Genetics, pathogenesis and clinical interventions in type 1 diabetes

Abstract

Type 1 diabetes is an autoimmune disorder afflicting millions of people worldwide. Once diagnosed, patients require lifelong insulin treatment and can experience numerous disease-associated complications. The last decade has seen tremendous advances in elucidating the causes and treatment of the disease based on extensive research both in rodent models of spontaneous diabetes and in humans. Integrating these advances has led to the recognition that the balance between regulatory and effector T cells determines disease risk, timing of disease activation, and disease tempo. Here we describe current progress, the challenges ahead and the new interventions that are being tested to address the unmet need for preventative or curative therapies.

Type 1 diabetes represents one of more than 80 diseases considered to have an autoimmune aetiology. The disease occurs as a consequence of the organ-specific immune destruction of the insulin-producing β-cells in the islets of Langerhans within the pancreas¹. The β-cells are elegant glucose ‘thermostats’, sensing glucose and releasing insulin to maintain physiologic glucose levels within a relatively narrow range. They thus comprise much more than just an insulin factory. Once those cells are destroyed, patients with type 1 diabetes lose blood glucose control, which can result in both acute conditions (for example, ketoacidosis and severe hypoglycaemia)² and secondary complications (including heart disease, blindness and kidney failure)—even with current insulin replacement therapies^3,4. Type 1 diabetes develops as a consequence of a combination of genetic predisposition, largely unknown environmental factors, and stochastic events. For many reasons, postulated to involve population hygiene, sun exposure, and other environmental factors, its incidence has increased dramatically over the last two decades, especially in children less than five years old⁵. Those under the age of 18 are most often afflicted⁶, but an equal number of adults over 18 are thought to develop the disease, although incidence in older people receives less media/research attention. In this review, we discuss our current understanding of the cellular/molecular mechanisms of disease aetiology and progression, the usefulness and limitations of rodent models of spontaneous diabetes, the factors that are influencing the current increased incidence and the clinical opportunities for those affected.

Pathophysiology of type 1 diabetes in mouse and human

Although the clinical picture of type 1 diabetes as a progressive loss of β-cell function over a period of years and the requirement for daily insulin treatment for patient survival has been apparent for over a century, the precise immunologic, genetic and physiologic events that control disease initiation and progression continue to be elucidated. During the last 25 years, two key animal models of type 1 diabetes—the inbred BioBreeding (BB) rat⁷ and non-obese diabetic (NOD) mouse^1,8—have been used to study the genetics, pathophysiology and environmental impact on the spontaneous form of this disease. The rodent models have many aspects in common with the human disease, including a number of similarities in genetic loci of susceptibility, influence of the environment and pathogenesis of disease. The studies in NOD mice have demonstrated that the disease occurs as a consequence of a breakdown in immune regulation, resulting in the expansion of autoreactive CD4⁺ and CD8⁺ T cells^9–11, autoantibody-producing B lymphocytes^12–14, and activation of the innate immune system that collaborate to destroy the insulin-producing β-cells^15,16. These attributes of the disease are consistent with studies of human type 1 diabetes. We note that of 26 loci identified through the genome-wide association study (GWAS¹⁷) of human type 1 diabetes, at least 6 loci are shared between the NOD mouse model and humans at risk for type 1 diabetes, and 19 are associated with immune regulation^17,18.

Although the presence of islet tissue-specific autoantibodies in sera from patients with type 1 diabetes was the first diagnostic of autoimmunity (Fig. 1a), there is overwhelming evidence in both the NOD mouse and human disease that autoreactive T cells play a dominant role in disease initiation and progression. CD11c⁺ dendritic cells and ER-MP23⁺ macrophages are the first cells to infiltrate the pancreas of NOD mice at approximately three weeks of age. At the same time, or shortly thereafter, potentially pathogenic T cells can be detected surrounding the islets (this is termed peri-insulitis) (Fig. 1b, c)¹. These T cells are presumably activated in the pancreatic draining lymph nodes as a result of high turnover of β-cells in the islets leading to antigen presentation¹⁹, although the molecular events that initiate the loss of tolerance in this setting are still speculative. Further islet damage leads to the release of self-antigens, leading to epitope spreading (that is, presentation of new autoantigens to the inflamed immune system, leading to newly activated T cells), and amplification by complex islet mononuclear cell infiltrates present at the time of disease onset. Both major histocompatibility complex (MHC) classes I and II restricted islet-antigen-reactive T cells have been identified in NOD mice and in the peripheral blood of type 1 diabetes patients. In many instances, these T cells have been shown to recognize islet autoantigens similar to those seen by autoantibodies (such as insulin, glutamic acid decarboxylase (GAD) and zinc transporter 8 (ZnT8)). The T cells also recognise other islet antigens, such as islet-specific glucose-6-phosphatase catalytic subunit-related protein (IGRP) and chromogranin A, in NOD mice and humans that have particular susceptibility alleles^10,20,21. In fact, autoreactive T cells are observed in very young NOD mice and are often found in the blood of susceptible individuals before disease onset. A recent study used an elegant retrogenic mouse T-cell receptor system to demonstrate directly that islet retention of T cells is antigen-specific and cell-intrinsic²². These studies suggest that a limited number of primary islet autoantigens recognized by early-infiltrating T cells may be responsible for disease initiation (Fig. 2). These findings raise the critical question of whether there is a single self-protein responsible for initiating type 1 diabetes.

