Rodent Ly49 exhibit allele-specific MHC I recognition, yet the interaction site, site 2, encompassing the area below the MHC peptide-binding groove, the α3 domain, and associated β2 microglobulin, is highly conserved among rat and mouse MHC I alleles. We previously demonstrated that allele-specific Ly49 recognition can be affected by polymorphisms specifically in the peptide anchor-binding and supertype-defining B pocket of MHC I, possibly through differential conformations assumed by solvent-exposed interaction residues when articulating with this pocket. Through mutagenesis of RT1-A1c and H-2Dd, we map for the first time the interaction site(s) on rat MHC I mediating rat Ly49i2 recognition and the previously unexamined Ly49GBALB/c interaction with H-2Dd. We demonstrate that rat Ly49i2 and mouse Ly49G use both unique and common interactions at three MHC I H chain subsites to mediate functional binding and allele-specific recognition. We find that the F subsite, formed by solvent-exposed residues below the more conserved C-terminal anchor residue-binding F pocket, acts as an anchoring location for both Ly49i2 and Ly49G, whereas these receptors exhibit distinctive reliance on solvent-exposed residues articulating with the polymorphic anchor-binding and supertype-defining pocket(s) at subsite B, as well as on interaction residues at subsite C in the MHC I α3 domain. Our findings, combined with previous Ly49A/H-2Dd and Ly49C/H-2Kb cocrystal data, suggest how allele-specific MHC I conformations and Ly49 polymorphisms may affect Ly49 placement on MHC I ligands and residue usage at site 2, thereby mediating allele-specific recognition at the highly conserved MHC I interface.
Natural killer cells are large granular lymphocytes that play an important role in mediating innate resistance to transformed and virally infected cells, through both the secretion of soluble mediators and cytolytic activity (1). Interactions between NK cells and potential target cells occur through a myriad of activating and inhibitory receptors of lectin-like and Ig-related receptor families expressed on the NK cell surface (2). In murine species, such as the rat and mouse, these include members of the Ly49 lectin-like receptor family (3, 4, 5).
Ly49 receptors belong to polygenic families comprised of numerous members both in the mouse and the rat (6, 7). Inbred rat and mouse strains each have a characteristic Ly49 haplotype (8, 9, 10), and allelic variation between Ly49 has been identified between mouse strains (11, 12). Both activating and inhibitory Ly49 receptor types are found, with some activating/inhibitory hybrid receptors identified in the rat (7). Class I MHC (MHC I)3 is the ligand for most well-studied activating and inhibitory murine Ly49 (4, 13). Additionally, the viral MHC I homologue, m157 from mouse CMV, is a ligand for the mouse-activating Ly49H and inhibitory Ly49I receptors (14).
Recognition of MHC I by Ly49 is allele specific, with each Ly49 receptor able to recognize only one or a small repertoire of MHC I alleles. In inbred rodent strains, each NK cell expresses only a small subset of the available genes in its Ly49 haplotype (15), but often at least one inhibitory Ly49 capable of recognizing an autologous MHC I allele product, which then transmits a dominant suppressive signal to the NK cell, thereby maintaining NK cell tolerance to self (16). Loss or lack of self-MHC I (missing self) through viral infection or transformation would then release the NK cell from this dominant suppression, precipitating NK effector functions toward the target cell (17). Human killer Ig-related (KIR) receptors, although structurally unrelated to rodent Ly49, are functionally equivalent in detecting missing self, demonstrating variegated expression on NK cells, and displaying MHC I allele specificity (18, 19). Ly49 and KIR resemble each other in another general sense, in that they display MHC I promiscuity, yet definitive allele specificity. This differs from the TCR on mature T cells that is normally restricted to recognizing Ag presented by a single MHC allele product.
The MHC I specificities of KIR2DL isoforms are for one or the other of two groups of HLA-C allele products (18). The two specificities are dictated by which polymorphic residues at positions 77 and 80 the HLA-C allele encodes (20). Through mutagenesis studies and examination of KIR/MHC I cocrystal structures, it has been demonstrated that KIR bind MHC I on the top surfaces of the α1 and α2 helices of MHC I, and this principally involves interaction of polymorphic KIR residues with HLA-C residues 77 and 80 (21, 22). The molecular interaction between mouse Ly49 and MHC I has also been studied by mutagenesis (23, 24, 25, 26, 27, 28, 29) and examination of cocrystal structures of Ly49A/H-2Dd and Ly49C/H-2Kb (30, 31), and occurs primarily at a large interface known as site 2 that encompasses MHC I residues below the peptide-binding groove, within the α3 domain, and on the β2-microglobulin (β2m). What has remained poorly understood is the mechanism by which this site 2 interface, which is highly conserved both within and between rat and mouse MHC I alleles, can engender the observed allele specificity of Ly49 receptors.
