Inhibition of a CII-reactiveTcell clone by an anti-CII antibody

Eur. J. Immunol. 1992. 22: 1063-1067

Harald Burkhardt, Tan Yan, Barbara Broker, Annette Beck-Sickinger., Rikard Holmdahle, Klaus Von der Mark and Frank Emmrich Max-Planck-Gesellschaft, Klinische Arbeitsgruppen fur Rheumatologie an der Medizinischen Klinik I11 der Universitat Erlangen-Niirnberg, Erlangen, Institute of Organic Chemistry., University of Tubingen, Tubingen and Department of Medical and Physiological Chemistrye, University of Uppsala, Uppsala

1063

Antibody binding to a collagen type41 epitope gives rise to an inhibitory peptide for autoreactive T cells* It is well documented that antigen recognition by T cells requires small peptides which are generated by protein cleavage in antigen-presenting cells. These peptides have to associate with major histocompatibility complex (MHC) molecules in order to be recognized. An inhibitory peptide may bind to the same site of the MHC-encoded protein but is not recognized by the Tcell. Here we describe a stimulatory and an inhibitory peptide sequence within human collagen type I1 (CII) as defined by means of the same autoreactive human T cell clone. Most interestingly, the inhibitory peptide is not generated by regular processing in peripheral blood mononuclear cells but only in the presence of an antibody that binds to the same domain and thereby seems to protect the inhibitory sequence. This finding may indicate that certain autoantibodies have the potential to block autoreactive T cells with specificity for a distinct epitope on the same antigen.

1 Introduction

2 Materials and methods

Cartilage components are considered a source of antigens that could continously fuel an autoimmune arthritis [l]. Collagen type I1 (CII), being restricted to cartilage, is one of the most interesting candidates for an autoantigen. Human T cell clones with reactivity to human CII (HCII) have been established [2]. To analyze HCII determinants that are relevant t o T cell recognition we have established a humanTcell clone from a healthy blood donor and, for the first time, have characterized the stimulatory epitope.

2.1 Antigens

Some antigenic proteins contain both stimulatory and inhibitory peptide sequences that have been described to compete for binding to MHC molecules but are not recognized by the Tcell receptor [3, 41. In the model systems investigated so far the inhibitory peptides were derived from protein sequences and synthesized in vitro [4]. The inhibitory peptides are either not generated or generated but not binding to MHC molecules during regular antigen processing (for unknown reasons). We report here the inhibitory effect of a monoclonal anti-HCII antibody on the proliferative response of a collagen-specificT cell clone. We provide experimental evidence for the interference of the antibody with the selection of peptide domains for presentation to the Tcell receptor. This may be an additional mechanism of control of autoreactive T cells.

[I 99301

* The Clinical Research Units for Rheumatology at the University of Erlangen-Niirnbergare supported by The German Ministry of Research and Technology. This work was partly supported by Deutsche Forschungsgemeinschaft, SFB 263, project C3. Correspondence: Harald Burkhardt, Max-Planck-Gesellschaft, Klinische Arbeitsgruppen fur Rheumatologie an der Medizinischen Klinik 111 der Universitat Erlangen-Nurnberg, Schwabachanlage 10, D-8520 Erlangen, FRG Abbreviation: CII: Collagen type I1

0 VCH Verlagsgesellschaft mbH, D-6940 Weinheim, 1992

CII was prepared from fetal human epiphyseal cartilage by extraction with guanidinium hydrochloride for removal of proteoglycans, digestion with pepsin and subsequent purification by fractional salt precipitation [5].After cyanogen bromide cleavage the CNBr fragments of the a1 (11) chain (CB8, CB9, CB10, CB11, CB12) were separated by cation exchange chromatography according to Miller and Lunde [6]. Final purification of the CNBr fragments was achieved by fast performance liquid chromatography (FPLC) in 0 . 1 ~ammonium acetate, pH 5 on a Superose 12 column. Peptides were synthesized by the solid-phase method [7] on a SMPS 350 multiple peptide synthesizer (Zinsser Analytic, Frankfurt, FRG) using the Fmoc/tert.butyl strategy and diisopropylcarbodiimide / l-hydroxybenzotriazole activation [8]. After purification by preparative HPLC, the identity of the peptides was proved by amino acid analysis and IonSpray mass spectrometry. 2.2 Cloning of human CII-specific T cells PBMC were separated from heparinized blood by centrifugation on a Ficoll-metrizoate density gradient (Lymphoprep, Nycomed, Oslo, Nonvay).The PBMC were washed in PBS and cultured in RPMI 1640 (Gibco) containing L-glutamine, 25 mM Hepes (Gibco), 100 U/ml penicillin G , 100 pg/ml streptomycin (Gibco), 10% human AB serum (HS) and 10 pg/ml human CII at 37 "C in a humidified 5% C 0 2 atmosphere. After 3 days of culture, 50 U/ml human rIL-2 (HrIL-2, Eurocetus, Amsterdam) were added. After an additional 7 days of culture the cells were collected and separated by centrifugation on a Ficoll-metrizoate density gradient to remove cell debris. They were plated in 96-well round-bottom microculture plates (Nunc, Roskilde, Denmark) at cell numbers ranging from 156 to 20000 cells/well (24 replicates each) with autologous irradiated (3000 rad) PBMC at 5 x lo4cells/well as antigen-presenting cells, and 0014-2980/92/0404-1063$3S O + ,2510

