Characterization of the Interactions Between MHC Class I Subunits: A Systematic Approach for the Engineering of Higher Affinity Variants of β₂-Microglobulin¹

The ability to generate cell-mediated immune responses is critical in the host defense against viral infections and malignant transformation. At the center of these responses is the presentation of viral or tumor-derived Ags to CD8⁺ T cells by MHC class I molecules. Through a variety of mechanisms, some viruses and tumors have devised ways to prevent the priming of cell-mediated immune responses, in some cases by interfering with the efficient expression of MHC class I molecules on the cell surface (1, 2, 3, 4, 5, 6, 7, 8, 9). As the density and half-life of MHC-bound peptides are important in the in vivo priming of cell-mediated immune responses, these effects on expression may serve to block the induction of specific CTL. The half-life of peptide-loaded MHC I molecules on the cell surface can be seconds to days and depends on the affinity of the bound peptide (10). In contrast, empty, or peptide-receptive, molecules are rapidly lost from the cell surface (11, 12, 13), ensuring that under normal circumstances these molecules will not acquire exogenously derived peptides.

Although many peptide-loaded class I molecules possess long half-lives, considerable evidence exists demonstrating that there can be substantial exchange of peptide as well as β₂m on the surface of cells in culture (14, 15). In fact, long before the three-dimensional structure of class I molecules was solved, the exchange of endogenous β₂m on murine L cells with bovine β₂m present in the serum added to cell culture medium was reported (16). Although some reports have suggested that peptide ligand exchange appears to occur independently of β₂m exchange (14, 15) or that they are non-co-operative or even antagonistic processes (17), the presence of β₂m in the culture medium is necessary for effective peptide pulsing of cell surface murine class I molecules to occur (18, 19, 20, 21). Consistent with this, empty murine MHC I molecules can be stabilized by the addition of exogenous human β₂m (hβ₂m)³ (12, 13), making them receptive to loading with exogenous peptides and creating stable MHC I complexes that can stimulate CD8⁺ T cells. Whether the target molecules for these effects of β₂m are pre-existing cell surface molecules that have lost their endogenous peptides, newly emerging peptide-receptive molecules, or even recirculating molecules is not clear. Further, the molecular basis for this effect of hβ₂m remains poorly understood, but probably relies on the greater affinity of human compared with murine β₂m (mβ₂m) for murine MHC heavy chains (22).

Advances in defining the peptide binding motifs of various MHC I molecules, improvements in the ability to predict immunodominant peptide motifs from viral proteins, and sequencing the peptides eluted from the MHC I of isolated tumors provide a basis for the design of peptide-based vaccines. To this end, identifying the molecular nature of the interactions between residues of human or murine β₂m and murine class I heavy chains that contribute to affinity differences will provide the means by which to engineer higher affinity variants of mβ₂m with superior peptide loading ability to be used in animal vaccine models.

Various strategies could be used to create a higher affinity β₂m mutant, including 1) random mutagenesis, 2) interspecies analysis of murine and hβ₂ms to define residues and regions that differ and therefore might affect the interaction with MHC heavy chains, and 3) modeling/mutating amino acid residues based on known crystal structures. Random mutagenesis has been conducted previously on hβ₂m to modulate HLA-B27 activity (23). Preparations of randomly mutated hβ₂m were screened for binding to HLA-B27 with a peptide-sensitive mAb and using functional T cell assays. However, due to the nature of the mutagenesis approach, many of the mutants had multiple changes, which complicates the interpretation of the data with respect to the contributions of individual residues. A second method to create a higher affinity β₂m is to take advantage of serendipitous “experiments” of nature. For example, it has been shown that hβ₂m’s affinity for murine MHC I heavy chains is higher than that of mβ₂m (22). Since 30 of 99 amino acid residues differ between the two, there are various candidate residues that could be responsible for the higher affinity. The generation of chimeric molecules can assist in localizing the region(s) most likely to account for the affinity difference. Furthermore, individual residues can be changed to determine their specific involvement. Finally, the existing three-dimensional crystal structures can be exploited to predict point mutations in β₂m that could lead to more stable interactions between the MHC heavy chain and β₂m. A combination of the latter two approaches has been taken here to better define the interactions between MHC heavy chains and β₂m and to lay the foundation for the rational design of higher affinity hβ₂ms. Using functional information obtained with human:murine chimeric molecules and an obvious conformational difference in the α carbon trace of human and murine β₂ms, an individual amino acid residue has been identified that, when mutated to destroy ionic bond(s), caused predictable functional consequences. Furthermore, a gain of function mutant hβ₂m is described that was engineered to create an ionic bond between β₂m and the MHC heavy chain.

RNA transcription plasmids LdCITE and DdCITE were generated by subcloning H-2Ld cDNAs into the transcription vector pCITE 2a (Novagen, Madison, WI). Murine β₂m cDNA was cloned by PCR and inserted into pCITE 2a as previously described (13). hβ₂m cDNA was a gift from Dr. Ken Parker (National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD) and was similarly subcloned into pCITE 2a. The chimeric β₂m constructs HHM (encoding amino acid residues 1–69 of hβ₂m and residues 70–99 of mβ₂m) and MMH (encoding amino acid residues 1–69 of mβ₂m and residues 70–99 of hβ₂m) were generated by taking advantage of a common EcoRI site found in both human and murine β₂m cDNAs at the position corresponding to amino acid residue 69. Identical restriction digests of both parent vectors were performed, and the inserts were purified and religated into the complementary vector to generate HHM and MMH.

