Mycoplasma Superantigen Is a CDR3-dependent Ligand for the T Cell Antigen Receptor

CD4⁺ clones were obtained after stimulation of fresh PBL with MAM by limiting dilution in 96-well plates. Every 2 wk clones were stimulated by sodium periodate–treated non-T PBL as previously described (18) and maintained in RPMI 1640 with 10% FCS and 5% IL-2 (Pharmacia Biotech, Piscataway, NJ).

2 × 10⁴ T cells and 1.5 × 10⁴ irradiated (15,000 rads) lymphoblastoid 8866P cells were incubated with 10-fold dilutions of superantigens (from 1 to 10⁻⁵ ng/ml) in triplicate microwells for 48 h, then labeled overnight with [³H]-thymidine and harvested. Reactivity was scored as cpm greater than twofold over background. Superantigens were purchased from Toxin Technology, (Sarasota, FL); MAM, purified to homogeneity, was provided by Dr. B.C. Cole (University of Utah School of Medicine, Salt Lake City, UT) and prepared according to reference 19.

cDNAs obtained from human T cell clones were amplified with primers specific for human Vβ families (18) and Vα families (20), cloned into a pCRII vector (Invitrogen Corp., Carlsbad, CA) and sequenced by the Sequenase 2 method (United States Biochemical, Cleveland, OH). Expression of Vβ 5.1, Vβ 8.1/2, Vβ12.2, and Vβ17.1 was confirmed by staining with anti-Vβ–specific monoclonal antibodies (21).

The retroviral cassette pTbeta was constructed by subcloning a SnabI-SalI fragment of mouse Cβ2 gene into the pBabe Puro vector (22). PCR products containing the leader sequence, Vβ region, N, Dβ, and Jβ were derived from T cell clones N17 and J17. Mutants were generated by overlapping extension PCR (23) cloned into pTbeta and transfected into the packaging cell line BOSC by the calcium phosphate method (24). TCR β chain–deficient T hybridoma cell lines DS 23.27 (mouse Vα2; reference 25) and YLβ- (mouse Vα3.1; reference 26) were infected with each virus, puromycin-selected, and sorted for Vβ17 expression using the C1 monoclonal antibody (21).

Activation of the TCR transfectants was tested according to reference 27. 10⁵ cells in a 100 μl/well were mixed with an equal number of 8866P lymphoblastoid cells. 10-fold dilutions of each superantigen were tested in triplicates. After 24 h, supernatants were frozen at −20°C. Portions of these supernatants were thawed and incubated for 24–30 h with IL-2– dependent HT-2 cells at 4,000 cells/well. Viability of HT-2 cells was evaluated by the MTT (3-(dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide) assay and analyzed on an ELISA reader at 570 nm (28). The amount of IL-2 produced was determined in comparison with the standard curve established with recombinant IL-2 (BRMP units; Becton Dickinson).

To characterize MAM-reactive TCRs, we tested a panel of 16 human T cell clones for proliferative responses to 7 different superantigens (Table 1, Fig. 1). Clonal T cells were incubated with APCs in the presence of gradual 10-fold dilutions of each superantigen, then pulsed with [³H]-thymidine and harvested. Among T cell clones using the same Vβ genes, similar patterns of reactivity were observed with most superantigens, whereas reactivity to MAM often differed (Fig. 1, shaded field). T cell clones expressing either Vβ5.1, Vβ8, Vβ12, or Vβ17 could respond to MAM, but within each of these Vβ subsets there were both reactive (MAM⁺) and nonreactive (MAM⁻) clones. Lack of reactivity to MAM was not dose dependent, as we were unable to find concentrations of MAM that would activate MAM⁻ clones (Table 1). The MAM⁻ cells were competent to respond to TCR-mediated signals from other superantigens (Fig. 1). In addition, the sequences of TCR Vβ regions from MAM⁺ and MAM⁻ T cell clones were identical, ruling out the possibility that Vβ allelic variants resulted in differences in MAM reactivity.

