Type 1 diabetes is a spontaneous organ-specific

Transcription

1 Immunological Characterization and Therapeutic Activity of an Altered-Peptide Ligand, NBI-6024, Based on the Immunodominant Type 1 Diabetes Autoantigen Insulin B-Chain (9 23) Peptide David G. Alleva, 1 Amitabh Gaur, 1 Liping Jin, 1 Dale Wegmann, 2 Peter A. Gottlieb, 2 Anil Pahuja, 1 Eric B. Johnson, 1 Theresa Motheral, 2 Amy Putnam, 2 Paul D. Crowe, 1 Nicholas Ling, 1 Stefen A. Boehme, 1 and Paul J. Conlon 1 The nonobese diabetic (NOD) mouse is a good model for human type 1 diabetes, which is characterized by autoreactive T-cell mediated destruction of insulin-producing islet -cells of the pancreas. The 9 23 amino acid region of the insulin B-chain [B (9 23) ] is an immunodominant T-cell target antigen in the NOD mouse that plays a critical role in the disease process. By testing a series of B (9 23) peptide analogs with single or double alanine substitutions, we identified a set of altered peptide ligands (APLs) capable of inhibiting B (9 23) -induced proliferative responses of NOD pathogenic T-cell clones. These APLs were unable to induce proliferation of these clones. However, vaccinations with the APLs induced strong cellular responses, as measured by in vitro lymphocyte proliferation and Th2 cytokine production (i.e., interleukin [IL]-4 and IL-10, but not -interferon [IFN- ]). These responses were cross-reactive with the native antigen, B (9 23), suggesting that the APL-induced Th2 responses may provide protection by controlling endogenous B (9 23) -specific Th1 (i.e., IFN- producing) pathogenic responses. One of these APLs that contained alanine substitutions at residues 16 and 19 (16Y3 A, 19C3 A; NBI-6024) was further characterized for its therapeutic activity because it consistently induced T-cell responses (e.g., T-cell lines and clones) that were of the Th2 type and that were cross-reactive with B (9 23). Subcutaneous injections of NBI-6024 to NOD mice administered either before or after the onset of disease substantially delayed the onset and reduced the incidence of diabetes. This study is the first to report therapeutic activity of an APL derived from an From 1 Neurocrine Biosciences, Inc., San Diego, California; and the 2 University of Colorado Health Sciences Center, Barbara Davis Center for Childhood Diabetes, Denver, Colorado. Address correspondence and reprint requests to Dr. David G. Alleva, Neurocrine Biosciences, Inc., Science Center Dr., San Diego, CA dalleva@neurocrine.com. Received for publication 4 December 2001 and accepted in revised form 3 April D.G.A., L.J., A.P., E.B.J., N.L., S.A.B., and P.J.C. are employed by and hold stock in Neurocrine Biosciences, Inc.; A.G. holds stock in Neurocrine Biosciences, Inc. APL, altered peptide ligand; B (9 23), 9 23 amino acid region of the insulin B-chain; CA, -cell target antigen; CFA, complete Freund s adjuvant; EAE, experimental autoimmune encephalomyelitis; ELISA, enzyme-linked immunosorbent assay; ELISPOT, enzyme-linked immunosorbent spot; IFA, incomplete Freund s adjuvant; IFN-, -interferon; IL, interleukin; MBP, myelin basic protein; MHC, major histocompatibility complex; MS, multiple sclerosis; NT, neurotensin; PLP, proteolipid protein; SI, stimulation index; SWM ( ), sperm whale myoglobin ( ); TCR, T-cell receptor. islet -cell specific antigen in type 1 diabetes. Diabetes 51: , 2002 Type 1 diabetes is a spontaneous organ-specific autoimmune disease that is triggered by the interaction of several genetic and environmental factors (1). The nonobese diabetic (NOD) mouse is a good model for type 1 diabetes; it resembles the human disease by sharing predisposing genetic factors and characteristics of disease initiation and progression (2,3). Insulitis is the initial stage of the disease process, in which inflammatory leukocytes infiltrate the pancreas and are responsible for lesions within islets. Overt disease is manifested only when insulin-producing -cells are destroyed within these islets, leading to impaired glucose metabolism and attendant complications characteristic of type 1 diabetes. The destruction of islet -cells in the NOD mouse is mediated by the activation of autoreactive T- cells, which recognize several islet -cell target antigens ( CAs), including insulin, GAD, heat shock protein 60, and some uncharacterized CAs (1 4). These antigen specificities have been defined in primary T-cell assays and by the generation of T-cell lines and clones from lymph nodes, spleens, and pancreata of NOD mice (5). Although T-cell reactivity to several autoantigens has been demonstrated, the majority of pathogenic CD4 T-cell clones derived from pancreata of NOD mice with insulitis or frank diabetes react specifically with the 9 23 amino acid region of the insulin B-chain [B (9 23) ] (5 8). Moreover, screening of an NOD islet -cell antigen cdna library demonstrated that as much as 87% of CD8 T-cells in the pancreata from young NOD mice recognized the region of the insulin B-chain (9). The B (9 23) epitope also appears to be involved in the human disease as well, in which the majority of type 1 diabetic subjects show positive Th1 responses to B (9 23) in the enzyme-linked immunosorbent spot (ELISPOT) and proliferation assays (10). Thus, the B (9 23) region of insulin appears to contain critical epitopes recognized by autoreactive T-cells in type 1 diabetes and, because it is a fragment of insulin, it is the only diabetes-associated autoantigen that has an expression limited to islet -cells and is the most abundantly produced protein by that tissue DIABETES, VOL. 51, JULY 2002

