Immunology Letters 91 (2004) 133–139

Molecular basis of antigen recognition by insulinspecific T cell receptor

Shigeru Sugiyamaa, Masako Kohyamaa, Masayuki Odaa, Takachika Azumaa,Joan E. Witherb, Nobumichi Hozumia,∗

a Research Institute for Biological Sciences (RIBS), Tokyo University of Science, 2669 Yamazaki, Noda, Chiba 278-0022, Japanb The Arthritis Center of Excellence, Toronto Western Research Institute, University Health Network,

and Departments of Medicine and Immunology, University of Toronto, Toronto, Ont., Canada

Received 21 October 2003; received in revised form 21 October 2003; accepted 5 November 2003

Abstract

The TCR�/� chains recognize antigen peptides bound to the groove of the MHC class II molecule. The crystal structure analyses of theTCR/peptide/MHC class II complexes have revealed that the V� chains play a significant role in antigen recognition. However, moleculardetails which amino acid residues of the V� chain are able to contribute to fine antigen specificity are not clearly understood. Previously, wehave classified a panel of T hybrids specific for insulin isotypes from different species of animals into four groups based on response profilesto these antigens. In particular, the group III (pork insulin≥ beef insulin hierarchy of responsiveness) and IV (pork insulin� beef insulinhierarchy of responsiveness) T hybrids are interesting, since these TCR�/� chains with marked different antigen specificities demonstrateidentical gene usages and very similar sequences. To specifically address the molecular requirements for insulin recognition by TCR, the TCR� and� chain genes from these group III and IV T hybrids were transfected into 58�−�− T hybrid. The experiments suggested that CDR3�dictates the fine antigen specificity. Then, we have introduced a series of mutations into position 95 of CDR3�. The mutation experimentsclearly indicated that position 95� determines the antigen specificity of the group III and IV T hybrids.© 2003 Elsevier B.V. All rights reserved.

Keywords: TCR CDR3�; Insulin peptide; In vitro mutagenesis; MHC class II

1. Introduction

The T cell receptor (TCR) of CD4+ T cells recognizesantigenic peptides bound to the groove of major histocom-patibility complex (MHC) class II molecule displayed onthe membrane surface of antigen presenting cells (APCs)[1]. The TCR is composed of� and� chains, which are as-sembled from V, D (for� only), and J segments[2]. The Vregions of the� and� chains include three complementaritydetermining regions (CDRs), referred to as CDR1, CDR2and CDR3, with a structure analogous to immunoglobulin(Ig). CDR1 and CDR2 are encoded for by the germline V

Abbreviations: APC, antigen presenting cell; BI, beef insulin; CDR,complementarity determining region; Ig, immunoglobulin; MHC, majorhistocompatibility complex; PI, pork insulin; pMHC, peptide/MHC; TCR,T cell receptor

∗ Tel.: +81-4-7121-4091; fax:+81-4-7121-4099.E-mail address: nobhozmi@rs.noda.tus.ac.jp (N. Hozumi).

gene segments[3], however unlike Ig, somatic mutations arerarely introduced in these regions[4]. Instead the majorityof TCR variability is located in CDR3 which is somaticallycreated by rearrangement of V, D and J segments. The di-chotomy between the less variable CDR1 and CDR2, andmore variable CDR3 led to speculation that CDR3 contactsantigenic peptide while the CDR1 and CDR2 contact MHCclass II[3–5].

Crystallography of three different TCR/peptide/MHCclass II (TCR-pMHC II) complexes (one mouse and twohuman) [6–8] and a number of TCR/peptide/MHC classI (TCR-pMHC I) complexes[9–12] has provided furtherinsight into the structural basis of TCR recognition. Thesestudies indicated that the TCR molecule sits on top of thepMHC complex diagonally rather than in the orthogonalposition originally proposed. Although there is considerablevariability in the angle of TCR-pMHC interactions for dif-ferent TCR-pMHC complexes, the limited crystallographicdata on TCR-pMHC II complexes suggest that the TCR

0165-2478/$ – see front matter © 2003 Elsevier B.V. All rights reserved.doi:10.1016/j.imlet.2003.11.001

