The β-Cell in Type 1 Diabetes Pathogenesis: A Victim of Circumstances or an Instigator of Tragic Events?

β-Cell failure in type 1 diabetes creeps forward at a pace that is defined by a dynamic relationship between immune cells and β-cell intrinsic responses (1,2). This relationship is affected by host factors such as genetics, age, and by poorly defined environmental triggers. In some individuals, autoimmunity is initiated but does not progress further, notably in the presence of a single autoantibody (aAb) (3). This may reflect an aborted immune process or responses of β-cells, or other cells, that prevent progression to the next stage of autoimmune amplification and β-cell destruction (4). However, in others, particularly young individuals, β-cell destruction is more rapid once autoimmunity (stage 1) or hyperglycemia (stage 3) appears (5,6), with little evidence of remission phases.

The precise molecular and cellular mechanisms that account for β-cell death remain unclear. Direct cytotoxic lysis by CD8⁺ T cells reactive to β-cell antigens such as preproinsulin is postulated. However, immune infiltrates in the pancreas are sparse at the time of diagnosis. While the late timing of these histological studies, i.e., years after autoimmune initiation and even after β-cell destruction and clinical onset, may account for these findings (2), other mechanisms capable of mediating β-cell killing at a distance have been suggested. Inflammatory molecules are produced by several cell types, including β-cells themselves (7), and may cause dysfunction or killing. Findings of shrinkage of the exocrine pancreas before type 1 diabetes diagnosis suggest that inflammation of the whole organ may occur and that β-cells are perhaps killed by soluble mediators produced by cells that are not present in the islets. Indeed, findings of a recent trial of anti–TNF-α antibody showed how neutralizing this cytokine may arrest progression in patients with new-onset type 1 diabetes (8). However, the way in which these acute inflammatory mediators affect the chronic progression remains undefined.

A number of investigators have described, as part of this immune/β-cell dynamics, how β-cells may respond to the autoimmune attack in ways that may promote their demise or protect them (9). The question of whether β-cells die by homicide versus suicide was originally proposed by G. F. Bottazzo (10), but newer studies call for revisiting this question. Clearly, both T cells and β-cells are active disease players, but is one the initiator and the other the victim or do both accelerate β-cell death? We here review this dynamic interplay in the light of recent studies.

The increasing evidence that T cells, together with other immune players, are not the sole culprits in type 1 diabetes pathogenesis has recently been reviewed (2). Briefly, arguments for their nonexclusive involvement include the following. First, experimental autoimmune diabetes cannot be induced in animal models by immunization with islet extracts or β-cell antigens alone (2), as is the case for other autoimmune diseases such as experimental autoimmune encephalomyelitis (a model of multiple sclerosis). Second, human T cells adoptively transferred into immunodeficient mice can cause insulitis but are not diabetogenic in the absence of strong priming conditions, e.g., streptozotocin administration to induce some β-cell death and self-antigen release (11,12). Third, histopathological, age-related endotypes (13,14) mirroring clinical endotypes defined by age and aAb targets (5,6) indicate that a number of other factors may determine rates of progression. Finally, a universal state of “benign” autoimmunity is found in the circulation of all individuals, as shown by the presence of islet-reactive (and largely naive) CD8⁺ T cells (15–17) and islet-reactive CD4⁺ T cells (18,19). T cells with the same antigen specificities are enriched in the pancreas of donors who had type 1 diabetes (15,16), pointing to local factors that may affect the phenotypes of T cells and attract them into the target organ. Although two case reports of type 1 diabetes transfer, on grafting of bone marrow from a donor who had diabetes (20) and grafting of islets into a recipient with diabetes (21), argue for a central role of T cells, an alternative explanation, regarding the contribution of β-cells, is possible, as explained below.

The active role of β-cells in type 1 diabetes pathogenesis has been linked to their specialized function of insulin secretion (1)—a paradigm that may apply more generally to other endocrine cells. Namely, the high endoplasmic reticulum (ER) and oxidative stress imposed by the requirements for rapid and considerable insulin biosynthesis, the rich islet vascularization—which exposes cells to inflammatory soluble mediators and cells, and the secretion of insulin and other antigenic granule contents directly in the bloodstream with high amplitude (22) are all features that may render β-cells more vulnerable to autoimmunity.

