.ebi.ac.uk/ipd/kir/).

Phylogenetic analysis of the coding sequence of 26 KIR2DL1 and 7 KIR2DS1 alleles distinguishes four clades of KIR (Fig. 1A, Supplemental Fig. 1). The amino acid substitutions that distinguish representative members of the four clades are shown in Fig. 1B. Clade 1 comprises 10 KIR2DL1*001-like alleles (Fig. 1A), 7 of which have been mapped to KIR haplotypes (29, 38–44). Five of the seven are present in Cen A, the centromeric region of KIR A haplotypes (Fig. 2). In contrast, KIR2DL1*022 and KIR2DL1*026 are present in Cen B, the centromeric region of KIR B haplotypes (44). Clade 2 comprises 10 KIR2DL1*003-like alleles, of which 5 have been mapped to KIR haplotypes (Fig. 1A). These alleles segregate with Cen A (Fig. 2). Clade 3 comprises six KIR2DL1*004-like alleles, whose products are distinguished from those of clades 1 and 2 by a 4-aa sequence motif comprising threonine 154, asparagine 163, and arginine 182 of the D2 domain and glutamate 216 of the stem domain (Fig. 1B, Supplemental Fig. 2). This motif is shared with KIR2DS1, suggesting that KIR2DL1 acquired this motif from KIR2DS1 by recombination (Fig. 1C) (26). Clade 3 comprises six KIR2DL1 alleles, of which three have been mapped to Cen B and one (KIR2DL1*011) to Cen A (Fig. 2). Clade 4 comprises the seven KIR2DS1 alleles. KIR2DS1 is distinguished by having an additional residue in the transmembrane region, threonine 237, as well as lysine 233 that forms a noncovalent association with DAP12, a disulfide-bonded homodimer that contains an ITAM (45) (Fig. 1B). KIR2DS1 alleles are all found in the telomeric region of KIR B haplotypes (Tel B). Four of these alleles have been identified in the course of high-resolution KIR haplotype studies (Fig. 2).

The KIR2DL2/3 gene encodes receptors that recognize HLA-C allotypes carrying the C1 epitope. The alleles of KIR2DL2/3 form two distinctive groups named KIR2DL2 and KIR2DL3. Phylogenetic analysis of 36 KIR2DL2/3 sequences identified four clades of alleles (Fig. 3A, Supplemental Figs. 1, 2B). The amino acid substitutions that distinguish representative members of the four clades are shown in Fig. 3B. Clades 1 and 2 comprise the 24 KIR2DL3 alleles; clades 3 and 4 comprise the 11 KIR2DL2 alleles (Fig. 3C). Clade 1 consists of 12 KIR2DL3*001-like alleles, and clade 2 consists of 12 KIR2DL3*004-like alleles. Of the 11 KIR2DL3 alleles mapped to KIR haplotypes, 8 are in Cen A and 2 (2DL3*014 and 2DL3*018) in Cen B (Fig. 2). Although usually present in Cen A, KIR2DL3*001 has been found in Cen B, but rarely (41).

The KIR2DL2 proteins encoded by clades 3 and 4 are distinguished from KIR2DL3 mainly by differences in the transmembrane region and the cytoplasmic tail (Fig. 3B). In this carboxyl-terminal part of the molecule, KIR2DL2 is more similar to KIR2DL1 than to KIR2DL3, indicating that KIR2DL2 acquired this sequence from KIR2DL1 by recombination (Fig. 3C) (21). Clade 3 comprises 10 of the 12 KIR2DL2 alleles. Five clade 3 KIR2DL2 alleles have been mapped to KIR haplotypes and shown to be present on Cen B (Fig. 2). Additionally, rare examples of 2DL2*003 and 2DL2*006 on Cen A have been reported (Fig. 2) (40, 41). Clade 4 comprises the KIR2DL2*004 and KIR2DL2*011 alleles. Their protein products are distinguished from other KIR2DL2/3 at positions 41, 167, 269, and 317 (Fig. 3B, Supplemental Fig. 2B). The clade 4 alleles are both present on Cen B (Fig. 2).

