HLA class I molecules that represent ligands for the inhibitory killer cell Ig-like receptor (KIR) 3DL1 found on NK cells are categorically defined as those HLA-A and HLA-B allotypes containing the Bw4 motif, yet KIR3DL1 demonstrates hierarchical recognition of these HLA-Bw4 ligands. To better understand the molecular basis underpinning differential KIR3DL1 recognition, the HLA-ABw4 family of allotypes were investigated. Transfected human 721.221 cells expressing HLA-A*32:01 strongly inhibited primary human KIR3DL1+ NK cells, whereas HLA-A*24:02 and HLA-A*23:01 displayed intermediate potency and HLA-A*25:01 failed to inhibit activation of KIR3DL1+ NK cells. Structural studies demonstrated that recognition of HLA-A*24:02 by KIR3DL1 used identical contacts as the potent HLA-B*57:01 ligand. Namely, the D1–D2 domains of KIR3DL1 were placed over the α1 helix and α2 helix of the HLA-A*24:02 binding cleft, respectively, whereas the D0 domain contacted the side of the HLA-A*24:02 molecule. Nevertheless, functional analyses showed KIR3DL1 recognition of HLA-A*24:02 was more sensitive to substitutions within the α2 helix of HLA-A*24:02, including residues Ile142 and Lys144. Furthermore, the presence of Thr149 in the α2 helix of HLA-A*25:01 abrogated KIR3DL1+ NK inhibition. Together, these data demonstrate a role for the HLA class I α2 helix in determining the hierarchy of KIR3DL1 ligands. Thus, recognition of HLA class I is dependent on a complex interplay between the peptide repertoire, polymorphisms within and proximal to the Bw4 motif, and the α2 helix. Collectively, the data furthers our understanding of KIR3DL1 ligands and will inform genetic association and immunogenetics studies examining the role of KIR3DL1 in disease settings.
Natural killer cells form an important arm of the innate immune response and have been implicated in responses to viruses such as HIV (1) and cancer (2). The immune surveillance function of NK cells uses a diverse array of receptors, in which the balance of activating and inhibitory signals determine NK cell responsiveness. In what is termed the “missing-self hypothesis,” inhibitory signaling from the binding of killer cell Ig-like receptors (KIR) to HLA class I molecules abrogates NK cell activation, whereas loss of HLA class I leads to cytotoxicity and the release of cytokines and chemokines by the NK cell. Because both the HLA and KIR loci exhibit enormous genetic variation, the presence of alleles encoding specific receptor/ligand pairs has the potential to impact clinical outcomes.
Subsets of HLA-A and HLA-B and all HLA-C allotypes form ligands for particular inhibitory KIR, with signature residues on the HLA class I α1 helix acting as primary specificity determinants. However, within these broad groupings of KIR ligands (C1/C2/Bw4), KIR demonstrate hierarchical interactions, and not all HLA molecules are equally capable of inhibiting NK cell activation (3, 4). For example, the serological Bw4 motif (residues 77–83 on the α1 helix) contains a dimorphic residue at position 80, with HLA-Bw4 molecules containing an isoleucine at position 80 commonly considered to be higher-affinity ligands for KIR3DL1 than those that possess threonine (1, 5). However, as analyses have expanded to cover a broader range of HLA-Bw4 allotypes and the impact of KIR3DL1 polymorphism is better understood, increasing evidence suggests that, at a population level, there exists a complex interaction matrix in which some HLA-Bw4 molecules, such as HLA-B*57:01, are strong ligands regardless of KIR3DL1 allotype, whereas others, such as HLA-A*24:02, are preferentially recognized by particular KIR3DL1 allotypes and some are not recognized at all (3, 6–8).
Among HLA-A allotypes, four contain the Bw4 motif, namely HLA-A*23, HLA-A*24, HLA-A*25, and HLA-A*32. Although HLA-A*32 is efficiently recognized by KIR3DL1, HLA-A*24 and HLA-A*23 show KIR3DL1 allotype-specific differences in recognition, and HLA-A*25 has not been reported to interact with KIR3DL1 (3, 6, 9, 10). Early serological analyses of HLA-A molecules originally defined six broad families, with HLA-A*24:02 and HLA-A*23:01 belonging to the A9 family, HLA-A*25:01 belonging to the A10 family, and HLA-A*32:01 belonging to the A19 family (11, 12). The presence of the Bw4 motif in four HLA-A allotypes across three serologically defined families derived from two ancient lineages is proposed to be the result of interallelic microconversion (13–15). HLA-A*25:01 is considered to have arisen by a single gene conversion event between HLA-A*26:01 and a HLA-Bw4+ locus (16), and differences in the A19 family are believed to have arisen from point mutations or gene conversion with a HLA-B or HLA-C locus (11). The potential importance of HLA-A alleles carrying the Bw4 motif is underscored by their retention in all populations studied to date, reaching high frequencies across Asia, Melanesia, and Oceania (17, 18).
