T-cell autoantigens in the non-obese diabetic mouse model of autoimmune diabetes

Abstract

The non-obese diabetic (NOD) mouse model of autoimmune (type 1) diabetes has contributed greatly to our understanding of disease pathogenesis and has facilitated the development and testing of therapeutic strategies to combat the disease. Although the model is a valuable immunological tool in its own right, it reaches its fullest potential in areas where its findings translate to the human disease. Perhaps the foremost example of this is the field of T-cell antigen discovery, from which diverse benefits can be derived, including the development of antigen-specific disease interventions. The majority of NOD T-cell antigens are also targets of T-cell autoimmunity in patients with type 1 diabetes, and several of these are currently being evaluated in clinical trials. Here we review the journeys of these antigens from bench to bedside. We also discuss several recently identified NOD T-cell autoantigens whose translational potential warrants further investigation.

Introduction

The non-obese diabetic (NOD) mouse is a widely studied model for spontaneous autoimmune (type 1) diabetes that is polygenic and T-cell-mediated.¹ However, those new to the field are often unaware of its serendipitous origin.^2,3 In the 1970s, outbred Imperial Cancer Research mice were being selectively bred to derive a strain characterized by cataracts [which later became the cataract Shionogi (CTS) strain]. At the sixth generation, two sublines were separately maintained, one exhibiting fasting hyperglycaemia and the other euglycaemia. Ironically, a single female in the latter line developed diabetes at the 20th generation. Her progeny, born during her prediabetic period, were intercrossed, and mice having one or two parents that developed diabetes were selected for subsequent rounds of breeding. This selective breeding for diabetes eventually led to the establishment of the now inbred NOD strain. The genetic loci that contribute to disease in these mice have recently been catalogued and discussed.⁴ Here we will focus our attention on the diabetes-related autoantigens that are targeted by T cells in the NOD strain. For at least 20 years, identification of these antigens has been the goal of numerous laboratories, including our own. These efforts have yielded 18 antigens to date (Table 1),⁵ supporting the idea that multiple antigens contribute to the pathogenesis of type 1 diabetes. Here we will discuss primarily those that have been used in antigen-specific type 1 diabetes prevention or intervention trials in humans. Several recently identified NOD T-cell antigens whose potential for translation has not yet been fully explored will also be highlighted.

Table 1.

Diabetes-relevant T-cell antigens in non-obese diabetic (NOD) mice

Antigen	Reference
Heat-shock protein 60	6
Glutamic acid decarboxylase 65	26
Carboxypeptidase E	26
Peripherin	26
Insulin	67
Glutamic acid decarboxylase 67	87
Islet cell autoantigen 69	88
Insulinoma-associated protein 2β	89
Glucose transporter 2	90
Insulinoma-associated protein 2	91
Hepatocarcinoma-intestine-pancreas/pancreatic associated protein	70
S100β	92
Glial fibrillary acidic protein	92
Islet-specific glucose-6-phosphatase catalytic subunit-related protein	93
Dystrophia myotonica kinase	94
Regenerating gene II	68
Pancreatic duodenal homeobox 1	72
Chromogranin A	73

Antigens are listed in chronological order of their description as T-cell targets in NOD mice.

NOD T-cell antigens: from bench to bedside

Heat-shock protein 60 (Hsp60)

Two decades ago,⁶ Cohen and colleagues reported the presence in NOD mice of serum antibodies to the 65-kD heat-shock protein of Mycobacterium tuberculosis (Mtb Hsp65). Proliferation of splenic T cells from NOD mice in response to Mtb Hsp65 was also observed, and CD4 T-cell clones specific for the protein could induce diabetes when transferred to young (4-week-old) female NOD mice.⁶ The investigators speculated that these antibody and T-cell reactivities reflected an autoimmune response to a cellular protein related to Mtb Hsp65. Subsequent work⁷ revealed that the diabetogenic Mtb Hsp65-specific T-cell clones recognized a peptide derived from human Hsp60. The peptide corresponds to amino acids 437–460 of human Hsp60 and is referred to throughout the literature as p277. Murine and human p277 differ at only a single position (Fig. 1), and NOD-derived T cells respond equivalently to the two peptides.⁸ Remarkably, a single administration of human p277 in incomplete Freund's adjuvant to 4-week-old NOD mice reduced the incidence of spontaneous diabetes at 32 weeks of age from 80% to 31%.⁷ Significant benefit was observed even when the mice were treated later in the disease process (i.e., at 12, 15 or 17 weeks of age).⁹

Figure 1.