2 D.G. ALLEVA AND ASSOCIATES Identification of critical autoantigenic epitopes is key to developing antigen-based therapies for autoimmune diseases such as type 1 diabetes, multiple sclerosis (MS), and rheumatoid arthritis. Most notably, altered peptide ligands (APLs) are peptide analogs of immunodominant autoantigenic epitopes, which have the ability to competitively inhibit pathogenic autoreactive T-cell clones from recognizing native peptide epitopes (11,12). Therapeutic APLs have been studied in the murine model of MS, experimental autoimmune encephalomyelitis (EAE), in which APLs are derived from immunodominant epitopes of the neuroantigens myelin basic protein (MBP) and myelin proteolipid protein (PLP) (13 18). These APLs competitively inhibited the recognition of native peptides by T-cell clones, but by themselves, did not stimulate these T-cells. Hence, therapeutic APLs maintain a strong binding affinity to major histocompatibility complex (MHC) molecules while engaging in a weak nonproductive interaction with the T-cell receptor (TCR) to avoid full activation of the pathogenic T-cells. However, the feature most responsible for APL therapeutic activity appears to be the induction of a small population of antigen cross-reactive regulatory (i.e., Th2) T-cells after vaccination (12,19,20). The therapeutic activity of APL-induced T-cell responses is demonstrated by adoptive transfer of APL-specific Th2 cell lines into naive hosts that provide protection from EAE (12,19). These APL-specific Th2 cells (i.e., interleukin [IL]-4, IL-5, and IL-10 producing cells) appear to protect the animals based on an antigen cross-reactive mechanism that dampens the pathogenic Th1 cytokine (i.e., -interferon [IFN- ]) T-cell response against native neuroantigenic peptides (21). Although it is unclear how APLs induce Th2 phenotypes, there appear to be both qualitative and quantitative differences between native peptides and their APLs in the TCR-associated signals induced during antigen presentation (11,21). Because insulin B (9 23) appears to be a dominant epitope associated with disease in the NOD mouse, we produced several APLs based on this epitope and, by in vitro characterization, identified one APL, NBI-6024, as a therapeutic candidate for in vivo studies. NBI-6024 strongly reduced disease incidence if given either before or after the onset of diabetes and induced a regulatory Th2 cytokine production phenotype. Results of this study are now presented as the first report of a therapeutic activity of an APL derived from an islet -cell specific antigen in type 1 diabetes. RESEARCH DESIGN AND METHODS Animals. Female 3- to 4-week-old NOD mice were purchased from The Jackson Laboratories (Bar Harbor, ME) and housed in a vivarium under pathogen-free conditions at Neurocrine Biosciences (San Diego, CA). Animal experimentation was approved by Neurocrine Biosciences Institutional Animal Care and Use Committee. Cell culture media and regents. Mouse cells were cultured in Dulbecco s minimal essential medium with high glucose supplemented with 2 mmol/l L-glutamine, 10 mmol/l HEPES (Cellgro, Herndon, VA), nonessential amino acids (Sigma, St. Louis, MO), 1 mmol/l sodium pyruvate, 50 g/ml gentamicin, 126 g/ml L-arginine, 20 g/ml L-aspartic acid (Life Technologies, Grand Island, NY), 50 mol/l 2-mercaptoethanol (Sigma), 100 units/ml penicillin, and 5 g/ml streptomycin (Life Technologies). Heat-inactivated 10% fetal bovine serum (HyClone, Logan, UT) was added to the medium before use. Human T-cell lines were cultured in RPMI-1640 medium (Cellgro) with the supplements described above. All peptides, including the native insulin B (9 23) (SHLVEALYLVCGERG) and several alanine-substituted APLs of B (9 23), sperm whale myoglobin ( ) [SWM ( ) ], and biotinylated herpes simplex virus peptide (HSV)-2 VP16 ( ) (EEVDMTPADALDDFD), were synthesized in house by Merrifield s solid-phase method (22) and purified by high-performance liquid chromatography (HPLC) (purity 95%). T-cell clone proliferation assays. B (9 23) -specific pathogenic T-cell clones PD , PD , PD12-4.1, and PD were isolated from the pancreatic islets of NOD mice with insulitis (6,7) and were maintained by stimulation with B (9 23) peptide (25 g/ml; Sigma) and irradiated (3,000 rad) syngeneic spleen cells for 4 days, followed by a resting phase of 5 days with the addition of recombinant human IL-2 (10 units/ml; Boehringer Mannheim, Mannheim, Germany). To assess reactivity to peptide antigens, T-cells (10 5 /well) and irradiated syngeneic spleen cells ( /well) were cultured in 96-well flat-bottom tissue-culture plates (Costar) in the presence of several concentrations of the APL, B (9 23) native peptide, or control peptides in a 7.5% CO 2 incubator. In some cultures, conditioned medium was removed after 48 h of incubation and analyzed for cytokine levels by enzyme-linked immunosorbent assay (ELISA) (monoclonal antibodies from PharMingen, San Diego, CA; anti IL-2: JES6-1A12/JES6-5H4; anti IL-4: BVD4-1D11/BVD6-24G2; anti IL-5: TRFK5/TRFK4; anti IL-10: JES5-2A5/JES5-16E3; anti IFN- : R4-6A2/ XMG1.2). In other cultures, T-cell proliferation was assessed by pulsing cultures with 1 Ci [ 3 H]thymidine specific activity 25 Ci/mmol; Amersham Life Sciences, Arlington Heights, IL) for the last 18 h of a 72-h incubation. Cells were harvested onto glass fiber lined plates (Unifilter-96 GF/B; Packard, Meriden, CT), and the amount of radioactivity incorporated into de novo synthesized DNA was measured in a scintillation counter (Top Count NXT; Packard). Vaccinations. Different doses of peptide antigens were diluted in a soluble vehicle (i.e., 10 mmol/l acetate buffer with 30 mg/ml D-mannitol, ph 6.0) and injected subcutaneously at different intervals in prediabetic female NOD mice (i.e., 4-week-old, insulitis-free). Peptide doses and frequency of administration are indicated in the figure legends. Mice were usually killed 3 days after the last injection and tested for responsiveness to antigen by culturing single-cell suspensions of lymph node or spleen cells ( /well) with antigen, and proliferation and cytokine production