Peptide Motif II Molecule Identifies an Immunogenic Type I Diabetes ...

Determination of Glutamic Acid Decarboxylase 65 Peptides Presented by the Type I Diabetes-Associated HLA-DQ8 Class II Molecule Identifies an Immunogenic Peptide Motif1 Ann E. Herman,* Roland M. Tisch,† Salil D. Patel,† Sarah L. Parry,† Jennifer Olson,†‡ Janelle A. Noble,¶i Andrew P. Cope,† Brett Cox,† Mauro Congia,† and Hugh O. McDevitt2†§ Particular HLA class II allelic sequences are associated with susceptibility to type I diabetes. To understand the mechanism, knowledge of the molecular nature of the specific TCR/peptide/class II interactions involved in the disease process is required. To this end, we have introduced the diabetes-associated human class II HLA-DQ8 allele (DQA1*0301/DQB1*0302) as a transgene into mice and analyzed T cell responses restricted by this molecule to an important Ag in human diabetes, human glutamic acid decarboxylase 65. Hybridomas were used to determine the particular peptides from this Ag presented by HLA-DQ8 to T cells and to map the core minimal epitopes required for T cell stimulation. Analysis of these core epitopes reveals a motif and relevant features for peptides that are immunogenic to T cells when presented by HLA-DQ8. The major immunogenic epitopes of glutamic acid decarboxylase 65 do not contain a negatively charged residue that binds in the P9 pocket of the HLA-DQ8 molecule. PBMC from HLA-DQ81 diabetic and nondiabetic individuals respond to these peptides, confirming that the mouse model is a useful tool to define epitopes of autoantigens that are processed by human APC and recognized by human T cells. The Journal of Immunology, 1999, 163: 6275– 6282.

T

ype I diabetes is a disease characterized by autoimmune destruction of the b cells in the pancreatic islets, resulting in a lack of ability to produce insulin (1). The disease has a complex etiology involving genetic susceptibility and environmental factors. Over 19 loci are found to correlate with susceptibility, the most striking genetic association being with the individual’s HLA class II genotype (2, 3). The HLA class II a- and b-chains form a peptide binding groove that presents peptides to CD41 T cells (4). Analysis of the gene sequences encoding these chains in diabetic individuals revealed that the strongest association with disease lies in the presence or the absence of an aspartic acid residue at position 57 of the HLA-DQ b-chain (b57),3 (5). Alleles encoding a noncharged serine, valine, or alanine residue at this position, such as the HLA-DQ2 (DQ2) and HLA-DQ8 (DQ8) alleles, confer susceptibility to type I diabetes, while closely related alleles, such as HLA-DQ7 (DQ7) and HLA-DQ9 (DQ9) encoding b57 aspartic acid, are not associated with susceptibility. *Program in Immunology, Departments of †Microbiology and Immunology, ‡Pediatrics, and §Medicine, Stanford University School of Medicine, Stanford, CA 94305; ¶ Children’s Hospital Oakland Research Institute, Oakland, CA 94609; and iDepartment of Human Genetics, Roche Molecular Systems, Alameda, CA 94501 Received for publication June 23, 1999. Accepted for publication September 14, 1999. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 This work was supported by Grants DK51667 and DK53005 from the National Institutes of Health. A.E.H. was supported by a predoctoral fellowship from the National Science Foundation. 2

Address correspondence and reprint requests to Dr. Hugh McDevitt, D345 Fairchild Building, Stanford University, Stanford, CA 94305-5124. E-mail address: [email protected]

3 Abbreviations used in this paper: b57, HLA-DQ b-chain position 57; DQ2, HLADQ2 (DQA1*0501/DQB1*0201); DQ7, HLA-DQ7 (DQA1*0301/DQB1*0301); DQ8, HLA-DQ8 (DQA1*0301/DQB1*0302); DQ9, HLA-DQ9 (DQA1*0301/ DQB1*0303); GAD65, glutamic acid decarboxylase 65; I-Ab0/0, homozygous targeted disruption of the I-Ab locus; NOD, nonobese diabetic; non-Tg, nontransgenic; Tg, transgenic.

Copyright © 1999 by The American Association of Immunologists

Exceptions to this have been noted, and other residues in HLA-DR as well as HLA-DQ molecules have been shown to play a critical role in susceptibility, for example in other ethnic groups (1, 2). The observation that specific allelic polymorphisms in an HLA molecule lead to disease susceptibility suggests that the role of these molecules in presenting unique sets of peptides to T cells could be the critical factor for determining susceptibility. Class II/self peptide complexes are involved in both positive and negative selection events in the thymus, where particular allele differences could lead, for example, to a lack of negative selection of autospecific CD41 T cells (6). Alternately, diabetes-associated HLA class II molecules might present distinct disease-promoting self peptides to CD41 T cells in the periphery (7). The P1 through P9 pockets in an HLA class II molecule correspond to the anchoring residues of the typically 9-aa core peptide required for binding to the molecule (8). By homology to the known crystal structure of HLA-DR1, an allele without an aspartic acid residue at b57 is unable to form a salt bridge with arginine at position 76 of the DR a-chain (9). This salt bridge contributes to the shape of the P9 pocket. The lack of this salt bridge has been postulated to affect disease susceptibility by selecting for negatively charged residues in the peptide at P9 (10, 11). Such residues in the peptide might substitute for the lack of aspartic acid at b57 by replacing this salt bridge and conferring increased molecular stability. Although peptide motifs for the DQ8 molecule have been defined by elution studies (12, 13), specific autoantigenic peptides recognized by human T cells in the context of the DQ8 molecule are poorly defined. Due to the difficulty of obtaining DQ8-restricted human T cell clones, no clones specific for autoantigens such as glutamic acid decarboxylase 65 (GAD65) are available (14). GAD65 is a protein expressed in neural and pancreatic tissue and is a target of the autoimmune process in type I diabetes (15). Several groups have been able to show that bulk cultures of human PBMC from diabetic individuals respond to GAD65 (16). The importance of GAD65 in the disease process is evidenced by the 0022-1767/99/$02.00