134 S. Sugiyama et al. / Immunology Letters 91 (2004) 133–139

may adopt a more orthogonal position on pMHC II thanseen for TCR-pMHC I complexes. The consequence of thismore orthogonal position is that CDR1� and CDR1� in-teract with both peptide and MHC II, whereas CDR2� andCDR2� interact mainly with the MHC II helices and havelittle contact with peptide[6]. CDR3� and CDR3�, whichare located over the peptide binding groove, make the mostextensive contacts with peptide. In the crystal structuressolved to date the V� domain makes a greater number ofcontacts with the pMHC II ligand than the V� domain,suggesting that it plays a dominant role in TCR recognition[6–8]. Several reports have suggested that the V� chainplays a predominant role in TCR repertoire developmentand mature T cell activation[13–15]. Nevertheless, it re-mains unclear whether the dominant role of the V� chainin T cell recognition reflects the limited number of antigensexamined or a fundamental aspect of the TCR-pMHC IIcomplex.

We have been investigating the molecular mechanismsinvolved in TCR recognition using insulin peptides asmodel antigens. Insulin, a heterodimeric molecule, is linkedby two interchain disulfide bonds between the 21-residue Achain and 30-residue B chain (Cys7-Cys7 and Cys20-Cys19)[16]. A third intra-chain disulfide bond is found in theA-chain (Cys6-Cys11) and has been described as the A-loopregion. The minimal immunogenic peptide (A1–13) of in-

Fig. 1. Amino acid sequences of PI, BI and the TCR CDR3 of the group III and IV T hybridomas. (A) Amino acid sequences of A (1–14) of PI and BI.(B) Amino acid sequences of TCR CDR3 of the V� and V� chains. Numbering of the amino acid sequences is as proposed by Kabat and coworkers[19,20]. Note position 99� is deleted.

sulin molecules binds to the MHC II groove in an extendedconformation as a result of intracellular reduction of theA-loop disulfide bond[17].

We have previously generated a panel of insulin Achain-specific I-Ad-restricted T cell hybridomas and shownthat they expressed a restricted repertoire of TCR� and�chains[16]. The hybridomas could be classified into fourgroups based on reactivity patterns to species variants ofinsulin, that differ in their A-loop regions (Fig. 1A). Dif-ferences in antigen reactivity were shown to correlate withsequence differences in CDR3� and CDR3�, suggestingthat the TCR CDR3 bind the insulin peptide. By assessingthe ability of mutant MHC II molecules to present in-sulin species variants to the hybridomas, we proposed thatinsulin-specific TCRs were oriented on the insulin-I-Ad

complex in a more orthogonal orientation than MHC Irestricted TCRs[18], similar to the orientation later pro-posed when the mouse TCR/conalbumin/I-Ak complexthat was crystallized[6]. Our study further indicated thatTCR-peptide contacts play a dominant role in T cell recog-nition. Although we demonstrated that CDR3� contactedpeptide and defined specificity, we were unable to defini-tively identify a role for CDR3� using this approach[18].Here we have used an alternate approach to identify the roleof CDR3� in insulin-peptide recognition. Our approachfocuses on two insulin-specific TCRs that differ in their

S. Sugiyama et al. / Immunology Letters 91 (2004) 133–139 135

ability to recognize beef insulin (BI). The type III hy-bridoma, GBD51 recognizes pork insulin (PI) and BI,whereas the Type IV hybridoma, B8C3X only recognizes PI[16]. The TCR of these two hybridomas differ by a singleamino acid at position 95 of CDR3� (Position 94� in ourprevious paper was numbered to 95� in this paper followingthe sequence numbering of Kabat and coworkers)[19,20]and two amino acids at position 100 and 102 of CDR3�(Fig. 1B). To define the relative role of CDR3� and CDR3�in mediating the fine specificity differences between thesetwo hybridomas, we have performed gene transfer experi-ments with various endogenous and mutagenized TCR�/�gene pairs. We show that the amino acid difference at po-sition 95 within CDR3� is solely responsible for the finespecificity differences and that this residue plays a criticalrole in T cell recognition. These results are compatible withthe current model of TCR-pMHC II interaction and takentogether with our previous results[16,18], suggest that bothCDR3� and CDR3� are likely to contact the antigenicpeptide and define the fine specificity of the TCR.