KIR3DL1 was originally identified for its ability to prevent NK cell lysis of target cells expressing HLA-B molecules bearing the Bw4 motif (19–21), and its ability to recognize HLA-ABw4 molecules has received less attention. Indeed, there has not been a consistent approach to the inclusion of HLA-ABw4 molecules in genetic association studies regarding KIR3DL1. This is partly because Bw4 status is often established on the basis of HLA-B allele typing, as seen for early genetic association studies between HLA-Bw4 and KIR3DL1 in the control of HIV (1, 22). Similarly, HLA-ABw4 alleles were not incorporated in studies linking KIR3DL1 with lung transplantation (23) and the risk of acute lymphoblastic leukemia (24), yet both HLA-ABw4 and HLA-BBw4 alleles were considered in studies assessing KIR3DL1 association with outcomes following hematopoietic stem cell transplantation (HSCT) (25). All HLA-ABw4 were actively excluded from studies into KIR3DL1 and HIV control by Boudreau et al. (4), whereas Marra et al. (26) considered HLA-ABw4 alleles to encode high-affinity ligands (apart from HLA-A*25:01) on the basis of possessing Ile80 when assessing relapse following autologous HSCT and KIR3DL1 in acute myeloid leukemia. Notwithstanding the compounding issue of HLA-A molecules containing favorable leader sequences for HLA-E that affords effective NKG2A/CD94–mediated NK cell education and effector functions (unlike most HLA-BBw4 molecules) (27), these studies and others clearly demonstrate that consensus is required over how to categorize HLA-ABw4 alleles in KIR3DL1 association studies.
Although the influence of KIR3DL1 allotype and peptide have been partially explored, particularly with regard to HLA-A*24:02 (8), the underlying explanation for the disparate recognition of the HLA-ABw4 ligands by KIR3DL1 is unknown. Therefore, to better describe the signature residues defining potent, weak, or non- KIR3DL1 ligands, a systematic assessment of the molecular mechanisms underpinning the differential recognition of the HLA-ABw4 ligands by KIR3DL1 was undertaken. In comparison with the archetypal KIR3DL1 ligand HLA-B*57:01, KIR3DL1 recognition of HLA-A*24:02 was more sensitive to changes at KIR3DL1 contact residues. Mutational analyses showed that the inhibition of KIR3DL1+ NK cells was highly influenced by residues present on the α2 helices of HLA-A*24:02 and HLA-A*25:01, but were absent from the “HLA-B–like” α2 helix of HLA-A*32:01. Extending these observations to HLA-BBw4 molecules further emphasized the impact of α2 helix residues, particularly residue 145. Together, these results highlight the importance of the α2 helix in providing a platform for efficient KIR3DL1 docking and suggest that the definition of HLA-encoded ligands for KIR3DL1 should encompass polymorphic residues found on both the α1 and α2 helices.
Materials and Methods
Mutation of HLA class I molecules
Codon-optimized GeneArt Strings encoding HLA-A*23:01, HLA-A*25:01, HLA-A*32:01, and HLA-B*13:01 (Thermo Fisher Scientific), as well as cDNA for HLA-A*11:01, were cloned into pcDNA3.1(−). HLA-B*57:01- and HLA-A*24:02-encoding constructs have been described previously (3, 28). Full-length HLA class I constructs in pcDNA3.1(−) were subjected to site-directed mutagenesis. Reciprocal mutations between HLA-B*57:01 and HLA-A*24:02 were made at residues 95/97, 116, 113/114/116, 144, and 151; reciprocal mutations between HLA-A*24:02 and HLA-A*25:01 were made at residues 90, 149, and 152; and reciprocal mutations between HLA-B*57:01 and HLA-B*13:01 were made at residue 145. Alanine residues were substituted at positions 16, 17, 19, 72, 76, 79, 80, 83, 84, 89, 142, 145, 146, and 151, and glycine at residues were substituted at 149 and 150 in HLA-A*24:02 and HLA-B*57:01 (for all primers see Supplemental Table I).
721.221 cells (221) expressing HLA-B*57:01 and HLA-A*24:02 have been described previously (3). Additional wild-type and mutant HLA class I constructs were electroporated into the HLA-A– and HLA-B–deficient lymphoblastoid cell line 221 at 200 V and 975 μF and selected with 0.5 mg/ml geneticin (Life Technologies), as described previously (28). All cells were cultured in RPMI media with 10% FBS plus supplements. Expression of class I was confirmed by indirect flow cytometry, staining with pan class I (W6/32) or Bw4-specific Ab [RM7.9.63 (29)] supernatants followed by anti-mouse IgG-FITC, and analyzed with FlowJo software.
NK cell purification and KIR3DL1 subtyping
PBMCs from healthy blood donors were used with approval from the University of Melbourne Human Research Ethics Committee. NK cells were purified from PBMC using the EasySep Human NK Cell Enrichment Kit (STEMCELL Technologies). NK cells were either allowed to rest overnight in RPMI media with 10% FBS and 100 U/ml recombinant human IL-2 (Miltenyi Biotec), or where indicated, KIR3DL1+ NK cells were sorted and expanded with irradiated feeder cells in the presence of 100 U/ml recombinant human IL-2 and 1.5 ng/ml PHA (Life Technologies) as previously described (3). Genomic DNA was isolated with the DNeasy Blood and Tissue Kit (QIAGEN), and the KIR3DL1 subtype was initially assigned using a multiplex PCR as previously published (30), followed by further amplification and sequencing with the following previously published primers: 3DL1 exon 3 (D0) F, 3DL1 exon 3 (D0) R, 3DL1 exon 5 (D2) F, 3DL1 exon 5 (D2) R, 3DL1-SEQ-E5F1, and 3DL1-SEQ-E5R (14, 31).