Importantly, new emerging data extend this picture to alterations of the exocrine pancreas (23). A reduced pancreas organ weight was reported in donors positive for a single aAb (24) and even in aAb-negative first-degree relatives (25). Investigators of other studies reported reduced serum trypsinogen and lipase levels starting at stage 1 (two or more aAb) type 1 diabetes (26,27) and enriched exocrine immune infiltration (28) and HLA class II expression in ductal cells from donors who had type 1 diabetes (29). In another notable recent single-cell epigenomics study (30), high expression was documented of novel type 1 diabetes risk gene variants (e.g., chymotrypsinogens, lipase, and cystic fibrosis transmembrane conductance regulator) in acinar and ductal cells, which is postulated to promote subclinical pancreas inflammation favoring later immune infiltration.

The pathogenic role of β-cells could theoretically be called into question at three sequential stages (Fig. 1).

1. Self-initiated homicide: are β-cells the original drivers of T-cell autoimmunity?

2. Self-amplified homicide: do β-cells enhance T cell–mediated killing?

3. Self-finalized suicide: do β-cells under attack eventually die by suicide?

Studies by Larger, Boitard, and colleagues (31) nearly 30 years ago showed that β-cells are needed to initiate autoimmune diabetes in NOD mice. These results support a primary role of β-cells in developing an autoimmune repertoire endowed with β-cell specificity, since splenocytes from β-cell–deficient mice could not transfer diabetes but the mice were otherwise immune competent and developed other autoimmune manifestations, i.e., sialitis. This observation pointed to an early critical time window in which T cells are primed by β-cell antigens and trigger diabetes. Of further note, this priming occurs primarily in pancreatic lymph nodes and is an early event, as pancreatic lymphadenectomy at 3 weeks, but not at 10 weeks of age, led to near-complete diabetes protection in NOD mice (32). Pancreatic lymph nodes may also harbor stem-like CD8⁺ T cells that can proliferate and generate effector T cells that invade islets (33).

Several lines of experimental evidence suggest that β-cell aberrancies, which may include increased antigen exposure and neoantigen expression, may be the first drivers of islet autoimmunity. Islet autoimmunity does not develop if β-cells are not visible to the immune system due to reduced antigen expression, i.e., if immune ignorance is maintained. An emblematic example is provided by NOD mice carrying an insulin B16 mutation that abolishes the immunogenicity of the initiating epitope insulin B9–23, which are completely protected from diabetes (34). Serreze et al. (35) found that initiation of autoimmune diabetes in NOD mice is dependent on MHC class I. β2-microglobulin (β2m)^−/− mice, which are MHC class I deficient, did not develop insulitis or diabetes, whereas splenocytes from diabetic NOD donors transferred diabetes into both wild-type and β2m^−/− NOD/scid recipients. In contrast, splenocytes from young prediabetic NOD donors transferred diabetes only to wild-type NOD/scid recipients, while β2m^−/− NOD/scid recipients required prior grafting with wild-type islets to become diabetic. The authors concluded that islet expression of MHC class I (i.e., β-cell visibility) was needed to prime T cells, which could then mediate β-cell destruction independent of MHC class I (35). Other adoptive T-cell transfer models suggest that, in addition to antigen presentation, β-cell alterations such as those imprinted in the NOD background or induced by streptozotocin are needed to enable antigen-driven T-cell priming and β-cell killing (2). On the same lines, T-cell receptor transgenic models in which T cells recognize a foreign antigen transgenically expressed in β-cells show that provision of costimulatory or inflammatory signals is required to activate these T cells (36). These mouse studies may provide an interpretation for some observations in patients. In a case report of identical twin-to-twin transplantation of a nondiabetic pancreas into a recipient with type 1 diabetes, the transplantation led to rapid autoimmune relapse (21); conversely, bone marrow transplantation from a sibling with type 1 diabetes into an HLA-identical recipient without diabetes triggered type 1 diabetes (20). In both cases, the previous priming of the autoimmune T-cell repertoire prior to transplantation may have bypassed the need for a β-cell trigger.