In summary, these analyses (Figs. 1–3) show that for both KIR2DL1 and KIR2DS1 and KIR2DL2/3 there are four phylogenetic clades of alleles that correlate with genomic location in the Cen A, Cen B, or Tel B regions of KIR haplotypes.

We investigated whether the receptors encoded by Cen A and Cen B alleles differ in their capacity to bind to a panel of nine C1 and seven C2 allotypes. All 10 KIR2DL1 allotypes encoded by Cen A (Fig. 2) have strong C2 specificity (Supplemental Fig. 3A). Of the seven KIR2DL1 encoded by Cen B (Fig. 2), six are specific for C2 (Supplemental Fig. 3A). The exception, KIR2DL1*022, has a methionine to lysine substitution at position 44 that gives it C1 specificity. From a synthesis of these binding data, we demonstrate that KIR2DL1 encoded by Cen A alleles binds C2 with greater avidity than does Cen B–encoded KIR2DL1, and the difference is statistically significant (two-tailed t test, p < 0.01) (Fig. 4, left panel). Receptor function of three Cen B–associated KIR2DL1 allotypes is further weakened by substitutions in the transmembrane or cytoplasmic domain. The presence of cysteine 245 in KIR2DL1*004 and *007 (Supplemental Fig. 2A) reduces their signaling capacity (31), and that of KIR2DL1*026 is abrogated as a result of premature termination at codon 246 that eliminates the cytoplasmic tail (44, 46). Consistent with a persistent selection pressure to reduce the functionality of Cen B–encoded KIR2DL1, some allotypes are weakened by substitutions in both the ligand binding and signaling domains (Table I).

We performed similar analyses of the KIR2DL2/3 allotypes encoded by Cen A and Cen B. The seven KIR2DL2 alleles generally associate with Cen B, although rare examples of 2DL2*003 and 2DL2*006 on Cen A have been reported (Fig. 2) (40, 41). Of these, KIR2DL2*004 is inactivated (47) and does not encode a functional C1 receptor (44). The eight KIR2DL3 alleles with defined haplotype association are all on Cen A. Among these is KIR2DL3*008N that does not encode a functional receptor (29). KIR2DL3*014 and 2DL3*018 are associated with Cen B and display strong binding to C1 and cross-reactivity with C2 (Supplemental Fig. 3B). These binding properties are more like KIR2DL2 allotypes than the Cen A–encoded KIR2DL3 allotypes. The Cen A–encoded KIR2DL2 and KIR2DL3 receptors have lower avidity for C1 than do the Cen B–encoded KIR2DL2 and KIR2DL3 receptors, and this is statistically significant (two-tailed t test, p < 0.0001) (Fig. 4, right panel). Thus, Cen B–associated KIR2DL2/3 have higher avidity for C2 than do Cen A–associated KIR2DL2/3 (two-tailed t test, p < 0.0001) (Fig. 4, center panel). This pattern is the opposite of that seen for KIR2DL1, where the Cen A–encoded allotypes have higher avidity for C2 than do those encoded by Cen B.

To determine the basis for the weaker C2 avidity of Cen B–encoded KIR2DL1, we compared KIR2DL1*003 and KIR2DL1*004, respectively, the most common Cen A– and Cen B–associated alleles (29, 38–44). In the D1 and D2 domains that form the ligand-binding site, KIR2DL1*003 and KIR2DL1*004 differ only in D2, at positions 114, 154, 163, and 182 (Fig. 5A). We made KIR-Fc fusion proteins from KIR2DL1*003, KIR2DL1*004, and the set of 14 KIR2DL1 mutants that represents all possible combinations of the four dimorphic positions. These 16 KIR-Fc fusion proteins were tested for binding to the panel of 97 HLA class I allotypes.