Two hundred and twenty-one transfectants were incubated with purified NK cells at a 1:1 ratio in the presence of anti-CD107a (PE or PE-Cy5; H4A3; BD Pharmingen) and monensin (BD Biosciences). Four (or five where indicated) hours later, cells were stained for KIR3DL1 (anti-NKB1-FITC; DX9; BD Pharmingen) and CD56 (anti-CD56-allophycocyanin; B159; BD Pharmingen) with or without CD3 (anti-CD3; SK7; PerCP/PECy7/allophycocyanin-Cy7; BD Biosciences/BD Pharmingen) and fixed with 2–4% paraformaldehyde. Where indicated, cells were further permeabilized and intracellularly stained for IFN-γ (anti–IFN-γ–AF700, B27; BD Pharmingen). Cells were analyzed by flow cytometry, gating on CD56+, CD3−, and KIR3DL1+ NK cells and assessed using FlowJo software with analysis performed using GraphPad Prism version 8.4.3.
Protein production for structural studies
KIR3DL1*001 and KIR3DL1*005 (residues 1–299) containing a N-terminal hexa-His tag were expressed in HEK 293S GnTI− cells and purified as described previously (3, 32). Briefly, expressed protein in culture supernatants was purified using loose Ni-NTA Agarose Resin (QIAGEN) followed by size-exclusion chromatography using a Sepharose S200 16/60 column in conjunction with an ÄKTA fast protein liquid chromatography (GE Healthcare Life Science). Purified KIR3DL1 protein was treated with Endoglycosidase H (New England Biolabs) prior to crystallization. HLA-A*24:02 H chain and β2-microglobulin (β2m) were produced separately in the BL21(DE3) strain of Escherichia coli as inclusion bodies, refolded in the presence of a molar excess of the HIV Nef peptide RYPLTFGW (RW8), and purified as previously described (33).
Crystallization and structure determination
KIR3DL1*001 or KIR3DL1*005 were concentrated to 10 mg/ml, combined with HLA-A*24:02/RW8 at a 1:1 M ratio, and crystallized at 294 K by the hanging-drop vapor-diffusion method in a solution comprising 14% PEG 3350, 2% Tacsimate *(pH 5.0), and 0.1 M trisodium citrate (pH 5.6). Crystals of sufficient size for diffraction experiments were equilibrated in crystallization solution supplemented with 35% PEG 3350 as a cryoprotectant and flash-frozen in liquid nitrogen at 100 K. X-ray diffraction data were collected at the MX2 beamline (Australian Synchrotron). Data were scaled using MOSFLM and Scala from the CCP4 program suite (34). Details of the data processing statistics are summarized in Supplemental Table II. Both complex structures were determined by molecular replacement as implemented in Phaser (35) using KIR3DL1*001 [Protein Data Bank (PDB) identifier (ID): 3VH8 (32)] or KIR3DL1*005 [PDB ID: 5B38 (3)] and HLA-A*24:02 [PDB ID: 3QZW (36)] with peptide atoms omitted as search models. The KIR3DL1*001 complex contained two noncrystallographic (NCS) copies in the asymmetric unit. The KIR3DL1*005 complex contained one NCS copy in the asymmetric unit. Accordingly, NCS torsion angle restraints were applied to the KIR3DL1*001 complex during refinement. Models were refined through iterative rounds of manual building in Coot (37), refinement in PHENIX (38), and validation with MolProbity (39). Translation, libration, and screw restraints were determined using the automated model-partitioning tool phenix.find_tls_groups and applied during refinement as implemented in PHENIX. Refinement statistics are summarized in Supplemental Table II. Visualizations of structural data were generated using PyMOL v2.3.4 (Schrödinger). Coordinates, structure factors, and validation reports are available from the PDB (https://www.rcsb.org/) under accession codes 7K80 and 7K81, respectively.
HLA-A*32:01, HLA-A*24:02, and HLA-A*23:01 are bona fide KIR3DL1 ligands with varying potency
KIR3DL1 molecules recognize HLA class I allotypes that possess the Bw4 motif (21). Although this motif spans residues 77 and 83 of the α1 helix, the KIR3DL1/HLA class I binding interface is much broader and includes residues within both the HLA class I α1 and α2 domains (Fig. 1) (32). To ascertain whether HLA-ABw4 molecules deviate from the archetypal HLA-B*57:01 ligand in these regions, the protein sequence of the α1 and α2 domains of HLA-B*57:01 was aligned against those of four HLA-ABw4 molecules (Fig. 1A). Each HLA-ABw4 molecule contained identical residues at the majority of established KIR3DL1 contact sites identified on HLA-B*57:01 as determined by x-ray crystallography (32, 40) (Fig. 1A, red): namely, Gly16, Arg17, Gly18, and Glu89 that contact the KIR3DL1 D0 domain; Gln72, Glu76, Arg79, Ile80, Arg83, and Tyr84 on the α1 helix that define KIR3DL1 specificity through the D1 domain; and residues Ile142, Arg145, Lys146, and Ala150 on the α2 helix that interact with the KIR3DL1 D2 domain (32). The notable exceptions to these were the presence of His151 in HLA-A*24:02 and HLA-A*25:01 and Thr149 in HLA-A*25:01. Most of the remaining differences between HLA-ABw4 molecules and HLA-B*57:01 were located on the sides and the floor of the peptide binding cleft (Fig. 1A, 1B).
To first assess the capacity of HLA-ABw4 molecules to act as ligands for KIR3DL1, the HLA-A– and HLA-B–deficient cell line 221 was transfected with plasmids encoding HLA-A*23:01, HLA-A*24:02, HLA-A*25:01, or HLA-A*32:01. Each of the transfected HLA-ABw4 molecules was expressed similarly at the cell surface, as assessed by flow cytometry staining with the pan class I Ab W6/32 (Fig. 2A). Staining with a Bw4-specific mAb RM7.9.63 (29) also showed comparable levels of staining across cells expressing each HLA-A allotype, with the exception of HLA-A*25:01, as previously reported (Fig. 2A) (6), suggesting that an allele-specific polymorphism impacts the conformation of the Bw4 epitope.
