Reciprocal Transfer of Class I MHC Allele Specificity between Activating Ly-49P and Ly-49W Receptors by Exchange of β4–β5 Loop Residues ¹

Natural killer cells constitute an important component of the innate immune response by directly eliminating tumor cells and virally infected cells, as well as secreting cytokines (1, 2). NK cell function is regulated by a diverse array of activating and inhibitory receptor types expressed by NK cells (3, 4). NK cells express a variety of inhibitory receptors for classical (class Ia) and nonclassical (class Ib) MHC-encoded proteins (3). Mouse NK cell receptors that recognize classical class I MHC molecules belong to the Ly-49 multigene family. The Ly-49 gene family is encoded within the NK cell gene complex along with genes encoding a variety of other lectin-like receptors (5, 6). There are variable numbers of conserved as well as potentially unique Ly-49 genes found in different inbred mouse strains, and Ly-49 genes display extensive allelic variation (7, 8, 9, 10). These differences result in substantial Ly-49 receptor repertoire diversity among mouse strains and presumably for the species as a whole. Similar repertoire diversity is observed in the structurally distinct, but functionally equivalent, killer Ig-related receptor genes in humans (11).

Ly-49Rs are expressed as disulfide-bonded homodimers on NK cells, NKT cells, and some subsets of memory CD8⁺ T cells (3, 12, 13). Ly-49Rs are generally class I allele specific and can be expressed as one of two forms, either inhibitory or activating. The inhibitory Ly-49Rs contain an immunoreceptor tyrosine-based inhibitory motif in their cytoplasmic tails. Upon engagement by class I ligands, inhibitory Ly-49Rs recruit the Src homology 2 domain-containing tyrosine phosphatase-1, which disrupts membrane-proximal activating signaling events (14). Inhibitory Ly-49Rs such as Ly-49A, -C, and -G, participate in the missing-self mechanism of NK cell regulation, whereby healthy cells expressing normal class I levels are spared by engaging the inhibitory Ly-49Rs, whereas cells expressing reduced class I, such as tumor cells or virally infected cells, do not send an Ly-49-dependent inhibitory signal and are destroyed (15). Activating Ly-49Rs possess a charged residue in their transmembrane segments that serves as a presumed docking site for association with the disulfide-bonded homodimeric immunoreceptor tyrosine-based activating motif-containing signaling adapter protein KARAP/DAP12 (16, 17). A number of activating Ly-49Rs, such as Ly-49D, -P, and -W, can recognize class I MHC molecules (9, 18, 19, 20, 21). In contrast, another activating Ly-49 receptor, Ly-49H, has not been found to recognize class I, but instead binds the mouse CMV m157 molecule and plays a direct role in strain-specific resistance to mouse CMV (22, 23).

Several studies have approached the problem of defining the molecular basis of class I MHC recognition by Ly-49Rs (24, 25, 26, 27, 28, 29, 30, 31, 32). The cocrystal structure of Ly-49A and its H-2D^d ligand is of particular interest, because it indicates two distinct areas of interaction termed site 1 and site 2 (26). Site 1 principally involves the region near the N and C termini of the α1 and α2 α helices, respectively, whereas site 2 involves a larger contact area beneath the peptide binding groove encompassing the cavity bordered by the class I α1/α2/α3 domains and β₂-microglobulin (26). Site-directed mutagenesis of class I and Ly-49Rs in conjunction with binding and functional studies are largely consistent with this model, with greater support for the importance of site 2 (27, 29, 33). Despite these studies, what determines differences in class I allele specificity from the perspective of Ly-49Rs is not known.

We have previously cloned a number of mouse Ly-49A- and -G-related activating and inhibitory receptors and some of their allelic forms, for which we have examined class I MHC specificities (9, 20, 34). These receptors have similar amino acid sequences, but demonstrate varying class I MHC allele specificities. We have allowed natural amino acid substitutions in polymorphic Ly-49Rs to serve as a guide in defining residues that may alter Ly-49 receptor specificity for distinct class I MHC allele products. In this report, we demonstrate that specificity for the H-2D^k class I MHC allele product can be transferred from one Ly-49 receptor, Ly-49W, to another, Ly-49P, by the exchange of 3 aa of the Ly-49W β4–β5 loop. Furthermore, we find that replacement of these same amino acids in Ly-49W, by corresponding residues in Ly-49P, abolishes H-2D^k recognition while still preserving H-2D^d recognition. These results provide the first evidence indicating the importance of a specific site on Ly-49Rs, the variable β4–β5 loop, in determining the class I allele specificity of these two activating Ly-49Rs, with potential relevance to the class I allele specificity of other Ly-49Rs.

