Hypothesis: Induction of Autoimmunity in Type 1 Diabetes—A Lipid Focus Free

Type 1 diabetes (T1D) is an autoimmune disease specifically targeted to the pancreatic β-cells within the islets of Langerhans (1). T1D develops within the context of a bidirectional interaction between immune cells, which invade the islet and release a variety of chemokines and cytokines, and putative immunogenic signals produced by injured or dying β-cells (1–3). This dialogue is shaped by several factors that contribute to T1D risk and heterogeneity, including demographic features, such as age; environmental factors, such as infections; and host genetic architecture (4,5). Information describing specific molecular details of this autoimmune process has been obtained but has not led to effective preventive or treatment interventions (2,3,6). At present, neither the trigger nor the mechanism for autoimmunity or destruction of healthy β-cells by the immune system in T1D is well understood.

Over 90% of people with newly diagnosed T1D have measurable antibodies against specific β-cell proteins, including insulin, GAD, islet antigen 2, and zinc transporter 8, among others (7,8). Birth cohort studies have demonstrated that the presence of two or more serum autoantibodies in children is associated with an 84% risk of clinical T1D by the age of 18 years (9). Based on these findings, stage 1 T1D is defined now as the presence of two or more autoantibodies, while stages 2 and 3 are defined as the progression of metabolic abnormalities from abnormal glycemia to overt diabetes (10).

Previous studies have focused on the triggers, genes and proteins that differentiate individuals with T1D from those without diabetes with a focus on the β-cell as a target of immune destruction and blood glucose as the main abnormality. Our focus is on metabolic communication as an early instigator with the β-cell as an active participant together with the immune cells. We present a testable hypothesis for the induction of autoimmunity: We propose that the combination of elevated levels of cytokines and free fatty acids (FFA) with accompanying cytokine-mediated inhibition of FFA oxidation (11,12) and increased reactive oxygen species (ROS) production (13) increases granule trafficking and basal insulin secretion (14) leading to overworked β-cells and possibly presentation of granule proteins to immune cells. We are referring here not to normal daily fluctuations in FFA in response to meals or overnight fasting but, rather, to increases above that normal range in response to prolonged fasting (15) or lipolytic stimuli.

Ex vivo treatment of human islets from donors without diabetes with proinflammatory cytokines like TNFα, IL1β, and IFNγ produces transcriptional signatures that are remarkably similar to those observed in islets from organ donors with T1D. Cytokines can be elevated in response to increased immune cell infiltration within the islet, in response to infections (16) and accompanying nutrient-induced inflammation (17). In this regard, infections have long been suspected to trigger autoimmune reactions (4). Much consideration has been given to a potential viral etiology but with conflicting evidence identifying any specific virus (18). Higher incidence of T1D is also correlated with less frequent early childhood infections resulting in less immunity in developed countries. Major infections, that most often precede development of T1D, increase circulating inflammatory cytokines and increase FFA levels, due to stimulation of lipolysis in response to lack of nutrient intake. Interestingly, serum cytokine levels, particularly TNFα, are significantly elevated in all T1D patients (19).

Cytokine-mediated inhibition of FFA oxidation also favors glycolytic-dependent inflammatory immune cells over lipid-oxidizing anti-inflammatory immune cells. Interestingly, certain cells of immune system, T helper (Th)1, Th2, and Th17 cells, derive their energy from glycolysis, while regulatory T cells (Tregs) require a high level of lipid oxidation (20,21). In this way, T cell differentiation can be manipulated, as inhibition of glycolysis blocks Th17 and promotes Treg differentiation (22). More recently, inhibition of glycolysis has been shown to drive T cell exhaustion and protect against T1D development in mice (23). Notably, teplizumab, an anti-CD3–based drug, was shown recently to delay the onset of stage 3 T1D by a median of 32.5 months in individuals at high risk (24,25). Teplizumab was associated with the acquisition of an exhausted T cell phenotype defined by an increased frequency of KLRG⁺TIGIT⁺CD8⁺ T cells. Finally, the inflammatory M1 phenotype of macrophages depends on glycolysis, while the anti-inflammatory M2 phenotype uses lipid oxidation (26). Hence, modulation of FFA oxidation, by cytokines or drugs, may impact both innate and adaptive immune profiles and participate in the triggering event. Further investigation is required regarding a possible role of FFA receptor modulation (27) in determining the response of immune cells and β-cells to elevations in FFA. In addition, that specific nutrient ingestion may also impact the immune system requires more detailed consideration.