Recently, we demonstrated that one of the polymorphic pockets within the peptide-binding groove of MHC I, which serves to secure the peptide in the groove through specific anchor residues in the peptide, and which defines the supertype of the MHC I molecule, can determine allele-specific recognition of MHC I (32). A reliance on supertype-defined conformations for allele specificity would allow NK cells to survey for the presence of MHC I supertypes on the cell surface. About nine supertypes are found in species as diverse as rodents and primates, with similar supertypes presumably arising through convergent evolution (33, 34), and which appear critical for CTL recognition and function (35, 36).
To investigate whether allele-specific Ly49 differentially interact with solvent-exposed residues at site 2, possibly due to conformations influenced by anchor-binding and supertype-defining pockets, and to investigate whether rat Ly49 recognition of MHC I occurs in a similar manner to mouse, we examined both a rat and a mouse Ly49/MHC I allele-specific receptor and ligand combination, creating mutations of solvent-exposed residues at site 2 that articulate directly with the anchor-binding pockets of MHC I. Because cocrystal structures of Ly49A with H-2Dd (30) and Ly49C with H-2Kb (31) predict that both of these Ly49 interact with different residues in the α3 domain of their MHC I ligands, we also mutated α3 domain residues to determine whether conformational differences in MHC I might also affect how Ly49 are oriented in relation to the α3 domain, generating altered dependency on residues in this domain for recognition.
Our data lead us to present a model in which three H chain subsites, at the highly conserved site 2 of MHC I, function in combination to determine rat and mouse Ly49 binding and allele specificity for MHC I. The F subsite, which involves solvent-exposed residues articulating with the more conserved anchor-binding F pocket, acts in both species as an anchor point and interacts with highly conserved residues in Ly49. The B subsite, involving residues that articulate with the polymorphic, supertype-defining, and N-terminal peptide anchor-binding pocket(s) on MHC I, directly interacts with the polymorphic L3 loop of Ly49, but may additionally contribute to allele specificity by altering placement of the Ly49 at site 2, most likely owing to variable β2m orientation. Altered positioning at site 2 may make residues in the α3 domain (C subsite) of MHC I more or less available for interaction with polymorphic β4-β5 and β2-β2′ loops in Ly49. This model appears to apply in both the rat and mouse, and suggests convergent evolution in these two species to produce innate recognition systems possibly capable of scrutinizing cells for the presence of MHC I supertypes, which are so crucial to CTL function and adaptive immunity.
Materials and Methods
Hybridomas and mAbs
Hybridomas producing the Abs 4D11 (rat IgG2a), anti-Ly49GBALB/c (37), 34-5-8S (IgG2a), and anti-H-2Dd were obtained from American Type Culture Collection. Abs were prepared from ammonium sulfate precipitates, as described (38). Purified STOK2 (rat IgG2a), anti-Ly49i2 Ab (39, 40) was purchased from BD Pharmingen. RT1-A1c-specific YR5/12 (rat IgG2b) hybridoma supernatant (41) was purchased from Serotec. The R-PE-conjugated secondary Abs, AffinPure F(ab′)2 donkey anti-rat IgG, and AffinPure F(ab′)2 donkey anti-mouse IgG were purchased from Jackson ImmunoResearch Laboratories.
YB2/0, a nonsecreting rat myeloma, was obtained from American Type Culture Collection and was maintained in DMEM supplemented with 10% FCS, l-glutamine, penicillin, streptomycin, 1 mM sodium pyruvate, and 0.1 mM nonessential amino acids. RNK-16, a spontaneous F344 rat strain NK cell leukemia cell line (42), was maintained in RPMI 1640 supplemented with 10% FCS, l-glutamine, penicillin, streptomycin, and 5 × 10−5 M 2-ME. Ly49i2 and Ly49GBALB/c transfectants of RNK-16 have been described previously (32). All transfected RNK-16 and YB2/0 cells were maintained under G418 selection. Transfected cells were grown in the absence of G418 for at least 48 h before cytotoxicity assays.
Mutagenesis of MHC I
RT1-A1c, previously cloned in this laboratory from PVG rat spleen (43) and H-2Dd from S49.1 T lymphoma cells (38), were mutated using a QuikChange Mutagenesis Kit (Stratagene). Mutagenic primers were designed to mutate to alanine, potential solvent-exposed interaction residues directly below or near the B pocket (residues 6, 8, 10, 23, 27, and 98), below the F pocket (residues 115 and 122), and in the α3 domain of the MHC I H chain (residues 223, 232, 243, 262) as defined by Ly49A/H-2Dd and Ly49C/H-2Kb cocrystals (30, 31). Double mutants in the B subsite (R6A F8A), F subsite (Q115A D122A), C subsite (E232A Y262A, D223A K243A, E232A K243A, Y262A K243A), and between subsites (R6A D122A, R6A K243A, D122A K243A) were generated through an additional round of mutagenesis on RT1-A1c constructs already containing the primary mutation. All mutations were verified by DNA sequencing. Mutagenesis was performed on RT1-A1c directly within the H-2Dk leader-enhanced GFP (EGFP) fusion vector that was previously generated in this laboratory (43). H-2Dd was subcloned into the EGFP fusion vector with its native leader sequence intact and mutagenesis performed directly on this construct. YB2/0 cells were stably transfected with each construct, as previously described (38).
