The I-A^g7 MHC Class II Molecule Linked to Murine Diabetes Is a Promiscuous Peptide Binder

DNA and bacterial strains

Phage vector pJCM13-88 containing a second synthetic copy of gene VIII has been described elsewhere (24). The random 12-mer peptide M13 phage display library JCM13-88-X12 was constructed in our laboratory in collaboration with Dr. C. Glass (R.W. Johnson Pharmaceutical Research Institute, La Jolla, CA). Plasmid pBSK-GAD65 containing the murine GAD65 coding sequence previously cloned from a C57BL/6 brain cDNA library was provided by Dr. J. Fehling (Basel Institute, Basel, Switzerland). The OVA coding sequence (25) subcloned into pCMU4 was obtained from Dr. K. Frueh (R.W. Johnson Pharmaceutical Research Institute). Bacterial strains Escherichia coli XL1-Blue and E. coli XLOLR were obtained from Stratagene (La Jolla, CA).

Construction, expression, and purification of I-A^g7 molecules

The wild-type cDNA for the I-A^g7 β-chain was modified to generate two constructs to express empty and peptide-filled molecules, as previously reported for I-A^d (26, 27). For the expression of empty molecules, the cDNA was interrupted after the codon for the last tryptophan before the transmembrane segment and extended by a linker (SSADL), a thrombin site (LVPRGS), a basic zipper, and a hexa-histidine tail. To tether peptides to the β-chain, a hybrid cDNA was constructed by substituting the natural signal sequence of I-A^g7 by the HLA-A2 signal sequence followed by the first four residues of the α1 domain of HLA-A2, a multiple cloning site, a glycine-serine linker, and the β1 domain of I-A^g7. Oligonucleotides corresponding to the peptide sequence were cloned between the SacII and BglII sites of the polylinker. The coding sequence of the extracellular domain of the human invariant chain (Ii) was cloned in frame with the HLA-A2 signal sequence followed by six histidine codons (26). The I-A^d α-chain was also modified by the addition of a BirA biotinylation sequence following the acidic zipper and proceeding the histidine tag (28). All constructs were subcloned into the metallothionein promoter-based fly expression vector pRMHa3 (29). Modified β-chains were transfected along with the soluble I-A^d acid zipper α-chain or soluble I-A^d acid zipper biotin α-chain (26, 27) to produce empty and peptide-filled I-A^g7 molecules. I-A^g7/Ii complexes were expressed by transfecting cDNAs for the α- and β-chain without histidine coding sequences and the human Ii construct. The various cDNA combinations were cotransfected into D. melanogaster SC2 cells by calcium phosphate precipitation. A neomycin-resistance gene was added to the transfection at a 1:200 ratio to allow the selection of stable cell lines in the presence of 0.5 mg/ml of G418. Expression was tested at 4 wk after a 3-day induction with 0.7 mM CuSO₄ and immunoprecipitation with nickel-agarose beads (Qiagen, Chatsworth, CA). Large-scale preparations were made using serum-free media (Insect-Xpress; BioWhittaker, Walkersville, MD) in roller bottles. Supernatants were concentrated and buffer exchanged against PBS by tangential flow. Recombinant proteins were purified at 4°C by batch binding to nickel-agarose beads and elution with a step gradient of imidazole. Empty molecules were further purified by gel filtration over a Superdex 200 column. Peptide-filled molecules were purified to homogeneity by ion exchange chromatography over a MonoQ10/10 column at pH 8.4 (Pharmacia, Piscataway, NJ). I-A^g7 Ii complexes were produced in Schneider’s media, 10% FCS, in triple-layered flasks. Metal induction was conducted in the presence of 5 mM pepstatin A and 5 mM leupeptin. Following the initial nickel-agarose step, purification was conducted by a succession of anion exchange (MonoQ10/10), hydrophobic interaction chromatography (Phenylsuperose 10), and gel filtration (Superdex 200). Purified proteins were kept in PBS, pH 7.4, containing 0.02% NaN₃ at 4°C.