6276

GAD65 PEPTIDES PRESENTED BY DIABETES-ASSOCIATED HLA-DQ8

presence of GAD65 autoantibodies in .70% of patients (17–19). Autoantibodies to GAD65 in combination with anti-IA-2 and antiinsulin Abs in individuals with a susceptible HLA haplotype have a high predictive value for subsequent development of diabetes. To analyze peptides from human GAD65 presented by the DQ8 molecule, we have studied the T cell response to this protein in a DQ8 transgenic (Tg) mouse model. In this system, the Tg DQ8 molecule is the only class II restriction element due to targeted mutation of the mouse class II I-Ab locus (I-Ab0/0), and a natural deletion in the I-E a gene promoter region. HLA Tg mice have been used successfully in other systems to analyze responses restricted by human class II molecules in lieu of studying responses directly in humans (20). DQ8 Tg mice have been used as a model for collagen-induced arthritis and to generate GAD65-specific mouse T cell clones (21, 22). An initial characterization of regions of GAD65 presented by the DQ8 molecule in Tg mice was published during preparation of this manuscript (23). We independently generated a similar model to determine GAD65-derived peptides that are processed and presented by the DQ8 molecule and are immunogenic to T cells. Furthermore, we have performed careful analyses of the GAD65 epitopes determined in DQ8 Tg mice, leading to the identification of an unexpected peptide motif without a negatively charged residue at P9 of the peptide. Additionally, we observed special features for these peptides, including a 10 mer minimal core epitope. To validate that these studies identify epitopes relevant to human type I diabetes, we showed that human PBMC from DQ81 diabetic and nondiabetic individuals proliferate to the human GAD65 epitopes defined in our DQ8 Tg mouse system.

Materials and Methods Production and characterization of transgenic mice A DNA construct for Tg expression of DQ8 was generated using genomic sequences for the DQA1*0301 (provided by Dr. Jack Strominger) and DQB1*0302 genes, each under the control of the I-E a promoter in the vector pSV5neo (provided by Drs. Diane Mathis and Christophe Benoist, Strasbourg, France). The I-E a promoter, in which SalI and XhoI sites were destroyed, was engineered upstream of the genomic sequence of DQA1*0301 and DQB1*0302. These two fragments were then subcloned into pSV5neo using engineered 59 XhoI and 39 SalI sites (see Fig. 1A). The XhoI fragment containing the promoters and transgene sequences was microinjected into (FVB 3 nonobese diabetic (NOD))F1 mouse embryos according to standard Tg mice production protocols (24). Eight separate founder Tg mice expressing DQ8 on the surface of PBL were identified by isolating mouse PBMC on a Lympholyte-M gradient (Cedarlane, Ontario, Canada), staining with a pan-anti-HLA-DQ Ab (Leu 10, Becton Dickinson, San Jose, CA), and analyzing cells by flow cytometry. Cells from spleen, lymph node, and thymus were also analyzed by flow cytometry (FACScan, Becton Dickinson, San Jose, CA). Mice were sacrificed, and organs were removed and prepared as single-cell suspensions. The cells were preblocked with FcBlock (PharMingen, San Diego, CA); stained with antiHLA-DQ and/or Abs specific for I-Ag7 (OX6), B220 (RA3-6B2), CD4 (GK1.5), or CD3 (145-2C11; PharMingen, San Diego, CA); and analyzed as described above. Tg mice (at N5 backcross to the NOD background) were crossed to I-Ab0/0 mice on the B6 background (provided by Drs. Diane Mathis and Christophe Benoist), and intercrossed to obtain mice expressing human Tg DQ8 as the only class II molecule. Mice were screened for this phenotype by flow cytometric analysis of PBMC, stained with anti-HLA-DQ or Abs specific for I-Ag7 and mouse CD4, as described above.

Production of human GAD65 Recombinant human GAD65 was produced using a baculovirus expression system (Invitrogen, San Diego, CA) as previously described (25). Histidine 6-tagged human GAD65 was purified using NTA-Ni21 agarose (Qiagen, Valencia, CA). For immunization, the protein was purified further by gel electrophoresis and electroelution of the 65-kDa species. Purified protein was dialyzed into PBS for injection and/or hybridoma screening.

Generation and screening of GAD-specific T cell hybridomas (NOD 3 B6)F2, homozygous DQ8 Tg, I-Ab0/0 mice were immunized in each hind footpad and at the base of the tail with 50 mg of purified recombinant human GAD65 emulsified in IFA (Difco, Detroit, MI). T cell hybridomas were generated from the draining lymph nodes, as previously described (25, 26), and propagated in complete medium: RPMI 1640, 10% FCS, 10 mM HEPES, 2 mM L-glutamine, 100 U/ml each of penicillin and streptomycin, and 50 mM 2-ME, with 13 hypoxanthine/aminopterin/thymidine (Life Technologies-BRL, Gaithersburg, MD). Cells were first screened for specificity to whole human GAD65 by culturing T cell hybridomas and APC with 20 mg/ml human GAD65 or medium only, as previously described (25). Human EBV-transformed B cells of the Priess line, homozygous for DQ8, were used for APC. Supernatants were tested by immunoassay for IL-2 production, using a europium-based detection system (Wallac, Gaithersburg, MD), (25). Cells specific for whole human GAD65 were then tested for peptide specificity in the same assay, using peptide in pools of 12 or singly at 6.7 mg/ml instead of whole Ag. Ab blocking studies were performed by adding 50 mg/ml IVD12 (American Type Culture Collection, Manassas, VA), specific for DQ8, to the cultures of hybridomas, APC, and specific peptide described above and testing for IL-2 production.

Peptides Peptides were synthesized using the PIN method to generate 117 15 mers overlapping by 10 aa, spanning the entire 585-aa sequence of human GAD65 (Chiron Technologies, San Diego, CA), as previously described (25). Truncation peptide sets of 11 mers shifting the register by 1 aa were produced in the same way.

Determination of core minimal epitope for hybridoma stimulation Truncation sets of 11 mer peptides were included in cultures of T cell hybridomas and irradiated human DQ81 Priess APC as described above. IL-2 production was assayed using eight concentrations of peptide at 4-fold dilutions ranging from 0 –100 mM, in duplicate wells for each experiment.

