Natural killer (NK) cells express receptors that recognize major histocompatibility complex (MHC) class I molecules and regulate cytotoxicity of target cells. In this study, we demonstrate that Ly49A, a prototypical C-type lectin–like receptor expressed on mouse NK cells, requires species-specific determinants on β2-microglobulin (β2m) to recognize its mouse MHC class I ligand, H-2Dd. The involvement of β2m in the interaction between Ly49A and H-2Dd is also demonstrated by the functional effects of a β2m-specific antibody. We also define three residues in α1/α2 and α3 domains of H-2Dd that are critical for the recognition of H-2Dd on target cells by Ly49A. In the crystal structure of the Ly49A/H-2Dd complex, these residues are involved in hydrogen bonding to Ly49A in one of the two potential Ly49A binding sites on H-2Dd. These data unambiguously indicate that the functional effect of Ly49A as an MHC class I–specific NK cell receptor is mediated by binding to a concave region formed by three structural domains of H-2Dd, which partially overlaps the CD8 binding site.
NK cells are a population of lymphocytes with an ability to spontaneously kill tumor cells and infected cells 1. Target recognition by NK cells involves MHC class I molecules on target cells 2. NK cells express C-type lectin–like or Ig-like receptors for MHC class I molecules 3,4. Engagement of these MHC class I receptors by its ligands inhibits or activates NK cells, depending on a motif found in the cytoplasmic region or a positively charged amino acid residue in the transmembrane segment of the receptors 3,5,6.
Mouse NK cells express receptors of the Ly49 family, comprised of >10 members. These molecules are homodimers of type II transmembrane proteins with C-type lectin–like domains in the extracellular region 7,8,9,10,11,12. Ly49A, the prototype member of this family, is an inhibitory receptor specific for the mouse MHC class I molecules H-2Dd and H-2Dk 13. MHC class I is a ternary complex of a heavy chain, which consists of α1/α2 and α3 domains, β2-microglobulin (β2m), and a peptide bound to a groove in the α1/α2 domain 14. Several lines of evidence suggested the involvement of the α1/α2 domain in the recognition of H-2Dd by Ly49A. The 34-5-8S antibody, which recognizes the α1/α2 domain of H-2Dd, but not an antibody against the α3 domain (34-2-12S) inhibits functional and physical interaction between Ly49A and H-2Dd 13,15. Ly49A recognizes the natural mutant MHC class I molecule dm-1, which has the α1 and NH2-terminal half of the α2 domain of H-2Dd with the rest of the molecule derived from H-2Ld, which is not a ligand for Ly49A 16. Ly49A recognizes only the peptide-bound form of H-2Dd molecules, but there is no apparent specificity for peptides as long as they have the anchoring residues required to bind H-2Dd 17,18. Despite the homology of Ly49A to C-type lectins, the ability of Ly49A to bind certain carbohydrates like fucoidan or dextran sulfate 19, and the presence of two Asn-linked carbohydrates in H-2Dd, binding of Ly49A to H-2Dd does not depend on carbohydrates on H-2Dd 20. Our previous study using Dd/Kd chimeric molecules has shown that the polymorphic determinant of H-2Dd that restricts Ly49A reactivity lies in the NH2-terminal halves of α1 and α2 regions of H-2Dd that form the bottom of the α1/α2 domain 20. Recently, Tormo et al. 21 resolved the crystal structure of the Ly49A/H-2Dd complex, providing two possible Ly49A binding sites on H-2Dd, and suggested that site 1, which includes the NH2 terminus of α1 α-helix and COOH terminus of α2 α-helix, is the functional binding site for Ly49A rather than site 2, a concave region formed by α1/α2 and α3 domains, and β2m. However, experimental data explaining which represents the functional Ly49A binding site on H-2Dd that leads to inhibition of NK cell cytotoxicity had been missing.
In this study, we first focused on β2m, which had not been thought to be involved in the recognition of H-2Dd by Ly49A, and demonstrate the essential role of β2m in recognition. This result led us to explore a panel of individually Ala-substituted mutants of H-2Dd, in which mutations were introduced into α1/α2 or α3 domains, to functionally interact with Ly49A. We found specific residues in the α1/α2 and α3 domains of H-2Dd that are critically important for Ly49A interaction, as determined by soluble Ly49A (sLy49A) binding and inhibition of cytotoxicity by Ly49A+ NK cells. Surprisingly, our results indicate that site 2 rather than site 1 is the functional binding site of Ly49A on H-2Dd that results in the inhibition of NK cell cytotoxicity.
