The Assembly and Stability of MHC Class II-(αβ)₂ Superdimers¹

CD4⁺ Th cells recognize peptide fragments of Ags bound to class II molecules presented on the surface of APCs (reviewed in 1 . MHC class II molecules consist of a heterodimer of noncovalently associated polymorphic α and β subunits, each type I integral membrane glycoproteins, with molecular mass of approximately 34 and 29 kDa, respectively (2, 3, 4). Upon synthesis in the endoplasmic reticulum, the α- and β-chains immediately associate and assemble noncovalently with a third protein, the invariant chain (Ii),⁴ forming a nine-chain (αβIi)₃ complex (5, 6). The Ii serves both to prevent peptides from binding to the class II molecules en route to the appropriate subcellular site for peptide binding (7, 8, 9), and to direct the class II molecules to that site (10, 11, 12). Ii is proteolytically removed from the class II molecules in the peptide-loading compartments (13, 14, 15), allowing the class II molecules to bind antigenic peptides (16, 17, 18). Peptide binding completes the protein-folding process and maturation of class II molecules (19, 20). Upon antigenic peptide binding, the association between the α- and β-chains becomes remarkably strong, withstanding the denaturing effects of SDS-PAGE at room temperature in the absence of reducing agents (21, 22, 23).

Studies over the last several years have yielded a detailed picture of the structure of class II molecules (reviewed in 24 . The three-dimensional structures of the human class II molecules, HLA-DR1 and HLA-DR3 (25, 26, 27, 28, 29, 30), and the mouse class II molecule I-E^k (31) have been determined by x-ray crystallography. The binding site for CD4, the Th cell coreceptor for class II molecules, has been mapped to a loop in the β₂ domain (32, 33, 34) and possibly the α₂ domain (25, 35). One striking feature of the class II crystal structures is that the αβ heterodimer dimerized with itself to form an (αβ)₂ complex. The fact that these dimers of αβ heterodimers, or superdimers, appeared in several different types of DR crystals, as well as in I-E^k crystals, suggests that the tendency to superdimerize may represent an intrinsic property of MHC class II molecules (reviewed in 36 . It has been postulated that the superdimerization of class II molecules may serve to both facilitate the activation of T cells by cross-linking TCRs and signaling the APC to express essential costimulatory molecules for T cell activation (25). These authors envisioned a synergistic process in which a pair of TCRs interacts with a pair of class II molecules, resulting in stable, dimerized class II-TCR complexes. Indeed, the x-ray crystal structures of the β-chain (37) and the Vα domain (38) of the TCR have been solved, and the Vα domain shows a tendency to form pairs of homodimers potentially able to interact with two class II molecules (38); however, homodimers were absent in the heterodimeric TCR-αβ structures (39, 40). Recently, Reich et al. (41) showed that TCR/peptide-MHC class II complexes oligomerized in solution in a ligand-dependent fashion to form supramolecular structures containing two to six TCR/peptide/MHC class II complexes. These results provided direct evidence for models of T cell signaling based on specific multimerization of TCR/peptide/MHC complexes. Others have hypothesized that CD4 oligomerization may promote MHC class II and TCR oligomerization (42). The model of CD4-mediated oligomerization is attractive because it does not depend upon two MHC class II molecules containing identical or closely related antigenic peptides required for TCR dimerization.

Our previous studies provided evidence for the existence of both a 60-kDa αβ heterodimer and a 120-kDa (αβ)₂ superdimer of the mouse I-E^k molecules, in detergent lysates of B cells (43). A mAb with demonstrated specificity for the (αβ)₂ superdimer inhibited a low affinity, but not a high affinity T cell response, suggesting that the superdimers exist on the B cell surface and that they may play a role in presenting low affinity Ags (43). Subsequently, Roucard et al. (44) provided biochemical evidence that HLA-DR existed as a superdimer in detergent solutions. Very recently, Cherry et al. (45), using single-particle fluorescence imaging, showed that HLA-DR exists as both dimers and superdimers on the surfaces of human fibroblasts transfected with the genes encoding HLA-DR α and β and Ii, and estimated that 25% of class II molecules were in the superdimer form. In this study, using chemical cross-linking, we demonstrate that the mouse I-E^k superdimer exists on the surface of B cells in the absence of detergent. We also provide evidence that both αβ and (αβ)₂ are derived from a single population of I-E^k molecules in B cell lysates and that purified I-E^k in detergent solution assumes both αβ dimers and (αβ)₂ superdimers. The class II superdimers are present in significant amounts, are free of components other than α- and β-chains, and have the same general properties of SDS stability as αβ complexes.

The mouse mAbs 17-3-3S (46) and Y-17 (47), which recognize different but overlapping determinants on the I-E^kβ-chain (48, 49, 50, 51), were obtained from hybridomas provided by American Type Culture Collection (ATCC, Manassas, VA). 17-3-3S is of the IgG2a,κ isotype, and Y-17 is of the IgG2b,κ isotype. The mAbs were purified by protein A-Sepharose chromatography. The Ii-specific rat mAb IN-1 (isotype IgG2b,κ) was kindly provided by Dr. N. Koch (University of Bonn, Bonn, Germany). The isotype control mAb IgG2a,κ was purchased from PharMingen (San Diego, CA), and rat IgG was purchased from Jackson ImmunoResearch (West Grove, PA). The horseradish peroxidase (HRP)-conjugated secondary Abs, rabbit Abs specific for mouse IgG2a, and goat Abs specific for rat IgG heavy and light chains were obtained from Zymed Laboratories (South San Francisco, CA).