1064

H. Burkhardt, T. Yan, B. Broker et al.

10 pg/ml human CII in RPMI 1640 containing 10% HS. After 3 days, the cultures were supplied with 50 U/ml HrIL-2. Ten days later, the cells were washed, split and tested for their proliferative response to autologous irradiated PBMC with or without addition of 10 yglml human CII .

-

Eur. J. Immunol. 1992. 22: 1063-1067 CPM x 10 -3

. 0

HCII DN-HCII CBll

CB8 I Human CII-specific T cell lines were purified over a Ficoll CB9,12 gradient and cloned by limiting dilution. Cell numbers CBlO I ranging from 10 to 0.5 per well were plated in 96-well PEP24 I round-bottom microculture plates with 5 x lo4 autologous PEP25 irradiated PBMC and 10 pg/ml CII in RPMI 1640 containing 10% HS. After 3 days of incubation at 37 "C, 5% COz, 50 U/ml HrIL-2 was added and cultures were subsequently supplied with fresh HrIL-2 medium if necessary. After 14 days, the cloned cells were washed, split and tested for their 0 proliferative response to CII. or1 (11)

'

CNBrfragments

2.3 T cell proliferation assay T cells were examined for collagen-induced proliferation by a [3H]thymidine incorporation assay. Prior to testing, they were cultured without CII, HrIL-2, or autologous irradiated PBMC for 5 to 7 days in order to remove CII and to increase their specific response. T cells were tested in 96-well round-bottom microculture plates at concentration of 1 x lo4 cells/well together with 5 x lo4 cells/well autologous irradiated PBMC and antigen. For the analysis of MHC restriction, a monoclonal antibody specific for MHC class I1 (109d6, kindly provided by Dr. R.Winchester, N.Y.) was used. After 3 days of incubation, 0.5 pCi/well [3H]thymidine (18.5 kBq, Amersham, Braunschweig , FRG) was added for a further 12 h. The cultures were then harvested onto glass fiber filters with an automatic multiple harvester (PHD System, Cambridge, MA) and the incorporated radioactivity was assessed by liquid scintillation counting. The results were expressed as mean cpm 5 SD of duplicate cultures. In some experiments antigen (HCII) and anti-CII mAb were used as complexes coupled to microspheres which can be ingested by blood monocytes. For this purpose 10 pg HCII were preincubated with the mAb (2 pg) for 2 h at room temperature. The mAb-HCII complexes were precipitated by DynabeadsTM(4 pm) coated with goat anti-mouse IgG (Dynal via Dianova, Hamburg, FRG). The loaded microspheres were extensively washed and used for the proliferation assays at a concentration of 5 x i04/weii. Uptake by monocytes was examined microscopically.

3 Results and discussion 3.1 Fine specificity of a Tcell for human CII A human CD4+ T cell clone (TC9) was established from a healthy blood donor. It is inhibited by an anti-MHC class I1 antibody and uses HLA-DR7 as restriction element. TC9 is unreactive to human collagen type 111, IV, V and VI but shows minor cross-reaction to human collagen type I [9]. As shown in Fig. 1, theT cell epitope is located in the cyanogen bromide fragment 11(CB11) of HCII and could be mapped by sequential testing with enzymatically cleaved CII fragments and synthetic peptides [9]. The whole stimulatory

