The potential of CD4+ T cells for cross-recognition of self and foreign Ags has important implications for the understanding of thymic selection, lymphocyte survival, and the occurrence of autoimmune diseases. Here, we define the extensive flexibility of Ag recognition for three human CD4+ autoreactive T cell clones (TCC) by using ligands with single and multiple amino acid (aa) substitutions. Our results demonstrate that the spectrum of tolerated ligands and the resulting stimulatory potency of peptides for a TCC can be predicted by the relative influence of each aa. Using this approach, we have identified stimulatory ligands not sharing a single aa in corresponding positions with the Ag used to establish the TCC. These results argue for an independent contribution of each aa in the peptide sequence to the affinity of the MHC/peptide complex to the TCR.
During recent years, research in T cell immunology has focused attention on delineating rules for the recognition of Ag by the TCR and the functional consequences of differential TCR ligation (1, 2, 3, 4). Early reports on cross-recognition between foreign and self Ags, based on the assumption of sequence identity or sequence homology, led to the hypothesis of molecular mimicry, which has remained one of the major models for the development of autoimmune diseases (5, 6). However, cross-reactivity occurs not only among largely homologous peptides but also among ligands that share only a few critical residues (7, 8). This observation stimulated an intensive search for rules of cross-recognition, resulting in new models for TCR interaction with MHC/peptide complexes (9). As an extension, the model recently proposed by Allen and co-workers points out the importance of one amino acid (aa)4 as a primary TCR contact, which is strictly necessary for recognition of the antigenic peptide, whereas other aa contacting the TCR modulate the response (1, 10).
When the structures of two class I-restricted TCRs and their complexes with antigenic peptide and MHC were resolved (11, 12), it became clear that the portion of the TCR that contacts the MHC/peptide complex forms a rather flat interface, with only one pocket where the Vα and Vβ chains join each other. Surprisingly, most of the contact surface between TCR and MHC/peptide complex is made up by TCR and MHC molecule directly rather than between the TCR and the peptide. For one of the two TCR, it was found that peptide residue Y5 which contacted the TCR in the pocket could be replaced by the nonconservative substitution A without losing the ability to form stable complexes (13). These biophysical observations indicate that a preexisting affinity between TCR and MHC may allow for at least part of the flexibility in TCR Ag recognition.
To address the question as to how many modifications of the antigenic peptide a TCR can tolerate and to define rules for how multisubstituted peptides are recognized, we studied in detail the responses of three well-characterized human CD4+, HLA class II-restricted autoreactive T cell clones (TCC). The response of the TCC to a large panel of single aa-substituted peptides was determined and compared with the response to multisubstituted peptides.
Materials and Methods
Myelin basic protein peptide 87–99 (MBP (87–99)), VHFFKNIVTPRTP, and derivatives thereof were synthesized using Merrifield’s solid phase methodology as described (14). All peptides have purity of >95%. The one-letter amino acid code is used throughout the article.
T cell clones
TCC were generated from PBL by limiting dilution technique as described (15). Clonality was proven by analysis of the Vα and Vβ TCR chain expression using family-specific primers for RT-PCR as described. The restriction elements are DRB5*0101 for TCC TL3A6, DRB1*0404 for TCC GP52, and DRB1*1302 for TCC GDBP.
T cell proliferation
Cell proliferation was measured by standard [3H]TdR incorporation as described (15). TCC were rested for 8 to 12 days, washed, and resuspended at 1 × 105 cells/ml in complete medium Iscove’s modified Dulbeccols medium containing 5% human serum, 1% penicillin/streptomycin, 0.2% gentamicin (all Whittaker Bioproducts, Gaithersburg, MD)). One hundred microliters of this cell suspension were added to each well of 96-well U-bottom plates containing 5 × 104 irradiated (3000 R) PBL and varying concentrations of peptide. Cells were cultured for 72 h at 37°C. During the last 6 h of culture, 1 μCi of [3H]TdR was added to each well. Cells were then harvested, and incorporated radioactivity was measured by scintillation counting. For the antagonist assays (15), PBL were prepulsed with 25 μg/ml MBP (87–99), washed twice, and then incubated with different concentrations of antagonist peptide.