The B cell lymphoma CH27 (52) was characterized and kindly provided by Dr. G. Haughton (University of North Carolina, Chapel Hill, NC). The mouse T cell hybridoma TPc9.1 generated in this laboratory is specific for cytochrome c presented by I-E^k-expressing APC, and secretes IL-2 upon activation (53, 54). The CTLL-2 cell line (55), obtained from ATCC, is an IL-2-dependent cell line. All cells were grown in complete medium (CM) (56) containing 15% FCS (15% CM). The CTLL-2 cells were maintained in 15% CM containing 10% T-Stim (Collaborative Biomedical Products, Bedford, MA) as an IL-2 source.

CH27 cells (2.5 × 10⁶ cells/ml) were incubated in Met⁻ 5% CM for 30 min, labeled with [³⁵S]Met,Cys (NEN, Wilmington, DE) at 150 μCi/ml for 30 min, and washed in 15% CM. The cells were cultured in 15% CM at 3 × 10⁵ cells/ml for 4 h. At the conclusion of the chase, cells were washed twice in cold PBS and lysed at 5 × 10⁶ cells/ml in ice-cold Nonidet P-40 lysis buffer (50 mM Tris-HCl, pH 7.6, 150 mM NaCl, 5 mM EDTA, and 0.5% Nonidet P-40) containing 200 μg/ml PMSF, 25 μg/ml aprotinin, and 10 mM iodoacetamide. The lysate was divided into two equal portions of 5 × 10⁶ cell equivalents each, centrifuged at 16,000 × g for 30 min at 4°C, and precleared twice with 200 μl protein A-Sepharose (Pharmacia, Piscataway, NJ) slurry. I-E^k was immunoprecipitated with mAbs 17-3-3S or Y-17 and 60 μl protein A-Sepharose slurry at 4°C. Supernatants were precipitated a second and third time with 30 μg mAb and 60 μl protein A-Sepharose slurry. Beads were washed three times with ice-cold Nonidet P-40 wash buffer (50 mM Tris-HCl, pH 7.6, 150 mM NaCl, 5 mM EDTA, and 0.2% Nonidet P-40). Bound proteins were eluted with 80 μl of nonreducing SDS-loading mixture (250 mM Tris-HCl, pH 6.8, 50% (v/v) glycerol, 5% (w/v)). Eluates were subjected to 10% SDS-PAGE and fluorography using Fluoro-Hance (Research Products International, Mt. Prospect, IL).

CH27 cells (1.6 × 10⁶ cells/ml) were labeled with [³⁵S]Met,Cys (150 μCi/ml) for 14 h at 37°C. The cells were washed with ice-cold PBS at pH 8.5 (PBS 8.5). The cells were resuspended in PBS 8.5 at a concentration of 25 × 10⁶ cells/ml, and aliquots (200 μl) were incubated with 2 μl of a 200 mM solution of a cross-linking reagent in DMSO or 2 μl of DMSO alone for 2 h at 4°C. The cross-linking reagents tested included: dithiobis[succinimidylpropionate] (DSP); 1,5-difluoro-2,4-dinitrobenzene (DFDNB); 3,3′-dithiobis[sulfosuccinimidylpropionate] (DTSSP); ethyleneglycolbis[succinimidylsuccinate] (EGS); bis[sulfosuccinimidyl]suberate (BS³); dimethyladipimidate.2HCl (DMA); dimethylpimelimidate.2HCl (DMP); disulfosuccinimidyltartarate (sulfo-DST); and dimethyl 3.3′-dithiobispropionimidate.2HCl (DTBP) (Pierce, Rockford IL). The cross-linking was quenched by adding 20 μl of 1 M glycine and incubating at 4°C for 5 min. The cells were pelleted and lysed in 1 ml Triton X-100 lysis buffer (50 mM Tris-HCl, pH 7.6, 150 mM NaCl, 5 mM EDTA, 0.02% sodium azide, and 1% Triton X-100) containing 200 μg/ml PMSF. I-E^k was immunoprecipitated with mAbs Y-17 or 17-3-3S, as described above. The bound I-E^k was eluted from the protein A-Sepharose with 40 μl of SDS-loading mixture alone or in the presence of either β-mercaptoethanol (2% v/v) or iodoacetamide (10 mM). The samples were centrifuged to pellet the protein A-Sepharose, the eluates were removed and either boiled for 5 min or not, and subjected to 10% SDS-PAGE. Where indicated, bands were excised from the gel, rehydrated as described below, and subjected to 8% SDS-PAGE under reducing conditions.

CH27 cells were grown, lysed, and I-E^k purified, as described (54), with the following modifications. The mAbs 17-3-3S and Y-17 were coupled to CNBr-activated Sepharose 4B (Pharmacia Biotech, Uppsala, Sweden) and used for affinity chromatography of I-E^k. I-E^k was eluted using β-octylglucoside (β-OG) elution buffer (15 mM triethanolamine, pH 11.5, 140 mM NaCl, and 30 mM β-OG), neutralized with a measured amount of 1 M 2-(N-morpholino)ethanesulfonic acid, pH 5, dialyzed against PBS 30 mM β-OG, concentrated in a centrifuge microconcentrator (Centricon 10; Amicon, Beverly, MA), and centrifuged through a 0.2-μm filter (Schleicher & Schuell, Keene, NH). Protein concentration was determined by bicinchoninic assay (Micro BCA Assay; Pierce, Rockford, IL).