Biochemical analysis, SDS-PAGE, and isoelectric focusing

Purifications were monitored by SDS-PAGE followed by Coomassie Brilliant Blue R 250 staining. Two-dimensional gel electrophoresis was performed as described (26) using 3.5–10.0 nonequilibrium pH gradient electrophoresis (NEPHGE) in the first dimension and 12% SDS-PAGE in the second dimension followed by Coomassie Brilliant Blue staining.

Vector constructs and random cDNA fragment libraries

Control phage constructs pJCM13-88-OVA and pJCM13-88-ROI coding for the OVA peptide 323-339 (ISQAVHAAHAEINEAGR) and the synthetic peptide ROI (VHAAHAVHAAHAVHA), respectively, were cloned as fusions to the N terminus of gene VIII by a GGGGS linker sequence. Vector pORFES IV was digested with NheI/HindIII and the stuffer replaced by the following annealed oligonucleotides: 5′-CTAGCGTAGCTCAGGCCGGTGGCGGTATATCTCAAGCTGTCCATGCAGCACATGCAGAAATCAATGAAGCAGGCAGAGGTGGCGGTGGCTCCCA-3′, 5′-AGCTTGGGAGCCACCGCCACCTCTGCCTGCTTCATTGATTTCTGCAT GTGCTGCATGGACAGCTTGAGATATACCGCCACCGGCCTGAGCTAC G-3′ (pORFESIVOVA); 5′-CTAGCGTAGCTCAGGCCGGTGGCGGTGTTC ACGCTGCTCACGCAGTTCACGCTGCACACGCTGTTCACGCAGGTGG CGGTGGCTCCCA-3′ and 5′-AGCTTGGGAGCCACCGCCACCTGCGTGA ACAGCGTGTGCAGCGTGAACTGCGTGAGCAGCGTGAACACCGCCAC CGGCCTGAGCTACG-3′ (pORFESIV-ROI). The ligated vectors were digested with XbaI/HindIII, and the inserts were directionally cloned into JCM13-88.

The construction of cDNA fragment phage display libraries will be described in details elsewhere (T. Stratmann, L. Teyton, and A. S. Kang, manuscript in preparation). Briefly, plasmids pCMU4 and pBSk-GAD65 were treated with BamHI and SacI/HindIII to release the GAD65 and OVA coding sequence, respectively. Random DNA fragmentation was established by digesting the inserts with DNase I and, subsequently, with mung bean nuclease to obtain blunt ends. The DNA fragments were ligated into pORFESIV that had been cleaved with NaeI and transformed into E. coli XLOLR cells, resulting in two primary libraries of 6 × 10⁵ (OVA) and 2.6 × 10⁶ (GAD65) independent clones. The fragments restoring the reading frame of the selection marker β-lactamase were rescued by treatment of the recovered plasmids with XbaI/HindIII and separated by agarose gel electrophoresis. Inserts of ∼160–210 bp of length (corresponding to 40–90 bp of random DNA fragment length) were isolated and cloned into pJCM13-88 to express peptide fragments at the N terminus of the synthetic copies of gene VIII of bacteriophage M13 upon transformation into E. coli XL1-Blue cells. Bacteriophages were propagated overnight at 30°C and prepared by standard methods. The sizes of the final libraries were 1.7 × 10⁶ independent clones for the OVA library, JCM13-88-rOVA, and 7 × 10⁶ clones for the GAD65 library, JCM13-88-rGAD65.

Phage inhibition assay and selection of M13 random 12-mer peptide and cDNA fragment display libraries

