The recognition of MHC class I molecules by killer cell immunoglobulin-like receptors (KIR) is central to the control of NK cell function and can also modulate the CTL activation threshold. Among KIR receptors, KIR3DL2 is thought to interact with HLA-A3 and -A11, although direct evidence has been lacking. In this study, we show that HLA-A3 and -A11 tetramers specifically bind to KIR3DL2*001 transfectants and that this recognition is peptide-specific. Single amino acid substitutions in the nonamer peptide underline a critical role for residue 8 in recognition of KIR3DL2. However, the role of this interaction in vivo still remains to be established.
Natural killer (NK) cells are a key component of innate immune defense 1, 2, and they may promote the establishment of appropriate adaptive immune responsesthrough cross-talk with dendritic cells (DC) 3. They secrete cytokines and chemokines in response to infection and contribute to the elimination of infected or transformed cells.NK cell effector functions are regulated by activating and inhibitory cell surface receptors that are differentially expressed on NK cells. In humans, the recognition of MHC class I molecules by killer cell Ig-like receptors (KIR) contributes to the array of receptor-ligand interactions controlling NK cell responses to a target. Two families of KIR can be identified based on the number of extracellular Ig domains, and they have either short (S) or long (L) intracytoplasmic tails that transduce activating and inhibitory signals, respectively. KIR2D (with 2 Ig domains) interact with HLA-C allotypes, while KIR3D (with 3 Ig domains) interact with HLA-B molecules that display the Bw4 epitope (KIR3DL1) or with HLA-A alleles (KIR3DL2). The specificity of the KIR3DL2 receptor is, however, still controversial. It has been difficult to show MHC class I-mediated inhibition of NK cell cytotoxicity through KIR3DL2 interaction 4. Two groups have reported that expression of HLA-A3 and -A11 on EBV-transformed B cells was sufficient to protect them from lysis by KIR3DL2+ NK cell clones 5, 6. However, they could not totally exclude the possibility that other unidentified receptor-ligand interactions may be responsible for the inhibition. The likelihood of a specific interaction of HLA-A3 with KIR3DL2 was further strengthened by the observation that the class I-deficient target cell 721.221 transfected with HLA-A3 was protected from lysis and was weakly stained by a soluble KIR3DL2-IgG1 fusion protein 5. To provide direct evidence that KIR3DL2 interacts with specific HLA-A alleles, we generated a panel of MHC class I tetrameric complexes that were tested for binding to KIR3DL2-expressing cell lines.
2.1 Direct binding of HLA-A3 to KIR3DL2 is peptide-specific
Cytotoxicity of KIR3DL2+ NK clones was previously shown to be inhibited by target cells expressing HLA-A3, and a soluble form of KIR3DL2 weakly stained HLA-A3-expressing cells 5, 6. To provide direct evidence that HLA-A3 interacts with KIR3DL2, we generated HLA-A3 tetramers refolded with several antigenic peptides and tested their binding to KIR3DL2*001 transfectants (Table 1). Interestingly, only the HLA-A3 tetramer (A3-RLR) refolded with a peptide from the Epstein Barr Virus (EBV) EBNA3A (603–611) viral protein was recognized by the KIR3DL2 receptor (Fig. 1A) 7. No binding to KIR3DL2 transfectants could be detected with the other four HLA-A3 tetramers refolded with previously described HLA-A3-restricted CTL epitopes or with one self peptide eluted from HLA-A3 8. We excluded that the lack of staining was due to poor quality/stability of the tetramers, as they could bind to ILT-2 transfected Baf3 cells (Fig. 1A) and EBV-specific CTL (data not shown). To further demonstrate direct interaction of HLA-A3 with KIR3DL2, we used a blocking anti-KIR3DL2 mAb (DX31). While isotype Ig control did not affect staining of KIR3DL2 transfectants with A3-RLR tetramer, DX31 blocked the staining (Fig. 1B). In addition, binding of the tetramer to KIR3DL2 was partially abolished when incubated with anti-HLA-A3 antibody (GAP-A3) or anti-class I antibody (DX17) (data not shown). Altogether, these results show direct binding of HLA-A3 to KIR3DL2, confirming previous studies. Importantly, this interaction seems to be highly dependent on the peptide present in the groove.