HMM (encoding amino acid residues 1–34 of hβ₂m and residues 35–99 of mβ₂m) was created using splicing by overlap extension (24) using the splicing oligonucleotide (5′-CAT CCA TCC GAC ATT GAA ATC CAA ATG CTG-3′), which encodes residues 30 to 34 of hβ₂m and residues 35–39 of mβ₂m, and its complimentary oligonucleotide (5′-CAG CAT TTG GAT TTC AAT GTC GGA TGG ATG-3′). MHM was also generated by splicing by overlap extension using HHM in pCITE 2a as template and splicing oligonucleotides that encode residues 30 to 34 of mβ₂m and residues 35 to 39 of hβ₂m (5′-G TTC CAC CCG CCT CAC ATT GAA GTT GAC TTA C-3′) and its complimentary oligonucleotide (5′-G TAA GTC AAC TTC AAT GTG AGG CGG GTG GAA C-3′).

For bacterial expression, human, murine, and chimeric β₂m cDNAs were subcloned from pCITE 2a into pET21-d (Novagen). Specifically, a BspHI site and initiation ATG were added immediately 5′ of the first codon of the mature β₂m protein by PCR. This full-length PCR fragment was isolated, digested with BspHI and BamHI (engineered into the 3′ untranslated region), and ligated into pET21-d that had been digested with NcoI and BamHI. All chimeric cDNAs were confirmed by sequence analysis using standard techniques.

The hβ₂m 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(+) (Novagen). The antisense oligonucleotides used for mutagenesis were: D53N, 5′-TAA ATT TGA ATG CTC CAC TTT TTC AAT TCT CTC-3′; D53V, 5′-TAA TAC TGA ATG CTC CAC TTT TTC AAT TCT CTC-3′; and K58E, 5′-TTG TCT TTC AGC GAG GAC TGG TCC TTC-3′. All constructs were confirmed by sequence analysis using standard techniques.

β₂m cDNAs cloned into the bacterial expression vector pET21-d were used to transform BL21 Escherichia coli. Cultures (50–200 ml) were grown to an OD_600nm of 0.6, and bacterial expression was then induced with 1 mM isopropyl β-d-thiogalactoside. Four hours following induction, bacteria were pelleted and washed in 200 mM Tris-HCl (pH 7.6)/2 mM EDTA and lysed by digestion with lysozyme followed by sonication, and the inclusion bodies were isolated by centrifugation. After washing in cold 200 mM Tris/2 mM EDTA (pH 7.6), the inclusion bodies were solubilized at room temperature for at least 1 h in 3 to 5 ml of 6 M guanidine-HCl containing 0.3 M DTT, 100 mM Tris (pH 8.0), and a mixture of antiproteases (5 μg/ml leupeptin, 0.5 mM 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride (AEBSF), and 1% aprotinin). Following overnight dialysis against 2 l of 6 M guanidine (pH 2.0), the recombinant protein was refolded over 72 h in 0.4 M arginine, 5 mM oxidized glutathione, 100 mM Tris, and 2 mM EDTA at 15°C. Following refolding, the protein was dialyzed exhaustively against PBS at 4°C. Recombinant β₂m was judged to be 80 to 95% pure based on analysis by SDS-PAGE and analytical HPLC. Finally, the protein was concentrated using Centriplus-3 concentrating units (Amicon Corp., Danvers, MA) and purified by preparative fast protein liquid chromatography on a Superdex 75 gel filtration column. Following purification, equivalent β₂m concentrations were calculated based on OD_280nm readings.

cDNA clones coding for specific H-2 gene products were linearized 3′ of the cDNA insert by digestion with BamHI. RNA was transcribed from 5 μg of the linearized plasmid with T7 RNA polymerase using the Ribomax T7 RNA transcription system (Promega, Madison, WI) following the manufacturer’s protocol. Following transcription, the cDNA template was digested with RNase-free DNase (Promega), extracted with phenol/chloroform/isoamyl alcohol (25/24/1), and precipitated and washed in ethanol.

RNA was translated using Flexi-Lysate rabbit reticulocyte lysate supplemented with canine pancreatic microsomes (Promega) in a final volume of 50 μl containing 50 μCi of [³⁵S]methionine (SJ.1015, Amersham, Arlington Heights, IL) and 100 mM KCl (to optimize translation of RNA containing the encephalomyocarditis virus 5′ untranslated region found in pCITE transcription vectors) following the manufacturer’s protocols. Reactions were incubated for 90 min at 26°C, were terminated by addition of 2 vol of ice-cold 0.75 M KCl, 20 mM Tris-HCl (pH 7.6), and 10 mM EDTA, and were placed on ice. Individual RNAs were first titrated to determine the relative amounts of input RNA needed in the translation reaction.