The 16 KIR2DL1-Fc bound only to the seven C2 allotypes, with the mean values being shown in Fig. 5B. Substitution at all four positions affects receptor avidity for C2. Single mutation at three of the positions (154, 163, and 182) showed that the KIR2DL1*003 residue increases and the KIR2DL1*004 residue decreases avidity. For position 114 the opposite effect was seen; that is, the KIR2DL1*003 leucine decreases avidity and the KIR2DL1*004 proline increases avidity. Receptors with higher C2 avidity than any natural KIR2DL1 allotype were made by introducing proline 114 into KIR2DL1*003 and histidine 182 into KIR2DL1*004, whereas the weakest of all the mutants tested is the KIR2DL1*004 mutant with leucine 114. In contrast to the striking differences seen among the single mutations, the mutants that differ by two substitutions from KIR2DL1*003 and KIR2DL1*004 are less varied and are all weaker C2 receptors than KIR2DL1*003. Notably, the combination of histidine 182 and proline 114, which individually give the strongest receptors, is a weaker receptor than KIR2DL1*003.

Each of the residues that contributes to the difference between KIR2DL1*003 and KIR2DL1*004 is present in 8 of the 16 KIR-Fc tested. By averaging the binding of the eight receptors containing a particular residue we see that the mean values are very similar, except for that between the groups of receptors having leucine or proline at position 114 (Fig. 5C). These results are consistent with all four positions contributing to the binding and being affected by the particular residues present at the other positions. Thus, the four dimorphic sites have not coevolved to produce the strongest and weakest C2 receptors possible. Instead, there exists a more moderate balance between a stronger and a weaker C2 receptor. The dimorphism of arginine (KIR2DL1*003) and cysteine (KIR2DL1*004) at position 245 at the end of the transmembrane domain (Fig. 1B) also adds to the functional difference between KIR2DL1*003 and KIR2DL1*004. The cysteine 245 reduces the capacity of KIR2DL1*004 to develop inhibitory signals (31) and educate NK cells (33).

KIR2DL1*004 is one example of a Cen B–associated allotype that is functionally affected by polymorphism at the junction between the transmembrane and cytoplasmic domains. Another is KIR2DL1*026 that has a termination codon at position 246 (Supplemental Fig. 2A). This substitution has two effects. First, it eliminates the cytoplasmic tail with its ITIMs that mediate inhibitory signaling function (46). Second, it reduces the cell surface expression of KIR2DL1*026, which is 32% that of KIR2DL1*012, the progenitor of KIR2DL1*026 (44). This observation raised the question: How much of the transmembrane domain is necessary for KIR2DL1 to become cell surface associated?

To address this question we introduced termination codons at five evenly spaced positions within the sequence encoding the transmembrane domain (residues 225–245) of KIR2DL1*003, as well as at position 224 in the stem domain and positions 246, 250, and 256 in the cytoplasmic tail (Fig. 6A). After incorporation of N-terminal 3× FLAG-tags, the 12 mutant constructs and wild-type KIR2DL1*003 were transiently transfected into HeLa cells. After 48 h of culture the amounts of KIR2DL1 on the surface of the transfected HeLa cells were measured using anti-FLAG Ab and flow cytometry.

Termination in the N-terminal half of the transmembrane region (residues 225–235) completely abrogated cell surface expression of the mutant KIR2DL1*003. In contrast, termination in the carboxyl-terminal half of the transmembrane (residues 236–245) permitted cell surface expression of mutant KIR2DL1*003, with levels corresponding to 53–72% of the wild-type (Fig. 6A). These results show that the N-terminal half of the transmembrane domain is required for membrane association and cell surface expression of KIR2DL1. Mutants that terminate at positions 238, 242, and 246, and that lack a cytoplasmic tail, were expressed at slightly higher levels than mutants 250 and 256 that have a short, truncated cytoplasmic tail (p = 0.0317). Terminating at position 246, KIR2DL1*026 is in the former category.

