Interactions between HLA-C ligands and inhibitory killer cell Ig-like receptors (KIR) control the development and response of human NK cells. This regulatory mechanism is usually described by mutually exclusive interactions of KIR2DL1 with C2 having lysine 80, and KIR2DL2/3 with C1 having asparagine 80. Consistent with this simple rule, we found from functional analysis and binding assays to 93 HLA-A, HLA-B, and HLA-C isoforms that KIR2DL1*003 bound all C2, and only C2, allotypes. The allotypically related KIR2DL2*001 and KIR2DL3*001 interacted with all C1, but they violated the simple rule through interactions with several C2 allotypes, notably Cw*0501 and Cw*0202, and two HLA-B allotypes (B*4601 and B*7301) that share polymorphisms with HLA-C. Although the specificities of the “cross-reactions” were similar for KIR2DL2*001 and KIR2DL3*001, they were stronger for KIR2DL2*001, as were the reactions with C1. Mutagenesis explored the avidity difference between KIR2DL2*001 and KIR2DL3*001. Recombinant mutants mapped the difference to the Ig-like domains, where site-directed mutagenesis showed that the combination, but not the individual substitutions, of arginine for proline 16 in D1 and cysteine for arginine 148 in D2 made KIR2DL2*001 a stronger receptor than KIR2DL3*001. Neither residue 16 or 148 is part of, or near to, the ligand-binding site. Instead, their juxtaposition near the flexible hinge between D1 and D2 suggests that their polymorphisms affect the ligand-binding site by changing the hinge angle and the relative orientation of the two domains. This study demonstrates how allelic polymorphism at sites distal to the ligand-binding site of KIR2DL2/3 has diversified this receptor’s interactions with HLA-C.
Natural killer cells are lymphocytes of the innate immune response that provide an important defense against infection, particularly viral infections (1, 2). NK cells also function in reproduction by controlling the remodeling of maternal blood vessels that occurs during implantation (3, 4). A wide variety of activating and inhibitory receptors determine the NK cell response and create both heterogeneity and repertoire in NK cell populations (5, 6). Common to several families of NK cell receptors is their ligand specificity for MHC class I and related molecules. In humans, polymorphic determinants of the ubiquitously expressed, classical MHC class I molecules are recognized by members of the killer cell Ig-like receptor (KIR)3 family (7, 8, 9). Of these, the inhibitory KIR are proven to have a strong regulatory effect on the development and effector functions of NK cells, whereas the functions and significance of the activating KIR remain enigmatic (10, 11, 12).
Although epitopes of HLA-A and HLA-B are recognized by inhibitory KIR3DL1 and KIR3DL2, only ∼50% of HLA haplotypes encode such epitopes, and consequently ∼25% of the human population lack them altogether (13). In contrast, with very few exceptions, all humans carry both an HLA-C allotype and a cognate KIR, and they use them to regulate and diversify their NK cell population (14). It thus appears that HLA-C, a locus that is recently evolved and specific to humans and apes (6, 15), dominates HLA-A and HLA-B in the regulation of NK cells.
Beginning with analysis of the specificity of alloreactive NK cells, two different groups of HLA-C receptors have been recognized and correlated with sequence dimorphism at position 80 of the α1 domain (7, 8). Of 312 HLA-C allotypes currently defined, half of them have asparagine at position 80, the C1 epitope, and half have lysine, the C2 epitope (16). Studies to correlate KIR expression with functional activity indicated that KIR2DL1 has C2 specificity and that KIR2DL2 and KIR2DL3 have C1 specificity (7, 8, 17). Furthermore, dimorphism at position 44 in the D1 domain of the KIR correlates with the receptors’ C-specificities, and mutation at this position is sufficient to convert a receptor’s specificity from C1 to C2 and vice versa (18).
Although the KIR2DL2 and KIR2DL3 cDNA sequences were given different names because they were first thought to represent different genes (17, 19, 20), subsequent genomic analysis and population studies showed that KIR2DL2 and KIR2DL3 segregate as alleles of one locus that can be designated as KIR2DL2/3 (21). Comparisons of gene sequence and gene organization within KIR haplotypes are consistent with KIR2DL2 having evolved from a KIR2DL3-like ancestor by nonreciprocal recombination with a KIR2DL1-like ancestor (22). As a consequence, KIR2DL2 is very similar to KIR2DL3 in the ligand-binding Ig-like domains, but less so in the stem, transmembrane, and cytoplasmic domains, where it more closely resembles KIR2DL1. Pointing to the functional importance of the polymorphic differences between KIR2DL2 and KIR2DL3 was correlation of the combination of C1 and KIR2DL3, but not C1 and KIR2DL2, with resolution of acute hepatitis C virus infection (23). These observations were interpreted in terms of earlier observations by Winter et al., which suggested that KIR2DL3 is a weaker and more specific C1 receptor than is KIR2DL2 (24). To define precisely the functional differences between KIR2DL2 and KIR2DL3 and their molecular basis, we combined mutagenesis analysis with functional studies and direct binding studies of KIR2DL1, KIR2DL2, and KIR2DL3 on an extensive array of HLA class I allotypes.