Aliquots not exceeding 75 μl from the terminated translation reactions were layered onto 100 μl of an ice-cold sucrose cushion (0.5 M sucrose, 20 mM Tris-HCl (pH 7.6), and 10 mM EDTA) and centrifuged in a Beckman Airfuge (Beckman, Palo Alto, CA) at 22 psi (∼100,000 × g) using an A-100-18 rotor for 15 min. Supernatants were aspirated completely, and pellets were lysed into 50 μl of ice-cold lysis buffer (150 mM NaCl, 50 mM Tris (pH 7.6), 1% Nonidet P-40, 0.5% aprotinin, 0.5 mM AEBSF, 10 μg/ml leupeptin, and 0.03 M iodoacetamide).

Solubilized microsomes (10–15 μl) were aliquoted into tubes containing purified mAbs (150 μg/ml final concentration) in lysis buffer containing 3% OVA (Sigma Chemical Co., St. Louis, MO) in a final volume of 100 μl. Tubes were incubated on ice for 1 h, and 30 μl of protein A-Sepharose was added in lysis buffer. Samples were incubated on a rotator for 30 min at 4°C. Protein A-Sepharose was then pelleted by centrifugation at 12,000 × g, and supernatants were removed by aspiration. Pelleted beads were then washed three times with 1 ml of lysis buffer at 4°C. Finally, pelleted washed beads were resuspended in sample loading buffer (0.05 M Tris-HCl (pH 6.8), 2% SDS, 10% glycerol, 100 μM 2 ME, and 0.01% bromophenol blue), heated in a boiling water bath for 5 min, and electrophoresed on 12.5% SDS-polyacrylamide gels. Gels were subsequently fixed in methanol (7%, v/v)/acetic acid (5%, v/v)/glycerol (10%, v/v), soaked in Enlightning (New England Nuclear-DuPont, Boston, MA), dried, and evaluated by autoradiography. Kodak XAR-5 film (Eastman Kodak, Rochester, NY) was exposed for varying times at −80°C using enhancing screens (DuPont, Wilmington, DE).

Exposed films were developed using an X-OMAT automated developer (Kodak). All autoradiographs were evaluated quantitatively by densitometry. Reference to quantities immunoprecipitated in Results and Discussion derives from such scans. Quantitative densitometry was performed using a Molecular Dynamics scanning densitometer (Sunnyvale, CA) equipped with ImageQuant software (Molecular Dynamics). Quantitation of autoradiographs was conducted on multiple exposures representing different intensities to ensure that scans were within the linear response range of the film. Autoradiographs selected for reproduction in the figures do not necessarily represent the same exposures that were used in densitometry. Calculations of molar ratios in immunoprecipitates were adjusted for the number of methionine residues in the MHC heavy chains, hβ₂m and mβ₂m, respectively.

L cells (DAP-3) and the TAP-1 defective cell line E-3 (EE2H3) (25) were transfected by the calcium phosphate method with H-2L^d H-2D^d, and H-2L^dE9V encoding expression plasmids as previously described (13, 26). B4.2.3 is an H-2D^d-restricted murine T cell hybridoma that is specific for the HIV gp160 envelope protein-derived peptide, p18-I-10 (RGPGRAFVTI) (27). Cells were propagated in DMEM augmented with 10% FCS, 20 mM HEPES, 2 mM l-glutamine, 1% nonessential amino acids, 1% Pen-strep (Biofluids, Rockville, MD), and 0.04 mg/ml of gentamicin sulfate. SKT 4.5 is a DAP-3 cell line stably transfected with genomic H-2D^d (28). The mAbs 28-14-8 and 34-2-12 recognize the α3 domains of H-2L^d and H-2D^d, respectively. The presence of this domain is generally regarded as representative of the total number of molecules, as its formation is independent of association with either β₂m or peptide (29). The mAbs 30-5-7 and 34-5-8 recognize the α2 domains of H-2L^d and H-2D^d, respectively, which typically depend on both association with β₂m and peptide. All mAbs were provided by Dr. David Margulies (National Institute of Allergy and Infectious Diseases, National Institutes of Health).

Peptides pMCMV (corresponding to residues 168–176 of the murine CMV pp89 early regulatory protein YPHFMPTNL) and p18-I-10 (derived from the HIV type 1 isolate envelope glycoprotein 120, residues 318–327, RGPGRAFVTI) were provided by Dr. David Margulies (National Institute of Allergy and Infectious Diseases, National Institutes of Health). All peptides were purified by reverse phase HPLC and were >95% pure as determined by analytical HPLC.

Transfected cells, grown at 37°C with 5% CO₂, were detached from tissue culture flasks by treatment with versene for 2 min and washed with ice-cold serum-free tissue culture medium. Following a 30-min incubation at 37°C, cells were washed and then resuspended in serum-free medium in the presence or the absence of purified recombinant β₂m. The cells were incubated for 2 h at 37°C and then washed with ice-cold FACS buffer (PBS containing 0.2% bovine albumin and 0.025% NaN₃) before staining with conformationally dependent Abs to class I molecules.

