KIR3DS1-Specific D0 Domain Polymorphisms Disrupt KIR3DL1 Surface Expression and HLA Binding

Natural killer cells are lymphocytes of the innate immune system that are important for the host immune defense against viral pathogens and malignant cells. NK cells are capable of cytotoxic function and can also secrete cytokines that stimulate the adaptive immune system (1, 2). In contrast to T and B lymphocytes, NK cells do not express Ag-specific receptors, but rather express a variety of activating and inhibitory receptors, including the killer Ig–like receptors (KIRs). Activation of NK cells is modulated by a balance of signaling events induced by these receptors at the immune synapse (3). Many of the inhibitory receptors, including KIR, recognize HLA as a marker of “self” to prevent NK cell–mediated autoreactivity (4, 5). Because virally infected cells or malignant cells downregulate HLA on the cell surface to evade the adaptive immune system, these cells become more susceptible to NK cell–mediated lysis (6, 7).

The polymorphic KIR gene family is located on human chromosome 19q13.4 and encodes both inhibitory and stimulatory receptors that are stochastically expressed by NK and T cells. The KIR genes encode type I integral membrane proteins that contain either three (KIR3D) or two extracellular (KIR2D) Ig-like domains. In general, KIRs with a long cytoplasmic tail (KIR3DL and KIR2DL) containing two ITIMs function as inhibitory receptors; upon recognition of self-HLA molecules (8–10), the phosphatases SHP1 and SHP2 are recruited to the ITIMs, resulting in the induction of inhibitory signaling (11, 12). In addition to inhibition, the interaction between KIR and HLA is vital to determining the activation potential of an NK cell. Through a mechanism referred to as “licensing” or “education,” NK cells that express an inhibitory KIR for a cognate self-HLA molecule display a lower threshold for activation and a greater functional response to target cells (13, 14).

Nearly all individuals are positive for the KIR3DL1/S1 gene (15, 16). KIR3DL1 encodes an inhibitory receptor that binds to HLA-A and HLA-B molecules that carry the Bw4 epitope designated by aa residues 77 and 80–83 in the α-1 helix of HLA (17–19). Polymorphisms in the extracellular domains between KIR3DL1 and KIR3DS1 are relatively limited, yet KIR3DS1 demonstrates a unique surface expression profile compared with KIR3DL1 and, with the exception of the rare KIR3DS1*014, is unable to bind HLA-Bw4 (20). The KIR3DL1 alleles have developed extensive diversification through gene recombination events and point mutations. KIR3DL1*009, which is found in 2% of the general population, is a product of a gene conversion event involving the common inhibitory allele KIR3DL1*001 and a 1.5-kb sequence of the common activating allele KIR3DS1*01301, which contains the exons that encode the D0 extracellular domain (21, 22). Resulting from this conversion, KIR3DL1*009 is identical to KIR3DL1*001 both upstream and downstream of this KIR3DS1*01301 insert. KIR3DL1*009 presents a unique opportunity to determine the critical polymorphic residues within the D0 domain that affect ligand recognition and the surface phenotype of KIR3DL1/S1. We demonstrate that the combined polymorphisms at amino positions 58 and 92 within the D0 domain of KIR3DL1*009 result in reduced surface expression and ligand binding compared with KIR3DL1*001. We further show that these polymorphic residues impact other KIR3DL1 alleles, because KIR3DL1*042, which resulted from a conversion between KIR3DS1*013 and KIR3DL1*005, also demonstrates reduced binding to HLA. These findings provide further molecular mechanisms that can explain the lack of reactivity between KIR3DS1*013 and HLA-Bw4 and contribute new data that can help understand the impact of the KIR–HLA interaction on clinical outcomes.

Primary PBMCs were collected anonymously from buffy coats obtained from the New York Blood Center (New York, NY), where donors provided written, informed consent. Additional consent from these donors was waived by the Memorial Sloan Kettering Cancer Center (MSKCC) institutional review board. All PBMCs were cryopreserved before experimental analysis. PBMCs were thawed and cultured overnight in RPMI 1640 supplemented with 250 U/ml IL-2. The parental, HLA-A, -B, -C negative 721.221 cell line as well as 721.221 cells stably expressing HLA-B*44:03 were kind gifts from Dr. Carolyn K. Hurley (Georgetown University, Washington, DC) and were maintained in RPMI 1640. The human embryonic kidney cell line, HEK293T, was maintained in DMEM. All media used to maintain the primary cells and cell lines described earlier were supplemented with 10% FBS, 1 mM l-glutamine, 10 mM HEPES buffer, and 1 mM sodium pyruvate. The Expi293F cell line was cultured in the proprietary Expi293 Expression medium as recommended by the manufacturer (Life Technologies, Carlsbad, CA).