CTL assays were performed as reported previously (16). Target cells (5 × 105) were labeled overnight at 37°C in 500 μl of CTL medium (RPMI + 5% FCS + 1% glutamine) with 50 μCi of 51Cr (DuPont-New England Nuclear, Boston, MA). 51Cr-labeled cells were incubated with either no Ag or 100 μg/ml peptide for 2 h and washed twice. Targets (2 × 103) were plated into 96-well U-bottom microtiter plates containing 1 × 104 or 1 × 105 T cells (E:T ratios, 5:1 and 50:1). After 4 h of incubation (37°C), supernatants were counted in a gamma counter (ME Plus, ICN Micromedic, Huntsville, AL). Specific lysis was calculated according to the following formula: 100% × [test release (cpm) − spontaneous release (cpm)]/[total incorporation (cpm) − spontaneous release (cpm)].
Epitope mapping using single aa peptide scans
Previously described human CD4+ TCC (15) were tested for their response to a large panel of modified ligands derived from MBP (87–99). First, the TCC were assayed for their proliferative response to a set of single aa-substituted peptides based on their native ligand MBP (87–99). Within the core sequence of the epitope, which was defined by the response to truncated variants (Table I⇓), at least eight substitutions (either nonconservative or conservative) and, for positions outside of the core, at least five substitutions were used for each aa within the MBP (87–99) peptide. The dose-response curve to the set of modified ligands was compared with that obtained with MBP (87–99). As shown for position 92 (N in the native peptide) (Fig. 1⇓), some modifications completely abrogated the response (i.e., F for TCC TL3A6), whereas others decreased the response by 1 (i.e., D for TCC GDBP) or 2 (i.e., Q for TCC GDBP) orders of magnitude. In agreement with our previous studies, we also identified modifications resulting in ligands that were as potent as (i.e., G for TCC TL3A6) or even more potent than (i.e., L for TCC TL3A6) the native ligand (16, 17). Such superagonist peptides elicited a similar functional response at Ag concentrations up to 2 orders of magnitude lower than the native ligand.
The effect of a single aa modification was determined by the relative shift in the dose-response curve of a given TCC and quantified by the concentration required to induce 20% of the maximal response (EC20) to MBP (87–99). Substitutions that shift the dose-response curve to higher Ag concentrations are termed negative modifications (increase of the EC20 by 1 (−1) or 2 (−2) or more (>−2) orders of magnitude), whereas substitutions shifting the dose-response curve to lower concentrations are termed positive (decrease of EC20 by 1 (+1) or 2 (+2) orders of magnitude). Substitutions without a measurable effect are termed neutral (0). The results of experiments involving all modified peptides were used to establish TCC-specific recognition patterns. The effect of each substitution was defined by the relative change in the response to the modified peptide in comparison with the native Ag (Table I⇑). The chemical properties of the modifications were related to the T cell response elicited. In some instances, only conservative substitutions were tolerated. For TCC TL3A6, R in position 97 could be replaced only by the highly conservative aa K, which shares charge and size. The replacement by H, which differs mainly by its aromatic ring, was not tolerated, nor was any other substitution. Similarly, TCC GDBP only tolerated conservative aa L, Y, and W to replace F in position 89. Effects of modifications of P in position 96 were more complex. While TCC TL3A6 responded partially to F, TCC GP52 was stimulated only by A, suggesting that the kink induced by P was important (imitated by the bulky aromatic ring of F) in the first case, whereas the size and the lack of polarity defined the tolerated aa in the latter case. Another example of structural influences on recognition was observed at position 88 for TCC GDBP, which best tolerated those aa that shared an aromatic ring (F, Y). However, in several instances, we did not observe a relationship between the chemical characteristics of aa side chains and the influence on peptide recognition. As shown for TCC GDBP in position 95, K induced a stronger response than the native aa T; N a similar response; E, I, and S a reduced response; and G and A an even weaker response (Table I⇑). For TCC GP52, the chemically unrelated aa F and A were tolerated in position 91 as replacement for K, whereas the most conservative aa R was not (Table I⇑). These findings demonstrate that the chemical properties and the functional response relate to each other in some positions (18), but not in others. With respect to the location relative to the epitope core, although positions outside of the core can negatively or positively affect recognition, the overall effects are minor and fade with increasing distance from the core (Table I⇑). The number of tolerated aa was lower in core positions, although many alterations resulted in productive ligands (13 substitutions of 85 tolerated by TCC TL3A6, 28 of 84 by TCC GDBP, and 24 of 85 by TCC GP52) (Table I⇑).