Results of recent elegant studies have revived interest in this hypothesis of self-initiated homicide in showing that the induction of an early and transient β-cell dedifferentiation protects NOD mice from diabetes. A first model of NOD mice with high β-cell proliferation rates and decreased antigen expression (37) was obtained either with a liver-specific insulin receptor genetic knockout or by pharmacological treatment with an insulin receptor antagonist from 4 to 6 weeks of age. In a second model (38), a conditional knockout of the inositol-requiring enzyme 1α, which activates the unfolded protein response (UPR) in β-cells, was induced by tamoxifen treatment during the neonatal period (1–5 days of life), resulting in a similar dedifferentiated β-cell phenotype. Both NOD models are protected from diabetes.

Can all these extreme murine models that “blind” the immune system from β-cells be taken as evidence for an active role of β-cells in initiating autoimmunity? Other arguments from human genetic studies need to be considered. First, the contribution of type 1 diabetes–associated gene variants expressed in β-cells to genetic susceptibility is marginal in comparison with HLA class II polymorphisms (2)—a feature that type 1 diabetes shares with several other autoimmune diseases. Second, the major influence of HLA class II haplotypes is on the antigen specificity and rate of aAb seroconversion (i.e., autoimmune initiation) rather than on the rate of actual β-cell destruction (i.e., clinical progression) (2,39,40), suggesting a driving effect on pathology. This may further suggest that β-cell–related non-HLA gene variants may act as modulators of this later clinical progression rather than drivers. Third, monogenic forms of neonatal diabetes causing β-cell defects, including mutations (e.g., in INS and EIF2AK3 genes) that induce severe ER stress, show little evidence of islet autoimmunity, even in the context of high-risk HLA class II haplotypes, with only 5.4% positivity for islet aAb, mostly (92%) for a single aAb (41). This is observed in spite of a dysregulated expression of genes (and proteins) associated with monogenic forms of diabetes resulting from ER stress in the pancreas of donors who had type 1 diabetes and were aAb⁺ (42), suggesting a role of primary β-cell alterations as disease contributors rather than drivers.

However, an alternative explanation could be that the β-cell contribution to type 1 diabetes pathogenesis may simply be less genetically imprinted and more environmentally driven than the immune contribution. This hypothesis may invite the search for environmental triggers that may compromise β-cell health rather than immune tolerance per se, e.g., islet-tropic viral infections and xenotoxic agents.

Altogether, this literature suggests that β-cell stress/dysfunction may play a role at initiating islet autoimmunity, and then significantly contribute to its amplification, as discussed in the next section.

A key observation suggesting the role of β-cell–driven initiation and amplification mechanisms is the heterogeneity of insulitis across islets, which do not appear to be attacked and destroyed at the same time (43). There are multiple mechanisms by which β-cells amplify the autoimmune attack (Fig. 2). The inflammatory environment of insulitis, particularly interferons, stimulates several active responses in β-cells that go beyond the passive effect of inducing ER stress, dysfunction, and apoptosis. First, inflammation drives β-cells themselves to release chemokines (e.g., CCL5, CXCL2, CXCL9, CXCL10) and cytokines (e.g., IL-1β, IL-6) (7), which further attract immune cells to the pancreas and contribute to their activation in situ.