Next, the ability of each HLA-ABw4 molecule to inhibit KIR3DL1+ NK cells was examined and compared with target cells expressing the potent ligand, HLA-B*57:01 (3, 4). Upon culture with transfected target cells, KIR3DL1− NK cells exhibited similar patterns of degranulation (as detected by the expression of CD107a) and IFN-γ production, irrespective of HLA-Bw4 expression on target cells (Fig. 2B, lower panel). Incubation of KIR3DL1+ NK cells with the 221 parental cell line resulted in both degranulation and IFN-γ production, whereas the expression of HLA-B*57:01 on 221 cells efficiently inhibited their activation (Fig. 2B–D). Target cells expressing HLA-A*32:01 similarly inhibited CD107a and IFN-γ expression from KIR3DL1+ NK cells (Fig. 2B–D). This inhibition was primarily mediated by KIR3DL1, as addition of the KIR3DL1 blocking Ab DX9 resulted in efficient degranulation of the NK cells toward 221.B5701 and 221.A3201 target cells (Supplemental Fig. 1).
Relative to HLA-B*57:01, HLA-A*24:02 demonstrated a weaker capacity to inhibit the activation of KIR3DL1+ NK cells, which varied markedly across donors as previously reported (3). Specifically, NK cells expressing KIR3DL1*005 allotypes (gray symbols) were more inhibited by the expression of HLA-A*24:02 than other KIR3DL1 allotypes examined in this study (Fig. 2C, 2D), which is consistent with the broader HLA-Bw4 specificity of KIR3DL1*005 (3, 7, 8). A similar situation was seen for the 221 cell line expressing HLA-A*23:01, although this HLA allotype had a reduced capacity relative to HLA-A*24:02 to inhibit KIR3DL1+ NK cell degranulation and IFN-γ production. Again, the increased activation observed following the addition of DX9 demonstrated 221.A2402-associated inhibition to be mediated by KIR3DL1, with a similar but NS trend seen for 221.A2301 targets (Supplemental Fig. 1). Unlike other HLA-ABw4 molecules, HLA-A*25:01 did NS inhibit KIR3DL1+ NK cell degranulation or IFN-γ expression, even by KIR3DL1*005+ donors (Fig. 2C, 2D). Blocking with DX9 also had no effect on the degranulation response of KIR3DL1+ NK cells (Supplemental Fig. 1), supporting previous observations that despite possessing the Bw4 motif, HLA-A*25:01 does not function as a KIR3DL1 ligand (6, 9). Together, these observations confirm previous reports that HLA-A*24:02 and HLA-A*32:01 and to a lesser extent HLA-A*23:01 but not HLA-A*25:01 are bona fide KIR3DL1 ligands that vary in potency (6, 9).
KIR3DL1 recognition of HLA-A*24:02 is highly sensitive to variation across HLA contact residues
The relative importance of each KIR3DL1 contact residue in HLA-A*24:02 for recognition was next compared with HLA-B*57:01. Single-site alanine mutations (or glycine where an alanine was present in the wild-type HLA class I) were introduced into both HLA-A*24:02 and HLA-B*57:01 at residues shown to directly contact KIR3DL1 (32) along with position 19, a dimorphic residue located within a conserved loop that interacts with the D0 domain, and the constructs expressed in 221 cells. The 32 mutant HLA molecules were expressed at similar levels on the cell surface as detected with W6/32, whereas Glu89Ala and Arg83Ala showed reduced staining with anti-Bw4, indicating that these mutations perturbed the Bw4 epitope (Supplemental Fig. 2). KIR3DL1+ NK cells were expanded from three donors who possessed KIR3DL1*005, an allotype that robustly interacts with HLA-A*24:02. These KIR3DL1*005+ NK cells were then incubated with the 221 cells expressing the alanine mutants and degranulation (CD107a expression) was compared with that observed following culture with 221 cells expressing the wild-type HLA class I molecules.
The majority of substitutions in the context of HLA-B*57:01 had little impact on the extent of inhibition observed, with only the Tyr84Ala, Lys146Ala, and Ala149Gly mutations abrogating recognition (Fig. 3A). In contrast, the introduction of alanine mutations into HLA-A*24:02 had a much more marked impact on its capacity to inhibit NK cell activation (Fig. 3B). KIR3DL1+ NK inhibition was sensitive to multiple residues across the α1 helix of HLA-A*24:02, including Glu76Ala, Ile80Ala, Arg83Ala, and Tyr84Ala, and most notably for residues Lys146Ala and Ala149Gly on the α2 helix, as for HLA-B*57:01. In the context of HLA-A*24:02, the Ile142Ala mutation resulted in activation levels comparable with that observed with the Tyr84Ala and Ala149Gly mutations, yet in the context of HLA-B*57:01, the same mutant was indistinguishable from wild-type HLA-B*57:01. These data suggest that, although HLA-A*24:02 possesses all of the required residues to interact with KIR3DL1, the overall interface is more easily perturbed by mutation, with unique dependencies on the α2 helix.
