Structural Basis of Specificity and Degeneracy of T Cell Recognition: Pluriallelic Restriction of T Cell Responses to a Peptide Antigen Involves Both Specific and Promiscuous Interactions Between the T Cell Receptor, Peptide, and HLA-DR¹

The ligand recognized by TCR on CD4⁺ T cells consists of a class II MHC molecule complexed with a peptide fragment of a protein Ag on the surface of an APC. The structure of MHC molecules imposes constraints on the nature of peptides that they can bind, and therefore the selection of antigenic determinants that can be presented to T cells (1). Each MHC class II molecule can bind and present an extremely large number of structurally diverse peptides based on motif residues which are anchored by complementary pockets within the MHC molecule (2, 3, 4, 5, 6, 7). Similarly, a given peptide can be bound and presented by many class II molecules with varying efficiencies for recognition by T cells (2, 7, 8).

The interaction between αβ TCRs and peptide-MHC class II ligands involves direct contact between the TCR and residues on both the MHC molecule and the bound peptide (9, 10). Analyses of the crystal structures of MHC class II molecules complexed with antigenic peptides (11, 12, 13, 14, 15) and studies using substituted peptides and mutated MHC molecules (6, 16, 17) have identified putative TCR contact residues on both the antigenic peptides and the α-helical regions of the MHC class II molecules. Two recent crystal structures of class I MHC-peptide-TCR complexes (18, 19, 20) have demonstrated that the αβ TCR recognizes its ligand through three variable loops (complementarity-determining regions (CDRs⁴)) on each polypeptide that form a relatively flat surface that fits diagonally over the MHC-peptide complex. All three CDRs of the TCR α-chain and CDR3 of the β-chain contact the MHC α helices, while the CDR1 and CDR3 loops of both the TCR α- and β-chains contact the bound peptide. Based on sequence and structure comparisons between peptides complexed with MHC class I and class II molecules, a similar mode of TCR interaction with peptide-MHC class II ligands has been predicted (19, 21).

While the overall conformation of the peptide-MHC-TCR ternary complex is thought to be similar for all MHC class I and class II molecules and all αβ TCRs (19, 21), the nature of the signal generated upon TCR ligation can vary greatly, leading to activation, apoptosis, or nonresponsiveness of the T cell and positive or negative selection in the thymus (22, 23). TCR activation can, in turn, result in a range of different effector functions, including T cell proliferation, cytolysis, and cytokine secretion of the Th0, Th1, or Th2 profiles. Factors that contribute to the outcome of TCR ligation include the nature and cell surface density of the peptide-MHC ligand, the avidity of its interaction with the TCR, the presence or absence of other costimulatory molecules, and the local cytokine milieu (22, 24, 25, 26). Specific contacts between amino acid residues on the peptide, MHC, and TCR molecules have profound effects on the nature of T cell stimulation. Single substitutions of MHC or peptide residues predicted to contact the TCR can enhance, antagonize, or anergize a T cell response or selectively inhibit some but not all effector functions (27, 28, 29, 30, 31, 32).

The TCRs of human CD4⁺ T cells primarily recognize antigenic peptides presented by MHC class II (HLA-DR, DQ, or DP) molecules. HLA-DR molecules are heterodimeric glycoproteins composed of a nonpolymorphic DRα and a polymorphic DRβ chain. The polymorphic amino acid residues in DRβ polypeptides are mostly clustered in three hypervariable regions (HVRs) which line the peptide binding groove (33). HVR1 (residues 9–14) and HVR2 (residues 26–33) are located on the floor of the groove and predominantly influence the specificity of peptide binding, while HVR3 (residues 67–74) is located on the α-helical portion of the DRβ molecule and can affect both peptide binding and TCR interaction through direct contacts with both molecules (11, 13, 14). HVR3 amino acid sequences can vary greatly within serologically related DR types but are conserved among serologically distinct alleles. For example, the ILEDE amino acid sequence at DRβ positions 67–71 is present on some but not all allelic variants of the DR1, DR4, DR11, and DR13 specificities. Another sequence at this position, LLEQRRAA, which is encoded by certain DR1, DR4, and DR14 alleles, constitutes the “shared epitope” thought to play a role in the development of rheumatoid arthritis (RA) (34, 35).