Materials and Methods
NKL, a leukemia-derived cell line with NK cell-like properties, has previously been characterized and was maintained as described (25, 26). Individual HLA transfectants were generated in the HLA-A-, HLA-B-, and HLA-C-deficient cell line 221, as previously described, with critical leader peptide residues mutated to abrogate HLA-E expression (27).
Human subjects and blood samples
Blood samples were obtained from healthy individuals; written consent was obtained from all donors. Procedures followed the approval by the Stanford University Institutional Review Board on Human Subjects. PBMC were prepared by Ficoll gradient separation (GE Healthcare). KIR gene-content typing was performed by sequence-specific polymorphism-PCR-based typing as described previously (11).
Class I-mediated inhibition of NK cell subsets
HLA-C1- and HLA-C2-mediated inhibition of NK subsets was assessed as previously described with modifications (11, 28). PBMC (1 × 106) were cocultured with HLA class I-deficient 221 cells, or 221 cells transfected with HLA-C, at an optimized effector/target ratio of 5:1 for 6 h. Brefeldin A (BD Biosciences) was added for the last 5 h of coincubation. Intracellular staining for IFN-γ using mAb PE-Cy7-anti-IFN-γ (4S.B3, BD Biosciences) on gated CD56+CD3− cells measured the NK cell response. mAbs used in flow cytometric analysis were FITC-anti-2DL2/2DL3/2DS2 (CHL, BD Biosciences), PE-Cy5-anti-LILRB1 (GHI/75, BD Biosciences), allophycocyanin-anti-2DL1/2DS1 (EB6, Beckman Coulter), APC-Cy7-anti-CD3 (S4.1, BD Biosciences), and PE-Cy5.5-anti-CD56 (MEM-188, Invitrogen). Dead cells were excluded from analyses. EB6+CHL−, EB6−CHL+, and EB6−CHL− subsets in the CD56dimCD3−LILRB1− NK cell population were independently assessed. Data analysis was performed using FlowJo software (TreeStar). Inhibition was calculated as described previously (11, 28).
Generation of NKL-KIR
NKL expressing wild-type and mutant KIR were generated as previously described (26). Full-length coding regions of KIR2DL1*003, KIR2DL2*001, and KIR2DL3*001 were amplified by PCR and cloned into the pBMN retroviral vector (a gift from Garry Nolan, Stanford University, Stanford, CA). Site-directed mutagenesis was performed using the QuikChange mutagenesis kit (Stratagene) according to the manufacturer’s instructions. Domain-swap mutants were generated using two-step recombinant PCR. Recombinant amphotrophic retrovirus was generated by transfection into Phi-NX cells using standard protocols. Supernatants were used to infect growing NKL cells, and stable, KIR-expressing cells were sorted for equivalent cell-surface expression levels using a FACStar cell sorter (BD Biosciences).
Killing of transfected and untransfected 221 cells by NKL transductants was assayed as described (26). Briefly, effector cells were mixed with 51Cr-loaded target cells for 4 h at 37°C at ratios ranging from 20:1 to 5:1. Following incubation, supernatants were harvested and 51Cr quantified using a Wallac β-scintillation counter. Specific lysis was calculated using the formula (specific release − spontaneous release)/(total release − spontaneous release). Experiments were conducted in triplicate for each condition, and each experiment was independently replicated three or more times.
Cold target competition assays were at a fixed 20:1 E:T ratio using untransfected 221 cells as the 51Cr-labeled target. Briefly, effector cells were coincubated for 30 min at 37°C with unlabeled target lines (parental 221 or HLA-C transfectants thereof). 51Cr-labeled parental 221 cells were then added and incubated for 4 h at 37°C and harvested as described above.
Generation of KIR-Fc fusion proteins
Regions encoding the Ig-like domains and stem of KIR2DL1, KIR2DL2, and KIR2DL3 were amplified by PCR and fused in-frame with the region encoding the Fc portion of the human IgG1 H chain. This chimeric product was inserted into the transfer vector pACgp67 and cotransfected with linearized baculovirus (BD Biosciences) into Sf9 cells using Cellfectin according to the manufacturer’s directions (Invitrogen). One to two further rounds of amplification were necessary to produce high-titer virus. KIR-Fc fusion proteins were produced by infection of Hi5 insect cells for 60 h. Supernatants were harvested by centrifugation and sterile filtration. Supernatants were then neutralized with HEPES buffered saline (final concentration 150 mM HEPES, 20 mM NaCl (pH 7.2)) and incubated overnight with protein A-Sepharose beads (Invitrogen). The protein was harvested, washed, and eluted with 0.1 M glycine (pH 2.7) and immediately neutralized using 0.2 M Tris (pH 9.0).