Figure 1. Markers of diabetes.

a, Typical child followed from birth until development of diabetes in the DAISY (Diabetes Auto Immunity Study in the Young; http://www.uchsc.edu/misc/diabetes/Teddy/DAISY/DAISY_home.htm) study (M. Rewers, unpublished work), expressing multiple autoantibodies (GAD, the islet cell antibody ICA512 and low-level insulin autoantibodies). b, Islet invasion by lymphocytes of NOD mice is asynchronous during progression to diabetes, often with a mixture of normal islets, peri-insulitis, intra-islet insulitis, and complete β-cell destruction. c, Pathology of the pancreas in a long-term type 1 diabetic (the nPOD program, pancreas 6028, see http://www.jdrfnpod.org/). The section shows lobular areas in which all β-cells (insulin-producing) in all islets have been destroyed (pseudoatrophic islets in which only glucagon, somatostatin, and pancreatic polypeptide cells are present) juxtaposed with regions in which all islets contain insulin-containing β-cells.

Figure 2. Immunologic history of type 1 diabetes.

An as-yet-undefined immunologic insult occurs in an individual with genetic predisposition and initiates a chronic low-grade immunologic process (priming). The initiating events involve infiltration of innate immune cells (such as monocytes and natural killer cells with autoreactive B cells) (orange ovals) into the pancreatic islets. The principal site of antigen presentation is thought to be the pancreatic lymph node where islet antigens are presented by antigen-presenting cells (white ovals) to T cells (brown dots). Blue ovals are antigen-presenting cells loaded with islet antigens. B cells (green dots) and dendritic cells may be among the early antigen-presenting cells. The cellular infiltration of islets ensues but the insulitis is uneven. Islets with infiltration may be situated near to islets without cells. The process specifically targets insulin-producing β-cells (light blue circles), while other endocrine cells (red circles) within the islet are spared. In the lymph nodes, the cycle of antigen presentation, activation of adaptive immune cells, licensing of effector T cells and epitope spreading continues with the loss of β-cells over time. There is evidence for a regenerative attempt of β-cells in the midst of the islet inflammation (dark blue circles). Tertiary lymphoid organs are thought to develop within the islets, which may lead to amplification of the adaptive immune response. Regulatory T cells (yellow dots) may arrest this process in its early and late stages but are not able to contain the amplified process in the late stages despite an increase in their numbers. With continued loss of β-cells, hyperglycaemia can be detected. The loss of metabolic function at presentation may be both functional and anatomic, because immune therapies can restore cells that have lost the capacity to produce insulin but have not been destroyed. Without intervention, however, β-cell loss continues.

In this regard, there is strong evidence that defects in thymic T-cell negative selection related to insulin reactivity itself is key to the genetic predisposition towards the disease^23,24. First, approximately 20% of individuals with spontaneous mutations of the autoimmune regulator gene AIRE develop type 1 diabetes with other autoimmune diseases, which is thought to reflect their inability to select against islet antigen reactivity during T-cell development²⁵. Insulin is an AIRE-regulated islet protein ectopically expressed in the thymic medullary epithelial cells. In addition, polymorphisms in the insulin promoter that map to a diabetes susceptibility allele control expression of insulin in the thymus, potentially regulating the autoimmune repertoire. Second, a high percentage of pathogenic T cells isolated from early islet infiltrates recognise insulin in the NOD mouse model. In fact, it appears that there is a unique relationship between the expression of certain T-cell receptor Vα regions and the recognition of an insulin B-chain peptide (B9–23)²⁶. However, preclinical evidence supports the notion of a sequential loss of tolerance to multiple epitopes, suggesting that multiple self-antigen specificities are probably involved in disease progression. For example, tolerance to proinsulin prevents induction of IGRP-reactive T cells in NOD mice, whereas diabetes is seen in mice tolerant to IGRP in spite of the absence of IGRP-reactive cells.