The cDNA encoding KIR3DL1*001 was a kind gift of Dr. Daniel McVicar (National Cancer Institute, Bethesda, MD). The KIR3DL1*001 cDNA was cloned in the pcDNA3-Clover plasmid (a kind gift from Dr. Michael Lin, Addgene plasmid 40259) upstream and in-frame with the Clover cDNA (23). The stop codon was mutated to create a C-terminally tagged full-length KIR molecule. Constructs encoding KIR3DL1*005, KIR3DL1*009, and KIR3DL1*001 mutants were created via site-directed mutagenesis as previously described with the modification of using high-fidelity Platinum Taq (Life Technologies) and associated buffers (24). To generate KIR-Fc constructs, we cloned cDNA consisting of the IL-2 signal peptide, exons encoding the extracellular domains of KIR3DL1*001, and the Fc region of human IgG1 into the mammalian expression vector pCDNA3.4. All other KIR-Fc constructs were created via site-directed mutagenesis. All constructs were prepared as per manufacturer’s instructions using the HiSpeed Plasmid Maxi Kit (Qiagen, Valencia, CA). The cDNA sequences were confirmed by the MSKCC DNA Sequencing Core Facility (New York, NY).

DNA from the New York Blood Center donors was isolated using the Qiagen DNeasy Blood and Tissue kit. Donors were first screened using PCR methods previously described to identify donors potentially positive for KIR3DL1*009 (21, 25, 26). DNA samples from donors identified as positive for KIR3DL1*009 by PCR were subsequently evaluated by sequencing as previously described (27).

Receptor expression was detected on the surface of primary NK cells (CD56⁺, CD3⁻) after staining of PBMCs with the KIR3DL1-specific Abs, DX9 (BD Biosciences, Franklin Lakes, NJ) and Z27 (Beckman Coulter, Brea, CA), conjugated to PE and allophycocyanin, respectively. To detect KIR3DL1 surface expression on HEK293T cells, we transiently transfected 2 × 10⁵ cells with 1.0 μg plasmids encoding the full-length cDNA of KIR3DL1 alleles under the control of a CMV promoter using the XtremeGENE HP DNA transfection reagent (Roche, Nutley, NJ). Forty-eight hours posttransfection, KIR3DL1 surface expression was analyzed on viable, Clover⁺ cells using the KIR3DL1-specific Ab, DX9. Surface expression data are presented as a ratio of KIR3DL1 surface expression (DX9 mean fluorescent intensity [MFI]) to total KIR3DL1 expression (Clover MFI). Flow-cytometry experiments were assessed using an LSR Fortessa (BD Biosciences) and analyzed using FlowJo software version 9.7.6 (Tree Star, Ashland, OR).

The internalization of KIR3DL1 on transfected HEK293T cells was analyzed as previously described (28). In brief, 16 h posttransfection, the transfectants were probed with PE-conjugated DX9 and cultured at 37°C for 0, 20, 60, or 120 min. At each time point, samples were either untreated or treated with an acidic solution (pH 2.5) containing 10 mM glycine to strip away externally bound Abs. The amount of internalized receptor was analyzed by flow cytometry. The percentage of internalized KIR3DL1 was calculated using the following equation: 100 × (MFI_exp − MFI_zero)/(MFI_total − MFI_zero). MFI_exp represents the MFI of the internalized KIR3DL1 at a specific time point. MFI_total and MFI_zero represent KIR3DL1 expression in samples either not treated with the acidic solution or samples stained with DX9 and immediately treated with the acid, respectively.

PBMCs were cultured overnight as described earlier and probed with the KIR3DL1-specific Ab, DX9. The KIR3DL1⁻, low- and high-density populations were sorted by the MSKCC Flow Cytometry Core Facility (New York, NY). RNA was isolated from each population using the RNeasy Mini kit as per manufacturer’s instruction (Qiagen) and converted to cDNA using the high-capacity cDNA reverse transcription kit (Life Technologies). KIR3DL1*001 and KIR3DL1*009 were amplified using the respective forward primers, 5′-GCTATACAAAGAAGACAGAATCCACA-3′ and 5′-CAAAGAAGACAGAATCCACG-3′. The reverse primer for both reactions was 5′-GGGAGCTGACAACTGATAGGG-3′. The conditions and control primers for each reaction have been previously described (21).