2. Materials and methods

2.1. Cells

The cellular and molecular characterization of the T cellhybridomas, GBD51 and B8C3X, has been previously de-scribed[16]. The T cell hybridoma, 58�−�− which lacksTCR � and� chains and CD4, and variant transfected withthe CD4 gene, 58�−�− CD4+, were generous gifts from Dr.Ed Palmer[21]. M12C3, a mutant B cell lymphoma lackingI-E surface expression and the I-A�d chain was transfectedwith the I-A�d gene as described[18].

2.2. Insulins and culture media

BI and PI were purchased from Sigma (St. Louis, MO).RPMI 1640 medium was supplemented with 10% fetalcalf serum, 2 mM glutamine, 5�M 2-mercaptoetanol and100�g/ml each of penicillin and streptomycin.

2.3. Antigen presentation and IL-2 assays

Antigen presentation assays were performed using theM12C3 I-A�d transfectant as APC[18,22]. For assays, 1×105 T cells and 1× 105 APCs were co-cultured for 24 hat 37◦C in wells containing various concentrations of PIor BI. IL-2 production was assayed by addition of 50�l ofcell culture supernatant to 1× 104 CTLL cells. All assayswere performed in triplicate, and the results are expressedas the arithmetic mean of the counts harvested following[3H]-thymidine incorporation by CTLL cells. In general, thestandard deviation (S.D.) was within±10% of the meanvalues.

2.4. Flow cytometory

T cell expression of CD4 and TCR V�8 was determinedby flow cytometry. Approximately, 105 cells were stainedwith monoclonal antibody for 30 min at 4◦C in total vol-ume of 100�l. CD4-PE conjugated monoclonal antibodywas obtained from Pharmingen (San Diego, CA, USA).Cells stained with F23.1 monoclonal antibody, which rec-ognizes V�8.1, V�8.2 and V�8.3, were washed and stainedwith a goat anti-mouse IgG flourescein-conjugated sec-ondary reagent. Samples were analyzed using a FACScan(Becton–Dickinson).

2.5. Construction of TCR expression vectors

Full-length cDNAs encoding the TCR� and� chain genesof GBD51 (group III) and B8C3X (group IV) were clonedinto pcDNA3, an eukaryotic expression vector containingthe�-actin promoter and a neomycin resistance cassette. TheTCR V� chain gene of group IV (cloned into the pcDNA3vector) was used as a template for site-directed mutagenesis.Point mutations were introduced using the Quick Changesite-directed mutagenesis kit (Stratagene, Heidelberg, Ger-many). The following primers were used for the introductionof the point mutations;

95Arg (GCTGCAAGTGGTCGACGAGGTTCAGCC),R95Arg (GGCTGAACCTCGTCGACCACTTGCAGC),95Glu (GCTGCAAGTGGGGAAAGAGGTTCAGCC),R95Glu (GGCTGAACCTCTTTCCCCACTTGCAGC),95Ala (GCTGCAAGTGGGGCCCGAGGTTCAGCC),R95Ala (GGCTGAACCTCGGGCCCCACTTGCAGC).

Mutated TCR chains were sequenced using ABIPRISMTM Dye Terminator Cycle Sequencing Ready Reac-tion kits (Perkin-Elmer) and cloned into the TCR expressionvectors.

2.6. Transfection of TCR expression vectors into mouse Thybridoma cells

The 58 �−�− T hybridoma cells(1 × 107) were sus-pended in phosphate-buffered saline, mixed with the lin-earized plasmid DNA, and electroporated (350 V, 960�F)using a Gene Pulser apparatus (Bio-Rad). Cells were seededinto flat-bottom 96-well microtiter plates, and stable trans-fectants were selected by the addition of G418 (2 mg/ml).Drug-resistant colonies were analyzed for the expression ofcell surface TCR/CD3 complexes by FACS. Transfectants,selected for high TCR expression levels, were used in sub-sequent experiments.

