The Effect of Human β2-Microglobulin on Major Histocompatibility Complex I Peptide Loading and the Engineering of a High Affinity Variant

The ability to directly load cell surface major histocompatibility complex (MHC) class I molecules with peptides provides a potentially powerful approach toward the development of vaccines to generate cell-mediated immunity. We demonstrate that exogenous β2-microglobulin (β2m) stabilizes human cell surface MHC I molecules and facilitates their loading with exogenous peptides. Additionally, using three-dimensional crystal structures and known interaction sites between MHC I heavy chains and β2m, we engineered variants of human β2m (hβ2m) with a single serine substitution at residue 55. This alteration was predicted to promote hydrophobic interactions at the MHC I heavy chain/β2m interface and displace an ordered water molecule. Compared with hβ2m, the serine to valine substitution at residue 55 had improved ability to bind to cell surface HLA-A1, HLA-A2, and HLA-A3 molecules, facilitate exogenous peptide loading, and promote recognition by peptide-specific T cells. The inclusion of hβ2m or higher affinity variants when pulsing cells with MHC-restricted peptides increases the efficiency of peptide loading 50–80-fold. Therefore, the inclusion of hβ2m in peptide-based vaccines may increase cell surface antigen densities above thresholds that allow recognition of peptide antigens by the immune system, particularly for cryptic, subdominant, or marginally antigenic peptides.

In the field of vaccine development, it has been relatively simple to induce humoral responses to injected antigens. However, one of the major challenges in the treatment of tumors and viral infections is the generation of vaccines that stimulate cell-mediated immune responses to these pathogens. Specific cytolytic responses are generally mediated by CD8ϩ T cell recognition of antigenic peptides in the context of major histocompatibility complex (MHC) 1 class I molecules. Recent ad-vances in defining "supermotif" antigens capable of being presented by multiple MHC I alleles (1)(2)(3)(4) and immunodominant epitopes (5-9) will undoubtedly have a significant impact on realizing these goals. However, beyond defining appropriate antigenic peptides, a second challenge lies in establishing effective methods with which to deliver these antigens to the MHC I loading pathway. Typically, MHC I molecules acquire peptides generated by the degradation of endogenous proteins by the proteasome. These peptides are transported into the endoplasmic reticulum, where they are bound by MHC I⅐␤ 2microglobulin (␤ 2 m) complexes and finally transit to the cell surface (10 -12). Although this pathway is the dominant means of loading MHC I molecules, other methods of delivering antigenic peptides to MHC I have been described including direct incorporation of DNA into cells (13)(14)(15)(16), phagocytosis-dependent representation of antigens (17,18), and infection by bacterial (19) or viral vectors (20,21). All of these approaches attempt to introduce peptide into the endogenous loading pathway. Alternatively, direct cell surface loading of MHC I molecules in vitro has also been demonstrated but is inefficient in the absence of serum (22,23). Thus, peptide antigens can be bound to MHC I molecules, but they must either be introduced into the endogenous pathway or be loaded exogenously.
Some active and approved clinical trials involve the direct injection of free peptide (usually in an oil-based adjuvant) into patients to induce cell-mediated immunity (e.g. Protocol ID numbers NCI-95-C-0145D, NCI-T95-0031N, NCI-95-C-0143B, NCI-T95-0039N, MSKCC-95052, NCI-H95-0769, LAC-USC-10M954, NCI-T95-0069D, NCI-97-C-0141, NCI-T96-0078, NCI-98-C-0023, NCI-98-C-0022, 94-C-0159). 2 The most likely mechanism for the formation of peptide-MHC I complexes in these cases is the exogenous binding of free peptide to cell surface MHC I molecules. Successful exogenous loading requires that added peptides bind to cell surface MHC I molecules by exchange of their endogenously derived peptides. Since MHC I molecules that have lost their endogenously loaded peptides are inherently unstable (24), efficient loading of exogenous peptide presents a significant challenge. We, as well as others, have demonstrated that empty MHC I molecules can be stabilized with exogenous human ␤ 2 m (h␤ 2 m) (23,25,26). Thus, the generation of higher affinity variants of ␤ 2 m could both quantitatively and qualitatively affect the efficiency of exogenous peptide loading by increasing both the amount of peptide loaded as well as the length of time that the peptide remains bound. These qualitative and quantitative effects may allow the immune system to generate cellular responses to lower affinity peptides. Responses to lower affinity antigens may in turn lead to recognition and destruction of neoplastic or virally infected cells that would otherwise be below the threshold of recognition by the immune system.