I-E^k purified by immunoaffinity chromatography on a Y-17 mAb column was divided into aliquots of 8 μg each, diluted 1/1 in nonreducing SDS-loading mixture, and subjected to 10% SDS-PAGE without boiling. Proteins were transferred to 0.2-μm-pore nitrocellulose membrane (Schleicher & Schuell) at 0.5 amps for 2 h. The membrane was Ponceau S stained and cut into strips. The strips were soaked in 150 mM NaCl, 10 mM sodium phosphate, pH 7.4 (PBS), containing 5% (w/v) nonfat dry milk and 0.01% NaN₃, washed in PBS containing 0.1% (v/v) Tween-20 (PBS Tween), and soaked 2 h in a 3 μg/ml solution of one of the following primary mAbs: 17-3-3S, mouse IgG2a,κ, IN-1, or rat IgG at 3 μg/ml. These primary mAbs were diluted in PBS Tween containing 1% (w/v) BSA. The strips were washed four times in PBS Tween and soaked in a 1/10,000 dilution of either HRP-rabbit anti-mouse IgG2a or HRP-goat anti-rat IgG for 30 min. The strips were washed four times in PBS Tween, and developed using enhanced chemoluminescence (ECL; Amersham, Arlington Heights, IL).

Gels were fixed in 30% ethanol, 10% acetic acid, then soaked in 1 mg/ml sodium thiosulfate, 30% ethanol, and 10 mM sodium acetate, pH 6, for 30 min. The gels were then washed thoroughly in distilled water, and soaked in 0.1% silver nitrate, 0.01% formaldehyde for 30 min. The gel was developed in 2.5% sodium carbonate, 0.02% formaldehyde, and the development was arrested by glacial acetic acid to yield a final concentration of 5% acetic acid. Gels were soaked in glycerin and stored in plastic.

I-E^k complexes were analyzed after electrophoresis by excising bands and electroeluting the proteins from the polyacrylamide for 100 V-h into 300 μl of 4× SDS-PAGE electrode buffer using a SixPac GE200 Gel Eluter (Hoefer Scientific Instruments, San Francisco, CA). This volume was concentrated in a Centricon 10 to a final volume of 50 μl, diluted 1/1 with SDS-loading mixture containing 4% (v/v) β-mercaptoethanol, boiled 10 min, subjected to SDS 10%-PAGE, and silver stained.

Peptide mapping of I-E^k complexes was performed by the method of Cleveland et al. (57), as modified by Lamb and Choppin (58). [³⁵S]Met, Cys-labeled I-E^k was synthesized by growing CH27 cells for 21 h at 1 × 10⁶ cells/ml in Met⁻Cys⁻ 5% CM containing 45 μCi/ml [³⁵S]Met,Cys. The cells were washed and lysed, and the I-E^k was purified, as described above. A 10-μg aliquot of I-E^k was electrophoresed without reducing or boiling. The gel was dried and exposed to film, and the 60- and 120-kDa I-E^k bands were excised. The gel pieces were rehydrated by soaking in sample buffer (SB, 0.125 M Tris-HCl, pH 6.8, 0.1% SDS, and 1 mM EDTA) containing 0.5% DTT for 45 min, and inserted into the wells of a 12% SDS polyacrylamide gel and overlaid with 20 μl SB containing 20% glycerol/bromphenol blue and 0.5% DTT. The gel pieces were further overlaid with 20 μl of 200 μg/ml Staphylococcus aureus V8 protease (Promega, Madison, WI) in SB containing 10% glycerol, and the gel was electrophoresed. When the dye front passed the stacking gel/separating gel interface, the current was turned off for 30 min to allow the protease to digest the I-E^kin situ. The gel was then run for the remainder of the electrophoresis as usual and subjected to fluorography.

For presentation of purified I-E^k, I-E^k, purified by immunoaffinity chromatography on either 17-3-3S or Y-17 mAb columns, was serially diluted in PBS across a 96-well tissue culture plate starting at 1 μg/well, in a final volume of 0.1 ml. The plate was incubated for 4 h at 25°C, then washed three times with 0.15 ml citrate buffer (100 mM sodium citrate, pH 4.5). A peptide corresponding to the C terminus of tobacco hornworm moth cytochrome c residues 82–103 (THMc 82–103) was synthesized, as detailed previously (59). THMc (82–103) was added to each well at a concentration of 1 μM in citrate buffer, and incubated at 25°C for 5 h. The wells were then washed three times with 0.15 ml 5% CM. For presentation by APCs, CH27 cells were cultured at 1.3 × 10⁶ cells/ml with 2 μM THMc (82–103) for 2 h at 37°C, in a 5% CO₂ atmosphere. The CH27 cells were fixed by incubation in 0.1% glutaraldehyde at 1 to 5 × 10⁶ cells/ml in DMEM at 4°C for 30 s. The cell suspension was diluted twofold with 0.2 M lysine in DMEM and washed in 5% CM. The fixed CH27 cells were serially diluted across a 96-well tissue culture plate starting at 2 × 10⁵ cells/well. TPc9.1 cells were added to the wells containing either the purified I-E^k or the CH27 cells at 5 × 10⁴ cells/well in 200 μl and incubated at 37°C for 24 h. Supernatants (75 μl) were harvested and tested for IL-2 content by their ability to support the growth of the IL-2-dependent T cell line CTLL-2, as previously described (60).

For thermal stability experiments, the aliquots were diluted 1/1 with nonreducing SDS-loading mixture and incubated at 25°, 37°, 50°, or 65°C for 30 min. For pH stability experiments, the aliquots of Y-17 affinity-purified I-E^k were diluted in Tris/acetate buffers to give the final conditions of 25 mM Tris-HCl, 25 mM acetate, 1% (w/v) SDS, and 15% (v/v) glycerin, pH 3, 4, 5, 6, 7, 8, or 9. One aliquot was diluted in nonreducing SDS-loading mixture, pH 7.4, and one was diluted in β-OG elution buffer, pH 11.5. The final volume of each sample was 50 μl. The samples were incubated at these pH levels for 30 min at 37°C and neutralized with 50 μl nonreducing SDS-loading mixture. Following treatment, all samples were electrophoresed on 10% SDS polyacrylamide gels and silver stained.