To assess specific binding of phages JCM13-88-OVA and JCM13-88-ROI to empty I-A^g7molecules, phages were incubated with 20 ng of recombinant I-A^g7 for 2 h at 37°C in 100 μl of 50 mM NaAc, pH 5.0, and 0.05% Tween 20 containing 0.02% NaN₃ in a micro tube in duplicates in the presence or absence of the ROI peptide at various concentrations. After neutralization with 1 M Tris, samples were transferred to Reacti-Bind metal chelate microtiter plates (100 μl per well; Pierce, Rockford, IL) that had been blocked with 0.5% BSA in PBS containing 0.02% NaN_3, and incubated at room temperature for 2 h under gentle agitation. Each well was washed 10 times with 300 μl of PBS containing 0.05% Tween 20 (PBST), and binding phages were eluted with 100 μl of 100 mM glycine, pH 2.2, for 10 min at room temperature. After neutralization with 1 M Tris, the phage titer was determined. Selection of the libraries JCM13-88-X12, JCM13-88-rOVA, and JCM13-88-rGAD65 was as described above, with the difference that 200 ng instead of 20 ng of I-A^g7 was used and that no free peptide was added. Eluted phages were propagated in E. coli XL1-Blue cells, and the selection was repeated three or four times. After selection, DNA from randomly chosen individual clones was prepared and analyzed by sequencing.

Peptide competition assay to MHC class II molecules

A modified competition assay (ELISA) was used to determine binding of peptides to I-A^g7 and I-A^d molecules using biotinylated OVA_323–339, or in some cases, biotinylated ROI peptide and unlabeled peptides (13, 30). Microtiter plates (96-well, Maxisorb; Nunc, Naperville, IL) were coated with anti-I-A β-chain-specific mAb M5.114 (100 μl/well, 5 μg/ml in 100 mM NaCO₃, pH 9.6; American Type Culture Collection, Manassas, VA) overnight at 4°C. Plates were washed five times with PBST, blocked overnight with 200 μl of PBS containing 5% BSA at 37°C, and washed. In a second set of plates, 200 ng of empty I-A^g7 or I-A^d was incubated in 100 μl binding buffer (6.7 mM citric phosphate, pH 10.4, 0.15 M NaCl, 2% Nonidet P-40, 2 mM EDTA) containing 300 nM of biotinylated OVA_323–339 peptide with varying concentrations of test peptide at 37°C overnight. The samples were transferred to the coated plates and incubated for 4 h at room temperature. The plates were washed 10 times with PBST for >1 h, followed by five additional washes with PBS for >30 min. Plates were incubated with 100 μl/well of PBS containing streptavidin-conjugated alkaline phosphatase (0.7 μg/ml) for 1 h at room temperature and washed as before. Biotinylated molecules were colorimetrically detected by overnight incubation with para-nitrophenyl phosphate at 4°C, and the absorbance at λ_{405 nm} was determined using a 96-well plate reader. Although we observed differences between batches of I-A^g7 and I-A^d molecules in terms of maximal binding capacity and half-lives of the preparations, their IC₅₀ values and the relative binding capacity of I-A^g7 vs I-A^d were reproducible.

Biotinylation and FACS analysis

MHC class II molecules with the biotinylation sequence were purified, as described above, and biotinylated with the BirA enzyme according to the manufacturer instructions (Avidity, Denver, Co). Biotinylated molecules were tetramerized with PE-labeled streptavidin (BioSource International, Camarillo, CA) and used for FACS. Staining was performed at 4°C for 2 h with 10 μM of labeled tetramers. Analysis was performed on a FACScaliber instrument (Becton Dickinson, Mountain View, CA). Data were analyzed using CellQuest software (Becton Dickinson). GAD65_524–538-specific hybridomas were provided by Dr. P. Reich (Anergen, Redwood City, CA).