KIR2DL1*014 differs from KIR2DL1*003 by substitution of glycine for serine at position 179 in the D2 domain (Supplemental Fig. 2A). Serine 179 prevents cell surface expression of KIR2DL1*014 and the binding of KIR2DL1*014-Fc to HLA class I (44). These properties suggest that serine 179 prevents KIR2DL1*014 from folding properly, thereby leading to a denatured protein and its intracellular retention.

To test this hypothesis, confocal microscopy was used to examine the cellular localization of FLAG-tagged KIR2DL1*014 in transiently transfected HeLa cells (Fig. 6B). KIR2DL1*012, an allotype expressed highly at the cell surface, served as the control. Phalloidin, which binds to intracellular actin, was used as an independent marker of the underside of the cell surface (Fig. 6Bb). In cells expressing FLAG-tagged KIR2DL1*012, the distribution of the anti-FLAG Ab (Fig. 6Bc, upper) overlapped with that of phalloidin. This is further seen in a merge of the two images (Fig. 6Bd, upper) and determination of the extent of colocalization (Fig. 6Be, upper). In the cells expressing KIR2DL1*014 the anti-FLAG Ab was detected inside the transfected cells (Fig. 6Bc, lower) but not at the cell surface similar to phalloidin (Fig. 6Bb, lower). On merging the two images (Fig. 6Bd, lower) there was no detectable colocalization (Fig. 6Be, lower). Quantification of the extent of the colocalization of phalloidin and anti-FLAG Ab in transfected HeLa cells expressing KIR2DL1*012 and KIR2DL1*014 is shown in Fig. 6C. This analysis demonstrates that transfected cells expressing KIR2DL1*014 make the protein but do not transport it to the cell surface. These data are consistent with KIR2DL1*014 not folding properly.

Analysis of the three-dimensional structure of KIR2DL1*001 (PDB:1NKR) (48) showed that position 179 of the D2 domain is buried beneath the binding site for HLA-C (Fig. 6D, left panel). We modeled the effect of replacing glycine 179 of KIR2DL1*001 with the serine 179 of KIR2DL1*014. In this model, the side chain substitution of a hydrogen atom for a methyl group displaces the tyrosine at position 134 (Fig. 6D, right panels). We hypothesize that this displacement is incompatible with proper folding of the KIR2D molecule. Supporting this interpretation, tyrosine 134 and glycine 179 are conserved in hominoid lineage III KIR (Supplemental Fig. 4).

Analysis of hominoid KIR sequences demonstrated that positions 16 (D1 domain) and 148 (D2 domain) have been subject to positive selection during hominoid evolution (26). These residues juxtapose within the hinge region that connects the D1 and D2 domains (Fig. 7A) (49). Polymorphism at residues 16 and 148 of KIR2DL2/3 has been proposed to vary the angle between the D1 and D2 domains and account for KIR2DL2*001 and KIR2DL3*001 having different avidity and specificity for HLA-C (21, 50). Having proline 16 and arginine 148, KIR2DL3*001 is a C1-specific receptor of moderate avidity. In contrast, arginine 16 and cysteine 148 give KIR2DL2*001 high avidity for C1 and cross-reactivity with C2 (21, 32).

Sequence comparison of 61 KIR2DL1, KIR2DL2, and KIR2DL3 variants showed that arginine and proline are the only residues present at position 16, whereas arginine, proline, and cysteine can occur at position 148 (Fig. 7B). Five of the six possible combinations of these residues are present in human lineage III KIR. Absent is the combination of arginine 16 with proline 148 (Supplemental Fig. 4). The other five combinations are represented in the 36 KIR2DL2/3 allotypes, but only two combinations are represented in the 33 KIR2DL1 and 2DS1 allotypes (Fig. 7C). To see how the variability at positions 16 and 148 influence avidity for HLA-C, we made 18 KIR-Fc fusion proteins in which all six combinations of the natural residues at positions 16 and 148 were introduced into KIR2DL1*003, KIR2DL2*001, and KIR2DL3*001. These KIR-Fc fusion proteins were tested for binding to HLA class I.