To generate the soluble KIR-Fc proteins, we transfected Expi293F cells (7.5 × 10⁷) with 30 μg of the appropriate plasmid using the Expifectamine 293F transfection reagent (Life Technologies) following the manufacturer’s protocol. Seven days posttransfection, the supernatants containing the secreted recombinant proteins were harvested and the concentration of soluble KIR-Fc was determined using the Easy-titer human IgG assay according to the manufacturer’s protocol (Pierce Biotechnologies, Rockford, IL). All recombinant proteins were analyzed for proper folding using conformation-specific Abs, as described previously (29). In brief, protein A–coated microspheres (Bangs Laboratories, Fishers, IN) were incubated with 1.0 μg of the KIR-Fc proteins for 1 h at 4°C. KIR-Fc molecules were conformationally tested to determine that the recombinant proteins are folded similarly to how they are on the cell surface by flow cytometry using the conjugated KIR3DL1 conformational specific Abs, DX9 and Z27 (Supplemental Fig. 1). As further confirmation of conformational specificity, recombinant KIR3DL1*001 was denatured at 70°C for 2 min before conjugation to the microspheres.

The parental 721.221 cell line or 721.221 cells stably expressing HLA-B*44:03 (2 × 10⁵) were incubated with either a single concentration (4.0 μg/ml) or a range of concentrations (0.4–4.0 μg/ml) of soluble KIR-Fc proteins for 1 h at 4°C. The samples were washed twice with cold PBS, probed with a PE-conjugated goat Ab specific for human IgG (One Lambda, Canoga Park, CA), and analyzed by flow cytometry. The KIR-Fc proteins were also tested for binding to a panel of single-Ag HLA molecules conjugated to uniquely identifiable microspheres (One Lambda) as previously described in the presence of the blocking Ab, DX9 (4.0 μg/ml), or an isotype control (30). The KIR-Fc proteins (4.0 μg/ml) were incubated with 5 μl beads for 1 h at room temperature while rotating at 300 rpm. The beads were washed three times with wash buffer and probed with a PE-conjugated goat Ab specific for human IgG. The beads were washed twice with wash buffer and the samples were analyzed on a LABScan 100 using the Xponent software (Luminex, Austin, TX). To account for potential variation of HLA density on each bead, we probed the beads with the PE-conjugated pan-HLA class I Ab, W6/32 (eBioscience, San Diego, CA). The MFI values obtained for each interaction between KIR and HLA were first background subtracted for binding to the negative control bead, followed by normalization to the amount of HLA detected for each individual bead region.

After overnight culture in 10³ U/ml IL-2, 5 × 10⁵ PBMCs were coincubated for 4 h with either 5 × 10⁴ parental 721.221 cells or 721.221 cells stably expressing HLA-B*44:03 in the presence of an allophycocyanin H7-conjugated CD107a-specific Ab (BD biosciences). To block inhibition imparted by the KIR3DL1, HLA-B*44:03 interaction, we added DX9 (20 μg/ml) to the coincubation. After this incubation, the cells were probed with conjugated Abs specific for CD56 (Beckman Coulter), CD3 (Biolegend), KIR2DL2/L3 (BD Biosciences), KIR2DL1/S1 (Beckman Coulter), NKG2A (Beckman Coulter), ILT-2 (Beckman Coulter), and KIR3DL1 (Z27; Beckman Coulter). Using the differential surface expression phenotypes of the two KIR3DL1 alleles, the percent CD107a⁺ of each KIR3DL1 population was determined and adjusted for the percent CD107a⁺ of the KIR⁻/NKG2A⁻ population to control for differences in the activation potential of the target cells. The data for each allele are expressed as a ratio comparing the degranulation responses observed against each target cell with the CD107a results observed after coincubation with the HLA⁻ 721.221 cells.