Sequence comparison of human and mouse heat-shock protein 60 (Hsp60) p277, DiaPep277, and the related peptide from Mycobacterium tuberculosis Hsp65. Dashes indicate identity to human p277, which corresponds to amino acids 437–460 of Hsp60. DiaPep277 is the version of human Hsp60 p277 that is used in clinical trials.

These encouraging preclinical results, coupled with the finding that type 1 diabetes patients show T-cell proliferative responses to Hsp60 peptides including p277,^9,10 inspired a series of still-ongoing clinical trials to explore the efficacy of p277 in the treatment of type 1 diabetes in humans. In the clinical trials, a modified version of the peptide (DiaPep277) is employed in which the two cysteines are substituted with valine for improved stability (Fig. 1).¹¹ The immunological activity of DiaPep277 is identical to that of the native peptide.¹² Results of a phase II trial of DiaPep277 were first reported in 2001.¹¹ Trial participants were newly diagnosed (within 6 months) with type 1 diabetes and ranged in age from 16 to 58 years. Patients were treated at study entry, and at 1, 6 and 12 months, with 1 mg DiaPep277. While mean glucagon-stimulated C-peptide concentrations dropped in the placebo group, they were maintained in the DiaPep277 group at both 10 months¹¹ and 18 months¹³ after study entry, suggesting preservation of beta cell function. Similar findings were obtained by others in patients receiving 2·5 mg of the peptide.¹⁴ While studies in children with type 1 diabetes have been less encouraging,^15,16 benefit in some children has been suggested.¹⁶ The potential of DiaPep277 to preserve beta cell function in type 1 diabetes patients is currently being further explored in two ongoing phase III trials (NCT00615264 and NCT01103284).¹⁷

The mechanism of action of DiaPep277 appears to be multifaceted. The peptide can bind the NOD class II major histocompatibility complex (MHC) molecule I-A^g7,¹⁸ and presumably also human leucocyte antigen (HLA)-DQ8, as these two MHC molecules bind a similar spectrum of peptides.¹⁹ Subcutaneous administration of DiaPep277 to NOD mice results in transient activation of peptide-specific T cells that produce the T helper type 2 (Th2) cytokines interleukin (IL)-4 and IL-10.²⁰ However, additional mechanisms of action for DiaPep277 have also been suggested.^21,22 Treatment of purified mouse or human T cells with the peptide results in inhibition of T-cell chemotaxis, and this effect is dependent on Toll-like receptor 2 (TLR2) signalling.²¹ The peptide also enhances the suppressive activity of human regulatory T cells in a TLR2-dependent manner.²²

Glutamic acid decarboxylase (GAD)

GAD was one of the earliest identified autoantigens in type 1 diabetes.²³ GAD is an enzyme involved in the production of the neurotransmitter γ-aminobutyric acid (GABA), and is expressed exclusively in the brain and pancreas. It exists as one of two distinctly encoded isoforms, GAD65 and GAD67, with GAD65 reactivity primarily associated with type 1 diabetes. The physiological role of GAD in the pancreas is unclear, although GABA is thought to be involved in the regulation of insulin secretion.²⁴

GAD was initially identified as an autoantigen using serum antibodies from patients with type 1 diabetes.²³ It was later identified as a major early target of islet-reactive T cells in NOD mice.^25,26 Furthermore, studies have demonstrated that GAD65-reactive cells can directly damage the pancreatic islets.²⁷ The early involvement of GAD65 in the disease course suggests that GAD may play an important and possibly fundamental role in the disease pathogenesis. The importance of GAD65 in type 1 diabetes has been further reinforced by a large set of in vivo studies performed in NOD mice. As the inflammatory process begins in earnest at about 3–4 weeks of age in NOD mice, immunotherapy of young mice with GAD65 protein or peptides was thought to be a promising approach. Indeed, numerous studies have demonstrated the efficacy of GAD65 treatment in preventing both insulitis and diabetes.^{25,26,28–30} The beneficial effect of treatment with GAD65 is not limited to young mice, but extends even to older mice in which the inflammatory process is well underway.^31,32 Additionally, novel strategies for delivering GAD have been developed, including plasmid DNA vaccination,³³ recombinant vaccinia virus administration,³⁴ and a gene gun-mediated DNA vaccination approach.³⁵

Understanding the role of GAD65 is a critical question in type 1 diabetes and it has been theorized that the Th1 subset of CD4 T cells is responsible for much of the beta cell damage, while the Th2 subset is associated with immunoprotective qualities. Many of the treatment studies with GAD65 have demonstrated that the CD4 T-cell profile shifts from Th1 to Th2 following treatment,^26,30,31,36 and the protection from insulitis and diabetes is thought to be a consequence of this shift, at least in part. Additionally, GAD65 treatment appears to induce regulatory T-cell populations that can suppress disease.^{30–32,36–38} It appears that, while certain populations of the GAD65-reactive CD4 T cells are indeed diabetogenic, others play an important immunoregulatory role.