In this study, we examined the molecular basis of the interactions between the TCR, MHC class II, and peptide molecules, using a herpes simplex type 2 virus (HSV-2)-specific human CD4⁺ T cell clone, ESL4.34, which recognizes a peptide determinant corresponding to residues 393–405 of the virion protein VP16, in the context of several HLA-DR alleles. This T cell, which elicits strong proliferative, cytolytic, and cytokine secretion responses, recognizes the VP16 393–405 peptide with a pattern of restriction that correlates with a specific amino acid sequence at positions 67, 70, and 71 of HVR3 of the DRβ polypeptide. By a combination of molecular modeling and assays of peptide binding and T cell stimulation using in vitro mutagenized DR molecules and substituted peptide analogues, we have dissected the structural interactions between DRβ HVR3, the peptide, and the TCR. The DRβ HVR3 was found to have a direct effect on both peptide binding and TCR recognition, suggesting an important role for this epitope in alloreactivity, in T cell repertoire selection, and in autoimmune disease.

HSV-1 strain E115, HSV-2 strain 333, and a recombinant virus consisting of VP16 of HSV-2 in an HSV-1 background, RP-2, were grown in human diploid cells and titered by plaque assay on Vero cells as previously described (36). Crude viral Ags were prepared from virus stocks by exposure to UV light for 30 min at 10 cm from a GT038 bulb (General Electric, Cleveland, OH), eliminating infectious virus, and used at a final dilution of 1:100 which corresponds to 10⁶ to 10⁷ plaque-forming units per ml before UV treatment. Peptides corresponding to amino acids 1–409 of VP16 of HSV-2, 13 amino acids long and overlapping by 9 amino acids, were obtained from Chiron Mimotopes, Clayton, Australia. The VP16 393–405 peptide (LVAPRMSFLSAGQ) and its analogues (see legend to figures and tables) were synthesized with an Applied Biosystems 432 Peptide Synthesizer (Foster City, CA). Peptides were biotinylated by amino-terminal addition of biotin onto ε-aminocaproic acid, as described by Kwok et al. (37), and used at the concentrations shown in the figure legends.

Mononuclear cells were prepared from the herpetic vesicle fluid from a DRB1*0402/*1301-positive patient with culture-proved recurrent HSV-2 infection by Ficoll-Hypaque density gradient centrifugation. The cells were stimulated with PHA (0.4 μg/ml PHA-P; Murex Diagnostics, Dartford, U.K.) and an equal number of irradiated (3300 rads γ-irradiation) allogeneic PBMC in T cell medium (RPMI 1640 containing 25 μM HEPES, 2 μM l-glutamine, 50 μg/ml streptomycin, 50 U/ml penicillin, and 10% heat-inactivated human male serum). Acyclovir (50 μM) was added for the first 2 weeks of culture to prevent viral replication (38). Human natural IL-2 (50 U/ml; Schiaperelli Biosystems, Columbia, MD) was added on day 3, and the cells were subsequently fed with IL-2 every 2 to 3 days. After 16 days of growth, 5 × 10⁵ cells were restimulated with crude HSV-2 Ags and 5 × 10⁵ autologous irradiated PBMC as APC and maintained as above. After 12 days, the cells were cloned at 1 cell/well using PHA, IL-2, and allogeneic feeders and subsequently restimulated every 12 to 14 days.

mRNA was prepared from T cell clones that were restimulated with lymphoblastoid cell lines (LCLs) as APC instead of PBMC using the QuickPrep Micro RNA Purification Kit (Pharmacia Biotech, Alameda, CA) and used for first strand cDNA synthesis using the Superscript Preamplification system (Life Technologies, Gaithersburg, MD). TCR Vα and Vβ gene segment usage was determined by PCR amplification of first strand cDNA (RT-PCR) using a set of subfamily specific 5′-primers (Vα1–29 and Vβ1–23) and constant region 3′-primers (kindly provided by Dr. Christine Vissinga, Virginia Mason Research Center, Seattle, WA) essentially as described by Choi et al. (39). Vβ typing was confirmed using the Clontech (Palo Alto, CA) TCR Amplimer Kit. Nucleotide sequences of the V-(D)-J junctions were determined by cycle sequencing of both strands of the PCR products obtained as above using the Prism Ready Reaction DyeDeoxy Terminator Cycle Sequencing Kit (Applied Biosystems) and analyzed on an Applied Biosystems 373 DNA Sequencer. Two different Vα7 genes were detected by sequence analysis; therefore, the Vα7 PCR products were cloned into the pCRII vector (Invitrogen, San Diego, CA), used to transform One Shot (INVαF′) cells (Invitrogen) and selected on antibiotic-containing L-agar plates. Plasmid DNA was prepared from overnight cultures of isolated colonies using the QIAprep Spin Miniprep Kit (Qiagen, Santa Clarita, CA) and amplified as above by PCR. The nucleotide sequences of both strands of the PCR products were determined by cycle sequencing as above.