The second major outcome of islet inflammation is HLA class I upregulation, which results in the exposure of a larger and more diversified repertoire of antigenic peptides (16), thus increasing the visibility and vulnerability of β-cells to infiltrating T cells. Not surprisingly, peptides derived from proteins of secretory granules (e.g., insulin, chromogranin A, secretogranin-5, urocortin-3, proconvertase-2) are highly represented across different human and murine MHC restrictions (16,17) and are targeted by diabetogenic CD8⁺ T cells, which induce disease when transferred into NOD/scid mice (17). This pathogenic role may rely on features that granule proteins share with (pro)insulin (44). All of these antigens are synthesized as proprotein precursors, which are subsequently enzymatically cleaved, mostly by proconvertases, to generate their bioactive products. Hence, the impaired proinsulin processing by proconvertases described in β-cells from patients with type 1 diabetes (44,45) may extend to all of these proteins, possibly diverting their degradation toward the MHC class I presentation pathway. In addition, all of these proteins are released in the bloodstream with insulin exocytosis, both in intact form and as antigenic degradation products (22). They can thus be taken up by extrapancreatic antigen-presenting cells and prime T cells at a distance on uptake by extrapancreatic antigen-presenting cells (1).

The MHC peptide repertoire displayed by β-cells further includes neoantigens, i.e., peptide sequences that are not templated in the genome and may therefore be regarded as nonself and trigger autoimmunity, since they have not participated in the physiological mechanisms of immune tolerance. These neoantigens are produced through different mechanisms: alternative mRNA splicing (16,17), out-of-frame protein translation due to defective ribosomal scanning under high insulin demand (46), posttranslational modifications (47), and transpeptidation (16,48), i.e., the fusion of noncontiguous peptide fragments from the same protein or from two distinct partners.

Other amplification mechanisms may be related to β-cell sensitivity to stress-induced senescence. In a recent study (49), the accumulation of senescent β-cells in NOD mice was documented during a critical early time window that marks the transition from peri-insulitis to destructive insulitis. Although these senescent β-cells increase the expression of prosurvival factors such as Bcl2, they also have a key role in amplifying the autoimmune response because they produce a specific secretome (the so-called SASP [senescence-associated secretory phenotype]) enriched in factors (e.g., IGFBP3, serpin E1) that propagate this senescent phenotype to neighboring β-cells, as well as in cytokines/chemokines (e.g., IL-6, CXCL10) that attract and activate immune cells in the pancreas. Senescence markers (CDKN1A, serpin E1, IL-6) were also observed in the pancreas of patients with type 1 diabetes (49). Importantly, when senescence was inhibited by an anti-Bcl2 small molecule that drives apoptosis specifically in these cells, diabetes was prevented, thus documenting the pathogenic role of the senescent phenotype. This observation identifies a novel therapeutic target involving a rather counterintuitive strategy of inducing apoptosis of selected β-cell subsets. This also suggests that the physiologic process of clearing senescent β-cells, or hypersecreting mutants (adenomas) (50), is misdirected in autoimmunity. It will be important to clarify whether the relationship of β-cell dysfunction, senescence, and apoptosis is sequential, mutually exclusive, and/or naturally reversible.

Besides inciting autoimmunity via all these immunogenic signals, β-cells may further amplify these mechanisms by ceasing to secrete protective antiapoptotic factors. A notable example is growth/differentiation factor (GDF)15, a transforming growth factor-β superfamily member that was found to be downregulated in β-cells from individuals with type 1 diabetes and on in vitro exposure to IL-1β and IFN-γ (7). Although its mechanisms of action remain unclear, GDF15 protects NOD mice from diabetes (7), presumably through anti-inflammatory effects, and exerts additional effects in the hindbrain (reducing food intake and promoting weight loss) and in peripheral tissues (hepatocytes, myocytes, and adipocytes), by increasing insulin sensitivity.

Further compensatory mechanisms mounted by β-cells may be self-detrimental (“suicidal”) in the long-term, acting through intrinsic mechanisms that do not require the participation of other cells (Fig. 2).