A model of KIR3DL1*009 was based on the crystal structure of KIR3DL1*001-pHLA-B*5701 complex, PDB ID 3VH8 (31). A fragment of KIR3DL1 (residues 7–197) with mutations introduced by means of the interactive program O was subject to molecular dynamics simulation with Desmond/Maestro suite for 1.2 ns using the default parameters (neutralization with Cl⁻ ions, NaCl concentration 0.15 M, force field OPLS_2005, temperature 300 K, pressure 1.01 bar, Coulombic cutoff 9.0 A) (32).

All experiments were performed in triplicate and reproduced in at least two independent experiments. The results were analyzed as indicated by unpaired Student t test or by a one-way or two-way ANOVA followed by a Tukey’s multiple comparisons test. Unless noted, the error bars in figures represent the SEM. All statistical analyses were performed with GraphPad Prism 6.0 software (La Jolla, CA).

Amino acid polymorphisms among the KIR3DL1 alleles can impact the surface expression phenotype of the encoded receptor, whereas the mRNA expression of these alleles remains similar (33, 34). KIR3DL1*001 and KIR3DL1*009 encode for receptors that differ by three amino acids, all of which lie within the D0 extracellular domain (Table I). We compared surface expression of KIR3DL1*009 with two KIR3DL1 allotypes with known high and low expression patterns. Constructs encoding KIR3DL1*001, KIR3DL1*005, or KIR3DL1*009 fused C-terminally to the fluorescent protein, Clover, were transiently transfected into HEK293T cells. KIR3DL1 surface expression on Clover⁺ cells showed that the surface phenotype of KIR3DL1*009 was similar to KIR3DL1*005, which is known to be expressed at a low surface density (Fig. 1A) (35). In contrast, KIR3DL1*009 surface expression was lower compared with the known high-density receptor, KIR3DL1*001. To control for total KIR3DL1 expression, we expressed the data as a ratio of KIR3DL1 surface expression (DX9 MFI) to total KIR expression within the Clover⁺ cells (Clover MFI; Fig. 1B). This quantification further illustrates the similarity in surface expression between KIR3DL1*009 and KIR3DL1*005 and the difference in surface expression observed between KIR3DL1*009 and KIR3DL1*001. The difference in surface expression is not due to differential affinity for the DX9 Ab as bead-conjugated recombinant KIR3DL1*001 and KIR3DL1*009 proteins similarly bound DX9 as determined by flow cytometry (Supplemental Fig. 1).

Mutations were made to the construct encoding wild type KIR3DL1*001 to identify which of the three amino acid polymorphisms contributes to the low surface expression phenotype of KIR3DL1*009. Mutation constructs were created to encode receptors carrying each possible combination of the amino acid polymorphisms between KIR3DL1*001 and KIR3DL1*009. Surface expression of these mutant receptors along with wild type KIR3DL1*001 and KIR3DL1*009 were analyzed on transiently transfected HEK293T cells (Fig. 1C). The individual polymorphisms at aa positions 47 (I47V) and 92 (V92M) had minimal impact on surface expression compared with wild type KIR3DL1*001. Similarly, these two polymorphisms (I47V, V92M) in combination did not dramatically affect KIR3DL1*001 surface expression, although the addition of I47V appeared to restore the modest loss of surface expression associated with the single V92M mutant (p < 0.001). In contrast, a single change from serine to glycine at position 58 (S58G) caused a significant reduction in surface expression, comparable with that of the wild type KIR3DL1*009. When the S58G mutation was combined with V92M (S58G, V92M), surface expression remained significantly reduced. When the S58G mutation was combined with I47V (I47V, S58G), however, the reduction in surface expression was not as dramatic, further suggesting that the valine at position 47 helps to rescue surface expression. These results implicate the glycine at position 58 as the key polymorphism underlying the low surface expression of KIR3DL1*009 and suggest that the additional polymorphism at position 92 can counteract the rescuing effect imparted by the valine at position 47.