While there appears to be no question that GAD65-reactive T cells play an important role in the disease process, it remains unclear if GAD is an essential antigen. The oft-replicated prevention of type 1 diabetes through the treatment of GAD65 suggests that reactivity to GAD65 plays a critical role in the disease pathogenesis. Furthermore, a transgenic NOD mouse model lacking GAD65 expression in the islets was prevented from developing insulitis and diabetes.³⁹ However, further studies have indicated that the disease process proceeds normally in NOD mice even in the absence of GAD65 expression.⁴⁰ Also, widespread expression of GAD65 and the presumed development of tolerance did not affect insulitis or diabetes development.⁴¹ When taken together, these studies suggest a need for further clarification of the contribution of GAD65 to the pathogenesis of type 1 diabetes.

GAD65 (and not GAD67) is expressed in human islets, and many of the observations in NOD mice are proposed to apply to type 1 diabetes in humans as well. As the initial detection of GAD65 as an antigen occurred in humans, translating the treatment successes from NOD mice back to humans would allow the research arc to come full circle. Diamyd, recombinant human GAD65 formulated with alum, has successfully progressed though phase II trials in patients having latent autoimmune diabetes in adults (LADA)^42,43 or recent-onset type 1 diabetes⁴⁴ and has shown promise in preserving beta cell function. When type 1 diabetes patients, ranging in age from 10 to 18 years, were treated with Diamyd within 6 months of diagnosis (two subcutaneous injections, 1 month apart), they showed less of a decline in both fasting and stimulated C-peptide levels after 30 months than did the placebo-treated group.⁴⁴ Diamyd is currently in large-scale phase III trials in newly diagnosed type 1 diabetes patients in the USA (NCT00751842) and Europe (NCT00723411).

Preproinsulin

Insulin is a major autoantigen in both NOD mice and human patients.⁴⁵ There are two insulin genes expressed in mice, Ins1 and Ins2. T-cell responses are generated against peptides from both the mature insulin and the cleaved signal and C peptides.⁴⁶ Although both genes are expressed in the islets, only Ins2 appears to be expressed in the thymus.⁴⁷ This differential expression is believed to be responsible for the findings that Ins1 knockout NOD mice are protected from disease⁴⁸ while Ins2 knockout mice have accelerated disease.^48,49

Insulin autoreactivity is essential for diabetes development in the NOD mouse,⁵⁰ and approximately 15 CD4 and CD8 T-cell epitopes for insulin have been identified in NOD mice.⁴⁶ While there are numerous autoantigens targeted by T cells in type 1 diabetes, reactivity to insulin appears to precede reactivity to other major autoantigens such as islet-specific glucose-6-phosphatase catalytic subunit-related protein.⁵⁰ This suggests that reactivity to insulin is a precipitating event that causes beta cell damage and release of other autoantigen targets. In particular, the peptide Ins B9-23 must be present for diabetes to develop in NOD mice.⁵¹ This peptide is recognized by autoreactive CD4 T cells⁵² and contains Ins B15-23, a target of CD8 T cells in the NOD mouse.⁵³ Ins B9-23 has also been shown to be a target of autoreactive CD4 T cells in type 1 diabetes patients.⁵⁴ These findings suggest that insulin-targeted antigen therapy has great potential for success in both NOD mice and patients.