Female CBA/J (H-2^k), DBA/2J (H-2^d), C57BL/6 (H-2^b), B10.BR (H-2^k), B10.D2 (H-2^d), and B10 (C57BL/10; H-2^b) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). All animal experiments were approved by the Animal Welfare and Policy Committee of the University of Alberta.

RNK-16, a spontaneous F344 rat strain NK cell leukemia, has been described (35). RNK-16 cells were maintained in RPMI 1640 supplemented with penicillin, streptomycin, l-glutamine, and 50 μM 2-ME (RNK medium). Ly-49P-expressing RNK-16, clone P1A3, and Ly-49W-expressing RNK-16, clone W10G5, were generated previously (9, 20). Daudi, a human B lymphoblast cell line (36), was obtained from American Type Culture Collection (Manassas,VA) and maintained in RNK medium.

Hybridomas producing the following Abs were obtained from American Type Culture Collection (Manassas,VA): 4D11 (rat IgG2a; anti-Ly-49G/W) (9), A1 (IgG2a; anti-Ly-49A/P) (20), 34-5-8S (IgG2a; anti-H-2D^d) (37), M1/42 (rat IgG2a; anti-pan-mouse MHC I) (38), BB7.1 (IgG1; anti-HLA-B7) (39), and B27M1 (IgG2a; anti-HLA-B27, B7) (40). The Cwy-3 (IgG1; anti-Ly-49G/W) (41)-producing hybridoma was generated in this laboratory. Abs were prepared by ammonium sulfate precipitation and PBS dialysis of tissue culture supernatants obtained from hybridomas grown in protein-free hybridoma medium. Purified OX-8 (IgG1; anti-rat CD8α) (42) and 15-5-5S (IgG2a; anti-H-2D^k) (43) were purchased from BD PharMingen (San Diego, CA). FITC-coupled rat anti-mouse IgG and mouse anti-rat IgG were purchased from Jackson ImmunoResearch Laboratories (West Grove, PA).

The cDNAs encoding Ly-49W1 and Ly-49P1 were previously cloned from the nonobese diabetic mouse into the p3T vector (Miltenyi Biotec, Göttingen, Germany) (9, 20). Ly-49W and Ly-49P were mutated using the QuikChange Mutagenesis kit (Stratagene, La Jolla, CA). Mutagenic primers were designed to mutate codons of interest. Residues 245–248 in Ly-49P corresponding to the sequence NcDQ were mutated to DcGK, where “c” represents the conserved cysteine residue and the underlined sequences represent exchanged residues. Residues 250–253 in Ly-49W corresponding to the sequence DcGK were mutated to NcDQ, DcDK, NcDK, and DcDQ. All mutated cDNAs were verified by DNA sequencing.

The coding regions of Ly-49P^DcGK, Ly-49W^NcDQ, Ly-49W^DcDK, Ly-49W^NcDK, and Ly-49W^DcDQ were inserted into the XhoI-XbaI sites of the bicistronic expression vector BSRαEN and transfected into RNK-16 cells, as described previously (14). Four million cells were transfected with 20 μg of plasmid by electroporation at 180 mV and 960 μF. Transfected cells were transferred into 96-well microtiter plates in RNK medium supplemented with 1 mg/ml G418. After 10 days in selection medium, viable cells were analyzed for Ly-49 expression. RNK-16 cells transfected with Ly-49P^DcGK were incubated with the A1 mAb, or B27M1 as the isotype control. The cells were then stained with FITC-labeled rat anti-mouse IgG secondary Abs. Cells transfected with Ly-49W^NcDQ, Ly-49W^DcDK, Ly-49W^NcDK, and Ly-49W^DcDQ were incubated with 4D11 or Cwy-3 mAb, while in parallel, M1/42 or BB7.1, respectively, were used as isotype controls. Cells were then incubated with FITC-conjugated mouse anti-rat IgG secondary Abs, fixed with paraformaldehyde, and expression of the Ly-49Rs was analyzed by FACScan (BD Biosciences, Mountain View, CA).

