Although amino acid motifs for peptides that bind to individual class I and some class II MHC molecules have been well defined (5, 6), the rules that govern binding of peptides to I-A^g7 are still unclear. Reich et al. (7) eluted and sequenced several naturally processed peptides from I-A^g7 and concluded that binding may require an acidic residue in the COOH terminus of the peptide. Carrasco-Marin et al. (8) found that I-A^g7 either on the surface of antigen-presenting cells or in SDS-PAGE after its purification was unstable and that the binding of known I-A^g7–restricted T cell epitopes or the peptides eluted by Reich et al. (7) was difficult or impossible to demonstrate. This led them to hypothesize that weak peptide binding by I-A^g7 militated against elimination of autoreactive T cells in the NOD mouse. Amor et al. (9) investigated the fine specificity of peptides from myelin oligodendrocyte glycoprotein (MOG) or proteolipid protein (PLP) for the induction of experimental allergic encephalomyelitis (EAE) in Biozzi AB/H mice and suggested a core motif for I-A^g7 binding peptides.

In this study, we used the I-A^g7–restricted T cell epitope, hen egg white lysozyme (HEL) amino acids 9–27, as a template with which to analyze the amino acid sequence of peptides that bind to purified, native I-A^g7 and activate a T cell hybridoma. This has enabled us to define general rules that identify most known I-A^g7 binding peptides.

Purification of I-A^g7. I-A^g7 protein was affinity-purified from detergent lysates of 4G4.7 B cell hybridoma cells by desorption from OX-6 mouse monoclonal antibody. The 4G4.7 B cell hybridoma was derived by polyethylene glycol (PEG)-induced fusion of NOD mouse T cell–depleted splenocytes with the HAT-sensitive A20.2J lymphoma line (10). OX-6 is a mouse monoclonal IgG1 antibody against an invariant determinant of rat Ia, which also recognizes I-A^g7 but not I-A^d (11, 12). Approximately 15 mg of OX-6 antibody was first bound to 4 ml of protein A–Sepharose 4 Fastflow (Pharmacia, Uppsala, Sweden) and then chemically cross-linked to the protein A with dimethyl pimelimidate dihydrochloride (Sigma Chemical Co., St. Louis, MO) in sodium borate buffer, pH 9.0. After 60 min at room temperature (RT), the reaction was quenched by incubating the Sepharose in 0.2 M ethanolamine, pH 8.0, for 60 min at RT. The suspension was washed thoroughly in PBS and stored in PBS, 0.02% sodium azide (NaN₃).

4G4.7 cells were harvested by centrifugation, washed in PBS, resuspended at 10⁸ cells/ml of lysis buffer, and then allowed to stand at 4°C for 120 min. The lysis buffer was 0.05 M sodium phosphate, pH 7.5, containing 0.15 M NaCl, 1% (vol/vol) NP40 detergent and the following protease inhibitors: 1 mM phenylmethylsulphonyl fluoride, 5 mM ε-amino-n-caproic acid and 10 μg/ml each of soybean trypsin inhibitor, antipain, pepstatin, leupeptin and chymotrypsin. Lysates were cleared of nuclei and debris by centrifugation at 27,000 g for 30 min and stored as such if not immediately processed further. To the postnuclear supernatant was added 0.2 vol of 5% sodium deoxycholate (DOC). After mixing at 4°C for 10 min, the supernatant was centrifuged at 100,000 g at 4°C for 120 min, carefully decanted, and filtered through a 0.45-μm nylon membrane. The lysate of 5 × 10¹⁰ 4G4.7 cells was gently mixed overnight at 4°C with 4 ml of OX6–protein A–Sepharose, and the suspension then poured into a column and washed with at least 50 vol each of buffers A, B, and C. Buffer A was 0.05 M Tris, pH 8.0, 0.15 M NaCl, 0.5% NP40, 0.5% DOC, 10% glycerol, and 0.03% NaN₃; buffer B was 0.05 M Tris, pH 9.0, 0.5 M NaCl, 0.5% NP-40, 0.5% DOC, 10% glycerol, and 0.03% NaN₃; buffer C was 2 mM Tris, pH 8.0, 1% octyl-β-d-glucopyranoside (OGP), 10% glycerol, and 0.03% NaN₃. Bound I-A^g7 was eluted with 50 mM diethylamine HCl, pH 11.5 in 0.15 M NaCl, 1 mM EDTA, 1% OGP, 10% glycerol, and 0.03% NaN₃, and immediately neutralized with 1 M Tris.