Single-Ag bead analysis of KIR-Fc specificity
Binding of KIR-Fc fusions to a broad panel of HLA-A, HLA-B, and HLA-C allotypes was assessed using commercially available LABScreen single-Ag bead sets (One Lambda). Cumulatively, the three sets encompass 29 HLA-A, 50 HLA-B, and 16 HLA-C allotypes. KIR-Fc fusion proteins at concentration ranging from 400 to 0.1 μg/ml were incubated with LABScreen microbeads for 30 min at room temperature. Beads were then washed three times and labeled with anti-human Fc-PE (One Lambda). Fluorescent intensity and identification labels of the individual beads were visualized on a Luminex 100 reader (Luminex). A minimum of 200 events per Ag were collected. Results shown are mean fluorescence and are expressed as relative fluorescence ratios, calculated using the formula (specific binding − control bead binding)/(positive binding − control bead binding). The W6/32 (anti-HLA class I) and BBM.1 (anti-β2M) Abs were used as positive controls and to account for bead-to-bead differences in the amount of HLA class bound to each bead. Bead nos. 66 and 90, coated with HLA-B*5102 and HLA-C*1203, respectively, gave low levels of binding with the Ab controls and were excluded from the analysis.
HLA-C2 is a weak ligand for KIR2DL2 and KIR2DL3 expressed by blood NK cells
In exploratory experiments, HLA-C-mediated inhibition of the IFN-γ response was compared for NK cells expressing KIR2DL1, KIR2DL2, and KIR2DL3 (Fig. 1). Analysis of KIR2DL-expressing NK cells from five donors homozygous for KIR A haplotypes showed that cells expressing KIR2DL1 were strongly inhibited by target cells expressing C2 (Cw*0401) but not C1 (Cw*1202). In contrast, NK cells expressing KIR2DL3 exhibited strong inhibition by C1 and a weaker inhibition by C2, but the results were statistically significant compared to those obtained with cells lacking KIR2DL1, KIR2DL2, and KIR2DL3 (Fig. 1,A). For the single donor homozygous for KIR B haplotypes, similar inhibition of KIR2DL2-expressing cells was observed (Fig. 1,B). Analysis of three AB heterozygotes revealed the same trend (Fig. 1 C). These trends for AB and BB donors did not reach statistical significance, possibly because of the small number of donors and/or the general inability to discriminate cells expressing inhibitory KIR2DL2 and/or activating KIR2DS2 with available mAb. The results suggested that KIR2DL2 and KIR2DL3 are less specific for C1 than KIR2DL1 is for C2; they also emphasized the inherent difficulty in dissecting KIR2D reactivity using primary NK cells from donors with all but the simplest KIR AA genotypes.
Interaction between C2 and KIR2DL2 strongly inhibits NKL cell cytotoxicity
To avoid the confounding effects of activating KIR and the coexpression of multiple inhibitory KIR, we used an in vitro system that analyzed the killing of target cells expressing single HLA-C allotypes by effector cells expressing single KIR2DL. NKL cells were transduced with KIR2DL1*003, KIR2DL2*001, and KIR2DL3*001, representing the most common alleles. Although there is no detectable KIR on NKL cell surfaces, they do express the CD94:NKG2A inhibitory receptor that is specific for complexes of HLA-E bound to peptides derived from HLA-A, HLA-B, and HLA-C leader sequences. To eliminate CD94:NKG2A-mediated inhibition of NKL cells, the HLA class I alleles used to transfect 221 cells were purpose-made mutants having leader sequences that do not permit HLA-E binding to CD94:NKG2A.
In cytotoxicity assays, NKL cells expressing KIR2DL1 were inhibited by target cells expressing C2 allotypes (Cw*0401 or Cw*1503) but not by target cells expressing C1 allotypes (Cw*0304 or Cw*0803). Reciprocally, NKL cells expressing KIR2DL3 were inhibited strongly by target cells expressing C1, but to little or no extent by target cells expressing C2 (Fig. 2,A). In contrast, NKL cells expressing KIR2DL2 were strongly inhibited by target cells expressing either C1 or C2 (Fig. 2,B). Our further observations that KIR2DL2-expressing NKL cells were not inhibited by target cells expressing single HLA-A (A*0201 or A*2403) or HLA-B (B*0801 or B*5701) allotypes demonstrated that C2-mediated inhibition by KIR2DL2 is not a nonspecific artifact (Fig. 2 C).