Spleens of various mouse strains were harvested and homogenized, and RBC were lysed by ammonium chloride treatment. Splenocytes were cultured at 5 × 10⁶ cells/ml in RNK medium with 3.0 μg/ml Con A (Sigma-Aldrich, St. Louis, MO) for 2 days. Blast cells were recovered and used as target cells for cytotoxicity assays.

Target cells were labeled at 37°C at 100–150 μCi of Na⁵¹CrO₄ (⁵¹Cr; Mandel/NEN Life Science Products, Guelph, Canada) for 1 h if tumor cells, or for 1.5 h if Con A blast cells. The cells were then extensively washed, and 1 × 10⁴ target cells were incubated for 4 h at 37°C with varying E:T ratios with RNK-16 cells in 96-well V-bottom microtiter plates. Following incubation, plates were centrifuged for 5 min. Supernatants (25 μl) were transferred to 96-well plates, and 100 μl of scintillant (OptiPhase SuperMix; Wallac, Loughborough, England) was added. The plates were analyzed in a beta counter (MicroBeta; Wallac). Percent specific lysis was calculated as follows: [(experimental release) − (spontaneous release)/(maximal release) − (spontaneous release)] × 100. All cytotoxicity assays were performed a minimum of three separate times.

Abs were preincubated with a soluble mixture of recombinant proteins A and G (Calbiochem, San Diego, CA) at 1 μg per microgram of mAb for 15 min to block the Ab Fc region. This mixture was then incubated with either target or effector cells for a further 30 min to mediate receptor blocking before use in a 4-h cytotoxicity assay.

Daudi cells were labeled with ⁵¹Cr and then incubated with 40 μg/ml specific Ab (Cwy-3) at room temperature for 30 min. The RNK-16 effector cells expressing Ly-49W^DcDQ were added to the preincubated target cells, and a 4-h cytotoxicity assay was performed.

We have previously cloned and characterized the Ly-49P and Ly-49W receptors from the nonobese diabetic mouse (9, 20). Both are activating receptors that lack an immunoreceptor tyrosine-based inhibitory motif on the cytoplasmic domain and also contain an arginine residue in the transmembrane segment, which permits association with the signaling adaptor molecule DAP12 (16, 17). These activating Ly-49Rs show high sequence identity in the extracellular domain to Ly-49A (Fig. 1,A) and Ly-49G (not shown) inhibitory receptors (9). Ly-49P shares ∼93% amino acid sequence identity in the carbohydrate recognition domain (CRD) with Ly-49A, whereas Ly-49W shares ∼98% sequence identity in the CRD with Ly-49G (9, 20). Amino acid sequence alignment shows that Ly-49P and -W also share significant sequence identity. Ly-49P and -W have ∼82% sequence identity in the CRD (Fig. 1,A); however, Ly-49P and -W display differences in specificity for class I MHC alleles and Ab reactivity (Fig. 1,B and Refs. 9 and 20). RNK-16 cells transfected with Ly-49P are able to recognize and kill target cells expressing H-2D^d (20), whereas Ly-49W-expressing RNK-16 cells display specificity for both H-2D^k and H-2D^d (9). We sought to define the molecular basis of the class I MHC specificity differences of Ly-49P and -W. The many amino acid identities throughout Ly-49W and -P strongly suggest that Ly-49W and -P will fold into a very similar structure to that defined for Ly-49A. Therefore, we used the cocrystal structure between Ly-49A and H-2D^d as a model for Ly-49P and -W class I interaction (26). The cocrystal structure showed that Ly-49A can potentially bind at two topologically distinct sites on class I MHC, termed site 1 and site 2 (26). Using Ly-49A residues involved in the interaction at both sites as a template, we focused on the β4–β5 loop involving three natural nonconservative amino acid residue substitutions (Fig. 1 A). In Ly-49P, these residues are N245, D247, and Q248, whereas in Ly-49W, these residues are D250, G252, and K253. We hypothesized that the differences in these residues between Ly-49P and Ly-49W may confer their differences in class I allele specificity.