Some cytokines (e.g., IL-1β, IL-6) that are secreted by β-cells themselves (7) can be proapoptotic in an autocrine/paracrine loop. Their secretion may be particularly important during the extended period of time—as long as 5 years—when insulin secretion may be impaired but clinically silent. The ability to compensate for metabolic demands becomes critical during this time for β-cell survival. As metabolic demands increase, reactive oxygen species are needed for glucose-induced insulin secretion (51) and mass expansion (52). However, the low expression of antioxidant enzymes such as catalase and superoxide dismutase exposes β-cells to cell-intrinsic oxidative damage, particularly in the ER (53). Even β-cell identity may be affected by this oxidative stress, since exposure to hydrogen peroxide can induce loss of nuclear localization of MafA, a transcription factor associated with mature β-cells that may enhance the β-cell antioxidant defenses, protecting them against deterioration during hyperglycemia (54–56). Increased metabolic demands also upregulate the UPR, which is constitutively active in β-cells due to their basal ER stress from high protein (mostly insulin) biosynthesis rates. This increased UPR can drive apoptosis when reaching a critical threshold (1). Indeed, UPR upregulation was observed in β-cells from two type 1 diabetes mouse models and from patients (57), as identified by the expression of activating transcription factor 6 (ATF6) and spliced X-box binding protein 1 (sXBP1). The deleterious effect of what is initially a compensatory response is suggested by the observation that surviving β-cells display lower ATF6 and sXBP1 expression and that relieving UPR with tauroursodeoxycholic acid protects mice from diabetes (57).

We recently showed that Tet2 is increased in β-cells from NOD mice and in human insulitis (58). Its deleterious β-cell intrinsic outcomes were documented in several ways. In β-cells that survive autoimmune attack, Tet2 expression was reduced. Despite unchanged insulitis and T-cell diabetogenicity, Tet2^−/− NOD mice were protected from diabetes, even after grafting of Tet2^+/+ bone marrow or transferring diabetogenic splenocytes. Tet2 affects the epigenetic control of signaling molecules such as STAT1, nuclear factor-κB, and IRF2. While its loss was associated with CXCL10 downregulation by β-cells, there was also reduced expression of the CXCR3 homing receptor on β-cell–reactive T cells in pancreatic lymph nodes.

Protective mechanisms are also mounted by β-cells (Fig. 2). During progression of autoimmunity in NOD mice, we described degranulation, loss of Ins1/Ins2 gene expression, and acquisition of a more undifferentiated phenotype, with expression of other islet hormones (e.g., Gcg, Sst) and stem cell markers (Sox9, L-Myc, Oct-4) in residual β-cells (59). In addition to insulin, there was reduced expression of other diabetes autoantigens (GAD, ZnT8, IA-2). In line with these results, Damond et al. (60) compared β-cells of patients with new-onset or long-standing type 1 diabetes with those of individuals without diabetes by image cytometry pseudo-timing using Network for Pancreatic Organ donors with Diabetes (nPOD) specimens. β-Cell destruction was preceded by downregulation of identity markers (INS, proinsulin, IAPP, PTPRN) with (pseudo)time, without changes in β-cell transcription factors PDX1 and NKX6-1. Loss of lineage-specific markers may thus be an early event, before further disease progression driven by β-cell death. In our study (59), dedifferentiated β-cells also expressed immunoregulatory ligands (PD-L1, Qa-2) and resisted immune killing by islet infiltrates or cytokines—unlike their mature counterparts. PD-L1 is induced on human and murine β-cells by IFN-γ and IFN-α (61). β-Cells also secrete antimicrobial peptides (e.g., CRAMP), a process triggered by in-vitro exposure to IL-1β or LPS (62) that is impaired in NOD mice (63). Besides inducing tolerogenic antigen-presenting cells and regulatory T cells (63), these peptides reduce β-cell apoptosis and production of inflammatory prostaglandin E2 (62). These mechanisms, while protective against killing in specific experimental settings, clearly fail to arrest the autoimmune response in most clinical settings.

Altogether, the available literature suggests that, following autoimmune insult, β-cells may also contribute in a direct way to their own killing. This concept of self-detrimental versus protective β-cell responses has clinical relevance. Successful immunotherapies (e.g., teplizumab, antithymocyte globulin, alefacept, rituximab) share a mechanism of modulation/deletion of inflammatory cells. While treatment is often followed by C-peptide stabilization, a decline subsequently occurs in many patients (64), suggesting either that the conditions that led to clinical onset have recurred or that β-cell intrinsic mechanisms may drive this decline. As an example, even continuous administration of CTLA-4 Ig at type 1 diabetes onset did not avoid a decline in β-cell function after an initial improvement (65). This suggests that maintaining β-cell intrinsic protective mechanisms together with immunomodulation may limit this later decline.