Flow cytometric analysis
Successful transfection of MHC I/EGFP fusion constructs was determined by EGFP fluorescence intensity of transfected YB2/0 cells relative to untransfected YB2/0 cells in 96-well black-walled plates using a FLA-5100 Imaging System (Fujifilm). Proper folding and surface expression levels were determined using specific primary Ab to RT1-A1c (YR5/12) and H-2Dd (34-5-8S) plus relevant PE-conjugated secondary Ab. Expression of the Ly49i2 and Ly49GBALB/c receptors by RNK-16 cells was monitored compared with untransfected RNK-16 cells using primary Abs STOK2 and 4D11, respectively, followed by FITC-conjugated relevant secondary Ab. Surface expression was assessed using a FACScan flow cytometer (BD Biosciences). Target cells were assayed on the same day as cytotoxicity assays for matched surface expression of mutant and control MHC I.
Target cells were labeled with 100–150 μCi of Na51CrO4 (51Cr) (Mandel) at 37°C for 1 h. After washing targets three times with RPMI 1640, they were plated at 1 × 104 51Cr-labeled cells/well and mixed with RNK-16 or transfected RNK-16 cells in V-bottom microtiter plates at indicated E:T ratios in triplicate. After a 4-h incubation at 37°C, plates were centrifuged for 5 min at 1500 rpm, and 25 μl of supernatant was collected and counted in a MicroBeta TriLux liquid scintillation counter (PerkinElmer). Percent specific lysis was determined as (experimental release – spontaneous release)/(maximal release – spontaneous release) × 100.
Ly49G recognition of H-2Dd is disrupted by mutagenesis of single residues articulating with anchor-binding pockets at subsite B, whereas Ly49i2 recognition of RT1-A1c is unaffected
Solvent-exposed residues at site 2 of the MHC I molecule are highly conserved between MHC I alleles in both mouse and rat, and even between the two species, making it difficult to understand how Ly49 receptors exhibit allele specificity for MHC I when interacting with this site. Recently, we demonstrated that the polymorphic, P2 peptide anchor residue-binding and supertype-defining B pocket is important in determining Ly49 allele specificity for MHC I (32). We proposed that this may occur through polymorphism-induced conformational alterations of side chains on solvent-exposed amino acids below and articulating with the B pocket, thereby altering their availability for interaction with Ly49. In support of this, Ly49 recognition has been shown to be reliant on R6 of MHC I, a residue that lies below and articulates directly with the B pocket, in a number of allele combinations (24, 32). These observations suggested that allele-specific MHC I recognition by Ly49 may involve variable usage of solvent-exposed MHC I residues below and articulating with the polymorphic B (or N-terminal peptide anchor-binding) pocket(s), at what we have designated as the B subsite (Fig. 1 A).
When Dam et al. (31) solved the cocrystal of H-2Kb and Ly49C, they showed that the Ly49C/H-2Kb interface, although having some Ly49/MHC I residue interactions in common with the Ly49A/H-2Dd cocrystal interface (30), also involves a unique set of Ly49/MHC I residue interactions to mediate recognition. This appears to be due to both a different orientation of Ly49C on H-2Kb and polymorphisms in Ly49C. The unique orientation of Ly49C on H-2Kb seemed to be due to a different mode of dimerization for Ly49C compared with Ly49A. In the Ly49C cocrystal, the Ly49C dimer was more open, allowing one Ly49C receptor to symmetrically bind two H-2Kb molecules simultaneously, whereas in the Ly49A cocrystal, Ly49A dimerization was more closed, resulting in an asymmetrical binding orientation on a single H-2Dd molecule (25, 26). Ly49A has since been seen to assume a similar open dimerization state, according to nuclear magnetic resonance solution studies, capable of binding two H-2Dd molecules simultaneously (44), making variable Ly49 dimerization states unlikely mediators of allele specificity. Regardless, variable placement of the Ly49 receptor on the MHC I molecule and allele-specific residue usage at the B subsite and two other subsites that we have designated on the MHC I H chain (Fig. 1, A and B) may contribute to allele specificity at the highly conserved site 2 of MHC I. These two additional subsites include subsite C, involving interaction residues on the α3 domain of the MHC I H chain and subsite F, involving solvent-exposed residues beneath and articulating with the anchor-binding F pocket, which is typically more conserved in amino acid composition than the B pocket (Fig. 1, A and B). We set out to investigate the usage of residues at each subsite by two different Ly49/MHC I allele combinations, for comparison with data previously collected on the Ly49A/H-2Dd and Ly49C/H-2Kb allele combinations, to determine whether differential or distinctive MHC I residue usage at one or more subsites could mediate MHC I allele-specific interaction with Ly49 receptors. One combination, H-2Dd and the BALB/c allele of Ly49G2, allowed us to compare residue usage at each subsite by a Ly49 that is related to, but distinct from Ly49A, and exhibits similar specificity for the H-2Dd allele. The other allele combination included a classical rat MHC I molecule, RT1-A1c with the rat Ly49i2 inhibitory receptor, and allowed us to map the residues mediating MHC I recognition by a Ly49 in this species. RNK-16 cells transfected to express the rat Ly49i2 or mouse Ly49G inhibitory receptors (Fig. 1,C) were used as effector cells toward YB2/0 target cells and YB2/0 transfected to express their respective MHC I ligands, RT1-A1c and H-2Dd (Fig. 1,D) in 4-h cytotoxicity assays (Fig. 1,E). RNK-16 effector cells expressing the inhibitory Ly49i2 or Ly49G receptors lyse untransfected YB2/0 cells, but are inhibited from killing transfected YB2/0 targets through recognition of their respective RT1-A1c and H-2Dd MHC I ligands (Fig. 1 E), as previously reported (12, 40).