Site-directed mutagenesis of MHC I heavy chains has identified residues that affect peptide association, T cell recognition, and stability with ␤ 2 m (27)(28)(29). Additionally, a great deal of work has been done identifying, analyzing, and modifying peptides bound to MHC molecules. In contrast, however, much less work has been done engineering and comparing different ␤ 2 m molecules, with a few notable exceptions comparing murine with human ␤ 2 m and mutagenizing h␤ 2 m (26, 30 -35). Further, comparisons of h␤ 2 m and murine ␤ 2 m (m␤ 2 m) reveal that h␤ 2 m binds with higher affinity to murine MHC I heavy chains than m␤ 2 m (31), and we have identified a region of h␤ 2 m (the S5 strand) responsible in part for this effect (26). Using this information and the available three-dimensional crystal structures of MHC I molecules, we engineered a variant of h␤ 2 m to create an ionic bond that stabilized interactions with the murine MHC I heavy chain H-2D d to a greater extent than wild-type h␤ 2 m (26).
We have now extended this work to human MHC I molecules. In this report we have utilized four independent means to compare variant h␤ 2 m molecules: 1) an assay in which exogenous ␤ 2 m and peptide stabilize cell surface MHC molecules that have been stripped of their endogenous ␤ 2 m and peptide (36,37); 2) an inhibition assay we have developed that directly compares the ability of different ␤ 2 m species to bind to cell surface MHC molecules; 3) direct peptide binding to MHC I molecules using a FITC-labeled peptide; and 4) a cytotoxic T lymphocyte (CTL) lysis assay, which indirectly measures the loading of antigenic peptides onto target cells based on the lysis of those target cells by CTL clones. We have engineered a higher affinity variant of h␤ 2 m that, by the above criteria, stabilizes human MHC I molecules to a greater extent than wild-type h␤ 2 m, enhances peptide binding, and enhances the ability of CTL to recognize peptide-loaded target cells. Further, this stabilizing activity was observed with all three HLA alleles examined: HLA-A1, HLA-A2, and HLA-A3.

EXPERIMENTAL PROCEDURES
Site-directed Mutagenesis h␤ 2 m cDNA in Bluescript SK (Stratagene, La Jolla, CA) was mutated using the ExSite mutagenesis system (Stratagene) according to the manufacturer's protocol and subcloned into the bacterial expression vector pET-21d(ϩ) (Novagen, Madison, WI) using engineered NcoI and BamHI sites at the 5Ј-and 3Ј-end of the mature protein sequence, respectively. The D53 variants were described previously (26). The oligonucleotides used for the S55 mutagenesis were as follows: sense S55I, 5Ј-ATT TTC AGC AAG GAC TGG TCT TTC-3Ј; sense S55V,  5Ј-GTG TTC AGC AAG GAC TGG TCT TTC-3Ј; common antisense,  5Ј-TAA GTC TGA ATG CTC CAC TTT TTC-3Ј. The oligonucleotides used for the insertion of the 9E10 epitope (amino acid sequence EQKLISEEDLN) at the amino terminus of wild-type ␤ 2 m to create the myc-␤ 2 m were as follows: sense myc, 5Ј-TCC GAG GAG GAC CTG AAC ATC CAG CGT ACT CCA AAG ATT CAG G-3Ј; Antisense myc, 5Ј-AAT AAG CTT CTG CTC CAT GGC CTC GAG GCC AGA AAG AGA GAG-3Ј. Mutagenized or added regions are underlined. Constructs were confirmed by sequence analysis using standard techniques.