8

4

0

12

I

16

20 I

I

GEPGPAGPQWGPA

-i-

124

402

552

-

897

1014 aa

1 length

100

150

200

250

300 m

I

Figure 1. Tcell epitope localization on human CII. Shown are the proliferative Tcell responses to native HCII, heat-denatured CII (DN-HCII), cyanogen bromide fragments of HCII: CB 8, CB 9, CB 10, CB 11, CB 12, and two synthetic peptides (PEP 24 25) with the sequences of the triple helical region of HCII: PEP 25 = Gly-Glu-Pro-Gly-Pro-Ala-Gly-Pro-Gln-Gly-AlaPro-Gly-Pro-Ala (aa 271-285, shown in one letter code) and PEP 24 = Gly-Leu-Ala-Gly-Pro-Lys-Gly-Ala-Asn-Gly-Asp-Pro-GlyArg (aa 337-350). The binding sites for antibody C1 and D3 were determined according to [lo] and [12] and are depicted in the figure.

+

activity for TC9 resides within the sequence Gly-GluPro-Gly-Pro -Ala-Gly-Pro-Gln-Gly -Ala-Pro-Gly-Pro- Ala corresponding to amino acid residues (aa) 271-285 of the triple helical region of HCII. This sequence is unique for HCII. All experiments described below were performed with clone TC9. 3.2 Fine specificity of antibodies to human CII To investigate whether HCII-specific antibodies would influence the proliferative response of TC9 we used two murine monoclonal antibodies (C1 and D3). Both are of the same isotype (IgGZ,, x ) and bind to different epitopes on CII [ll].Although generated by immunizing with chicken CII both antibodies bind efficiently to HCII. The antibody epitopes were further characterized by statistical analysis of the distances from the antibody binding site to the N terminus in electronmicrographs after rotary shadowing of 70 antigen antibody complexes [12]. The localization of the binding domains of antibody C l to C B l l (89nm k 8 nm from the amino-terminal end) has been published recently [12]. Antibody D3 binds to CBlO (189 nm k 9 nm) which is shown by staining a Western blot of the corresponding cyanogen bromide fragment together with the statistical analysis of the rotary shadowing experiment (Fig. 2). A partial renaturation of the fragments after blotting on a nitrocellulose membrane was required to see

z o r

-

1065

Inhibition of a CII-reactive Tcell clone by an anti-CII antibody

Eur. J. Immunol. 1992. 22: 1063-1067

50

1 DB+DN-HCII D3+HCll

30 Cl+DN-HCII

I NH2 -+4,6,12l

11

I 8

1

I

k.,*i 10

CNBr

-

9

fragment

Figure2. Average distance of the D3 epitope in nm from the amino end of HCII. Enlarged photographs of rotary shadowed preparations of collagen-antibody complexes (n = 72) were used to determine the position of the D3 binding site (method described in [12]).The binding site is 189 nm + 9 nm along the triple helix. For comparison the position of the cyanogen bromide fragments is shown. The binding region is located in the CB 10 fragment. The insert of the figure shows a Western blot with the positive staining for the CB 10, whereas the other fragments (their typical localization on the filter was identified by Ponceau staining after blotting) remained negative. The additional band ( 0 )is due to incomplete CNBr cleavage resulting in a fragment of higher M, than CB 10.

-6

0 : . I -0.5 0.0

noantigen .

I

0.5

1

Cl+HCII

I

.

.

1.0

I

1.5

antibody

.

,

.

2.0

I

I

.

2.5

3.0

(pglrni)

Figure 3. Influence on the proliferative T cell response by monoclonal antibodies to HCII. HCII or DN-HCII (10 pg) were preincubated with varying amounts (0.25-2 pg/ml) of C1 or D3 antibody for 1h before addition to autologous mononuclear cells and to the T cell clone TC9.

the lack of inhibition by D3 could be confirmed (Fig. 4). It should be noted that antigen presentation by the phagocytosing mononuclear cells was impaired as can be seen by the reduced proliferative response to 10 pg HCII (compare Figs. 1 and 4).

binding as both antibodies C1 and D3 do not bind to heat-denatured antigen (> 50°C, 1 h) in solution [12]. Using tryptic fragments of cross-reactive chicken and rat C B l l we could identify an immunoreactive peptide for antibody C1 which was isolated by HPLC and sequenced on a gas phase sequencer [12]. By sequence comparison with the human al(I1) chain [13] the relevant binding region for antibody C1 was localized to aa 316-333, which matches precisely the localization predicted by rotary shadowing (90 nm, corresponding to aa 310 k 30).

