Role of Conserved Glycosylation Site Unique to Murine Class I MHC in Recognition by Ly-49 NK Cell Receptor¹

In many systems, expression of class I MHC molecules on target cells inversely correlates with their susceptibility to NK cell lysis (1, 2, 3, 4). The “missing self” hypothesis states that NK cells distinguish between normal and aberrant cells by recognizing and killing targets that lack self class I MHC (1, 5). Recent findings of MHC-specific inhibitory receptors on NK cells have provided the molecular basis for this hypothesis. In mice, the Ly-49 family of type II transmembrane proteins containing C-type lectin domains has been identified to be NK receptors for class I MHC. Of the nine Ly-49 molecules (Ly-49A-I) thus far identified, Ly-49A and Ly-49C are the best characterized. Ly-49A has been shown by Ab inhibition studies to bind specifically to the H-2D^d and H-2^k haplotypes, resulting in the inability of Ly-49A⁺ NK cells to kill targets of these haplotypes in vitro and in vivo (6, 7, 8). The interaction between Ly-49A and class I MHC has also been demonstrated by the adhesion of Ly-49A⁺ lymphoma cells to purified and immobilized D^d and D^k (9). Similarly, Ly-49C binds certain class I MHC molecules, including H-2D^d, K^d, and K^b (10, 11), and functions as an inhibitory receptor for NK cells (12). Binding studies using various blocking mAb have suggested that Ly-49 binds to the α1/α2 domains of class I MHC (13).

Ly-49 contains a putative carbohydrate recognition domain (14, 15, 16), and previous studies have shown that Ly-49A and -C bind sulfated polysaccharides (17, 18). Furthermore, treatment of cells expressing class I MHC with tunicamycin to inhibit glycosylation (17) or with fucosidase (18) inhibited binding of class I MHC to Ly-49A or -C. These results suggested an involvement of carbohydrates in the recognition of class I MHC. However, the precise role of carbohydrates on class I MHC in its interaction with the Ly-49 family of NK inhibitory receptors has not yet been determined. The murine class I MHC molecules have two conserved asparagine N-linked glycosylation sites, amino acid residues at positions 86 in the α1 domain and 176 in the α2 domain. In this study, we report that the glycosylation at residue 176, but not residue 86, is required for the binding of class I MHC to Ly-49A and Ly-49C. We also report that other structural features are likely to be involved in this interaction. Interestingly, unlike residue 86, which is invariably conserved among all class I MHC molecules thus far sequenced (19, 20, 21, 22), the glycosylation site at residue 176 is conserved only among murine class I MHC and not those of other species. These findings suggest that the Ly-49 family of murine NK inhibitory receptors appear to have evolved simultaneously with murine class I MHC molecules.

The 5E6 (anti-Ly-49C and -I) and YE1/48 (anti-Ly-49A) mAbs have been described (23, 24). The 34-5-8S, 34-2-12S, and 34-4-20S hybridomas (anti-D^d) were obtained from the American Type Culture Collection (ATCC, Manassas, VA). Flow cytometric analysis was done as described (11). Briefly, 5 × 10⁵ cells were incubated for 30 min on ice with the appropriate hybridoma supernatants, followed by two rinses with HBSS (containing 2% FCS) and an incubation with an FITC-conjugated secondary Ab. After a final rinse with HBSS/FBS, cells were stained with propidium iodide for detection of nonviable cells. Flow cytometric analysis was subsequently performed on the FACScan (Becton Dickinson, Mountain View, CA).