Synthesis and Purification of Recombinant ␤ 2 m
Recombinant ␤ 2 m expression and purification has been described previously (26). Briefly, ␤ 2 m constructs in pET-21d(ϩ) were transformed into the BL21(DE3) strain of Escherichia coli. At an A 600 of 0.6, cultures were induced with 1 mM isopropyl-1-thio-␤-D-galactopyrano-side for 4 h, and inclusion bodies were isolated by centrifugation after sonication of bacteria in 200 mM Tris, 2 mM EDTA, 10% Triton X-100, pH 7.6, and washing in 200 mM Tris, 2 mM EDTA, pH 7.6. Inclusion bodies were solubilized in 6 M guanidine HCl containing 0.3 M dithiothreitol, 100 mM Tris, pH 8.0, and a mixture of protease inhibitors (5 g/ml leupeptin, 0.5 mM 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride, 1% aprotinin). Following overnight dialysis in 6 M guanidine, pH 2.0, recombinant protein was refolded over 72 h in 0.4 M arginine, 5 mM oxidized glutathione, 100 mM Tris, 2 mM EDTA at 10°C. Following refolding, preparations were dialyzed exhaustively against 0.4 M arginine, 100 mM Tris, 2 mM EDTA, pH 8.0, and then PBS at 4°C. Preparations were purified as a single peak by preparative fast protein liquid chromatography on a Superdex 75pg gel filtration column (Amersham Pharmacia Biotech), concentrated using Centriprep-3 concentrating units (Amicon, Beverly, MA), and sterile-filtered, and concentrations were calculated based on A 280 readings. Recombinant ␤ 2 m was judged to be Ն95% pure based on analysis by SDS-polyacrylamide gel electrophoresis, and analytical fast protein liquid chromatography. Independent preparations have been compared for specific activity and shown to be identical at a molar level with no loss in activity when stored at Ϫ70°C for Ͼ6 months and at 4°C for Ͼ2 months. Recombinant h␤ 2 m was also shown to be identical in activity to native h␤ 2 m purified from urine (Sigma) (38).

Cytotoxic T Lymphocyte Lysis Assays
Cytolytic activity was assessed using a modified Europium release assay (50) as follows. Hmy2.C1R cells transfected with HLA-A2 or HLA-A3 were resuspended to 4 ϫ 10 6 cells/ml in complete Dulbecco's modified Eagle's medium supplemented with 20 M BATD (which forms a fluorescent chelate with Europium; Wallac, Gaithersburg, MD) and incubated at 37°C for 30 -60 min. Cells were resuspended in 10 ml of serum-free (SF) CTL medium (Iscove's modified Dulbecco's medium (IMDM) supplemented with 5 mg/ml bovine serum albumin (BSA; Sigma), 2 mM L-glutamine, 1.25 mM sulfinpyrazone (Sigma), and 1% penicillin/streptomycin), centrifuged, and washed once more with SF CTL medium. Cells were then pulsed with peptide with or without ␤ 2 m in SF CTL medium for 60 -90 min at 37°C. Cells were washed twice in SF CTL medium, resuspended in CTL medium (5% fetal calf serum in lieu of BSA), and combined at the designated effector:target ratio with CTL clones in round bottom microtiter plates. Plates were gently centrifuged at 100 ϫ g for 2 min and then incubated at 37°C for 2 h. Finally, plates were centrifuged at 300 ϫ g, and 20 l/well was transferred to 200 l of 0.3 M acetic acid, 60 mM sodium acetate, 7.5 g/ml Europium (Aldrich, Milwaukee, WI), and the plate was read on a Wallac 1234 DELFIA fluorometer. The percentage of specific lysis was calculated with the following equation: 100 ϫ ((experimental Ϫ blank) Ϫ (spontaneous Ϫ blank))/((maximum Ϫ blank) Ϫ (spontaneous Ϫ blank)). Triplicate samples were used, and error bars indicate S.D. values.

FITC-conjugated Peptide Binding to Cell Surface MHC I Molecules
Hmy2.C1R-A2 cells were resuspended in SF IMDM (IMDM supplemented with 2.5 mg/ml BSA, 20 mM HEPES, 2 mM L-glutamine, 1% nonessential amino acids, 1% penicillin/streptomycin, and 0.04 mg/ml Gentamicin sulfate) with the indicated concentrations of peptide with or without recombinant ␤ 2 m and incubated at 37°C for 17 h. Cells were washed twice with PBS plus 2 mg/ml BSA (PBSA) and resuspended in PBSA followed by flow cytometric analysis on a FACS analyzer (Becton Dickinson, Mountain View, CA). 8000 -20,000 live (propidium iodideexcluding) events were collected per sample, and values are expressed as mean fluorescence intensity. ED 50 values were calculated using a sigmoid logistic fit.