Materials And Methods
C57BL/6J mice were obtained from Clea. This study was approved by the Animal Experiment Review Board of the Faculty of Pharmaceutical Sciences at The University of Tokyo, Tokyo, Japan.
Cells and Antibodies.
C1498 cells, which have C57BL/6 origin with H-2b and β2mb, and Daudi cells were obtained from American Type Culture Collection. Ly49A-transfected Chinese hamster ovary (CHO) cells and H-2Dd-, Kd-, or Ld-transfected C1498 cells were established as described 13,19,20. Ly49A+ IL-2–activated NK cells were prepared from C57BL/6 mouse splenocytes as described 13. S19.8 (anti–mouse β2mb; reference 22), BBM1 (anti–human β2m; reference 23), 34-5-8S (anti–H-2Dd α1/α2; reference 24), 34-2-12S (anti–H-2Dd α3; reference 24), and A1 (anti-Ly49A; reference 25) were purified from culture supernatants. Fab and F(ab′)2 fragment of antibodies were prepared with standard methods. Because both of the anti-β2m antibodies are mouse IgG2b isotypes, it is difficult to make F(ab′)2 fragments; Fab fragments of these antibodies were instead used in cell-mediated cytotoxicity assays to avoid antibody-dependent cellular cytotoxicity (ADCC). Daudi cells were transfected by electroporation with mouse β2mb or human β2m cDNA (a gift from Dr. R.K. Ribaudo, Molecular Applications Group, Silver Spring, MD) that was cloned into pApuro vector 26 together with wild-type H-2Dd cDNA cloned into pHβAPr-1neo vector, and stable transfectants were established as described 20.
Cell-mediated Cytotoxicity Assay and Cell–Cell Adhesion Assay.
Cell-mediated cytotoxicity of Ly49A+ NK cells against H-2Dd–transfected C1498 cells was tested by a 4-h 51Cr-release assay as described 20. All cytotoxicity assays were done in triplicate. When indicated, target or Ly49A+ NK cells were preincubated for 15 min with anti–MHC class I or anti-Ly49A antibodies, respectively. To prevent ADCC, anti-β2m and anti–H-2Dd antibodies were used as Fab and F(ab′)2 fragments, respectively. Intact antibodies and F(ab′)2 fragments were used at 10 μg/ml and Fab fragments were used at 40 μg/ml. Binding of H-2Dd–transfected cells to Ly49A-transfected CHO cells was examined as described previously 15,20. When indicated, target cells or Ly49A+ NK cells were preincubated for 15 min with anti–MHC class I antibodies or anti-Ly49A antibody, respectively.
Preparation of sLy49A Tetramer.
sLy49A was prepared as described elsewhere (Matsumoto, N., K. Tajima, M. Mitsuki, and K. Yamamoto, manuscript submitted for publication). In brief, the extracellular domain of Ly49A with NH2-terminal biotinylation sequence tag 27 was expressed in Escherichia coli using an efficient T7 RNA polymerase-based system 28. The recombinant protein was in vitro refolded by dilution 29 and purified by cation exchange and gel filtration column chromatography. The sLy49A was biotinylated by biotin ligase BirA (Avidity). sLy49A tetramer was formed by incubating the biotinylated sLy49A with R-PE–conjugated streptavdin (BD PharMingen) at a molar ratio of 4:1.
β2m Replacement Studies.
H-2Dd–transfected C1498 cells were cultured for 16 h in the presence or absence of 4 μM human β2m (purified from plasma; Calbiochem) in RPMI 1640 free from FCS at 37°C. Then the cells were used for flow cytometry or cell-mediated cytotoxicity assay. After 16 h of culture under FCS-free condition, >99% of the cells were viable.
Site-directed Mutagenesis and Stable Transfection of Cells.