Analysis of responses from human peripheral blood Blood was isolated from recent-onset diabetic patients at the Stanford Pediatric Endocrinology Clinic headed by Dr. Darrell Wilson at their first clinical visit after diagnosis or from healthy control volunteers. PBMC were isolated by density centrifugation. Blood was layered 1/1 on a Histopaque (Sigma, St. Louis, MO) gradient and centrifuged at 2200 rpm for 20 min at room temperature. Cells were harvested and washed in complete medium (as described for hybridomas, except without FCS). Cells were then incubated with autologous serum at 1 3 106 cells/ml in 1.0-ml cultures in the presence or the absence of human GAD65 peptides at 20 mg/ml or with 20 mg/ml tetanus Ag as a positive control. After 5 days, wells were split into five wells of 250 ml, and 1 mCi/well of [3H]thymidine was added to the cultures for 8 –12 h. Cells were harvested, and counts were determined on a Tomtec harvester (Orange, CT). Data were calculated as the mean 6 SD of five replicate cultures. HLA typing was performed using a PCR method, as previously described (27).

Results Generation and characterization of DQ8 Tg mice Transgenic mice were produced that express the diabetes-associated DQ8 molecule. The genomic sequences containing the DQA1*0301 and DQB1*0302 genes were each engineered to be under the control of the mouse I-E a promoter (Fig. 1A). This approach was taken after generation of DQ Tg mice using human genomic promoters led to inappropriate expression of human class II molecules (data not shown). Spleen lymphocytes from Tg mice established using the I-E a promoter-DQ8 construct were analyzed by flow cytometry for HLA-DQ and mouse class II I-A expression (Fig. 1B). DQ8 Tg mice, but not nontransgenic (non-Tg) littermates, express HLA-DQ on the same cells that express endogenous mouse class II molecules. We tested spleen, thymus, lymph node, and peripheral blood and found the DQ8 product expressed on the same cells as mouse class II I-A (Fig. 1B; data not shown). DQ8 Tg mice at the N5 backcross to the NOD strain were crossed to B6 strain I-Ab0/0 mice. Heterozygotes were intercrossed

The Journal of Immunology

FIGURE 1. DQ8 Tg mice were established that express the human DQ8 molecule on cells that normally express mouse class II I-A. A, The DNA construct used to generate Tg mice is shown. An XhoI fragment, containing the genomic sequences of the DQA1*0301 and DQB1*0302 genes, each controlled by an I-E a promoter, was injected into (NOD 3 FVB)F1 embryos. Three Tg lines were established from eight founders. B, Flow cytometric analysis of splenic lymphocytes from DQ8 Tg mice (right panel) and non-Tg littermates (left panel) stained with Abs specific for HLA-DQ and the mouse I-A molecule (I-Ag7 in the NOD strain). C, Flow cytometric analysis is shown for PBL from a Tg DQ8, I-Ab0/0 mouse stained with Abs specific for HLA-DQ. PBMC were isolated and prepared as described in Materials and Methods. D, Expression of mouse CD4 vs I-A in PBL from the same mouse as that in C is shown.

to produce (NOD 3 B6)F2 I-Ab0/0 pups with homozygous Tg DQ8 as the only class II molecule. Flow cytometric analysis of PBMC from a representative mouse confirms the DQ8 Tg phenotype (Fig. 1C). PBMC from the same mouse were analyzed for mouse class II expression (I-A) and mouse CD4, showing that the human DQ8 Tg is able to select mouse CD41 T cells in the absence of mouse class II (Fig. 1D). The addition of a human CD4 transgene to the DQ8 Tg mice described above did not enhance the number of mouse CD41 T cells in peripheral blood (data not shown; human CD4 Tg mice provided by Dr. Dan Littman). In the (NOD 3 B6)F2 background, CD41 T cells make up 20 –30% of PBL in homozygous Tg DQ8, I-Ab0/0 animals (Fig. 1D). Determination of human GAD65 epitopes in Tg mice We have used the Tg mice described above as a means to identify the molecular nature of the DQ8-restricted T cell response to human GAD65, an important diabetes autoantigen expressed in the pancreas. DQ8 Tg I-Ab0/0 mice (see Fig. 1, C and D) were immunized with recombinant human GAD65 protein. Responding T cells were used to generate T cell hybridomas, which were screened for reactivity to whole human GAD65 protein. Different hybridoma clones responded to whole GAD65 with IL-2 production ranging from 1.5- to 9-fold over medium alone (data not shown). Specific epitopes were determined by testing for IL-2 pro-

6277

FIGURE 2. DQ8-restricted human GAD65-specific hybridomas recognize at least nine epitopes spanning almost the entire molecule. Thirty-five GAD-specific, DQ8-restricted T cell hybridomas were generated from two Tg DQ8, I-Ab0/0 mice and tested for increased IL-2 production in the presence of Ag. T cells were cocultured with DQ81 human APC in the presence of whole human GAD65, 1 of 10 pools of GAD65 peptides, or medium alone, and IL-2 production was measured, as described previously (25). A positive response to Ag was considered to be a 2-fold or greater increase in IL-2 production over medium and APC alone. The number of T cell hybridoma clones responding to each of 10 pools of human GAD65 peptides is plotted.

duction by the GAD-specific hybridomas in response to 15 mer peptides from a peptide set spanning the whole protein. The set consisted of 15 mer peptides overlapping by 10 residues to represent every possible 10 mer epitope in the protein. A total of 35 DQ8-restricted hybridomas was obtained from two independent fusion events. The panel of hybridomas represents reactivity to at least nine epitopes within human GAD65 (Fig. 2). Most clones respond to one of three epitopes, in regions 51–120, 111–180, or 521–585. T cell responsiveness was identified in multiple regions of the protein. Six epitopes have been characterized to the level of a 15 mer peptide (Figs. 2 and 3; data not shown). Six hybridomas responded to 101–115, six hybridomas responded to 126 –140, one hybridoma responded to 206 –220, and nine hybridomas responded to 536 –550. Two hybridomas each responded to 431– 445 and 461– 475. Fig. 3A shows an example of the IL-2 response of one hybridoma to whole human GAD65 and a series of 10 pools of peptides using a DQ81 human EBV-transformed B cell line as APC. Hybridoma 20D4.1 recognizes whole human GAD65, specifically an epitope in the C-terminal region. The response to this region can be mapped to the peptide 536 –550 (Fig. 3B). This response is DQ8 restricted, since the human APC and Tg mice have only DQ8 in common. Additionally, an Ab specific for DQ8 (IVD12) included in the cultures with APC and specific peptide could completely inhibit the production of IL-2 from all hybridomas tested, specific for five of the six epitopes identified in this study (data not shown). Truncation analysis to determine an immunogenic motif To examine the interaction of peptide, DQ8, and TCR, we determined the core minimal sequence required for stimulation of the T cell hybridomas by truncation analysis. Fig. 4 shows the IL-2 responses of four hybridomas to their specific 15 mer peptides and