Peptide Synthesis. Peptides were synthesized with a multiple peptide synthesizer (model 396; Advanced ChemTech, Louisville, KY) using Fmoc chemistry and solid phase synthesis on Rink Amide resin. All acylation reactions were effected with a threefold excess of activated Fmoc amino acids, and a standard coupling time of 20 min was used. Each Fmoc amino acid was coupled at least twice. Cleavage and side chain deprotection was achieved by treating the resin with 90% trifluoroacetic acid, 5% thioanisole, 2.5% phenol, 2.5% water. The indicator peptide for the binding assay was biotinylated before being cleaved from resin by coupling two 6-aminocaproic acid spacers on the NH₂ terminus and one biotin molecule sequentially, using the above-described procedure. Individual peptides were analyzed by reverse-phase HPLC and those used in this study were routinely 85% pure.

T Cell Hybridoma. Hybridoma 2D12.1 was generated by PEG- induced fusion of HEL-immune lymph node cells from a NOD mouse with the TCR-α/β–negative variant of the BW5147 thymoma, as described previously (13). Reactivity of 2D12.1 to HEL peptides was assayed by incubating 2.5 × 10⁵ NOD spleen cells and HEL peptides (0.3 nM to 10 μm) with 5 × 10⁴ T hybridoma cells/well. Culture medium was RPMI 1640 supplemented with 10% FCS, 2 mM l-glutamine, 50 μg/ml gentamicin, and 50 μm 2-mercaptoethanol. After 24 h of culture, 50 μl of supernatants were transferred to culture wells containing 10⁴ IL-2– responsive CTLL-2 cells. During the final 4 h of a 24-h culture, CTLL-2 cells were pulsed with 1 μCi [³H]thymidine. Thymidine incorporation was measured by scintillation spectrometry. The concentration of peptide that caused 50% of maximum stimulation is referred to as SC₅₀.

I-A^g7 Peptide-binding Assay. Peptides were dissolved at 10 mM in DMSO and diluted into 20% DMSO/PBS for assay. Indicator I-A^g7 binding peptide, HEL 10–23, was synthesized with a biotin molecule and two spacer residues at the NH₂ terminus. Approximately 200 nM of this biotinylated HEL peptide and each test peptide in seven concentrations ranging from 50 μM to 50 pM, were coincubated with ~200 ng of I-A^g7 protein in U-bottomed polypropylene 96-well plates (Costar Serocluster, Costar Corp., Cambridge, MA) in binding buffer at RT. The binding buffer was 6.7 mM citric phosphate, pH 7.0, with 0.15 M NaCl, 2% NP-40, 2 mM EDTA, and the protease inhibitors as used in the lysis buffer. After a minimum of 24 h, each incubate was transferred to the corresponding well of an ELISA plate (Nunc Maxisorp, Nunc, Roskilde, Denmark) containing prebound OX-6 antibody (5 μg/ml overnight at 4°C, followed by washing). After incubation at RT for at least 2 h, and washing, bound biotinylated peptide–I-A^g7 complexes were detected colorimetrically at 405 nm after reaction with streptavidin–alkaline phosphatase and paranitrophenolphosphate. Competition binding curves were plotted and the affinity of peptide for I-A^g7 was expressed as an inhibitory concentration 50 (IC₅₀), the concentration of peptide required to inhibit the binding of bio-HEL 10–23 by 50%.

I-A^g7 Purification and Binding Assay. Approximately 2 mg of protein, estimated by Coomassie blue binding (Bio Rad Protein assay), was purified from 5 × 10¹⁰ 4G4.7 cells. In SDS-PAGE, the majority (>95%) of the protein was resolved as two bands of molecular weight ~33,000 and ~28,000 that correspond to the α and β subunits, respectively, of mouse class II MHC molecules (data not shown). The competition binding assay with purified I-A^g7 was sensitive and specific (Fig. 1), and highly reproducible; in 15 separate assays the mean ± SD of the IC₅₀ for competition between biotinylated and unlabeled HEL 10–23 was 295 ± 72 nM.

Figure 1 Examples of competition between biotinylated HEL peptide (amino acids 10–23) and unlabeled peptides for binding to purified I-Ag7, measured by ELISA (see Materials and Methods). Unlabeled HEL 10–23 (□) was used as an internal (more ...)

Carrasco-Marin et al. (8) were unable to demonstrate direct binding of HEL 11–25 to purified I-A^g7 and proposed that I-A^g7 was inherently unstable. We found that purified I-A^g7 stored at −70°C for more than 1 yr reproducibly bound HEL 10–23 with high affinity. Therefore, our results do not support their hypothesis that I-A^g7 is inherently unstable, which they postulated would impair its ability to bind and induce tolerance to autoreactive peptides.

Truncation Analysis of HEL 9–27. Peptides representing sequential truncations of HEL 9–27, from either the NH₂ or COOH-terminus, were each assayed in parallel for binding to I-A^g7 and for their ability to activate the 2D12.1 hybridoma. Inspection of these data (Table 1) reveals that the minimum T cell epitope is M12-R21, and the minimum binder is M12-Y20 or K13-R21.