We previously reported the existence of both 60-kDa dimers and 120-kDa superdimers of the mouse class II I-E^k molecules in detergent lysates of B lymphocytes by immunoprecipitation using the I-E^k-specific mAb Y-17, which differentially recognized the 120-kDa superdimeric form of I-E^k (43). We further showed by Western blotting of the B cell lysate that the existence of the 120-kDa superdimer was not dependent on the presence of the Y-17 mAb (43). To determine whether the superdimer exists on the surfaces of B cells in the absence of detergents, B cells were treated with chemical cross-linking reagents to covalently cross-link the components of the superdimer. Because it is difficult to predict the ability of a chemical cross-linking reagent to covalently link the components of an oligomeric structure, several chemical cross-linkers were tried, including: DTSSP, DSP, DFDNB, EGS, BS³, DMA, DMP, sulfo-DST, and DTPB. Treatment of B cells with three cross-linkers (DSP, DTSSP, and DFDNB) resulted in some degree of covalent cross-linking of the α- and β-chains within the 60-kDa dimer. However, only DFDNB showed covalent cross-linking of the (αβ)₂ superdimer. The results for DFDNB and DSP are shown (Fig. 1). DSP is a long (12A°) homobifunctional NHS-ester NH₂-linker that cross-links through −NH₂ amines. The presence of a disulfide bond within DSP allows for reversal of cross-linking with a reducing reagent (61). DFDNB is a small (3A°) bifunctional aryl halide that reacts with −NH₂ groups to form a nonreducible bond (62).

B cells were treated with the cross-linking reagents and subjected to immunoprecipitation using either the Y-17 mAb (Fig. 1,A) or the 17-3-3S mAb (Fig. 1,B), which differentially immunoprecipitate the 120-kDa superdimer or the 60-kDa dimer, respectively. The immunoprecipitates were subjected to SDS-PAGE under conditions in which the 60-kDa dimer and 120-kDa superdimer are stable (without reducing or boiling) or under conditions in which both the dimer and superdimer are unstable and migrate as α- and β-chains (boiling with or without reduction). Immunoprecipitation of either untreated cells or cells cross-linked with DSP or DFDNB using the Y-17 mAb yields the predicted 120-kDa (αβ)₂ superdimers when the samples are not boiled (Fig. 1,A). In contrast, the 17-3-3S immunoprecipitates contain only the 60-kDa dimers and no 120-kDa superdimers (Fig. 1,B). When the samples are boiled, but not reduced, the Y-17-immunoprecipitated class II molecules from untreated cells migrate as free α- and β-chains, and the class II molecules from DSP- and DFDNB-treated cells migrate as 60-kDa αβ dimers (Fig. 1,A), indicating that both DSP and DFDNB cross-linked the α- and β-chains within the heterodimer. DFDNB appears to be a more efficient cross-linker as compared with DSP, as the immunoprecipitates from DFDNB-treated cells have less free αβ-chains compared with the immunoprecipitates from the DSP-treated cells. The DSP- and DFDNB-cross-linked αβ dimers migrate slightly differently, suggesting that the cross-links introduced into the αβ dimers are not the same for the two different chemical cross-linkers. In addition to the cross-linked 60-kDa αβ dimer, a 240-kDa band is present in the immunoprecipitates from the DFDNB-treated cells. This 240-kDa band is the correct size to be as dimer of the superdimer or (αβ)₄. The 240-kDa band is only present in the Y-17 superdimer-specific immunoprecipitates, and not in the 17-3-3S immunoprecipitates (Fig. 1,B), indicating that the 240-kDa band is related to the superdimer. When the Y-17 mAb immunoprecipitates from DFDNB-treated cells are run under reducing conditions with boiling, the 240-kDa band disappears and the 120-kDa superdimer appears (Fig. 1 A), suggesting that the 240-kDa band is a disulfide-linked dimer of DFDNB-cross-linked (αβ)₂ superdimers. Reduction of the Y-17 immunoprecipitate also reduces the DSP chemical cross-links in the 60-kDa dimers, resulting in free α- and β-chains. By densitometry, the cross-linked (αβ)₂ superdimers in the reduced and boiled Y-17 immunoprecipitates represent approximately 18% of the total population of class II molecules.

To further characterize the 240-kDa band, the band was excised, rehydrated, and subjected to SDS-PAGE under reducing conditions. The 240-kDa band resolves into 240-, 120-, and 60-kDa bands (Fig. 1,C), indicating that the 240-kDa band is a disulfide-linked oligomer of DFDNB-cross-linked 60-kDa (αβ) dimers and 120-kDa (αβ)₂ superdimers. When the Y-17 immunoprecipitates from the DFDNB-treated cells are reduced, all of the 240-kDa resolves as 120-kDa superdimers and 60-kDa dimers (Fig. 1,A). The failure to completely reduce the excised, rehydrated 240-kDa band (Fig. 1,C) may be due to some degree of irreversible denaturation of the material in the gel. A disulfide-linked oligomer of class II molecules was unexpected because the predominant chemical cross-linking reaction of DFDNB is through amines at the pH used in this study (pH 8.5), although it is possible for DFDNB to cross-link through sulfhydryl groups. However, the only free cysteines in the I-E^k molecules are in the transmembrane region or buried in the peptide binding site, making it unlikely that the 240-kDa (αβ)₄ was the result of DFDNB sulfhydryl cross-linking. To determine whether disulfide bonds formed between DFDNB-cross-linked αβ and (αβ)₂ in the immunoprecipitates, iodoacetamide was added to the SDS SB to block free sulfhydryls. The 120-kDa, but not the 240-kDa band is present in iodoacetamide-treated Y-17 immunoprecipitates of DFDNB-treated cells (Fig. 1 A), indicating that the disulfide bonds were formed between DFDNB-cross-linked superdimers in the immunoprecipitate. In the Y-17 immunoprecipitates, which are neither boiled nor reduced, the 240-kDa band does not appear, suggesting that either the disulfide bonds are formed during boiling or that in the absence of boiling the (αβ)₄ may aggregate and not enter the gel. The fact that the 240-kDa (αβ)₄ was not present in immunoprecipitates from untreated cells or DSP-treated cells suggests that the formation of the disulfide bond was favored between DFDNB-cross-linked αβ and (αβ)₂.