3. Results and discussion

CD4 plays a role in TCR specificity and modulates TCRmediated signaling[23–26]. Initially, we examined whether

136 S. Sugiyama et al. / Immunology Letters 91 (2004) 133–139

Fig. 2. Fine specificity of�(III ) × �(III ) CD4+ (A) and �(III ) × �(III )CD4− (B) transfectants. The CD4+ group III TCR (�, �) and CD4−group III TCR (�, �) transfectants were tested for IL-2 production inresponse to various concentrations of PI (filled circle) and BI (opensquare). The results are expressed as3H-thymidine incorporation (cpm)by IL-2 dependent CTLL. Each point is the arithmetic mean of triplicateexperiments. The results are representative of four different CD4− andCD4+ T clones with similar results.

expression of CD4 in the 58�−�− cell line was neces-sary to obtain strong insulin-specific T cell responses fol-lowing transfection with TCR� and� genes. To this end,CD4+ and CD4− 58 �−�− cells were transfected with theTCR � and� genes from GBD51, denoted�(III) and �(III)to indicate the specificity of the original hybridoma. Flowcytometric analysis of cells stably transfected with TCR� and � chains showed similar TCR levels on CD4+ andCD4− transfectants (data not shown). Although the CD4+transfectants demonstrated increased production of IL-2 inresponse to BI and PI at all antigen concentrations, anti-gen specificity was not significantly affected by CD4 andthe dose response curve was not shifted (Fig. 2). We haveobtained the similar results using three other independentCD4− and CD4+ clones. These results contrast with the re-quirement for CD4 in primary T cells, where it has been

suggested that engagement of CD4 promotes immunologi-cal synapse formation and enhances T cell activation[27].It is possible that the T cell hybridomas used in our ex-periments may not require CD4 for signaling, since thecellular activation machinery is different from that of pri-mary T cells. Alternatively, our transfectants may have anincreased number of the TCR molecules expressed on thecell surface or have a higher affinity TCR than the pri-mary TCR transgenic T cells that were previously exam-ined, obviating the need for CD4. Based on these results,we decided to use the parental CD4− T cell line for furtherexperiments.

As outlined previously group III and group IV hybridomasare both PI-reactive but differ in the ability to respond to BI(group III but not IV are BI-reactive)[16]. To determine con-tribution of CDR3� and CDR3� to this altered specificity,we generated a series of transfectants expressing differentcombination TCR� and� genes of TCR genes from thesehybridomas. The combinations used were�(III ) × �(III ),�(IV ) × �(III ), �(III ) × �(IV ), and�(IV ) × �(IV ). Trans-fectants expressing similar TCR levels, as revealed by flowcytometry (data not shown), were assessed for their abilityto respond to PI and BI. Although all four sets of trans-fectants responded similarly to PI, only�(III ) × �(III ) and�(III ) × �(IV ) transfectants responded to BI (Fig. 3). Theresults clearly indicate that CDR3� dictates the specificityof the immune response to PI and BI.

Since the TCR�(III) and � (IV) differ only at position95, it is likely that this residue is critically involved in therecognition of the insulin peptide. To further assess this pos-sibility, position 95 of�(IV) was mutated from Gly (smallestnonpolar amino acid) to several different amino acids. Thepresence of Asp (negatively charged smallest amino acid) atthis position in�(III) suggested that a negative charge mayfacilitate interaction to BI, therefore Gly95� was mutated toAla, Glu, and Arg as representative, small nonpolar, nega-tively charged and positively charged amino acid substitu-tions. These mutants were transfected along with�(IV) geneand stable transfectants expressing similar TCR levels cho-sen for analysis (data not shown). Surprisingly, all three ofthe induced CDR3� mutants were unresponsive to both BIand PI (Fig. 4). These results indicate that V�95 is criticalfor insulin recognition.

Although the precise structure of insulin peptide/I-Ad hasnot been determined, several lines of evidence suggest thatVal3 of PI and BI corresponds to P1 pocket of I-Ad (seeFig. 1A): (1) truncation of the A1-14 peptide to A3-14 resultsin a markedly decreased ability to stimulate T cells, withfurther truncation leading to complete loss of reactivity[17];(2) alkylation of Cys7 (P5) does not affect binding of theA1–14 peptide to I-Ad, while alkylation of Cys6 (P4) andCys11 (P9) abrogates binding, consistent with the role of P4and P9 as anchor residues[28]; (3) a free thiol at Cys7 (P5)is required for T cell recognition by some insulin-specificI-Ad-restricted hybridomas[28]; (4) Val3 at P1, Cys6 at P4,and Cys11 at P9 are compatible with the requirement for