Point mutations were introduced by primer extension with T4 DNA polymerase using the Altered Sites® II system (Promega) or by sequential PCR steps as described by Cormack 30. To introduce point mutation by the PCR-based technique, we introduced individual mutations into the 5′ fragment of H-2Dd cDNA, which encodes a signal sequence and α1/α2 domain, or 3′ fragment, which encodes the rest of H-2Dd, by PCR using the following terminal primers: 5′-CCTGCAGGTCGACTCTAGAG-3′ and 5′-GTTCTTAAGAGCGTAGCATTCCCGTTC-3′ for the 5′ fragment and 5′-CTCTTAAGAACAGATCCCCCAAAGGC-3′ and 5′-GGATCCACACCAGGCAGCTG-3′ for the 3′ fragment. These primers contain SalI, AflII, or BamHI site (indicated in italics). The sequence of primers used in the first PCR step will be provided on request. All the PCR reactions were performed using KOD-Plus-DNA polymerase (Toyobo). Each of the 5′ or 3′ fragments was subcloned into the SmaI site of pBluescript® II SK+ (Stratagene), and the sequence was confirmed by reading both strands using an LS-2000 sequencer (LI-COR). Then, each of the mutant 5′ fragments and wild-type 3′ fragment or wild-type 5′ fragment and each of the mutant 3′ fragments was directionally cloned into pHβApr-1neo vector and the constructs were used for the transfection of C1498 cells by electroporation as described 20.
Cells were stained with 10 μg/ml of indicated primary antibodies and then with FITC-goat anti–mouse IgG F(ab′)2. For sLy49A staining, cells were stained with PE-labeled sLy49A tetramer and then fixed with 0.5% paraformaldehyde in PBS. In both cases, the stained cells were analyzed using a FACScalibur™ with CELLQuest™ software. Binding of the sLy49A tetramer to each mutant H-2Dd was calibrated, with the expression of H-2Dd detected by 34-2-12S or 34-5-8S antibodies in the following formula: sLy49A tetramer binding index = (mean fluorescence intensity [MFI] of sLy49A tetramer stained cells − MFI of streptavidin-PE stained cells)/(MFI of 34-2-12S or 34-5-8S stained cells − MFI of control antibody stained cells). sLy49A tetramer binding of each H-2Dd mutant is expressed as the relative value of the binding index when wild-type H-2Dd is adjusted to 100. Because introduction of E227A mutation into H-2Dd abrogated the 34-2-12S epitope, we evaluated the expression of E227A mutant by reactivity with the 34-5-8S antibody.
Anti–Mouse β2mb Antibody Inhibits Recognition of H-2Dd by Ly49A.
To investigate the possible involvement of β2m in the recognition of H-2Dd by Ly49A, we examined the effect of the S19.8 antibody, which reacts with the b allele of mouse β2m, on the interaction of Ly49A with H-2Dd in target cell killing assays. In these experiments, antibody fragments lacking the Fc region were used to avoid the potential effect of target lysis by ADCC mediated by NK cells through their Fc receptors. F(ab′)2, or Fab for IgG2b isotypes, fragments were used. As reported previously 13, Ly49A+ LAK cells were unable to kill H-2Dd–transfected C1498 (H-2b) lymphoma cells efficiently (Fig. 1 A). Importantly, the addition of S19.8 Fab fragments to the killing assay, as well as the positive control anti–H-2Dd α1/α2 antibody (34-5-8S) F(ab′)2 fragments or intact anti-Ly49A antibody (A1), reversed the H-2Dd–mediated inhibition of the target cell killing by Ly49A+ LAK cells (Fig. 1 A). By contrast, negative control anti–H-2Dd α3 antibody (34-2-12S) F(ab′)2 fragments had no effect. We also investigated the effect of S19.8 antibody on the physical binding of H-2Dd–transfected C1498 cells to Ly49A-transfected CHO cells (Fig. 1 B). Anti–mouse β2m antibody, as well as anti–H-2Dd α1/α2 or anti-Ly49A antibody, completely abrogated the binding of H-2Dd–transfected C1498 cells to Ly49A-transfected CHO cells (Fig. 1 B). These results clearly demonstrate inhibition of the functional and physical interaction between H-2Dd and Ly49A by the anti-β2mb antibody S19.8 and suggest the possible involvement of β2m in Ly49A binding to H-2Dd.
Failure of Ly49A To Recognize H-2Dd Complexed with Human β2m.