The EBV-transformed LCLs YAR (DRB1*0402), MT (*0404), HHKB (*1301), JVM (*1102), TISI (*1103), and MAT (*0301) were obtained from the VIIIth and Xth International Histocompatibility Workshop Panels. The LCL 8854 (*1303) was generated by EBV transformation of locally characterized PBMC. The MHC class II-deficient BLS-1 line was kindly donated by Dr. Janet Lee, Sloan-Kettering Memorial, New York, NY. LCLs were cultured in RPMI 1640 supplemented with 10% heat-inactivated FCS (HyClone, Logan, UT) and sodium pyruvate, HEPES, l-glutamine, streptomycin, and penicillin as above. mAbs used were L243 (anti-DR; American Type Culture Collection, Rockville, MD), SPV-L3 (anti-DQ; kindly provided by Dr. Hans Yssel, DNAX, Palo Alto, CA), B7/21 (anti-DP; Becton Dickinson, San Jose, CA), anti-αβ TCR and anti-Vβ8 (PharMingen, San Diego, CA), NFLD.D1, and NFLD.D11 (anti-DR4 and -DR4w4; kindly provided by Dr. Sheila Drover, University of St. Johns, St. Johns, Canada). Cell surface expression of HLA-DR and TCR was assayed by mAb staining as described previously (40) and analyzed on a Becton Dickinson FACSort flow cytometer.

HLA-DR genes and mutagenized constructs were introduced into MAT LCLs by retroviral infection using the methods of Kwok et al. (41). The construction of the retroviral vectors containing wild-type and mutagenized DRB1*0404 genes which were altered at codons 67 (L to I), 70 (Q to D), 71 (R to E), and both 67 and 71 (L67 to I67 and R71 to E71) has been described previously (40). For the purpose of clarity, these mutants are named according to their differences from the DRB1*0402 amino acid sequence in this report, i.e., DRB1*0402-Q70,R71, DRB1*0402-L67,R71, DRB1*0402-L67,Q70, and DRB1*0402-Q70, respectively. Surface expression of the transgenes was monitored by flow cytometry using the NFLD.D1 and NFLD.D11 mAbs.

Stimulator cells (PBMC irradiated with 3,300 rads or LCL irradiated with 20,000 rads γ-irradiation) were pulsed with Ag (peptide concentrations shown in figure legends) for 2 to 4 h at 37°C, 5% CO₂, followed by three washings with supplemented RPMI medium. PBMC (10⁵) or LCL (2.5 × 10⁴) were plated in 96-well U-bottom plates in T cell medium containing 10⁴ cloned T cells or bulk lesion-derived cells as responders in a final volume of 150 μl. All assays were performed in triplicate. After incubation for 48 h, [³H]thymidine (1 μCi/well; NEN, Boston, MA) was added for 12 to 16 h, the cells were harvested using a Tomtec 96 Mach III Harvester (Hamden, CT), and ³H incorporation was measured on a 1450 Microbeta Plus Liquid Scintillation Counter (Wallac, Gaithersburg, MD). Stimulation indices were calculated as (cpm of sample − cpm in absence of responders)/cpm in absence of stimulators. To determine restricting HLA molecules, mAbs L243, SPV-L3, and B7/21, which recognize HLA-DR, DQ, and DP, respectively, were used as previously described (38).

Target LCLs (5 × 10⁵) were pulsed with peptide (concentrations shown in figure legends) and simultaneously loaded with ⁵¹Cr (25 μCi; NEN) for 8 to 12 h at 37°C, 5% CO₂, followed by gentle washing three times with supplemented RPMI medium. Labeled targets (1–2 × 10³) were incubated with effector T cells at various E:T ratios in a final volume of 150 ml in T cell medium in 96-well U-bottom plates. After 4 h, cell supernatants (50 μl) were assayed for ⁵¹Cr release in a 1450 Microbeta Plus Liquid Scintillation Counter. Percent specific lysis was expressed as (cpm of sample − cpm of spontaneous release)/(cpm of maximum release − cpm of spontaneous release).

Binding of biotinylated peptides to HLA-DR on whole cells, with subsequent capture of the peptide-class II complex, was performed using a europium-labeled streptavidin assay as described previously (37), except that the L243 Ab was used to immobilize the HLA-DR molecules. A 10 μM peptide concentration was used in all experiments, this concentration having been previously determined to be submaximal for binding. The class II-deficient LCL, BLS-1, served as a negative control for nonspecific binding.