Alterations of the polymorphic and supertype-defining B pocket in MHC I disrupt Ly49 recognition (32), possibly by altering the conformation of solvent-exposed resides that specifically articulate with the N-terminal peptide anchor-binding pocket(s) of MHC I. We therefore mutated to alanine, individual solvent-exposed residues that articulate directly with the anchor-binding B pocket of RT1-A1c (R6, F8) (Fig. 2,A) and the anchor-binding B and C pockets of H-2Dd (R6, F8, M98) (Fig. 2,B), along with residues that may also contribute to Ly49 interaction below the peptide-binding groove, but that do not directly articulate with anchor-binding pockets. For RT1-A1c, these residues included I10, I23, Y27, and M98 (Fig. 2,A) and for H-2Dd, residues T10, M23, and Y27 (Fig. 2,B). Because both the B and C pockets act as the supertype-defining and anchor-binding pockets of H-2Dd, we have grouped/designated solvent-exposed residues below and articulating with these pockets of H-2Dd as the B subsite. Each mutant was expressed on YB2/0 cells to similar levels compared with wild-type RT1-A1c (Fig. 3,A) and H-2Dd (Fig. 3,B). RT1-A1c, and each of its mutants, exhibited a similar ability to inhibit cytolysis of YB2/0 through Ly49i2 recognition of the transfected MHC I molecules (Fig. 3,C). In contrast, Ly49G recognition of H-2Dd was greatly reduced by the R6A mutant, as previously seen for Ly49A recognition of this molecule (24), and was completely disrupted by the F8A and M98A mutations of H-2Dd (Fig. 3,D). These three residues lie directly beneath and normally articulate with the anchor-binding B and C pockets of H-2Dd (Fig. 2,B). The Y27A mutant of H-2Dd exhibited a partial loss in ability to inhibit cytolysis in repeated assays (Fig. 3,D and data not shown), possibly reflecting its position below, but on the periphery of the B pocket (Fig. 2,B), and therefore a less crucial interaction residue at the B subsite. Mutagenesis of residues T10 and M23, which are positioned in the middle of the peptide-binding platform and away from the anchor-binding B and C pockets, showed no effect on Ly49G recognition of H-2Dd, as would be expected if only residues below and articulating with the anchor-binding pockets were required for recognition at this subsite (Fig. 3,D). In summary, Ly49i2 recognition of RT1-A1c and Ly49G recognition of H-2Dd differ in their reliance on individual solvent-exposed residues at subsite B with Ly49i2 recognition being resistant to single mutations at this subsite, despite a demonstrated reliance on the B pocket for recognition (32), and Ly49G recognition being specifically sensitive to alterations in individual residues directly below and articulating with the anchor-binding B and C pockets of H-2Dd (Fig. 2).