HLA-A–specific residues on the α2 helix impact KIR3DL1 interaction
HLA-A*24 uniquely contains a lysine at position 144 among HLA-Bw4+ allotypes, whereas His151 is also found in HLA-A*25 and some HLA-A*32 family members (Fig. 1A). Because these two residues are located within the KIR3DL1 D2 docking region on the α2 helix (Fig. 1B), reciprocal mutations between HLA-A*24:02 and HLA-B*57:01 were generated at positions 144 and 151, and the resulting mutants expressed in 221 cells. Surface staining of these cell lines with W6/32 showed the wild-type and mutated class I molecules to be expressed at equivalent levels (Supplemental Fig. 2). When incubated with purified NK cells, 221.B5701, 221.B57 Q144K, and 221.B57 R151H each effectively inhibited activation of KIR3DL1+ NK cells (Fig. 4A). In contrast, whereas 221.A2402 and 221.A24 H151R weakly inhibited KIR3DL1+ NK cell degranulation, the Lys144Gln mutation improved the inhibitory capacity of HLA-A*24:02. Combined with the impact of the alanine substitution at residue Ile142 (Fig. 3B), these results further point to residues within the α2 helix impacting KIR3DL1 recognition.
Like HLA-A*24:02, HLA-A*25:01 contains unique residues within the α2 helix as well as possessing Asp90, which lies in close proximity to the KIR3DL1 contact Glu89 (Fig. 1). To ascertain whether these polymorphisms explained the apparent inability of HLA-A*25:01 to function as a KIR3DL1 ligand, reciprocal mutations at positions 90, 149, and 152 were made between HLA-A*25:01 and HLA-A*24:02 and expressed in 221 cells (Supplemental Fig. 2). Although Asp90 in HLA-A*25:01 had been previously hypothesized to impair KIR3DL1 recognition (9), substitution of alanine for an aspartic acid at position 90 in HLA-A*24:02 did not appreciably impact the ability of HLA-A*24:02 to inhibit KIR3DL1+ NK cell degranulation (Fig. 4B). The corresponding Asp90Ala substitution in HLA-A*25:01 also failed to alter its inhibitory potential, despite the substitution of this residue impacting the reactivity with the RM7.9.63 anti-Bw4 but not W6/32 Abs (Supplemental Fig. 2).
Residue 152 sits on the α2 helix near the KIR3DL1 D2 domain contact region (Fig. 1B), with both glutamate and valine residues found among HLA-Bw4 molecules. The Glu152Val substitution in HLA-A*25:01, however, had no impact on the extent of inhibition of KIR3DL1+ NK cells (Fig. 4B). The introduction of a Val152Glu substitution into HLA-A*24:02 resulted in an intermediate phenotype, with 221.A24 V152E attenuating recognition of HLA-A*24:02 by KIR3DL1+ NK cells. Strikingly, mutation of Ala149 to threonine in HLA-A*24:02 completely abrogated its capacity to inhibit activation of KIR3DL1+ NK cells, whereas the reciprocal Thr149Ala mutation conferred inhibitory potential on HLA-A*25:01, with the mutant inhibiting KIR3DL1+ NK cells as efficiently as HLA-B*57:01 (Fig. 4B). Substitution of Val152 into HLA-A*25:01 may have had the potential to impact binding but was unable to overcome the detrimental effect of Thr149. Thus, the poor recognition of HLA-A*25:01 by KIR3DL1 is likely due to Thr149, a polymorphism found in other HLA-A molecules but unique to HLA-A*25:01 among those that possess the Bw4 motif.
Recognition of HLA-BBw4 molecules is sensitive to α2 helix polymorphisms
To determine whether KIR3DL1 was also sensitive to unique polymorphisms within the α2 helix of HLA-BBw4 molecules, Leu145 of the poor ligand HLA-B*13:01 was also investigated. KIR3DL1*001 forms direct contacts with Arg145 of HLA-B*57:01, including a salt bridge to KIR3DL1 residue Asp230 and hydrogen bonds to Ser228 (32). Reciprocal mutations at position 145 between HLA-B*57:01 and HLA-B*13:01 were therefore generated and expressed in 221 cells (Fig. 5A). Although wild-type HLA-B*13:01 was expressed at a slightly reduced level compared with HLA-B*57:01, the Leu145Arg mutation resulted in improved surface expression of HLA-B*13:01, perhaps stabilizing the molecule. Recognition of HLA-B*13:01 containing a Leu145Arg mutation also successfully imparted inhibition by this usually poor ligand, whereas the Arg145Leu mutation in HLA-B*57:01 compromised its ability to inhibit KIR3DL1+ NK cells (Fig. 5B). These results further indicate that the D2 docking platform on the α2 helix is critical for KIR3DL1 recognition of HLA-Bw4 ligands of both HLA-A and HLA-B groups.
Analogous contacts enable KIR3DL1 to dock onto HLA-A*24:02
To determine whether KIR3DL1 docking onto HLA-ABw4 molecules differed from HLA-B*57:01, the crystal structures of KIR3DL1*001 and KIR3DL1*005 in complex with HLA-A*24:02 presenting the KIR3DL1-permissive Nef peptide RYPLTFGW (RW8) (41) were solved to 2.4 and 2.0 Å resolution, respectively (Fig. 6, Supplemental Table II). Comparison with the previously solved structures of KIR3DL1*001 and KIR3DL1*005 in complex with HLA-B*57:01/LF9 (32), both of which were crystallized in the same space group (P1) and to similar resolution (2.0–2.4 Å), showed KIR3DL1 bound HLA-A*24:02 in a similar manner as HLA-B*57:01 (Fig. 6A). The D0 domain engaged the side of the peptide binding groove proximal to β2m, whereas the D1 and D2 domains bound over the α1 and α2 helices, respectively. Polymorphic differences between KIR3DL1*001 and KIR3DL1*005 are known to produce distinct angles of binding across HLA-B*57:01, whereby KIR3DL1*005 engaged its ligand with a more obtuse D1–D2 binding angle that was associated with increased domain dynamics at the D1–D2 hinge (3). In complex with HLA-A*24:02, similar subtle differences in D1 and D2 domain angles were observed between KIR3DL1*001 and KIR3DL1*005. Notably, the KIR3DL1*001 in complex with HLA-A*24:02 more closely resembled the conformation in complex with HLA-B*57:01 (root-mean-square deviation of 0.8 Å) compared with the equivalent complexes with KIR3DL1*005 (root-mean-square deviation of 1.4 Å) (Fig. 6A). Nevertheless, all the key contacts are conserved between HLA-A*24:02 and HLA-B*57:01 to either KIR3DL1*001 or KIR3DL1*005. Thus, the subtle structural differences between the KIR3DL1*001 and KIR3DL1*005 complexes are likely due to the KIR3DL1 polymorphisms, as previously reported (3, 7). This further demonstrates that the dynamics of KIR3DL1*005 may account for the permissiveness of this allele to engage more effectively to a more diverse range of Bw4 containing HLA allomorphs and peptide repertoires. In addition, the consistency of the KIR3DL1/HLA contacts allowed for robust analysis of residues of interest on the α2 helix of the HLA.
