New therapeutic approaches to type 1 diabetes: from prevention to cellular or gene therapies.

Abstract

The burden of the disease, the inadequacy of treatment in the prevention of chronic complications and the risk of severe hypoglycaemia in type 1 diabetes justify research into preventive strategies and directed towards cellular alternatives to exogenous insulin administration. Although studies such as those of The Diabetes Control and Complications Trial Research (DCCT) Group (1993) and the UK Prospective Diabetes Study (UKPDS) Group (1998) demonstrate the relationships between close blood glucose control and reduced risk of chronic complications, it is clear that intensive insulin therapy is not applicable to all diabetic patients and that physiological control of blood glucose excursions is difficult to achieve. The quest for prevention of diabetes has been made feasible by the unravelling of the immunogenetics of the disease and the identification of at-risk subjects by immunological markers. The existence of a long pre-diabetic period offers a chance for intervention. Control of anti-beta cell immunity also seems to be critical for long-term success of both islet and pancreatic transplantation. Gene therapy might be applicable for key factors that regulate the immune response or for ectopic insulin production. In this review, we report on the different strategies to prevent or palliate the beta-cell loss. Type 1 diabetes is thought to result from an interaction between environmental and genetic factors that trigger selective beta-cell destruction through a T-cell-mediated phenomenon (Baisch & et al., 1990; Castano & Eisenbarth, 1990; Chao et al., 1999; Davies et al., 1999). The hallmark of the autoimmune process is a chronic inflammatory mononuclear cell infiltration of the islets, termed insulitis (Gepts, 1965; Foulis et al., 1991). This destructive process is associated with an excess of pro-inflammatory T helper 1 (Th1) cytokines interferon γ (IFN-γ), tumour necrosis factor β (TNF-β), interleukin 2 (IL-2) and lower expression of Th2 cytokines (IL-4 and IL-10) (Rabinovitch, 1998) as shown in Fig. 1. At time of clinical diagnosis, it is generally accepted that more than 80% of the beta cells have been destroyed. However, comprehensive models of beta-cell loss are lacking to support this hypothesis. The persistence of a significant number of quiescent beta cells might explain the capacity of intensively treated patients to develop clinical remissions (Shah et al., 1989). The availability of experimental models of autoimmune diabetes such as the non-obese diabetic (NOD) mouse and the bio-breeding (BB) rat has greatly enhanced our understanding (Mordes & Rossini, 1981). Circulating autoantibodies against multiple beta-cell autoantigens are present in the serum long before a person develops diabetes and herald the pre-diabetic phase in humans. Risk of diabetes and time to onset of diabetes in relatives correlate directly with the number of autoantibodies to beta-cell constituents such as insulin, glutamic acid decarboxylase 65 (GAD65) and two related transmembrane proteins of the tyrosine phosphatase family, islet-associated antigen 2 (IA-2) and a related protein IA-2β (Verge et al., 1998; Bingley et al., 1999). Once islet cell autoimmunity has begun, progression to islet cell destruction is variable with some patients progressing rapidly to diabetes while others remain in a non-progressive state. Antigen or epitope spreading of the antibody response is an important feature of ongoing autoimmunity in pre-diabetes (Panicot et al., 1999; Ziegler et al., 1999) or during autoimmune recurrence in patients with pancreatic transplants (Thivolet et al., 2000). Primary antibody screening followed by quantification of risk by further antibody, genetic and metabolic testing applies to first-degree relatives of diabetic patients. Because of the low percentage of multiplex families (less than 5%) it is reasonable to apply a different strategy in countries with the highest incidence of the disease. Relatives enrolled in the BABYDIAB (German multicentre study investigating offspring of diabetic parents) (Ziegler et al., 1999), DAISY (diabetes autoimmunity study of the young) (Rewers et al., 1996) and DIPP (Finnish type 1 diabetes prediction and prevention project) (Nejentsev et al., 1999) studies are screened for the presence of high-risk HLA genes and followed prospectively for the appearance of autoantibodies. In these studies, the presence of circulating antibodies in genetically at-risk subjects are already present in the circulation after only a few months of life, suggesting that beta cell destruction might occur in very early childhood. Interactions between the immune system and the beta cells leading to type 1 diabetes. APC, antigen-presenting cell; TCR, T-cell receptor. The availability of animal models allowed the dissection of many immune events and helped the design of intervention strategies (Fig. 1). The present difficulties are to identify from the