To eliminate any variability caused by differences in lysis of the individual 51Cr-labeled target cells, we devised a cold-target competition assay. In this assay the lysis of 51Cr-labeled 221 cells (the hot target) by KIR2DL-transduced NKL was measured in the presence of variable numbers of unlabeled “cold targets”: either 221 cells or 221 cells transfected with HLA class I. Cold targets lacking a ligand for the KIR2DL expressed by NKL cells compete for lysis with the 51Cr-labeled 221 cells, and the latter are lysed less effectively. Conversely, cold targets that express a class I ligand for the NKL cell’s inhibitory KIR2DL do not compete with the 51Cr-labeled 221 cells. Using this assay, we found that lysis of 51Cr-labeled 221 cells by KIR2DL2-transduced NKL cells was not reduced by cold targets expressing either C1 or C2 (Fig. 3, top). These results confirm that interactions between KIR2DL2 and either C1 or C2 can effectively inhibit NK cell cytolysis. In contrast, the lysis by KIR2DL3-transduced NKL cells was strongly reduced by C1-expressing cold targets and only slightly by C2-expressing cold targets (Fig. 3, middle). Lysis by KIR2DL1-transduced NKL cells was reduced strongly by C2-expressing cold targets and weakly by C1-expressing cold targets (Fig. 3, bottom). This could be due to weak cross-reactivity of KIR2DL1 with HLA-C1, a possibility that seems less likely given the lack of any such reactivity in the direct cytotoxicity (Fig. 2) and direct binding assays (Fig. 4). Alternatively, the inhibition could arise from C1 interaction with LILRB1, which is expressed by NKL (25), or from some low-level inhibition through CD94:NKG2A caused by HLA-C-derived peptides bound to HLA-E. Such mechanisms might also account for C2-expressing targets being marginally less competitive cold targets than is 221 for NKL-2DL3. These uncertainties, however, do not affect our main conclusion that interactions of KIR2DL2 on NK cells with either C1 or C2 on target cells can inhibit NK cell function in a specific manner.
KIR2D-Fc fusion proteins bind to all HLA-C, two HLA-B, and no HLA-A allotypes
Bivalent Fc-fusion proteins were made from KIR2DL1*003, KIR2DL2*001, and KIR2DL3*001 and tested for binding to beads coated with single HLA class I allotypes (One Lambda single Ag beads). In these assays the binding of KIR2D-Fc proteins to each bead was normalized to the binding of a monomorphic anti-HLA class I mAb to the same bead. The binding to 29 HLA-A, 49 HLA-B, and 15 HLA-C allotypes was independently assessed (Fig. 4).
As expected, the KIR2D-Fc did not bind to any HLA-A allotype or to 47 of the 49 HLA-B allotypes tested. The two exceptional HLA-B allotypes that bound KIR2D-Fc have structural features in common with HLA-C that distinguish them from other HLA-B. B*4601, which acquired the C1 epitope by gene conversion, bound KIR2DL2-Fc and KIR2DL3-Fc, consistent with our previous analysis demonstrating that this allotype is a ligand for C1-specific KIR (29). An unanticipated observation was that B*7301, the most divergent HLA-B allotype (30), also bound KIR2DL2-Fc and KIR2DL3-Fc at levels similar to B*4601 (Fig. 4,A). This property correlates with B*7301 sharing valine 76 with all HLA-C allotypes and B*4601, a substitution known to contribute to the affinity of HLA-C for KIR2DL (31). Whereas only 2 of the 78 HLA-A and HLA-B allotypes bound a KIR2D-Fc fusion protein, all 15 HLA-C allotypes bound to one or more of the KIR2D-Fc fusion proteins (Fig. 4 B). In summary, the high degree of HLA-C locus-specificity of the KIR2D-Fc binding reactions and the clear structural basis for the cross-reactions with the two exceptional HLA-B allotypes give one confidence that the binding specificities of the KIR2D-Fc reflect those of their respective cell-surface KIR.
Direct binding assays show specific interactions of KIR2DL2 and KIR2DL3 with C2
KIR2DL1-Fc bound to all seven HLA-C allotypes having the C2 epitope and to none of the eight HLA-C allotypes having the C1 epitope (Fig. 4,B). The level of binding varied among the C2 allotypes, being strongest for Cw*0501 and Cw*0202 (Fig. 5, bottom). Thus, the KIR2DL1-Fc specificity for HLA-C in this direct binding assay is identical with that inferred from assays of cytokine production (Fig. 1) and cytotoxicity (Figs. 2 and 3).