Multiple genes within the MHC have been recognized for more than two decades as the dominant loci associated with disease in both the NOD model and human disease. The MHC class I and class II molecules have been suggested to account for both positive and negative selection of autoreactive T cells by virtue of creating the binding groove for antigens presented to T lymphocytes, the final effectors of disease. For instance, the NOD MHC I-A^g7 allele is essential for disease development while alternative MHC alleles (such as MHC I-E^k) act in a dominant fashion to protect NOD mice from disease occurrence. The human MHC is called the human leukocyte antigen (HLA) region. In humans, the presence of susceptibility loci within the HLA complex (the DRB1 0401, DRB1 0402, DRB1 0405, DQA1 0301, DQB1 0302 or DQB1 0201 alleles), which are common in North Americans of European ancestry and Europeans, result in a lod score (the logarithm of the odds of linkage) of 116 (which is 50-fold higher than the next susceptibility locus identified in GWAS studies)^17,27. However, strong protection from certain HLA alleles (DRB1 1501, 1401 or 0701 and DQB1 0602, 0503 or 0303 alleles) is dominant, suggesting that the HLA may have a more significant role of protection than predisposition. The key role of HLA genetics in disease susceptibility has proved useful in screening for prevention trials because certain alleles are used to identify patients at high risk of disease, and exclusion of individuals with resistance alleles can reduce the sample size needed for clinical prevention studies. Finally, although T cells reactive with MHC class II may be the driver for disease initiation, CD8⁺ T cells are likely also to be involved in disease pathogenesis because CD8⁺ T cells reactive with MHC class I-restricted antigens can be found in the blood and islets of mice and humans with type 1 diabetes.

The mediators of islet destruction have not been precisely defined. In studying cadaver pancreases from humans with type 1 diabetes, it has been rare to observe the degree of insulitis seen in the animal models or evidence of pathogenic T cells²⁸ (see http://www.jdrfnpod.org/). Notably, studies of the pancreas from both new-onset and long-term patients with residual β-cells indicate destruction of β-cells in lobules of the pancreas (likely to underlie slow progression) despite the presence of autoantibodies, typically for years before hyperglycaemia occurs²⁹. Circulating T cells that can be studied in the peripheral blood of patients may be an indirect reflection of the events that are occurring in the pancreas. The lack of easy accessibility to pancreatic tissue from patients except in unusual circumstances^28,30 has made analyses of human disease difficult. However, a recent study of cadaver pancreases isolated from patients with new-onset diabetes showed that CD8⁺ cells and CD68⁺ appeared to dominate the infiltrates³¹. Moreover, the creation of a nation-wide network, nPOD (http://www.jdrfnpod.org/), has allowed increased access to potentially informative samples (Fig. 1c).

Yet so far, most of the information we have on disease pathogenesis has been from the NOD mouse studies. Although T helper 1 cells, producing interferon gamma (IFNγ) can transfer disease, especially in CD4⁺ T-cell-receptor transgenic models, the role of IFNγ in the spontaneous disease remains controversial because IFNγ-deficient NOD mice develop the disease similarly to wild-type NOD mice. Interestingly, disruption of Tbx21, the gene encoding the transcription factor Tbet that controls IFNγ production, ameliorates disease; however, this may be due to the role Tbet plays in T helper 1 cell trafficking and dendritic cell function³². Similar discrepancies have also been observed in studies disrupting the gene encoding the cytokine interleukin (IL)-12, which is key to the T helper 1 cell lineage. In contrast, there is good evidence that a variety of cytotoxicity-inducing molecules (for example, the tumour necrosis factor TNFα, perforin and granzyme B) are critical in disease pathogenesis mediated by both the CD4⁺ and CD8⁺ T cells¹. The identification of IL-17-producing T helper 17 cells raised the possibility that this population (and the associated IL-23 cytokine) may be involved in the disease. However, recent studies have largely discounted this possibility³³.