Sepsis is a fairly common example of cytokine-mediated damage, with lethal consequences in ∼15–30% of cases, due to the body’s overactive response to infection. It is not known why this occurs in some but not others with similar severity of infection. Our studies with cytokines began with a goal of understanding the mechanism for sepsis and the variable outcomes (11). Links were established between inflammatory cytokines and altered FFA metabolism in hepatocytes and endotoxin-treated animal models (11,28,29). Cytokines markedly inhibit hepatic mitochondrial fat oxidation, thus promoting elevation of cytosolic lipids, formation of complex lipids, and increasing lipid stores.

Severe sepsis is characterized by dysregulated systemic inflammation in response to infection, with impaired metabolism and mitochondrial function (30–32). It is associated with an anorectic response that stimulates adipocyte lipolysis and elevates circulating lipids. At the same time, PPARα is downregulated, impairing the ability of mitochondria to use FFA. This in turn leads to energy deficiency, lipotoxicity, and mitochondrial dysfunction due to increased ROS production and inefficient ROS removal. Indeed, in a recent metabolic proteomic screen investigators identified significant differences in glucose and fatty acid metabolism pathways between sepsis survivors and nonsurvivors (33). Accumulating evidence suggests that long-chain FFA act as natural ligands for free fatty acid receptors (FFARs) that can impact immune cells as well as pancreatic islets: FFAR1 (GPR40), FFAR2 (GPR43), and FFAR4 (GPR120). Both inflammatory and anti-inflammatory actions have been reported as well as a role for short-chain fatty acids derived from dietary fiber and ultra-processed food (27,34,35). Observed pathology in sepsis, though more severe and more systemic, has similarities to changes that are also characteristic of T1D.

Our findings in sepsis led us to consider that the response of immune cells to FFA may also be important in the etiology of T1D, worthy of more detailed investigation. Indeed, lipid abnormalities characterize individuals susceptible to T1D, with many classes of lipids altered as early as 12 months of age (36,37). These differences are observed even before the first appearance of islet autoantibodies. During the first year of life, i.e., before the median age of seroconversion, most lipids measured in peripheral blood mononuclear cells (PBMCs) were downregulated in the groups with a single positive autoantibody and diabetes compared with control subjects. The occurrence of low levels of triglycerides (TGs) enriched with saturated fatty acids such as palmitate and myristate suggests diminished dietary fat consumption, a high-carbohydrate diet, or impaired de novo lipogenesis. At 24 months, a time coinciding with seroconversion to T1D in most children in the single autoantibody group, there was an increase in the level of lipids in the PBMCs in comparison with the control group. We also found increased lipid content in PBMCs from established T1D subjects (38). In another study, a negative correlation was found between FFA and C-peptide, thus supporting the concept that higher levels of FFA occurred in children with C-peptide deficiency at onset of T1D (39). Several additional small studies documented that antibody-positive children had higher levels of odd-chain TGs and polyunsaturated fatty acid–containing phospholipids (40), exaggerated increases in plasma glycerol and FFA during exercise with no abnormalities in glucose or insulin (41), and increased FFA and glycerol (42). Results of a study of adults who developed T1D showed that TG levels were positively associated with T1D risk (43). It is interesting to note that functional annotation analyses of islet gene networks from both T1D and type 2 diabetes revealed significantly enriched lipid metabolism pathways (44). Important areas of concern include inadequate focus on possible lipid abnormalities preceding disease diagnosis and mechanistic information differentiating β-cell responses to nutrient-derived FFA that stimulate insulin secretion and the much higher FFA levels that occur during fasting and elevate basal but do not impact stimulated secretion.

The similarities to sepsis in cytokine and FFA alterations in the development of T1D led us to compare responses to these agonists in cultured human fibroblasts obtained from individuals with or without T1D. We identified major signaling differences between cells from affected individuals and control subjects in response to both cytokines and lipids (Fig. 1). Cultured human fibroblasts from 10 unrelated subjects with T1D responded to the inflammatory cytokines (TNFα and IL1-β) very differently from cultured human fibroblasts from 10 subjects without diabetes, with almost no overlap between the two groups (45). Patient cells exposed to these cytokines exhibited significantly enhanced G-protein–coupled receptor–mediated Ca²⁺ transients and accumulated greater lipid stores compared with control subjects (38,45). No differences were observed in Ca²⁺ signals in the absence of cytokines. The addition of FFA alone or FFA plus cytokines amplified the differences between T1D and control subjects in both Ca²⁺ signals and lipid handling (Fig. 1). These findings led us to suspect that altered cytokine or metabolic responses might be an unrecognized predisposing characteristic of individuals susceptible to T1D, although the observation that 10 of 10 human subjects expressed this difference suggests it might be a common trait. Clearly, validation requires a much larger group of subjects as well as identification of the relevant genes. The possibility of differences in transcriptional regulation of cellular cytokine or cytokine receptor expression in response to infection or lipids need to be considered, as previously suggested (2,3). It will be of great interest to determine whether individuals with one or two autoantibodies exhibit decreased FFA oxidation or enhanced cytokine or FFA sensitivity and whether such sensitivity predicts development of T1D.