Expression of empty molecules

We have previously reported the successful production of large amounts of I-A^d molecules in D. melanogaster SC2 cells and their crystallization with covalently linked single peptides (14, 26, 27). Soluble, secreted, and empty I-A^g7 molecules were expressed in a similar manner. Their transmembrane and cytoplasmic domains were deleted and replaced by leucine zippers to enhance heterodimeric chain pairing (14, 27). Levels of expression in SC2 cells were comparable to I-A^d expression levels, and oscillated between 0.2 and 0.5 mg/L of final purified product. No aggregation of the empty molecules was noted at any stage of purification. The purification of these molecules was conducted by a combination of nickel-agarose affinity chromatography and gel filtration at neutral pH (Fig. 1⇓). The α- and β-chains were secreted at equimolar ratios (Fig. 1⇓A). To test the stability of these dimers, leucine zippers were removed by thrombin digestion at 37°C for 4 h and the digest was analyzed by gel filtration and SDS-PAGE after undigested molecules were removed using nickel-agarose beads. Uncut and cut material were incubated at 37°C for 60 min before gel filtration. Both samples had the same retention time (15 ml, 55 kDa) on the Superdex 200 profile (Fig. 1⇓B) indicating that, with or without zippers, these molecules had no tendency to dissociate into monomeric chains or to aggregate, suggesting that the loss of the β57/α76 salt bridge was not critical for the stability of the I-A^g7 heterodimer. Unlike Ab affinity chromatography (26), this purification procedure does not use extreme pH and has a major impact in term of assessing the binding of peptides to purified empty I-A^g7 or I-A^d molecules. Whereas Ab affinity-purified empty MHC molecules were really empty (26), molecules purified at neutral pH were found to be associated with a large collection of peptides that could be stripped at acidic pH and analyzed by mass spectrometry (Fig. 1⇓C). This same phenomenon has been described for empty MHC class I molecules expressed in the fly system (31). Thus, it would be proper to call these molecules “functionally empty”, instead of “empty”, because they are associated with low affinity peptides. These peptides could easily be displaced by higher affinity peptides in a binding assay (Fig. 2⇓). Direct binding of peptides and specificity of peptide binding (tested by competition) were shown for both I-A^d and I-A^g7 with the ROI peptide (VHAAHAVHAAHAVHA) (Fig. 2⇓, A and B).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

Expression of recombinant I-A^g7 molecules. A, Empty I-A^g7 molecules were purified from transfected SC2 cell culture supernatant and purified by metal chelate affinity chromatography followed by size exclusion chromatography. In addition, I-A^g7 molecules complexed to the Ii were purified by ion exchange chromatography. Eluted fractions were analyzed by one-dimensional SDS-PAGE and the pooled fractions by two-dimensional gel electrophoresis (NEPHGE followed by 12% SDS-PAGE). B, Size exclusion chromatography of empty I-A^g7 molecules before and after cleavage of the zipper. C, Mass spectrometry analysis of peptides obtained by acid elution from “empty” I-A^g7 molecules.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

Examples for binding of peptides by empty I-A^g7 and I-A^d molecules. A, I-A^g7 and I-A^d molecules were incubated with increasing concentrations of biotinylated OVA_323–339 peptide in a direct binding assay. B, For binding inhibition assays, I-A^g7 and I-A^d molecules were incubated with 300 nM of biotinylated OVA_323–339 peptide and increasing concentrations of test peptides. The inhibition using the ROI peptide is shown. Negative control peptides for I-A^g7 and I-A^d were HEL_50–62 and GAD_345–361, respectively.

Association of I-A^g7 molecules with the Ii

Correct folding of the recombinant I-A^g7 molecule in the fly expression system was further demonstrated by its ability to associate with the Ii. A soluble form of human Ii with an N-terminal hexa-histidine tag was cotransfected with I-A^g7 α and β constructs retaining zippers but no histidine tag. The cell line was cloned and the single cell clone 134, which expresses equivalent amounts of the three chains, was used for further studies. Complexes of the three chains, α-β-Ii, were isolated by nickel-agarose chromatography (binding only Ii) followed by gel filtration, hydrophobic interaction chromatography, and ion exchange chromatography. The presence of all three chains was demonstrated by SDS-PAGE and two-dimensional gel analysis of the final pooled fractions (Fig. 1⇑A). Again, the behavior of I-A^g7/Ii complexes was almost identical with the behavior of I-A^d/Ii complexes (26). Peptides could be loaded on these complexes at acidic pH in the absence of accessory molecules (data not shown, 26). These results confirmed the observation by Peterson and Sant (17) that I-A^g7 associates normally with the Ii, excluding possible gross defects of intercellular trafficking.