6278


FIGURE 3. Human GAD65 epitopes can be determined from hybridomas generated in DQ8 Tg mice. A, Hybridoma 20D4.1 was screened for Ag specificity. T cells were tested with DQ81 human APC in the presence of human GAD65, 1 of 10 pools of GAD65 peptides, or medium alone, as previously described (25). B, Hybridoma 20D4.1 was tested for IL-2 production with each of 11 peptides in the pool spanning aa 521–585 of human GAD65.

truncations for the following epitopes of human GAD65: 101–115 (Fig. 4A), 126 –140 (Fig. 4B), 206 –220 (Fig. 4C), and 536 –550 (Fig. 4D). Hybridoma 15H2 responds to the 101–115 epitope of human GAD65 in a dose-dependent fashion as well as responds to some truncations of that sequence spanning amino acids 102–112, 103–113, and 104 –114 (Fig. 4A). Fig. 4B shows the same type of analysis for hybridoma 17G12.7, specific for an epitope from 126 – 140. The dose-dependent IL-2 response of hybridoma 8G5.4 to the 206 –220 epitope is confirmed in Fig. 4C. Finally, the epitope found at 536 –550 elicits a response from representative hybridoma 20D4.1 (Fig. 4D). To map the smallest essential region necessary for stimulation of the T cell, we used the response to truncated peptides to determine the core minimal epitope that still yields IL-2 production from a specific clone. Response at any dose of a truncated peptide was interpreted to mean that that peptide contains the necessary core epitope. For clarity, Fig. 5 replots the responses of the hybridomas to the maximal peptide dose of 100 mM from Fig. 4, with the peptide sequences shown at the left. Truncating the glutamic acid at the N-terminal position 104 completely abrogates IL-2 production from clone 15H2 (Fig. 5A). C-terminal truncation of the glutamine at position 112 also abrogates IL-2 production, indicat-

FIGURE 4. Hybridomas respond to specific 15 mer peptides of human GAD65 as well as truncated variants of those peptides. The IL-2 response of DQ8-restricted T cell hybridomas to their specific peptides and truncations of those epitopes is shown. A, IL-2 production of hybridoma 15H2 responding to human APC and a dose-response curve from 0 –100 mM of peptide 101–115 and to the truncations of 101–115 shown in the legend at the right is shown. Similar data as in A are shown for a 126 –140 specific hybridoma 17G12.7 (B), for the 206 –220 specific hybridoma 8G5.4 (C), and for hybridoma 20D4.1 (D) responding to its specific epitope from 536 –550 and the indicated truncations of those epitopes.

ing a 9-aa core from glutamic acid 104 to glutamine at position 112. The aspartic acid at position 113 is preferred, but not required, for T cell stimulation. Hybridoma 17G12.7 responds to a 10-aalong core sequence, from the aspartic acid residue at position 127 to the histidine at position 136, as evidenced by the lack of response to 121–135 (Fig. 5B). The 8G5.4 hybridoma also sees a 10 mer core peptide, from a glutamic acid at 208 to a second glutamic


6279

FIGURE 5. Analysis of the response to truncated peptides reveals the core minimal epitope for T cell stimulation. The truncation peptides are listed to the left of the IL-2 response they elicit at 100 mM peptide, replotted from Fig. 4. A, Hybridoma 15H2 responds to some truncations of its specific epitope from 101–115, shown by the production of IL-2. Similar analysis for the 17G12.7 hybridoma specific for 126 –140 (B), 8G5.4 specific for 206 –220 (C), and 20D4.1 specific for 536 –550 (D).

acid residue at position 217 (Fig. 5C). Responsiveness is also lost when residues are present at 218 and 219, but returns with 206 – 217, suggesting that 208 –217 is the core. Truncation analysis of the 536 –550 epitope demonstrates that clone 20D4.1 recognizes a 10-aa core that extends from a glutamic acid at position 539 to the glutamine residue at 548 (Fig. 5D). Aligning the core minimal sequences determined in Fig. 5 reveals a striking pattern for peptides derived from human GAD65 that bind to DQ8 and are immunogenic to T cells (Table I). Alignment of other known DQ8 binding peptides, a l repressor peptide

and a herpes virus VP-16 peptide, confirms agreement of these sequences with a motif determined from the human GAD65 epitopes (10). Table I also shows a putative alignment for two epitopes found in this study from 431– 445 and 461– 475 that were not analyzed by truncation. At the amino terminus P1, we found a negatively charged acidic residue. There was a small polar or proline residue at P4. P6 and P7 were predominantly hydrophobic residues. No definitive pattern emerged at P9, which presumably would be the position affected by the aspartic acid/noncharged residue polymorphism at DQ8 b57. Interestingly, we did see a

6280


Table I. Alignment of the core epitope sequences determined from truncation analysesa Position Peptide