We made mutations to exchange the three β4–β5 loop residues in Ly-49W to the corresponding residues in Ly-49P (Ly-49W^NcDQ). The reciprocal change was also made in Ly-49P to generate Ly-49P^DcGK. These mutants were expressed in RNK-16, and the mAbs A1 and 4D11 still recognized the mutant Ly-49P and Ly-49W receptors, respectively (Fig. 2). These results suggest that the mutations did not result in a major change to the overall structures of the receptors, nor are these residues involved in specifying Ly-49 receptor recognition by these mAbs.

The RNK-16 cells transfected with the Ly-49P and -W mutants were used in cytotoxicity assays against Con A blasts derived from splenocytes of various inbred mouse strains expressing different MHC haplotypes (Fig. 3,A). Consistent with previous results, wild-type Ly-49W-expressing RNK-16 cells (clone 10G5) lyse target cells from both DBA/J (H-2^d) and CBA/J (H-2^k) mice but not from the C57BL/6 (H-2^b) mouse. In contrast, although Ly-49W^NcDQ-expressing cells are still able to kill target cells from the DBA/J (H-2^d) mouse, they lose the ability to lyse target cells from the CBA/J (H-2^k) mice. Although wild-type Ly-49P-expressing RNK-16 cells are able to lyse target cells from the DBA/J (H-2^d) mouse, but not from the CBA/J (H-2^k) or the C57BL/6 (H-2^b) mouse, Ly-49P^DcGK- expressing cells kill target cells from the DBA/J (H-2^d) mouse, but in addition, have gained the ability to lyse target cells from CBA/J (H-2^k) mice (Fig. 3 A). As a negative control, untransfected RNK-16 cells were assayed with the same Con A blasts, and they were unable to lyse target cells from the CBA/J (H-2^k), DBA/J (H-2^d), or C57BL/6 (H-2^b) mouse strains. Thus, the pattern of strain recognition by Ly-49P^DcGK is identical with that of the wild-type Ly-49W receptor, and the pattern of recognition by Ly-49W^NcDQ is identical with that of the wild-type Ly-49P receptor.

To verify that the mutant Ly-49 receptor recognition was MHC specific, we repeated the assays using MHC congenic mice (Fig. 3,B). Consistent with previous results, wild-type Ly-49W recognized Con A blasts derived from the B10.BR (H-2^k) and B10.D2 (H-2^d) mouse, but not from the B10 (H-2^b) mouse. Ly-49W^NcDQ-expressing RNK-16 cells were able to kill target cells from the B10.D2 (H-2D^d) mouse, but not from the B10.BR (H-2^k) or B10 (H-2^b) mouse, thus confirming that the results seen in Fig. 3,A were due to differences in target cell MHC. Wild-type Ly-49P-expressing RNK-16 cells were able to kill Con A blasts from B10.D2 (H-2D^d), but not from B10.BR (H-2^k) or B10 (H-2^b) mice. Ly-49P^DcGK-expressing RNK-16 cells lysed target cells from both B10.BR (H-2^k) and B10.D2 (H-2^d), but not from B10 (H-2^b) mice, which is again consistent with the results in Fig. 3,A. Untransfected RNK-16 cells are unable to kill Con A blast cells from B10.BR (H-2^k), B10.D2 (H-2^d), or B10 (H-2^b) mice. Thus, the results in Fig. 3 indicate that the ability of Ly-49W to recognize H-2^k targets is conferred within the three residues D250, G252, and K253 in the β4–β5 loop of Ly-49W.