Cells were incubated with purified mAbs 30-5-7, 28-14-8, 34-5-8, or 34-2-12 on ice for 1 h, washed with FACS buffer, incubated with FITC-conjugated goat anti-mouse IgG (Cappel Laboratories, Durham, NC), and incubated for an additional 30 min on ice. Cells were then washed and resuspended in 400 μl of ice-cold FACS buffer. Stained cells were treated with propidium iodide and analyzed on a FACScan II analyzer (Becton Dickinson) using CellQuest 1.2 software. Cells were gated for uniform forward and side scatter and negative propidium iodide staining. A minimum of 5000 data points/sample were counted, and all experiments were either performed in triplicate or repeated at least three times. Values are expressed as the mean fluorescence intensity of representative experiments. The relative mean fluorescence intensity values varied by <10% between individual replicate experiments.

The T cell hybridoma growth inhibition was performed essentially as previously described (30). Briefly, H-2D^d-transfected L cells (SKT 4.5) were used as APC to stimulate the T cell hybridoma B4.2.3. APC (2 × 10⁴ cells) were plated overnight in microtiter wells in 10% FCS-containing DMEM at 37°C in 5% CO₂. The next morning, the medium was removed, and the adherent cells were washed three times with PBS and preincubated for a minimum of 1 h at 37°C in 5% CO₂ in serum-free DMEM. The serum-free medium was then aspirated, and the cells were incubated with the indicated concentrations of exogenous β₂m and p18-I-10 (a range of 1 × 10⁻⁷ to 1 × 10⁻¹³ M) for 2 h at 37°C. After incubation with β₂m and peptide, the cells were washed three times with PBS as described above and incubated overnight with 2 × 10⁴ B4.2.3 cells. The following morning, wells were pulsed with 10 mCi of [³H]thymidine and incubated for 4 h at 37°C. Wells were harvested and counted on an LKB β-Plate scintillation counter (LKB, Rockville, MD). Values were expressed as the percentage of thymidine incorporation relative to that observed in the absence of β₂m and peptide (100%). Calculations of ED₅₀ values were performed from the raw data by curve fitting using the Sigmoid logistic ℱ(x) = (a − d)/(1 + (x/c)^b + d), where a is the minimal plateau value, b is the slope factor, c is the value that results in 50% maximal response, and d is the upper plateau value. Curve fitting was implemented by fixing values a and d based on the raw data, and iterating the logistic 20 to 50 times until no change in b, c, χ², or correlation coefficient values were observed. All correlation coefficients were 0.95 or greater.

We have previously developed an in vitro translation and assembly system to study the generation of murine MHC I molecules (29). Using this system, conditions necessary for stable association of murine β₂m with the heavy chain H-2L^d were established. Initial studies used this model system to explore the nature of the interactions between murine heavy chain and hβ₂m.

Immunoprecipitation of in vitro translated H-2L^d with the mAb 28-14-8 is indicative of a conformed α3 domain and proper intrachain disulfide bond formation (Fig. 1), which is a prerequisite for assembly of in vitro translated MHC I complexes (29). The mAb 30-5-7 recognizes a native α2 domain epitope on H-2L^d that typically depends on presence of β₂m and peptide (29, 31). Consistent with this, 30-5-7 immunoprecipitation of cotranslated H-2L^d and mβ₂m demonstrated a requirement for the presence of an H-2L^d binding peptide for both formation of the peptide binding domain and stable association of the two chains. As has been observed previously, this Ab coprecipitated mβ₂m in a molar ratio of approximately 10:1 (29, 32, 33). In contrast, cotranslation of H-2L^d with hβ₂m resulted in both peptide-independent association of the two chains and generation of a native α2 domain with a molar ratio of heavy chain to hβ₂m of approximately 2:1 in both the presence and the absence of peptide, consistent with a higher affinity interaction between the chains. Calculation of the molar ratio takes into account the fact that mβ₂m has five methionine residues that can be radiolabeled while hβ₂m has only one methionine residue, hence the relatively lighter bands in Figure 1, lanes 3, 4, 7, and 8.

The interaction of mβ₂m with the murine H-2L^d heavy chain is weaker than that with most other murine class I MHC heavy chains, resulting in relatively low total cell surface expression that can be increased by the addition of either hβ₂m or an H-2L^d binding peptide (29, 34). Based on the results shown in Figure 1, one possible mechanism by which hβ₂m exerts this effect is by inducing a native conformation in the peptide binding domain of cell surface molecules, which is generally believed to be critical for cell surface stability (35, 36).

To further explore these interactions, the H-2L^d-transfected L cell T1.1.1 was incubated with graded concentrations of human and murine β₂m in serum-free medium for 2 h, and cell surface expression was determined by flow cytometry (Fig. 2). As generation of the membrane-proximal α3 domain epitope occurs in the endoplasmic reticulum and is independent of association of the heavy chain with either peptide or β₂m, quantitation of this domain reflects the total amount of these molecules on the cell surface regardless of the state of their peptide binding α2 domain (29). Addition of exogenous mβ₂m had a small, but reproducible, effect on both total levels of H-2L^d on the cell surface (as determined by presence of the α3 epitope) and on the α2 domain epitope. hβ₂m had a more pronounced effect on the level of total cell surface H-2L^d compared with effects observed with mβ₂m and an even more dramatic effect on the folding of the α2 domain epitope, again consistent with its higher affinity association with murine heavy chains.