GAD GAD GAD GAD

b

104-113 127-136 208-217 539-548

DQ8 motif GAD 431-445c GAD 461-473c l rep 12-24c VP-16 433-445c

K

H

Y

L

E

1

2

3

4

5

6

7

8

9

10

E D E E

R R I Y

P S A G

T T P T

L K V T

A V F M

F I V V

L D L S

Q F L Y

D H E Q

D E

Y,I R

A,V M,F

V F,I

D A D D

L K A M

T G K D

G F A A

T P S G R T

Y T R P

D T L A

D,E Q,H D E I L

K A Y D

A H E D

L V K F

Q D

K

D

L

Highlighted peptide residues are required for T cell stimulation. The l repressor 12-24 peptide, which binds to DQ8 in vitro; and the HSV-2 VP-16 433-445 peptide, for which there is a DQ8-restricted human T cell clone, are aligned to the proposed motif (10, 35). b Residue 113 is preferred not required for T cell stimulation. c Putative alignment. a

trend indicating a negatively charged residue at P10, which is required for T cell stimulation of most clones (Table I and Fig. 5). Response of PBMC from diabetic individuals to the human GAD65 epitopes To ask whether the epitopes determined in DQ8 Tg mice are relevant in the human disease process, we stimulated PBMC from diabetic and nondiabetic control individuals with the human GAD65 peptides and tested for proliferation. Individuals with positive responses to GAD65 peptides and .3-fold increased response to tetanus are presented (Table II). Not all individuals responded to GAD65 peptides (data not shown). Overall, 11 of 16 recent-onset diabetic individuals and 9 of 16 control individuals show proliferative responses to at least one GAD65 peptide. All six epitopes from this study generate a proliferative response by PBMC from at least one diabetic individual (Table II). PBMC from control individuals also proliferate in response to the peptides. Some of these responses are not restricted to DQ8, because not all responders were DQ81. Fifteen of 16 diabetic and 4 of 16 control individuals were DQ81 (Table II and data not shown). Responses demonstrated in diabetic and control individuals could be restricted by DQ2 or DQA1*0101/DQB1*0501, both diabetesassociated HLA molecules lacking aspartic acid at b57 (2).

Discussion Key factors in generating immune responses in HLA Tg mice have been elucidated by previous work (20). The absence of mouse class II is essential for obtaining optimal human HLA-restricted responses in Tg systems. The addition of a human CD4 transgene can also be critical for the selection of CD41 T cells in some HLA-DR Tg systems (25). However, in the DQ8 Tg system, murine CD4 was sufficient to select a DQ8-restricted mouse CD4 T cell repertoire that was functional in the periphery (Fig. 1D). We were able to generate GAD65-specific hybridomas from DQ8 Tg mice, defining the presence of at least nine epitopes in human GAD65 presented by the DQ8 molecule (Fig. 2). Although human GAD65 is not a self Ag in these mice, mouse and human GAD65 are 96% identical at the amino acid level, and all core peptides determined in this study are identical between the mouse and human sequences (data not shown). We were able to specifically identify 101–115, 126 –140, 206 –220, 431– 445, 461– 475, and 536 –550 as epitopes. Interestingly, the 206 –220 epitope is the immunodominant epitope presented by the diabetes-associated IAg7 class II molecule in NOD mice (26, 28). Although the struc-

ture of the two molecules may be very similar (29), the DQ8restricted 206 –220 specific hybridoma does not recognize this epitope in the context of I-Ag7 (data not shown). Human GAD65 regions presented by the DQ8 molecule were determined in a similar DQ8 Tg mouse system in a study published during the preparation of this manuscript (23). This study also identified 121–140 and 201–220 as well as 231–250 and 471– 490 as regions eliciting proliferative responses from bulk cultures of splenocytes from human GAD65-immunized mice. The identification by both studies of 126 –140 and 206 –220 provides independent confirmation that these regions are presented by the DQ8 molecule. Our study also detected reactivity in the regions spanning 231–250 and 471– 490, although those hybridomas became nonresponsive before a specific epitope could be mapped (see Fig. 2). The hybridoma methodology may have allowed us to identify the additional epitopes from 101–115, 431– 445, 461– 475, and 536 –550 due to greater ability to detect specific T cells that are not detectable in bulk culture. The differences between the two studies could be due to the inclusion of a human CD4 transgene in the DQ8 Tg mice studied by Liu et al. (23), to the use of bacterial vs baculovirus-derived GAD65, or to the differences in immunization and detection of the T cell response between the two studies. Preserving the T cell specificities as hybridomas allowed us to prove that the six epitopes identified in this study can be processed by human APC. The IL-2 responses demonstrated to the whole human GAD65 protein are stimulated by human EBV-transformed DQ81 B cells presenting Ag (Figs. 2 and 3; data not shown). This underscores the usefulness of the mouse model, because it appears that the murine processing machinery is similar enough to generate many epitopes also generated by the human processing machinery. Analyses of the responses by hybridomas to their specific peptides and truncations shows several interesting characteristics for peptides presented by the DQ8 diabetes-associated molecule. One observation common to all clones is that they produce significant amounts of IL-2 with a minimum requirement of 6.25 mM of their specific peptide (Fig. 4). A similar high concentration of peptide is required to stimulate GAD65-specific hybridomas restricted to the diabetes-associated I-Ag7 molecule (C.-C. Chao, personal communication). In contrast, hybridomas restricted to HLA-DR class II alleles can respond to peptide at concentrations of ,0.3 mM (25) (A. Cope, unpublished observation). This observation could reflect the low affinity binding of these peptides to the DQ8 molecule, as has been documented for epitopes from autoantigens important in


6281

Table II. Proliferative responses of PBMC from recent-onset diabetic individuals and nondiabetic controlsa DQB1/DQB1 Diabetic *0201/*0302 *02/*0302 *0201/*0302 *02/*0302 *0201/*0302 *02/*0302 *0501/*0302 *0501/*0302 *02/*0609 Control *02/*0501d *02/*0302 *02/*0402 *0602/*0501 *0601/*0501 *02/*0301 *02/*0304d *02/*0302d

Tetanus

GAD 101-115

GAD 126-140

GAD 206-220

GAD 431-445

GAD 461-475

GAD 536-550

No Ag

71,431 6 7,168 23,768 6 8,163 133,637 6 12,218 9,711 6 1,897 24,604 6 3,500 22,129 6 4,279 7,723 6 1,626 14,514 6 5,207 44,489 6 16,939

229 6 102 944 6 157b 735 6 295 1,237 6 245 4,483 6 161 1,093 6 126b 1,524 6 263 3,194 6 3,115b 1,909 6 1,187

690 6 503 944 6 157b 4,340 6 2,485 581 6 169 5,390 6 1,154 1,093 6 126b 836 6 155 3,194 6 3115b 1,778 6 905

7,916 6 639 4,028 6 782 1,640 6 751 2,504 6 836 19,376 6 2,320 1,629 6 191 839 6 137 2,082 6 405 1,699 6 486