Consequently, the critical event leading to T cell inhibition should take place inside the antigen-presenting cells. In the presence of antibody C1 directed against sequence 316-333, the stimulatory peptide sequence 271-285 could have been processed in an abnormal way, being either totally digested or inefficiently processed. As an alternative explanation we considered the generation of an inhibitory peptide that would be degraded during antigen processing in the absence of the antibody C1.

3.3 Modulation of the proliferative T cell response to CII

To test this hypothesis we purified the tryptic fragment of CBll containing the C1 binding site as recently described

Fig. 3 shows the effect of both antibodies on the Tcell response to native and denatured HCII. Antibody C1 was a very efficient inhibitor whereas antibody D3 did not affect CTT-induced T cell proliferation. However, T cell proliferation remained unaffected after heat denaturation of CII prior to the addition of the antibody C1. Two conclusions can be drawn from these observations: (a) a direct effect of antibody C1 alone on either the antigen presenting cells or theT cells themselves is excluded and (b) inhibition of T cell proliferation by C1 requires native triple helical HCII. Moreover, the inability of the antibody D3 to exert an inhibitory effect like that of C1 reveals that a distinct epitope on HCII seems to be crucial for efficient Tcell blocking. In addition, this negative result with the isotypematched antibody D3 demonstrates that inhibition of antigen uptake after formation of CII-antibody complexes does not explain the C1 effect. By coupling the CII-C1 complex t o anti-mouse IgG-coated microbeads (diameter 4 pm) we could make sure by light microscopic control that endocytosis of the complex together with the beads was completed. Under these experimental conditions the inhibitory effect of the antibody C1 on Tcell proliferation and

F

C1+HCII HCll

0

1000

2000

3000

4000

5000

CPM

Figure 4. Proliferative T cell response of clone TC9 to antibodyHCII complexes coupled to Dynabeads (details of the coupling are gwen in Sect. 2.3). The first column (0) is the background stimulation by the goat anti-mouse IgG-coated beads (5 x 104/well)without HCILThe proliferative response to 10 pg HCII in the presence of the beads is the positive control (HCII). A marked decrease in [3H]thymidine incorporation is shown in the ease of C1-HCII complex-loaded Dynabeads (C1 + HCII), whereas D3-HCII-coated beads (D3 + HCII) cause a proliferative response in the range of the positive control.

H. Burkhardt, T. Yan, B. Broker et al.

1066

[12]. This fragment was able to inhibit the proliferative Tcell response to HCII (Fig. 5 ) . Subsequently, synthetic peptides were obtained and tested for their inhibitory potential. A peptide covering the sequence of the C1 binding region as well as smaller peptides representing its amino- and carboxy-terminal ends were compared with control peptides derived from collagen sequences. As shown in Fig. 5 the complete peptide (aa 316-333: GlyP he-Pro-Gly-Gln-Asp-Gly-Leu-Ala-Gly-Pro-Lys-Gly-~aPro-Gly-Glu-Arg), as well as its carboxy-terminal part (Ala-Gly-Pro-Lys-Gly-Ala-Pro-Gly-Glu-Arg) were able to inhibit the Tcell response to CII. This inhibition is very efficient and reached already 75% at a 2.5-fold molar excess of the peptide (0.5 pg of the 18mer in Fig. 5 ) over the a1 chain of CII. With the more restricted decapeptide a complete inhibition was found at a 10-50-fold molar excess (equivalent to 5 pg of the peptide). These results meet the ranges of molar excess required for highly effective inhibitory peptides capable of inhibiting T cell recognition and binding to MHC molecules [3, 141. Taken together, our data show a squence representing a very efficient inhibitory peptide for TC9, which is located carboxy terminal from the stimulatory epitope within CB11. However, during regular antigen processing of HCII by blood mononuclear cells such a peptide is either not generated or does not become effective. Inhibition of aTcell response by high concentrations (>50 pglml) of antibodies directed against the stimulatory T cell epitope has been reported previously [15, 161. Since in these studies T and Bcell epitopes were identical, the