Cell Surface MHC Stabilization of Acid-stripped Cells
The MHC stabilization was done essentially as described previously (36,37,51) with minor modifications. Briefly, Hmy2.C1R-A1, -A2, and -A3 cells were washed twice with PBS, resuspended in 0.13 M citric acid, 66 mM Na 2 HPO 4 , pH 2.9 (pH 3.2 for A2 cells), for 90 s at 4°C, washed with two 50-ml changes of IMDM, and resuspended in SF IMDM. 10 5 cells/well were added to a 96-well microtiter plate containing hybridoma supernatants, peptide, and ␤ 2 m dilutions in a total volume of 150 l. After a 4-h incubation at 23°C, cells were washed twice with PBS, 2 mg/ml BSA, 0.02% NaN 3 (FACS buffer) and stained with FITCconjugated goat anti-mouse IgG (H ϩ L) F(abЈ) 2 fragment (Cappel/ Organon Teknika, Durham, NC) for 1 h at 4°C. Cells were washed twice with FACS buffer and fixed in 1% formaldehyde in PBS followed by flow cytometric analysis on a FACScan II (Becton Dickinson, Mountain View, CA). Values are expressed as mean fluorescence intensity.

myc-h␤ 2 m Binding and Inhibition Assays
myc-␤ 2 m Binding Assay-Hmy2.C1R transfectant cells at 2.5 ϫ 10 5 cells/tube in a 500-l volume were incubated at 37°C for 16 h in SF IMDM with different concentrations of myc-␤ 2 m with or without peptide. Cells were washed three times in plain IMDM followed by incubation with 9E10 (anti-myc) hybridoma supernatant at 4°C for 1 h. After washing with IMDM, cells were stained for 1 h with FITC antimouse IgG at 4°C for 1 h. Cells were washed a final time in FACS buffer, and live cells (propidium iodide-excluding) were analyzed by flow cytometry.
myc-␤ 2 m Inhibition Assay-The inhibition assay is identical to the binding assay with the following modifications: 2.5 M myc-␤ 2 m was used in all cases, and different concentrations of non-myc-labeled recombinant ␤ 2 m were included to inhibit myc-␤ 2 m binding to cell surface MHC molecules. The percentage of inhibition was calculated by the following equation: (1 Ϫ ((experimental Ϫ background)/(no inhibitor Ϫ background))) ϫ 100. 10,000 -20,000 gated events/sample were counted, and all experiments were repeated at least twice. ID 50 values were calculated using a sigmoid logistic fit.

␤ 2 m Enhances CTL Recognition of Target Cells and Exogenous Peptide
Loading-The effectiveness of ␤ 2 m in facilitating exogenous peptide loading of MHC I molecules was measured using a CTL lysis assay. Target cell lysis not only correlates with the loading of a specific peptide antigen, but it also demonstrates that the peptide is bound in an immunologically relevant manner. Hmy2.C1R-A2 target cells (a human lymphoblastoid cell essentially null for HLA molecules except for the transfected HLA-A2.1) (40,41) were pulsed with a suboptimal concentration of HTLV-1 TAX peptide for 90 min in SF CTL medium in the absence or presence of increasing concentrations of purified, recombinant h␤ 2 m and then used as targets in a conventional lysis assay. The presence of h␤ 2 m dramatically increased the specific lysis by the TAX-specific CTL clone in a dose-dependent fashion (Fig. 1). Lysis at the maximal amount of h␤ 2 m in the presence of an irrelevant A2-binding peptide was at background levels. Using this suboptimal concentration of peptide, there was 20% lysis in the absence of ␤ 2 m. The addition of 8 M ␤ 2 m increased the lysis to the maximum observed at this effector:target ratio. In the absence of ␤ 2 m, a 50 -100-fold higher concentration of peptide would be required to achieve comparable levels of lysis. 3 To directly demonstrate the effectiveness of h␤ 2 m to enhance peptide loading of MHC I molecules, we measured the binding of a FITC-labeled, HLA-A2-restricted peptide (HIV RT 195-207) to Hmy2.C1R-A2 cells. As shown in Fig. 2a, the presence of h␤ 2 m shifted the peptide titration curve approximately 50fold compared with peptide loading in the absence of h␤ 2 m, consistent with the magnitude of shift observed for the augmentation of CTL lysis.