9,264 6 581 737 6 287 1,093 6 378 829 6 338 ND 1,222 6 73b 1,115 6 881 4,164 6 1,676b 3,767 6 2,720

2,236 6 397 705 6 133 1,836 6 340 840 6 330 ND 1,222 6 73b 965 6 224 4,164 6 1,676b 2,159 6 947

8,163 6 2,790 2,767 6 498 619 6 186 604 6 81 2,616 6 390 1,274 6 231 784 6 429 2,674 6 2,443 1,996 6 335

347 6 216 1,604 6 287 874 6 196 644 6 85 6,682 6 939 631 6 159 339 6 87 924 6 361 1,027 6 203

941 6 175 631 6 231 509 6 132 768 6 61 515 6 145 914 6 173 676 6 291b 2,611 6 900b

1,287 6 96 354 6 158 479 6 227 504 6 120 357 6 86 1,310 6 622 676 6 291b 2,611 6 900b

1,041 6 158 1,144 6 357 1,330 6 627 4,456 6 1,387 4,187 6 226 11,118 6 1,849 2,966 6 2,905 2,093 6 25

1,268 6 369 463 6 82b 615 6 227b 692 6 248 526 6 212 1,918 6 522 915 6 1,015 2,347 6 307b

2,018 6 169 463 6 82b 615 6 227b 1,154 6 622 683 6 283 4,758 6 1,023 915 6 1,015 2,347 6 307b

1,486 6 158 1,206 6 625 2,647 6 854 2,966 6 1,031 1,814 6 399 1,347 6 1,011 674 6 122 2,046 6 484c

330 6 159 332 6 35 535 6 277 617 6 332 586 6 438 1,480 6 359 581 6 337 808 6 248

10,325 6 1,040 16,934 6 2,805 2,671 6 392 4,628 6 743 13,600 6 1,364 75,726 6 19,183 2,183 6 3,074 2,213 6 404

a Proliferative responses of PBMC from recent-onset diabetic individuals and nondiabetic controls to the human GAD65 epitopes are presented in terms of counts of [3H]thymidine uptake. Responses are the mean 6 SD of five independent wells. Responses of stimulation index .2 over the media only control, and that were at least 1000 cpm, and where the individual also had a positive response to tetanus, are considered positive, and are in bold. Only individuals with positive response to at least one GAD65 epitope and a .3-fold response to tetanus are shown. HLA-DQ haplotypes are shown at the left. ND, not done. b Peptide pooled with adjacent peptide (101-115 with 126-140, or 431-445 with 461-475). c Peptide 536-550 was pooled with 526-540, a DRB1*0401 epitope, and response could be to either peptide. d Sibling or parent of a recent-onset diabetic individual.

other autoimmune diseases (30). Alternately, the high concentration of peptide required could reflect a relatively low avidity TCR interacting with the DQ8/peptide complex. Another feature is that a majority of the hybridomas require a 10-aa core peptide for stimulation, in contrast to a 9 mer core required by most class II-restricted T cell responses (Fig. 5) (8). Our data do not distinguish between the 10th residue being required for binding to the DQ8 molecule or for recognition by the TCR. The crystal structure of mouse I-Ak/hen egg lysozyme peptide, the closest HLA-DQ homologue determined, suggests involvement of peptide-flanking residues outside P1–P9 in TCR engagement (31). In this structure, residues that flank the core nine residues of the peptide on either end point up and angle toward the TCR. Positions P10 and P11 at the C terminus of the hen egg lysozyme peptide have been shown to be critical for TCR recognition of several clones (32). Additionally, naturally processed sets of peptides with differences in the number of peptide-flanking residues have been shown to act as antagonists for T cells dependent on these residues for recognition (33). A similar recognition pattern by GAD65-specific DQ8-restricted clones as a general feature could be biologically relevant and potentially important for Agspecific disease therapy. Clone 8G5.4, specific for the core epitope 208 –217, appears to be sensitive to the presence of additional flanking residues beyond P10 (Fig. 5C). It is unclear why the addition of residues 218 or 218 and 219 produces this effect, while the addition of residues 218 –220 does not. The pattern discerned from the alignment of the core epitopes determined in this study in Table I is similar in part to the pattern that emerges from DQ8 peptide elution studies or binding studies, but differs at the C terminus (12, 13). DQ8 peptide elution studies found predominantly a negatively charged residue at P1, confirming our result and the alignment of the P1 position (13). Unlike our study, the elution studies found negatively charged residues at P9. They demonstrated that peptides with negatively charged P9 residues bound with higher affinity than peptides without such residues. However, elution studies have the disadvantage of detecting predominantly epitopes of high affinity or high availability and do not take into account the T cell response to the peptides studied. By determining immunogenic T cell epitopes, we find many epitopes

from an Ag important in the disease process that do not contain a negatively charged residue at P9. This may again indicate that these GAD65 epitopes have low affinity interactions with the DQ8 molecule. This result is consistent with the idea that MHC/peptide complexes of high affinity, and therefore high availability, would delete specific T cells during negative selection in the thymus (34). Kwok et al. (10) found using the HSV-2 VP-16 peptide that the DQ8 molecule prefers negatively charged residues at the putative P9 residue, while the reverse is true for diabetes-nonassociated DQ9, which differs only at b57. Applying the DQ8 motif in Table I to this peptide puts the aspartic acid residue 433 in the P1 position, placing the aspartic acid residue 442 critical for the binding difference between DQ8 and DQ9 into the P10 position (see Table I). Using this motif with the l repressor 12–24 peptide also puts the glutamic acid 23 residue, critical for DQ8/DQ9 binding differences (35), in P10. Our results indicate that P10 is absolutely required for T cell stimulation for most epitopes, and P10 is commonly a negatively charged residue (Fig. 5 and Table I). Taken together, these studies suggest the possibility that a negatively charged residue at P10 is important in binding differences caused by a nonaspartic acid residue at b57. The ultimate test of the mouse model system was to determine whether human PBMC from diabetic individuals could respond to the GAD65 epitopes defined in DQ8 Tg mice. This is indeed the case, although responses are not restricted to diabetic individuals (Table II). Prior studies have shown that both individuals with autoimmune disease and healthy subjects have T cells that respond to self peptides (36, 37). Although responses are generated in many nondiabetic individuals, pathogenic responses could be qualitatively or quantitatively different. Alternately, the difference may be a lack of ability to regulate a pathogenic response in a diabetic individual. Another notable finding is that the response to these GAD65 peptides is not necessarily restricted to the DQ8 molecule, since in a number of cases DQ8-negative individuals can respond to these peptides. The products of DQ2 and other alleles may be able to present some of the same peptides from human GAD65 as the DQ8 molecule. In a Sardinian population, DQ2 homozygous diabetic individuals also respond to the 206–220 peptide, indicating that the DQ2 molecule can