In these experiments, transcription of each allele was controlled by a CMV promoter, suggesting the lower surface expression of KIR3DL1*009 could be caused by amino acid polymorphisms that reduce surface stability or processing and trafficking of the receptor to the surface as observed with other KIR3DL1 and KIR2DL molecules (24, 36–38). To determine whether the D0 polymorphisms impacted the surface stability of KIR3DL1*009, we monitored the internalization of the receptor on transfected HEK293T cells. We observed no difference in the internalization of KIR3DL1*009 compared with that of KIR3DL1*001 (Fig. 1D). Therefore, we hypothesized that intracellular retention of KIR3DL1*009 caused the reduction in surface expression. Similar intracellular retention caused by improper folding of the receptor has been demonstrated for KIR3DL1*004 (39). At physiological temperatures, the allotype KIR3DL1*004 is not expressed at the cell surface. However, culturing cells transfected with this allele at subphysiological temperatures resulted in detectable surface expression that was correlated with less intracellular retention of the receptor within the endoplasmic reticulum (39). To determine whether KIR3DL1*009 is intracellularly retained in a similar fashion, KIR3DL1 surface expression was analyzed on transfected HEK293T cells after culture at either 37°C or 25°C (Fig. 1D). As observed earlier, culture at 37°C results in a 2-fold difference between KIR3DL1*009 and KIR3DL1*001 surface expression. However, when the transfectants were cultured at 25°C, KIR3DL1*009 surface expression increased to a level nearly equivalent to that observed for KIR3DL1*001.

Resulting from the surface expression differences observed in the transfection model, we analyzed KIR3DL1 surface expression on primary NK cells from healthy donors heterozygous for KIR3DL1*001 and KIR3DL1*009. Two distinct KIR3DL1⁺ NK cell populations were distinguishable using the KIR3DL1-specific Abs, DX9 and Z27 (Fig. 2A). To determine whether KIR3DL1*001 and KIR3DL1*009 expression was segregated to specific populations, KIR3DL1 mRNA expression was analyzed in the KIR3DL1⁻, low- and high-density populations using primers that specifically amplify either KIR3DL1*001 or KIR3DL1*001 (Fig. 2B). These RT-PCR data reveal that KIR3DL1*001 and KIR3DL1*009 mRNA expression were detected exclusively in the high- and low-density KIR3DL1 populations, respectively, supporting the data from the transfection model that KIR3DL1*009 is expressed at a lower density compared with KIR3DL1*001. Also, because the Ab clone DX9 is capable of binding the KIR3DL1*009 population on primary NK cells, these D0 polymorphisms are not responsible for the lack of DX9 binding to KIR3DS1. To further analyze surface expression of KIR3DL1 on primary NK cells, we compared the KIR3DL1 expression of the heterozygous KIR3DL1*001/KIR3DL1*009 with KIR3DL1 surface expression from donors expressing either KIR3DL1*001 or KIR3DL1*005 alone (Fig. 2C). In accordance with previous reports, KIR3DL1*001 and KIR3DL1*005 donors exhibited high and low KIR3DL1 surface phenotypes, respectively. The flow-cytometry analysis showed that the higher density population from the heterozygous donor matched the surface expression of the KIR3DL1*001 donors, whereas the lower surface density population matched the phenotype of the KIR3DL1*005 donors. These data further support the findings from the transfection model that the polymorphisms present within KIR3DL1*009 cause the receptor to have a low surface phenotype that is similar to KIR3DL1*005.

The polymorphisms within the D0 domain of KIR3DL1*009 are not predicted to directly interact with HLA and, to date, have not been examined for their role in ligand recognition. To understand whether these polymorphisms impact the receptor’s ability to interact with HLA, we analyzed soluble KIR-Fc recombinant proteins containing the extracellular domains of KIR3DL1*009, KIR3DL1*001, KIR3DL1*042, KIR3DL1*005, or KIR3DS1*013 by flow cytometry for binding to HLA-B*44:03 stably expressed on the surface of 721.221 cells (Fig. 3A). KIR3DL1*042 and KIR3DL1*005 were analyzed, because KIR3DL1*042 shares the D0 polymorphisms of KIR3DS1*013 but is identical to KIR3DL1*005 for the remainder of the mature protein (Table I). As expected, KIR3DS1*013 exhibited no binding to HLA. Although KIR3DL1*009 and KIR3DL1*042 exhibited some evidence of binding to HLA-B*44:03 in comparison with KIR3DS1*013, the level of KIR3DL1*009 and KIR3DL1*042 binding was minimal compared with that of KIR3DL1*001 and KIR3DL1*005 (Fig. 3B). KIR3DL1*005 exhibited enhanced binding to HLA-B*44:03 compared with KIR3DL1*001, which could be attributed to the polymorphic residue at position 283 (Fig. 3B, Supplemental Fig. 2E) (40). The difference in binding to HLA-B*44:03 between KIR3DL1*001 and KIR3DL1*009 was consistent over a dose range of the KIR-Fc proteins (Fig. 3C). To demonstrate specificity of binding, all of the KIR-Fc binding to HLA-B*44:03 on the surface of the cells could be blocked by DX9 and Z27 (Supplemental Fig. 2A–D).