At varying level, KIR2DL3-Fc bound to all eight C1 allotypes and to three of the seven C2 allotypes: Cw*0501, Cw*0202, and, to lesser extent, Cw*0401 (Figs. 4,B and 5, middle). In particular, Cw*0501 bound stronger than did several of the C1 allotypes. KIR2DL2-Fc variably bound to the eight C1 allotypes, but with levels generally higher than those attained by KIR2DL3-Fc; KIR2DL2-Fc also bound the seven C2 allotypes (Fig. 5, top). Of these, Cw*0501 bound KIR2DL2-Fc and KIR2DL3-Fc at levels comparable to the C1 allotypes; Cw*0202 bound KIR2DL3-Fc similar to the weakest C1 allotypes, but its binding to KIR2DL2-Fc was lower than for C1 allotypes and was within the range of the other C2, excepting Cw*0501.
Two of the HLA-C allotypes examined in the cytotoxicity assay were tested in the direct binding assay (i.e., Cw*0304 and Cw*0401). The KIR2DL2 and KIR2DL3 fusion proteins both bound strongly to Cw*0304-coated beads, while KIR2DL1-Fc showed no binding. The Cw*0401 beads bound strongly to KIR2DL1-Fc, and their binding to KIR2DL2-Fc was among the strongest obtained with C2 allotypes, representing results consistent with those obtained in cytotoxicity assays (Fig. 2). The KIR2DL3 fusion protein bound Cw*0401, matching the weak cross-reactivity seen in the IFN-γ assay (Fig. 1). Different allotypes of HLA-Cw*08 were tested in the cytotoxicity (Cw*0803) and binding (Cw*0801) assays. These allotypes gave comparable patterns of interaction with KIR2D, consistent with their single amino acid difference being at position 175 located away from the site of HLA-C interaction with KIR and peptide. In contrast, two allotypes of HLA-Cw*15 gave significantly different results in the cytotoxicity (Cw*1503) and binding (Cw*1502) assays. Whereas Cw*1503 gave strong inhibition of cytoxicity mediated by KIR2DL2, the binding of KIR2DL2-Fc to beads coated with Cw*1502 was weak. Such a difference is consistent with the single difference between the two allotypes being at position 73 (threonine in Cw*1502 and alanine in Cw*1503), which is proximal to the site of KIR-HLA interaction and influences peptide repertoire selection through direct contacts in the C pocket of HLA-C (32).
As in the cytoxicity assays, we found that C2 allotypes can be ligands for KIR2DL2, but they are generally weaker than C1 allotypes. That the hierarchies with which KIR2DL2-Fc and KIR2DL3-Fc bind to the various HLA allotypes are very similar points to their binding differences being ones of avidity rather than specificity. In turn, this suggests that the more qualitative differences seen between KIR2DL2 and KIR2DL3 in the cytotoxic assays (Figs. 2 and 3) could arise because the strength of the interaction between KIR2DL2 and C2 exceeds a threshold necessary for functional inhibition that is not reached by the weaker interaction of KIR2DL3 with C2. On average, binding of KIR2DL2-Fc to C1 and C2 allotypes was enhanced by 25% and 108%, respectively, over KIR2DL3-Fc (measured at 25 μg/ml). The weaker enhancement for C1 allotypes is likely due to both fusion proteins approaching saturation at this concentration. We also find that certain C2 allotypes, notably Cw*0501 and Cw*0202, bind to KIR2DL1, KIR2DL2, and KIR2DL3, although to differing extents. That Cw*0202 and Cw*0501 have identical contact residues for p8 and p9, the residues of the bound peptide that influences HLA-C interaction with KIR (32, 33, 34), suggests peptide effects could contribute to the broader and stronger binding reactions of these two HLA-C allotypes. Not all HLA-C-binding peptides are permissive to KIR interaction (35, 36), raising the possibility that the differences we observe in the levels of saturation of KIR2D-Fc binding to the HLA-C allotypes is in part due to the different proportions of peptides they bind that are permissive to KIR interaction. Another possible cause of the differences is that the binding of anti-HLA class I mAb to beads does not correlate well with the accessibility of the KIR binding site on HLA-C because different allotypes tend to attach to the beads in different orientations.
Stronger interactions with C2 are a function of the extracellular domains of KIR2DL2
KIR2DL2 is a recombinant (22) in which the extracellular domains are structurally more similar to KIR2DL3, whereas the stem, transmembrane, and intracellular domains are much closer to KIR2DL1 (Fig. 6). Because C2-mediated inhibition is principally a property of KIR2DL1, we hypothesized that KIR2DL2’s capacity for C2-mediated inhibition derived from its KIR2DL1-like stem, transmembrane, and intracellular domains. To test this hypothesis, we made chimeric constructs in which the extracellular Ig-like domains of KIR2DL2 were paired with the stem, transmembrane, and intracellular domains of KIR2DL3, and vice versa. These constructs were transduced into NKL cells, and the chimeric KIR2DL were tested for their capacity to inhibit the killing of target cells expressing C1 or C2.