Any effects of exogenous β₂m on cell surface MHC I levels must be observed over the background of endogenous peptide-loaded natively folded molecules on the surface of normal cells. To eliminate this background, we used cell lines with a defect in the peptide transport system (TAP). The murine system allows us to take advantage of a unique property of murine MHC I expression. Unlike human MHC I, which are retained in the endoplasmic reticulum in TAP-negative cells, murine class I molecules accumulate on the cell surface when incubated at reduced temperature or in the presence of peptide and β₂m (37).

H-2L^d-transfected E-3 cells (LdE), which have a regulatory defect in the synthesis of one of the TAP components (13, 25, 38), were treated with increasing concentrations of murine or human β₂m, and the effects on both total (α3 domain) and α2 domain conformed molecules were determined. hβ₂m was more effective than mβ₂m in the generation of both total and α2 conformed MHC I molecules (Fig. 3). Moreover, its effects on the α2 domain epitope were again more pronounced than those on the α3 domain epitope. This suggests that hβ₂m has a qualitatively different effect with regard to the conformation of the α2 domain than does mβ₂m, and this is apparently independent of peptide. This is consistent with the observations with the in vitro translation assay (Fig. 1) and is indicative of hβ₂m’s ability to natively fold this domain in a peptide-independent fashion.

To identify the regions of hβ₂m responsible for the observed effects, we engineered two chimeric human:murine β₂m cDNAs, taking advantage of a conserved restriction site in the cDNAs of both human and murine β₂m. The abilities of these chimeric proteins to assemble with H-2L^d were evaluated by in vitro translation and assembly, focusing on the peptide-independent ability of hβ₂m to generate a native α2 domain. The chimeric β₂m containing the first two-thirds of hβ₂m (HHM) conformed the α2 domain to a similar extent as that observed with native hβ₂m (Fig. 4). Chimeric β₂m consisting of the first two-thirds murine and the last third human (MMH) was indistinguishable from murine β₂m in its effects on α2 domain conformation. Parallel experiments were performed with the murine heavy chain H-2D^d, and similar results were obtained (data not shown).

To more precisely determine the regions of hβ₂m responsible for these effects, additional chimeric cDNAs (Fig. 5) were expressed as recombinant proteins and evaluated for their effects on cell surface MHC I expression in TAP-defective cell lines. When LdE cells were incubated for 2 h at 37°C in the absence of β₂m (serum-free medium), H-2L^d expression decreased to approximately 50% of the initial levels in cells grown in 10% FBS (Table I, first two columns). This presumably reflects the inability of the TAP-defective cells to provide peptide-loaded molecules to offset the loss of unstable cell surface molecules that had accumulated in the presence of the bovine β₂m present in FBS. Addition of mβ₂m, HMM (whose first 34 residues are identical with those of hβ₂m), or MHM (which contains only the middle residues (35–69) of hβ₂m) prevented much of the loss of total H-2L^d from the cell surface over 2 h, while addition of hβ₂m resulted in a net increase in total cell surface H-2L^d. Evaluation of the effects of exogenous β₂m treatment on the α2 domain reveals a similar progressive rescue of H-2L^d. HHM was also more effective than either HMM or MHM, suggesting that there were contributions from both regions 1 to 34 and 35 to 69 in stabilizing cell surface H-2L^d. Addition of peptide resulted in dramatic increases in both the α2 conformed and total cell surface H-2L^d, but only in the presence of β₂m, demonstrating that not only can these molecules be preserved by addition of exogenous β₂m, but they also can be loaded (data not shown).

A number of laboratories have observed that the heavy chains of H-2L^d associate with β₂m with lower affinity than most other murine class I molecules, and the substitution of glutamic acid (E) for valine (V) at position 9 of H-2L^d has been implicated as contributing to this phenotype (13, 32). To determine whether the effects of exogenous β₂m on cell surface H-2L^d were in part due to H-2L^d’s lower affinity for β₂m, we tested chimeric β₂m preparations on an E-3 cell line transfected with H-2L^d containing the E9V substitution (LdE9V; Table I). Compared with the 50% loss observed with H-2L^d, there was a 35% loss of total H-2LdE9V from the cell surface during a 2-h incubation in serum-free medium consistent with its higher affinity for β₂m. Addition of either HHM or hβ₂m completely reversed this effect and resulted in an increase in total cell surface H-2L^d of about 35%, while mβ₂m and chimeric β₂m, with only the first or middle regions containing human sequence, had intermediate effects. These effects qualitatively parallel those on the α2 domain of H-2L^dE9V, suggesting a tight linkage between the integrity of this domain and stability on the cell surface. Interestingly, the relative effects of hβ₂m on the α2 domain of LdE9V (2-fold increase) were significantly greater than those on the α3 domain (1.4-fold increase). This result is consistent with a greater number of α3 conformed, α2 nonconformed H-2LdE9V molecules on the cell surface than observed with H-2L^d before any β₂m treatment and presumably reflects the higher affinity interaction between the heavy chain and the endogenous β₂m as a consequence of the E9V mutation.