Strikingly, all T cell clones that did not respond to MAM used the TCR Jβ2.1 gene segment, whereas MAM-reactive clones used other Jβ2 segments (Fig. 2). Alignment of the CDR3 regions showed that a tyrosine (Y), present within the conserved motif Q(Y/F)FG, correlated with responsiveness to MAM. Nonreactive clones had phenylalanine (F) at this position. The same correlation was observed in polyclonal cell lines obtained after repeated stimulation of fresh T cells with MAM. Dominant clones within these cultures (Fig. 2; SEL) represent T cells expanded in vitro after stimulation with MAM. All Vβ17 MAM⁺ and SEL clones have isoleucine at position 94 in addition to Y within the Q(Y/F)FG motif in their CDR3-β region.

Staining data indicated that in the polyclonal cell lines only the Vβ17⁺ subset was expanded in response to MAM, to ∼50% of all T cells, in agreement with previous data (21). The minor MAM-responsive phenotypes (Vβ5, Vβ8, and Vβ12) were not expanded in these polyclonal cell lines. However, within these minor Vβ subsets, dominant clones were also found after three to five stimulations with MAM. We hypothesize that these clones represented relatively infrequent cells within the Vβ5, Vβ8, or Vβ12 subsets, which may depend on the other TCR elements for MAM reactivity. Consistent with this view, Vα2 and Vα8 were frequently observed in clones of minor Vβ phenotypes (Fig. 2). Thus, reactivity may depend on Vα usage with these minor Vβ phenotypes, as previously described for low affinity TCR Vβ–superantigen interactions (29–31).

However, the major MAM-responsive Vβ17 phenotype is not α chain–dependent. To prove this point, we transfected β chains derived from clones N17 and J17 into two mouse T hybridoma cell lines, DS 23.27 and YLβ-, that are β chain–deficient (Table 2). Surface expression of the endogenous mouse TCR-α chain, either Vα2 or Vα3.1, respectively, is rescued by TCR complex formation with the transfected β chain. β chains N17 and J17 retained their differential MAM reactivity regardless of which α chain they were paired with (Table 2).

Presentation of superantigens to T cell clones can be sensitive to minor differences among MHC class II alleles (32) and the nature of the peptide in the MHC class II groove (33). Therefore, we tested the reactivity of MAM⁺ and MAM⁻ clones with three superantigens presented by lymphoblastoid cell lines expressing different MHC class II alleles (Fig. 3). In each of four groups (Vβ17, Vβ8, Vβ12, and Vβ5), MAM⁺ and MAM⁻ clones retained their specific reactivity regardless of the MHC class II alleles expressed on APCs. The magnitude of the proliferative responses differed from one APC type to another, probably due to different levels of expression of MHC class II or accessory molecules, but the pattern of reactivity remained the same. MAM⁺ clones were always MAM-reactive and MAM⁻ clones did not respond to MAM presented by other APCs.

Sequence alignment of TCR-β chains implicated a single residue (Y/F) of the CDR3 region in determining MAM reactivity (Fig. 2). To test whether Y103 of clone N17 (MAM⁺) and the equivalent residue, F106, of clone J17 (MAM⁻) determine MAM reactivity, we attempted to reconstitute reactivity of J17 by mutagenesis in the CDR3-β region (Fig. 4). Wild-type and mutant TCRs were expressed in DS23.27 cells and tested for IL-2 production in response to several superantigens. A single amino acid substitution (JY, F106Y) at position 106 was insufficient to restore reactivity to MAM. A second candidate amino acid was isoleucine in position 94. This position is encoded by the Vβ–Db junctional region and in all Vβ17⁺ clones reactive with MAM, it was represented by isoleucine (Fig. 2). Therefore, we introduced a second substitution at position D94 of J17. Strikingly, the combination of D94I and F106Y (transfectant JIY) resulted in full reconstitution of MAM reactivity (Fig. 4) while the single change, D94I, was insufficient (JIF).

To demonstrate that other regions of CDR3 were not involved, we tested three deletion mutants including either one or two residues between I94 and Y106 (JIY-SY, JIY-NE, JIY-E). These mutants would be expected to differ in their peptide/MHC reactivity, because recognition of conventional peptide antigen is extremely sensitive to changes in CDR3 (34, 35). All three deletion mutants retained MAM reactivity (Fig. 4). Together with the data on MAM presentation by different MHC class II alleles (Fig. 3) these experiments exclude a role for MHC II–bound peptides in these MAM–TCR interactions. A direct contact between TCR CDR3-β and MAM is the most likely explanation for dependence on residues I94 and Y106. Based on recent crystallographic data, these two amino acids are located at the base of the CDR3 loop lying in close proximity to one another (36). The side chain of Y106 extends towards the exposed surface of the TCR (Y107 in reference 2). It appears that residues I94 and Y106 are accessible for direct interaction with MAM. These interactions are specific for MAM since none of the tested mutations affected reactivity with staphylococcal superantigens (Fig. 4).