Promiscuous peptide motif for I-A^d and I-A^g7 molecules

A broad number of MHC class I and class II molecules binding motifs have been identified by direct sequencing of natural peptides and/or binding of random peptides and sequencing (32). For HLA-DQ and I-A molecules, the same strategy has not proven very fruitful, and most of these haplotypes have no clearly defined motif to date. In the case of I-A^g7, the number of studies addressing the issue is very limited and they have yielded rather controversial results (11, 12, 13). To try to address this question, we followed a strategy previously used by Hammer and colleagues to identify the HLA-DR1 class II molecule binding motif based on the selection of random peptide M13 phage display libraries with purified MHC molecules (33).

A random 12-mer peptide phage library was selected by four consecutive rounds of panning against empty I-A^g7 molecules. An increase of the absolute amount of eluted phages, combined with an increase of binding over background up to ∼2700-fold after the third round of selection indicated the enrichment of binding clones from the original library (Fig. 3⇓A). One hundred and two clones of the third and fourth round of selection and 107 clones from the naive library were further analyzed by direct sequencing (Table I⇓). In the selected library, four different clones were identified multiple times. A binding motif was sought by manual alignment and by using CLUSTAL W 1.7 (Human Center, Baylor College of Medicine, Houston, TX; Ref. 34), as well as PILEUP software (program of the Genetics Computer Group, Madison, WI). When the total numbers of all 20 aa from the enriched library (averaged to 100 clones) were compared with the total numbers of the same amount of sequences from the naive library (Table II⇓), only minor enrichments were found in the selected library for cysteine (65 vs 50), glutamate (41 vs 31), and phenylalanine (75 vs 51). However, these residues were randomly distributed along the sequence, and no motif was apparent.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

Selection phage display libraries against empty I-A^g7 and I-A^d molecules. A and B, A random 12-mer peptide phage display library was selected for four rounds against I-A^g7 molecules and for two rounds against I-A^d molecules. C and D, Random fragment libraries prepared from OVA and GAD65 cDNAs, respectively, were selected for four rounds against I-A^g7 molecules.

Synthetic peptides were then made for two of the sequences picked randomly from the I-A^g7-selected library, TSS31 (NYYPWSIIPSIG) and TSS39 (HVHYPAVLIPRD), and tested for binding to I-A^g7 and I-A^d. TSS31 bound equally well to both I-A molecules (0.08 and 0.15 μM for I-A^g7 and I-A^d, respectively), whereas TSS39 was a weak binder to I-A^d (0.14 μM) but a stronger binder to I-A^g7 (0.016 μM) (Table III⇓). These results suggested specific selection for I-A^g7 binding during panning. With respect to negatively charged residues that could compensate for the absence of the aspartic acid β57 in I-A^g7, aspartic and glutamic acids were found more frequently within the last four residues of the I-A^g7-selected library (30.4%, Table I⇑) than in the I-A^d (12.3%, Table IV⇓) or naive (22.4%, Table II⇑) libraries.

The same experiment of panning the random library on purified MHC molecules was repeated with empty I-A^d molecules (Fig. 3⇑B). A strong enrichment of binding clones was observed after one and two rounds of panning. A total of 57 clones were sequenced and analyzed for a possible binding motif (Table IV⇑). Two hydrophobic amino acids were enriched in the selected library, leucine (151 vs 117) and phenylalanine (72 vs 51), whereas lysine was also increased (44 vs 28) and serine diminished (116 vs 142). Again, no apparent motif could be determined, but the prevalence of hydrophobic residues satisfies the optimal requirement for such residues at the P4 and P6 positions in binding peptide to I-A^d, as deduced from the crystal structure of two I-A^d peptide complexes (14).