S. Sugiyama et al. / Immunology Letters 91 (2004) 133–139 137

Fig. 3. Fine specificity of�(III ) × �(III ) (A), �(IV ) × �(III ) (B), �(III ) × �(IV ) (C), and�(IV ) × �(IV ) (D) TCR transfectants. The TCR transfectantswere tested for IL-2 production in response to various concentration of PI (filled circle) and BI (open square). The results are expressed as3H-thymidineincorporation (cpm) by IL-2 dependent CTLL. Each point is the arithmetic mean of triplicate experiments.

small hydrophobic residues at P4 and P9 to bind to I-Ad [29];(5) Positioning Val3 at P1 places Glu4, the only differencebetween PI and mouse insulin, at P2 where it can contact theTCR providing an explanation for the ability of T cells inI-Ad mice to discriminate between these two species variantsof insulin and thus mount a response to PI[29,30]. Thismodel results in the A-loop amino acids that vary betweenthe different species variants of insulin being located at P6–8.

We have previously provided evidence that in someinsulin-specific hybridomas the TCR CDR3� binds to P8,providing the basis for discrimination between PI and BI,which have Ile and Val at this position, respectively[18].Here, we provide evidence that the basis for another subsetof hybridomas ability to discriminate between these twospecies variants of insulin resides in the TCR CDR3�.Based upon the recently published crystal structures ofmouse and human TCR-pMHC II complexes we proposethe following molecular mechanism for this interaction[6–8]. The crystal structure of the TCR, D10, complexedwith conalbumin peptide bound to I-Ak, has shown that

Ser100� and Phe101� of CDR3� interact with the P5-P8 re-gion of the peptide[6]. According to an alignment proposedby Chothia et al.[3,20], the CDR3� residues (from Thr93�

to Leu104�) of D10 correspond to the residues (from Ser93�

to Leu104�) of the group III and IV CDR3�, since the D10V� region is four residues shorter than insulin specificV�15. Thus, the D10 residues (Gly97�-Ser100�-Phe101�)may correspond to the insulin specific V�15 residues(Gly94�-Asp/Gly95�-Arg96�-Gly97�-Ser98�). We speculatethese V�15 residues recognize the P5-P8 region of the in-sulin peptide which contains the variable A-loop peptidesthat differ between BI and PI (seeFig. 1).

It is currently unclear how the Gly95� to Asp95� changepromotes binding to BI. Based upon the amino acid differ-ences between PI and BI together with the results of ourCDR3� mutagenesis experiments, it appears unlikely thatthe difference in charge alone is sufficient to explain theacquisition of BI reactivity. It seems more likely that theAsp95�/Gly95� differences modulate the CDR3� loop con-formation. As demonstrated in the recent crystal structure

138 S. Sugiyama et al. / Immunology Letters 91 (2004) 133–139

Fig. 4. Fine specificity of�(IV ) × �(IV ) (A), �(Gly95Ala) × �(IV ) (B), �(Gly95Glu) × �(IV ) (C), and�(Gly95Arg) × �(IV ) (D) TCR transfectants.The TCR transfectants were tested for IL-2 production in response to various concentrations of PI (filled circle) and BI (open square). The results areexpressed as3H-thymidine incorporation (cpm) by IL-2 dependent CTLL. Each point is the arithmetic mean of triplicate experiments.

analyses of TCRs, the CDR3� loop is extremely flexible[31]. It is probable, that the specific recognition of in-sulin is extremely sensitive to small perturbations of theconformation of the CDR3� loop so that even conservativemutations fail to maintain insulin-reactivity. This observa-tion suggests that while a given TCR has the capacity torecognize several structurally unrelated antigens, the interac-tion between the TCR and a given antigen is highly specific.

In summary, we have clearly demonstrated that for a sub-set of insulin-specific I-Ad-restricted TCR that CDR3� dic-tates specific insulin peptide recognition. This observation,taken together with our previous findings indicating thatall insulin-specific TCR are similarly positioned over theinsulin-I-Ad complex and that interactions between CDR3�also dictate insulin specificity, suggests that both CDR3�and CDR3� are positioned over P6–8. This structure isanalogous to the TCR-pMHC crystal complexes defined todate, providing evidence that these structures are likely tobe representative of T cell recognition involving diverseantigens.