In this study we show that the CDR3 region of TCR-β chain mediates specific recognition of the MAM superantigen. Complementarity-determining regions (CDR loops) (2, 37) represent hypervariable loops of the TCR that face the MHC during antigen recognition. On the β chain, both CDR1 and CDR2 as well as the HVR4 loops can be involved in interactions with superantigens. TCR-β point mutations that affect specific interaction with a particular superantigen are shown in Fig. 5. The cocrystals of TCR β chain with staphylococcal enterotoxins SEC2 and SEC3 revealed that these staphylococcal superantigens interact with CDR1, CDR2, FR3, and HVR4 but not with the CDR3 loop (38). Most contacts (47%) are formed between the superantigen and the CDR2 loop. Since Y106 of CDR3 (2) and exposed CDR2 residues are located at opposing faces of the TCR β chain (1), our data suggest that recognition of MAM and staphylococcal superantigens is substantially different. That CDR2 regions are not involved in MAM recognition is supported by the observation that the primary sequence of the homologous human Vβ17 and mouse Vβ6 chains, both MAM-reactive, differs in the CDR2 region.

Mutational and structural studies have shown that the CDR3 loops of the α and β chains are central in peptide recognition (1, 2, 34, 35, 39, 40). CDR3 regions are the most diverse parts of the TCR produced by extensive processing of end joints during antigen receptor rearrangement (3). Junctional diversity is increased by N-additions contributed by the lymphoid-specific enzyme terminal deoxytransferase (TdT; references 41–43). Although an N-region– depleted TCR repertoire from TdT-knockout mice was shown to be as efficient as normal TCR repertoire (44), the N-region-depleted TCRs were more promiscuous in peptide recognition. These TCRs recognized a significantly larger number of different peptides than did the wild-type TCRs when tested for reactivity against a library of random peptides (45).

The structural basis for the role of N-encoded TCR residues was revealed in a recently solved structure of the TCR–MHC interface (1). In this crystal, three interactions are formed between residue p5 of the peptide and the TCR. Two of these are contacts of p5 with N-encoded residues of the Vα–Dα and Vβ–Dβ junctional regions.

The specific interaction of MAM with the TCR appears to imitate the natural process of peptide/MHC recognition. During peptide recognition, CDR3 regions of α and β chains form contacts with MHC. The complex is stabilized by CDR3–peptide interactions (46, 47). Recognition of MAM is clearly Vβ-dependent, as is the case with other superantigens, but it appears to be stabilized by additional contacts with the CDR3-β region.

In all MAM-reactive Vβ17⁺ clones that we have sequenced, position 94 was represented by isoleucine. I94 is adjacent to the conserved end of Vβ region CASS and corresponds to position 2 of the CDR3 region (48, 49). In random CDR3 sequences, position 2 is rarely represented by isoleucine (49) but in Vβ17⁺ sequences I94 is more common, occurring in ∼20% of the sequences (50). This residue is often assumed to be part of the N-region, but in the case of isoleucine it may be encoded by the 3′ end of the germline Vβ17 sequence, codon ATA of the genomic sequence (51). However, in ∼80% of the cases this codon is removed during Vβ–Dβ recombination and replaced by N-additions.

The second CDR3-β position critical for interaction with MAM is tyrosine within the Q(Y/F)FG motif of Jβ2 (T cell clones using Jβ1 are rare and they were not analyzed in this study). At this position Y is encoded by Jβ2.3, Jβ2.4, Jβ2.5, and Jβ2.7. F, which does not allow MAM reactivity (Fig. 4), is encoded by Jβ2.1 and Jβ2.2. The Jβ2.1 segment, which was used by the MAM⁻ T cell clones studied here (Fig. 2), is used by ∼40% of all T cells from peripheral blood (52).