β2m bound to MHC class I molecules on the cells in culture can be replaced by exogenously added β2m 31,32. It is conceivable that some MHC class I molecules on C1498 cells and their transfectants were associated with bovine β2m, because those cells were maintained in the presence of FCS. The observation that anti–mouse β2m antibody completely abrogated the recognition of H-2Dd by Ly49A (Fig. 1) suggested that H-2Dd complexed with β2m from bovine or other species might not be recognized by Ly49A. To investigate the species-specific involvement of β2m in Ly49A recognition of H-2Dd, we used human β2m to which a serological reagent was available. Incubation of H-2Dd–transfected mouse C1498 cells with human β2m induced expression of human β2m epitope detected by BBM1 antibody (Fig. 2 A) and decreased expression of mouse β2m epitope by 34% in MFI compared with H-2Dd–transfected C1498 cells cultured in the absence of human β2m (Fig. 2A and Fig. B). Incubation of H-2Dd–transfected C1498 cells with human β2m also increased expression of H-2Dd by 12.4%, consistent with the reported ability of human β2m to stabilize surface expression of mouse MHC class I molecules 33. These data indicate that a substantial number of H-2Dd molecules were complexed with human β2m. These cells were then tested for killing by Ly49A+ NK cells in the presence or absence of various antibodies (Fig. 2 C). To evaluate the interaction of Ly49A with H-2Dd complexed with human β2m, H-2Dd complexed with mouse β2m was masked by anti–mouse β2m antibody. Addition of anti–mouse β2m antibody Fab fragments as well as the anti–H-2Dd α1/α2 antibody F(ab′)2 fragments reversed the inhibition of Ly49+ NK cell–mediated killing of H-2Dd–transfected C1498 cells cultured in the presence of human β2m (Fig. 2 C). Further addition of anti–human β2m antibody Fab fragments did not show any effect. These data imply that H-2Dd complexed with human β2m is unable to inhibit killing by Ly49A+ LAK cells; however, a firm conclusion could not be obtained due to the incomplete exchange of human β2m for mouse β2m.
To further address this issue, we transfected β2m defective human cell line Daudi with human or mouse β2m together with H-2Dd heavy chain (Fig. 3A and Fig. B). We recently prepared sLy49A tetramer, which specifically binds H-2Dd (Matsumoto, N., K. Tajima, M. Mitsuki, and K. Yamamoto, manuscript submitted for publication). sLy49A tetramer bound Daudi cells transfected with mouse β2m and H-2Dd (Fig. 3 C) but not those transfected with human β2m and H-2Dd (Fig. 3 D), despite the equivalent level of H-2Dd expression on those cells (Fig. 3A and Fig. B). These results clearly demonstrate the species-specific ability of mouse β2m to support Ly49A binding to H-2Dd and suggest the direct involvement of β2m in the recognition of H-2Dd by Ly49A.
Disruption of Ly49A Recognition of H-2Dd by the Introduction of Three Individual Mutations in α1/α2 or α3 Domains of H-2Dd.
Previous observations indicate the importance of the α1/α2 domain, especially the NH2-terminal halves of the α1 and α2 regions of H-2Dd that form the bottom of the α1/α2 domain, in Ly49A recognition of H-2Dd 16,34. Moreover, our analysis indicates that β2m may be directly involved in the recognition. These observations prompted us to prepare a panel of Ala substituted mutants of H-2Dd and to explore their interaction with Ly49A. Residues substituted with Ala were chosen from solvent-exposed residues in α1/α2 or α3 domain based on the crystal structure of H-2Dd 35,36. These sites include the residues involved in hydrogen bonding between H-2Dd and Ly49A at the two putative interaction sites in the crystal structure of the Ly49A/H-2Dd complex reported by Tormo et al. during the course of this study 21.