was assessed as described above. In some experiments, peptides were also administered subcutaneously with incomplete Freund s adjuvant (IFA) or complete Freund s adjuvant (CFA), which contained 1 5 mg/ml heat-killed Mycobacterium (Difco). Generation of APL-specific T-cell lines and clones. T-cell lines were generated by culturing spleen cells ( /well) from APL-vaccinated mice in 24-well culture plates (Costar) in the presence of 10 mol/l of the appropriate APL used for vaccination for 7 days. After washing, T-cells were restimulated in 25-cm 2 tissue culture flasks with irradiated (3,000 rad) NOD spleen cells as antigen-presenting cells in the presence of APL (10 mol/l) and recombinant human IL-2 (5 units/ml; Boehringer Mannheim) for another 7 days. After an additional eight stimulation cycles, T-cells were used in proliferation and cytokine assays. T-cell clones were derived from these cell lines by a limiting dilution procedure, as previously described (6). Monitoring for diabetes by blood glucose measurements. Blood glucose was monitored using a glucometer (Encore Glucometer; Bayer, Elkhart, IN) at weekly intervals, beginning at 10 weeks of age. Mice with blood glucose levels 200 mg/dl on two consecutive occasions were considered diabetic. The data are presented as the percentage of animals remaining diabetes-free over the course of the experiment. The differences between curves were tested using the log-rank test, which compared the distributions over the entire observation period. RESULTS Identification of functional APLs derived from the insulin B (9 23) autoantigenic epitope in type 1 diabetes. The B (9 23) epitope of insulin appears to be an immunodominant target antigen for pathogenic autoreactive T-cells in NOD mice (6 9). An APL derived from B (9 23) that lacks the ability to activate B (9 23) -specific pathogenic T-cells, but that can induce an antigen crossreactive regulatory T-cell population, may be a useful therapeutic agent. Therefore, we first determined the critical amino acid residues of the B (9 23) peptide involved in stimulating pathogenic T-cells by generating several APLs with single alanine residue substitutions at each position of B (9 23) and tested whether these APLs were able to stimulate a panel of five B (9 23) -specific pathogenic CD4 T-cell clones derived from pancreata of NOD mice DIABETES, VOL. 51, JULY

3 APL OF B (9 23) AUTOANTIGEN IN TYPE 1 DIABETES TABLE 1 APL activation of NOD pathogenic B (9-23) -specific T-cell clones Proliferation of B (9-23) -specific T-cell clones (cpm) B (9-23) APL* PD6-4.3 PD PD PD PD AA position 9 S A 12,861 42,234 1,000 18, H A 12,507 1, , L A 14,148 2, , V A 8, ,129 15, E A A NA 15 L A 161 1, Y A L A , ,460 4, V A , , C A 7,678 34,212 4,295 6, G A 2,440 38,685 1,323 13, E A 91 39,087 7,900 4, R A 6,555 51,722 1,313 12,635 1, G A 14,304 75,441 3,228 18, B (9-23) 17,963 32,221 3,794 14,820 3,614 No antigen *APL of the insulin B (9-23) peptide were generated by substituting each residue of B (9-23) with alanine. Pathogenic B (9-23) -specific T-cell clones isolated from the pancreata of diabetic NOD mice (6,7) were cultured in the presence of irradiated syngeneic spleen cells with or without APL or the B (9-23) native antigen, and proliferation was assessed by [ 3 H]thymidine incorporation. Values represent the mean cpm of triplicate cultures. SE was 15% for all mean values. SI (antigen-stimulated mean cpm)/(medium mean cpm); nonstimulatory peptides defined by SI 2. (6,7). Alanine was used as the test replacement amino acid because of its small neutral side-chain. Although several APLs lost their ability to stimulate at least one clone (i.e., stimulation index [SI] 2 for positive stimulation), those APLs with alanine substitutions at positions 13, 15, and 16 (i.e., A 13,A 15, and A 16, respectively) lost stimulatory activity on all five clones tested (Table 1), suggesting that these alterations disrupted either the TCR contact site or the MHC class II binding site of B (9 23). In an effort to minimize cysteine cysteine interactions, A 13, A 15, and A 16 were modified to increase their stability by replacing the free cysteine residue at position 19 with alanine (i.e., A 13,19, A 15,19, and A 16,19 ). The alanine substitution at position 19 had no effect on peptide activity because it did not cause APLs to stimulate pathogenic T-cell clones (Table 2) and, in most cases, did not substantially alter the stimulatory activity of the native B (9 23) (i.e., A 19 ) peptide (Tables 1 and 2). These stable double alanine substituted APLs were further characterized. Although APLs must bind the MHC class II molecule for stimulation of a protective regulatory T-cell population, they must not activate the TCR of B (9 23) -specific pathogenic T-cells. As an indirect assessment of whether A 13,19, A 15,19, and A 16,19 could bind MHC class II molecules, APLs were tested for their ability to competitively inhibit B (9 23) from activating a pathogenic T-cell clone. All three APLs competitively inhibited B (9 23) from activating clone , and this inhibition was specific because negative control peptides, SWM ( ) and neurotensin (NT), did not interfere with B (9 23) stimulatory activity (Fig. 1). In addition, competitive inhibitory activity by these APLs was specific for binding the TCR of the clone and not for binding MHC class II molecules on antigen-presenting cells because the control peptides [i.e., SWM ( ) and NT], which are known to bind the NOD MHC class II molecules by inducing immune responses in NOD mice (A.G., L.J., unpublished data), did not compete with B (9 23) for binding MHC. Thus, these APLs were able to bind MHC molecules of the NOD mouse and engage in a nonproductive (i.e., nonstimulatory) binding event with TCRs that competed with TCR binding of the native B (9 23) antigen. Induction of APL-specific