NKL cells expressing either chimeric KIR2D were, like wild-type KIR2DL2 and KIR2DL3, strongly inhibited by C1-bearing target cells (Fig. 7). Only wild-type KIR2DL2 and the chimera having the Ig-like domains of KIR2DL2 (mL2–L3), were inhibited by C2, demonstrating that polymorphisms that distinguish the Ig-like domains of KIR2DL2 and KIR2DL3 are responsible for the functional differences. Conversely, the observation that NKL cells expressing mL3–L2, which combines the Ig-like domains of KIR2DL3 with the stem, transmembrane, and cytoplasmic domains of KIR2DL2, were not inhibited by C2 (Fig. 7) shows that differences in the transmembrane and cytoplasmic domains of KIR2DL2 and KIR2DL3 do not contribute to the differences in C2-mediated inhibition. On the contrary, NKL cells expressing the mL3–L2 chimera showed increased lysis of C2-expressing target cells compared with NKL expressing wild-type KIR2DL3. This suggests that any contribution from the signaling domains to KIR-mediated inhibition is minor and is outweighed by the effect of the extracellular polymorphisms.
In conclusion, this analysis disproved our hypothesis that the stronger interactions of KIR2DL2 with C2 were due to its KIR2DL1-like stem, transmembrane, and cytoplasmic regions. Instead, the results show that this difference in C2-binding must lie with the polymorphic positions in the Ig-like domains.
Polymorphisms at positions 16 and 148 of KIR2DL2 and KIR2DL3 determine the differences in HLA-C binding
Distinguishing the Ig-like domains of KIR2DL2*001 and KIR2DL3*001 are four polymorphic positions: 16 and 35 in the D1 domain, and 148 and 200 in the D2 domain. For the two D1 positions, KIR2LD2*001 has the same residue as KIR2DL1*003, whereas for the two D2 positions, KIR2DL3*001 and KIR2DL1*003 are identical and KIR2DL2*001 is unique. To examine the contribution of each substitution to the interaction with C2, a series of mutants was made in which the residues present in KIR2DL2 were replaced with those in KIR2DL3, and vice versa.
NKL transductants expressing each mutant were studied for their capacity to lyse untransfected 221 cells and 221 cells expressing HLA-C allotypes carrying the C1 or C2 epitopes (Fig. 8). All of the NKL transductants expressing mutant KIR2D were inhibited by C1 in a manner similar to wild-type KIR2DL2 and KIR2DL3, showing that none of the substitutions affected the C1 specificity of either KIR (Fig. 8, A and C, left).
In contrast, single substitutions at positions 16, 35, and 148 weakened the interaction of KIR2DL2 with C2, whereas substitution at position 200 had no effect (Fig. 8, A and C, right). The reciprocal mutations at positions 16 and 148 in KIR2DL3 slightly increased the interaction with C2, whereas mutation at positions 35 and 200 had no effect. More substantial effects were seen for mutants combining several residue changes. Most importantly, mutation at positions 16 and 148 in KIR2DL2 completely abrogated interaction with C2, whereas the reciprocal mutations in KIR2DL3 conferred substantial, although not complete, interaction with C2 (Fig. 8, B and C). Triple mutation at positions 16, 35, and 148 had a similar effect as the double mutation at positions 16 and 148 (Fig. 8,C). Combined mutation of residues 16 and 35 or 35 and 148 in KIR2DL2 also resulted in partial but not complete loss of C2-mediated inhibition (Fig. 8 C). The reciprocal mutants of KIR2DL3 were slightly inhibited by C2, but not to the same level as wild-type KIR2DL2 or the arginine 16/cysteine 148 double mutant.
This mutational analysis showed that substitutions of arginine at position 16 in the D1 domain and cysteine at position 148 in the D2 domain are principally responsible for the differences in binding to HLA-C by KIR2DL2 and KIR2DL3 and for the strength of C2-mediated inhibition of cytolysis. In KIR2DL2, these substitutions comprise one that is shared with C2-specific KIR2DL1 (arginine 16) and one that is unique to KIR2DL2 (cysteine 148). Additionally, a minor contribution from glutamate 35 cannot be ruled out.
Polymorphisms at positions 16 in D1 and 148 in D2 of KIR2DL2/3 are predicted to change the angle between the two domains
Crystal structures of KIR2D alone and also bound to HLA-C show that none of the four polymorphic residues that distinguish the Ig-like domains of KIR2DL2 and KIR2DL3 contacts HLA-C directly (Fig. 9) (33, 34). Thus, the substitutions at positions 16 and 148 in KIR2DL2/3 must exert their effect on HLA-C binding in an indirect way. Residues 16 and 148 are situated near the hinge linking the D1 and D2 domains, where they face each other from complementary locations on the internal aspect of their respective domains. This juxtaposition is consistent with our observation that substitutions at positions 16 and 148 have a synergistic effect on C2 binding. The presence of residues 16 and 148 immediately below the hinge and adjacent to conserved residues that stabilize the hinge (positions 17 and 149, respectively) suggests that coordinated substitutions at positions 16 and 148 could change the angle of the hinge and/or its flexibility upon HLA-C interaction.