C1498 mouse lymphoma cells were stably transfected with the mutant H-2Dd cDNA constructs. The transfectants with a similar level of H-2Dd expression, as assayed by staining with 34-2-12S or 34-5-8S antibodies, were selected for further analysis (Fig. 4). Most of the H-2Dd mutants tested here were equally reactive with both 34-2-12S and 34-5-8S antibodies (data not shown). However, substitution of Glu227 in the α3 domain of H-2Dd with Ala (E227A) disrupted the epitope recognized by 34-2-12S but not of 34-5-8S as reported by Connolly et al. 37. The panel of H-2Dd mutant transfectants was then assayed for binding of the sLy49A tetramer (Fig. 5). Individual Ala substitution of the residues Arg6, Asp122 in the α1/α2 domain of H-2Dd, or Lys243 in the α3 domain completely abrogated the ability of H-2Dd to bind sLy49A tetramer, whereas substitution of Arg111 partially inhibited the binding. The rest of the mutants had similar capacities to bind the sLy49A tetramer as wild-type H-2Dd. The same panel of H-2Dd mutants was also tested for the ability to protect tumor cells from killing by Ly49A+ LAK cells (Fig. 6). Introduction of R6A, D122A, or K243A mutation into H-2Dd completely abrogated the protective activity of H-2Dd against killing by Ly49A+ LAK cells. None of the other mutations tested here significantly impaired the ability of H-2Dd to protect C1498 cells from killing by Ly49A+ LAK cells (Fig. 6). Neither one of the H-2Dd mutants tested here nor wild-type H-2Dd protected C1498 cells from killing by Ly49A− LAK cells (data not shown). These results indicate that Arg6, Asp122, and Lys243 are essential for the physical binding of Ly49A and also for the functional binding of Ly49A that leads to inhibition of NK cell cytotoxicity. Importantly, these residues are involved in hydrogen bonding to Ly49A in one out of two binding sites in the crystal structure of the Ly49A/H-2Dd complex (Fig. 7; reference 21).
Our data clearly indicate the functional Ly49A binding site on H-2Dd that is associated with inhibition of NK cell cytotoxicity. The crystal structure of the Ly49A/H-2Dd complex defined two possible binding sites for Ly49A on a single H-2Dd molecule (site 1 and site 2 in Fig. 7; reference 21). Site 2 spans the three structural domains that constitute the MHC class I molecule, α1/α2, α3, and β2m. Several new lines of evidence indicate that site 2, rather than site 1, is the functional binding site for Ly49A. (a) 34-5-8S antibody, which recognizes the α1/α2 domain, inhibits the interaction between Ly49A and H-2Dd (Fig. 1 and Fig. 2; reference 13). We recently mapped the epitope of 34-5-8S to a region containing Glu104 and Gly107 of the H-2Dd heavy chain (Matsumoto, N., W. Yokoyama, S. Kojima, and K. Yamamoto, manuscript submitted for publication), which is located in the neighborhood of site 2 (Fig. 7). (b) The anti-β2mb antibody S19.8 completely inhibited the functional and physical interaction between Ly49A and H-2Dd (Fig. 1). S19.8 recognizes an epitope containing Ala85 and His34 of mouse β2m of b allele 38,39, which is also juxtaposed to site 2 (Fig. 7). (c) To bind H-2Dd, Ly49A required a complex of H-2Dd with mouse β2m, but not with human or bovine β2m (Fig. 1,Fig. 2,Fig. 3). On the other hand, H-2Dd complexed with rat β2m can bind Ly49A as shown by Sundback et al. 40. β2m contributes 25% of the binding surface in site 2 of the Ly49A/H-2Dd complex structure 21. Importantly, the surface of β2m has the species-specific residues found in mouse and rat β2m but not in human or bovine β2m (data not shown). In particular, in the crystal structure of the Ly49A/H-2Dd complex, site 2 encompasses the side chains of Lys3, Gln29, and Lys58 of mouse β2m. Of these, Lys3 and Gln29 are replaced with Arg and Gly, respectively, in human and bovine β2m but are conserved in rat β2m, suggesting significance of Lys3 and Gln29 in species-specific contribution of β2m to the interaction between Ly49A and H-2Dd. However, we cannot exclude the possibility that the replacement in other residues in human or bovine β2m forces H-2Dd heavy chain to have a different conformation from H-2Dd complexed with mouse or rat β2m. Mutation analysis is in progress to identify β2m residues that account for the species specificity. (d) Ala substituted mutation into any one of the residues of H-2Dd heavy chain, Arg6 and Asp122 in α1/α2 and Lys243 in α3, completely abrogated the physical and the functional interaction between H-2Dd and Ly49A (Fig. 5 and Fig. 6). Introduction of R111A mutation into α1/α2 of H-2Dd partially inhibited the ability of H-2Dd to bind Ly49A to such an extent that the effect was not evident in the functional protection assay (Fig. 5 and Fig. 6). Importantly, these four residues are involved in hydrogen bonding to Ly49A in site 2 in the crystal structure of the Ly49A/H-2Dd complex (Fig. 7; reference 21). Interestingly, not all of the Ala substitution of the residues that