T-cell responses in vivo. The disease-protective activity of APLs is thought to exist because of their ability to induce regulatory T-cell popu- TABLE 2 Effects of alanine substitution at position 19 of B (9-23) on activation of NOD pathogenic B (9-23) -specific T-cell clones Proliferation of B (9-23) -specific T-cell clones (cpm) Peptide* PD PD PD No peptide B (9-23) 75,535 2,595 25,559 4,433 28,436 1,72 A 19 92,254 8,891 25,151 4,621 22, A A A A 13, A 15, A 16, SWM ( ) NT Data are means SE cpm of triplicate cultures. *APLs of the insulin B (9-23) peptide were generated by substituting amino acid residues 13, 15, 16, or 19 of B (9-23) with alanine. Irrelevant control peptides were SWM ( ) and NT. Pathogenic B (9-23) -specific T-cell clones isolated from the pancreata of diabetic NOD mice (6,7) were cultured in the presence of irradiated syngeneic spleen cells, and each APL or the B (9-23) native antigen (25 mol/l) and proliferation was assessed by [ 3 H]thymidine incorporation DIABETES, VOL. 51, JULY 2002

4 D.G. ALLEVA AND ASSOCIATES FIG. 1. Competitive inhibition of a B (9 23) - activated NOD T-cell clone by APLs. The NOD pathogenic B (9 23) -specific T-cell clone PD (10 5 /well) and irradiated syngeneic spleen cells ( /well) were cultured with B (9 23) (10 mol/l) in 96-well flat-bottom tissue-culture plates in the presence or absence of each APL (A 13,19,A 15,19, and A 16,19 ) or the positive control peptides SWM ( ) and NT. Cultures were pulsed with [ 3 H]thymidine for the last h of a 72-h incubation, and the amount of incorporated radioactivity was counted. Values are the mean cpm SE of triplicate cultures from one of four representative experiments. *Significantly (P < 0.05) different from cultures without APLs. lations in vivo that cross-react with their native antigen counterpart, as has been demonstrated by APLs of neuroantigenic epitopes involved in EAE (12 20). We, therefore, determined whether these B (9 23) -derived APLs could induce a specific T-cell response in vivo after subcutaneous administration of A 13,19,A 15,19, and A 16,19 in a buffered aqueous vehicle (soluble vehicle) to young prediabetic female NOD mice. The soluble vehicle, rather than CFA, was used in these experiments to promote the skewing of APL-induced T-cell responses toward a disease-protective Th2-like cytokine profile, as has been previously described for administration of native insulin peptides (23). Vaccination with each APL induced strong (SI 10) antigenspecific proliferative T-cell responses in the spleen (Fig. 2A). However, only modest cross-reactive responses to the native antigen, B (9 23), were apparent in the spleens of APL-vaccinated mice (Fig. 2B). The degree of these crossreactive responses (i.e., SI 2) was similar to those responses observed in naive age-matched NOD mice (Fig. 2C; SI 2), which makes it difficult to determine whether the APL induced the cross-reactive response. Note that vaccination with B (9 23) in either soluble vehicle or CFA induced antigen-specific responses that were stronger than those in naive mice (SIs were roughly 20 and 6, respectively; Fig. 2C), demonstrating the capacity of NOD mice to respond to the native antigen. These results show that B (9 23) -derived APLs themselves, given in a soluble vehicle, were able to induce antigen-specific responses in young prediabetic NOD mice, suggesting that this mode of administration could lead to regulatory T-cell responses that may protect against disease. Characterization of APL-derived T-cell lines. To further evaluate APL-induced cross-reactivity to B (9 23) and to determine the cytokine profile of APL-induced T-cell responses, APL-reactive T-cell lines were generated from spleens of vaccinated mice. After eight cycles of restimulation with the appropriate APLs (i.e., A 13,19,A 15,19, and A 16,19 ), all three APL-derived T-cell lines showed a strong proliferative response to their respective APL (Fig. 3A), while maintaining a small but significant cross-reactivity to B (9 23) (Fig. 3B). Although this cross-reactivity was greater than that seen immediately after vaccination (Fig. 2C), the response was nevertheless modest (i.e., SI for A 13,19 7.6, for A 15,19 2.7, and for A 16,19 2.5), which demonstrated that vaccination with APLs predominantly induces an APL-specific response. By stimulation with the appropriate APL, these cell lines showed Th0- or Th2-like cytokine profiles (Fig. 4). Although each cell line produced similar levels of the Th2-derived cytokines IL-4 and IL-5, A 13,19 produced large quantities of the Th1-derived cytokine IFN-, whereas A 15,19 and A 16,19 produced 8- to 10-fold less IFN- (Fig. 4). Therefore, the A 13,19 -specific cell line was characterized as having a Th0 phenotype, and those specific for A 15,19 and A 16,19 were characterized as having a Th2 phenotype. Frequency of cross-reactive and Th2-like T-cell clones generated from A 16,19 cell lines. To further investigate the cross-reactive and Th2-inducing nature of APLs, we used A 16,19 (NBI-6024) to generate T-cell clones from the A 16,19 -derived T-cell line because this APL showed the ability to induce both cross-reactive and Th2 responses. After challenging these A 16,19 -specific T-cell clones with B (9 23) for proliferative responses, 6 of 35 clones tested (i.e., 17.1%) scored positively in their response to B (9 23) using an SI 2 as a criterion (Table 3). In addition, 35 of 38 clones tested (i.e., 92.1%) showed a Th2 phenotype (stimulated with A 16,19 ) by producing substantial levels of IL-4 and IL-10, and 3 of 38 clones tested (i.e., 7.9%) showed a Th0 phenotype by producing substantial levels of IFN- in addition to IL-4 or IL-10 (Table 4). Criteria for identifying positive cytokine-producing clones were derived from the levels of cytokines produced by the B (9 23) -specific NOD Th1 pathogenic clones PD , PD12-4.1, and PD (Table 4). These results are con- DIABETES, VOL. 51, JULY