First, looking at differences on the α2 helix between HLA-A*24:02 and HLA-B*57:01, despite affecting NK activation, Lys144 of HLA-A*24:02 did not directly contact either KIR3DL1*001 or KIR3DL1*005 (Fig. 6Bi). Lys144 was located on the underside of the α2 helix, where it bound the Trp133 side chain and backbone atoms of Trp133–Ala135 that make up the connecting loop between the α2 helix and the binding groove floor. HLA-B*57:01 possesses Gln144 at this position, which makes polar contacts to the connecting loop 133–135 (Fig. 6Bii). Thus, the effect of the Lys144Gln mutation in HLA-A*24:02 on KIR3DL1 recognition may have been attributable to other factors, such as the overall positioning, stability, or flexibility of the α2 helix. In contrast, His151 of HLA-A*24:02 forms a salt bridge with Glu201 of both KIR3DL1*001 and KIR3DL1*005 (Fig. 6Bi) that is maintained in HLA-B*57:01, which possesses an arginine at position 151 and is consistent with the observation that reciprocal mutations between histidine and arginine at position 151 had little effect on KIR3DL1 recognition (Fig. 4A).
A defining feature of the nonpermissive HLA-A*25:01 on the α2 helix are the Thr149 and Glu152 residues. In both the KIR3DL1*001- and KIR3DL1*005-HLA-A*24:02/RW8 structures, Ala149 was central to the binding interface, making close interactions with the KIR3DL1 residues Ser227 and Phe276 (Fig. 6C). Because of the close contacts made across multiple KIR3DL1 binding loops, including the D1–D2 interdomain loop that is critical to the interaction (32), the polar side chain of Thr149 as found in HLA-A*25:01 would likely not be spatially accommodated by either KIR3DL1*001 or KIR3DL1*005. In contrast, Val152 of HLA-A*24:02 abuts the E pocket in which P6-Phe of RW8 is located but did not contact KIR3DL1*001 or KIR3DL1*005 (Fig. 6D). Thus, substitution of Glu152 into HLA-A*24:02 is likely to modulate peptide repertoire selection about the E pocket, which may then impact the key peptide position PΩ-2.
Involvement of peptide in KIR3DL1 engagement of HLA-A*24:02
KIR3DL1 molecules exhibit significant peptide selectivity in binding their HLA-Bw4 ligands (8, 42, 43). The structural data provided direct insights into the contribution of peptide to KIR3DL1 recognition in the setting of an HLA-A allomorph. Neither KIR3DL1*001 nor KIR3DL1*005 made direct contact with the RW8 peptide displayed by HLA-A*24:02, yet KIR3DL1*005 made a number of water-mediated contacts (Fig. 7A). This contrasted with recognition of the HLA-B*57:01/LF9 complex, in which KIR3DL1*001 Leu166 directly contacted Ser8 at PΩ-1 of the LF9 peptide, which was accompanied by an additional contact in KIR3DL1*005 with Met165 in binding to Thr6 at PΩ-3 (3, 32) (Fig. 7B). For HLA-A*24:02/RW8, Met165 and Leu166 in both KIR3DL1*001 and KIR3DL1*005 ternary complexes were positioned >5.6 Å from peptide atoms, owing to the presence of Gly7 at PΩ-1 and the surface of the shorter 8-mer RW8 peptide sitting relatively flush within the cleft. This flat peptide surface and the concomitant lack of contacts to KIR3DL1 may in fact facilitate binding in the case of RW8. Indeed, previous analyses of HLA-B*57:01 and HLA-B*57:03 presenting identical peptides showed that differences in their pocket structure resulted in surface exposure of the PΩ-2 side chain in HLA-B*57:03, which sterically impacted KIR3DL1 binding (44). Like HLA-B*57:03, HLA-A*24:02 possesses a tyrosine at residue 116, in comparison with the serine of HLA-B*57:01, raising the potential that bulky residues within the peptide binding cleft lead to increased exposure of the PΩ-2 side chain that may create more challenging peptide landscapes for KIR3DL1 docking.
To further explore whether the characteristics of the peptide binding pockets contributed to the distinct patterns of KIR3DL1 recognition of HLA-A*24:02 and HLA-B*57:01, a series of reciprocal mutations were therefore introduced into the peptide binding clefts. Positions 95 (in the F pocket) and 97 (C and E pocket), residue 116 in isolation (F pocket), and residues 113, 114, and 116 (D, E, and F pockets) in concert were all exchanged between the two HLA molecules. Following transfection of each construct into 221 cells, the mutant HLA class I molecules were expressed at similar levels, as assessed by flow cytometry (Supplemental Fig. 2).