Ly49i2 and Ly49G both exhibit sensitivity to alterations in solvent-exposed residues below and articulating with the F pockets of RT1-A1c and H-2Dd at subsite F
Because we have shown that allele-specific Ly49 recognition of MHC I appears to specifically rely on residues both within (32) and articulating with the anchor-binding and supertype-defining B pocket, the latter at least for H-2Dd recognition by Ly49G (Fig. 3,D), and because Ly49A/H-2Dd and Ly49C/H-2Kb recognition is apparently highly dependent on the solvent-exposed residues Q115 and D122 (24, 30, 31) located below the other peptide-anchoring F pocket at what we designated the F subsite, we investigated whether Q155 and D122 might also play a significant role in rat Ly49i2 and mouse Ly49G recognition of MHC I. RT1-A1c was mutated to alanine at Q115 and D122 (Fig. 2,A), and each mutant was expressed to a similar level as wild-type RT1-A1c on YB2/0 target cells (Fig. 4,A). H-2Dd was also mutated at residues Q115 and D122 (Fig. 2,B), but only the Q115A mutant could be expressed to comparable wild-type H-2Dd levels (Fig. 4,A and data not shown). The inability to express the D122A mutant was most likely due to the previously demonstrated reliance of H-2Dd on D122 for association with the peptide-loading complex (45). The Q115A mutant of RT1-A1c showed no loss of recognition by Ly49i2 compared with wild-type RT1-A1c (Fig. 4,B), whereas the single D122A mutant showed partial disruption of recognition by Ly49i2 in repeated assays using different clones, all with equivalent expression levels to wild-type RT1-A1c (Fig. 4,B and data not shown). In contrast, Ly49G recognition of H-2Dd was nearly completely disrupted through mutagenesis of the Q115 residue at subsite F and exhibited similar levels of cytolysis to untransfected YB2/0 cells in the same assay (Fig. 4,C). In contrast to the B subsite, it is evident that both Ly49i2 and Ly49G, similar to Ly49A and C, rely on solvent-exposed residues that articulate with the more conserved anchor-binding F pocket, at the F subsite, for recognition of MHC I, with Ly49G recognition of H-2Dd again exhibiting greater sensitivity to a single residue alteration at this subsite compared with Ly49i2 recognition of RT1-A1c (Fig. 2).
Specific single amino acid changes at subsite C in the MHC I α3 domain can completely abolish Ly49G recognition of H-2Dd, whereas Ly49i2 recognition of RT1-A1c is relatively unaffected
The cocrystal structures of Ly49C/H-2Kb (31) and Ly49A/H-2Dd (30) predict that each of these Ly49/MHC I allele combinations interacts with a different set of residues in the highly conserved α3 domain of MHC I. Therefore, we designated this as another potential subsite, subsite C, for the discrimination of MHC I allele products by Ly49. We mutated H-2Dd and RT1-A1c at amino acid positions corresponding to those previously predicted by cocrystal structures as being involved in either Ly49A recognition of H-2Dd (E232, K243) or Ly49C recognition of H-2Kb (E223, K243, Y262) (30, 31) to assess whether Ly49i2 and Ly49G recognition of their respective MHC I ligands also depended on the same or different sets of residues at subsite C. Each mutant in RT1-A1c (D223A, E232A, K243A, and E262A) (Fig. 2,A) was expressed in YB2/0 cells at similar cell surface levels compared with wild-type RT1-A1c, as were the H-2Dd C subsite mutants (E223A, E232A, K234A, E262A) (Fig. 2,B) compared with wild-type H-2Dd (Fig. 5,A). Cytolysis by RNK-16 cells expressing Ly49i2 remained inhibited through recognition of each RT1-A1c single mutant at the C subsite, compared with wild-type RT1-A1c expressed on YB2/0 cells. A possible exception was YB2/0 cells expressing the E232A mutant, which consistently showed a modest increase in cytolysis compared with wild-type RT1-A1c (Fig. 5,B). In contrast, Ly49G recognition of H-2Dd showed greater sensitivity to single mutants at this subsite. RNK-16 cells expressing Ly49G had continued, although slightly reduced, recognition and cytolytic inhibition of YB2/0 cells expressing the E223A and E262A H-2Dd C subsite mutants, compared with wild-type H-2Dd in the same assay (Fig. 5,C). However, Ly49G recognition of H-2Dd was totally disrupted by mutagenesis of either E232 or K243 residues to alanine (Fig. 5,C). Therefore, H-2Dd recognition by Ly49G shows a different and greater sensitivity to single mutations at this subsite compared with RT1-A1c recognition by Ly49i2, with Ly49i2 recognition of RT1-A1c slightly disrupted by mutagenesis at E232 and Ly49G recognition of H-2Dd slightly impaired through E223A and E262A mutations, and completely disrupted by mutations at E232 and K243 (Fig. 2).