6282


present this peptide (M. Congia, unpublished observation). Recent evidence from this laboratory using HLA-DQ6 (DQA1*0103/ DQB1*0601) Tg mice shows that this molecule can present at least one peptide from human GAD65 presented by the DQ8 molecule (G. Hwang, personal communication). Our data suggest that the Tg mouse model is a useful system for identifying epitopes from human autoantigens that are recognized by human T cells. Epitopes that are naturally processed from the whole protein can be identified (38). The model allows us to address questions that cannot be addressed in the human system due to the difficulty of human T cell cloning or the large volumes of patient blood needed for proliferative studies. In addition to examining peptide presentation, the HLA Tg mouse model provides an opportunity to study how T cell responses restricted to diseaseassociated class II molecules may differ from those restricted to class II molecules not associated with type I diabetes.

Acknowledgments We thank Dr. Henry Erlich for access to HLA typing of individuals in the PBMC proliferation studies; Dr. Darrell Wilson for access to recent-onset diabetic patients; Drs. Diane Mathis and Christophe Benoist for providing the class II-deficient mice, the I-E a promoter, and the pSV5neo vector; Dr. Jack Strominger for a construct containing the DQA1*0301 sequence; Dr. Dan Littman for human CD4 transgenic mice; May Koo, Peggy Sullivan, Mary Vadeboncoeur, Sharon Phillips, Jon Toma, and Robert Pesich for excellent technical assistance; Lou Hildalgo and Ricardo Salazar for animal care; and Dr. Mark M. Davis, Igor Brodsky, Dr. Iris Ferber, Marcos Garcı´a-Ojeda, Dr. Frances Hall, Dr. Sybil Munson, Dr. Joshua Rabinowitz, and Kristin Tarbell for critical reading of the manuscript.

References 1. Castano, L., and G. S. Eisenbarth. 1990. Type-I diabetes: a chronic autoimmune disease of human, mouse, and rat. Annu. Rev. Immunol. 8:647. 2. Nepom, G. T., and H. Erlich. 1991. MHC class-II molecules and autoimmunity. Annu. Rev. Immunol. 9:493. 3. Todd, J. A. 1997. Genetics of type 1 diabetes. Pathol. Biol. (Paris) 45:219. 4. Todd, J. A., H. Acha-Orbea, J. I. Bell, N. Chao, Z. Fronek, C. O. Jacob, M. McDermott, A. A. Sinha, L. Timmerman, L. Steinman, et al. 1988. A molecular basis for MHC class II-associated autoimmunity [published erratum appears in 1988, Science 241:888]. Science 240:1003. 5. Todd, J. A., J. I. Bell, and H. O. McDevitt. 1987. HLA-DQ b gene contributes to susceptibility and resistance to insulin-dependent diabetes mellitus. Nature 329: 599. 6. Jameson, S. C., K. A. Hogquist, and M. J. Bevan. 1995. Positive selection of thymocytes. Annu. Rev. Immunol. 13:93. 7. Kanagawa, O., B. A. Vaupel, G. Xu, E. R. Unanue, and J. D. Katz. 1998. Thymic positive selection and peripheral activation of islet antigen-specific T cells: separation of two diabetogenic steps by an I-A(g7) class II MHC b-chain mutant. J. Immunol. 161:4489. 8. Rammensee, H. G., T. Friede, and S. Stevanoviic. 1995. MHC ligands and peptide motifs: first listing. Immunogenetics 41:178. 9. Stern, L. J., J. H. Brown, T. S. Jardetzky, J. C. Gorga, R. G. Urban, J. L. Strominger, and D. C. Wiley. 1994. Crystal structure of the human class II MHC protein HLA-DR1 complexed with an influenza virus peptide. Nature 368: 215. 10. Kwok, W. W., M. E. Domeier, M. L. Johnson, G. T. Nepom, and D. M. Koelle. 1996. HLA-DQB1 codon 57 is critical for peptide binding and recognition. J. Exp. Med. 183:1253. 11. Wucherpfennig, K. W., and J. L. Strominger. 1995. Selective binding of self peptides to disease-associated major histocompatibility complex (MHC) molecules: a mechanism for MHC-linked susceptibility to human autoimmune diseases. J. Exp. Med. 181:1597. 12. Chicz, R. M., W. S. Lane, R. A. Robinson, M. Trucco, J. L. Strominger, and J. C. Gorga. 1994. Self-peptides bound to the type I diabetes associated class II MHC molecules HLA-DQ1 and HLA-DQ8. Int. Immunol. 6:1639. 13. Godkin, A., T. Friede, M. Davenport, S. Stevanovic, A. Willis, D. Jewell, A. Hill, and H. G. Rammensee. 1997. Use of eluted peptide sequence data to identify the binding characteristics of peptides to the insulin-dependent diabetes susceptibility allele HLA-DQ8 (DQ 3.2). Int. Immunol. 9:905. 14. Roep, B. O. 1996. T-cell responses to autoantigens in IDDM: the search for the Holy Grail. Diabetes 45:1147.