These results suggest that the effects of exogenous hβ₂m on cell surface stability and MHC folding are not unique to low affinity heavy chains such as H-2L^d. To establish that this is a more generalizable phenomenon, we tested the effects of the β₂m chimeras on H-2D^d, a naturally occurring murine MHC I molecule that exhibits higher affinity interactions with β₂m than does H-2L^d (32). LKD8 cells (H-2D^d-transfected E-3 cells) (12) were incubated with chimeric β₂m, total cell surface H-2D^d was determined by flow cytometry with the α3-specific mAb 34-2-12, and the levels of conformed α2 domain epitope were determined with the mAb 34-5-8 (Table I). In contrast to H-2L^d, in the absence of any source of exogenous β₂m, there was no appreciable drop in total cell surface H-2D^d over 2 h, consistent with H-2D^d’s higher affinity interaction with β₂m. However, there was little or no α2 domain epitope detectable under these conditions, consistent with these molecules having an unfolded peptide binding domain due to the absence of a functional peptide transporter (TAP). As the half-life of these molecules on LKD8 cells has previously been demonstrated to be only about 10 min (12, 38), their stability over time in serum-free medium probably reflects a steady state between molecules being removed from the surface due to β₂m dissociation and newly emerging molecules with endogenously bound β₂m.

The above data are consistent with exogenous β₂m stabilizing empty class I molecules on the cell surface in a state receptive to loading with exogenous peptide and support the hypothesis that hβ₂m folds the α2 domain even in the absence of peptide. To determine whether the relative effects of the different chimeric β₂m on cell surface class I stability and folding correlate with T cell recognition, we treated the H-2D^d-transfected L cell SKT 4.5 with an H-2D^d binding peptide in the presence of different chimeric forms of β₂m. These cells were then used to stimulate the H-2D^d-restricted, peptide-specific hybridoma B4.2.3, using growth inhibition as a measure of activation (Fig. 6). In the absence of any source of β₂m (SF), the ED₅₀ was approximately 3 × 10⁻⁹ M peptide. The addition of mβ₂m shifted the sensitivity of the response by about four- to fivefold (ED₅₀ = 7.2 × 10⁻¹⁰ M). Consistent with previous results, HMM and MHM resulted in similar increases in the efficiency of peptide loading (ED₅₀ = 8.3 and 9.4 × 10⁻¹⁰ M, respectively). However, the addition of either HHM or hβ₂m further improved the effectiveness of peptide loading another three- to fourfold (ED₅₀ = 1.9 and 3.2 × 10¹⁰ M, respectively), consistent with their ability to stabilize peptide-receptive molecules on the cell surface and make them available for peptide loading.

Our results suggest that the NH₂-terminal two-thirds of hβ₂m contribute to its increased affinity for heavy chain compared with mβ₂m. Alignment analysis of crystal structures for human and murine class I molecules reveal that the S4 strand of hβ₂m (residues 50–56) contains a β bulge, while the same region of mβ₂m is a continuous β strand (22, 39). Superimposing hβ₂m and mβ₂m α carbons reveals that even in this region there is very close alignment, with the notable exception of a single residue at position 53 (D53). The side chain of this residue in hβ₂m lies directly between two conserved arginine residues at positions 35 and 48 in the floor of the heavy chain, with the δ oxygen of D53 coming within 2.9 angstroms of the amino group of R35 and within 4 angstroms of the amino side chain of R48 of the heavy chain (Fig. 7,a), close enough to form ionic bonds. In contrast, the crystal structures of both H-2K^b (Fig. 7,b) and H-2D^b (data not shown) reveal that the side chains of D53 lie parallel to the floor of the heavy chain and come only within 4.7 angstroms of the amino group of R35 (H-2K^b) and >6.5 angstroms from R48. To directly assess whether this residue contributes significantly to the interaction between heavy chain and β₂m, we mutated it to either glutamine (D53N) to prevent the formation of ionic bonds or to valine (D53V) to also prevent the formation of hydrogen bonds with other side chains and determined their effects on H-2D^d expression in LKD8 cells (Fig. 8, a and b). As shown in previous figures, hβ₂m was much more effective in increasing the surface expression of natively folded H-2D^d than was mβ₂m. Significantly, the single mutation of aspartate 53 of hβ₂m to either asparagine or valine resulted in decreased levels of both total and α2 conformed H-2D^d, presumably due to its inability to form ionic bonds with residues in the heavy chain.