Insulin injections,⁵⁵ oral insulin,⁵⁶ and aerosolized insulin⁵⁷ have all been shown to reduce diabetes incidence in NOD mice when treatment is begun at 4–5 weeks of age. Furthermore, administration of insulin-coupled fixed splenocytes can reverse new-onset disease in NOD mice.⁵⁸ The success of insulin antigen therapy in NOD mice has inspired several clinical trials in humans. The Diabetes Prevention Trial-Type 1 Diabetes (DPT-1) evaluated the ability of parenteral administration of insulin in at-risk subjects to prevent or delay overt diabetes.⁵⁹ The subjects were relatives of patients with type 1 diabetes and were considered at high risk for diabetes because of the presence of islet cell antibodies and abnormal glucose tolerance test results. Over a median period of 3·7 years, no delay or prevention of diabetes was observed. A separate DPT-1 trial tested oral insulin administration in relatives of patients with type 1 diabetes who had an intermediate risk of developing the disease (positive for islet cell and insulin autoantibodies but normal glucose tolerance).⁶⁰ Over a median period of 4·3 years, there was no prevention or delay of disease with this treatment. However, hypothesis-generating subgroup analysis revealed that, in subjects with higher (≥ 80nU/ml) insulin autoantibodies, oral insulin therapy did show a beneficial effect, with a disease delay of 4·5 years. More recently, the Type 1 Diabetes Prediction and Prevention (DIPP) study examined the use of insulin administered intranasally to children with predicted risk because of their HLA genotype and autoantibody levels, but no beneficial effect was observed.⁶¹

While none of these trials was ultimately successful, they should not be considered failures. The oral insulin trial provided evidence for an effect in one patient subgroup, and a phase III trial to investigate this further is currently in progress (NCT00419562). Also, the trials generated valuable data that could help to evaluate whether adjustments in parameters such as dose, dose frequency, route of administration, and time of intervention could improve the outcome. Indeed, the ongoing pre-Primary Oral/intranasal INsulin Trial (pre-POINT) will study islet autoantibody-negative children having a high genetic risk for type 1 diabetes and will utilize two different routes of insulin administration (oral or intranasal) and multiple different doses.⁶² The conditions to be used in the subsequent disease prevention POINT study will be those that elicit an immune response having ‘characteristics consistent with protection’ in the pre-POINT study.⁶²

Studies in NOD mice have suggested possible ways for making insulin antigen therapy more successful. Insulin coupled to the B subunit of cholera toxin (CTB) has been shown to prevent islet infiltration in NOD mice when administered orally in small doses.⁶³ CTB aids in passage across the intestinal barrier, thereby reducing the necessary dosage. Our laboratory has shown that targeting peptides derived from islet autoantigens to steady-state dendritic cells (DCs) can induce tolerance of the corresponding autoreactive T cells in NOD mice.⁶⁴ This is achieved through the use of an antibody to DEC-205, an endocytic receptor expressed on DCs, which is covalently linked to the antigenic peptide. We are currently working to use this method to deliver insulin peptides and protein to steady-state DCs in NOD mice. Transgenic mice expressing human genes such as human insulin and human MHC susceptibility alleles are also being employed to permit the NOD model to more closely mimic the human disease. Finally, administration of insulin in combination with GAD65 has shown promise in protecting NOD mice from diabetes.⁶⁵ Indeed, the Immune Tolerance Network–Juvenile Diabetes Research Foundation Type 1 Diabetes Combination Therapy Assessment Group recently advocated the exploration of combination therapies, including administration of two different autoantigens.⁶⁶

Recently identified NOD T-cell antigens

Hsp60, GAD65, and insulin were all identified as T-cell antigens in NOD mice in the early 1990s and were among the first T-cell targets to be identified in this model (Table 1).^6,26,67 Thus, they have had the greatest amount of time to be tested in preclinical models and to transit to human trials. In this section, we will highlight several recently identified NOD T-cell antigens whose translational potential has yet to be explored.

Regenerating gene II (Reg II)

Reg II is one of the most recently identified autoantigens in the NOD mouse.⁶⁸ It is a member of the regenerating gene (Reg) family whose founding member was identified from a cDNA library prepared from rat islets undergoing regeneration in response to a 90% pancreatectomy and treatment with a poly(ADP-ribose) synthetase inhibitor.⁶⁹ Reg II was explored as a possible autoantigen in NOD mice not only because it was found to be expressed in beta cells,⁶⁸ but also because other Reg family members were known to be T-cell antigens in NOD mice⁷⁰ or the target of autoantibodies in patients with type 1 diabetes.⁷¹ When NOD mice were immunized with the N-terminal fragment of Reg II in alum, diabetes was accelerated, suggesting the presence of Reg II epitopes capable of stimulating pathogenic beta cell-specific T cells in NOD mice.⁶⁸ However, the therapeutic potential of the protein is suggested by the finding that administration of full-length Reg II or its C-terminal fragment delayed disease.⁶⁸