The concepts of β-cell responsiveness to autoimmune attack also suggest ways in which autoimmunity might be arrested with a combination of agents that target immune cells and β-cells. Evidence (33,66) for stem-like features of islet-reactive CD8⁺ T cells that can proliferate and are the source of short-lived effector cells that migrate to the pancreas highlights the challenge in purging the repertoire of effector T cells sufficiently to prevent disease (33). Intriguingly, results from teplizumab trials (67) suggest that immunotherapies may be most effective when β-cell stress is occurring, i.e., at stage 2 of rapid C-peptide decline, consistent with preclinical data in NOD mice. Although the β-cell changes occurring during this period of rapid progression may expose new and more antigen targets, the T cells that recognize them may represent the more differentiated ones that are amenable to treatment.

Modulating the β-cell and antigen targets of autoimmune T cells may thus be needed in addition to immunotherapies. Clinical studies are ongoing with agents that are postulated to limit β-cell stress responses, such as verapamil (clinical trial reg. no. NCT04545151, ClinicalTrials.gov), a calcium channel blocker that inhibits TXNIP, a mediator of oxidative stress (68), and tauroursodeoxycholic acid (NCT02218619), a chemical chaperone that mitigates ER stress (57). In a previous study of the tyrosine kinase inhibitor imatinib as a single agent, a significant effect on C-peptide responses was not sustained (69), but ongoing studies are evaluating effects on markers of β-cell stress. Other studies with agents such as JAK1/2 inhibitors, that can block critical signaling pathways in β-cells in addition to blunting T-cell activation, are under development. Finally, in a randomized four-arm placebo-controlled trial, anti–IL-21 antibodies and the GLP-1 receptor agonist liraglutide significantly reduced C-peptide decline at 54 weeks in patients with recently diagnosed type 1 diabetes compared with placebo (70). The finding that this decline was lower than with either agent alone suggests that immune and β-cell combination therapies may be synergistic in achieving clinical outcomes.

These studies highlight the dynamic relationships between immune effector cells and their β-cell targets. The findings have shown how both parties contribute to killing but also suggest that therapies that are directed only at one arm may be insufficient to completely arrest the ongoing process. Much work has been done to characterize the autoimmune repertoire, including the targets and features of T and B cells that can mediate disease. Attention to the attributes of β-cells that can instill protection from or participation in autoimmunity are now needed.

Acknowledgments. The authors apologize to colleagues whose publications could not be referenced due to space constraints. Figs. 1 and 2 were created with elements from Servier Medical ART (SMART) (https://smart.servier.com).

Funding. Work performed in the laboratory of R.M. is supported by The Leona M. and Harry B. Helmsley Charitable Trust (no. 1901-03689), Agence Nationale de la Recherche (ANR-19-CE15-0014-01), and Fondation pour la Recherche Medicale (EQU20193007831) and by the Innovative Medicines Initiative 2 Joint Undertaking under grant agreements 115797 and 945268, INNODIA and INNODIA HARVEST, which receive support from the European Union Horizon 2020 program, the European Federation of Pharmaceutical Industries and Associations, JDRF, and The Leona M. and Harry B. Helmsley Charitable Trust. C.H. was funded by an Année Recherche fellowship of University of Paris-Saclay. Work performed in the laboratory of K.C.H. is supported by National Institutes of Health grants DK057846, DK101122, DK104205, and DK045735; JDRF grant SRA2014-142-S-B; a Pilot Award from the Yale Diabetes Research Center, P30DK116577; and a gift from the Howalt family.

Duality of Interest. K.C.H. is co-inventor on a patent application for use of teplizumab. No other potential conflicts of interest relevant to this article were reported.