To verify that the lytic activity observed was indeed dependent on transfected mutant Ly-49 receptor interaction with class I MHC, we performed Ab inhibition assays. Ly-49W-expressing RNK-16 cell lysis of B10.BR and B10.D2 Con A blasts is mediated by interaction of the Ly-49W with H-2D^k and H-2D^d (Fig. 4,A and Ref.9). Lysis of Con A blast cells from B10.D2 mice by Ly-49W^NcDQ-expressing RNK-16 cells was mediated by the interaction of Ly-49W^NcDQ receptor with H-2D^d, because both Cwy-3 (anti-Ly-49W) and 34-5-8S (anti-H-2D^d) mAbs blocked lysis, whereas isotype Abs had little or no effect (Fig. 4,B). Lysis of B10.D2 Con A blast cells by Ly-49P-expressing RNK-16 cells was dependent on interaction of Ly-49P and H-2D^d, because A1 (anti-Ly-49A/P) and 34-5-8S blocked lysis (Fig. 4,C). Ly-49P^DcGK-mediated lysis of B10.BR- and B10.D2-derived target cells was dependent on the interaction of Ly-49P^DcGK with either H-2D^k or H-2D^d, because the A1, 15-5-5S (anti-H-2D^k), and 34-5-8S Abs blocked cytolysis (Fig. 4 D and data not shown). The results confirm that the differential cytolysis observed is mediated by the mutant Ly-49Rs, and thus, the three residues D, G, and K in the β4–β5 loop of Ly-49W confer the recognition of H-2D^k.

To determine which of the three identified Ly-49W residues, D250, G252, and K253, are involved in H-2D^k recognition, we began by making a single amino acid substitution. A previous study by our laboratory on the allelic variation of the inhibitory Ly-49G receptor showed that Ly-49G from the BALB/c mouse can recognize both H-2D^d and H-2D^k, whereas Ly-49G from the B6 mouse can recognize only H-2D^d targets (34). Examination of the residues in the β4–β5 loop indicates that the two Ly-49G alleles differ by only one residue in this location. The BALB/c allele is similar to Ly-49W in that it contains the residues D, G, and K. The B6 allele differs from both in that the G corresponding to residue 252 in Ly-49W and Ly-49G^BALB/c is replaced with D. We investigated whether mutation of G252 to D in Ly-49W was sufficient to cause a loss in recognition of H-2D^k. Ly-49W was mutated to Ly-49W^DcDK and expressed in RNK-16 cells (Fig. 5,A). These transfectants were then used in cytotoxicity assays against Con A blasts from various mouse strains. Ly-49W^DcDK-expressing cells were still able to lyse cells expressing H-2D^d and H-2D^k (Fig. 5,B), although there was a partial loss in recognition of H-2D^k targets relative to H-2D^d targets compared with that typically observed with wild-type Ly-49W (Fig. 3 A). Thus, a single amino acid exchange between Ly-49P and Ly-49W did not completely abrogate recognition of H-2D^k-expressing cells.

Considering that the single substitution of G252 to D resulted in a partial loss in H-2D^k recognition relative to H-2D^d targets, we examined the possibility that exchange of G252 to D in combination with either an exchange of D250 for N or K253 for Q of the identified three β4–β5 loop amino acids would result in a loss in H-2D^k recognition. We made mutations in Ly-49W to create Ly-49W^NcDK and Ly-49W^DcDQ. These mutants were then expressed in RNK-16 cells and used in cytotoxicity assays (Fig. 6, A and B). RNK-16 cells expressing Ly-49W^NcDK, like the single mutant Ly-49W^DcDK, only display a partial reduction in H-2D^k recognition relative to H-2D^d targets, suggesting that exchange of these two residues is not sufficient to completely eliminate specificity for H-2D^k (Fig. 6,A). Interestingly, RNK-16 cells expressing Ly-49W^DcDQ displayed a complete loss in recognition of targets expressing either H-2D^d or H-2D^k (Fig. 6,B). This result was found with all of several RNK-16 Ly-49W^DcDQ transfectant clones from multiple transfections (Fig. 6,B and data not shown). To show that this loss in recognition of targets was not due to a general inability to signal through the mutant Ly-49W receptor, we performed a rADCC assay using FcR-positive Daudi cells (Fig. 6 C). Daudi cells were preincubated with Cwy-3 or an isotype control, and subsequently Ly-49W^DcDQ-expressing RNK-16 cell were added. In the presence of Cwy-3, Daudi cells were lysed, indicating that the Ly-49^DcDQ receptor was still capable of transmitting activating signals. Thus, the exchange of all three residues, D250, G252, and K253, in the β4–β5 loop of Ly-49W is most effective at ablating H-2D^k recognition while maintaining recognition of another class I allele, H-2D^d. We conclude that residues in the β4–β5 loop determine H-2D^k allele specificity of the activating Ly-49W receptor.