The S55V Mutation Enhances MHC Molecule Stability on the Cell Surface-Based on previous work, the region of h␤ 2 m including the S5 strand was found to play an important role in the higher affinity of h␤ 2 m for murine MHC I heavy chains (26) and was the focus of our structural search for mutations to increase the affinity of h␤ 2 m. Using the HLA-A2 crystal structure (52), 4 we identified a hydrophilic serine residue at position 55 of h␤ 2 m that was buried at the h␤ 2 m/heavy chain interface and was situated directly adjacent to an ordered water molecule. This residue was substituted with hydrophobic residues of increasing mass (valine and isoleucine) in order to promote hydrophobic interactions and exclude the ordered water. The valine side chain was chosen, since it is approximately the volume of a serine side chain hydrogen-bonded to a water molecule. Serine 55 was also changed to isoleucine, since it was unclear if and how the hydrophobic residues would pack in relation to other side chains at the h␤ 2 m/heavy chain interface.
We initially used an HLA stabilization assay (36,37) to screen the ␤ 2 m variants. Since there can be significant allelespecific differences in the ability of h␤ 2 m point mutants to stabilize murine heavy chains 3 (26) a number of HLA alleles, HLA-A1, HLA-A2, and HLA-A3, were analyzed in order to determine the generality of any observed effects. Fig. 3 demonstrates the ability of the S55 substituted ␤ 2 ms to stabilize HLA-A1 (Fig. 3a), HLA-A2 (Fig. 3b), and HLA-A3 (Fig. 3, c and d) in the presence of a specific binding peptide and an appropriate HLA-specific antibody. The S55V variant (Fig.  3, ϫ) stabilized HLA-A1 and HLA-A3 approximately 2-and 3-fold better, respectively, than wild-type h␤ 2 m (diamonds) at a molar level, and effects on HLA-A2 stabilization by S55V were only slightly better than those observed with wild-type h␤ 2 m. The effects of S55I (triangles) varied depending on the allele: better along the upper end of the titration curve with HLA-A1, worse with HLA-A2, and marginally better with HLA-A3. We have previously demonstrated that a single mutation at residue 53 (D53N or D53V) of h␤ 2 m decreased the affinity of its interaction with murine heavy chains, presumably by disruption of a conserved, coordinated ionic bond (26). As was observed with murine heavy chains, both Asp 53 mutants (circles and stars) were deficient relative to wild-type h␤ 2 m at stabilizing all three human heavy chain alleles.
We observed a decline in the stabilization of all three MHC I alleles at the highest concentrations of ␤ 2 m tested when using mAb BB7.5 (Fig. 3, a-c). The choice of this mAb for the stabilization experiments in Fig. 3 was determined by the mAb's dependence on both ␤ 2 m and peptide for the formation of the epitope as well as the fact that it was reactive with all three alleles studied. Similar high dose inhibition was also observed even when freshly isolated cells were incubated on ice in the presence of BB7.5, 3 suggesting that this effect was due to inhibition of BB7.5 binding by free ␤ 2 m in solution, consistent with ␤ 2 m forming part of the combinatorial BB7.5 epitope (53). Using an alternative, HLA-A3-specific mAb to compare wildtype h␤ 2 m and S55V, no inhibition was observed (Fig. 3d). Further, the 2.5-fold molar increase in the ability of the S55V mutant to stabilize HLA-A3 was consistent with the 3-fold increase when using BB7.5 (Fig. 3c). Interestingly, the pan-HLA-ABC-specific mAb W6/32 could not be used due to its inability to bind MHC complexes containing recombinant ␤ 2 m, which have an additional methionine at the amino terminus (38). Fig. 3 were selected due to their dependence on both ␤ 2 m and peptide in order to detect "complete" molecules, i.e. heavy chain⅐␤ 2 m⅐peptide natively folded trimeric complexes. Since this binding assay required the presence of an antibody in addition to ␤ 2 m and peptide (37), there was formally the possibility that the antibodies exerted an effect that was specific for a particular ␤ 2 m mutant such as S55V. Due to previously reported concerns regarding the potential contribution of the antibodies to the stabilization of cell surface MHC I complexes (54), we extended the stabilization studies by developing a binding inhibition assay that directly measured the relative abilities of ␤ 2 ms to bind to MHC I molecules. To do this we engineered an epitope tag (myc) onto the amino terminus of h␤ 2 m so that its binding could be directly measured with the anti-myc mAb 9E10. The addition of this tag to the amino terminus was preferable to biotinylation, iodination, or other labeling procedures that could variably alter residues involved in heavy chain/␤ 2 m interactions and therefore require further purification (55).