The arginine residues at positions 35 and 48 are conserved among most human and murine class I heavy chains. One exception is H-2L^d, which has an arginine at position 35 but has a glutamine at position 48 (40). To determine whether this would result in a less dramatic effect of the D53N or D53V mutations on cell surface H-2L^d than was observed with H-2D^d, we next incubated LdE cells with these mutated forms of hβ₂m. As shown in Figure 8, c and d, when LdE cells were incubated with recombinant D53N, stabilization of both total and α2 conformed cell surface H-2L^d was as good as or better than that observed with hβ₂m. When these cells were incubated with the D53V form of hβ₂m, whose side chain at position 53 can form neither ionic nor hydrogen bonds with neighboring residues in the heavy chain, a slight decrease in its ability to induce both epitopes was observed compared with that of hβ₂m and D53N. These data establish two important points. First, the D53N mutation did not negatively effect MHC I expression nonspecifically by effecting the folding or secondary structure of the protein, as this mutant worked as well as hβ₂m in stabilizing cell surface H-2L^d, and second, careful analysis of available crystal structures of class I MHC molecules can facilitate the engineering of β₂m variants that predictably affect MHC I surface expression.

A number of “humanized” murine β₂m point mutants have been made by mutating amino acid residues in and around the S4 strand to those found in the human sequence. The only amino acid differences between murine and human β₂m sequences in the S4 strand itself (where D53 resides) are methionines 51 and 54 in the murine sequence, which are histidine and leucine, respectively, in hβ₂m. Therefore, the M51H and M54L double point mutant was made to determine the effects these residues would have on cell surface MHC molecules. Additionally, the two proline residues at positions 33 and 47 in murine β₂m were individually changed to serine and aspartic acid found in hβ₂m (P33S and P47E). When each of these three mutants was assessed for its ability to stabilize cell surface H-2D^d and H-2L^d, no difference was seen relative to wild-type mβ₂m (data not shown).

The effects of the position 53 mutation suggested that a single ionic bond may contribute measurably to the stability of heavy chain-β₂m interactions. Upon examination of the crystal structure of HLA-A2, a lysine residue at position 58 of β₂m (K58) was identified that comes in close proximity to a conserved arginine at position 6 of the heavy chain (R6). Therefore, site-directed mutagenesis was used to change the lysine at position 58 of hβ₂m to the negatively charged glutamic acid (K58E) to promote the formation of an ionic bond with R6 of the heavy chain. Since R6 is conserved across not only human but also murine class I molecules, we took advantage of our in vitro cell surface stabilization assay and well-defined mAbs to murine MHC molecules to examine the effect of recombinant K58E on LKD8 cells. Approximately twice the amount of hβ₂m as K58E was required to provide the same level of cell surface H-2D^d stabilization (Fig. 9,a). The enhanced stabilizing ability of K58E was also demonstrable in β₂m excess (plateau beyond 2.5 μM), suggesting that this effect is qualitatively different from that observed with wild-type hβ₂m and consistent with a higher affinity interaction. The effects on the α2 domain were less dramatic, but consistent with the α3 domain results (Fig. 9,b). In contrast to H-2D^d, cell surface stabilization of H-2L^d on LdE cells by K58E was essentially indistinguishable from that by hβ₂m (Fig. 9, c and d). The recently reported crystal structure of H-2L^d (41) reveals that the interatomic distance between the terminal amino groups of the K58 residue of mβ₂m and the R6 of the heavy chain is greater than 11 angstroms, which would make formation of a salt bridge very unlikely. Again these data demonstrate the ability to use structural data to facilitate the engineering of β₂m variants with predictable changes in their functional abilities to stabilize cell surface MHC I molecules.

While both mβ₂m and hβ₂m assemble with and promote cell surface loading of class I molecules, hβ₂m is substantially more effective at preserving cell surface class I expression and facilitating peptide loading (Figs. 3, 6, and 8). This property correlates with its ability to induce α2 domain folding even in the absence of peptide, a characteristic not observed with mβ₂m (Fig. 1). The effects of xenogeneic β₂m on murine class I heavy chain conformation have been reported previously (42, 43, 44, 45), but the work herein correlates the conformational changes induced on the peptide binding domain by hβ₂m and their effects on the stability of empty molecules.

If the ability to conform the peptide binding domain is a general feature of a higher affinity β₂m, it may have important implications for the generation of peptide-based vaccines. For example, it becomes important to determine the T cell response to empty heavy chains that have been stabilized by higher affinity β₂m. These α2 conformed molecules may be recognized as foreign, and therefore determining any effects of these molecules on subsequent immune responses becomes critically important. Recently, Orihuela and colleagues partly addressed this issue in their study of NK cell responses, which are directly affected by MHC I expression. They demonstrated that cell surface H-2D^d molecules on LKD8 cells, stabilized with exogenous hβ₂m in the absence of peptide, did not confer resistance to Ly-49A⁺ NK cells, while addition of an H-2D^d-restricted peptide did confer resistance (46). Although these experiments were only performed using hβ₂m at relatively low concentrations (17 μg/ml, or 1.4 μM), this would suggest that Ly-49⁺ NK cells can distinguish between natively folded empty MHC I molecules and peptide-loaded MHC I molecules. Whether this would also be the case with higher concentrations of exogenous hβ₂m remains to be determined. Study of T cell responses to these empty molecules may also provide insight into mechanisms for alloantigen recognition, specifically whether alloreactive T cells require the presence of peptide in the binding groove of class I MHC molecules to become stimulated.