hundreds of treatments that are effective in the NOD mouse those that might be the best candidates for human interventional therapy (Atkinson & Leiter, 1999). NOD mice share many features with human type 1 diabetics including insulitis, a long pre-diabetic phase and the importance of class II MHC molecules in the genetic predisposition to the disease. For mice, however, several important differences are present, including the lack of genetic diversity and the controlled conditions of housing, which might influence the capacity to respond to the manipulation of the immune system. Many strategies have been evaluated in mice for short periods of time and particular attention in human studies should be given to long-term side-effects of such therapies and non-specific effects on the immune system. Pharmacological immunosuppression, anti-inflammatory cytokines and antibodies against T-cell surface molecules can inhibit pathological immune responses in animal models and also interfere broadly with immune system function. Low doses of CD3 mouse antibody (mAb) were able to normalize blood glucose levels in overtly diabetic NOD mice but did not protect 4- and 8-week-old mice from diabetes (Chatenoud et al., 1997). This treatment was associated with the persistence of insulitis and did not influence the capacity of spleen cells to transfer diabetes. A non-depleting CD4+ Ab showed protective effects in NOD mice, with a decrease in the production of pro-inflammatory cytokines and the disappearance of infiltrating cells from the pancreas (Phillips et al., 2000). Selective targeting of T cells bearing high-affinity IL-2 receptor (IL-2R) is also an attractive therapy since IL-2R is present on recently activated but not on resting or memory T cells. In NOD mice, a cytolytic chimeric IL-2/Fc fusion protein has been shown to possess anti-diabetogenic effects (Zheng et al., 1999). Transfer to humans is being applied for other autoimmune diseases but this requires humanized antibodies. Targeting autoreactive T-cell populations more selectively is preferable, especially in pre-diabetic subjects. This has been shown to be effective in NOD mice by manipulation of conformational disease-associated, MHC-derived peptides (Dunsavage et al., 1999). This strategy is, however, difficult to apply to the genetically diverse human population whose autoreactive T-cell repertoire might change with disease progression. Although several dominant peptides have been characterized, their respective involvement in the course of the disease remains to be determined. Parenteral insulin administration has been shown to prevent beta-cell destruction in BB rats (Gottlieb et al., 1991) and NOD mice (Atkinson et al., 1990; Thivolet et al., 1991). Several hypotheses have been proposed to support these protective effects including modulation of antigen expression and beta cell rest. However, this hypothesis can be excluded in view of the capacity of an inactive analogue of insulin to provide protection (Karounos et al., 1997) and induction of peripheral tolerance is suspected. Antigen-based therapies consist of oral and nasal administration of candidate antigens to induce antigen-specific suppressor cells that could interfere with autoimmune disease progression. This has been tried successfully using insulin in the NOD mouse model (Zhang et al., 1991), but it failed in BB rats (Mordes et al., 1996). This form of peripheral tolerance implies peripheral Th2 type T cells that belong to the CD4 phenotype (Bergerot et al., 1994). However, the ability to prime the Th2 response declines with disease progression. Oral administration of antigens to overtly diabetic mice failed to block beta-cell destruction and could not normalize blood glucose levels. Another limitation of oral tolerance is the necessity to administer high and repeated doses of antigen in already immune hosts. We have shown that oral tolerance can be enhanced by the conjugation of insulin to mucosal adjuvant such as the B moiety of cholera toxin or CTB (Bergerot et al., 1997). Because of the limitations of antigen doses, CTB–insulin conjugates are an interesting alternative for human trials in pre-diabetic subjects. The intranasal route of delivery has also been investigated for insulin and GAD peptides in NOD mice, and led to immune tolerance similar to the oral route but requiring reduced amounts of antigen or peptides (Maron et al., 1998). The induction of oral tolerance can promote humoral autoimmune responses to other target tissue antigens. In type 1 diabetes, however, anti-beta-cell autoantibodies seem to be non-pathogenic. Concerns about the safety of oral antigen feeding have been raised by the observation that antigen feeding can induce cytolytic CD8+ cells and autoimmune disease in a highly polarized T-cell receptor (TCR) transgenic mouse model (Blanas et al., 1996). Doses and timing of antigen delivery are therefore critical factors that need to be investigated in humans to determine whether tolerance or immunization occurs. Other