Taken together, the results presented in this study indicate that treating cells with DFDNB results in covalently cross-linked 120-kDa (αβ)₂ superdimers, providing direct biochemical evidence for the existence of the 120-kDa superdimers on the B cell surface. Y-17 immunoprecipitation of DFDNB-treated cells also results in the formation of a disulfide-linked 240-kDa (αβ)₄ oligomer composed of DFDNB-cross-linked (αβ)₂ and αβ.

To determine whether class II molecules that form the 60-kDa dimer versus the 120-kDa superdimers exist as one single or two separate pools in cell lysates, exhaustive immunoprecipitations were performed. CH27 cells were metabolically labeled with [³⁵S]Met for 30 min, chased for 4 h, and lysed. The cell lysate was divided in two, lysate 1 and lysate 2 (Fig. 2). Lysate 1 was subjected to two successive immunoprecipitations using the αβ dimer-specific mAb, 17-3-3S, followed by a single immunoprecipitation using the (αβ)₂ superdimer-specific mAb, Y-17. Conversely, lysate 2 was immunoprecipitated twice using the Y-17 mAb, then once using the 17-3-3S mAb. The immunoprecipitates were subjected to SDS-PAGE without reducing or boiling, conditions under which peptide-loaded αβ dimers are stable. The 17-3-3S mAb immunoprecipitates I-E^k, which migrates as a 60-kDa αβ band. The 17-3-3S mAb depleted nearly all of the I-E^k from the lysate, leaving only a small amount to be immunoprecipitated by Y-17, which migrates as a 120-kDa (αβ)₂ band. The I-E^k immunoprecipitated by Y-17 from lysate 2 also migrates as a 120-kDa band. After two successive Y-17 immunoprecipitations, no detectable I-E^k remained to be immunoprecipitated by 17-3-3S (Fig. 2). This result indicates that the 60-kDa αβ heterodimers and 120-kDa (αβ)₂ superdimers are derived from the same pool of I-E^k molecules in the cell lysate, and that all class II αβ dimers can dimerize to form 120-kDa superdimers.

Our previous studies demonstrated the presence of 120-kDa I-E^k complexes in cell lysates both by immunoprecipitation and by immunoblotting (43). Cell lysates are complicated mixtures that potentially contain cellular components that could affect I-E^k superdimerization. To determine whether purified I-E^k has the intrinsic ability to form 120-kDa complexes in the absence of other cellular components, I-E^k was purified by Y-17 immunoaffinity chromatography and analyzed by SDS-PAGE without reducing or boiling. A portion of the gel was silver stained (Fig. 3, right panel), and the remainder was used to transfer the proteins to nitrocellulose for immunoblotting. Immunoblots were probed with the mAb 17-3-3S and its isotype control mAb mouse IgG2a,κ, or the Ii-specific mAb IN-1 and its isotype control mAb rat IgG2b,κ. The 17-3-3S immunoblots were stained with a IgG2a-specific Ab, and the IN-1 immunoblots were stained with rat IgG heavy and light chain-specific Ab and developed by ECL (Fig. 3, left and center panels). Two major bands are detected in the silver-stained gel having molecular mass of 60 and 120 kDa, and these bands stain positively for I-E^k in immunoblotting. Thus, the Y-17 immunoaffinity-purified I-E^k appears to be a mixture of both 60-kDa dimers and 120-kDa superdimers. It should be noted that in contrast to the preparations of Y-17 immunoaffinity-purified I-E^k, the Y-17-immunoprecipitated I-E^k is almost entirely in the superdimer form (see Fig. 3). This difference may be explained by the fact that during immunoaffinity purification, the I-E^k is eluted from the Y-17 column at high pH and, as will be shown below (Fig. 8), the superdimers dissociate at extreme pHs. Thus, it is likely that the mixture of dimers and superdimers in the Y-17 immunoaffinity-purified preparation represents the reassociation of I-E^k molecules at neutral pH.

In addition to the 60- and 120-kDa bands, a faint third band of 90 kDa is visible in the silver-stained gels and in the 17-3-3S immunoblot (Fig. 3). This band does not stain with the IgG2a-specific secondary Ab alone, but does stain with the heavy and light chain-specific secondary Ab. This 90-kDa band most likely represents small amounts of contaminating Y-17 mAb from the affinity column, which is of the IgG2b isotype. A faint 35-kDa band is also detected in immunoblots using the heavy and light chain-specific secondary Ab, and this is most likely the contaminating Y-17 light chain. Significantly, the 120-kDa I-E^k band does not stain with either of the secondary Abs alone, indicating that there is no Ab present in this band.