In the KIR2DL2 crystal structures, arginine 16 extends into the interdomain gap, bringing its side chain into close proximity to the CC′ loop of D2 (residues 141–148). The gap distance is small enough to support formation of an interdomain hydrogen bond with either the main-chain oxygen of alanine 145 or the sidechain of glutamic acid 147. Formation of such a bond is predicted to affect the interdomain hinge angle. Cysteine 148 is unique to KIR2DL2 and part of the CC′ loop of D2. In other KIR2DL, arginine 148 stabilizes the loop by forming an intradomain hydrogen bond with the main chain of proline 165. That cysteine 148 cannot form such a bond could explain the relative displacement of the CC′ loop in the structures of KIR2DL2 alone and bound to HLA-Cw3, one of few notable differences in the two structures (34). The CC′ loop also forms several interdomain contacts that stabilize the hinge (37). Our results from mutagenesis suggest that position 148 affects the CC′ loop’s conformation, which in turn influences the ability of arginine 16 to form an interdomain hydrogen bond. Taken together, these effects may synergize, allowing the KIR2DL2 hinge angle to assume a more acute conformation, like that seen in the structure of KIR2DL1 bound to HLA-Cw4 (33).
The interdomain hinge angle of KIR is characteristically acute compared with the closely related hemopoietic receptors and other Ig-superfamily receptors (37, 38). The angle also flexes on binding to HLA-C: increasing from 55 to 66° when KIR2DL1 binds HLA-Cw*04 (C2), and decreasing from 84 to 81° when KIR2DL2 binds HLA-Cw*03 (C1) (33, 34). The hinge angle of KIR2DL3 alone is 78°, more like KIR2DL2 than KIR2DL3, but the effect of ligand binding has yet to be determined.
Whereas interaction of KIR2DL with C2 narrows the interdomain hinge angle, it is widened by interaction with C1. Because the hinge residues are conserved in KIR2DL1, KIR2DL2, and KIR2DL3, polymorphism at other positions must influence the relative orientation of the two domains. The proximity of positions 16 and 148 to the hinge and the influence of their natural variation on HLA-C specificity make them likely candidates. We therefore propose that the interdomain interactions contributed by arginine 16 and cysteine 148 in KIR2DL2 favor an acute hinge angle that is more conducive to binding C2. This structural effect appears to strengthen interactions of KIR2DL2 with both C1 and C2 allotypes, relative to KIR2DL3.
The C1 and C2 ligands for inhibitory KIR2DL were first defined using NK cells from a few selected donors, at a time before the allelic polymorphism and haplotypic gene-content diversity of the KIR genes were appreciated (8, 39). KIR2DL1 was then described as being specific for HLA-C allotypes carrying the C1 epitope marked by lysine at position 80 (19, 39). Analysis here of the functional interactions of KIR2DL1*003, the most common KIR2DL1 allotype in a variety of populations, and its binding to 93 isolated HLA-A, HLA-B, and HLA-C allotypes demonstrated that KIR2DL1*003 is exquisitely specific for HLA-C allotypes bearing the C2 epitope. No interaction of KIR2DL1*003 with HLA-A, HLA-B, or C1-bearing HLA-C allotypes was observed. This simplicity to KIR2DL1 contrasts with the more complicated situation we observed for KIR2DL2 and KIR2DL3, originally described as specific for HLA-C allotypes having the C1 epitope marked by asparagine 80 (19, 39).
KIR2DL3*001 and KIR2DL2*001 are common allotypes of the KIR2DL2/3 locus. In addition to binding to all C1 HLA-C allotypes, KIR2DL2 and KIR2DL3 also bind C2 allotypes and two HLA-B allotypes, B*4601 and B*7301. No binding to HLA-A was detected. As for many HLA-B allotypes, B*4601 and B*7301 have asparagine 80, which is necessary but not sufficient to form the C1 epitope. Additionally, they also have valine 76, which is fixed in HLA-C but absent from other HLA-B. Presence of valine 76 and asparagine 80 is sufficient to form a functional C1 epitope (31), as demonstrated by functional analysis of B*4601 (29). In binding assays, KIR2DL2 shows some affinity for most C2 allotypes; for most, the binding is weaker than for C1 allotypes, but Cw*0501 and, to lesser extent, Cw*0202 bind similarly to the C1 allotypes. KIR2DL3 binds C1 and C2 allotypes with a hierarchy similar to KIR2DL2, but the levels of binding are consistently lower. These results, which are consistent with the preliminary observations of Winter et al. (24), indicate that the binding sites of KIR2DL2 and KIR2DL3 have similar specificity but different avidity for HLA class I.