Molecular models of DRB1*0402, *1102, and *1301 were constructed from the crystal structure coordinates of the HLA-DR1 molecule complexed with the influenza hemagglutinin peptide 306–318 (11) as previously described (42). Using the interactive graphics program PSSHOW (43), amino acid side chains were replaced and assigned to the preferred conformer from the Ponder and Richards rotamer database (44) that was closest to the side chain conformer in the crystal structure template. The HSV VP16 393–405 peptide was constructed from the hemagglutinin peptide coordinates (11), and several different alignments of the anchor residues with the VP16 393–405 sequence were examined to predict the binding motif. Peptide binding experiments were designed to test the predicted binding motif.

Recovery of cells from herpetic vesicle fluid and subsequent stimulation with whole HSV-2 Ag followed by cloning yielded a T cell clone, ESL4.34, that responded to HSV-2 and RP-2 but not HSV-1 in proliferation assays (Fig. 1), indicating a specificity for VP16 of HSV-2 but not HSV-1. The fine specificity of ESL4.34 was determined by testing its proliferative response to a set of 10 pools of overlapping peptides corresponding to amino acids 1–409 of VP16 of HSV-2. Only a pool spanning amino acids 361–409 (pool 10) gave a positive response (Fig. 1). Assay with individual peptides of this pool identified the epitope as one contained in overlapping peptides 389–401 and 393–405 (Fig. 1). Peptide VP16 393–405 was used in subsequent studies to characterize the residues required for MHC binding and T cell recognition.

The HLA locus that serves as the restriction element for ESL4.34 was determined in proliferation assays using HSV-2 and VP16 393–405, presented by autologous PBMC, in the presence and absence of mAbs to HLA-DR, -DQ, and -DP. Inhibition of proliferation was found in the presence of the anti-DR mAb only (data not shown), indicating that ESL4.34 is DR restricted. Oligonucleotide typing of autologous PBMC revealed positivity for DRB1*0402 and DRB1*1301. To determine which allele is the restriction element for ESL4.34 recognition of VP16 393–405, DR-homozygous LCLs expressing DRB1*0402 (YAR) and DRB1*1301 (HHKB) were tested as APCs for the peptide. Figure 2 shows that both DRB1*0402 and DRB1*1301 presented VP16 393–405 to ESL4.34, eliciting both proliferative and cytolytic responses.

ESL4.34 TCR α- and β-chain gene usage was determined by PCR amplification of cDNA using primers specific for Vα1–29 and Vβ1–23. PCR products corresponding to Vα7 and Vβ8 only were obtained. Flow cytometric analysis of ESL4.34 using anti-αβ TCR and anti-Vβ8 mAbs confirmed that all αβ TCR⁺ cells expressed Vβ8. Nucleotide sequence analysis of the Vβ8 PCR product in both directions revealed a unique V-d-J junctional sequence that encodes a TCR β-chain CDR3 (residues 92–117) amino acid sequence CASSERGDTDTQYFGPGTRLTVLEDL, which corresponds to Vβ8, Dβ (ERGD), Jβ2.3 (TDTQYFGPGTRLTVLEDL), and Cβ2. Sequencing of the Vα7 PCR product revealed two Vα7 TCR gene sequences, and cloning and sequencing of these two products identified distinct Vα7Jα junctional sequences. One of these contained a G to C nucleotide substitution which, if translated, would result in the substitution of the conserved cysteine residue at position 90 with a serine residue, resulting in a nonfunctional TCR α-chain (45). The viable TCR α-chain CDR3 region was encoded by Vα7.2, Jβ2.3, and a V-J junctional sequence (residues 90–106), which code for the amino acid sequence CAPRGAGRRALTFGSGT.