Other cell types are also involved in disease pathogenesis in NOD mice and are present in human insulitis. This includes natural killer cells, monocytes and potentially other cells of the innate immune system^34,35 (modelled in Fig. 2). In the case of B cells, their precise role is less clear. Although they may have a role in autoantibody production, the antigen-specific B cells are very efficient in presenting antigens and produce cytokines that can either promote or suppress immunity^36,37. Additionally, the innate immune cells also contribute to the inflammatory milieu, promoting islet cell destruction and immune activation. It is becoming increasingly clear that type 1 diabetes in NOD mice, and possibly in humans, is amplified by immune components of tissue inflammation, analogous to that described for type 2 diabetes, Alzheimer’s, atherosclerosis and other diseases^38,39.

Although there are a number of GWAS-identified susceptibility alleles that may contribute to T-cell-mediated islet cell destruction, these immune-response genes have not helped us to identify monogenic pathways that directly lead to disease. Rather, it appears that a global problem of immune regulation may underlie disease susceptibility. For example, mutations of genes encoded in several of the susceptibility loci—including Il2, Il2ra (CD25), Ctla4, PTPN22, and Pdca1 (PD-1)—in multiple animal models lead to the development of a diverse array of autoimmune diseases, including type 1 diabetes. The proteins encoded by these genes have been implicated in maintaining immune homeostasis either by directly tuning the signal strength of T and B cell receptor complex or indirectly through the control of regulatory cell populations critical to controlling autoimmunity.

Most prominent is the potential role of these pathways in the control of Foxp3⁺ regulatory T (T_reg) cells, which are essential suppressors of unwanted autoimmune responses^40,41. T_reg cells, whose T-cell-receptor repertoire is skewed towards self-protein recognition, are thought to modulate T-cell activation and promote tolerance by suppressing adaptive immunity through direct cell–cell interactions and production of immune modulatory cytokines such as the transforming growth factor TGF-β and IL-10 (ref. 42). However, other proposed T_reg cell functions target more generalized mechanisms of inflammation, such as the redox reaction, ATP utilization, tryptophan metabolism and the nitric oxide pathways. These results suggest that T_reg cells may play a more generalized role in β-cell survival and function as ‘innate’ regulators of immune-mediated tissue damage. T_reg cells exist in fat and control obesity-related inflammatory responses that alter insulin resistance^43,44. In fact, T_reg cells may affect tissue repair, linking elements of type 1 diabetes and β-cell stress⁴⁵. Mutations of Foxp3 result in the immune polyendocrinopathy enteropathy X-linked (IPEX) syndrome, which includes autoimmune diabetes within the first years of life. However, polymorphisms in Foxp3 alleles have not been found among the susceptibility loci for type 1 diabetes, so the manifestations of T_reg cell alterations controlled by the susceptibility alleles must be subtle.

Nonetheless, the critical importance of T_reg cells in the autoimmune setting has been documented. Several studies have shown that T_reg cell numbers and functions are altered during disease activity and the subset has been shown to be unstable in some settings, such as the inflamed pancreas of autoimmune mice⁴⁶. At the time of disease onset in NOD mice, the number and function of the T_reg cells, especially the stable CD25⁺ subset in the pancreas, is significantly reduced in the inflamed islet tissue, most probably owing to defects in IL-2 production⁴⁷. In fact, a subset of the T_reg cells loses its Foxp3 expression, turns on IFNγ, and when adoptively transferred can become pathogenic in its own right⁴⁶. A similar observation has been made in humans with type 1 diabetes, demonstrating a significant increase in the number of IFNγ-producing Foxp3⁺ cells in the circulation of new-onset type 1 diabetes patients, concomitant with slightly reduced Foxp3 expression in the circulating T_reg cell subset (S. A. McClymont, A. L. Putnam, M. R. Lee, J. H. Esensten, W. Liu, U. Baron, S. Olek, J.A.B. and T. M. Brusko, unpublished work). Coupling these observations with the findings in both the NOD mouse and humans with type 1 diabetes that T effector cells change over time to become resistant to T_reg-cell-mediated suppression^48,49, we are left with a model that suggests that ultimate disease progression is a direct consequence of the imbalance of T_reg cell to effector T cells⁴² (Fig. 3).

Figure 3. Immune system balance is key to disease pathogenesis.

This schematic illustrates the fine balance of immune regulation versus pathogenesis, highlighting a number of genes that are likely to influence the balance through effects on central and peripheral tolerance and the environmental factors that control immunity. The key cell types that affect the balance locally during immune responses are listed (with the regulatory cytokines and proteins given in parentheses). iNOS, inducible nitric oxide synthase. IDO, indoleamine-2,3 dioxygenase.