antigen-based forms of intervention have been attempted. A single subcutaneous injection of the B chain of insulin (Muir et al., 1995) or the immunodominant B chain peptide B9-23 in incomplete Freund's adjuvant (Daniel & Wegmann, 1996) prevented diabetes in NOD mice. NOD diabetes can also be inhibited by vaccination with a DNA construct encoding human heat shock protein 60 (HSP60), with a downregulation of T-cell responses to HSP60 and its peptide p277 and an increase in specific Ab responses suggesting a shift towards a Th2 response (Quintana et al., 2000). In a model of transgenic mice expressing lymphocytic choriomeningitis virus nucleoprotein (LCMV-NP) in their beta cells which develop diabetes after LCMV infection, the inoculation of plasmid DNA encoding the insulin B chain reduced the incidence of IDDM by 50% (Coon et al., 1999). The insulin B-chain DNA vaccination was effective through induction of regulatory CD4 lymphocytes that react with the insulin B chain, secrete IL-4, and locally reduce activity of LCMV-NP-autoreactive cytotoxic T lymphocytes in the pancreatic draining lymph nodes. Even if performed in an induced form of beta-cell autoimmunity, this study provides an elegant demonstration of the bystander effect of regulatory cells. The majority of autoimmune/inflammatory diseases are associated with excessive production of inflammatory cytokines such as IL-1, IL-12, TNF-α and IFN-γ and reduction of TGF-β, IL-4 and IL-10, which are most frequently protective. Blocking the Th1 pathway prevented the onset of diabetes in NOD females depending on the age of administration. Administration of anti-IL-12 mAb to female NOD mice between 5 and 30 weeks of age led to the suppression of both insulitis and diabetes (Fujihira et al., 2000). However, short-term administration of the same antibody at 2 weeks of age for 6 consecutive days resulted in the acceleration of the onset of diabetes. This study indicates that Th1 cytokines might have contradictory effects due in part to their involvement in both expansion and apoptosis of autoreactive cells. Gene therapy offers advantages for the immunotherapeutic delivery of cytokines or their inhibitors (Prud'homme, 2000). After gene transfer, these mediators are produced at relatively constant, non-toxic levels and sometimes in a tissue-specific manner, obviating limitations of protein administration. Genes encoding vectors that inhibit these cytokines, such as IFN-γ-receptor/IgG-Fc fusion proteins, are protective in NOD mice when injected in muscle tissue (Chang & Prud'homme, 1999). On the other hand, administration of IL-4 by repeated injections (Rapoport et al., 1993) or gene therapy with expression of murine interleukin 4 (mIL-4) cDNA (Cameron et al., 2000) prevented the onset of diabetes in NOD mice. Because of the large interference with the immune system, such strategies have limited application to humans without tissue selectivity. Many additional compounds might also interfere with Th1 function. Among them, vitamin D analogues stimulated suppressor cell activity in NOD mice and reduced both spontaneous and accelerated forms of diabetes (Mathieu et al., 1995). The risk of hypercalcaemia limits the use of such compounds in humans. In addition to interactions between the class II antigen complex on the surface of antigen-presenting cells (APCs) and the TCR on autoreactive T cells, several co-stimulatory molecules are necessary to activate the T cells (Fig. 1). Anti-CD40L mAb treatment of 3- to 4-week-old NOD females (the age at which insulitis typically begins) completely prevented insulitis and diabetes. In contrast, treatment of such mice with anti-CD40L at more than 9 weeks of age did not inhibit the disease process (Balasa et al., 1997). Blocking the CD40–CD154 co-stimulatory pathway by anti-CD154 antibody prevented allorejection and also recurrence of autoimmunity of islet allografts transplanted into chimeric NOD mice (Seung et al., 2000). On the other hand, disruption of the CD28/B7 pathway early in the NOD mouse strain, using CD28–/– and CTLA4Ig transgenic mice, promoted the development and progression of spontaneous autoimmune diabetes (Lenschow et al., 1996), underlining the role of this co-stimulatory pathway in autoreactive T-cell homeostasis. Fas ligand-mediated beta-cell destruction might play a role in promoting leucocytic infiltration of islets and beta-cell destruction in autoimmune diabetes in addition to toxic molecules such as nitric oxide or perforin (Thomas & Kay, 2000). Cytokines such as IL-1α, IL-1β and IFN-γ upregulate Fas expression on beta cells and facilitate Fas-dependent lysis by autoreactive CD4(+) T cells (Amrani et al., 2000). Nicotinamide, a derivative of the B vitamin niacin, significantly reduces nitric oxide accumulation in the NOD pancreas and protects islet cells against cytotoxic actions in vitro (Kolb & Burkart, 1999). This drug seems to reduce beta-cell apoptosis in rodent islets (O'Brien et al., 2000) and can protect human beta cells against radical-induced necrosis but