The above experiments indicate that ESL4.34 is a Vα7⁺Vβ8⁺ T cell clone specific for VP16 393–405 presented by both autologous DR alleles, DRB1*0402 and DRB1*1301, in both proliferation and cytotoxicity assays (Fig. 2). This clone, however, does not respond to the peptide presented by the allogeneic DRB1*0404 allele, which differs from *0402 only at three amino acid positions (residues 67, 70, 71) in the DRβ α helix (Fig. 2). To further investigate the pluriallelic restriction of ESL4.34, a set of LCLs expressing DRB1 allelic variants of three serologically defined DR types, DR4, DR11, and DR13, which differ in their α-helical amino acid sequences, were used as APCs for VP16 393–405 in proliferation and cytotoxicity assays. Figure 3, A and B, shows that VP16 393–405 was recognized by ESL4.34 in proliferation and cytotoxicity assays, when presented on LCL expressing DRB1*0402, DRB1*1301, and DRB1*1102. Each of these DRB1 alleles encode the same sequence at DRβ residues 67–71 (HVR3), namely ILEDE, whereas the other DR4, DR13, and DR11 alleles that did not support ESL4.34 responses had different HVR3 sequences, as summarized in Table I. Even the single isoleucine to phenylalanine substitution at DRβ position 67, that distinguishes DRB1*1102 and DRB1*1103, is sufficient to abolish ESL4.34 responses. Further analyses of ESL4.34 responses VP16 393–405-pulsed LCLs expressing other DRβ 67–71 ILEDE-negative DR4 (DRB1*0401 and *0405) and DR11 (DRB1*1101) alleles (data not shown) were consistent with the hypothesis that this motif at HVR3 is required for ELS4.34 recognition.

The qualitative differences in the proliferative and cytolytic responses of ESL4.34 to VP16 393–405 restricted by the different DRB1 alleles might be explained either by differences in the avidity of the TCR for the peptide-MHC complexes or by structural interactions between the peptide and MHC molecule alone. Figure 3,C shows that the relative binding of VP16 393–405 to DR expressed on the various APCs varied greatly for the different DRB1 alleles and showed no correlation with the proliferative and cytolytic responses of ESL4.34 to VP16 393–405. While DRB1*1303 and DRB1*1103 bound the peptide as well as, or better than, their stimulatory DRB1*1301 and DRB1*1102 counterparts, no proliferative or cytolytic responses were observed using these ligands. This suggests that the ability of ESL4.34 to respond to its peptide-MHC ligand is controlled in part by the presenting MHC molecule itself, through specific interactions between residues on the TCR and the DR molecule. This hypothesis, however, does not exclude the involvement of the peptide in ESL4.34 responses, since no proliferation or cytotoxicity was observed in the absence of VP16 393–405, and the responses increased proportionately with peptide concentration (Fig. 2).

The three DRB1 alleles (DRB1*0402, DRB1*1102, and DRB1*1301) that could present VP16 393–405 to ESL4.34 in proliferation and cytotoxicity assays are serologically and structurally distinct but share a common amino acid sequence, ILEDE, at DRβ positions 67 to 71, whereas the other DRB1 alleles tested in the present study encode different amino acid sequences at this region. This suggests a role for this “shared epitope” in association with VP16 393–405 in eliciting ESL4.34 responses. To test this hypothesis, we generated a panel of LCLs expressing mutagenized DRB1 molecules that distinguish the stimulatory DRB1*0402 allele from the nonstimulatory DRB1*0404 molecule, which differ only at the three polymorphic residues in this epitope, residues 67, 70, and 71. Surface expression of these mutant DR4 molecules was comparable with that of wild-type DRB1*0402 and DRB1*0404 in YAR and MT LCLs, respectively (Fig. 4). Figure 5 shows the proliferative and cytolytic responses of ESL4.34 to VP16 393–405 presented by these mutant APCs and the relative binding affinities of this peptide for the mutant DR molecules. Binding was drastically reduced by substitutions at 67 and 71 (DRB1*0402-L67, R71) and substitutions at 70 and 71 (DRB1*0402-Q70, R71), but changes at residues 67 and 70 (*0402-L67,Q70, and *0402-Q70) had less significant effects on VP16 393–405 binding (Fig. 5,C). In contrast, substitution of any of the DRB1*0402 residues 67, 70, or 71 completely abrogated ESL4.34 proliferative and cytolytic responses to VP16 393–405. These data, together with the responses of ESL4.34 to the DR4, DR11, and DR13 allelic variants (Fig. 3), confirm that DRβ residues I67, D70, and E71 are essential (and possibly sufficient) for presentation of VP16 393–405 to ESL4.34. The involvement of HVR3 in peptide binding and ESL4.34 responses does not exclude less significant contributions by the first and second HVRs of DRβ and/or the V/G dimorphism at position 86. Significant influences of the DRB3 and DRB4 gene products, which are encoded on DR11, DR13, and DR4 haplotypes, however, are unlikely, because these molecules have distinct HVR3 sequences.