To identify the residues that interfere with the interaction between KIR3DL1*009 and HLA, the construct encoding KIR3DL1*001-Fc was mutated to account for each combination of amino acid polymorphisms between KIR3DL1*001 and KIR3DL1*009. Binding of each of the mutant recombinant proteins to HLA-B*44:03 was compared with the binding of the wild type receptors (Fig. 3D). Mutation of KIR3DL*001 at positions 47 (I47V) or 58 (S58G), independently or in combination (I47V, S58G), had minimal impact on the receptor’s ability to bind HLA. However, the presence of a methionine at position 92 (V92M) resulted in a 60% reduction in binding to HLA compared with KIR3DL1*001. Although the combination of valine at position 47 and methionine at position 92 (I47V, V92M) further decreased binding compared with the single change at position 92 (V92M), the mutant still exhibited significantly more binding compared with the wild type KIR3DL1*009. It was the combination of glycine at position 58 and methionine at position 92 (S58G, V92M) that reduced binding of HLA to levels comparable with that of KIR3DL1*009.

To examine whether the reduced binding observed for KIR3DL1*009 and KIR3DL1*042 to HLA-B*44:03 was applicable to other HLA-Bw4 Ags, the wild type recombinant KIR-Fc proteins as well as the mutant protein containing the combined polymorphisms at position 58 and 92 (S58G, V92M) were tested for binding to 18 different HLA Bw4 and 31 noninteracting HLA-Bw6 Ags presenting a heterogeneous pool of peptides using a bead-based multiplex platform (Fig. 4). Consistent with the results for binding to HLA-B*44:03, the recombinant KIR3DL1*009 and KIR3DL1*042 showed significantly reduced binding to all HLA Bw4 molecules analyzed, compared with KIR3DL1*001 and KIR3DL1*005 (Fig. 4A). Specificity of the KIR-Fc proteins is demonstrated, because no binding to the HLA-Bw6 Ags was observed and the binding to the HLA-Bw4 Ags could be blocked by DX9. In support of the binding to HLA-B*44:03 on 721.221 cells, the presence of a glycine and methionine at positions 58 and 92, respectively, were responsible for the reduced binding of soluble KIR3DL1*009 to all of the HLA molecules tested (Fig. 4B). These data further demonstrate that the diminished binding exhibited by KIR3DL1*009 and KIR3DL1*042 is broadly applicable for a large number of HLA molecules.

We then sought to understand how the KIR3DL1*009 polymorphisms impact the structure and ligand interfaces between KIR and HLA. The polymorphic positions 47, 58, and 92 all reside outside of the direct KIR/HLA interface (Fig. 5A) (31). Position 92, however, is located on the final strand (G) of the D0 domain that leads into the D1 domain. Because V92M was identified as the key polymorphism that disrupts KIR3DL1*009 binding to HLA, we hypothesized this position may cause a conformational shift in the receptor that would indirectly impact the KIR/HLA interaction. To understand how the KIR3DL1*009 polymorphisms may impact the receptor structure, a molecular dynamic simulation was performed on the KIR3DL1*001 crystal structure after insertion of the KIR3DL1*009 polymorphisms (Fig. 5B). Consistent with the binding data, the polymorphisms at positions 47 and 58 did not impact the overall structure of KIR3DL1. However, V92M caused a significant change in the structure likely resulting from steric hindrance caused by the bulkier side chain of methionine. Interestingly, the change in conformation did not appear to impact the interface between D0 and HLA even though the amino acids involved at this interface (9F, 11S, 13W) are on a neighboring strand (A). In contrast, V92M caused a dramatic conformational shift in the overall protein structure. In particular, the hinge angle between the D0 and D1 domains defined as the angle between Cα atoms of residues 82, 178, and 197 is 92° in the crystal structure of KIR3DL1*001 but measures 102° after the molecular dynamic simulation of KIR3DL1*009. This conformational change causes a significant displacement of the loops between the C and C’ strands as well as the E and F strands that contain the residues (138G, 140S, 165M, 166L, 167A) that directly interact with HLA. These data support the observed binding data, suggesting that the reduced capability of KIR3DL1*009 to bind HLA is caused primarily by the polymorphism, V92M, which results in a conformational change of the D1 domain.