The cDNA encoding D^d was isolated by RT-PCR from the murine B lymphoma line A20. The D^d mut1 in which serine 88 is replaced by glycine residue was generated by a two-step PCR using Pfu I polymerase (Strategene, La Jolla, CA). The upstream PCR to generate a DNA fragment encoding the N-terminal portion of D^d that contained the mutation at residue 88 was conducted using oligonucleotide primers 5′-CTGGATCCCAGATGGGGGCGAT-3′ and 5′-CCCTGGTTGTAGTAGCGCAGCG-3′ (phosphorylated). The downstream PCR to generate a DNA fragment encoding the C-terminal portion of D^d was performed using primers 5′-CGCGGGCGGCTCTCACACACTC-3′ (phosphorylated) and 5′-TTGAATTCATGTGTTAGTCTTGGTGGATGAAGG-3′. The two PCR fragments were blunt-end ligated and used as a template for the second round of PCR that used the primers 5′-CTGGATCCCAGATGGGGGCGAT-3′ and 5′- TTGAATTCATGTGTTAGTCTTGGTGGATGAAGG3′. The D^d mut2 in which threonine in residue 178 is replaced by glycine was also generated in the same manner. The upstream PCR to generate the mutated DNA fragment was done using primers 5′-CTGGATCCCAGATGGGGGCGAT-3′ and 5′-CCCAGCATTCCCGTTCTTCAGGTATC-3′ (phosphorylated), and the downstream PCR was done using primers 5′-CTGCTGCGCACAGATCCCCC-3′ (phosphorylated) and 5′-TTGAATTCATGTGTTAGTCTTGGTGGATGAAGG-3′. The second round of PCR was done using the same primers as those for the first mutation. The DNA fragments from the second round PCR were digested with BamHI and EcoRI, subcloned into pBluescript, and sequenced.

The cell lines COS-1, RBL-1, GM979, and C1498 were obtained from ATCC and cultured in DMEM supplemented with 5% FCS. Ly-49A and -C cDNAs that had been cloned into the pAX142 expression vector were transfected into COS cells using Lipofectamine (Canadian Life Technologies, Burlington, ON) according to the manufacturer’s protocols. Transfection of RBL-1, GM979, and C1498 cells were performed by electroporation at 0.45 kV and 125 μFD, and stable transfectants resistant to G418 (Canadian Life Technologies) were subsequently selected for high expression by limiting dilution, panning, or cell sorting.

GM979 transfectants were surface biotinylated and lysed with lysis buffer (1% Triton X-100 lysis buffer containing 1% BSA, 150 mM NaCl, and 0.1% NaN₃ in 10 mM Tris-HCl, pH 7.5) as described (25). The clarified lysates were incubated overnight at 4°C with anti-D^d (34-5-8S)-coupled beads (50 μl of a 1:1 slurry). The beads were then washed four times with lysis buffer (without BSA and NaN₃), and bound proteins were eluted with 100 μl 2× SDS sample buffer containing 50 mM DTT and boiling for 5 min. The eluted proteins were subsequently analyzed by SDS-PAGE and transferred onto nitrocellulose for Western blotting. Following incubation with peroxidase-conjugated streptavidin, proteins were detected using the ECL (Amersham, Arlington Heights, IL) chemiluminescence method.

Following immunoprecipitation as above, D^d bound to anti-D^d-mAb-coupled beads were washed and resuspended with PBS, treated with two units of N-glycosidase F (Boehringer Mannheim Canada, Laval, QC) for 6 h or 18 h at 37°C, and finally eluted with 100 μl 2× SDS sample buffer.

Twenty-four hours after transfection, COS cells were trypsinized and replated on 6-cm dishes (Falcon, Oxnard, CA). Forty-eighty hours later, D^d transfectants were added to COS cells. For the adhesion of RBL-1 transfectants, dishes containing COS cells were pretreated with heat-inactivated BSA (0.1% in PBS) for 1 h at 37°C to reduce nonspecific binding. Adhesion assays were performed for 2 h at 37°C or 15 min at room temperature for GM979 and RBL-1 cells, respectively. The plates were subsequently rinsed six times with prewarmed media and photographed on a Nikon Diaphot microscope. For quantitative assays, GM979 cells were radiolabeled with ⁵¹Cr (Mandel Scientific, Guelph, ON), and equivalent amounts of radioactivity were subsequently added to COS cells. After the unbound cells were removed by rinsing with prewarmed media, the remaining bound cells were lysed with 10% Triton X-100 and counted for radioactivity.

Spleens from C57BL/6 mice were homogenized and passed through a nylon wool column (Polysciences, Warrington, PA). Flow-through cells were then further enriched for NK cells using the immunomagnetic separation method, StemSep (StemCell Technologies, Vancouver, BC). The NK cells were subsequently cultured for 3 days in 1000 U/ml of murine IL-2 before purification of the Ly-49A⁺ subpopulation using the MACS separation method (Miltenyi Biotec, Auburn, CA). After 6 additional days of culture in the presence of IL-2, the Ly-49A⁺ NK cells were used in standard ⁵¹Cr release cytotoxicity assays.