Molecular modeling of the VP16 peptide 393–405 complexed with DRB1*0402, DRB1*1102, and DRB1*1103 suggested a binding motif such that V394 occupies pocket 1 and R397 occupies pocket 4. This predicted binding motif was tested by synthesizing peptides with nonconservative substitutions at these anchor positions and measuring their relative binding to LCLs expressing DRB1*0402, DRB1*0404, DRB1*1301, DRB1*1303, DRB1*1102, and DRB1*1103 (Fig. 6). Substitution of valine at position 394 with lysine (P1 [V→K]) dramatically decreased the binding of this peptide to DRB1*0402, DRB1*1102, and DRB1*1103, whereas replacing leucine at position 393 with lysine (P-1 [L→K]) slightly increased the level of binding to these alleles (Fig. 6). This indicates that V394 is the first anchor residue. The increase in binding observed for L393K (P-1 [L→K]) may be due to favorable interactions between K393 and DRα E53. Substitution of arginine at VP16 393–405 position 397 with the negatively charged residue glutamic acid (P4 [R→E]) reduced peptide binding to the DRβ alleles containing negatively charged residues at positions 70 or 71, whereas binding to DRB1*0404 which has arginine at DRβ position 71 was increased by this substitution (Fig. 6). This clearly demonstrates the importance of the negatively charged DRβ residues D70 and E71 of DRB1*0402, DRB1*1301, DRB1*1102, and DRB1*1103 in influencing the nature of pocket 4. Substitution of this P4 residue with a nonpolar alanine (P4 [R→A]) resulted in a less dramatic reduction in binding to DRB1*0402, and surprisingly, an increase in binding to DRB1*0404 compared with that of the P4 (R→E) analogue peptide. The minimum binding epitope of VP16 393–405 was tested using truncated analogues with two, three, or four residues removed from the C terminus. The VP16 peptides 393–401 and 393–402 did not bind or bound very poorly to DRB1*0402 (data not shown), but VP16 393–403 did bind although with a lower affinity than the wild-type VP16 393–405 peptide (Fig. 6). These observations suggest that the two C-terminal residues of VP16 393–405 are not anchor residues but that they may further stabilize the association by interacting with residues on the DR molecule.

To further analyze this binding motif, VP16 393–405 analogues with amino acid substitutions of predicted TCR contact residues were synthesized and tested for their capacity to bind to DRB1*0402 and to stimulate ESL4.34 in proliferation and cytolysis assays. Table II compares the DRB1*0402-binding and ESL4.34-stimulatory capacities of the minimum VP16-binding epitope VP16 393–403 and two analogues of this peptide with substitutions at the predicted TCR contact positions 5 (P5 [M→NL]) and 7 (P7 [F→I]) of the truncated peptide. Although VP16 393–403 bound to DRB1*0402 more weakly than the full length VP16 393–405 peptide, the proliferation and cytolytic responses of ESL4.34 to the two peptides were similar, indicating that the two C-terminal residues of VP16 393–405 are not involved in T cell responses. The binding of VP16 393–403 analogues with a phenylalanine to isoleucine substitution at position 400 (P7 [F→I]), or a methionine to norleucine substitution at position 398 (P5 [M→NL]), was better or slightly reduced compared with that of the VP16 393–403 peptide. However, the responses of ESL4.34 to these APCs were greatly diminished or completely abrogated (Table II), indicating that these peptide residues are TCR contact residues that can modulate T cell recognition while playing only a minor role in binding to DR.