Studies with chimeric murine:human β₂m along with comparative structural analyses focused attention on the S4 strand (residues 50–56). As predicted by these analyses, mutation of aspartic acid at position 53 of hβ₂m to either asparagine or valine significantly affects its ability to stabilize cell surface H-2D^d molecules (Fig. 8, a andb). This is presumably due to the inability of the mutated residue to form ionic bonds with arginines at positions 35 (R35) and 48 (R48) of the heavy chain and also, in the case of D53V, an inability of the side chain of residue 53 to form hydrogen bonds. Although both hβ₂m and mβ₂m have aspartic acid at this position, residue differences in the neighboring regions may affect the relative orientations of D53.

The fact that the D53N and D53V variants of hβ₂m are still more effective than mβ₂m suggests, not surprisingly, that other interactions also play a role. In this regard, the observations with H-2L^d are particularly significant. hβ₂m is clearly much more effective than mβ₂m in increasing total and α2 conformed cell surface H-2L^d. However, unlike H-2D^d, H-2L^d contains glutamine at position 48, making this residue an unsuitable partner in forming a ionic bond with D53 (26), but a good candidate for forming hydrogen bonds. This may explain why the D53 mutations had much less of an effect on cell surface H-2L^d and why the D53N mutant had slightly better binding. Despite the glutamine at position 48 of H-2L^d, hβ₂m is still more effective than mβ₂m in stabilizing H-2L^d, presumably as a consequence of interactions of other residues in the β₂m:heavy chain interface.

The mutation of a single residue of hβ₂m (K58E) to promote the formation of a ionic bond improves its ability to fold and stabilize H-2D^d molecules on LKD8 cells (Fig. 9, a and b). This difference is seen throughout the titration range, most notably after saturation with β₂m, indicating the qualitatively different binding of this mutant. Whether the number or the quality of peptides capable of being stably loaded onto H-2D^d molecules in the presence of K58E is sufficient to influence T cell recognition and stimulation remains to be determined and is currently under examination. The stabilizing effect is not demonstrated with H-2L^d (Fig. 9, c and d), most likely reflecting the effects of heavy chain polymorphisms. However, since there are no crystal structures of hβ₂m bound to murine heavy chains, it is difficult to determine the structural basis for the difference between H-2D^d and H-2L^d. Considering the above data, another approach for creating higher affinity variants of β₂m could be directed at residues interacting with the α3 domain of the heavy chain, as it is the most conserved of the domains interacting with β₂m. Clearly, it will be important to determine the effects of the K58E and other mutations on human MHC I expression and the ability to facilitate peptide loading on human cells.

Until recently, attempts to generate CTL responses in vivo by immunization with peptides have required either traditional adjuvants such as CFA or IFA conjugation of peptides to a variety of lipids or other carriers, or prepulsing of the peptide Ags onto cultured dendritic cells in vitro (47, 48, 49). However, in a study by Rock and colleagues, in vivo priming of mice with peptides using only hβ₂m as an adjuvant was convincingly demonstrated (50). The results presented here suggest an underlying mechanism for these observations. Generally, in vivo peptide priming of primary CTL is conducted in the presence of adjuvants that generate local inflammatory reactions that are often associated with increased serum levels of β₂m (51, 52). If the loss of class I molecules from the cell surface is the result of the ordered process of 1) peptide dissociation, 2) β₂m dissociation, and 3) internalization of free heavy chain, then elevated levels of β₂m in inflammatory exudate could stabilize empty or peptide-receptive class I molecules on the cell surface long enough for peptide loading to occur. The addition of hβ₂m to the peptide inoculum provides a means to stabilize these molecules long enough for peptide binding to occur by virtue of its higher affinity interaction with murine MHC heavy chains and thereby generate loaded molecules capable of stimulating CD8⁺ T cells. Additionally, hβ₂m may have contributed to the in vivo priming by supplying a helper epitope in the form of processed hβ₂m that was represented by MHC class II molecules to CD4⁺ T cells.

Based on this model, one would predict that the use of syngeneic mβ₂m would not be as effective in promoting in vivo peptide priming in their system unless supplied with a source of T cell help. Considering the data herein, in vivo priming probably would have required significantly higher concentrations of mβ₂m and may not have worked at all. The 30% sequence disparity between hβ₂m and mβ₂m becomes an important consideration in their studies in particular and in the generation of practical vaccines in general. For general vaccine considerations, the likely necessity for multiple rounds of immunization or subsequent immunization with different agents seems to be incompatible with the probable antigenicity of a protein adjuvant with 30% difference in amino acid sequence than the host’s own β₂m. Consequently, engineering individual mutations that promote interchain stability, e.g., via the formation of ionic bonds, may lead to the generation of high affinity murine β₂m variants with minimal antigenicity for use in animal vaccination models. Hence, these and subsequent studies are the initial steps to provide a paradigm for the engineering of higher affinity hβ₂m variants for use in peptide-based vaccines in humans.

We thank Drs. Jonathan Ashwell, Allan Weissman, Jonathan Yewdell, Kelly Kearse, and David Margulies for their critical review of this manuscript.