We then evaluated how the observed lower surface expression and weaker ligand binding might impact KIR3DL1*009 receptor function. CD107a mobilization was monitored on KIR3DL1⁺ NK cells from a healthy donor heterozygous for KIR3DL1*001 and KIR3DL1*009 after incubation with HLA⁻ 721.221 cells or 721.221 cells stably expressing HLA-B*44:03 in the presence or absence of the DX9 blocking Ab. Based on the differential surface expression phenotypes of these alleles determined by Z27 staining, degranulation by KIR3DL1*001 and KIR3DL1*009 populations could be independently monitored. This donor was not positive for an HLA-Bw4 Ag, thereby negating a potential effect of licensing differences between these two populations. Accordingly, the response levels, as measured by CD107a mobilization, were equivalent between the KIR3DL1*001⁺ and KIR3DL1*009⁺ populations after stimulation with the HLA⁻ 721.221 cells, demonstrating that these two populations have equal thresholds for activation against target cells. Likewise, blocking the KIR3DL1, HLA-B*44:03 interaction with the DX9 Ab rescued the degranulation of each population to levels similar to the response against the HLA⁻ 721.221 cells (Fig. 6). However, in the absence of a blocking Ab, the KIR3DL1*009⁺ NK cells demonstrated significantly greater potential to degranulate compared with the KIR3DL1*001⁺ NK cells when coincubated with a HLA-Bw4⁺ target cell. Activation of the KIR3DL1*009⁺ population was 2-fold greater after coincubation with HLA-Bw4 targets in the absence of the DX9 Ab compared with the activation response of the KIR3DL1*001⁺ population. To understand whether this enhanced ability to respond was due to lower surface expression or to reduced binding strength, inhibition of KIR3DL1*005⁺ NK cells from a HLA-Bw4⁻ individual was monitored in parallel. Degranulation of the KIR3DL1*005 population against the 721.221 targets was equal to the responses observed for the KIR3DL1*001 and KIR3DL1*009 populations. Similar to the KIR3DL1*001 population, the KIR3DL1*005⁺ NK cells were inhibited significantly more by HLA-B*44:03 compared with the KIR3DL1*009 population. These data demonstrate that the reduced binding strength exhibited by KIR3DL1*009 allows the KIR3DL1*009⁺ NK cells to maintain greater activation against a Bw4⁺ target cell compared with KIR3DL1*001 and KIR3DL1*005.

KIR3DL1 functions as an inhibitory receptor that recognizes HLA-A and HLA-B molecules that carry the Bw4 epitope (17–19). The KIR3DL1/S1 gene has evolved under balancing selection into two distinct lineages of inhibitory KIR3DL1 allotypes and one lineage of activating KIR3DS1 allotypes (41). The two inhibitory receptors in this study, KIR3DL1*001 and KIR3DL1*009, segregate as members of the KIR3DL1*005-like lineage based on sequence homology (15). Differing from the extensive polymorphisms within the KIR3DL1 lineages, the KIR3DS1 allotypes have remained fairly conserved. Meiotic recombination events and mutations are the main contributors to the diversity of the KIR3DL1 lineages. The two most common regions for recombination involve the D0 domain polymorphisms and a dimorphism at position 283 of the D2 domain. KIR3DL1*009 and KIR3DL1*042 resulted from such a genetic conversion event involving KIR3DS1*013 with either KIR3DL1*001 or KIR3DL1*005, respectively (22). Although the protein sequences of KIR3DL1*009 and KIR3DL1*042 are more similar to their KIR3DL1 parental alleles, the minimal HLA binding exhibited by these receptors is more analogous to that observed for KIR3DS1*013. The findings of this study provide novel insight into the functional consequences of KIR3DS1-specific D0 polymorphisms to explain the lack of binding between KIR3DS1*013 and HLA-Bw4.