Despite the high degree of specificity of ligand recognition by TCRs, several reports have described T cells that can recognize and respond to multiple peptide-MHC complexes. Such multiple ligands can be provided by the presentation of multiple peptides by a single MHC molecule (46), presentation of a single peptide by multiple MHC molecules (47, 48), or ligands that are distinct in both the peptide and MHC portions (21, 49). TCR recognition involves specific molecular contacts with residues on both the MHC molecule and the bound antigenic peptide. Although the overall conformation of the peptide-MHC-TCR ternary complex is thought to be similar for all MHC and αβ TCR molecules (19, 21), the relative contributions of peptide and MHC to T cell specificity can vary greatly. We therefore have investigated the structural interactions between the TCR, MHC, and peptide for a CD4⁺ T cell clone, ESL4.34, which recognizes the VP16 393–405 peptide epitope of HSV-2 in the context of multiple distinct HLA-DR alleles. This clone, which was isolated from a DRB1*0402/*1301 heterozygous individual, responds to the peptide presented by both parental DR alleles, as well as the allogeneic DR11 allele, DRB1*1102, in proliferation and cytotoxicity assays. These three alleles are serologically and structurally distinct, but they share a common amino acid sequence motif, ILEDE, at positions 67–71 of the DRβ polypeptide. The role of this sequence motif in supporting ESL4.34 stimulation was confirmed by testing an extended panel of LCLs expressing DR4, DR13, and DR11 variants which were negative for this sequence and by site-directed mutagenesis experiments in which DRB1*0404-specific substitutions were introduced into the DRB1*0402 gene at codons 67, 70, and 71. We found that proliferative and cytolytic responses of ESL4.34 to VP16 393–405 exhibited an absolute dependence on the presence of DRβ I67, D70, and E71 (summarized in Table I). This restriction of T cell recognition appears to be due to a direct interaction between these residues and the TCR, since no correlation was found between the relative binding affinities of VP16 393–405 for HLA-DR and ESL4.34 reactivity, and some DRβ 67–71 ILEDE-negative alleles strongly bound the peptide but failed to stimulate the T cells. Thus, amino acid substitutions that had little or no effect on relative peptide binding, such as the single I67 to F67 substitution that converts DRB1*1102 to *1103; mutations that change DRB1*0402 to DRB1*0402-Q70 and DRB1*0402-L67,Q70; and substitutions that distinguish DRB1*1301 and DRB1*1303, totally abolished ESL4.34 recognition. This suggests that the responses of ESL4.34 are solely directed against HVR3 of the DRβ molecule, similar to allospecific responses, but our observations that no T cell recognition occurred in the absence of VP16 393–405 and the responses increased proportionately with peptide concentration (Fig. 2) confirm the Ag-specific nature of ESL4.34 stimulation.

Numerous studies using allelic variants of class II and in vitro mutagenized class II molecules with substituted peptide analogues or using sequence analysis of naturally processed peptides presented by class II, have identified residues on MHC molecules that are critical for peptide binding (2, 5, 6, 7) and T cell activation (9, 16, 17, 50) and have shown that TCR recognition is dependent both on residues necessary for peptide binding and residues with little or no effect on binding. Our results indicate that DRβ E71 is critical for the binding of VP16 393–405, since substitution of this residue with R in DR4 variants and in DRB1*0402 mutants (DRB1*0402-Q70,R71, and DRB1*0402-L67,R71) drastically reduced binding of the peptide. Less significant reductions in relative peptide binding were associated with substitutions of DRβ D70, while changes at position 67 alone had little or no effect on binding. This hierarchy of peptide binding dependence in the order E71 > D70 > I67 is consistent with the location of these residues in the crystal structures of HLA-DR molecules, which place residues 67, 70, and 71 at the highest point of the DRβ α helix, with residue 71 pointing “up” toward the TCR and into the peptide binding cleft interacting with peptide position 4, residue 70 pointing up, and residue 67 pointing up and away from the peptide binding cleft (11, 14). There is precedent for the importance of HVR3 in both peptide binding and Ag-specific T cell responses to DRB1*0402 (51, 52), DRB1*0401 (17, 50, 53, 54), DR13 (55), and DR11 (47, 56) and in alloreactive T cell responses (40, 57, 58).

The peptides present in the five MHC class II molecules for which crystal structures have been determined to date, which include DR1, DR3, DR4, and I-E^k, have remarkably similar conformations (11, 12, 13, 14, 15). In every case, the peptide binds in a extended conformation with residues at relative peptide positions P1, P4, P6, P7, and P9 projecting into pockets in the MHC class II molecule, and P-2, P-1, P2, P3, P5, P8, P10, and P11 exposed to the solvent. Molecular modeling of VP16 393–405 complexed with DRB1*0402 predicts a binding motif such that V394 occupies pocket 1 and R397 occupies pocket 4, as follows:

This motif was confirmed using substituted peptide analogues in peptide binding and T cell stimulation assays. Substitutions of V394 (P1) and R397 (P4) and removal of the C-terminal residues S402 (P9) substantially reduced peptide binding, while substitution of L393 (P-1) or removal of the last two C-terminal residues (G404 and Q405; P11 and P12) did not. Substitutions of M398 (P5) and F400 (P7) had small effects on peptide binding but substantially reduced ESL4.34 T cell responses, indicating that these changes influence TCR contact sites. The crystal structure of DRB1*0401 complexed with a human collagen II peptide (14) has confirmed previous predictions that the residue at DRβ position 71 influences the peptide that can be bound by forming a salt bridge with the side chain of the residue at peptide position P4 (5, 17, 50). Thus, while peptides with negatively charged P4 residues can bind to DRB1*0401 and other alleles with lysine or arginine at DRβ 71, peptides with positively charged P4 residues (lysine, arginine, or histidine) can bind to alleles expressing the ILEDE epitope at DRβ 67–71, such as DRB1*0402, DRB1*1102, and DRB1*1301. However, our observation that a nonpolar alanine residue at position P4 also supported good binding to DRB1*0404 suggests that size constraints are also important for this interaction.