KIR2DL1*003 and four of the five KIR2DL1 mutants have similarly high avidity and specificity for HLA-C, exhibiting <15% variability in the binding to any HLA-C allotype (Fig. 7D). In contrast, the mutant combining proline 16 with proline 148 retained high specificity for C2 but its avidity was reduced to 60% that of KIR2DL1*003.

In comparison with the relative insensitivity of KIR2DL1*003 to mutation, all the KIR2DL2 and KIR2DL3 mutants exhibit detectable differences (Fig. 7D). For KIR2DL2 the major effect of mutation was to change the avidity, whereas the specificity, of stronger binding to C1 and weaker binding to C2, was less affected. Arginine residues correlate with the highest binding. Thus, the mutant with arginine 16 and arginine 148 binds more strongly to HLA-C than does the wild-type KIR2DL2*001 that combines arginine 16 with cysteine 148. Alternatively, lower binding correlates with the presence of proline at positions 16 and 148, with the weakest receptor having proline 16 and proline 148. The effect of cysteine at position 148 is intermediate between that of arginine and proline. These general effects of arginine and proline are seen for KIR2DL1, KIR2DL2, and KIR2DL3 (Fig. 7E).

Three of the KIR2DL3 mutants have higher avidity for HLA-C than does wild-type KIR2DL3*001. What distinguishes KIR2DL3*001 from the five mutants is its minimal cross-reactivity with C2. It therefore appears that KIR2DL3*001 evolved under selection for C1 specificity, with avidity being of secondary importance. In contrast, the evolution of KIR2DL1*003 resulted in a receptor having both high avidity and high specificity for C2. The properties of KIR2DL2 are different again. KIR2DL2 recognizes both C1 and C2, making it a stronger C1 receptor than KIR2DL3 and a weaker C2 receptor than KIR2DL1. Thus, KIR2DL2 appears as an evolutionary compromise, or intermediate, between KIR2DL1 and KIR2DL3.

KIR2DS1 has specificity for C2 (20), similar to KIR2DL1, but is less polymorphic. KIR2DS1 alleles encoding seven KIR2DS1 allotypes have been described (37). KIR2DS1*002 is by far the most frequent allele, being present in every population that has been KIR typed at allele-level resolution (29, 38–44). To study the avidity and specificity of KIR2DS1, we constructed KIR-Fc fusion proteins for the four KIR2DS1 allotypes that differ in the amino acid sequences of the D1 and D2 domains that bind HLA class I (Fig. 8A).

The four KIR2DS1 allotypes all bound to the seven C2-bearing HLA-C allotypes (Fig. 8B) but did not bind to the nine C1-bearing HLA-C allotypes (data not shown). KIR2DS1*001 has the highest avidity for C2, which correlates with arginine at position 70, where the other KIR2DS1 allotypes have lysine. KIR2DS1*008 has the lowest avidity for C2, which correlates with serine 123, where the other KIR2DS1 allotypes have tryptophan. Having a similar intermediate avidity for C2 are KIR2DS1*002 and KIR2DS1*004. They differ at position 90, where the valine–leucine dimorphism has little effect on the receptors’ avidity. Thus, the observed differences in avidity can be attributed to the dimorphisms at position 70 in the D1 domain and 123 in the D2 domain. Position 70 has been subject to positive selection in hominoid KIR and has a dominant effect in modifying the avidity of KIR2DL1 and KIR2DL3 (51). Arginine at position 70 improves recognition of HLA-C as shown in this study for KIR2DS1 and previously for KIR2DL3 (51). Although, overall, KIR2DS1 allotypes are not as avid HLA-C2 receptors as KIR2DL1, there are KIR2DL1 allotypes that have similar HLA-C2 avidity to each of the KIR2DS1 allotypes (Fig. 8B).