Finally, the environment is likely to have a strong effect on the development and progression of type 1 diabetes. In the case of NOD mice, the incidence of diabetes dramatically decreases when mice are exposed to microbial stimuli, by injection with mycobacteria, or through contact with various microbial products^50,51. Similar observations have been noted in humans, with the development of the so-called “hygiene hypothesis”, which postulates that the increase in type 1 diabetes incidence is most notable in industrialized societies with reduced exposure to parasites⁵². In this regard, recent studies have shown that the type of “commensal” bacteria can protect NOD mice from developing type 1 diabetes⁵¹. Specifically, disruption of the Myd88 gene, encoding an adaptor for multiple innate immune receptors that recognize microbial stimuli, protects NOD mice from developing type 1 diabetes by altering the composition of gut microbiota⁵¹. Other environmental factors may also influence disease incidence, ranging from cows’ milk/bovine serum albumin, meat preservatives/N-nitroso compounds, and so on⁵². There is evidence of a role for vitamin D and its analogues, omega-3 fatty acids, and environmental stress and toxins^53–56. Finally, ongoing epidemiological and laboratory research efforts support a potential role for viruses in the disease pathogenesis, either owing to antigenic mimicry or to the possibility that certain viral infections break self-tolerance by activating innate immunity, potentially in concert with risk alleles influencing innate immune function^57,58. Each of Coxsackie B4, the seasonal incidence of rhinovirus and influenza has been implicated⁵⁶. It has been difficult to identify the most significant environmental factors involved because of the multitude of suggested causes. The TEDDY (The Environmental Determinants of Diabetes in the Young; http:teddy.epi.usf.edu/) study is prospectively analysing environmental factors that may be responsible for modifying the incidence of this disease.

Opportunities for preventing and treating type 1 diabetes

As highlighted above, our increased understanding of the pathogenesis of the disease and the identification of genes and environmental factors that control disease incidence have provided a wealth of potential targets for disease intervention (Fig. 4).

Figure 4. Targets of immune intervention in type 1 diabetes.

This schematic provides an overview of the pathogenesis of type 1 diabetes, highlighting a number of key pathways that are being targeted by current therapeutics. Although not exhaustive (see Supplementary Table 1), this figure shows that both non-specific and antigen-specific therapies are being tested, which inhibit effector cells and antigen presentation as well as boost regulatory pathways. Purple and green dotted arrows indicate the therapeutics, black arrows are immune and metabolic pathways; a green dotted arrow indicates a positive effect and a purple dotted arrow indicates a negative effect. In addition, these immunotherapies are being combined with drugs that promote β-cell survival to potentially replenish insulin-producing β-cells. The figure has been redrawn after ref. 87, with permission.

Clinical trials for new-onset disease are feasible and have been conducted over the past 20 years but have the following limitations. First, given the reduced level of β-cell numbers at the time of diagnosis, the goal of most clinical trials in type 1 diabetes is to improve functional residual β-cell mass, optimally through induction of immunologic tolerance, while preserving protective immune responses. By definition, this will rarely “cure” the disease because of the significant β-cell destruction that preceded the treatment. Second, since there are no biomarkers of the disease process that are reliably correlated with the pathogenic process, endpoints for studies have been limited to measures of β-cell function and clinical parameters after one or two years, which by definition does not “read-out” the immune pathogenesis but rather the metabolic consequences of the disease. Third, timing may be important. Metabolic and immunologic data from the Diabetes Prevention Trial-1 (DPT-1; (http://www.diabetestrialnet.org/publications/publications.htm#dpt1) and the European Nicotinamide Diabetes Intervention Trial (ENDIT; http://www.bristol.ac.uk/clinicalsciencenorth/diabetes/ri/endit.html), and studies after clinical presentation have been consistent with the model of a chronic autoimmune process that relentlessly leads to β-cell destruction. Human data are consistent with this notion, because the risk of diabetes exceeds 90% in first-degree relatives of patients who are positive for at least two biochemical autoantibodies, whereas it is less than 20% in relatives who are positive for only one autoantibody. On the basis of this model, therefore, antigen-specific tolerance would be predicted to have greatest success in the earliest stages of disease, whereas a broader approach would be needed at the time of diagnosis when there is a polyclonal autoreactive repertoire.