not against cytokine-induced apoptosis (Hoorens & Pipeleers, 1999). Additional strategies have been tested successfully, including the use of growth factors such as IGF-1 (Bergerot et al., 1995). To date, no established therapy to prevent or delay the onset of diabetes exists but safe candidate trials are being conducted in humans. The long pre-diabetic period holds hope that the clinical disease might be delayed through therapies that can interrupt the causal disease process. Prevention strategies can be classified into several categories. Primary prevention targets subjects at very early stages of the disease. Secondary prevention is aimed at delaying and possibly suppressing beta-cell destruction in high-risk individuals. Tertiary prevention is initiated after the clinical onset of diabetes. Primary prevention of type 1 diabetes could theoretically be implemented in a general population approach by altering the lifestyle and/or environmental determinants known to be risk factors. This form of intervention might also apply to high-risk individuals with a screening strategy based on specific HLA-DQB1 genotypes. This is the basis of several studies, such as the DIPP project in Finland (Nejentsev et al., 1999; Kupila et al., 2001). The general approach is attractive because more than 90% of newly diagnosed patients have no family history. Accordingly, population-based approaches are needed for strategies aimed at an effective decrease in the incidence of diabetes. However, the inductive events that trigger autoimmunity are largely unknown. In contrast to experimental models, the immune system in humans is shaped by multiple interactions with nutritional and/or infectious factors. Nevertheless, a national study to compare the effects of feeding cow's milk vs. a casein hydrolysate formula (the TRIGGR study) has been started in Finland. This study was designed in light of the inverse relationship between duration of breast feeding and prevalence of diabetes (Virtanen et al., 1991) as well as data indicating that specific immune responses to cow's milk proteins occur early in life in high-risk individuals (Paronen et al., 2000). Other candidates are suspected, such as enteroviruses or vaccination procedures, but no prospective studies support their role in the disease. Although the prevalence of type 1 diabetes varies according to country, the increase in incidence and the younger age at onset suggest strong interactions between environment and susceptible genetic alleles. Secondary prevention applies to high-risk individuals identified by genetic and antibody markers. This refers mainly to first- or secondary-degree relatives of diabetic patients. Antigen-based interventions are attractive strategies. Antigens can be introduced systematically or through mucosal routes. Parenteral insulin administration is known to prolong the honeymoon period in recent-onset diabetic patients (Shah et al., 1989). Encouraging results of a pilot study in antibody-positive relatives (Keller et al., 1993) have led to a large intervention study in the USA, the Diabetes Prevention Trial (DPT-1), with daily injections of low-dose insulin in individuals with a risk greater than 50% of diabetes over 5 years (DPT-1 study group, 1995). However, the results presented at the American Diabetes Association meeting in June 2001 showed no difference in the rate of diabetes development in subjects treated by parenteral insulin in comparison to controls. Although disappointing, it is important not to overinterpret these negative findings. Insulin doses that were used in the DPT-1 trial (0·2 unit/kg/day) were much less than the doses used in animal studies (40 units/kg/day). In this respect, inactive insulin analogues could be valuable to give a higher dose but avoid the side-effects of hypoglycaemia. Data from animal studies have also led to oral insulin trials. Relatives at intermediate risk (5–50% risk of diabetes over 5 years) were recruited in a 4-year oral trial (DPT-1 oral), evaluating the effects of a daily oral administration of 7·5 mg insulin vs. placebo. A 50% reduction risk is expected. In a recent randomized trial comparing the oral administration of 2·5 mg, 7·5 mg of human insulin vs. placebo in type 1 diabetic patients with recent onset, no significant changes in fasting and stimulated C-peptide levels were noted after 1 year of follow-up (Chaillous et al., 2000). This negative study does not indicate that oral tolerance does not apply to human diabetes but outlines the need to reconsider the modalities of tolerance induction and its immunological monitoring. As suggested by animal studies, induction of tolerance at early stages of the disease process might be more efficient to protect the remaining beta cells. Promising trials using nasal insulin to prevent diabetes are also being conducted in Finland (DIPP project) and Australia (INIT trial). Additional forms of human intervention are also being investigated. Nicotinamide, a free radical scavenger, is currently being used in a