The Peptide Dependence of h␤ 2 m Binding and Its Augmentation by S55V-The antibodies used in
Hmy2.C1R cells transfected with HLA-A1, -A2, or -A3 alleles were incubated overnight in SF medium with increasing concentrations of myc-␤ 2 m (Fig. 4). In the presence of an appropriate peptide, there was concentration-dependent myc-␤ 2 m binding for all alleles studied. However, when cells were incubated with myc-␤ 2 m in the absence of peptide, very little myc-␤ 2 m binding was observed. It should be noted that these experiments were conducted for 16 h, whereas comparable or even greater binding of human or murine ␤ 2 m to murine MHC I molecules occurred in only 2 h even in the absence of peptide. 5 This suggests some fundamental differences in the stability and behavior of murine and human MHC I molecules.
Having established conditions in which myc-␤ 2 m binding could be demonstrated, the relative abilities of wild-type h␤ 2 m and S55V to inhibit the binding of myc-␤ 2 m to HLA molecules were next compared (Fig. 5). Compared with wild-type h␤ 2 m, the S55V mutant inhibited myc-␤ 2 m binding about 2-fold better at a molar level for HLA-A1, -A2, and -A3. These results are consistent with a higher relative affinity of the S55V mutant compared with wild-type h␤ 2 m for HLA-A1, -A2, and -A3.
The S55V Mutant Enhances Peptide Loading and CTL Recognition-Having established the higher relative affinity of the S55V mutant for HLA-A1, -A2, and -A3, we next studied its ability to enhance the loading of peptide onto Hmy2.C1R-A2 cells directly with the FITC-labeled HIV RT peptide. In the presence of a limiting concentration of FITC-peptide (100 ng/ ml; see Fig. 2a), h␤ 2 m and S55V were titered onto cells to measure their ability to enhance exogenous peptide loading (Fig. 6). The S55V mutant was approximately 2-fold more effective on a molar basis than wild-type h␤ 2 m at promoting peptide binding. When binding was examined over a wide range of peptide concentrations, the ED 50 values in the presence of an almost saturating concentration of h␤ 2 m and S55V (10 M) were 54-and 77-fold greater, respectively, than observed in the absence of ␤ 2 m (Table I).
The augmented ability of the S55V variant to enhance exogenous peptide loading of cells suggested that these cells would also be better targets for CTL recognition. To test this possi-5 M. J. Shields and R. K. Ribaudo, manuscript in preparation. bility, two CTL clones, specific for an HTLV-1 TAX peptide in the context of HLA-A2 and an influenza nucleoprotein peptide in the context of HLA-A3, were used in a standard lysis assay of Hmy2.C1R transfectants pulsed with a suboptimal concentration of antigenic peptide. The S55V mutant was 4-fold more effective at a molar level than wild-type ␤ 2 m at enhancing target cell lysis for HLA-A2 (Fig. 7a) and 6-to 7-fold better for HLA-A3 (Fig. 7b). Controls with irrelevant HLA-A2 and HLA-A3 binding peptides with the highest concentrations of ␤ 2 m used resulted in only background levels of killing. Additionally, multiple TAX-specific A2-restricted clones displayed similar levels of S55V-enhanced killing relative to wild-type ␤ 2 m (data not shown). DISCUSSION MHC I molecules have evolved to present endogenously derived antigens and are relatively refractory to peptide loading unless ␤ 2 m is present (22, 23, 56 -58). Data presented in Figs. 1-3 demonstrate the extent to which ␤ 2 m facilitates peptide loading and MHC stability. In the absence of ␤ 2 m, up to 100fold higher concentrations of peptide are required to achieve similar levels of peptide loading and target cell lysis. Since peptides can only bind to MHC I molecules already devoid of peptide and since these empty molecules are normally very unstable (23), the mechanism of exogenous ␤ 2 m's effect on peptide binding is probably via the prolongation of the cell surface half-life of empty MHC I molecules. Therefore, if the exogenous ␤ 2 m concentration is high enough or, alternatively, of high enough affinity, then these empty "peptide-receptive" molecules may remain available on the surface long enough for exogenously added peptide to bind. This ␤ 2 m effect may even be more pronounced in vivo when peptide is administered as a vaccine, because diffusion and/or degradation of these peptides would greatly limit the amount available for cell surface binding. Additionally, a higher affinity ␤ 2 m could result in decreased dissociation rates of peptide⅐MHC complexes (a longer cell surface half-life) and, therefore, an increased antigen density of the relevant antigenic MHC complexes.