Table 1 Truncation Analysis of I-Ag7–restricted HEL 9–27 Epitope

Effect of Selected Substitutions on Binding and Bioactivity of HEL 12–22. Substitution of alanine (A) at each position in HEL 12–22 (Table 2) had no significant effect on binding, with the sole exceptions of positions L17 and Y20. Substitution at either of these two positions virtually abolished binding. On the other hand, while having no effect on binding, substitutions by A at K13, R14, H15, G16, and D18, and to a lesser extent at R21, abolished T cell activation. Removal of R21 (see Table 1) abolished T cell activation. Further substitutions of representative amino acids (D, K, P, Y, L, Q) at each position (Table 2) revealed varying levels of tolerance of specific residues/positions for binding (see below) and generally confirmed the results of the alanine substitutions on T cell activation. On the basis of these results, we can deduce that most residues in the minimal T cell epitope HEL 12–21 have TCR contacts and that two, L17 and Y20, are essential for binding to I-A^g7 (Fig. 2).

Table 2 Effect of Selected Amino Acid Substitutions on Binding and T Cell Activation of HEL 10–22*

Figure 2 Minimal T cell epitope, HEL 12(M)–21(R), showing TCR contact residues and L and Y primary anchors for binding to I-A97 at relative positions 6 and 9, and positions 3 and 8 at which specific residues are also not tolerated for binding.

Anchor Residues for Peptide Binding. The critical roles of L17 and Y20 in the HEL epitope, as model anchor residues for binding to I-A^g7, was demonstrated with poly(A) peptides (Table 3). The nonbinding poly(A) peptide, KAA AAAAAA, was converted to a super binder simply by incorporating L and Y at the same relative positions as in the HEL epitope. Either residue alone was not sufficient. Binding was reconstituted only when these two residues were appropriately spaced and in the correct order. In addition, this approach reveals the importance of the frame or context of the anchor residues. The LAAY sequence must be flanked by at least two As, an absence of which at the COOH terminus can be compensated for by at least three As on the NH₂ terminus, but not vice versa. This suggests that binding of these specific residues within the I-A^g7 groove requires stabilization by hydrogen bonding from nonspecific flanking residues, in particular at the NH₂ terminus. For the purpose of further analysis, the relative positions (p) of L and Y in the HEL epitope 12–21 are designated p6 and p9.

Table 3 Binding to I-Ag7 of L- and Y-substituted Poly(A) Peptides

Effect of Multiple Substitutions at p6 and p9. In addition to the selected substitutions at all positions (see Table 2), we investigated the effect on binding of all possible substitutions (except labile cysteine) at p6 or p9. A single residue substitution was classified as well tolerated, weakly tolerated, or nontolerated according to a threshold on its IC₅₀ value: well tolerated, <1,000 nM; weakly tolerated, 1,000– 10,000 nM; nontolerated, 10,000 nM. Although somewhat arbitrary, this classification corresponds to generally accepted notions of good binders, moderate binders, and weak to non-binders. The results, combined with those from Table 2, are presented in Table 4. Optimally, p6 is a large, hydrophobic residue (L, I, M, V), whereas p9 is aromatic and hydrophobic (Y, F) or positively charged (K, R). Most amino acids are not well tolerated at these anchor positions. Additionally, specific amino acids are not tolerated at other positions, namely F and E at p3 and W and Y at p8. This information allowed us to propose and test minimum rules for a motif for I-A^g7 binding peptides. These were that a binder must have the following: (a) two well-tolerated residues or one well-tolerated and one weakly-tolerated residue at anchor positions p6 and p9, (b) no nontolerated residues at positions p3 and p8, and (c) at least two residues flanking p6 and p9, or at least three residues NH₂-terminal of p6.

Table 4 Effects of Amino Acid Substitutions on Binding of HEL 12–22 to I-Ag7

Motifs in Peptides Known or Deduced to Bind I-A^g7. Relatively few peptides containing sequences that might bind to I-A^g7 have been reported in the literature. They include peptides that stimulate I-A^g7–restricted T cells or T cell hybridomas, compete for antigen presentation to T cell hybridomas, induce EAE in Biozzi AB/H mice, or have been eluted from I-A^g7 (listed in Table 5). It should be noted that apart from the present study and that of CarrascoMarin et al. (8), binding has not been determined by direct peptide interaction with purified I-A^g7, but either by elution from I-A^g7, competition with peptides that activate I-A^g7– bearing T cell hybridomas or induction of EAE. The motif we have defined correctly identifies 27/30 (90%) of the published sequences (Table 5). Interestingly, we found that one of these sequences, mouse serum albumin 560–574, that does not contain the motif, did not bind to I-A^g7 (data not shown).