The IN-1 immunoblot of the affinity-purified I-E^k preparations shows bands at 35 to 40 kDa and 65 kDa, indicating that Ii monomers and disulfide-linked dimers (5) are present in this I-E^k preparation. The 65-kDa Ii dimer migrates slightly higher than the bulk of the 60-kDa I-E^k. Significantly, the 120-kDa band does not stain with the IN-1 mAb, indicating that it does not contain Ii. Taken together, these results provide evidence that affinity-purified I-E^k forms both 60- and 120-kDa complexes in vitro, and that these complexes do not contain Ii or Ab.

Analysis of I-E^k molecules purified using the dimer-specific 17-3-3S mAb also showed these to be a mixture of dimers and superdimers (see Fig. 7 below). Moreover, the ratio of αβ dimers to (αβ)₂ superdimers in Y-17 and 17-3-3S affinity-purified preparations was the same. In addition, the I-E^k purified using the Y-17 or the 17-3-3S Ab were functionally equivalent in their ability to present peptide in vitro. To demonstrate this, the purified I-E^k preparations were adsorbed to tissue culture plates, incubated with the cytochrome c peptide THMc (82–103) at pH 4.5, and cultured with the I-E^k-restricted cytochrome c-specific T cell hybrid TPc9.1. The two sources of I-E^k were equivalent in their ability to stimulate the TPc9.1 cells (Fig. 4). The magnitude of stimulation is the same as that attained by whole CH27 cells and peptides.

To determine the composition of the 120-kDa band observed in preparations of purified I-E^k, both the 60- and the 120-kDa bands were excised from gels. The proteins were electroeluted, reduced, boiled, reelectrophoresed on a second gel, and silver stained (Fig. 5). Under these conditions, the constituent proteins in both of the 60- and 120-kDa bands migrate as two bands, α and β, of 34 and 29 kDa, respectively. The excised 120-kDa band may have also contained peptide that was not resolved in the system.

To further demonstrate the identity of the 60- and 120-kDa I-E^k complexes, peptide mapping was performed. CH27 cells were metabolically labeled, washed, and lysed, and the I-E^k was purified, as described above. Approximately 10 μg of ³⁵S-labeled I-E^k was loaded onto a 10% SDS polyacrylamide gel without reducing or boiling. The gel was electrophoresed and dried and exposed to film. The autoradiogram revealed the location of the 60- and 120-kDa I-E^k complexes, allowing the precise excision of a portion of each band. Care was taken to avoid the upper region of the 60-kDa band, as this is likely to contain some contaminating Ii dimers (see Fig. 3). The gel pieces were rehydrated in buffer containing DTT and inserted into the wells of a 12% SDS polyacrylamide gel. The proteins were partially digested with protease from S. aureus V8, separated by electrophoresis, and imaged by fluorography (Fig. 6). Two different exposures are shown, to allow for optimal analysis of the digestion products from each gel piece. The products of proteolytic digestion derived from the 60- and 120-kDa I-E^k complexes appear to be highly related and within the resolution of the method that they can be considered the same protein. Peptide mapping of these complexes using the protease chymotrypsin also produced closely related patterns of digestion products (data not shown). This result, together with the silver-stained gels in Figures 3 and 5, confirms that the 60- and 120-kDa I-E^k complexes are composed of the same constituent proteins, and that the 120-kDa complex does not result from the addition of extraneous protein to the 60-kDa complex. The proteolytic digestion products of the 60- and 120-kDa complexes shown in Figure 6, as well as the electroeluted α- and β-chains shown in Figure 5, are present in the same relative amounts. Therefore, the stoichiometry of the constituent proteins in each of the two complexes is the same, and the 120-kDa complex observed in this study represents a dimer of the αβ heterodimer, the superdimer (αβ)₂ complex.

Having established that purified I-E^k has the tendency to form αβ and (αβ)₂ complexes in detergent solution, we examined the stability of these complexes to SDS denaturation. The observation that these complexes exist during SDS-PAGE indicates that they are held together either by very stable noncovalent associations, or by covalent bonds, perhaps an interchain disulfide bond, which stabilizes the complex against SDS-mediated denaturation. To address this question, Y-17-purified I-E^k was incubated in the absence of reducing agents at various temperatures for 30 min, in the presence of 2.5% SDS at pH 6.8 before electrophoresis on a 10% SDS polyacrylamide gel (Fig. 7). At 25° or 37°C, the I-E^k migrates as αβ and (αβ)₂ complexes. However, at 50°C, there is partial denaturation of both the αβ and (αβ)₂ complexes into individual α- and β-chains. At 65°C, all of the I-E^k is denatured into free α- and β-chains. The ability of these complexes to become denatured solely by the addition of heat indicates that they are noncovalent complexes. Both the αβ and (αβ)₂ appear to be partially decomposed at 50°C, suggesting that the two complexes have similar thermal stabilities. Since thermal denaturation does not result in an increase in the amount of (αβ)₂, it is unlikely that this complex forms as a result of nonspecific aggregation of I-E^k.

The superdimer (αβ)₂ complex has previously been observed in immunoprecipitates of newly synthesized I-E^k that has acquired peptide (43). The intravesicular pH of the subcellular compartment in which peptide loading onto class II molecules occurs has been measured to be approximately 4.6 (13). To examine the pH stability of αβ and (αβ)₂ complexes, aliquots of affinity-purified I-E^k were incubated at various pH levels for 30 min at 37°C in the presence of 2.5% SDS. These aliquots were then neutralized, electrophoresed on a 10% SDS polyacrylamide gel, and silver stained (Fig. 8). Both αβ and (αβ)₂ complexes are stable at pH levels between 6 and 9. At pH 5, these complexes partially denature and run as individual α- and β-chains. At pH 4 and 3, the complexes completely dissociate into the free chains. The dissociation is accompanied by a conversion of a portion of the αβ heterodimers into a complex having a m.w. slightly higher than that of the major αβ band, the compact heterodimer. This may be the floppy I-E^k, a conformational intermediate observed by Dornmair et al. (22) during the denaturation of compact αβ heterodimers to individual α- and β-chains. At pH 11.5, the (αβ)₂ and compact αβ complexes are unstable (Fig. 8). Taken together, these studies demonstrate that αβ and (αβ)₂ complexes display similar pH and thermal stability to SDS denaturation.