To determine the role of the conserved N-linked carbohydrates of class I MHC in the recognition of class I MHC by Ly-49, two glycosylation-deficient mutants of D^d were generated. The D^d mut1 lacks the glycosylation site at residue 86 due to a replacement of serine by glycine at residue 88. In the D^d mut2, residue 176 is deficient in glycosylation due to a substitution of threonine at residue 178 by glycine. The wild-type and mutant D^d cDNA clones were subsequently transfected into the murine erythroleukemic cell line GM979, the rat basophilic leukemia line RBL-1, and the murine lymphoma cell line C1498, and transfectants expressing D^d were established. Flow cytometric analysis of the transfected cells using the conformation-dependent anti-D^d mAb 34-5-8 showed that wild-type D^d and both mutant D^d were expressed at comparable levels (Fig. 1,A). No differential binding of other mAb that recognize D^d, including 34-4-21, 34-2-12, and 34-4-20, to wild-type and mutant D^d was observed (data not shown), indicating that the mutations did not significantly disrupt the conformation of D^d. To verify that the expressed mutant D^d proteins were indeed deficient at one glycosylation site, D^d was immunoprecipitated from the transfected cells using the 34-5-8 mAb and analyzed by SDS-PAGE. The mutant D^d had lower apparent molecular masses than that of the wild-type (Fig. 1,B), presumably because lack of glycosylation at the mutated sites would reduce the overall mass. Treatment of immunoprecipitated wild-type D^d with N-glycosidase F for 6 h to cleave off N-linked carbohydrates yielded three bands in SDS-PAGE, corresponding to undigested, partially digested, and fully digested D^d (Fig. 1,B, upper panel). The apparent size of the two mutant D^d corresponded to the partially digested wild-type molecule, and treatment with N-glycosidase F reduced the sizes of mutant D^d to that of the fully digested wild-type D^d. Furthermore, prolonged treatment with the enzyme subsequently resulted in complete deglycosylation of all the D^d molecules to its lowest molecular mass (Fig. 1 B, lower panel). These results confirmed that the glycosylation of D^d was indeed disrupted by the mutations as expected.

Binding of mutant D^d to Ly-49 was determined by adhesion of D^d-transfected cells to Ly-49-transfected COS cells. Flow cytometric analysis showed high levels of Ly-49A and Ly-49C expression on the transfected COS cells (Fig. 2,A). GM979 transfected with wild-type D^d readily bound to Ly-49A-transfected COS cells (Fig. 2,B). Similarly, GM979 transfected with D^d mut1 also bound to Ly-49A-transfected COS cells, and no significant effect of this mutation was observed in the binding assay. In contrast, D^d mut2-transfected GM979 failed to bind to Ly-49A-transfected COS cells. This was not due to the expression level of this mutant D^d, because flow cytometric analysis (Fig. 1,A) showed the level of this mutant D^d to be slightly higher than that of D^d mut1. None of the GM979 lines bound to control COS cells transfected with vector alone, whereas untransfected GM979 cells did not bind to Ly-49A-transfected COS cells, indicating that the binding in this assay is mediated by D^d and Ly-49A. Quantitative analysis using ⁵¹Cr-labeled transfected GM979 cells in the presence and absence of YE1/32 (anti-Ly-49A mAb) also confirmed the binding of wild-type D^d and D^d mut1 to Ly-49A (Fig. 2 C). The binding of D^d mut2 was significantly lower even in the absence of the blocking Ab.

To examine the binding of mutant D^d to Ly-49C, the rat leukemia line RBL-1 was used, since endogenous class I MHC (H-2^s) on untransfected GM979 cells bound to Ly-49C-transfected COS cells. RBL-1 expressing wild-type D^d and D^d mut1 bound avidly to both Ly-49A⁺ and Ly-49C⁺ COS cells whereas D^d mut2 showed no significant binding above background (Fig. 3). In all cases, the binding assays were scored in a blind fashion, and the results were consistent in three independent experiments. Therefore, glycosylation at residue 176 but not residue 86, of D^d is important for the binding to Ly-49C as well as Ly-49A in this assay system.