In the diagonal binding mode observed in the class I MHC-peptide-TCR crystal structures, the CDR1 loops interact with the peptide termini, the CDR2 loops interact predominantly with the MHC α helices, and the CDR3 loops cover the central portion of the MHC-peptide ligand (18, 19, 20). In the absence of crystallographic data for class II MHC-peptide-TCR complexes, a similar diagonal binding orientation has been predicted based on shared properties of the class I and class II MHC molecules (19, 21). It is interesting to note the presence of three arginine residues in the CDR3 sequence of the ESL4.34 TCR α chain and an arginine at position 97 in CDR3 of the TCR β chain. A binding configuration analogous to the crystal structures of the class I MHC-peptide-TCR complexes would position the ESL4.34 CDR3 loops such that these arginines could potentially interact with residues of the negatively charged ILEDE epitope.

Our data indicate that although ESL4.34 displays promiscuous recognition of MHC class II molecules, its activation requirements are highly specific and depend on both the presence of the DRβ 67–71 ILEDE sequence and the VP16 393–405 peptide. Although multiple DR alleles expressing the ILEDE epitope were able to present this peptide to ESL4.34, conservative substitutions at this HVR eliminated T cell reactivity. Although the presence of the DRβ 67–71 ILEDE epitope has an influence on peptide binding, the relative affinity of the peptide-DR interaction does not have a major effect on the capacity of the ligand to stimulate ESL4.34. Instead T cell recognition appears to depend on specific interactions between the TCR and DR residues 67, 70, and 71, and peptide residues such as methionine at P5 and phenylalanine at P7, as discussed above. Other workers have also found that T cell responses do not depend on the affinity of peptide for MHC (59) and conservative substitutions in antigenic peptides and MHC molecules can have more dramatic effects on T cell responsiveness than nonconservative substitutions or unrelated peptides (27, 28, 29, 30, 31, 32). In this respect, the pluriallelic T cell recognition by clone ELS4.34 is reminiscent of recognition patterns of alloreactive T cells, in which specific recognition of particular DRβ HVR3 epitopes has been shown to be essential for the activation of many alloreactive T cell clones (40, 58, 59).

The peptide-self-MHC ligands recognized by developing thymocytes have an important influence on the selection of the peripheral T cell repertoire. In particular, the DRβ HVR3 residues have been shown to determine the specificity of positive selection and exert a strong bias on the nature of ligands recognized by T cells selected in an individual (60, 61). Promiscuous DR-restricted T cell recognition of Ag has been seen in the context of both parental alleles (48), consistent with the finding that ESL4.34 recognizes VP16 393–405 in the context of the parental alleles, DRB1*0402 and *1301. The ILEDE epitope, present on both alleles, likely amplified the selection bias for this (and related) specificities.

This observation may have important implications for the selection of autoreactive T cells. The third HVR of DRβ is thought to have a central role in susceptibility to autoimmune diseases including RA (34, 35) and autoimmune hepatitis (62). In RA, the genetic element most highly associated with this disease is a shared epitope consisting of the amino acid sequences LLEQRRAA or LLEQKRAA at DRβ positions 67–74, which are present on HLA alleles that are associated with RA (DRB1*0401, DRB1*0404, DRB1*0405, DRB1*0101, and DRB1*1402) in different populations (35). Interestingly, DRB1*0401/*0404-positive individuals, who are heterozygous for these two HVR3 sequences, have an increased risk of developing RA compared with the risk associated with possession of either HVR3 alone. This might reflect the influence of the HVR3s of both parental alleles in biasing the selection of T cells that exhibit promiscuous recognition of Ags in the context of both alleles, potentially contributing to the specificity of T cell responses that result in autoimmune activation in RA. We hypothesize that the HVR3 portion of the peptide-MHC ligand is a discrete site of T cell recognition that may itself function as a dominant selection element for autoreactive T cells.

We thank Mary Ellen Domeier and Ben Falk for technical assistance, and Nicky Ducommun for preparation of the manuscript.