Table 5 Motif in Reported I-Ag7 Binding Peptides

Two groups have suggested putative motifs for peptides that bind to I-A^g7. Reich et al. (7) found that several peptides eluted from I-A^g7 had an acidic residue at the COOH terminus. Their data also indicated that this residue was separated by three from a basic residue. Whereas basic residues are major p9 anchors in our motif, an acidic residue at the COOH terminus is not a uniform feature of other peptides deduced (Table 5) or shown (Table 6) to bind to I-A^g7. However, it is conceivable that in some cases, e.g., OVA 323–339 (see Table 5), a COOH-terminal acidic residue could compensate for a nontolerated residue at p9. Amor et al. (9) described a possible motif shared by encephalitogenic peptides in the Biozzi AB/H (I-A^g7) mouse. It contained hydrophobic (I or L), basic (K, R, or H), a small T cell contact (A or G) and large hydrophobic (L or F) residues within a 6– to 7–amino acid core. They studied the effect of K substitutions on the immunogenicity of phospholipid protein 56–70 (see Table 5), in which they had deduced a core sequence, NVIHAFQ, necessary for the induction of EAE. This sequence contains our motif I (p6) and F (p9). K substitutions at I, H, A, or F completely abolished the ability of the peptide to induce EAE. We would have predicted abolition of binding by the K substitution at I or A and, by analogy with HEL (see Table 2), a significant reduction in T cell activation by K substitution at H or F. Thus, the features of this encephalitogenic motif are contained within the expanded and generalized motif we have described.

Table 6 Overlapping Human Proinsulin Peptides Tested for Binding to I-Ag7

Presence of Motif in Overlapping Peptides from Human Proinsulin. We tested peptides overlapping by four residues and spanning the entire sequence of human proinsulin for binding to I-A^g7, and inspected them for presence of the binding motif (Table 6). All six (100%) good binders contained a motif. However, a motif was present in 4/13 (30%) weak or non-binders. Clearly, the motif rules do not fully account for the effects of residue combinations or flanking sequences. Proinsulin 5–19 has a well-tolerated V at p6 and a weakly tolerated L at p9 yet did not bind, but when this anchor pair moves towards the NH₂ terminus in the following 9–23 sequence, the peptide becomes a binder. Proinsulin 45–59 has a well-tolerated L at p6 and a weakly tolerated L at p9 and binds with high affinity, but when this anchor pair moves towards the NH₂ terminus in the following 49–63 sequence, the affinity of the peptide decreases. Human proinsulin 17–31 has a weakly tolerated Y at p6 and a well-tolerated K at p9, yet does not bind. Although this anchor pair is close to the COOH terminus, this does not preclude other peptides, e.g., human proinsulin 65–79, from binding with high affinity. However, human proinsulin 17–31 has a positively charged p9/COOH terminus, whereas the other binding peptides are generally neutral or tend to be acidic. Reich et al. (7) noted a bias towards acidic residues at the COOH termini of peptides they eluted from I-A^g7. When the anchor pair in this peptide moves towards the NH₂ terminus in the following 21– 35 sequence with an acidic COOH terminus, the peptide becomes a binder. Thus, in a small set of unbiased peptides, the motif appears to have high sensitivity and some degree of specificity. A similar degree of specificity was found for an I-E^k motif by correlating binding with the presence of the motif in a panel of ~150 peptides (23).

Although the core region length of class II MHC binding peptides is ~13 residues (24), analysis of binding motifs (25) indicates that only nine residues within the core region are essential for binding. The motif we describe conforms with this requirement. It is simplistic in the sense that each residue is assumed to contribute to binding independently of other residues and, when located at a given position, to contribute the same amount to binding even within different sequences. The rules that govern binding to class II MHC molecules are more complex and will have to take account of interactions between residues. Nevertheless, by defining tolerated and nontolerated residues for binding at key positions in HEL 12–22, we appear to have unearthed general rules that identify a large majority of known binders to I-A^g7, and discriminate most non-binders.

The high sensitivity of the motif for reported I-A^g7 binders or T cell epitopes is remarkable, but the utility of the motif will depend on its specificity, i.e., its absence in nonbinders. Specificity was 70% for the peptides in Table 6, but the database is small. Even if somewhat degenerate in its present form, the motif should considerably narrow the search for possible binders. The I-A^g7 binding assay we have described is robust and will enable the database of binders and non-binders to be enlarged progressively to further validate the motif. Experiments to fine tune the motif by using peptide libraries are in progress. Just as a motif for human class II DR4 (*0401) binding peptides was applied to scan candidate autoantigen proteins in rheumatoid arthritis for potential epitopes (26), so also might the motif for I-A^g7 binding be applied to identify potential autoepitopes for IDDM in NOD mice and CR-EAE in Biozzi AB/H mice.