The introduction of the Ser116Tyr and His113Tyr/Asp114His/Ser116Tyr mutations into HLA-B*57:01 did NS alter degranulation of KIR3DL1+ NK cells, yet a marginal diminution of inhibition was observed with cells expressing an Ile95Leu/Val97Met mutant of HLA-B*57:01 (Fig. 7C). Similarly, the introduction of reciprocal mutations into HLA-A*24:02 was assessed. In this case, mutant HLA-A*24:02 molecules carrying the pocket residues of HLA-B*57:01 (Leu95Ile/Met97Val, Tyr116Ser, and Tyr113His/His114Asp/Tyr116Ser) each had a slightly increased capacity to inhibit the activation of KIR3DL1+ NK cells, although this was only significant for 221.A24 Y116S and 221.A24 I95L/V97M. Thus, mutations in the peptide binding cleft did impact KIR3DL1 recognition of both HLA-A*24:02 and HLA-B*57:01, suggesting that differences in peptide repertoire or variation in peptide conformation, as a factor of HLA class I architecture, may contribute to the differential KIR3DL1 recognition of these two HLA-Bw4 allotypes.
Although it is established that all KIR3DL1 ligands carry the Bw4 motif, be they HLA-A or HLA-B allotypes, the extent to which each effectively mediates inhibition has been unclear. This is of particular note with regards to HLA-ABw4 molecules that have been variably incorporated into genetic association or immunogenetic studies examining linkages between KIR3DL1 and disease outcomes, as well as the persistent inclusion of common allotypes, such as HLA-B*13 as KIR3DL1 ligands, despite no reported interaction with KIR3DL1 to date. Although the primary specificity determinant of KIR3DL1 has been the presence of the Bw4 epitope, analyses of HLA-ABw4 molecule recognition by KIR3DL1 highlighted the considerable sensitivity of the interaction to polymorphic residues within the HLA class I α2 helix. Thus, differential recognition across HLA-Bw4 allotypes is therefore likely a product of polymorphic residues in both the α1 and α2 domains, differences in peptide repertoire and conformation, as well as KIR3DL1 allotype.
The HLA-ABw4 family provided a unique setting within which to investigate the contribution of the α2 helix to KIR3DL1 interactions, containing examples of Bw4 allotypes that were strong or weak KIR3DL1 ligands. Most remarkable was that HLA-A*25:01 failed to function as a KIR3DL1 ligand because of the presence of Thr149. This residue along with Glu152 are characteristic of both the A2 and A10 serotypes (16), within which only HLA-A*25 carries the Bw4 epitope. Lys144 is present in a number of HLA-A allotypes, yet among HLA-ABw4 members, it is unique to HLA-A*24, with KIR3DL1 sensitive to its mutation. The other prominent HLA-ABw4 allotype is HLA-A*32:01, which belongs to the A19 family that is reportedly distinguished by the presence of Gln144 and Arg151 in the α2 helix (11), two residues also found in HLA-B*57:01. Mutagenesis of these residues that define allotypic HLA class I families showed that Ile142, Ala149, and to a lesser extent Gln144 and Val152 were all important for maintaining an accommodating KIR3DL1 docking platform on the α2 helix. HLA-A*23:01 inhibits as weakly as HLA-A*24:02 and yet does not encode the cogent Lys144 or His151 residues on the α2 helix, perhaps further signifying the potential for peptide involvement. Nevertheless, distinct HLA-A–encoded α2 helix residues in HLA-A*24:02 and HLA-A*25:01 each drove the suboptimal binding of KIR3DL1, contrasting with HLA-A*32, which, while possessing an α1 helix closely related to HLA-A*24, has a more HLA-B–like α2 helix (45).
Structural analyses of KIR3DL1 binding to HLA-A*24:02 showed that the receptor docked in a very similar way and used the same contact residues as for the potent HLA-B*57:01 ligand. Alanine-scanning mutagenesis of these residues, however, revealed functional recognition to be more sensitive to changes in HLA-A*24:02 than HLA-B*57:01, suggesting that the interaction with HLA-A*24:02 was generally weaker, consistent with HLA-binding data (3). One explanation for this more finely balanced recognition of HLA-A*24:02 might be that the contribution of peptide to the interaction with KIR3DL1 differed from that of HLA-B*57:01. Such effects could be mediated by differences in the composition of the peptide repertoire because variation in the sequence of the HLA-bound peptide, particularly at positions 7 and 8, have been shown to impact KIR3DL1 recognition (42, 46). Similarly, differences in the pocket architecture may impact the exposure of P7 which has been shown to influence recognition by KIR3DL1 (44). In this study, structural analyses showed that the RW8 peptide presented by HLA-A*24:02 failed to make direct contacts with KIR3DL1, contrasting with previously determined structures of KIR3DL1 complexed with HLA-B*57:01/LF9 (32). However the extent to which this is true for other HLA-A24/peptide complexes remains unclear. Previous studies have demonstrated that altering the peptide binding cleft affected KIR3DL1 recognition, presumably by impacting the presented peptide repertoire (43, 47), and indeed, mutation to the peptide binding clefts of HLA-A*24:02 and HLA-B*57:01 had some impact on KIR3DL1 recognition. These data suggests that variation in the nature of the peptide repertoire or the way peptide is displayed may contribute to the differential KIR3DL1 recognition of HLA-A*24:02 and HLA-B*57:01. Nevertheless, further analyses are required to understand the extent to which differences in the global peptide repertoires present on the cell surface impact target cell recognition by KIR3DL1+ NK cells.