Despite determination of the Ly-49A-H-2D^d cocrystal structure (26) and a number of site-directed mutagenesis studies (25, 27, 30), the molecular features of Ly-49Rs that confer MHC allele specificity have remained unclear. Using a series of amino acid exchanges between the NK activating receptors Ly-49W and -P, we have determined that three residues, D250, G252, and K253, in the predicted β4–β5 loop of Ly-49W confer specificity for H-2D^k. When these residues are transferred into Ly-49P, they allow Ly-49P^DcGK to gain specificity for H-2D^k. When the corresponding residues from Ly-49P are exchanged into Ly-49W, Ly-49W^NcDQ loses the ability to recognize H-2D^k, but retains the ability to recognize H-2D^d. Exchange of one or two of these residues either resulted in a partial loss of H-2D^k recognition relative to H-2D^d targets or a total loss of both H-2D^d and H-2D^k recognition. To our knowledge, this is the first report to identify Ly-49 residues that participate in discrimination between specific class I MHC allele ligands.

High conservation of amino acid sequences in the CRD of Ly-49A, -P, and -W would predict that these proteins adopt highly similar structures. Thus, modeling interactions between Ly-49P and -W with H-2D^d and H-2D^k based on the published cocrystal structure between Ly-49A and H-2D^d is possible (26). There are 25 residues on Ly-49A that contact the H-2D^d at either site 1 and/or site 2. Of these residues, 18 are conserved among Ly-49A, -P, and -W. However, a cluster of amino acid differences between Ly-49A/P and Ly-49W lie within the solvent-exposed β4–β5 loop, and correspond to a subset of Ly-49A residues that interface with H-2D^d (Fig. 7). This difference in interface residues could account for the differences seen in the MHC allele specificity displayed by Ly-49P and -W. Our results indeed indicate that the β4–β5 loop plays an important role in class I allele recognition by Ly-49P and -W. However, it is less clear which allele-specific features of class I MHC are sensed by the β4–β5 loop. The ambiguity lies with the cocrystal structure between Ly-49A and H-2D^d where the β4–β5 loop of one Ly-49 monomer is involved in the site 1 interface, whereas the same residues of the β4–β5 loop from another monomer are involved in the site 2 interface (Fig. 7,B and Ref.26). Mutagenesis of class I residues in the site 2 interface have clearly demonstrated a role for site 2 in H-2D^d recognition by Ly-49A (27). However, it has not been determined whether class I residues of site 1 and/or site 2 contribute to activating Ly-49 receptor recognition, or for the Ly-49W-related inhibitory receptor Ly-49G. All class I MHC residues in the site 2 interface that have at least one atom within 10 Å of any atom of the β4–β5 loop are fully conserved between H-2D^d and H-2D^k (Fig. 7,B). The other residues of the site 2 interface come from β₂-microglobulin and are therefore also invariant. Accordingly, detection of class I allele-specific features by Ly-49 in site 2 would involve long-distance effects, most likely requiring conformational changes. In contrast, the β4–β5 loop in site 1 interacts directly with a polymorphic region of class I, K173-N174 in H-2D^d and E173-L175 in H-2D^k (Fig. 7 B). We have previously suggested that Ly-49W may specifically recognize E173 of D^k via a salt bridge formed with K253 in the β4–β5 loop. Ly-49P cannot make the same salt bridge, because it has a glutamine instead of the lysine (44). In this study, we show that the double mutant Ly-49W^NcDK indeed retains partial recognition of D^k, supporting a critical role for K253. We had further proposed that the asparagine to aspartate and aspartate to glycine substitutions in Ly-49W compared with Ly-49P (DcGK vs NcDQ, respectively, in the β4–β5 loop) would enhance affinity for H-2D^k by creating space for the formation of the K253-E173 salt bridge. In this study, we show that Ly-49P^DcGK indeed lyses D^k targets more efficiently than Ly-49W^NcDK, consistent with that possibility. The involvement of class I residues of site 1 or site 2 in determining MHC allele specific interaction with Ly-49Rs must await functional studies with class I mutants at these sites. The question of how the β4–β5 loop is conferring specificity for H-2D^k may be answered by the solution of a cocrystal structure between Ly-49W and H-2D^k.