Indeed, prevention trials, which enroll individuals at risk, identified after the development of autoantibodies have been uniformly unsuccessful. Prevention trials involving relatives of people with type 1 diabetes require the screening of a large number of individuals in order to identify a study group because the frequency of autoantibody positivity within this population is only approximately 3.5%. In the DPT-1 trial, parenteral insulin was administered to individuals with positive autoantibody levels and impaired insulin responses to intravenous glucose; this approach was based on studies in the NOD mouse and pilot human studies that showed this treatment to prevent disease progression (http://www.diabetestrialnet.org/index.htm)⁸⁸. However, the trial failed to show a significant effect of insulin administration on the progression of disease. The ENDIT trial, which proposed to arrest early stages of β-cell injury, enrolled subjects with positive autoantibodies and administered nicotinamide, but likewise showed no beneficial effects. The negative outcomes of these large trials have led to speculation regarding differences between the animal model (the NOD mouse) that is used for preclinical studies and human disease. However, a number of therapies described throughout this review have been translated to humans. So the lack of success in these trials may be due to the use of the wrong dose of the interventional agent or the timing of the intervention, given that the trials involved individuals in whom autoimmunity and β-cell destruction was already initiated.

The most extensive clinical interventions have been performed at the time of disease onset, when nearly all individuals have ‘clinically significant’ levels of C-peptide production (that is, at least 0.2 pmol ml⁻¹ following metabolic stimulation). Some 15 years after completion of the Diabetes Control and Complications Trial (DCCT; http://diabetes.niddk.nih.gov/dm/pubs/control/) there is still compelling evidence that improving short-term control of blood sugars, which can be achieved by preserving endogenous β-cell function, can lead to long-term reductions in diabetes complications, through a process referred to as “metabolic memory”⁴.

The first successful immunosuppressive agent used in a placebo-controlled, double-blind clinical trial for type 1 diabetes was cyclosporine A^59,60. Unfortunately, although β-cell function appeared to be preserved (on the basis of reduced insulin usage), its excessive nephrotoxicity forced termination of therapy in those studies after one year. Moreover, the protective effect was not extended beyond treatment cessation⁶¹. In the mid-1980s, a commercial anti-thymocyte globulin in conjunction with prednisone was shown to reduce insulin requirements in a small number of new-onset patients at a time when many were still questioning the autoimmune aetiology of type 1 diabetes⁶². Finally, based on the hygiene hypothesis and related studies described above, there have been attempts to use a variety of adjuvants, including the bacillus Calmette-Guerin, without success⁶³.

In 2002, there was a theoretical paradigm shift when Herold et al.⁶⁴ reported that treatment of newly diagnosed patients with type 1 diabetes with the anti-T-cell humanized anti-CD3 monoclonal antibody, teplizimab (mutated to prevent binding to Fc receptors) led to a sustained preservation of C-peptide (insulin) production. A second European report showed similar efficacy using a second anti-CD3 monoclonal antibody otelixizumab (similarly Fc-mutated)⁶⁵. In an example of mouse models of type 1 diabetes equating with the human disease, a short (that is, two weeks or less) course of an anti-CD3 monoclonal antibody drug (145-2C11) revealed that diabetes was permanently reversed in NOD mice, whereas in humans, C-peptide levels were sustained for at least one to two years in most of the patients and in some cases five years or more, suggesting that short-term therapy can have a long-term effect (of at least two years)^66,89. Further studies in mice and patients have suggested that by virtue of its function as a partial T-cell-receptor agonist, T_reg cells are selectively preserved by treatment with the anti-CD3 monoclonal antibody. In mice, these cells are localized to the pancreatic lymph nodes and require TGF-β, whereas in humans, both CD4⁺ and CD8⁺ T_reg cells have been described^67,68. The probable mechanism of anti-CD3 monoclonal antibody therefore involves the preservation of T_reg cells that may regulate antigen-specific responses involved in the disease. The duration of this effect is still unknown, as is whether the T_reg preservation can be boosted by repeated drug administration. These questions are being addressed in clinical trials. In addition, an effect of anti-CD3 monoclonal antibody on the ability of residual effector T cells to be regulated has not been studied.