Ly49i2 recognition of RT1-A1c is sensitive to combined alanine mutations at two different subsites
Because Ly49i2 recognition of RT1-A1c showed only modest or no disruption by mutagenesis of single residues at each of the B, F, and C subsites, there was the possibility that rat Ly49/MHC I interactions do not rely on the same interaction site or subsites as mouse Ly49/MHC I. Based on our previous results showing that Ly49i2 recognition of RT1-A1c relies on residues within the anchor-binding B pocket (32) and the partial loss of recognition seen with the D122A (Fig. 4,B) and E232A mutants (Fig. 5,B), it seemed unlikely that Ly49i2 recognition relied on entirely different subsites on MHC I compared with mouse Ly49. Therefore, we made combined alanine mutations in RT1-A1c at two different subsites. We chose the residues R6, D122, and K243 that Matsumoto et al. (24) previously mutated in H-2Dd, demonstrating that Ly49A recognition occurs at site 2, because these residues are found at subsites B, F, and C, respectively (Fig. 2,A). The three double mutants generated, R6A D122A, D122A K243A, and R6A K243A, were expressed on YB2/0 cells to similar levels as wild-type RT1-A1c (Fig. 6,A). The R6A D122A mutant, which represents mutations at both the B and F subsites, completely disrupted recognition by Ly49i2 (Fig. 6,B). This indicated that the partial loss of recognition seen with the single D122A mutant at subsite F (Fig. 4,B) could be augmented through mutagenesis of the B subsite that previously appeared to have no effect. Similarly, pairing mutations at the F and C subsites as with the D122A K243A mutant also produced an augmentation of the partial loss of recognition seen with the single D122A mutant (Fig. 4,B) and resulted in complete loss of recognition (Fig. 6,B), indicating that, despite the ineffectiveness of single mutations at the C subsite, this subsite indeed plays a role in the recognition of RT1-A1c by Ly49i2. The most compelling evidence for the importance of B and C subsites in mediating recognition by Ly49i2 came with the B/C subsite mutant R6A K243A, which was also able to disrupt recognition of RT1-A1c (Fig. 6,B) despite the inability of single mutations at either subsite to disrupt recognition individually (Figs. 3,C and 5,B). Therefore, rat Ly49i2 recognition, like mouse Ly49 recognition of MHC I, also occurs at site 2 and relies on the B, F, and C subsites in combination, for recognition of RT1-A1c (Fig. 6 C).
Ly49i2 recognition of RT1-A1c is sensitive to double alanine mutations at each individual subsite
Having established that Ly49i2 recognition of RT1-A1c, like mouse Ly49 recognition of MHC I, also relies on site 2 and the same B, F, and C subsites, we wanted to assess more fully the usage of residues at each individual subsite for comparison with the information already gathered for mouse allele products. Because we were able to detect partial loss of Ly49i2 recognition using single mutations at subsites F and C of RT1-A1c (Figs. 2,A, 4,B, and 5,B) and because concurrent mutations at two different subsites were able to disrupt recognition (Fig. 6), it appeared that Ly49i2, although relying on similar subsites as mouse Ly49, might interact with greater overall affinity for its MHC I ligand. Therefore, to assess more fully the usage of residues at each subsite, we produced double alanine mutations at each subsite of RT1-A1c that could possibly then overcome this evidently greater affinity of Ly49i2 for its ligand, providing additional information on the usage of each individual subsite on RT1-A1c in Ly49i2 recognition (Fig. 7 A).
At the B subsite, we mutated residues R6 and F8 (Fig. 7,A) because both of these residues strongly affected Ly49G recognition of H-2Dd (Fig. 3,D), and because both articulate closely with the anchor-binding B pocket of the RT1-A1c molecule. Additionally, the R6 mutation was able to disrupt recognition in combination with mutants at other subsites (Fig. 6,B). At the F subsite, we chose D122 because it has a partial effect as a single mutant in RT1-A1c (Fig. 4,B), and combined it with Q115 (Fig. 7,A), which disrupted Ly49G recognition of H-2Dd (Fig. 4,C). Lastly, at the C subsite, we chose mutant combinations that would either affect a Ly49A- or Ly49G/H-2Dd-type interaction (30) that uses residues E232A and K243A in the α3 domain for recognition (Fig. 7,B) or a Ly49C/H-2Kb-type interaction (31) in which residues E223A and Y262A, and not E232, act in addition to K243 to mediate recognition (Fig. 7,C). Each subsite double mutant was expressed on YB2/0 cells to similar cell surface expression levels as wild-type RT1-A1c (Fig. 8,A) and used as targets in 4-h cytotoxicity assays, with RNK-16 effector cells expressing the Ly49i2 inhibitory receptor (Fig. 8,B). The B subsite continued to show a reduced importance for recognition of RT1-A1c by Ly49i2 because YB2/0 cells expressing the B subsite double mutant, R6A F8A, showed only a partial loss of recognition compared with YB2/0 cells expressing wild-type RT1-A1c in the same assay (Fig. 8,B). In contrast, at the F subsite, the partial disruption caused by the single D122A RT1-A1c mutant seen in Fig. 4,B was augmented by the addition of the Q115A mutation, that alone had no effect (Fig. 4,B), resulting in complete loss of recognition by Ly49i2 expressing RNK-16 cells (Fig. 8 B).