putatively disrupt hydrogen bonds between Ly49A and H-2Dd in site 2 resulted in loss in binding and function: E227A and E232A mutants of H-2Dd were fully functional in binding and protection from killing (Fig. 5 and Fig. 6). However, detailed examination of the crystal structure of the Ly49A/H-2Dd complex revealed that multiple hydrogen bonds are provided by each side chain of the residues of which Ala substitution completely (R6, D122, K243) or partially (R111) abolished the interaction between Ly49A and H-2Dd. By contrast, each side chain of the residues E227 and E232 forms only single hydrogen bond (data not shown). These differences could account for the observed absence in binding and functional effects of the E227A and E232A mutants. (e) Another possible Ly49A binding site on H-2Dd (site 1) was also revealed by the crystal structure of the Ly49A/H-2Dd complex 21. None of the Ala substitutions that were expected to disrupt hydrogen bonds between Ly49A and site 1 in the crystal structure showed any effect on sLy49A tetramer binding, or on the functional interaction of H-2Dd with Ly49A (Fig. 5,Fig. 6,Fig. 7). However, our data do not exclude the possibility that Ly49A interacts with H-2Dd through site 1 so weakly that the interaction was not detectable by sLy49A tetramer staining. The weak interaction through site 1 might be associated with cis interaction between Ly49A and H-2Dd on NK cells that leads to modulation of the Ly49A receptor as observed by Kåse et al. 43.
Collectively, our current results combined with the crystal structure of the Ly49A/H-2Dd complex 21 unambiguously indicate that the functional binding site of Ly49A lies in a concave region formed by the bottom of the α1/α2 and α3 domains, and β2m. These results also explain the previous observation that a single chain H-2Dd molecule, in which β2m is covalently linked to H-2Dd heavy chain through a peptide spacer, fails to interact with Ly49A 41. The peptide spacer is expected to cross in the middle of site 2, thereby interfering with the binding of Ly49A to site 2. Also, the current data provide a basis for the observation that the Ly49A binding site on MHC class I is distinct from the binding site for T cell receptor, which binds the top of the α1/α2 domain 42. Similarly, the identification of the binding site for Ly49A provides an understanding of other observations.
Functional binding of Ly49A to site 2 of H-2Dd well explains the observation that binding of Ly49A to H-2Dd requires the presence of peptide in the groove of H-2Dd α1/α2 domain 17,18. In the crystal structure of the Ly49A/H-2Dd complex, the side chains of Thr238 and Arg239 of Ly49A-1, which has a smaller contact area with H-2Dd than the other Ly49A subunit Ly49A-2 (Fig. 7), are hydrogen bonded to main chain carbonyl oxygens of Tyr85 and Asn86, respectively, both of which are located in the COOH-terminal end of α1 α-helix 21. The side chain of Ser192 of Ly49A-1 is hydrogen bonded to the amide groups of Met138 and Ala139, both of which are in the NH2-terminal end of α2 α-helix. Binding of a peptide to the peptide-binding groove of H-2Dd would bring two α-helices of H-2Dd to a position where Tyr85, Asn86, Met138, and Ala139 are available for hydrogen bonding to the residues in Ly49A-1. The notion that mutations in the peptide-binding groove of H-2Dd deteriorate the interaction between H-2Dd and Ly49A is remarkable in this context (Matsumoto, N., W. Yokoyama, S. Kojima, and K. Yamamoto, manuscript submitted for publication; references 45, 46). It is also noteworthy that the binding of Ly49C and Ly49I to H-2Kb and Kd, respectively, is restricted by the peptide bound to MHC class I 44,47. Interestingly, position 7, which is proximal to the end of the groove, of H-2Kb–bound peptide is critically involved in the peptide specificity of Ly49C binding to H-2Kb 47. Provided that Ly49C binds to a similar site on H-2Kb as Ly49A functionally binds H-2Dd (site 2), the peptide bound to H-2Kb might affect the conformation of the end regions of the α-helices and thereby influence the binding of Ly49C.
The functional Ly49A binding site on H-2Dd is located beneath the α1/α2 domain (Fig. 7), and it partially overlaps the CD8 binding site 48,49. CD8 has a stalk region of 30–50 residues, which is highly O-glycosylated 50, and its extended structure enables the unique Ig-like domain to reach MHC class I on the opposing target cell 49. Similarly, Ly49A has a stalk region of 68 residues, which connects the C-type lectin–like domain to the transmembrane domain and has three potential N-glycosylation sites 7,8. N-glycosylation on these sites might keep the stalk region of polypeptide in an extended conformation to enable the C-type lectin–like domain of Ly49A to reach the recognition surface beneath the α1/α2 domain of H-2Dd. It is of note that two of the three N-glycosylation sites were highly conserved among other Ly49 members (data not shown).