The data from both primary NK cells and a transfection model revealed KIR3DL1*009 is expressed at a low surface density similar to the expression levels previously characterized for KIR3DL1*005 (35). Analysis of mRNA expression from sorted KIR3DL1 populations on primary cells from a heterozygous KIR3DL1*001/*009 individual revealed that transcription of KIR3DL1*001 and KIR3DL1*009 segregated into two distinct populations defined by their KIR3DL1 surface phenotype. This segregation of KIR3DL1 allele expression at the transcript level supports previous studies that have demonstrated allele-specific gene expression of KIR3DL1 (42, 43). Site-directed mutagenesis of the wild type KIR3DL1*001 revealed a glycine residue at aa position 58 is the main contributor to KIR3DL1*009 phenotype, whereas a methionine residue at position 92 is necessary to completely reproduce the reduced surface expression phenotype of KIR3DL1*009. The higher conformational freedom of glycine at this position may increase the flexibility of the polypeptide chain, causing greater difficulty to properly fold and mature the receptor, ultimately resulting in intracellular retention of the receptor. Consistent with this interpretation, the surface expression of KIR3DL1*009 was enhanced to levels similar to KIR3DL1*001 on cells cultured at subphysiological temperatures. These data are analogous to previous findings for KIR3DL1*004, which is completely retained within the endoplasmic reticulum at physiological temperatures but is capable of surface expression at reduced temperatures (39). We also observed that once at the cell surface, the stability of KIR3DL1*001 and KIR3DL1*009 was similar. Taken together, these results suggest KIR3DL1*009, similar to KIR3DL1*004, is likely retained within the cell caused by the presence of a glycine and methionine at aa positions 58 and 92.

The KIR3DS1-specific D0 polymorphisms also disrupted the interaction between KIR3DL1 and HLA-Bw4. The data presented within this manuscript identify two novel polymorphisms at positions 58 (S58G) and 92 (V92M) that, in combination, cause a significant reduction in the capacity of KIR3DL1 to interact with HLA. Varying from the surface expression data, the methionine at position 92 is the key residue, whereas the glycine at position 58 further augments the effect. Although aa position 92 does not directly interact with HLA, molecular dynamic analysis indicates that 92M causes steric hindrance within the molecule, resulting in a significant conformational shift within the D1 domain of KIR3DL1*009. This conformational shift significantly displaces the D1 domain amino acid residues that directly interact with HLA. The D0 domain has previously been shown to have an enhancer effect for HLA binding (44). To our knowledge, our findings are the first to show that natural polymorphisms within this domain are sufficient to significantly reduce HLA binding. In combination with the D1 polymorphisms of KIR3DS1*013, the effects imparted by the D0 polymorphisms provide the molecular mechanism by which KIR3DS1*013 has lost the ability to interact with HLA (20). The loss of ligand binding of KIR3DL1*009 and KIR3DL1*042 may have provided a protective effect in regions where pathogens caused an increase in the strength of the interaction between KIR3DL1 and HLA, causing less NK cell–mediated killing of the pathogen similar to evasive mechanisms described for HIV-specific peptides and inhibitory KIR function (45).

KIR3DL1*009⁺ NK cells showed a greater capability of response against a HLA-B*44:03⁺ target cell compared with KIR3DL1*001⁺ or KIR3DL1*005⁺ NK cells, suggesting that they have a weak sensitivity for inhibition. Given that KIR3DL1*009 is expressed at a surface density similar to KIR3DL1*005, the reduced ligand binding, rather than the reduced surface expression, contributed to the lack of inhibition imparted by HLA-B*44:03. Although the inhibition observed for KIR3DL1*001 and KIR3DL1*005⁺ NK cells was similar, this could be explained by the significantly greater strength of interaction observed between KIR3DL1*005 and HLA-B*44:03. It is therefore possible that the reduced surface expression of KIR3DL1*005 and other KIR3DL1 alleles may impact the functional capacity of the receptors when confronted with an HLA Bw4 Ag that exhibits weaker binding strength.

This study illuminates key functional effects on ligand recognition and receptor surface expression caused by D0 polymorphisms that are conserved in all KIR3DS1 alleles and are present in KIR3DL1*009, KIR3DL1*042, and KIR3DL1*057. The data presented further demonstrate the complexity of the KIR3DL1–HLA interaction and provide insight to how KIR3DL1 and HLA combinations may impact NK cell responses to viral pathogens and malignant cells.

We thank Jean-Benoit Le Ludeuc, Christopher Forlenza, Xia-Rong Liu, and Frances Weis-Garcia for technical assistance. We also thank Addgene for providing the pcDNA3-Clover vector from Dr. Michael Lin.

The authors have no financial conflicts of interest.