We have demonstrated that the three identified residues of the predicted β4–β5 loop clearly play a role in its recognition of H-2D^k by Ly-49W; however, it is less clear what role these β4–β5 loop residues play in its recognition of H-2D^d. Recognition of H-2D^d occurred with both triple residue exchange mutants, Ly-49W^NcDQ and Ly-49P^DcGK, indicating that either H-2D^d binding is common to both sets of residues, or else specificity is conferred elsewhere on the Ly-49 molecule. Mutational studies of Ly-49A have shown that numerous residues outside the β4–β5 loop are crucial for the interaction with H-2D^d (27). All of these residues are conserved in both Ly-49P and -W (Fig. 1,A) and could account for maintenance of H-2D^d recognition despite changes in H-2D^k binding. However, an alanine mutation of Ly-49A resulting in the sequence NcAQ yielded a loss in H-2D^d binding (27), suggesting that the β4–β5 loop of Ly-49A is involved in H-2D^d recognition. Interestingly, the Ly-49W^DcDQ mutant led to a loss in both H-2D^d and H-2D^k recognition, suggesting that residues in the Ly-49W β4–β5 loop influence interaction with H-2D^d as well as H-2D^k. The Ly-49W^DcDQ mutant is unable to recognize H-2D^d targets, yet the Ly-49W^NcDQ mutant recognizes H-2D^d targets effectively. These mutants differ by one residue, namely the aspartate for an asparagine at position 250. In the Ly-49A interaction at site 1, the corresponding N244D substitution would not be expected to have consequences for class I interaction. In addition, the fact that Ly-49W and the mutant Ly-49W^DcDK recognize H-2D^d-expressing target cells effectively (Fig. 5), indicates that an aspartate at position 250 can be tolerated. However, the differential behavior of the Ly-49W^DcDQ and Ly-49W^DcDK mutants might be due to effects in the site 2 interface. The aspartate at position 250 of Ly-49W would be positioned in between E232 of the MHC class I molecule and D252 of the Ly-49W β4–β5 loop. In the Ly-49W^DcDQ mutant, there are no positive charges nearby to compensate the charge-charge repulsion, and lack of killing of both H-2D^d and D^k could thus be due to a weakened interaction in site 2. In contrast, the K253 of the Ly-49W^DcDK mutant might make an extra salt bridge to E232 and thereby enhance affinity to a functional level. Although, the Ly-49W^DcDQ receptor is still able to transduce signals into the cell (Fig. 6 C), the sequence DcDQ has not been found in any known Ly-49 sequences thus far. Whether this mutant receptor is capable of recognizing any MHC molecules is unknown. Finally, a Ly-49W^NcGQ β4–β5 loop mutant was not examined in this study, because a Ly-49A mutant with the sequence NcAQ (27) and the Ly-49W^DcDQ mutant both cannot even recognize H-2D^d. It would be predicted that Ly-49W^NcGQ could also not recognize H-2D^d, and likely H-2D^k.