The higher affinity of h␤ 2 m for murine MHC heavy chains (31) increases the amount and stability of cell surface MHC I complexes that can be exogenously loaded with antigenic peptide relative to m␤ 2 m (26). The resulting antigenic peptide⅐ murine MHC⅐h␤ 2 m complexes that are longer lived would be more likely to induce cell-mediated immune responses. In fact, peptide-specific MHC I-restricted CTL have been successfully elicited in mice by in vivo injection of peptide in the presence but not the absence of h␤ 2 m (59). Murine ␤ 2 m is far less efficient at stabilizing cell surface MHC I, presumably due to its lower affinity interaction with heavy chains (26,31). It might therefore be predicted that in vivo immunization of mice with peptide and murine ␤ 2 m would have been much less Although some studies demonstrate that peptide and ␤ 2 m exchange can be independent or even antagonistic (60 -62), these data clearly demonstrate the enhancement of peptide binding by h␤ 2 m. It should also be noted that the system reported here employs human ␤ 2 m and human MHC I molecules, while previous reports dealt exclusively with the murine equivalents (62) or h␤ 2 m with murine MHC I molecules (60,61).
There are notable differences between the abilities of h␤ 2 m to exchange onto human MHC I molecules versus m␤ 2 m exchanging onto murine MHC I molecules. The myc-␤ 2 m binding experiments in Fig. 4 require 16 h incubations in the presence of peptide to demonstrate significant ␤ 2 m binding. In contrast, comparable binding of m␤ 2 m to murine MHC I molecules is observed within 2 h, even in the absence of peptide (26). Another study demonstrated that up to 5% of HLA molecules on normal T cells exchanged directly fluoresceinated h␤ 2 m (35). The authors noted that HLA-A2 on transfected C1R lymphoblastoid cells (such as those used in this study) exchanged much less h␤ 2 m than did normal human T cells (35). Based on our study, this exchange would likely be greatly increased in the presence of an A2-restricted peptide. The relatively small amount of peptide-independent exchange demonstrated in the human system was in great contrast to that noted for murine cells (35), consistent with our findings (26). Additionally, others have shown that HLA-A3 and HLA-B27 transfected into the TAP-deficient T2 or RMA-S cells are refractory to peptide and ␤ 2 m loading, while the murine alleles H-2K b and H-2D b are effectively stabilized (63).
If the exogenous loading characteristics of murine and human MHC I molecules are fundamentally different, it may be advantageous or even necessary to use mice transgenic for human MHC I and h␤ 2 m (64 -70) in the development of human peptide-based vaccine protocols. The synergy of loading peptide in the presence of higher affinity ␤ 2 ms and the requirement of h␤ 2 m to induce CTL in mice (59) indicates the importance of testing immunization regimens with native h␤ 2 m and higher affinity variants of ␤ 2 m. Therefore, the engineering of higher affinity variants of human ␤ 2 m represents a significant complementary approach to defining immunodominant epitopes and MHC supermotifs for the development of peptide-based vaccines for the treatment of tumors and viruses.