The binding of D^d on target cells to Ly-49A on NK cells has been shown to inhibit cytotoxicity of Ly-49A⁺ NK cells (7). Therefore, we tested whether the very low level of binding of D^d mut2 to Ly-49A results in defective inhibition in NK cytotoxicity. For this study, we were unable to use RBL-1 or GM979 cells, because the former labeled poorly with chromium and the latter cells were NK-resistant. Therefore, we transfected wild-type and mutant D^d into the murine leukemic cell line C1498, which is highly sensitive to NK cytotoxicity. The expression levels of D^d on the transfectants were comparable (Fig. 4,A). In cytotoxicity assays, untransfected C1498 cells were readily lysed by Ly-49A⁺ NK cells. C1498-expressing wild-type D^d or D^d mut1 were resistant to Ly-49A⁺ NK cells, whereas D^d mut2-transfected C1498 cells were only partially resistant to Ly-49A⁺ NK cells (Fig. 4 B). In four independent experiments, the cytotoxic killing of D^d mut2-transfected C1498 was consistently higher than that of wild-type D^d or D^d mut1 transfectants, while no statistically significant difference was seen between wild-type D^d and D^d mut1. The cytotoxicities were restored to the level of control untransfected C1498 cells in the presence of mAb to Ly-49A (YE1/48) or F(ab′)₂ fragment of anti-D^d (34-5-8), demonstrating that the resistance of these cells to NK killing is due to the specific recognition of D^d by Ly-49A. These results indicate that the recognition of D^d mut2 by Ly-49A is less efficient than that of wild-type D^d or D^d mut1.

The results presented in this report have demonstrated that the conserved glycosylation site at residue 176 of D^d is involved in the interaction with the murine NK receptors Ly-49A and -C. In cell adhesion assays using D^d-transfected leukemic cells and Ly-49-transfected COS cells, the mutant D^d (D^d mut2) lacking this glycosylation site did not significantly bind to Ly-49A or -C, whereas wild-type D^d and the mutant D^d (D^d mut1) lacking the other conserved glycosylation site at residue 86 readily mediated cell adhesion. In cytotoxicity assays, D^d mut2 only partially inhibited cytotoxicity of Ly-49A⁺ NK cells. The level of this inhibition was significantly lower than that of wild-type D^d or D^d mut1, which almost completely inhibited cytotoxicity of Ly-49A⁺ NK cells. The difference seen in the two assays is probably due to low affinity interaction between D^d and Ly-49A or -C. It is likely that binding of D^d-transfected cells to Ly-49-transfected COS cells requires a large number of Ly-49 molecules binding to D^d, allowing high enough avidity interaction to maintain cell-cell binding during the washing procedure. In contrast, inhibition of NK cytotoxicity may require much less receptor (Ly-49A)-ligand (D^d) interaction. Thus, while the affinity of D^d mut2 for Ly-49A seems too low to mediate cell-cell adhesion, it is sufficient to partially inhibit cytotoxicity of Ly-49A⁺ NK cells. It should be noted that the expression levels of D^d and the two mutant D^d on the transfected cells in these studies were comparable, and the mutations did not seem to affect the conformation of D^d as determined by the binding of conformation-dependent mAbs. Therefore, these results suggest that carbohydrates linked to residue 176, but not residue 86, modulate the affinity of D^d for Ly-49.

The view that asparagine 176-linked carbohydrates on D^d plays a role in the interaction with Ly-49 is consistent with previous findings. Ly-49A and -C have been shown to bind certain polysaccharides (16, 17, 18). Furthermore, binding of class I MHC to Ly-49C is also inhibited by treatment of class I MHC-positive cells with tunicamycin (17) or fucosidase (18). These results indicate that carbohydrates on class I MHC are involved in the recognition by Ly-49. On the other hand, it is unlikely that the binding of class I MHC to Ly-49 is mediated solely by carbohydrates. It has been shown that recognition of class I MHC by Ly-49 depends on the conformation of the former since Ly-49A is unable to recognize class I MHC lacking bound peptides (13). Therefore, Ly-49 most probably recognizes a combination of carbohydrates and the peptide backbone structure of class I MHC. This is consistent with our previous finding that the binding of Ly-49 to class I MHC involves not only the carbohydrate recognition domain of Ly-49 but also a portion of the stalk region immediately adjacent to the carbohydrate recognition domain (14, 16). The involvement of carbohydrates of class I MHC in the binding to Ly-49 implies that NK cytotoxicity may be regulated not only by the expression level of class I MHC on target cells but also by the pattern of glycosylation of class I MHC.