This study investigated the genetic and functional diversity of human HLA-C receptors. We showed that the inhibitory KIR2DL1, activating KIR2DS1, and inhibitory KIR2DL2/3 alleles form distinctive phylogenetic clades that associate with specific KIR haplotypes. Typifying KIR Cen A haplotypes are KIR2DL1 alleles that encode strong inhibitory C2 receptors and KIR2DL2/3 alleles encoding weak inhibitory C1 receptors. In striking contrast, Cen B haplotypes combine KIR2DL1 alleles that encode weak inhibitory C2 receptors with KIR2DL2/3 alleles encoding strong inhibitory C1 receptors.

Our results are in accordance with the first descriptions of the genomic organization of KIR haplotypes (28, 52, 53). These studies identified the segregation of KIR2DL3 with KIR A haplotypes and KIR2DL2 with KIR B haplotypes. Subsequent work highlighted the functional differences between the receptors encoded by these allelic variants (21, 32). High-resolution KIR analysis of white populations consistently associated KIR2DL1*004 with the KIR B haplotype, and in positive linkage disequilibrium with KIR2DL2 (30). Functional differences between the KIR2DL1 alleles of KIR A and B haplotypes were likewise discovered when NK cells expressing KIR2DL1*004 were found to be hyporesponsive in comparison with NK cells expressing KIR2DL1*003 (33). Dimorphism at position 245 in the transmembrane region of KIR2DL1 was demonstrated to be one mechanism that causes such functional differences (31).

We show that such attenuated function is not limited to KIR2DL1*004, or even to the group of KIR2DL1 allotypes having cysteine 245, but it is a defining characteristic of the KIR2DL1 encoded by Cen B–associated alleles. A variety of mechanisms cause functional attenuation, with mutations in the ligand-binding domains regulating the avidity and specificity of KIR2DL1 for HLA-C. Such changes can synergize with those in the transmembrane domain (which predominantly regulate cell surface expression and signaling function) to produce a broad range of functionally distinct receptors that share the common feature of being weakened in comparison with the KIR2DL1 allotypes associated with the KIR A haplotype.

Examination of KIR2DL2/3 allotypes revealed a range of functional properties, similar to those seen for KIR2DL1. In contrast to KIR2DL1, the weaker C1 receptors are associated with KIR A haplotypes and the stronger receptors are associated with KIR B haplotypes. Additionally, the weaker KIR A–associated receptors are more specific C1 receptors than the KIR B–associated receptors, which to varying degree exhibit cross-reactivity for C2. Unlike KIR2DL1, most functional differences between KIR2DL2/3 allotypes arise from amino acid substitutions in the ligand-binding domains.

Of interest in this regard are KIR2DL2*008 and KIR2DL2*010, two alleles for which haplotype associations have yet to be determined. The ligand-binding domains are identical to those of the KIR2DL2*001 prototype. Where they differ from KIR2DL2*001, and all other KIR2DL2/3 allotypes, is in the transmembrane domain. They have cysteine instead of arginine at position 245. Given the functional attenuation caused by cysteine 245 in KIR2DL1 (31), it is likely that cysteine 245 in KIR2DL2*008 and KIR2DL1*010 weakens their function by reducing cell surface expression and inhibitory signaling. Supporting this hypothesis, KIR2DL2 and KIR2DL1 have almost identical transmembrane and cytoplasmic domains (21). We further predict that KIR2DL2*008 and KIR2DL2*010 associate with Cen A haplotypes, not Cen B haplotypes, similar to other KIR2DL2 alleles. A precedent for such “haplotype infidelity” is that KIR2DL3*014 and KIR2DL3*018 are strong C1 receptors that cross-react with C2 and associate with Cen B.