The ability of ␤ 2 m to bind and stabilize MHC I molecules is clearly related to its relative affinity for the MHC I heavy chains (26). For the development of murine vaccination models, the availability of a naturally occurring higher affinity ␤ 2 m (h␤ 2 m) provides this essential component (31,59). In contrast, there are no known naturally occurring ␤ 2 ms available with higher affinity for human MHC I heavy chains than h␤ 2 m. The availability of three-dimensional structural data for a number of MHC I molecules as well as our previous studies examining chimeric m␤ 2 m/h␤ 2 m molecules (26) led to the identification of a specific region and, ultimately, a single residue of ␤ 2 m (serine 55) to mutate in order to increase its affinity for human MHC I heavy chains.
The identification of the S55V mutant was aided in large part by the availability of a three-dimensional crystal structure for MHC I molecules and underscores the importance and utility of advances in the field of structural biology. In the HLA-A2 structure, serine 55 of ␤ 2 m (at the ␤ 2 m/heavy chain interface) is directly adjacent to an ordered water molecule (52). 4 Our earlier work suggested that residue changes that promoted protein/protein interactions (e.g. formation of salt bridges) could predictably improve stability between the two MHC I subunits. Since hydrophobic interactions contribute   7. Enhanced target cell lysis mediated by the S55V mutant. The S55V mutant enhances CTL recognition better than wildtype h␤ 2 m. Hmy2.C1R-A2 (a) or Hmy2.C1R-A3 (b) target cells were pulsed with BATD (Europium chelator) for 30 min and washed in SF CTL medium followed by incubation with a suboptimal dose of TAX peptide (9.3 pM) or control A2-binding HIV gag peptide at 1 nM with the indicated concentrations of ␤ 2 m for 90 min (a) or with a suboptimal dose of NP 265-273 peptide (100 pM) or control A3-binding pn2a.A3 peptide at 1 M with the indicated concentrations of ␤ 2 m for 60 min (b). Target cells were washed and then combined with the TAX-specific HLA-A2restricted CTL clone N1218 at an effector:target ratio of 4:1 (a) or the NP-specific HLA-A3 restricted CTL clone 2711 at an effector:target ratio of 2:1 (b) for 2 h. 20 l of supernatant was harvested, added to 200 l of Europium solution, and read on a Wallac 1234 DELFIA fluorometer. All incubations were done at 37°C. significantly to the overall stability of proteins (71,72) and the exclusion of water molecules from an otherwise hydrophobic interface would result in a gain of entropy and make the interaction more thermodynamically favorable (72), we mutated the Ser 55 residue to the hydrophobic residues valine and isoleucine. By making these variants, we maximized the opportunity to occupy the hydrophobic cavity without sterically hindering the ␤ 2 m/heavy chain interaction.
By four different criteria (MHC complex stabilization, ␤ 2 m binding, peptide binding, and CTL activity), we have demonstrated that the S55V mutant of human ␤ 2 m binds to human MHC I molecules and enhances the loading of antigenic peptide to a greater extent than wild-type h␤ 2 m. Moreover, this higher affinity has been shown to be a general feature for the HLA-A alleles, since it has been extended to all three HLA-A alleles studied. The combination of HLA-A1, -A2, and -A3 is represented in 30 -70% of the population, depending on factors such as ethnicity and country of origin (73,74), suggesting that the S55V mutant may find broad clinical utility in combination with antigenic HLA supermotif peptides that have been selected based on their ability to bind multiple HLA alleles (1)(2)(3)(4).
It will be interesting to determine the effect the S55V mutant ␤ 2 m will have on the stability of lower affinity peptides. The qualitative and quantitative effects of the higher affinity of S55V may prolong the stability of MHC complexes exogenously loaded with lower affinity antigenic peptides on the cell surface, thereby allowing for the induction of cell-mediated responses to cryptic, subdominant, and marginally antigenic peptides. The generation of these responses could serve to eliminate neoplasms and some virally infected cells (e.g. HIV or herpes) that otherwise might evade detection. Coupled with an appropriate regimen to induce a strong cell-mediated immune response, a patient could be cleared of tumors and acquire protective immunity to viruses without the necessity of immunization with the neoplasms/pathogens themselves. Thus, in concert with efforts identifying antigenic T cell epitopes (5)(6)(7)(8)(9) and antigenic MHC supermotif peptides (1-4), h␤ 2 m and higher affinity variants may find global utility in the development of effective, novel peptide-based vaccine protocols to treat both tumors and viral infections.