15. Lernmark, A. 1996. Glutamic acid decarboxylase: gene to antigen to disease. J. Intern. Med. 240:259. 16. Schloot, N. C., B. O. Roep, D. R. Wegmann, L. Yu, T. B. Wang, and G. S. Eisenbarth. 1997. T-cell reactivity to GAD65 peptide sequences shared with coxsackie virus protein in recent-onset IDDM, post-onset IDDM patients and control subjects. Diabetologia 40:332. 17. Hawa, M., R. Rowe, M. S. Lan, A. L. Notkins, P. Pozzilli, M. R. Christie, and R. D. Leslie. 1997. Value of antibodies to islet protein tyrosine phosphatase-like molecule in predicting type 1 diabetes. Diabetes 46:1270. 18. Pietropaolo, M., M. Peakman, S. L. Pietropaolo, M. M. Zanone, T. P. Foley, Jr., D. J. Becker, and M. Trucco. 1998. Combined analysis of GAD65 and ICA512(IA-2) autoantibodies in organ and non-organ-specific autoimmune diseases confers high specificity for insulin-dependent diabetes mellitus. J. Autoimmun. 11:1. 19. Verge, C. F., R. Gianani, E. Kawasaki, L. Yu, M. Pietropaolo, R. A. Jackson, H. P. Chase, and G. S. Eisenbarth. 1996. Prediction of type I diabetes in firstdegree relatives using a combination of insulin, GAD, and ICA512bdc/IA-2 autoantibodies. Diabetes 45:926. 20. Cope, A. P., S. D. Patel, F. Hall, M. Congia, H. A. Hubers, G. F. Verheijden, A. M. Boots, R. Menon, M. Trucco, A. W. Rijnders, et al. 1999. T cell responses to a human cartilage autoantigen in the context of rheumatoid arthritis-associated and nonassociated HLA-DR4 alleles. Arthritis Rheum. 42:1497. 21. Taneja, V., and C. S. David. 1998. HLA transgenic mice as humanized mouse models of disease and immunity. J. Clin. Invest. 101:921. 22. Wen, L., F. S. Wong, L. Burkly, M. Altieri, C. Mamalaki, D. Kioussis, R. A. Flavell, and R. S. Sherwin. 1998. Induction of insulitis by glutamic acid decarboxylase peptide-specific and HLA-DQ8-restricted CD4(1) T cells from human DQ transgenic mice [published erratum appears in 1998, J. Clin. Invest. 102:following 1463]. J. Clin. Invest. 102:947. 23. Liu, J., L. E. Purdy, S. Rabinovitch, A. M. Jevnikar, and J. F. Elliott. 1999. Major DQ8-restricted T-cell epitopes for human GAD65 mapped using human CD4, DQA1*0301, DQB1*0302 transgenic IA(null) NOD mice. Diabetes 48:469. 24. Hogan, B. 1986. Manipulating the Mouse Embryo: A Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor. 25. Patel, S. D., A. P. Cope, M. Congia, T. T. Chen, E. Kim, L. Fugger, D. Wherrett, and G. Sonderstrup-McDevitt. 1997. Identification of immunodominant T cell epitopes of human glutamic acid decarboxylase 65 by using HLADR(a1*0101,b1*0401) transgenic mice. Proc. Natl. Acad. Sci. USA 94:8082. 26. Chao, C. C., and H. O. McDevitt. 1997. Identification of immunogenic epitopes of GAD 65 presented by Ag7 in non-obese diabetic mice. Immunogenetics 46:29. 27. Noble, J. A., A. M. Valdes, M. Cook, W. Klitz, G. Thomson, and H. A. Erlich. 1996. The role of HLA class II genes in insulin-dependent diabetes mellitus: molecular analysis of 180 Caucasian, multiplex families. Am. J. Hum. Genet. 59:1134. 28. Zechel, M. A., J. F. Elliott, M. A. Atkinson, and B. Singh. 1998. Characterization of novel T-cell epitopes on 65 kDa and 67 kDa glutamic acid decarboxylase relevant in autoimmune responses in NOD mice. J. Autoimmun. 11:83. 29. Acha-Orbea, H., and H. O. McDevitt. 1987. The first external domain of the nonobese diabetic mouse class II I-A b chain is unique. Proc. Natl. Acad. Sci. USA 84:2435. 30. Lee, C., M. N. Liang, K. M. Tate, J. D. Rabinowitz, C. Beeson, P. P. Jones, and H. M. McConnell. 1998. Evidence that the autoimmune antigen myelin basic protein (MBP) Ac1–9 binds towards one end of the major histocompatibility complex (MHC) cleft. J. Exp. Med. 187:1505. 31. Fremont, D. H., D. Monnaie, C. A. Nelson, W. A. Hendrickson, and E. R. Unanue. 1998. Crystal structure of I-Ak in complex with a dominant epitope of lysozyme. Immunity 8:305. 32. Carson, R. T., K. M. Vignali, D. L. Woodland, and D. A. Vignali. 1997. T cell receptor recognition of MHC class II-bound peptide flanking residues enhances immunogenicity and results in altered TCR V region usage. Immunity 7:387. 33. Carson, R. T., D. D. Desai, K. M. Vignali, and D. A. Vignali. 1999. Immunoregulation of Th cells by naturally processed peptide antagonists. J. Immunol. 162:1. 34. Ashton-Rickardt, P. G., and S. Tonegawa. 1994. A differential-avidity model for T-cell selection. Immunol. Today 15:362. 35. Nepom, B. S., G. T. Nepom, M. Coleman, and W. W. Kwok. 1996. Critical contribution of b chain residue 57 in peptide binding ability of both HLA-DR and -DQ molecules. Proc. Natl. Acad. Sci. USA 93:7202. 36. Boyton, R. J., T. Lohmann, M. Londei, H. Kalbacher, T. Halder, A. J. Frater, D. C. Douek, D. G. Leslie, R. A. Flavell, and D. M. Altmann. 1998. Glutamic acid decarboxylase T lymphocyte responses associated with susceptibility or resistance to type I diabetes: analysis in disease discordant human twins, non-obese diabetic mice and HLA-DQ transgenic mice. Int. Immunol. 10:1765. 37. Trotter, J. L., C. M. Pelfrey, A. L. Trotter, J. A. Selvidge, K. C. Gushleff, T. Mohanakumar, and H. F. McFarland. 1998. T cell recognition of myelin proteolipid protein and myelin proteolipid protein peptides in the peripheral blood of multiple sclerosis and control subjects. J. Neuroimmunol. 84:172. 38. Viner, N. J., C. A. Nelson, B. Deck, and E. R. Unanue. 1996. Complexes generated by the binding of free peptides to class II MHC molecules are antigenically diverse compared with those generated by intracellular processing. J. Immunol. 156:2365.