5 APL OF B (9 23) AUTOANTIGEN IN TYPE 1 DIABETES FIG. 2. Induction of antigen-specific immune response in NOD mice by B (9 23) -derived APLs. B (9 23) -derived APLs A 13,19,A 15,19, and A 16,19 (20 mg/kg in 100 l) in aqueous vehicle (10 mmol/l acetate buffer with 30 mg/ml D-mannitol, ph 6.0) were administered subcutaneously at days 0, 6, and 12 to 4-week-old prediabetic female NOD mice in groups of three; these mice were killed 3 days after the last injection (A and B). Responsiveness to the immunizing antigen was tested by culturing pooled spleen cells ( /well) in the presence or absence of the appropriate APL or the native B (9 23) antigen (0 25 mol/l), and T-cell proliferation was assessed by pulsing each culture with [ 3 H]thymidine for the last h of a 72-h incubation and counting the amount of incorporated radioactivity (A C). Some mice were vaccinated with B (9 23) (20 mg/kg) in aqueous vehicle (same protocol as above) or CFA (second booster injection at 3 days in IFA), and spleen cells were cultured with B (9 23) (C). Values in A C are the mean cpm SE of triplicate cultures from one of four representative experiments. *Significantly (P < 0.05) different from cultures without peptide. sistent with our characterization of the parental A 16,19 - specific T-cell line, in that the frequency of cross-reactive clones (17.1%) derived from this line was similar to the degree of cross-reactivity induced by A 16,19 vaccination (Fig. 2) or to the degree of cross-reactivity in the A 16,19 T-cell line (Fig. 3). Furthermore, the frequency of Th2 (i.e., IL-4 and IL-10) versus Th0 (i.e., Th2 plus IFN- ) clones, which was 92.1 vs. 7.9%, respectively, was similar to the relative proportions of these cytokines produced by the A 16,19 cell line (Fig. 4). In vivo efficacy of A 16,19. Because A 16,19 induced a Th2 phenotype in addition to a T-cell population that crossreacted with the native B (9 23) antigen, we chose this APL FIG. 3. Generation of B (9 23) -derived APL-specific T-cell lines from NOD mice. Young prediabetic female NOD mice were vaccinated with B (9 23) -derived APLs A 13,19, A 15,19, and A 16,19 (20 mg/kg in aqueous vehicle, as describe in Fig. 2), and spleen cells were isolated 3 days after the last injection. Spleen cells ( /well) were cultured in the presence of the appropriate APL (10 mol/l) for 5 7 days, at which time they were washed and restimulated with irradiated spleen cells plus APL for another 7 days. After eight cycles of restimulation, T-cells (10 5 /well) plus irradiated spleen cells were cultured in the presence or absence of APL or B (9 23), and proliferation was assessed by [ 3 H]thymidine incorporation during the last 18 h of a 72-h incubation. Values are the mean cpm SE of triplicate cultures from one of four representative experiments. *Significantly (P < 0.05) different from cultures without peptide. FIG. 4. Cytokine production phenotype of APL-specific T-cell lines from NOD mice. APL-specific T-cell lines were generated from NOD mice vaccinated with the B (9 23) -derived APLs A 13,19,A 15,19, and A 16,19, as described in Fig. 3. The T-cell lines were cultured (10 5 /well) with irradiated spleen cells ( /well) plus APL (25 mol/l) for 48 h, and culture-conditioned medium was removed and analyzed for cytokine levels by ELISA. Values are the mean cpm SE of triplicate cultures from one of three representative experiments in which baseline cytokine levels for spleen cells plus the T-cells in the absence of APL were subtracted DIABETES, VOL. 51, JULY 2002

6 D.G. ALLEVA AND ASSOCIATES TABLE 3 Frequency of A 16,19 -derived T-cell clones that cross-react with B (9 23) T-cell clone* T-cell proliferation (mean cpm) Medium A 16,19 SI B (9 23) SI 1A , A , A , B , B6 90 6, B8 9, , , C , , E , F F , F9 1, , , G , G A , A , A , C , D , E , E , E , , E , F , F , F , F , G , H , H , B , C , D4 69 7, E , G3 1,347 23, , G , *NOD splenic T-cell clones derived from A 16,19 vaccination were cultured with irradiated syngeneic splenocytes and 25 mol/l of either A 16,19 or B (9 23) for 72 h, and [ 3 H]thymidine was added for assessment of T-cell proliferation. Data are the mean values of triplicate cultures; SE 10% for all mean values. SI (antigenstimulated mean cpm)/(medium mean cpm); cultures with an SI 2 were considered positive. TABLE 4 Cytokine phenotype of T-cell clones derived from APL (A 16,19 )- immunized NOD mice Cytokine production (ng/ml) T-cell clone* IL-4 IL-10 IFN- Phenotype 1A Th2 1A Th2 1B Th2 1B Th2 1D Th2 1E Th2 1F Th2 1F Th2 1F Th2 1G Th2 1H Th2 2A Th0 2A Th2 2C Th2 2C Th2 2C Th2 2C Th2 2D Th2 2D Th2 2E Th2 2E Th2 2E Th2 2F Th2 2F Th2 2G Th2 2G Th2 2H Th2 3A11a Th2 3A Th2 3B Th2 3B Th2 3B Th0 3C Th2 3D Th2 3D Th2 3D Th2 3E Th2 3G Th0 PD Th1 PD Th1 PD Not done 30.1 Th1 *T-cell clones were generated from the splenic T-cell lines from NOD mice vaccinated with A 16,19 by limiting dilution and cultured with irradiated syngeneic splenocytes with 25 mol/l A 16,19 for 48 h, and conditioned medium was analyzed for cytokine levels by ELISA. Data are the mean values of triplicate cultures with the mean cytokine values of spleen cells plus T-cells subtracted; SE 10% for all mean values; the means SE for background cytokine levels were ng/ml for IL-4, ng/ml for IL-5, ng/ml for IFN- ; 0 denotes 0.02 ng/ml. Each clone was scored as having a Th0 (IFN- and IL-4 or IL-10), Th1 (IFN- but not IL-4 or IL-10), or Th2 (IL-4 or IL-10 but not IFN- ) phenotype. Positive cytokine levels were determine by the following: IL ng/ml, IL-10 1 ng/ml, IFN- 2 ng/ml. Th1 T-cell clone specific for B (9 23) (6) and activated with B (9 23) for cytokine production. for its ability to modulate disease in NOD mice. The optimal dose of this APL was first determined by injecting single doses ranging from to 20 mg/kg into