T cell responses to multisubstituted peptides are predictable
Next, we addressed how combinations of multiple aa modifications within one peptide would affect recognition. The combination of two positive alterations resulted in an even stronger ligand for TCC TL3A6 (Fig. 2⇓ first panel), whereas the combination of a negative and a positive alteration resulted in a peptide with the same potency as MBP (87–99) (Fig. 2⇓, second panel). Peptides with aa modifications that led to altered functional T cell responses such as antagonism or partial activation (15) were turned into stimulatory ligands when multiple positive modifications were introduced (Fig. 2⇓, third and fourth panels). Interestingly, for TCC TL3A6, ligands that elicited partial agonist responses (anergy and IL-2R up-regulation in the absence of proliferation) turned into stronger ligands than peptides with mixed partial agonist (anergy induction in the absence of proliferation)/antagonist properties (Fig. 2⇓; Table II⇓) or even pure antagonist properties (Table II⇓) after introduction of multiple superagonist modifications. These findings support the concept of a hierarchy in the stimulatory capacity that groups antagonists below partial agonists and full agonists. However, it also indicates that the induction of antagonism or partial agonism relates to the affinity of the TCR for the entire MHC/peptide complex (19) rather than being a feature of modifications at particular positions of the ligand (18, 20).
We then extended these studies to a set of 16 multisubstituted peptides (Table II⇑). The proliferative response was determined and the EC20 defined for each peptide. The sum of the relative changes induced by multiple substitutions was compared with the effect of single aa modifications (Table II⇑). With one exception, the response to single aa-modified peptides permitted the prediction of the relative changes in the response to a multisubstituted peptide with remarkable accuracy (within the range of 1 order of magnitude).
Identification of agonist ligands with no sequence homology to MBP (87–99)
Following these observations and on the basis of the response of TCC GP52 to single aa mutations, we designed a peptide that differed from the MBP (87–99) peptide in all 13 positions. The final peptide consisted of four positive substitutions (L89, L90, A96, K98), three neutral substitutions (G87, G88, A99), four substitutions that induced a slight decrease in the response (H92, V93, I94, S95), and two aa that strongly decreased the response (A91, K97). The resulting peptide induced proliferation at concentrations 2 to 3 orders of magnitude higher than that of MBP (87–99), within a range predicted by adding the effects of single aa modifications (Fig. 3⇓A). The recognition was confirmed by CTL assays, which showed that the TCC lysed MHC-matched target cells pulsed with the nonhomologous peptide, but not the mismatched targets loaded with the same peptide (Fig. 3⇓B). This proved that the recognition of the peptide was dependent on MHC presentation, excluding unspecific mechanisms of T cell activation by the peptide. Similar results were obtained for TCC TL3A6. On the basis of results obtained with single amino acid modifications (Table I⇑) and a decapeptide combinatorial library in the positional scanning format (16) (B. Hemmer, manuscript in preparation), we identified a peptide ligand (WYALLPSCKG) that stimulated the TCC at high Ag concentration (EC20 2 orders of magnitude higher than with the MBP peptide) (Fig. 3⇓C). This ligand incorporated three neutral aa modifications (W89, Y90, G98), three that increased (L92, P94, K97), and four that decreased the response (A91, L93, S95, C96).
Several important conclusions can be drawn from these observations: 1) TCR recognition is highly degenerate; 2) the chemical properties of aa side chains in the peptide sequence do not always allow predictions on tolerated aa substitutions; 3) similar to what has been shown for peptide/MHC interactions (21), TCR Ag recognition of peptides mutated at multiple sites can be predicted on the basis of the effects of single aa mutations. Although exceptions are found in our (Table II⇑) and other experimental systems (18, 22), this strategy allows the reliable identification of multisubstituted cross-reacting Ags; 4) as a result of this approach, stimulatory peptides can be identified, with aa sequences entirely different from that of the Ag used to establish and propagate the TCC.