H-2Dd has two N-glycosylation sites at Asn86 and Asn176. Results from investigations on the role of carbohydrate moieties in recognition of H-2Dd by Ly49A were controversial 20,51,52. However, it was established that Ly49A does not require carbohydrates on H-2Dd to interact with H-2Dd 20,42. Because site 2 is located in the neighborhood of the N-glycosylation site at Asn86, the carbohydrate attached to this site might influence the Ly49A binding. We modeled H-2Dd with a high mannose-type glycochain on Asn86 by transplanting that from human CD2, of which dynamic structure including the carbohydrate moiety was determined by nuclear magnetic resonance 53 (Matsumoto, N., H. Iijima, and K. Yamamoto, unpublished data). The carbohydrate in any conformation found in CD2 is well accommodated in the interface of Ly49A at site 2. Moreover, the model raises the possibility that the carbohydrate might interact with the surface of Ly49A that corresponds to the carbohydrate-recognition surface found in typical carbohydrate-binding C-type lectins 54. This could account for the finding that optimal binding of H-2Dd–expressing cells to immobilized Ly49A is compromised by a sulfation inhibitor 51. The modeling of the H-2Dd with carbohydrate moieties on Asn86 also provides an insight into the stoichiometry of Ly49A binding to H-2Dd. From the model provided by the crystal structure of the Ly49A/H-2Dd complex, one Ly49A dimer could associate with two H-2Dd molecules. However, the carbohydrate modeled on Asn86 of H-2Dd occupied the space where the H-2Dd molecule that interacts through site 1 is supposed to fill. Therefore, the stoichiometry of binding of the Ly49A dimer to the N-glycosylated H-2Dd is postulated to be one to one; when a single Ly49A dimer on NK cells binds H-2Dd on target cells via site 2, the same Ly49A molecule would not be able to interact with MHC class I on NK cells via site 1.
Ly49A distinguishes polymorphic MHC class I molecules (Matsumoto, N., K. Tajima, M. Mitsuki, and K. Yamamoto, manuscript submitted for publication; reference 44). However, the critical residues identified in this study are conserved among mouse MHC class I molecules, including H-2Dd and Dk, which are ligands for Ly49A, and H-2Db, Kb, and Kd, which are not ligands for Ly49A (data not shown). Some of the polymorphic residues exposed on the surface of non-Ly49A ligand MHC class I that correspond to site 2 (data not shown) might determine the reactivity with Ly49A. In this context, we previously reported that NH2-terminal halves of α1 and α2 regions of H-2Dd are critically important for the recognition of H-2Dd by Ly49A by analyzing H-2Dd/Kd chimeric molecules 20. Sundback et al. 40 also reported the inability of the H-2Db α2 domain to support recognition of H-2Dd by Ly49A by exon shuffling between H-2Dd and Db. Mutational studies on H-2Dd revealed that polymorphic residues inside and outside of the peptide-binding groove affect the recognition of H-2Dd by Ly49A 45,46 (Matsumoto, N., W. Yokoyama, S. Kojima, and K. Yamamoto, manuscript submitted for publication). Of particular importance, we recently found that the substitution of polymorphic Asn30 of H-2Dd with Asp, which is found in non-Ly49A ligands Kb and Kd, partially abolished the functional as well as physical recognition of H-2Dd by Ly49A (Matsumoto, N., W. Yokoyama, S. Kojima, and K. Yamamoto, manuscript submitted for publication). The effect of the mutations in these sites may be conformational, since these residues identified by mutational studies are not found in the interface between Ly49A and H-2Dd that was functionally identified in this study (site 2). It should also be noted that a comparison of the crystal structure of MHC class I molecules revealed variation of the relative orientation of α3 domain or β2m to α1/α2 domain among mouse and human MHC class I molecules 36. Configuration of these domains might be critical for the interaction with Ly49A because Ly49A interacts with the surface of H-2Dd that spans the α1/α2 and α3 domains, and β2m in site 2 (Fig. 7). Therefore, the structural basis for MHC class I allele specificity of Ly49A remains to be examined in detail by site-directed mutagenesis.