young prediabetic NOD female mice and determining APL-induced proliferative responses 7 days later (Fig. 5). Although 0.005, 0.05, and 0.5 mg/kg did not induce detectable splenic T-cell proliferative responses, a dose-dependent induction of responses was observed with 5, 10, and 20 mg/kg, in which maximum responsiveness was achieved with 20 mg/kg (Fig. 5). Lymph node T-cell responses were less pronounced than splenic responses, a feature that is consistently observed with APL vaccinations in the soluble vehicle, whereas stronger lymph node responses were observed using CFA as a vehicle (data not shown). To determine whether A 16,19 could effectively modulate the onset of disease in NOD mice, 20 mg/kg of APL was administered to young (4-week-old) predisease NOD female mice at weekly intervals for 12 weeks; one injection was administered every 2 weeks thereafter until termination of the experiment (i.e., 35 weeks), and spontaneous disease was monitored (i.e., glucose levels 200 mg/dl on two consecutive weekly readings). Females were used because disease incidence is dramatically greater in female DIABETES, VOL. 51, JULY

7 APL OF B (9 23) AUTOANTIGEN IN TYPE 1 DIABETES FIG. 5. Dose response to single injection of APL. Prediabetic 4-weekold female NOD mice were administered a single injection of the B (9 23) -derived APL A 16,19 at different doses ( mg/kg) in an aqueous vehicle in groups of three and were killed 7 days later. Responsiveness to the immunizing antigen was tested by culturing spleen cells pooled from all five mice within a group ( /well) in the presence or absence of several doses of A 16,19, and T-cell proliferation was assessed by [ 3 H]thymidine incorporation during the last 18 h of a 72-h incubation. Values are the mean cpm SE of triplicate cultures from one of three representative experiments. NOD mice than in male NOD mice (i.e., 80 90% in female cohorts vs % in male cohorts) (2), in which the onset of frank diabetes usually occurs by weeks. In this experiment, the onset of disease occurred at 15 weeks of age in the control peptide group, in which disease incidence reached 50% by 20 weeks and % by 30 weeks in both control groups (Fig. 6A). However, disease onset in A 16,19 -treated mice was significantly delayed (20% incidence at weeks) and the maximum incidence achieved was reduced to 47% (53% diabetes-free), even at weeks (Fig. 6A). This therapeutic activity prompted us to determine whether APL treatment of mice with severe insulitis or new-onset diabetes could modulate disease progression, an approach that would mimic most clinical settings with recently diagnosed new-onset diabetic subjects. APL was administered to 12- to 20-week-old female diabetic mice with severe insulitis in a single injection emulsified in IFA to maximize the delivery of the APL without inducing a Th1 response, as would be induced with CFA. Strikingly, treatment with a single dose of the APL reduced disease incidence from 81 to 53% in the control peptide-treated group (tetanus toxoid) (Fig. 6B). These results demonstrate that APL therapy may be more versatile than expected in clinical settings, where it could be administered to patients either before or after diagnosis. DISCUSSION Here, we report the first demonstration of a therapeutic activity of an APL derived from an islet -cell specific antigen in type 1 diabetes. This APL, A 16,19 (i.e., NBI-6024), was based on the B (9 23) autoantigenic immunodominant epitope in the NOD mouse and was identified by generating a series of peptides containing single alanine substitutions at each amino acid residue of B (9 23) and testing their ability to stimulate or competitively inhibit proliferation of B (9 23) -specific NOD pathogenic T-cell clones. These experiments allowed us to identify at least three critical TCR contact residues at positions 13, 15, and 16 of B (9 23) that prevented the activation of T-cell clones while allowing these peptides to maintain their binding to MHC class II molecules (i.e., a requirement for competitive inhibition). To increase peptide stability and to avoid adverse reducing effects of the cysteine at position 19, this position was substituted with alanine in all three of the APLs and conveyed a neutral function. These APLs did not appear to compete with B (9 23) in MHC class II binding because the control peptides SWM ( ) and NT, which are known to bind the NOD MHC, did not alter the ability of B (9 23) to stimulate T-cell clones. Additional evidence that these APLs could bind NOD MHC class II molecules for interaction with T-cells was demonstrated by their ability to induce immune responses in vivo, which led to the generation of APL-specific T-cell lines and clones. Others have recently reported the development of APLs of other type 1 diabetes antigens involved in the human disease, such as GAD65 (24) and Imogen 38 (25). These APLs have been characterized by their ability to bind MHC class II molecules and to activate T-cell clones. Nepom et al. (24) identified an APL of the native GAD65 ( ) peptide with an alanine substitution at position 561 (methionine) that resulted in complete inhibition of proliferation of a T-cell clone specific for GAD ( ). This APL maintained binding to two HLA-DR4 haplotypes, *0404 and *0405, which are highly associated with type 1 diabetes. An APL based on the region of the mitochondrial 38-kDa pancreatic islet autoantigen, Imogen 38, was shown to block proliferation of an Imogen 38 specific Th1 cell clone from a type 1 diabetic patient (25). However, without in vivo efficacy data on their ability to modulate disease, the therapeutic activity of these APLs remains to be determined. The B (9 23) -based APLs in this study did not activate pathogenic B (9 23) -specific T-cell clones but did induce a regulatory Th2 cell population that was cross-reactive with the native B (9 23) antigen, a requisite for APL therapeutic activity (12,16,18 21). This cross-reactive population was relatively small compared with the non cross-reactive APL-specific population because 1) the cross-reactive response of spleen cells after vaccination was weak, 2) the cross-reactive proliferative response of APL-induced T-cell lines was modest, and 3) the frequency of cross-reactive APL-induced T-cell clones was only 17.9%. This small cross-reactive population induced by APL vaccinations is consistent with that induced by another APL, Q144, based on the neuroantigenic peptide, PLP ( ) (i.e., W144), used in EAE (20). The number of cross-reactive T-cell clones derived from vaccinations with the Q144 APL constituted a small fraction of the total number of Q DIABETES, VOL. 51, JULY 2002