Eur. J. Immunol. 1992. 22: 1063-1067

antibodies very likely have interfered with the binding of the peptide to MHC molecules or with the binding of the Tcell receptor to the peptidelMHC complex. This is different from our system because the antibody C1 does not bind to the stimulatory Tcell epitope but is directed to a domain remote from it. The hypothesis of determinant protection by MHC has been proposed to explain why peptides escape total enzymatic degradation during antigen processing [17]. Antibodies can protect their epitopes from enzymatic degradation, as has been demonstrated by Manca et al. [18] who showed protection of P-galactosidase from pepsin digestion using monoclonal antibodies. In a similar approach proteolysis of the antibody-antigen complexes has been utilized to map protected antibody epitopes of cytochrome c [19]. It is therefore possible that the antibody C1 protects its epitope from degradation in the endosomes thus leading to generation of the inhibitory peptide. It has already been discussed that antibodies may either lead to positive selection of their epitope during processing or conversely inhibit epitope presentation to the Tcell receptor by steric hindrance of peptide binding to MHC molecules [18,20,21]. Steric hindrance by antibody C1 of binding of the stimulatory peptide (aa 271-285) t o the MHC molecule is not very likely because of a distance of 30 amino acid residues on the rod-like and rather rigid CII molecule between the stimulatory epitope and the C1 binding site (aa 316-333). This distance corresponds approximately to 8.5 nm, which is about 3.5 times the total length of the binding groove of a HLA molecule (2.5 nm, [22]). Moreover, we could demonstrate that only the C1 epitope but not other CII-derived sequences were able to inhibit clone TC9 (Fig. 5). Therefore, we suggest the

cornpelillan

20pg1rn1

I

II

1

CPM x 10

-3

Figure 5. Inhibition of the proliferative Tcell response of clone TC9 by a tryptic CBll fragment and synthetic peptides. The first column (0) represents the background stimulation in the absence of HCII, whereas the second column is the positive control of TC9 proliferation to 10 pg/ml HCII. The tryptic peptide (aa 316-333) containing the C1 binding site was purified as described previously [ 121. Several synthetic peptides (sequences are shown in one-letter code) were derived from the CB11 sequence of HCII including the sequence of the C1 binding site (aa 316-333: Gly-Phe-Pro-Gly--

Figure 6. Hypothetical explanation of the suppressive effect of antibody C1 on stimulation of HCII-specific Tcell clone TC9 by Gln-Asp-Gly-Leu-Ala-Gly-Pro-Lys-Gly-Ala-Pro-Gly-Glu-Arg). native antigen. We hypothesize that the antibody protects its For a more precise investigation the decapeptide covering the epitope on CII from lysosomal degradation in the antigencarboxy terminal part of this sequence (aa 324-333: Ala-Gly- presenting cells. The protected peptide competes with a partially Pro-Lys-Gly-Ala-Pro-Gly-Glu-Arg) was compared to the amino homologoous peptide within the CB11 sequence for binding to terminal part (aa 316-322: Gly-Phe-Pro-Gly-Gln-Asp-Gly) and MHC class 11, but is not recognized by the Tcell receptor. In the some control peptides with sequence similarities. For the Tcell absence of the C1 antibody, the C1 epitope is degraded intracelproliferative assays 10 pg/ml HCII was used and the synthetic lularly and thus cannot compete out the stimulatory Tcell epipeptides were added at concentrations indicated. tope.

Eur. J. Immunol. 1992. 22: 1063-1067

Inhibition of a CII-reactive Tcell clone by an anti-CII antibody

generation of an intrinsic inhibitory peptide by protection of a cryptic determinant as explanation of our data and as a novel possibility of altered processing (Fig. 6). Once generated, the inhibitory peptide would compete out the stimulatory epitope for binding to HLA-DR7. Therefore, experiments were performed to demonstrate the competitive binding to MHC on autologous non-fixed monocytes by incubation with radiolabeled peptide 271-285 according to Ceppelini et al. [23]. However, peptide binding to MHC was not detectable above background indicating low affinity of the antigen for MHC as was reported in several other systems [24, 251. The results do not prove whether after ingestion of the CII-C1 complex the macrophage MHC contained the inhibitory peptide. However, comparison of the inhibitory decapeptide domain with the corresponding stimulatory sequence revealed that 7 out of 10 amino acid residues are identical and may provide a common motif for binding to MHC class I1 molecules (as boxed in Fig. 6). Moreover, this common motif occurs only at those two locations where it was identified experimentally, despite of a relatively simple and repetetive structure (Gly-x-y) and a considerable length (1014 aa) of HCII.