Lastly, RT1-A1c double mutants at the C subsite also demonstrated a role for this subsite in recognition by Ly49i2 (Fig. 8,B). In examination of the Ly49C/H-2Kb interaction-type mutants, D223A E262A and D223A K243A, it appeared that Ly49i2 did not use this type of interaction for recognition of RT1-A1c, because YB2/0 cells expressing both of these C subsite double mutants were recognized to a similar degree as wild-type RT1-A1c-expressing YB2/0 cells (Fig. 8,B). Expression of the Ly49A,G/H-2Dd interaction-type mutant, E232A K243A, on YB2/0, resulted in a loss of recognition by Ly49i2-expressing RNK-16 cells, suggesting that the Ly49i2 receptor may follow a similar pattern as Ly49A and G for recognition of MHC I at the C subsite (Fig. 8,B), and is substantiated by the partial loss of recognition of the single E232A mutant seen in Fig. 5,B. Further investigation with the Ly49C/H-2Kb interaction-type mutant, E262A K243A, revealed that Ly49i2 also relies on this residue combination for recognition at the C subsite, because YB2/0 cells expressing this mutant were lysed at similar levels to untransfected YB2/0 cells (Fig. 8,B). Therefore, Ly49i2 recognition of RT1-A1c appears to have a Ly49A/G and Ly49C hybrid-type dependence on residues at subsite C (Fig. 7). The use of double mutants at each subsite reveals that like the mouse Ly49/MHC I allele combinations, rat Ly49i2 also interacts with MHC I at site 2 and uses a combination of residues at the B, F, and C subsites in recognition of the RT1-A1c MHC I allele product (Fig. 7 A).
In this study, we performed extensive MHC I mutagenesis coupled with receptor-mediated NK cell inhibition assays to discern for the first time the molecular basis of interaction of an MHC I allele-specific rat Ly49 with its MHC I ligand (Ly49i2 and RT1-A1c) and a new mouse Ly49/MHC I allele combination, Ly49GBALB/c and H-2Dd. We focused on site 2, the known interaction site for Ly49 binding to mouse MHC I and, in particular, the potential contribution of individual site 2 subsites, corresponding to residues in the α3 domain or areas of solvent-exposed residues that articulate with either the N-terminal (B or C pocket) or C-terminal (F pocket) peptide anchor-binding pockets. We designated these subsites C, B, and F, respectively. For the mouse Ly49G/Dd interaction, all three subsites were found to contribute, with single amino acid changes at any of the three MHC I subsites disrupting receptor recognition. Importantly, we found that the rat Ly49i2 interaction with its MHC I ligand, like the mouse Ly49G/Dd combination, also involves all three MHC I subsites, although this was only demonstrated with double mutations at an individual subsite or combined single mutations at two subsites, suggesting that the Ly49i2/A1c interaction may be of higher affinity than that of the mouse combination. With an evolutionary distance of 20–40 million years between rats and mice, and the use of the same site 2 subsites by the nonorthologous mouse and rat Ly49 when interacting with their nonorthologous MHC I ligands, these results suggest the possibility of convergent evolution. Simultaneous interaction at multiple subsites by Ly49 receptors in both species may be advantageous in providing identity and/or flexibility in determining MHC I allele specificity.
Our analysis of mutant MHC I recognition by Ly49i2 and Ly49G combined with previous mutagenesis and crystallographic data available for Ly49C recognition of H-2Kb (31, 46, 47, 48) and Ly49A recognition of H-2Dd (23, 24, 25, 27, 30, 49) allowed us to compare the usage of solvent-exposed residues at subsites on MHC I in four different Ly49/MHC I allele combinations. The subsite F residues Q115 and D122 are shared by rat RT1-A1c and mouse MHC I, including H-2Dd and Kb. From the cocrystal structures of Ly49A and C bound with H-2Dd and Kb, respectively, these residues contribute very significantly to Ly49/MHC I interaction, particularly D122, which, for example, contributes 23 atomic contacts between H-2Dd and Ly49A. The MHC I D122 interacts with a conserved 236, 238, and 239 triad in mouse Ly49 (Fig. 9, A and B), and Dam et al. have suggested that these residues provide the majority of free energy to the Ly49/MHC I interaction (31), which is supported by Ly49A mutagenesis studies. The importance of subsite F for mouse Ly49/MHC I interaction is further supported by our results in this study, demonstrating that mutation at Q115 of H-2Dd abolished its interaction with Ly49G. Furthermore, the same or similar 236, 238, 239 Ly49 triads would be predicted for many rat Ly49, including Ly49i2, and our results showed that Ly49i2 recognition RT1-A1c was most strongly reliant on subsite F, because the D122A mutant of RT1-A1c was the only single mutant able to produce a significant loss in recognition. Our findings support the concept that subsite F of MHC I is an essential site for anchoring Ly49 interaction, and extend it to include interactions of a rat Ly49 with its MHC I ligand.