8 D.G. ALLEVA AND ASSOCIATES FIG. 6. APL modulation of spontaneous diabetes in NOD mice. A: Groups (n) of 4-week-old female NOD mice were administered weekly injections of the B (9 23) -derived APL A 16,19 at 20 mg/kg in an soluble vehicle for 12 weeks and then one injection every 2 weeks until mice reached 39 weeks of age. B: Groups (n 21) of 12- to 20-week-old female NOD mice with insulitis received a single subcutaneous injection (5 mg/kg) of APL or control peptide [tetanus toxin ( )], emulsified in IFA. Blood glucose levels were monitored weekly; mice with levels >200 mg/dl on 2 consecutive weeks were considered diabetic. The log-rank test was used to assess whether the results of the two treatment groups were significantly different. The data represent three experiments. derived T-cell clones, in which Anderson et al. describe this heterogeneity in TCR specificity as unfocused crossreactivity rather than focused cross-reactivity (20). Focused cross-reactivity would entail that each T-cell in an immune response to an antigen demonstrate cross-reactivity with another related antigen, which could lead to autoimmunity. However, unfocused cross-reactivity in which only a small subset of the T-cell population shows cross-reactivity would ensure immune diversity without the threat of a hyper-expansion of potentially harmful autoreactive T-cells (20). In the case of therapeutic APLs, the large non cross-reactive APL-specific T-cell population may provide bystander Th2 cytokines for the small cross-reactive population being activated because of the close proximity between these populations during antigen presentation. The subsequent expansion of this crossreactive population would influence both the Th2 development and activity of B (9 23) -specific pathogenic T-cells. Thus, the therapeutic activity of the APL may work in a two-stage fashion in which a small cross-reactive population is first activated and expanded into a Th2 population in the draining lymph nodes near the site of vaccination. This Th2 population could then migrate to draining lymph nodes of the pancreas and dampen the activation of or promote the Th2 development of endogenous B (9 23) - specific pathogenic T-cells using its cross-reactive capabilities. In addition, these regulatory Th2 cells could migrate directly to the inflamed pancreas where they suppress the activity of pathogenic Th1 cells. Indeed, we have observed, in a phase II clinical trial with MS patients receiving APL of MBP (83 99), an initial APL-specific Th2 response immediately after dosing, which was modestly cross-reactive and was followed by a strong Th2 response to the native MBP (83 99) antigen several weeks after dosing (18). Perhaps this APL-induced mechanism may shed light on how aerosol or soluble forms of administration of the native insulin B-chain or B-chain peptides induce protective Th2 immunity in NOD mice (8,26 28). That is, the induction of any Th2 population via a soluble vehicle that reacts with B (9 23) should be able to provide protection via a bystander mechanism involving Th2 cytokine production that dampens pathogenic B (9 23) -specific Th1 cell activity. This Th2 population could be either an unfocused crossreactive population, such as that induced by APL, or a non cross-reactive B (9 23) -specific population induced with B (9 23) in a Th2-promoting vehicle. To ensure that Th2 DIABETES, VOL. 51, JULY

9 APL OF B (9 23) AUTOANTIGEN IN TYPE 1 DIABETES responses were induced by APLs in this study, the peptides were administered in a soluble vehicle instead of CFA because CFA is known to induce strong Th1 responses, in part because Mycobacterium constituents in the adjuvant stimulate antigen-presenting cell production of the Th1-promoting cytokine IL-12. Alternatively, the absence of adjuvants usually leads to induction of weak, if any, T-cell responses, especially to foreign antigens. It was striking that even a single injection of APLs in the absence of adjuvant (i.e., soluble vehicle only) could induce strong T-cell responses in the spleen of NOD mice. Although lymph node responses were not readily detectable using a soluble vehicle, we have shown that injection of APLs emulsified in CFA induces robust responses in both the spleen and lymph nodes (A.G., L.J., unpublished data), demonstrating that adjuvants enhance the immunogenicity of APLs. Yet these APLs appear to have greater immunogenicity than conventional foreign antigens because of their elevated potency in soluble vehicles. Through the use of the ELISPOT assay and short-term T-cell lines, we demonstrated that only type 1 diabetic subjects in the study had endogenous Th1 responses to B (9 23) (10), which supports that B (9 23) may be one of the immunodominant autoantigenic epitopes in humans as well as in the NOD mouse. In addition, we demonstrated that A 16,19 APL could bind the disease-associated HLA haplotype HLA-DQB1*0302 and was effective in competitively inhibiting B (9 23) from stimulating short-term T-cell lines from a type 1 diabetic patient (P.A.G., unpublished