4 Concluding remarks In the past years the interest in defined blocking peptides for T cell recognition has been growing considerably. Here we describe an inhibitory peptide derived from the CII sequence which does not seem to be generated during regular stimulation of T cells. Inhibition of a CII-reactive Tcell clone was seen, however, if an antibody to the corresponding CII sequence was present during processing. The inhibitory peptide is similar in sequence to the stimulatory epitope for the Tcell clone. We have defined and synthesized the stimulatory as well as the inhibitory peptide. We suggest protection by the antibody of the inhibitory peptide during antigen processing thus allowing appearance of the inhibitory peptide on the antigenpresenting cell. It remains to be established whether this epitope protection by an autoantibody may play a regulatory role for the immune response to putative autoantigens. We thank L. Serokin for reading the manuscript and E. Bauer for excellent technical assistance.

Received September 23,1991; in final revised form December 19, 1991.

1067

5 References 1 Holmdahl, R., Andersson, M., Goldschmidt,T. J., Gustafsson, K., Jansson, L. and Mo, J. A., lrnrnunot. Rev. 1990. II8: 193. 2 Londei, M., Savill, C. M.,Verhoef, A., Brennan, F., Leech, Z . A., Duance,V., Maini, R. N. and Feldmann, M., Proc. Natl. Acad. Sci. USA 1988. 86: 636. 3 Adorini, L., Appela, E., Doria, G. and Nagy, Z . A., J. Exp. Med. 1988. 168: 2091. 4 Gammon, G., Shastri, N., Cogswell, J., Wilbur, S., SadeeghNasseri, S., Krzych, U., Miller, A. and Sercarz, E., Imrnunol. Rev. 1987. 98: 53. 5 Von der Mark, K. ,Van Menxel, M. and Wiedemann, H., Eur. J. Biochem. 1982. 124: 57. 6 Miller, E. J. and Lunde, L. G., Biochemistry 1973. 12: 3153. 7 Merrifield, R. B., J. Am. Chem. SOC. 1963. 85: 2149. 8 Schnorrenberg, G. and Gerhardt, H., Tetrahedron 1989. 45: 7759. 9 Yan,T.,Burkhardt,H., Ritter,T., Broker, B.,Mann, K. H.,Von der Mark, K. and Emmrich, E , Eur. J. Immunol. 1992. 22: 51. 10 Shotton, D. M., Burke, B. and Branton, D., J. Mol. Biol. 1979. 131: 303. 11 Holmdahl, R., Rubin, K., Klareskog, L., Larssen, E. and Wigzell, H., Arthr. Rheum. 1986. 29: 400. 12 Burkhardt, H., Holmdahl, R., Deutzmann, R., Wiedemann,

H.,Von der Mark, H., Goodman, S. and Von der Mark, K., Eur. J. Immunol. 1991. 21: 49. 13 Su, M. W., Lee, B., Ramirez, F., Machado, M. and Horton,W., Nucleic Acids Res. 1989. 17: 9473. 14 Buus, S., Sette, A,, Colon, S. M., Miles, C. and Grey, H. M., Science 1987. 235: 1353. 15 Corradin, G. and Engers, H. D., Nature 1984. 308: 547. 16 Lamb, J. R., Zanders, E. D., Lake, P.,Webster, R. G., Eckels, D. D. ,Woody, J. N., Green, N., Lerner, R. A. and Feldmann, M., Eur. J. Imrnunol. 1984. 14: 153. 17 Donermeyer, D. L. and Allen, P. M., J. Immunol. 1989. 142: 1063. 18 Manca, F., Fenoglio, D., Kunkl, A., Cambiaggi, C., Sasso, M. and Celada, F. J., J. Immunol. 1988. 140: 2893. 19 Jemmerson, R. and Paterson,Y., Science 1986. 232: 1001. 20 Berzofsky, J. A., Surv. Immunol. Rev. 1983. 2: 223. 21 Ozaki, S. and Berzofsky, J. A., J. Imrnunol. 1987. 138: 4133. 22 Bjorkman, I? J., Saper, M. A., Samraoui, B., Bennett, W. S., Strominger, J. L. and Wiley, D. C., Nature 1987. 329: 506. 23 Ceppellini, R., Frumento, G., Ferrara, G. B.,Tosi, R., Chersi, A. and Pernis, B., Nature 1989. 339: 392. 24 Shimonkevitz, R., Colon, S., Kappler, J. W., Marrack, P. and Grey, H. M., J. Imrnunol. 1984. 133: 2067. 25 Babbit, B. l?, Allen, P. M., Matsueda, G., Haber, E. and Unanue, E. R., Nature 1985. 317: 359.