Pancreatic duodenal homeobox 1 (Pdx1)

The transcription factor Pdx1 is a newly identified autoantigen in NOD mice.⁷² Pdx1 autoantibodies were found in the serum of NOD mice as young as 5 weeks of age, but were not detectable in C57BL/6 or BALB/c mice. The immunodominant B-cell epitope resides in the C-terminal portion of the protein (amino acids 200–283). Splenic T cells from NOD mice proliferated in response to this fragment of the protein, suggesting that a T-cell epitope is also located within this region, but further study is warranted. In this same report, type 1 diabetes patients and at-risk individuals were shown to harbour autoantibodies to Pdx1, and two immunodominant epitopes were suggested (residues 159–200 and 200–283). Whether Pdx1 is targeted by T cells in type 1 diabetes patients remains to be investigated.

Chromogranin A

Chromogranin A is the newest addition to the list of diabetes-related NOD mouse T-cell antigens.⁷³ It was recently identified as the target for the highly diabetogenic NOD-derived CD4 T-cell clone BDC-2·5,⁷⁴ whose specificity had remained elusive for over 20 years. Chromogranin A is not specific to the beta cell but rather is found in the secretory granules of a variety of endocrine cells and neurons and is processed into multiple peptides by prohormone convertase enzymes.⁷⁵ The epitope for BDC-2·5 was found to be one of these naturally processed peptides, designated WE14.⁷³ The peptide binds in a non-traditional manner to the class II MHC molecule I-A^g7, leaving empty the N-terminal half of the peptide-binding groove. The authors of this work speculate that WE14 may be post-translationally modified in vivo (as a partially purified beta cell secretory granule protein preparation stimulates BDC-2·5 better than does synthetic WE14) and that a pancreas-specific modification could explain why BDC-2·5 mediates beta cell autoimmunity while apparently sparing other endocrine cells that also express chromogranin A. Multiple laboratories are undoubtedly currently investigating whether chromogranin A is also an autoantigen in type 1 diabetes patients.

The diverse benefits of antigen identification

The NOD mouse can be used not only to identify new antigens targeted in the disease, but also to evaluate alternate methods for therapeutic antigen delivery once important antigens are discovered. From recent discussions of therapeutic strategies being explored in NOD mice,^5,76,77 it can be seen that diverse forms of antigen have been considered (Fig. 2). These include proteins,²⁵ peptides or altered peptide ligands,^78,79 antigens linked to cholera toxin B,⁶³ cell-bound antigens (e.g. peptide-pulsed immature DCs⁸⁰ or antigen-coupled fixed splenocytes⁵⁸), multimerized peptide/MHC complexes (e.g. class II MHC dimers,⁸¹ class I MHC-coated nanoparticles,⁸² or toxin-coupled class I MHC tetramers⁸³), and antigen-linked antibody molecules (e.g. peptide-linked anti-DEC-205⁶⁴). While here we have discussed NOD T-cell antigens primarily from the point of view of their value in the development of antigen-specific therapeutic strategies, it should be noted that multiple other benefits derive from antigen discovery. Examples are the ability to image the islet infiltrate in vivo,⁸⁴ monitor beta cell autoimmunity in type 1diabetes patients and islet transplant recipients using T-cell-based assays,^85,86 and gain new insights into disease pathogenesis.⁷³

Figure 2.

T-cell antigen delivery methods employed in non-obese diabetic (NOD) mice. In addition to the administration of free proteins or peptides, a variety of other antigen delivery strategies have been utilized in NOD mice. These are illustrated schematically and are not drawn to scale. See text for referenced examples of each delivery strategy.

Conclusions

Of the 18 known diabetes-relevant T-cell antigens in NOD mice, 12 are also targeted by T cells in patients with type 1 diabetes.⁵ This illustrates the value of the NOD mouse model in the identification of antigens relevant to the human disease. Indeed, as we have discussed here, three NOD T-cell antigens (Hsp60, GAD65 and insulin) continue to be evaluated in a series of ongoing clinical trials. Furthermore, the usefulness of identified antigens for immune monitoring in the context of intervention trials will increase in the future as T-cell assays for autoreactive specificities in patients continue to improve. The recent identification of several novel T-cell antigens in NOD mice indicates that the model continues to represent fertile ground for antigen discovery research.

Disclosures

The authors have no potential conflicts of interest to declare regarding this article.