We used the three-dimensional structure of I-A^g7 complexed with GAD_207–220 to rationalize these binding studies. Like I-A^d, I-A^g7 does not use large side chains to anchor in its pockets for peptide binding. Instead, the binding energy is provided mainly by interaction of the peptide backbone with the peptide-binding groove. From the structural data, we were able to suggest the following binding motif: P1, degenerate; P4 optimal, P, L, V, I, G, A, and C; P4 nonoptimal, M, S, T, N, Q, and H; P6 optimal, L, V, I, P, S, T, G, A, C, and N; P6 nonoptimal, D; P9 optimal, D, E, S, G, and A; P9 nonoptimal, M, L, I, V, P, C, N, H, Q, and T (15). Using that motif, we could align the vast majority of peptides known to bind to I-A^g7. Peptides from the unselected, as well as the selected I-A^g7 and I-A^d peptide libraries, respectively, were then aligned based on that motif. In the unselected (naive) library, 36% of the sequences could satisfactorily be fitted using the optimal alignment for I-A^g7 (P4, P6, and P9 using optimal side chains). This number rose to 74% if one nonoptimal residue was allowed only at one position. After selection against I-A^g7, these numbers changed for round 3 to 27 and 65% (optimal and nonoptimal fit peptides, respectively) and to 32 and 64% for round 4. Two clones, which were identified multiple times in round 3, were eliminated under these strict rules for motif-containing clones (Table I⇑). However, it is likely that these sequences bind to I-A^g7; this hypothesis will be tested later using synthetic peptides. For I-A^d, 54 and 55% could be aligned in the unselected and the selected library using the optimal and nonoptimal motif, respectively. In summary, essentially no enhancement of sequences containing this binding motif was observed after selection. This result can be easily explained by the high numbers of potential binding clones in the unselected library. For comparison, <1% of clones of the naive library contained a motif for I-E^b or I-A^k (33, 35). Selection of the same library for binding to these particular MHC class II molecules should lead to a clear enrichment of motif-containing clones.

Isolation of OVA and GAD65 peptides able to associate to I-A^g7 molecules

The number of peptides known to bind I-A^g7 molecules is limited. The number of known self-peptides is even smaller and, for most of them, their affinity is unknown. Because functional and structural studies of MHC molecules require the identification of the appropriate peptide, we decided to design a general method to identify I-A^g7 binding peptides. Several groups, including ours, have previously shown that Ab-specific epitopes can be retrieved from random cDNA fragment phage display libraries (36, 37, 38). Generally, these libraries consist of 15–30 aa peptide fragments that are generated by random cDNA fragmentation and expressed as fusions to one of the surface proteins of bacteriophage M13. A similar strategy was used to isolate peptides that bound to empty MHC class II molecules. The approach was tested by constructing two phages displaying peptides known to bind I-A^g7 and I-A^d: OVA_323–339 (phage JCM13-88-OVA) and ROI (phage JCM13-88-ROI) (13, 39). These peptides were fused to the N terminus of protein VIII and a four glycine-one serine linker. Specificity of binding was demonstrated by incubating JCM13-88-OVA and JCM13-88-ROI phages with empty I-A^g7 molecules with increasing concentrations of free ROI peptide. Binding of phages was dose-dependently decreased by increasing doses of ROI but not by a non-I-A^g7-specific control peptide (data not shown). Based on this result, two cDNA fragment libraries were constructed from the OVA and murine GAD65 cDNAs called JCM13-88-rOVA and JCM13-88-rGAD65, respectively. GAD65 was chosen because it has been identified as one of the major autoantigens in IDDM in both humans and mice (40, 41, 42). OVA was used as a control. Both libraries were subjected to four successive rounds of selection (Fig. 3⇑, C and D). After the fourth round, binding of the JCM13-88-rOVA library had doubled as compared with the first round of selection, and binding was 650 times more than background. To identify binding peptides, phage clones were randomly chosen from the third and fourth rounds of selection and sequenced (Table V⇓). Of 60 clones analyzed, 13 clones were coded for OVA peptides. Alignment of these sequences revealed two major consensus sequences: GISSAESLKISQA, OVA_314–326 and SVSEE FRADHPFL, OVA_353–365. The previously described I-A^g7 and I-A^d epitope OVA_323–339 sequence was found in only one clone. Both consensus sequences could be aligned multiple times using our predicted I-A^g7 motif. The affinity of OVA_323–339 and OVA_353–365 synthetic peptides was measured directly for empty I-A^g7 molecules (Table III⇑). Interestingly, OVA_353–365 did not bind to I-A^d but, at 25 nM, was one of the best I-A^g7 binders that we identified in this study compared with 0.1 μM for OVA_323–339. Thus, these data suggest that I-A^g7 and I-A^d can bind both similar and haplotype-specific epitopes from the same protein.