In this study, we further characterize the dimers and superdimers of I-E^k in B lymphocytes. We initially observed the 120-kDa superdimer of class II molecules using the mAbs 17-3-3S and Y-17, which immunoprecipitated I-E^k αβ dimers and (αβ)₂ superdimers, respectively (43). The molecular basis of the apparent specificity of these two mAbs is not known. Our previous flow-cytometric analysis showed that these mAbs bind simultaneously to I-E^k molecules expressed on cell surfaces and show no competition for binding to I-E^k (43). Mutational studies of I-E^k have shown that the epitopes recognized by the 17-3-3S and Y-17 mAb are largely distinct (48, 51). However, the epitopes for these two mAbs must be somewhat overlapping, because immobilized purified 17-3-3S/I-E^k complexes were shown not to bind to Y-17 by plasmon resonance (49). The best assessment of the epitopes that these two mAbs recognize is that 17-3-3S reacts with the β₁ domain, while Y-17 reacts either with an epitope on β₁, which is dependent on the proximal α₁ domain for conformation, or on an epitope including both the β₁ and α₁ domains (50).

The exhaustive immunoprecipitation analysis presented in this work confirms the specificity of the 17-3-3S and Y-17 mAbs for the αβ heterodimers and (αβ)₂ superdimers, respectively, and demonstrates that the αβ and (αβ)₂ complexes are derived from the same pool of I-E^k in cell lysates. Thus, all of the I-E^k molecules in the cell are capable of forming superdimers, under the conditions used in this study. Exhaustive immunoprecipitation also illustrates that the mAbs themselves can influence the conformation of the I-E^k. The form of the I-E^k complex observed following SDS-PAGE (αβ or (αβ)₂) depends solely on which mAb was used for immunoprecipitation (17-3-3S or Y-17). Thus, it is likely that the mAbs actually drive the I-E^k conformation toward one form or the other during immunoprecipitation. The 17-3-3S epitope may block superdimer formation by interfering with the β₁-β₁ interactions (25) at the most membrane-distal part of the (αβ)₂ superdimer interface. Y-17 may avoid such interference, and may actually bind to a β₁-α₁ epitope of an αβ heterodimer with one F(ab′)₂ arm and to another β₁-α₁ epitope with another Fab arm, thus bringing two αβ heterodimers together to form the (αβ)₂ superdimer.

Because of the influences these mAbs may have on I-E^k conformation, it was important to demonstrate the presence of the I-E^k superdimer in the absence of the Y-17 mAbs. This was accomplished in three ways. First, as previously shown, both 60-kDa αβ and 120-kDa (αβ)₂ complexes were detected by immunoblotting of untreated whole cell lysates (43). Second, both αβ and (αβ)₂ complexes were observed in preparations of affinity-purified I-E^k. Third, the 120-kDa superdimer can be chemically cross-linked on the cell surface in the absence of the Y-17 mAb. Therefore, while (αβ)₂ superdimer formation can be facilitated by the Y-17 mAb, I-E^k has the intrinsic ability to superdimerize and does not require Y-17 to do so.

It was also important to demonstrate that the superdimers exist in the absence of detergent, given the potential for detergent-induced artifactual associations. The fact that the superdimer did not appear in all class II preparations, in particular in the 17-3-3S immunoprecipitates, argues against detergent-induced class II dimerization. If exposure to SDS induces dimerization, then the 17-3-3S-immunoprecipitated class II molecules, which as discussed above have the inherent ability to dimerize, should dimerize. In addition, our previous studies, showing a differential effect of the Y-17 and 17-3-3S mAbs on the presentation of peptide by B cells, suggested that the superdimers did indeed exist on B cell surfaces (43). In this study, we provide direct biochemical evidence for the presence of superdimers in the absence of detergent by showing that the (αβ)₂ superdimer can be chemically cross-linked on B cell surfaces using the cross-linker DFDNB. DFDNB chemically cross-linked the α- and β-chains in αβ dimers as well as portions of (αβ)₂. In addition, we observed a phenomenon seen only with chemically cross-linked αβ and (αβ)₂, which was further oligomerized in vitro to form (αβ)₄ stabilized by disulfide bonds. Because the (αβ)₄ did not form in the 17-3-3S immunoprecipitates, we assume that the Y-17 mAb held the (αβ)₂ in a position that facilitated the disulfide bond formation.

The presence of (αβ)₂ class II superdimers was confirmed very recently by Cherry et al. (45), who used single-particle fluorescence imaging to demonstrate the presence of HLA-DR superdimers on the surface of a fibroblast cell line transfected with the genes encoding HLA-DR α- and β- and Ii chains. Their results also indicated the existence of class II oligomers larger than (αβ)₂. They report that approximately 25% of class II molecules on the cell surface are present as superdimers. By densitometry of the DFDNB-cross-linked (αβ) and (αβ)₂, we estimated that approximately 18% of class II molecules on the cell surface are present as superdimers in good agreement with the estimate of Cherry et al.