large European trial (ENDIT) in first-degree relatives (Gale, 1996). This drug has been associated with protective effects in humans in a population-based study in New Zealand (Elliott et al., 1996). Although nicotinamide has not shown its efficacy in newly diagnosed patients or in the Deutsche Nicotinamide Intervention Study (DENIS) (Lampeter et al., 1998), the outcome of ENDIT is expected in 2003. A cure for diabetes has long been sought using several different biological approaches (Fig. 2), including islet transplantation, regeneration of beta cells and insulin gene therapy. Reintroduction of living beta cells has been attempted in order to control many of the complications associated with chronic hyperglycaemia and to improve quality of life by reducing severe hypoglycaemic episodes and the burden of insulin treatment. Beta-cell replacement suffers from several caveats such as the difficulty to isolate large numbers of islet cells from a single pancreatic gland and the fragility of purified beta cells. Given the pathophysiology of type 1 diabetes mellitus, the barriers to successful pancreatic allotransplants are both alloimmunity and recurrence of autoimmunity. Advances in surgical techniques as well as progress in the management of chronic immunosuppression have improved the results of pancreatic transplantation (Dubernard et al., 1998; Sutherland et al., 1999). More than 75% of the pancreatic grafts are still functioning at 1 year in patients undergoing simultaneous pancreatic and kidney allotransplantation. With the reduction of short-term complications associated mainly with difficulties of pancreatic duct drainage, infections and vascular thrombosis, the identification of several patients with pancreatic graft loss in the absence of histological signs of chronic rejection has led to the hypothesis of autoimmune recurrence (Sutherland et al., 1984; Tyden et al., 1996). Similar observations have been made in patients harbouring allogeneic islets (Stegall et al., 1996) despite chronic immunosuppression. Results from islet cell transplantation have greatly improved in the past few years. Thus far, this experimental procedure was restricted to patients with end-stage renal failure, with or awaiting a kidney graft. In this group of severely impaired patients, the results of islet transplantation were modest. By the end of 1999, the insulin-independence rate reported in the International Registry was 8% at 1 year. However, the most active transplantation centres reached an average of 20% insulin independence and 50–70% islet function rate (Secchi et al., 1997; Oberholzer et al., 2000). The recently reported data coming from Shapiro et al. (2000) open a new era in islet transplantation. The achievement of 100% insulin independence in a consecutive series of seven non-uraemic patients represented a major breakthrough. Patients were included because of severe metabolic disturbances. This study clearly emphasized the critical role of islet mass as all patients required at least two donor pancreas and an average 11 000 islet equivalents per kilogram of body weight. Finally, the improvement in beta-cell survival might also result from a glucocorticoid-free immunosuppressive regimen consisting of sirolimus or rapamycin, tacrolimus or FK-506, and daclizumab, an anti-IL-2R Ab. This combination of drugs might be associated with a limited rate of beta-cell apoptosis. Induction of specific forms of tolerance as well as cellular alternatives to beta cells might be useful to control the recurrence of autoimmunity, to enhance the long-term results of cell transplantation and to prevent any shortage in donor organs. Schematic representation of the different therapeutic strategies in type 1 diabetes. Insulinoma cell lines Engineered insulinoma cell lines (Newgard, 1994; Macfarlane et al., 2000) might represent an alternative to isolated islets for transplantation therapy of type 1 diabetes due to the shortage of available pancreatic glands. The use of immortalized cell lines for control of type I diabetes in humans will require their encapsulation and transplantation in non-native sites where relative hypoxia and cytokines might threaten their survival. Success of this approach would require development of cell lines that can withstand cytokine-mediated damage (Dupraz et al., 1999; Chen et al., 2000). Beta-cell neogenesis and pancreatic stem cells Stimulating beta-cell neogenesis is another attractive strategy for beta-cell replacement. Mature rat beta cells have a lifespan of approximately 48–56 days and are replaced by the replication of pre-existing beta cells and by the differentiation and proliferation of new beta cells derived from the pancreatic ducts. Expansion of human ductal tissue in vitro, and its subsequent differentiation to islet cells, has been observed using ductal cells overlaid by a thin layer of Matrigel (Bonner-Weir et al., 2000). At present, the amount of islet-like clusters obtained with this technique is not adapted for transplantation purposes