The structure of the human Ig-like NK cell receptor KIR2DL2 in complex with its MHC class I ligand, HLA-Cw3, has been determined 55. Our analysis clearly demonstrated that Ly49A recognizes the surface of MHC class I that is distinct from the surface recognized by KIR2DL2. The interaction of KIR2DL2 with MHC class I is abrogated by individual amino acid substitutions of the residues in KIR2DL2 that disrupt hydrogen bonds between KIR2DLD and MHC class I 55. Similarly, the presence of hydrogen bonds critical for Ly49A/H-2Dd association was shown in our assays (Fig. 5 and Fig. 6). Thus, despite the structural difference and the difference in the binding sites on MHC class I, binding of the functionally similar receptors Ly49A and KIR2DL2 to MHC class I is critically mediated by hydrogen bonds and is very sensitive to individual disruption of hydrogen bonds.
The identification of the functional Ly49A binding site on H-2Dd provides a molecular basis for understanding the recognition of the MHC class I or related molecules not only by other members of the Ly49 family but also by other C-type lectin–like NK cell receptors, including HLA-E or Qa-1 recognition by CD94/NKG2A 56 and MIC-A, B, and RAE recognition by human and mouse NKG2D, respectively 57,58. While our studies predict that other C-type lectins, such as CD94/NKG2A, may interact with MHC class I–related molecules in the same way, several findings suggest that there may be differences. CD94/NKG2A is a heterodimer of two related chains with unique C-type lectin–like domains, whereas Ly49A is a homodimer. CD94/NKG2A recognizes the nonclassical MHC class I molecule HLA-E (Qa-1 in mouse), whereas Ly49A recognizes the classical MHC class I molecules like H-2Dd and Dk. Recognition of the MHC class I ligand by CD94/NKG2A is dependent on the sequence of the MHC class I–bound peptide 59,60, whereas Ly49A has no apparent specificity for MHC class I–bound peptide 17,18. CD94 and NKG2A have relatively short stalk regions, 28 residues in human CD94 and 24 residues in human NKG2A, compared with Ly49A, which has a stalk region of 68 residues. One might argue that the short stalk region is not compatible with the idea that CD94/NKG2A interacts with the similar site on HLA-E as Ly49A functionally interacts with H-2Dd. However, the stalk regions with 24–28 residues are able to stretch for at least 8 nm in an extended conformation and would be capable of placing C-type lectin–like domains for CD94/NKG2A to bind the similar site on HLA-E. Biochemical as well as structural studies on the interaction between other members of the C-type lectin–like NK cell receptors and their ligands are needed to show whether the similar sites on MHC class I or its related molecules are used as receptor binding interface.
Structurally based studies such as this work together with the recently resolved crystal structures of the Ly49A/H-2Dd and the KIR2DLD/HLA-Cw3 complexes 21,55 have unveiled the mode of recognition of MHC class I molecules by MHC class I–specific NK cell receptors of the two structurally different families. These studies are also important with respect to NK cell biology in general. The missing-self hypothesis predicts that NK cells monitor the expression of MHC class I molecules and kill the cells with aberrant expression of MHC class I associated with such events as tumorigenesis or infection 2. The structural studies suggest how NK cell receptors can sense the aberrant expression of MHC class I molecules, in addition to global loss of expression.
We thank Drs. R.K. Ribaudo and M.J. Shields for their helpful discussions and reagents, Dr. D.H. Margulies for the coordinates for the Ly49A/H-2Dd complex, and Dr. H. Iijima for his critical reading of the manuscript and modeling of the Ly49A complexed with N-glycosylated H-2Dd.
This work was supported by Grants-in-Aid for Scientific Research from the Ministry of Education, Science, and Culture of Japan (12672107), by a grant for Research on Health Sciences focusing on Drug Innovation from the Japan Health Science Foundation (12259), and by grants from the National Institutes of Health to W.M. Yokoyama, who is an Investigator for the Howard Hughes Medical Institute.
Abbreviations used in this paper: ADCC, antibody-dependent cellular cytotoxicity; β2m, β2-microglobulin; CHO, Chinese hamster ovary; MFI, mean fluorescence intensity; sLy49A, soluble Ly49A.
N. Matsumoto and M. Mitsuki contributed equally to this work.