Initial preclinical studies that showed transfer of diabetes by purified T cells, independently of B lymphocytes, cast doubt on a role of B cells late in the disease course. It was shown, however, that membrane-bound immunoglobulin could partially restore disease to mice deficient in secreted immunoglobulin, suggesting that although there was not an essential role of soluble immunoglobulins (that is, autoantibodies), B cells might be important for antigen presentation or other functions⁹⁰. Hu et al. used mice expressing the human CD20 transgene to demonstrate that depletion of cells could prevent, and even reverse, diabetes when the cells were depleted at diagnosis⁶⁹. Preclinical studies suggested that B cells with regulatory function were induced with anti-CD20 monoclonal antibody treatment^69,70. More recently, in a blinded placebo-controlled clinical trial, the anti-human CD20 antibody, rituximab, given in four weekly doses was shown to improve provoked C-peptide responses three months after diagnosis⁷¹. Subtle differences in B cell subsets were identified in the clinical trial and correlated with clinical responses. The long-term effects of rituximab treatment have not been reported, but a number of observations are of interest. The greatest difference in responses between drug- and placebo-treated subjects occurred three months after entry into the study (which occurred within three months of diagnosis), and the decline in C-peptide production by six months after the treatment began and thereafter was similar in drug and placebo groups. In addition, the duration of immune effects could be problematic, because immunoglobulin M levels remained depressed long-term. The application of this therapy to clinical practice, therefore, requires careful weighing and further study because, in general, long-term immune suppression is not acceptable as a therapeutic option.

The safety concerns and adverse effects of antigen non-specific interventions, as well as the lack of permanent remission of disease with any agent tested to date have heightened interest in antigen-specific interventions that might modulate the disease. Even a partial response, requiring repeated administration of a drug, might be preferable to a broad immune-suppressive agent in this patient population. The polyclonal nature of the autoimmune response, reflected by the presence of multiple autoantibodies and multiple T-cell epitopes that are recognized by peripheral blood cells, might suggest that this goal cannot be achieved, but immunologic tolerance mechanisms, even by antigen-specific cells, are not restricted to a single antigen. The notion of “bystander suppression” mediated by soluble products or cell–cell contact explains how T cells with regulatory properties could modulate the function of T cells specific for other antigens^72–74. Cytokines such as IL-10, TGF-β and even IL-4, produced by T cells in response to antigen or antigen-presenting cells, can modulate the function of effector T cells in the vicinity of the antigen. In this way, activated T_reg cells exert their function by entering the site of the incipient immune response and need not directly recognize the antigen(s) recognized by effector T cells. By modulating the differentiation of other antigen-specific T cells, pathogenicity could be modified permanently in cells otherwise destined to become effector T cells (“infectious tolerance”)^75,76. In addition, CD4⁺CD25⁺FoxP3⁺ and other regulatory T cells are able to modulate the function of effector T cells through contact-dependent mechanisms. We have shown that polyclonal diabetogenic T cells may be inhibited by antigen-specific T_reg cells in vitro and in an adoptive transfer model of diabetes⁷⁷.

Hence, two trials, the oral insulin arm of the DPT-1 and the GAD65 immunization trial in patients with new-onset disease (http://clinicaltrials.gov/ct2/show/NCT00435981?term=GAD65&rank=4), support the concept that an antigen-specific intervention may have broad immunologic effects, even late in the immune progression, provided the antigen selected is directly involved in the disease pathogenesis and/or regulation. A second arm of the DPT-1 study enrolled patients who had positive insulin autoantibodies but did not show the same impairment in insulin responses as those enrolled in the parenteral arm, and administered oral insulin based on the hypothesis that an oral antigen would induce a tolerogenic (producing immunological tolerance) response. The oral insulin administration yielded intriguing results in a subset of patients with high levels of insulin autoantibodies, a finding now being studied in detail by Trialnet⁷⁸. Immunization with alum GAD65 was also shown to attenuate the loss of C-peptide in individuals treated within six months of diagnosis⁷⁹. These studies are a prelude to a number of additional antigen-specific therapies that are in late pre-clinical or clinical development. This includes a number of additional insulin-specific therapies that target tolerogenic pathways, other potential regulatory autoantigens and combination therapies that include autoantigens as part of the tolerogenic cocktail (see Fig. 4 and Supplementary Tables 1 and 2). By more directly targeting the pathogenic response with regulatory cells, combined with adoptive immune therapy, we might find an approach, which, if found to be safe, could be repeated on a regular basis without the need to treat patients with broadly active immunosuppressive drugs and put them at risk of treatment-associated adverse events. This strategy is now being tested in type 1 diabetes using CD4⁺CD25⁺Foxp3⁺ T_reg cells as well as T regulatory 1 cells IL-10+) adaptive cells⁸⁰.