|HLA molecule||Peptide sequence||Pathogen/protein (residues)||Binding of the tetramer to|
|Baf3-KIR3DL2 cells||Baf3-ILT2 cells||Baf3-CD94/NKG2A cells|
|ILRGSVAHK||Influenza NP (265–273)||–||+||–|
|KLFEKVYNY||Eluted self peptide||–||+||–|
|HLA-E*0101||VMAPRTLFL||HLA-G signal sequence (3–11)||–||+||+|
|HLA-G*0101||RIIPRHLQL||Eluted self peptide||–||+||–|
2.2 HLA-A11 can be refolded with the EBV EBNA3A peptide and binds to KIR3DL2
HLA-A11 has also been proposed to interact with KIR3DL2 because of its expression on a B-EBV target cell line, which was protected from killing by KIR3DL2+ NK cell clones 6. However, no further demonstration has been reported. To investigate the binding specificity of HLA-A11, we constructed HLA-A11 tetramers refolded with several antigenic peptides. While they all bound to ILT-2-transfected Baf3 cells, the tetramers did not bind to KIR3DL2*001-Baf3 cells or negative control CD94/NKG2A-Baf3 transfectants (Table 1). HLA-A3 and HLA-A11 have been reported to belong to the same supertype family, which is characterized by overlapping peptide-binding repertoires and structural similarities in the antigen-binding groove 9. We therefore attempted to refold HLA-A11 with the EBV EBNA3A peptide previously shown to interact with the KIR3DL2 receptor when complexed to HLA-A3 (Fig. 1A). We successfully generated A11-RLR tetramers, consistent with the subsequent finding that EBV EBNA3A (603–611)-specific, HLA-A11-restricted CTL can be found in EBV-infected patients (unpublished data). A11-RLR tetramer bound both ILT-2 and KIR3DL2 transfectants but not the negative control CD94/NKG2A-Baf3 cells (Fig. 2 and data not shown). The binding to KIR3DL2 was specific, because it could be blocked by addition of DX31 mAb (Fig. 2). In addition, NK cell clones expressing KIR3DL2 could be stained with both HLA-A3 and HLA-A11 tetramers refolded with the EBV EBNA3A (603–611) peptide, and the staining was blocked by DX31 mAb (Fig. 3). These results suggest that KIR3DL2 recognizes a structure common to HLA-A3 and -A11 and confirms that this interaction is highly peptide-specific. We also tried to refold HLA-A6801, an other HLA belonging to the HLA-A3-like supertype, with the EBV EBNA3A peptide (603-611), but we did not succeed (data not shown). Several other HLA-A, -B, -E and -G tetramers refolded with CTL epitopes or self peptides were tested for binding to KIR3DL2, but none of them stained (Table 1). Although we cannot exclude that other HLA allotypes bind to KIR3DL2, this study suggests that only a restricted set of peptides refolded with HLA-A3 and -A11 can be recognized by KIR3DL2.
2.3 Substitution at position 8 abolishes KIR3DL2 interaction with HLA-A3
In an attempt to identify the amino acid residues critical for KIR3DL2 recognition, we synthesized four EBV EBNA3A (603–611) peptides with single amino acid substitutions (Fig. 4A). The mutations we tested were at positions 7 and 8, which were previously reported to be important for contact with other KIR 10–13, and position 5, where the side chain is exposed to solvent rather than buried in the groove 14. Interestingly, replacing the glutamic acid residue at position 5 with an alanine allowed refolding of HLA-A3 (A3RLR-(E5A)) but did not disrupt binding to KIR3DL2 (Fig. 4B). This interaction was specific, as DX31 mAb abolished the staining. Because position 7 is an auxiliary anchor residue in peptides binding to HLA-A3, we first substituted the glutamine with asparagine, a related residue. We could not refold HLA-A3 with this peptide, whereas a more drastic substitution with alanine allowed refolding of HLA-A3 (A3RLR-(Q7A)). Similarly to the mutation at position 5, (A3RLR-(Q7A)) tetramer bound to KIR3DL2*001 transfectants, and the staining was blocked by DX31 (Fig. 4B). The last substitution tested was at position 8 where the valine was replaced by an arginine. HLA-A3 could be refolded with that mutated peptide, but while the tetramer (A3RLR-(V8R)) bound to ILT-2 transfectants, it failed to bind to KIR3DL2*001-Baf3 cells (Fig. 4C). Altogether, these results demonstrate that position 5, where the side chain is usually exposed, is not involved in the interaction with KIR3DL2, while position 8 (but not 7) is critical for KIR3DL2 recognition. These two residues were reported to be involved in the interaction of MHC class I with KIR3DL1 and KIR2DL1 receptors 10, 11.