Of interest is the potential contribution of different site 2 subsites to the MHC I allele specificity of Ly49 interaction. Because subsite F conformations are very similar, if not identical, on MHC I molecules, and subsite interactions with various Ly49 are highly similar and occur with every Ly49/MHC I pair examined, this makes it unlikely that this subsite will provide a significant contribution to allele specificity. Although subsite interactions are a common and a minimum requirement for stable Ly49 binding, subsites B and C may determine allele specificity. Conformational differences at subsites B and C relevant to allele-specific interaction could be established primarily through the highly polymorphic anchor pockets in MHC I that bind the N-terminal peptide anchor residues, and we have already shown in the case of RT1-A1c, this can control MHC I allele specificity (32). This would not only alter the direct availability of solvent-exposed residues below the pocket(s) for interaction at subsite B, but may variably orient β2m (31, 50, 51), altering the placement of polymorphic Ly49 loops on the α3 domain of MHC I, and possibly on β2m itself. Such a mechanism would put particular emphasis on the polymorphic L3, β2-β2′, and β4-β5 regions of Ly49 (Fig. 9, A and B) in mediating allele specificity, whereas the conserved 236, 238, 239 triad of Ly49 (Fig. 9, A and B) would provide a stable interaction with the more conserved F pocket and solvent-exposed residues of the F subsite. In support of this model, mutagenesis of subsite B residues F8 and M98, residues that articulate specifically with both the N-terminal anchor-binding pockets of H-20d and B2m and therefore could affect B2m orientation (Fig. 9,C), strongly affects by 49G recognition of H-20d. Furthermore, our results with Ly49i2 and Ly49G combined with previous findings for Ly49A and C indicate that each of the four Ly49 uses a different combination of residues at subsite C of MHC I for recognition, suggesting differential placement of Ly49 on the α3 domain of MHC I at subsite C. These findings, combined with the different conformations that the same residues, such as E232 and E223, at subsite C of different MHC I allele products can in some cases assume (Figs. 2 and 7), provide support for subsite C contributing to allele specificity of Ly49 interaction. Furthermore, our previous findings showed that substituting a sequence within the β4-β5 loop of Ly49W for that of another Ly49 receptor, Ly49P, allows the mutant Ly49P to recognize a new MHC I allele product (43, 52). Because the β4-β5 loop of Ly49P and W, similar to Ly49A (Fig. 9), are predicted to interact at subsite C, this is entirely consistent with subsite C contributing to allele specificity of Ly49 interaction.
Our mutagenesis results indicate the importance of individual solvent-exposed residues of H-2Dd that articulate with the B pocket of the peptide-binding groove, such as R6, F8, and M98 in Ly49G recognition, and not residues below other pockets between the B and F pockets, such as T10 and M23. Furthermore, a double mutation of R6 and F8 in RT1-A1c also reduced Ly49i2 recognition. These findings, combined with our previous results indicating that alteration of the residues forming the B pocket in RT1-A1c affect MHC I allele-specific Ly49 recognition, suggest the possibility that bound peptide, particularly the anchor residue in the B pocket, may affect the conformation of solvent-exposed residues at subsite B, and possibly β2m orientation (Fig. 9 C), which may both influence Ly49 recognition. The influence of different peptide anchor residues bound in the B and/or C pocket of H-2Dd or the B pocket of RT1-A1c on allele-specific Ly49 interaction may be informative.
A potentially interesting comparison is how KIR of the Ig superfamily distinguish between MHC I allele products and how Ly49 of the unrelated lectin-like receptor family may achieve this same function. In the case of KIR2DL isoforms, their specificities for HLA-C allele products are determined by direct interactions with polymorphic HLA-C residues 77 and 80 (20). These interactions occur on one site on the α1 helix on the top face of the HLA-C molecule and involve polymorphic residues at positions 44–46 and 67–70 in domain 1 of KIR receptors (21, 22, 53, 54). In contrast, MHC I allele specificity of Ly49 receptors appears to involve recognition of different conformations or positions of side chains of the same conserved amino acid residues at two distinct sites, subsites B and C, below the peptide-binding groove and on the α3 domain, respectively. Furthermore, allele-specific interactions seem to involve multiple Ly49 polymorphic loops with these subsites, including L3, β2-β2′, and β4–β5, indicating the potential for significant complexity in determining specificity, compared with KIR2DL interactions with HLA-C molecules. In line with this possibility, in this study we have demonstrated differences between Ly49i2 and Ly49G in their interactions at subsites B and C, detecting differences in the residues involved at individual B and C subsites, and the relative contribution of the B and C subsites to interaction. Additional examples of Ly49/MHC I interactions may need to be analyzed to fully understand the basis of allele specificity at the level of solvent-exposed residue interactions. Until that time, grouping MHC I allele products based on supertype, given the significant influence that polymorphic peptide anchor residue-binding pockets may have on MHC I conformation, and therefore specificity, may be a useful first prediction.
We thank Dong-Er Gong for excellent technical assistance.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by an operating grant from the Canadian Institutes for Health Research (to K.P.K.). K.J.L. is supported by Canadian Institutes for Health Research and Alberta Heritage Foundation for Medical Research studentships. K.P.K. is an Alberta Heritage Foundation for Medical Research scientist.
Abbreviations used in this paper: MHCI, Class I MHC; KIR, killer Ig related; β2m, β2-microglobulin; EGFP, enhanced GFP.