Lastly, it has been established that persistent dysregulated inflammation contributes to most chronic diseases, including vascular diseases, cancer, neurodegenerative diseases, metabolic diseases such as type 2 diabetes and autoimmune diseases such as type 1 diabetes. Importantly, it is this dysregulated inflammation that is critical in breaking immune tolerance by disabling key regulatory pathways, such as T_reg cells and tolerogenic dendritic cells. Thus, it is not surprising that drugs that target the inflammatory responses are actively being tested in the clinic on the basis of strong preclinical data showing an ability to moderate the progression of type 1 diabetes⁸¹. Current trials using anti-cytokine drugs such as IL-1 and IL-15 antagonists, as well as anti-chemokine antagonists, are aimed at reducing inflammation in an attempt to allow re-establishment of immune homeostasis. In fact, these anti-inflammatory drugs have been used preclinically, in combination with other pro-tolerogenic therapies. In this regard, several Food and Drug Administration (FDA)-approved small molecules and biologics have shown very promising results in reversing and inducing long-lasting remission of diabetes in the NOD mouse. Aralast NP (an a1 anti-trypsin used to treat the genetic deficiency)⁸² and imatinib (Gleevec)⁸³, a cancer drug that inhibits a variety of receptor tyrosine kinases such as c-abl, PDGF and VEGF, have been shown to target the inflammatory macrophages and other innate immune cells that are critical for disease pathogenesis. Both of these drugs are about to enter clinical trials to examine their effect on prolonging residual β-cells in recent-onset diabetics.

Next steps

There are still many unanswered questions that need to be addressed before we are likely to develop an effective and acceptably safe cure for type 1 diabetes. First, to target the T cells involved in the development and progression of the disease, it is essential that we determine the precise specificity of the primary pathogenic T cells. This prerequisite is so that the disease-causing cells can be monitored and the various interventions and complementary strategies can be implemented to eliminate pathogenic cells while enhancing immune regulation. In this regard, recent efforts to access tissues from cadavers of patients with type 1 diabetes may provide a source of antigen-specific T cells at the site of disease activity. Second, linking genotypes with phenotypes and the impact of environment factors will be valuable in determining the best diagnostics and treatments. At present, we have only scratched the surface of understanding the biological pathways of the genes that have been shown to modulate disease activity. Finally, there is a need to broaden our study of disease pathogenesis beyond the T cells. In particular, a better understanding of the role of inflammation and the innate immune response will undoubtedly be important in any comprehensive approach to disease treatment. The recent links between inflammation and multiple diseases ranging from autoimmunity to heart disease, Alzheimer’s and type 2 diabetes suggest that core immune system parameters, including natural killer cells, macrophages and even T_reg cells, may act to control tissue damage as a means of modulating the autoimmune phenotype driven by the antigen-specific T cells and perhaps B cells.

It is likely, that, like most autoimmune diseases, combination therapies that take advantage of targeting different arms of the immune system will be most effective at ameliorating disease. A provocative report⁸⁴ showed that CD34⁺ stem cell transplantation following myeloablation with cyclophosphamide and conditioning with anti-thymocyte globulin and cyclophosphamide resulted in non-insulin-requiring remissions in 14 of 15 subjects for an average of 16 months—a response that has not been achieved with any single immune modulator. Although the approach is controversial owing to the potentially highly toxic nature of the therapy, its outcome suggests that an aggressive approach may be able to reverse the disease and that certain elements of the protocol (for example, anti-thymocyte globulin and granulocyte/macrophage-colony stimulating factor (GM-CSF) conditioning) may be useful alone or in combination with other agents. More acceptable alternatives may be to combine more generalized immunosuppressives (for example, anti-CD3, anti-B-lymphocytes, and anti-inflammatory drugs) with antigen-specific DNA vaccines and antigenic peptides administered via the mucosa⁸⁵. A suggested mechanism of this combination involves induction of antigen-specific T_reg cells when the antigen is administered under the umbrella of the immune modulator. Moreover, efforts to combine the immunomodulatory therapies with islet-preserving drugs, such as glucagon-like peptide-1 (GLP-1) agonists or islet growth factors, may further enhance therapeutic efficacy⁸⁶.

Conclusion

The last decade has seen extraordinary progress in our understanding and treatment of type 1 diabetes. The pathophysiology and the genetics are increasingly clear. There are over a dozen novel therapies being tested in individuals with this disease at all stages, including pre-clinically. A better understanding of the mechanisms of disease, combined with a more complete set of results from the clinical trials currently under way, may help to isolate the active components in the disease and eliminate the need for broad immunosuppressive treatment. We are confident that the next decade will bring new insights into the linkages between genotype and phenotype, new sources of islets that can be used to replace the destroyed β-cells, and combination therapies that will selectively target antigen-driven pathways to re-reestablish tolerance in this and other autoimmune settings.