Previous binding and functional analyses have shown that recognition of HLA-Bw4 allotypes is impacted by allotypic variation within KIR3DL1 (3, 48). Structural and biophysical analyses have linked the broader HLA-Bw4 specificity and peptide tolerance of KIR3DL1*005 with an altered D1–D2 angle and increased flexibility of the ligand binding regions owing to the presence of Leu283 (3, 7). Consistent with this, recognition of HLA-B*57:01 by NK cells expressing KIR3DL1*005 was largely unaffected by numerous mutations in the α1 domain of HLA-B*57:01, including positions 76, 80, 83, and 89, as well as Ile142 on the α2 helix, unlike binding analyses between KIR3DL1*001 and HLA-B*57:01 (32). This broader tolerance, however, did not extend to key residues within the α2 docking platform such as Lys146 and Ala149 or even Tyr84 within the α1 helix. KIR3DL1*005 also displayed stronger recognition of both HLA-A*24:02 and HLA-A*23:01 (data in this study and Refs. 3 and 6); yet, although a difference in the contact residues was not apparent in the structural data for HLA-A*24:02, heightened plasticity, as noted previously (3), may compensate for the suboptimal docking platform.
The identification of key residues within the α2 helix of HLA-ABw4 for recognition by KIR3DL1 also provided broader insight into recognition of HLA-BBw4 ligands. In particular, the presence of a leucine at position 145 largely accounted for the poor KIR3DL1 recognition of HLA-B*13:01. Indeed, recognition of HLA-BBw4 molecules has previously been shown to be functionally sensitive to mutations to the Bw4 motif itself, as well as residues of the peptide binding pocket, and even more distant residues, including residues 67 and 194 in given allotypes (28, 43, 47). Given the intermediate phenotype induced by HLA-A*24:02 carrying a Val152Glu mutation and that the HLA-Bw4 binding specificity of the evolutionary intermediate, orangutan Popy2DLA, was found to strongly correlate with the combined presence of Ile80 and Val152 (49), the impact of this dimorphic residue among HLA-BBw4 on KIR3DL1 binding also warrants further analysis. Val152 is frequently found among the strongest ligands for KIR3DL1*001 and KIR3DL1*015, including HLA-B*57:01/03, HLA-B*58:01, HLA-B*53:01, and HLA-B*38:01, with HLA-B*49:01 being the notable exception (3). Whether such polymorphism within the α2 helix could similarly impart differential recognition of HLA class I ligands by other KIR remains open to investigation.
The data in this study have implications for correlative studies examining the relationship between the presence of KIR3DL1 and its ligands with clinical outcomes. Foley et al. (6) have previously estimated that if HLA-A*32 and HLA-A*24 but not HLA-B*13:01 were considered in defining an individual’s HLA-Bw4 status (as opposed to only HLA-BBw4 alleles) up to 15% of people would need reclassification, impacting donor selection for haploidentical HSCT. These impacts may be even more keenly felt in populations in which HLA-ABw4 molecules significantly contribute to the KIR3DL1 ligand pool. Indeed, HLA-A*24:02 is highly prevalent in Oceanic and Southeast Asian regions, and at 37%, forms the most prominent HLA-A allele and KIR3DL1 ligand in the Māori population (18). HLA-A*25:01 appears at much lower frequencies worldwide than HLA-A*24:02 and is primarily found in Europe, whereas HLA-A*23:01 is associated with African populations (50). Interestingly haplotypes containing both HLA-BBw4+ and HLA-ABw4+ alleles are rare. Understanding the contribution of HLA-ABw4 alleles to clinical outcomes is also complicated by their carriage of Met−21, which unlike the Thr−21 found in most HLA-BBw4 alleles, creates peptides that bind well to HLA-E (27). Consequently, HLA-ABw4 molecules are able to mediate NK cell education and effector functions more readily through NKG2A/CD94 than HLA-BBw4 molecules (27). Indeed, high expression of HLA-A in HIV+ individuals was associated with poorer control, presumably owing to enhanced NKG2A-mediated inhibition (51). Whereas these observations may suggest there is reason to exclude HLA-ABw4 molecules from association studies with KIR3DL1 [and potentially the HLA-BBw4 allele HLA-B*38:01 which also possesses Met−21 (27)], it conversely would not account for the impact of alleles, such as HLA-A*32:01, that are strong KIR3DL1 ligands. Indeed, exclusion of HLA-Bw4 allotypes that are nonfunctional KIR3DL1 ligands, such as HLA-A*25:01 and HLA-B*13:01, may allow for a more robust understanding of the impacts of HLA–KIR pairings in disease and transplantation scenarios.
We thank the Melbourne Cytometry Platform (Doherty Institute node) for provision of flow cytometry services. This research was undertaken, in part, on the MX2 beamline at the Australian Synchrotron, part of Australian Nuclear Science and Technology Organisation.
This work was supported by a program grant from the National Health and Medical Research Council of Australia (1113293) and the Australian Research Council (ARC) (CE140100011); J.R. is supported by an Australian ARC Laureate Fellowship.
The coordinates, structure factors, and validation reports presented in this article have been submitted to the Protein Data Bank (https://www.rcsb.org/) under accession numbers 7K80 and 7K81.
The online version of this article contains supplemental material.
Abbreviations used in this article:
hematopoietic stem cell transplantation
killer cell Ig-like receptor
Protein Data Bank.
The authors have no financial conflicts of interest.