The major subset of T cells targeted by MAM is the Vβ17 subset (21). Yet not all Vβ17⁺ T cells respond to MAM, and the response is limited to cells expressing two appropriate residues at the base of the CDR3-β loop. This suggests that T cell activation by MAM is dependent on TCR junctional diversity, thus limiting the number of potential MAM-reactive T cell clones. The frequency of T cells responding to this superantigen could be significantly lower than to other superantigens (for example SEB, SEC1, SEC2, and SEC3). Among the minor MAM-responsive TCR phenotypes (Vβ5, Vβ8, and Vβ12), the frequency is even lower than within the Vβ17 subset, probably because Vα usage is an additional requirement as suggested by the data in Fig. 2. As previously shown, MAM-responsive T cells expressing Vβ5, Vβ8, and Vβ12 can only be detected by testing individual clones and these Vβ phenotypes do not expand in polyclonal cultures exposed to MAM (18). However, TCRs of the minor phenotype also appear to use the Jβ-encoded tyrosine for MAM recognition (Fig. 2).

It is possible that Jβ usage can affect low affinity interactions of TCR with other superantigens. Incomplete deletion of Vβ8⁺ or Vβ6⁺ cells in mice carrying Mls-1a (Mtv-7) revealed that TCRs that escaped deletion used distinct Jβ regions (53, 54). In this experimental system, Jβ1.2 seemed to protect Vβ6⁺ TCR from interaction with a retroviral superantigen, Mls-1a. However, these studies did not find any conserved motifs in the CDR3 region that could be implicated in unresponsiveness to Mls-1a (54). Staphylococcal superantigens seem to be CDR3-independent (Fig. 4 and reference 55). However, in the case of SEB and Urtica dioica superantigens, it was suggested that the Jβ segment, but not the CDR3 region, can affect superantigen binding by influencing the quaternary structure of the TCR-β chain (55).

The major MAM-responsive Vβ phenotype is strikingly dependent on discrete residues of CDR3. This fact clearly distinguishes MAM from other superantigens. Thus, recognition of MAM is dependent on junctional diversity of the TCR β chain.

Mycoplasma arthritides is a microorganism that causes a disease in rodents that is remarkably similar to human rheumatoid arthritis (RA). It is also the only mycoplasma that produces MAM. It should be emphasized that Mycoplasma arthritides is not a human pathogen. Surprisingly, preferentially expanded T cell clonotypes found in human RA often belong to the MAM-reactive TCR phenotype. Dominant clones from RA patients, characterized in three independent studies, were found to use Vβ17 followed by isoleucine at position 94 (50, 56, 57). In one of the studies, these T cells were shown to be autoreactive against a lymphoblastoid cell line expressing HLA-DR4 (50). Similar autoreactivity was reported in a TCR transgenic mouse model of RA (58). Inflammatory sinovitis in mice was triggered by the transgenic TCR recognition of host MHC class II (A^g7). This transgenic TCR-β chain used a mVβ6 followed by isoleucine (CASSI) (D. Mathis, personal communication). Murine Vβ6 is the closest homologue of human Vβ17, and mouse Vβ6⁺ T cells are also MAM-reactive (15). Thus, both murine and human TCRs associated with RA appear to express a conserved isoleucine at the COOH-terminal end of the Vβ region. The role of the conserved isoleucine in the CDR3-β region of arthritogenic TCRs remains to be determined. Is it necessary for interaction with MHC class II or a putative joint-specific peptide? Is it possible that tolerized self-reactive clones can be reactivated by MAM in mice infected by Mycoplasma arthritides?

There is a strong genetic link between certain DR alleles (class II MHC) and RA in humans (59, 60). Position 71 of the particular DR4 or DR1 β chains, a lysine residue, is associated with susceptibility for RA (61). This residue determines the nature of the peptides in the class II groove by interacting with P4/P5 of the peptide (for review see reference 62). Conserved CDR3-β motifs that include isoleucine 94 have been proposed as evidence for antigen-driven immune response in RA (50). Whether superantigens like MAM have a role in RA pathogenesis is still an open question.

We are grateful to N. Bhardwaj for her clones of human T cells, B.C. Cole for purified MAM protein, J.P. Morgenstern for the pBabe vector, W.S. Pear for the BOSC cell line, and D. Kostyu for HLA typing. We also thank G. Kelsoe, A. Chervonsky, and S. Santagata for discussion of the manuscript.

This work was supported by an Aaron Diamond Foundation fellowship to A.S. Hodtsev and by National Institutes of Health grants to D.N. Posnett (AI-31140 and AI-33322) and to Y. Choi (CA-59751).