Our study provides direct evidence that HLA-A3 and -A11 can bind to KIR3DL2, while other HLA allotypes, including HLA-A2, -B7, -B8, -B27, -B58, -E and -G, did not interact. Interestingly, thisinteraction is highly dependent on the peptide present in the groove, as only one out of the eight CTL epitopes we tested permitted KIR recognition. In addition, HLA-A3 refolded with a self peptidedid not interact with KIR3DL2. This finding is somewhat surprising, as HLA-A3 molecules associated with self peptides are expected to prevent NK cell lysis of autologous healthy cells through interaction with KIR3DL2. However, we cannot exclude that only a restricted set of self peptides permits KIR3DL2 recognition. The restricted peptide specificity of KIR3DL2 we observed may explain the difficulty of demonstrating MHC class I-mediated inhibition through KIR3DL2 and the previous controversy as to whether it recognizes HLA-A allotypes 4–6. The CTL epitope allowing KIR3DL2 recognition is derived from the EBV protein EBNA3A (603–611) 7, which suggests that HLA-A3 and -A11 interactions with KIR3DL2 may be particularly relevant during EBV infection. Like all other latent viral proteins, EBNA3A is expected to be expressed by all EBV-transformed B cell lines, consistent with the identification of KIR3DL2 interaction with HLA-A3 or -A11 using such targets. Peptide specificity in the recognition of MHC class I by NK cell clones has previously been reported for inhibitory KIR receptors. HLA-B*2705 interaction with KIR3DL1 shows some degree of peptide specificity, with residues at positions 7 and 8 identified as critical for the recognition 10, 15. Binding of soluble KIR2DL1 to HLA-Cw4 only occurred when the MHC class I molecule was loaded with peptides, and certain substitutions at position 7 and 8 abolished the interaction 11. HLA-Cw*0304 and -Cw7 interactions with KIR2DL2 were also modulated by the peptide present in the groove 12, 16. Similarly, we found that substitution at position 8 in the EBV EBNA3A peptide abolished binding to KIR3DL2. It therefore seems that the mode of interaction of HLA-A3/A11 with KIR3DL2 is similar to the other KIR interacting with their respective MHC class I ligands. This is also consistent with the crystal structure of KIR2DL2 in complex with HLA-Cw3, which provided direct evidence that KIR contact requires position 8 of the peptide to be a small residue 13. A model of KIR3DL1 binding to the HLA-B allele harboring the Bw4 epitope has also been described and suggests that the peptide participates in the interaction with KIR3DL1 17. It is intriguing that KIR3DL2 displays such restricted ligand specificity. Interestingly, KIR3DL2 polymorphism is also strikingly different from the other KIR. There is evidence for recombination in the generation of KIR3DL1 diversity, while KIR3DL2 polymorphism results from point mutations 18. Whether this reflects differences in the selection imposed by pathogens or structural constraintsin the genomic organization of KIR haplotypes is not known and needs to be investigated in relation to the restricted ligand specificity we observed.