and the ability of these cells to restore blood glucose levels in vivo is still unproven. In fact, the capacity to obtain large amounts of insulin-secreting cells from already differentiated cells is low. The use of precursor cells differentiating into mature beta cells has been proposed as an alternative to provide unlimited sources of insulin-secreting cells. The characterization of transcription factors and the role of growth factors in pancreatic development and cell proliferation have provided an important guide to produce islet cells in vitro. For example, it has been shown recently that the insulinotropic hormone glucagon-like peptide 1 (GLP-1), which is produced by the intestine, enhances the pancreatic expression of the homeodomain transcription factor IDX-1 that is critical for pancreatic development and the transcriptional regulation of the insulin gene (Stoffers et al., 2000). The capacity of embryonic stem cells to differentiate along a pancreatic endocrine path once transfected with an insulin promoter is exciting (Soria et al., 2000), but the fact that this procedure gave rise to proliferating cells needs further investigation before clinical application. Xenogeneic islet cells Porcine islets have been considered as an unlimited supply of insulin-producing cells for transplantation. However, the immunological barrier to xenogeneic grafts appears greater than for allo-transplantation and the development of transgenic pigs that express human genes and lack immunodominant xenoantigens has been undertaken. The disadvantage of such therapies is the possible contamination of the recipient by porcine endogenous retroviruses. These concerns have led the Food and Drug Administration (FDA) to halt xenogeneic islet grafts in the USA until the risk of retroviral infection is evaluated in patients who have already received a graft. A gene therapy approach to treatment of type 1 diabetes could be used either to supply insulin delivery or to induce the differentiation of insulin-secreting cells. Such possibilities have been limited, however, in the past decade due to the difficulties in achieving efficient gene transfer and long-term gene expression. Effective and safe insulin gene therapy will require regulation of transgenic insulin secretion to avoid hypoglycaemia. In order to regulate a liver-targeted insulin transgene, the expression vector needs inclusion of glucose-responsive elements within its promoter. Transgenic insulin production has been obtained in a rodent model of diabetes mellitus using a liver-targeted insulin transgene which included glucose-responsive elements (Thule & Liu, 2000). More recently, a recombinant adeno-associated virus (rAAV) that expresses a single-chain insulin analog was used in streptozotocin-induced diabetic rats or NOD mice and corrected diabetes for periods of 8 months (Lee et al., 2000). This analog has 20–40% of the activity of native insulin without the need for enzymatic conversion. The gene encoding the insulin analog was placed under the control of the hepatocyte-specific L-type pyruvate kinase promoter, which regulated the insulin analog expression in response to blood glucose levels. However, the application of this promising technique to humans remains to be determined. The differences in hepatic physiology between rodents and humans, as well as the delayed and prolonged insulin secretion obtained with the viral construct, would certainly lead to hypoglycaemia. Administration of high levels of insulin through an irreversible gene delivery system is also a matter of concern. Insulin delivery should be modulated in response to changing needs such as exercise, food intake or fasting periods. Gene therapy might also be applied to reprogramming liver stem cells into insulin-producing cells. The protein encoded by the pancreatic and duodenal homeobox gene 1 (PDX-1) is central in regulating pancreatic development and insulin gene expression. Recombinant-adenovirus-mediated gene transfer of PDX-1 to the livers of BALB/C and C57BL/6 mice led to an increase in immunoreactive insulin levels by reprogramming liver cells in a beta-cell phenotype (Ferber et al., 2000). Although premature, this study provides an interesting approach for generating surrogate beta cells. The recent breakthroughs in the control of immune responses obtained in animal models of autoimmune diabetes and the strategies to replace the deficient beta cells are very exciting. Understanding the pancreatic beta-cell ontogeny is an important challenge to support the development of cell therapies using either duct cells or appropriate stem cells. Research in islet-cell and stem-cell transplantation is set to develop rapidly. Safe suppression of autoimmunity must also be achieved to block the cellular mechanisms leading to diabetes or to autoimmune recurrence on grafted cells in already diabetic subjects. The real challenge for the new millennium is to indicate whether and when the encouraging results obtained in animal models will also apply to humans.