The formal proof that the 120-kDa band seen in this study in preparations of I-E^k purified by immunoaffinity chromatography is an (αβ)₂ complex was achieved by a comparison of the 60- and 120-kDa complexes following fully denaturing reelectrophoresis and peptide mapping. The similarity of the polypeptide patterns derived from each complex by these two methods verifies that they are composed of α- and β-chains in the same stoichiometric amounts. Therefore, the 120-kDa complex can only be a dimer of the 60-kDa complex.

The nature of the 120-kDa (αβ)₂ I-E^k complex remains to be determined. The structure of this complex may be similar to that which has been determined by x-ray crystallography of I-E^k (31). Alternatively, the 120-kDa complex may be stabilized by hydrophobic interactions between the transmembrane regions of the α- and β-chains. These two scenarios are not mutually exclusive, as it has been established that specific residues in the transmembrane domain of I-A^k facilitate assembly of the αβ heterodimer (63). The 120-kDa complex could also be the result of two I-E^k molecules binding to a single long peptide. The ability of the Y-17 mAb to drive total 120-kDa complex formation would argue against this possibility, because of the improbability of having enough peptides present that are capable of binding all I-E^k molecules in pairs, without having any monomers, trimers, tetramers, etc.

The (αβ)₂ superdimers observed in affinity-purified I-E^k preparations seem to have a stability to SDS denaturation similar to that of the αβ heterodimer. Both complexes partially decomposed at 50°C, with total dissociation into α- and β-chains at 65°C. This heat-sensitive nature of I-E^k under nonreducing conditions is evidence that the interactions involved are noncovalent, and the data in this study are similar to those in previous studies that showed that compact αβ heterodimers in β-OG micelles are stable at 50°C, but begin to dissociate at 65°C (22). Thus, it would appear that the stability of the (αβ)₂ superdimer to SDS denaturation depends on the stability of the constituent αβ heterodimers themselves.

The pH stability exhibited by the (αβ)₂ complexes derived from affinity-purified I-E^k appears to correlate with previously measured pH stabilities of I-E^k-peptide complexes (22). Dissociation into free α- and β-chains most likely reflects the loss of the heterodimer-stabilizing peptide. Recent studies have confirmed the longstanding model of a pH-dependent conformational change in MHC class II to facilitate peptide binding (64, 65). Furthermore, the intravesicular pH of the subcellular compartment, where peptide loading onto MHC class II molecules occurs, has been measured to be approximately 4.6 (13). While these conditions used to test the stability of αβ and (αβ)₂ complex described in this study are certainly different from those found in the peptide-loading compartment in vivo, these findings suggest that (αβ)₂ complexes may indeed be able to form intracellularly upon peptide binding to I-E^k.

A class II superdimer would have potentially important implications for the mechanism of T cell stimulation. One possibility is that a pair of class II molecules might interact with a pair of TCRs to help strengthen the rather weak affinity of MHC class II for the TCR (25). This model is supported by the crystal structure of the TCR Vα domain, which suggests an (αβ)₂ TCR conformation complementary to the HLA-DR1 structure (38). However, it is statistically unlikely that both MHC class II molecules in a superdimer would be filled with identical or closely related peptides, given the large number of peptides available in an APC. Even if the superdimer did contain two of the same peptides, simple dimerization of the TCR alone may provide only weak activation. Reich et al. (41) recently provided evidence for ligand-specific oligomerization of TCR/peptide/MHC complexes to form complexes containing two to six ternary complexes. The superdimers we observed may not be required for TCR dimerization, but may indicate the intrinsic ability to oligomerize when containing ligands for peptide-specific TCR.

Others have proposed that MHC class II superdimers may play a role in T cell signaling through CD4, a mechanism that would be independent of the peptides bound to MHC class II. These models depend largely on the ability of the superdimer to cross-link or oligomerize CD4 (reviewed in Refs. 42 and 66). The models are supported by evidence for multiple sites on MHC class II that are important for CD4 binding. Originally, the CD4 binding site was shown to lie only on the β₂ domain of HLA-DR (32, 33, 67). More recently, however, studies have shown that there is a second site of interaction with CD4 on the α₂ domain, oriented such that it cannot interact simultaneously with the same CD4 molecule as the β₂ site (35). To accommodate these data, either monomeric CD4 would cross-link MHC class II monomers, or MHC class II superdimers would interact with CD4 oligomers. It has recently been demonstrated that oligomers of CD4 are required for stable interaction with MHC class II (68), and an interaction between CD4 tetramers and MHC class II superdimers has been modeled (42). Thus, the weight of the data is consistent with a role for MHC class II superdimers in the binding of CD4. Such interactions of CD4 with the superdimer could contribute either to the overall adherence of the T cell and the APC and/or stabilize the TCR/class II dimer interaction. In this regard, it is interesting to note that the superdimer-specific mAb Y-17 (43) had the same effect as a CD4-specific mAb (53) on an Ag-specific T cell response, blocking the low affinity, but not high affinity, T cell response. Finally, a mutational analysis of residues on the putative dimer-dimer interface of HLA-DR3, identified on the basis of the residues that participate in the HLA-DR1 superdimer (25), has not supported a critical role for these residues in maintaining DR3 conformation peptide binding, or SDS-stable heterodimer formation (69). However, this study did not directly measure the ability of these mutant DR3 molecules to form superdimers. In addition, these two residues represent only a small part of the overall dimer-dimer interface, which encompasses more than 1300 Ų of surface area and includes seven salt bridges (69).

The data presented in this study have verified that the full-length mouse MHC class II molecule I-E^k has the intrinsic ability to superdimerize in detergent micelles and remain stable during SDS-PAGE under nonreducing, nonboiled conditions. Superdimerization may only be one step in the formation of larger order MHC class II/TCR/CD4 multimers, but its effects should nevertheless be significant for both T cell and APC signal transduction.