data), a cell line that we previously characterized as having a Th1 phenotype (10). Therefore, the activity of A 16,19 in NOD mice demonstrates a potential therapeutic value, and, thus, APL therapies directed at this autoantigenic response might be beneficial in controlling type 1 diabetes in humans. Indeed, such efforts are currently underway in clinical trials. ACKNOWLEDGMENTS The authors thank Taisho Pharmaceutical for their critical review of the manuscript and Dr. Lawrence Steinman for his helpful discussions and scientific expertise. REFERENCES 1. Atkinson MA, Maclaren NK: The pathogenesis of insulin-dependent diabetes mellitus. N Engl J Med 331: , Delovitch TL, Singh B: The nonobese diabetic mouse as a model of autoimmune diabetes: immune dysregulation gets the NOD. Immunity 7: , Wicker LS, Todd JA, Peterson LB: Genetic control of autoimmune diabetes in the NOD mouse. Annu Rev Immunol 13: , Durinovic-Bello I: Autoimmune diabetes: the role of T cells, MHC molecules and autoantigens. Autoimmunity 27: , Haskins K, Wegmann D: Diabetogenic T-cell clones. Diabetes 45: , Wegmann DR, Norbury-Glaser M, Daniel D: Insulin-specific T cells are a predominant component of islet infiltrates in pre-diabetic NOD mice. Eur J Immunol 24: , Daniel D, Gill RG, Schloot N, Wegmann D: Epitope specificity, cytokine production profile and diabetogenic activity of insulin-specific T cell clones isolated from NOD mice. Eur J Immunol 25: , Daniel D, Wegmann DR: Protection of nonobese diabetic mice from diabetes by intranasal or subcutaneous administration of insulin peptide B-(9 23). Proc Natl Acad Sci U S A 93: , Wong FS, Karttunen J, Dumont C, Wen L, Visintin I, Pilip IM, Shastri N, Pamer EG, Janeway CA Jr: Identification of an MHC class I-restricted autoantigen in type 1 diabetes by screening an organ-specific cdna library. Nat Med 5: , Alleva DG, Crowe PD, Jin L, Kwok WW, Ling N, Gottschalk M, Conlon PJ, Gottlieb PA, Putnam AL, Gaur A: A disease-associated cellular immune response in type 1 diabetics to an immunodominant epitope of insulin. J Clin Invest 107: , Sloan-Lancaster J, Allen PM: Altered peptide ligand-induced partial T cell activation: molecular mechanisms and role in T cell biology. Annu Rev Immunol 14:1 27, Nicholson LB, Kuchroo VK: T cell recognition of self and altered self antigens. Crit Rev Immunol 17: , Brocke S, Gijbels K, Allegretta M, Ferber I, Piercy C, Blankenstein T, Martin R, Utz U, Karin N, Mitchell D, Veromaa T, Waisman A, Gaur A, Conlon P, Ling N, Fairchild PJ, Wraith DC, O Garra A, Fathman CG, Steinman L: Treatment of experimental encephalomyelitis with a peptide analogue of myelin basic protein. Nature 379: , Karin N, Mitchell DJ, Brocke S, Ling N, Steinman L: Reversal of experimental autoimmune encephalomyelitis by a soluble peptide variant of a myelin basic protein epitope: T cell receptor antagonism and reduction of interferon gamma and tumor necrosis factor alpha production. J Exp Med 180: , Franco A, Southwood S, Arrhenius T, Kuchroo VK, Grey HM, Sette A, Ishioka GY: T cell receptor antagonist peptides are highly effective inhibitors of experimental allergic encephalomyelitis. Eur J Immunol 24: , Gaur A, Boehme SA, Chalmers D, Crowe PD, Pahuja A, Ling N, Brocke S, Steinman L, Conlon PJ: Amelioration of relapsing experimental autoimmune encephalomyelitis with altered myelin basic protein peptides involves different cellular mechanisms. J Neuroimmunol 74: , Santambrogio L, Pakaski M, Wong ML, Cipriani B, Brosnan CF, Lees MB, Dorf ME: Antigen presenting capacity of brain microvasculature in altered peptide ligand modulation of experimental allergic encephalomyelitis. J Neuroimmunol 93:81 91, Kuchroo VK, Greer JM, Kaul D, Ishioka G, Franco A, Sette A, Sobel RA, Lees MB: A single TCR antagonist peptide inhibits experimental allergic encephalomyelitis mediated by a diverse T cell repertoire. J Immunol 153: , Young DA, Lowe LD, Booth SS, Whitters MJ, Nicholson L, Kuchroo VK, Collins M: IL-4, IL-10, IL-13, and TGF-beta from an altered peptide ligand-specific Th2 cell clone down-regulate adoptive transfer of experimental autoimmune encephalomyelitis. J Immunol 164: , Anderson AC, Waldner H, Turchin V, Jabs C, Prabhu Das M, Kuchroo VK, Nicholson LB: Autoantigen-responsive T cell clones demonstrate unfocused TCR cross-reactivity toward multiple related ligands: implications for autoimmunity. Cell Immunol 202:88 96, Constant SL, Bottomly K: Induction of Th1 and Th2 CD4 T cell responses: the alternative approaches. Annu Rev Immunol 15: , Merrifield RB: Solid-phase peptide synthesis. Adv Enzymol Relat Areas Mol Biol 32: , Kwok WW, Nepom GT, Raymond FC: HLA-DQ polymorphisms are highly selective for peptide binding interactions. J Immunol 155: , Nepom GT, Lippolis JD, White FM, Masewicz S, Marto JA, Herman A, Luckey CJ, Falk B, Shabanowitz J, Hunt DF, Engelhard VH, Nepom BS: Identification and modulation of a naturally processed T cell epitope from the diabetes-associated autoantigen human glutamic acid decarboxylase 65 (hgad65). Proc Natl Acad Sci U S A 98: , Geluk A, van Meijgaarden KE, Roep BO, Ottenhoff TH: Altered peptide ligands of islet autoantigen Imogen 38 inhibit antigen specific T cell reactivity in human type-1 diabetes. J Autoimmun 11: , Tian J, Chau C, Kaufman DL: Insulin selectively primes Th2 responses and induces regulatory tolerance to insulin in pre-diabetic mice. Diabetologia 41: , Muir A, Peck A, Clare-Salzler M, Song YH, Cornelius J, Luchetta R, Krischer J, Maclaren N: Insulin immunization of nonobese diabetic mice induces a protective insulitis characterized by diminished intraislet interferongamma transcription. J Clin Invest 95: , Maron R, Melican NS, Weiner HL: Regulatory Th2-type T cell lines against insulin and GAD peptides derived from orally- and nasally-treated NOD mice suppress diabetes. J Autoimmun 12: , DIABETES, VOL. 51, JULY 2002