Acknowledgements

This work was supported by grants from the Ministry ofEducation, Science, Sports and Culture in Japan and theJapan Health Sciences Foundation. The authors thank Dr.Akinori Sarai of the Institute of Physical and Chemical Re-search (RIKEN) for technical support and helpful discus-sion.

References

[1] Babbitt BP, Allen PM, Matsueda G, Haber E, Unanue ER. Nature1985;317:359–61.

[2] Kronenberg M, Sui G, Hood LE, Shastri N. Annu Rev Immunol1986;4:529–91.

[3] Chothia C, Boswell DR, Lesk AM. EMBO J 1988;7:3745–55.[4] Davis MM, Bjorkman PJ. Nature 1988;334:395–402.[5] Claverie JM, Prochnicka Chalufour A, Bougneleret L. Immunol To-

day 1989;10:10–4.

S. Sugiyama et al. / Immunology Letters 91 (2004) 133–139 139

[6] L Reinherz E, Tan K, Tang L, Kern P, Liu J, Xiong Y, et al. Science1999;286:1913–21.

[7] Hennecke J, Carfi A, Wiley DC. EMBO J 2000;19:5611–24.[8] Hennecke J, Wiley DC. J Exp Med 2002;195:571–81.[9] Garcia KC, Degano M, Stanfield RL, Brunmark A, Jackson MR,

Peterson PA, et al. Science 1996;274:209–19.[10] Garboczi DN, Ghosh P, Utz U, Fan QR, Biddison WE, Wiley DC.

Nature 1996;384:134–41.[11] Ding YH, Smith KJ, Garboczi DN, Utz U, Biddison WE, Wiley DC.

Immunity 1998;8:403–11.[12] Ding YH, Baker BM, Garboczi DN, Biddison WE, Wiley DC.

Immunity 1999;11:45–56.[13] Brandle D, Burki K, Wallace VA, Rohrer TW, Mak B, Malissen B,

et al. Eur J Immunol 1991;21:2195–202.[14] Brandle D, Birduscha-Riem K, Hayday AC, Owen MJ, Hengartner

H, Pircher H. Eur J Immunol 1995;25:2650–5.[15] Yokosuka T, Takase K, Suzuki M, Nakazawa Y, Taki S, Takahashi

H, et al. J Exp Med 2002;195:991–1001.[16] Wither J, Pawling J, Phillips L, Delovitch T, Hozumi N. J Immunol

1991;146:3513–22.[17] Williams DB, Ferguson J, Gariepy J, McKay D, Teng Y-T, Iwasaki

S, et al. J Immunol 1993;151:3627–37.[18] Wither J, Vukusic B. J Immunol 1999;162:2113–22.

[19] Kabat EA, Wu TT, Perry HM, Gottesman KS, Foeller C.Bethesda, MD: US Department of Health and Human Services,1991.

[20] Machius M, Cianga P, Deisenhofer J, Ward ES. J Mol Biol2001;310:689–98.

[21] Letourneur F, Malissen B. Eur J Immunol 1989;19:2269–74.[22] Vukusic B, Poplonski L, Phillips L, Pawling J, Delovitch T, Hozumi

N, et al. Mol Immunol 1995;32:1355–67.[23] Ballhausen WG, Reske-Kunz AB, Tourvieille B, Ohashi PS, Parnes

JR, Mak TW. J Exp Med 1988;167:1493–8.[24] Veillette A, Bookman MA, Horak EM, Bolen JB. Cell 1988;55:301–

8.[25] Vignali DAA, Strominger JL. J Exp Med 1994;179:1945–56.[26] Vidal K, Daniel C, Hill M, Littman DR, Allen PM. J Immunol

1999;163:4811–8.[27] Grakoui A, Bromley SK, Sumen D, Davis MM, Shaw AS, Allen

PM, et al. Science 1999;285:221–7.[28] Jensen PE. J Exp Med 1991;174:1121–30.[29] Scott CA, Peterson PA, Teyton L, Wilson IA. Immunity 1998;8:319–

29.[30] Jensen PE, Kapp JA. J Immunol 1985;135:2990–5.[31] Reiser J-B, Darnault C, Gregoire C, Mosser T, Mazza G, Kearney

A, et al. Nature Immunol 2003;4:241–7.