Considering the large contact surface between TCR and MHC demonstrated by the crystal structure (11, 12) and the independent contribution of each aa to the stimulatory capacity of the peptide, we propose a model for the interaction of the TCR with the MHC/peptide complex. The TCR itself has a defined affinity to the MHC molecule that may even be sufficient for T cell activation as demonstrated in some models of peptide-independent allorecognition (23). In most cases, however, the interaction requires the contribution of the antigenic peptide. The additive contribution (positive, negative, or neutral) of each aa in the sequence to (1) peptide binding to the MHC (21) and (2) affinity of the complex for the TCR is indeed consistent with the linear structure of the MHC-bound antigenic peptide. Therefore, each residue of the peptide can influence ligand density on the APC (by MHC binding) and the overall affinity of the MHC/peptide complex to the TCR. Although aa in MHC binding position contribute more to MHC affinity and less to TCR affinity, whereas aa in TCR contact positions add more to TCR affinity than to MHC binding, each aa contributes to both interactions, thus ultimately determining the number of engaged TCR molecules and the functional activation of the T cell (24, 25). This integration model for TCR Ag recognition is strongly supported by recent reports, which demonstrate that a set of combinatorial peptide libraries each with only one defined aa allow the identification of high potency ligands for these (B. Hemmer, C. Pinilla, B. Gran, N. Ling, P. Conlon, H. F. McFarland, R. Houghten, and R. Martin, manuscript in preparation) and other TCC (16, 26). Although certain aa in the peptide sequence seem to be more important than others, none of them is strictly required for T cell recognition. Consistent with observations regarding other protein-protein interactions (27), we propose for the interaction of the TCR with its ligand that the combination of positive and negative effects of individual aa in the antigenic peptide determines whether the resulting affinity of the MHC/peptide ligand for the TCR is high enough to trigger TCR-dependent signaling events (16).
This flexibility in TCR Ag recognition may explain recent findings in thymic selection that (1) unrelated peptides can select the same TCR (28) and (2) the same MHC/peptide complex can select T cells with very different Ag specificities (29).
With respect to molecular mimicry, our results clarify why a search algorithm for mimicry peptides that incorporated only few aa in specific MHC and TCR contact positions identified only a small number of cross-reactive peptides and why the predictive value of this approach was limited (8). The concept of T cell recognition discussed here argues for extensive degeneracy in TCR-MHC/peptide interactions and seems to question the exquisite specificity of the cellular immune response. Although many ligands are indeed recognized by a specific TCR, it should be noted that under physiologic conditions only the few best fitting MHC/peptide complexes will fully activate a given T cell. During natural immune responses such as those to viral infections, only these high affinity interactions will lead to expansion of T cells. These observations not only extend previous models for TCR Ag recognition but also provide new directions for examining cross-reactivity between Ags during autoimmune responses.
We thank W. Biddison and A. Tzou, Neuroimmunology Branch, National Institute of Neurological Diseases and Stroke, National Institutes of Health, for discussion and critical comments and T. Wong, L. Tranquill, G. Afshar, and H. Maloni for technical support.
↵1 B.H. was supported in part by a postdoctoral fellowship of the Deutsche Forschungsgemeinschaft (He 2386/1-2) and by a Fogarty Fellowship. This work was supported in part by a Small Business Innovation Research grant from the Department of Health and Human Services, Public Health Service.
↵2 B.H. and M.V. contributed equally to the study.
↵3 Address correspondence and reprint requests to Dr. Roland Martin, Neuroimmunology Branch, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bldg. 10 Room 5B16, 10 Center Drive MSC 1400, Bethesda, MD 20892-1400. E-mail address:
↵4 Abbreviations used in this paper: aa, amino acid; TCC, T cell clone; MBP, myelin basic protein peptide 87–99; EC20, concentration required to induce 20% of the maximal response.
- Received October 31, 1997.
- Accepted February 11, 1998.
- Copyright © 1998 by The American Association of Immunologists