The importance of residues in the predicted β4–β5 loop in the determination of specificity in other Ly-49Rs has yet to be determined. There are 20 fully sequenced putative Ly-49 family members along with about 14 allelic variations published in GenBank, and the corresponding amino acid residues in the β4–β5 loop of these receptors show a high degree of sequence variability, with 11 different sequences. The variability in this loop may contribute to or define the fine specificities of MHC allele recognition. Examining other Ly-49Rs with known class I MHC specificities, the Ly-49G^BALB/c allele recognizes H-2D^k very well and shares the same DcGK β4–β5 loop sequence with Ly-49W (34). By contrast, the Ly-49G^B6 allele displays an altered residue sequence in the β4–β5 loop, DcDK, and does not recognize H-2D^k (34). In addition, we would predict that, because the Ly-49G^129/J allele has the sequence DcGK in the β4–β5 loop, it should recognize H-2D^k. Tetramer binding studies with H-2D^k show that this is the case, albeit the binding is relatively weak (45). Ly-49A shares the sequence NcDQ in the β4–β5 loop with Ly-49P, which recognizes H-2D^d but not H-2D^k. We have previously shown that Ly-49A poorly recognizes H-2D^k relative to H-2D^d, indicating that the NcDQ in the β4–β5 loop is not ideal for H-2D^k recognition (46). Ly-49C recognizes H-2^s cells much more strongly than Ly-49I (47). Ly-49C and Ly-49I differ at residue 247 in the β4–β5 loop, where Ly-49C possesses an isoleucine and Ly-49I has a threonine. The position of these residues is adjacent to the D250 of the Ly-49W β4–β5 loop. Interestingly, the Ly-49C mutant I247T exhibited a reduction of binding to GM979 cells (H-2^s) but retained its ability to bind IC-21 cells (H-2^b), thus also implicating the β4–β5 loop in the determination of specificity for selective class I MHC molecules (25). However, determination of the role of I247 in the differential binding to H-2^s cells in that study was incomplete, because the authors did not further define class I MHC molecules that may be involved, nor did they determine whether the reciprocal mutation, T247I, transferred enhanced binding to H-2^s onto Ly-49I. In another study, Ly-49C residues involved in recognition of multiple class I ligands were identified (30). However, no residues were identified in Ly-49C that discriminated between class I MHC alleles, nor was there acquisition of a new class I specificity, such as for H-2D^d, upon mutagenesis of this receptor to selected residues from Ly-49A (30). In contrast, our study has demonstrated that the DcGK sequence in the β4–β5 loop plays an important role in H-2D^k allele recognition. The β4–β5 loop we have identified for class I MHC allele specificity by activating Ly-49Rs is likely to also play a role in the allele specificity of inhibitory Ly-49Rs. Thus, overall, related activating and inhibitory Ly-49Rs such as Ly-49W and Ly-49G, respectively, may not be fundamentally different in class I MHC recognition (34).

Ly-49Rs belong to the C-type lectin superfamily that also includes the selectins and the collectins (48). The β4–β5 loop may also play a functional role in defining specificity of other C-type lectins. The cocrystal structure between P-selectin and a peptide fragment from P-selectin glycoprotein ligand-1 showed that residues in the His¹⁰⁸-His¹¹⁴ loop in P-selectin, which is equivalent to the Ly-49 β4–β5 loop, forms a surface for ligand contact (49). Differences exist in the residues of the corresponding loops of P-selectin and E-selectin and may contribute to the fine specificity for ligand binding of P-selectin vs the broad specificity of E-selectin.

The precise function of activating Ly-49Rs remains to be determined, and it is not clear whether germline-encoded class I molecules or viral MHC-like products are their primary ligands. It is well known that inhibitory Ly-49Rs can recognize self class I MHC molecules on target cells and suppress lytic activity, thus maintaining NK cell tolerance to self. There is overlap of expression of Ly-49 activators and inhibitors on NK cells (18, 50). This presumably ensures that inhibitory Ly-49Rs suppress the function of activating Ly-49Rs upon encounter with normal self cells (50). Potentially, activating Ly-49Rs that recognize class I ligands become active when expression of a nonoverlapping class I ligand of a coexpressed inhibitory Ly-49 receptor is lost. Class I MHC allele-specific activating Ly-49Rs could allow NK cells to detect the loss of expression of even a single class I allele, particularly in the context of pathological challenge such as with viruses. Additional studies are necessary to define the function of activating Ly-49Rs. The present study provides insight into how two activating Ly-49Rs discriminate between class I ligands and may also have relevance to the differentiation of class I ligands by related inhibitory Ly-49Rs. This knowledge will be helpful in discerning the influence of Ly-49Rs on the balance of signals regulating NK cell functions.

We thank Dr. Mary Nakamura (University of California, San Francisco, CA) for RNK-16 cells and helpful advice. We also thank Dr. Andrey Shaw (Washington University, St. Louis, MO) for the BSRαEN expression vector.