From the GAD65 library, 54 clones were sequenced. Seventeen clones (8/27 in round 3, 9/27 in round 4) coded for known GAD65 peptide sequences (Table VI⇓). The most represented sequence was YEIAPVFVLLE, GAD65 residues 207-217, which appeared in 10 clones (59%) and which mapped within a sequence previously identified as the major GAD65 epitope in vivo (43, 44, 45). The second consensus sequence, GAD65_105–119, had not been described as being a functional epitope, but its binding to I-A^g7 molecules was confirmed in a binding assay (Table III⇑). The second major functional epitope identified by Chao and McDevitt as well as Zechel et al. (43, 44, 45), GAD65 residues 221–235, partially overlapped with clone TSS116. Therefore, it is noticeable that, in vitro, we were able to identify two of the three major functional determinants for GAD65.

The various peptides that were identified and chemically synthesized, as well as a collection of peptides that had been shown previously by other groups to stimulate I-A^g7-restricted autoreactive T cells (40, 43, 44, 45), were tested for their ability to bind empty I-A^g7 and I-A^d molecules (Table III⇑). From the compilation of these data, three main features emerged: 1) none of the peptides had very high affinity (<1 nM) for I-A^d or I-A^g7 molecules, 2) at the cutoff of 10 μM (medium affinity), I-A^g7 bound slightly more peptides (16/20) than I-A^d molecules (9/20), and 3) specificity was high with only 9 of 20 peptides binding to both haplotypes (<10 μM). Also, as recently reported (46), the binding of the full-length class II-associated invariant chain peptide (81-104, human sequence) to I-A^g7 was found to be of relatively low affinity (1.3 μM) compared with its binding to I-A^d (0.18 μM).

Expression and crystallization of a series of I-A^g7 peptide complexes

Because one of our goals was to carry out functional and structural studies of I-A^g7 molecules, a series of peptide sequences were cloned in frame into the HLA-A2-I-A^g7 β-chain hybrid construct (see Materials and Methods) and transfected with the I-A^d α-chain into SC2 cells. The functionality of the recombinant product was tested for its ability to stimulate T cell hybridomas in a plate assay (26 , and data not shown) and to stain T cells after biotinylation and tetramerization with PE-labeled streptavidin (Fig. 4⇓). All molecules were purified by a succession of nickel-agarose chromatography, ion exchange, and gel filtration (isoelectric point values ranged from 5.5 to 7.5). The two-dimensional analysis of some of these molecules is shown in Fig. 5⇓. Protein yields ranged from 0.2 to 2 mg/L and correlated very closely with peptide affinities. The crystallization of four complexes was successful: OVA_323–339, ROI, GAD_207–220, and GAD_221–235. However, the I-A^g7 OVA_323–339 and I-A^g7 ROI complex crystals were needles that could not be grown to sufficient size for crystallographic analysis. But, data quality crystals could be obtained, after removal of the leucine zipper with thrombin digestion, for I-A^g7 GAD_207–220 and GAD_221–235 complexes. Diffraction data to 2.6 and 3.0Å were obtained for I-A^g7 GAD_207–220 and GAD_221–235, respectively, upon cryocooling of the crystals, and these structures have been determined by x-ray crystallography (15).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4.

Flow cytometric analysis of three different murine T cell hybridomas specific for GAD_524–538. Hybridomas were stained with biotinylated I-A^g7 molecules tethered to the GAD_524–538 peptide and tetramerized using PE-labeled streptavidin (dotted lines). Empty I-A^g7 tetramers were used as a reference for staining (solid lines).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 5.

Expression of peptide-filled I-A^g7 molecules. A series of peptide sequences was tethered to the β-chain of I-A^g7. Expression of both chains was analyzed by two-dimensional gel electrophoresis (NEPHGE followed by a 12% SDS-PAGE) using purified molecules.