Our mutational analysis of KIR2DL2/3 demonstrates that the combination of polymorphisms at positions 16 in the D1 domain and 148 in the D2 domain determines the avidity difference between the KIR2DL2 and KIR2DL3 binding sites. The combination of arginine 16 and cysteine 148 strengthens the KIR2DL2 binding site, whereas proline 16 and arginine 148 weakens the KIR2DL3 binding site. Consistent with their role in modulating avidity rather than specificity, residues 16 and 148 are distal to the MHC class I binding site and are close to the flexible hinge that connects D1 and D2 (33, 34). Residues 16 and 148 are at complementary sites on the D1 and D2 domains and they face each other in close juxtaposition. This location in the structure, combined with our finding that residues 16 and 148 act synergistically to alter binding strength, as well as crystallographic evidence that the KIR2D hinge angle increases on binding C2 and decreases on binding C1, mutually support a model in which arginine 16 and cysteine 148 give KIR2DL2 additional hinge flexibility that permits effective binding to both C2 and C1. Furthermore, independent evidence for the functional importance of variation at positions 16 and 148 is analysis showing that it was produced by natural selection during evolution of the hominoids (>0.95 by Bayesian posterior probability, L. Abi-Rached, personal communications), the only species that have MHC-C and cognate KIR (40).
A key question to emerge from our study is the extent to which C2 interactions with KIR2DL2 and KIR2DL3 have any influence on the development and response of NK cells. Using an in vitro system involving NKL cells expressing one KIR2DL and target cells expressing single HLA class I allotypes, we demonstrated strong inhibition of cytotoxicity mediated by C2 and KIR2DL2. Moreover, such inhibition was not limited to those C2 allotypes that bound most strongly to KIR2DL2. In contrast, no significant inhibition of cytotoxicity could be attributed to interaction of C2 with KIR2DL3. In this situation we think that the lower avidity of KIR2DL3 for C2 was insufficient to reach the threshold necessary to trigger inhibitory signals. As such, this could represent another functional difference between KIR2DL2 and KIR2DL3, one wherein the differential avidity becomes in effect a difference in specificity. With a more sensitive assay to measure IFN-γ production by peripheral blood NK cells, we found that interaction between C2 and KIR2DL3 mediated a low (∼20%) but statistically significant inhibition, and effects of a similar magnitude were seen for C2 and KIR2DL2. In conclusion, these results suggest that interactions between C2 and both KIR2DL2 and KIR2DL3 could well have physiological roles in the development of the NK cell repertoire and the NK cell response to infection, malignancy, and allogeneic cells.
Phylogenetic comparisons have favored an evolutionary model in which MHC-C allotypes carrying C1 and their cognate inhibitory KIR evolved before C2-bearing allotypes and their cognate KIR (41). Knowing now that KIR2DL2/3 has a weak, broad affinity for C2 and that KIR2DL1 has no affinity for C1, we can explain how the C2 epitope and its cognate KIR evolved under natural selection after the C1 epitope and C1-specific KIR were already in place. The process required that the C2 epitope evolved before the C2-specific receptor, because the C2 epitope could then have been selected for its functional interactions with the preexisting C1-specific KIR. This, in turn, could have set the stage for the selection of novel KIR variants that became increasingly C2-specific. The alternative, that C2-reactive receptors evolved first, is less likely because they would not have been able to use that function in the absence of C2 epitopes. One might speculate that the system of HLA-C ligands and cognate inhibitory KIR has been evolving toward a system of mutually exclusive ligand–receptor interactions. Even if this is true, however, the results presented herein indicate that the system has yet to reach that state, and while KIR2DL1 is impressively C2-specific, KIR2DL2/3 has measurable interactions with both C1 and C2.
We thank Jar-How Lee of One Lambda Inc. for supplying the Luminex reader.
The authors have no financial conflicts of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by National Institutes of Health Grant AI022039 (to P.P.), Stanford Graduate Fellowships (to A.K.M. and M.G.), a National Science Foundation Graduate Fellowship (to A.K.M.), a Howard Hughes Medical Institute Pre-doctoral Fellowship (to M.G.), and a Lymphoma Research Foundation Fellowship (to P.J.N.).
Abbreviations used in this paper: KIR, killer cell Ig-like receptor; NKL, a leukemia-derived cell line with NK cell-like properties; SSP-PCR, sequence-specific polymorphism-PCR.