In healthy individuals, KIR3DL2 is expressed by about 20–30% of circulating NK cells and ∼5% of T cells (mainly CD8+). It remains to be established whether the interaction of KIR3DL2 with HLA-A3 and -A11 complexed with a restricted set of peptides is physiologically relevant. KIR3DL2 expression was recently associated with CD4+ cutaneous T cell lymphomas and could contribute to their resistance to activation-induced cell death (AICD), although no solid evidence has been provided 19, 20. Using the HLA-A3 and -A11 tetramers, we could stain NK cell clones expressing quite high levels of KIR3DL2 receptors, but we failed to stain NK or T cells from PBMC, suggesting that the density of the receptor on circulating NK and T cells may not be sufficient to obtain binding (data not shown). This discrepancy between staining of PBMC and cell lines expressing high levels of NK receptors has previously been reported (21 and unpublished data). Alternatively, the interaction of KIR3DL2 with HLA-A3 and -A11 may have a relatively lower affinity compared to other KIR or CD94/NKG2A. KIR3DL2 polymorphism may also modulate the affinity of the interaction, although both KIR3DL2*001 and KIR3DL2*002 alleles were previously reported to interact with HLA-A3 and -A11. As KIR3DL2*001 and KIR3DL2*002 are expressed by 50% and 36% of the population, respectively, and because we stained eight different donors, it is unlikely that the lack of staining of PBMC was due to specific binding to some KIR3DL2 alleles and not others.
More recently, a novel ligand for KIR3DL2, which consists of HLA-B*2705 heavy chain homodimers lacking β2-microglobulin and peptides, was identified 22. Expression of these structures seems to be elevated in patients with spondylarthritis, and they may contribute to disease pathogenesis. Interestingly, classical HLA-B*2705 tetramers refolded with β2-microglobulin and peptide bound to KIR3DL1 but not KIR3DL2, while the homodimers bound to both receptors. The HLA-B*2705 homodimers must therefore adopt a structure resembling HLA-A3 or -A11 complexed with theEBV EBNA3A peptide while retaining its ability to interact with KIR3DL1. This suggests similarities in the mode of binding to the two receptors. Whether this interaction also occurs in vivo, along with the contribution of KIR3DL2 signaling to the regulation of NK and T cell functions, remain to be established.
4 Materials and methods
4.1 Cell lines and antibodies
Baf3 cells transfected with KIR3DL2*001 (also known as NKAT4), ILT-2 and CD94/NKG2A have been described previously and were maintained in RPMI medium supplemented with 10 ng/ml mouse IL-3 21, 23. Expression of the NK receptors was monitored by flow cytometry using the following monoclonal antibodies: anti-KIR3DL2 (DX31), anti-ILT-2 (VMP55)24 and anti-CD94 (DX22).
4.2 Generation of HLA tetramers and flow cytometry
HLA-peptide tetrameric complexes ("tetramers") of HLA-A*0301 and -A*1101 complexed with epitope peptides as outlined in Table 1 were produced as previously described 25. FACS stainings were performed as previously described 26. Briefly, 2.5×105 cells were spun down and resuspended with 2 μl titrated phycoerythrin (PE)-conjugated tetramers or Extravadin-PE control and then incubated for 20 min at 37°C. Additional staining was performed using DX31 or Ig isotype control antibodies. Cells were washed and stored in Cell FixTM buffer (Becton Dickinson, BD) at 4°C until analysis. Samples were analyzed on a BD FACSCalibur flow cytometer using Cellquest software. Blocking experiments were performed using DX31, which specifically binds to KIR3DL2. Ba/F3-KIR3DL2 cells were incubated with 2 μg DX31 on ice for 30 min, washed and stained with tetramers as outlined above. HCA2 antibody was used as an isotype Ig control in these experiments.
4.3 Generation of NK cell clones
Peripheral blood mononuclear cells (PBMC) were isolated from blood samples by Ficoll-Hypaque gradient separation. CD3–/CD56+ NK cells were negatively selected by using anti-CD3-, anti-CD14- and anti-CD19-coated magnetic beads (Dynal Biotech, Oslo, Norway). Purified NK cells were subsequently co-cultured with irradiated allogeneic PBMC in RPMI 1640 supplemented with 10% heated inactivated human AB serum (H10) and recombinant human IL-2 (rhIL-2) at 100 U/ml for 2 weeks. NK cell clones were generated by limiting dilution and further cultured in H10 medium supplemented with rhIL-2 at 100 U/ml and rhIL-15 at 10 ng/ml in the presence of feeder cells.
We thank DNAX and Lewis Lanier for the Baf3 transfectants. This work was supported by grants from the CNRS and Sidaction (V. M. B., H. A.) and the Medical Research Council of the UK (P. H., T. D., H. T., M. W., C. W., S. R.-J.).