Our results indicate that polymorphisms in all of the structural domains of the mature protein can impact the function of KIR with respect to the initiation and propagation of inhibitory or activating signals. Substitutions that change the avidity of KIR2DL1 for HLA-C are usually in the D2 domain, whereas substitutions that change the avidity of KIR2DL2/3 for HLA-C are usually in the hinge region connecting the D1 and D2 domains. The only site where substitution has significant impact on the specificity of KIR2DL1 for C2, or KIR2DL2/3 for C1, is position 44 in the D1 domain. Substitutions that affect the cell surface expression of KIR2DL are more often in the transmembrane or the cytoplasmic domain, but there are also substitutions in the extracellular domains that affect receptor function in this way. An extreme example is KIR2DL1*014, which is completely retained inside the cell as a consequence of having serine, rather than glycine, at position 179 in the D2 domain. We suspect that this substitution prevents proper folding of KIR2DL1*014.

Similar to KIR2DL1*014, KIR2DL1*013N, KIR2DL1*026, KIR2DL2*004, and KIR2DL3*008N are inhibitory lineage III KIR that are unlikely to be functional receptors (Table I). We examined the extent to which inactivation of these receptors affected the capacity of their haplotypes to provide inhibitory receptors that recognize C1 and C2. From the initial description of KIR A and B haplotypes it was appreciated that they both encode inhibitory receptors that recognize C1 and C2 (28, 52). Subsequent studies established the combination of KIR2DL1 and KIR2DL3 on Cen A and of KIR2DL1 and KIR2DL2 on Cen B (25, 30, 40–42). Overlaying our functional binding data onto the high-resolution KIR haplotypes of seven populations, we find that every KIR2DL1 allotype encoded by Cen A recognizes C2 and that every KIR2DL3 allotype encoded by Cen A recognizes C1. The exception is a KIR A haplotype of Yucpa Amerindians, which has a frequency of 7.4% and combines a strong, C2-specific receptor with the nonfunctional KIR2DL3*008N allele (29). Consequently, the receptors encoded by this haplotype cannot recognize C1, although the frequency of the haplotypes is such that <1% of individuals in the population would be deficient in C1 recognition.

Cen B haplotypes that encode KIR2DL1*022 and KIR2DL1*026 (neither of which can mediate inhibition via engagement with C2) are in strong linkage disequilibrium with KIR2DL2*003 and are specific to the KhoeSan and immediately neighboring populations in southern Africa (44). Cen B haplotypes that lack the KIR2DL1 gene (KIR2DL1 blank) are widespread and associated with KIR2DL2*003, KIR2DL2*001 and KIR2DL2*005 (Table I) (38–44). These Cen B haplotypes retain some capacity to recognize C2 because their associated KIR2DL2 allotypes cross-react with C2 (Supplemental Fig. 3B). The functional importance of these cross-reactions needs to be assessed in cellular assays of NK cell education and effector response.

Although the relative frequency of KIR A and B haplotypes varies substantially across the world, as do C1 and C2 frequencies, there is a strong correlation between C2 and KIR B and a corresponding inverse correlation between C2 and KIR A (14). Thus, it appears that in populations with high C2 frequency, such as those in Africa, there has been selection for weak C2 receptors and strong C1 receptors (found on KIR B haplotypes) whereas in populations with high C1 frequency, such as those in Asia and the Americas, there has been selection for weak C1 receptors but strong C2 receptors (found on KIR A haplotypes). Underlying these observations, and implicating a strong inhibitory C2 receptor–ligand interaction in their pathogenesis, are correlations with pregnancy syndromes. Thus, women who are pregnant with a fetus expressing C2 are at increased risk of pre-eclampsia (14, 15). Those same strong inhibitory KIR–ligand interactions are, however, vital for the development of well-educated NK cells that are both self-tolerant and responsive to virally infected and malignantly transformed cells. Thus, a pattern emerges in which KIR haplotypes with contrasting functional properties are subject to selection in response to the relative abundance of HLA-C ligand. In this way, the KIR system may be considered a buffering mechanism by which optimal NK cell function is preserved, despite fluctuations in the frequency of available ligand.

The authors have no financial conflicts of interest.