While this manuscript was under review, Matsumoto et al. reported similar studies with different results (26). In their studies, mutations at the conserved glycosylation sites of D^d had no detectable effects on inhibition of NK cytotoxicity or binding of D^d-transfected C1498 cells to Ly-49A-transfected Chinese hamster ovary cells. The reasons for the discrepancies are unknown at this time, but they may be due to differences in cells used, mutations (substitutions of asparagines with glutamines in their studies vs substitution of serine or threonine with glycine in this study), or expression levels of D^d on target cells. Our studies have shown that mutation of the glycosylation site at residue 176 reduces the binding of D^d to Ly-49, but the binding is sufficient to induce partial inhibition of cytotoxicity of Ly-49A⁺ NK cells. It is conceivable that, if the expression level of this mutant D^d on target cells is high enough, it can protect the targets from cytotoxicity.

It is of interest that glycosylation at residue 176, but not residue 86, of class I MHC is involved in the binding to Ly-49. Residues 86 and 176 are located at the end of the α1 and α2 domains, respectively, and are outside of the peptide binding groove. The residue 86 glycosylation site is invariably conserved among all class I MHC thus far sequenced (19, 21). In contrast, the residue 176 glycosylation site is conserved only among murine class I MHC (Fig. 5). A survey of various class I MHC sequences in the data base showed that this glycosylation site in the α2 domain is conserved among all classical murine class I MHC including K^d, K^b, K^k, K^f, D^d, D^b, D^k, and D^f but is not found in any others examined. In addition to the class I MHC sequences listed in Figure 5, those of human (10 HLA-A, 30 HLA-B, 7 HLA-C), six rat, 13 chimpanzee, 5 orangutan, 14 gorilla, 3 baboon, 3 gibbon, 11 rhesus monkey, 6 cat, 3 rabbit, 3 cow, 5 horse, 2 cheetah, 2 squirrel, 2 frog, and 4 fish were examined. The glycosylation site at residue 176 was not found among any of the nonmurine class I MHC. This indicates that this glycosylation site in the α2 domain likely arose after the divergence of mice from other species and has became fixed in all murine class I MHC genes.

Carbohydrates on residue 86 are thought to be important for the binding of class I MHC to the chaperone calnexin during the assembly of class I MHC α-chain with β₂ microglobulin and peptides (19, 27, 28). On the other hand, the functional role of carbohydrates on residue 176 of murine class I MHC has not yet been described. Since it is conserved among murine class I MHC and not found among others, its function should be uniquely attributed to mice. The results in this study show that carbohydrates on residue 176 function as a part of the ligand structure for the murine NK receptor Ly-49. Therefore, it is possible that murine class I MHC conserved this glycosylation site during evolution to maintain the function to serve as a ligand for the Ly-49 family. The complex and highly polymorphic Ly-49 family likely evolved in response to the rapid evolution of murine class I MHC. The lack of this N-linked glycosylation in class I MHC of other species suggests that the murine Ly-49 family diverged from other NK receptors with respect to how they recognize class I MHC on target cells. Although Ly-49 has also been found in rats (29, 30), direct interaction between rat Ly-49 and rat class I MHC has not been demonstrated. The absence of glycosylation sites in the α2 domain of rat class I MHC suggests that binding specificities of rat Ly-49 may be quite different from murine Ly-49. Moreover, efforts to isolate homologous human Ly-49 genes using mouse probes have not been successful. While a human Ly-49 gene family may exist, our results support the hypothesis that it may be significantly diverged from mouse Ly-49, reflecting the rapid evolution of genes involved in immune functions.

We thank Dr. V. Kumar for the 5E6 hybridoma.