Most islet autoantibodies are directed against proteins located within the secretory pathway of pancreatic β-cells. Indeed, most are directly associated with the insulin secretory granule itself (7). In addition to antibody responses, recent studies show that both CD4⁺ and CD8⁺ T cell responses are directed against a number of secretory pathway and granule localized proteins, including chromogranin, PC2, and urocortin 3 (46). Once an individual reaches a diagnosis of stage 3 T1D, immunomodulatory interventions show limited efficacy (47). Thus, the critical events that initiate β-cell–directed immune responses and β-cell destruction may be long past. The hypothesis that we are presenting here focuses on the earliest events that precede overt T1D. Clarifying these early events is important given the success of teplizumab in the trial in stage 2 T1D, which suggests that appropriately timed interventions can delay the onset of disease in certain individuals at high risk (24). Early interventions that improve β-cell health and decrease β-cell immunogenicity may have therapeutic efficacy in similar disease prevention strategies, especially when used in combination with immune-modulatory interventions.

In β-cells, endotoxin or IL-1β potentiates insulin secretion by increasing islet sensitivity to glucose (a left-shift) but does not increase glucose-stimulated insulin secretion (GSIS) (48). This is similar to the relationship we have previously reported demonstrating that as lipid stores increase, sensitivity to glucose is left-shifted while GSIS and insulin stores in islets and clonal β-cells are diminished (49). T1D involves T cell–mediated immune destruction of β-cells. High insulin secretion at basal glucose has been implicated as causing insulin resistance (14,50). Interestingly, there is also evidence that there is some degree of insulin resistance during progression to T1D (51). Insulin resistance and higher BMI have been associated with increased progression to T1D in the TrialNet Pathway to Prevention cohort (52–54). Insulin resistance, possibly caused by basal hypersecretion or impaired lipid metabolism, may play a larger role in T1D that has received little consideration (51,55).

Our hypothesis evolved from efforts to understand the implications of the differences in cytokine and lipid responsiveness between patient and control cells. Like fibroblasts and immune cells, the pancreatic β-cell responds to elevated FFA and inflammatory cytokines with increased ROS generation (56). ROS in the presence of elevated lipid increases trafficking of secretory granules to and from the β-cell plasma membrane to increase membrane capacitance (57) and/or elevate basal insulin secretion (14,58). Indeed, we have shown that basal insulin secretion correlates directly with cellular lipid content under conditions where fuel is excessive and ROS production is high (49). In fact, H₂O₂ alone increases basal insulin secretion in a concentration-dependent manner (58). Furthermore, such increased secretion is accompanied by increased cycling of secretory granule proteins and increased surface presentation of granule proteins to the immune system (Fig. 2). Continued secretion requires recycling of granule proteins that involves S-acylation and deacylation and is linked to and dependent on both lipid availability and redox state (59,60). S-acylation results from the enzymatic addition of long-chain lipids, most typically palmitate, onto intracellular cysteine residues of soluble and transmembrane proteins via a labile thioester linkage. Addition of lipid results in increases in protein hydrophobicity that can impact protein structure, assembly, maturation, trafficking, and function. S-acylation is reversible and controls protein trafficking and protein-protein interactions. We have proposed that S-acylation ultimately targets specific β-cell granular proteins (14), which cycle to the cell surface, for immune destruction. It should also be noted that S-acylation influences many cellular responses in other cell types including cells of the immune system (61). Furthermore, the TNFα receptor is palmitoylated at multiple cysteines, dynamic S-palmitoylation at transmembrane-proximal Cys248 regulates its plasma membrane localization, and depalmitoylation of the activated receptor is essential for NF-κB signaling (61).

A relevant finding in diabetes-prone BB rats, but not in the control groups, is that sustained elevation of circulating FFA significantly decreased β-cell function, an effect that could be blocked by ROS scavenging with N-acetylcysteine (21).

Cytokines, in addition to generating ROS, have been reported to deplete the endoplasmic reticulum (ER) Ca²⁺ stores, an effect that might exacerbate impaired GSIS or protein misfolding (62), and we have found that long-chain acyl-CoA, the activated form of FFA, also decreases cytosolic free Ca²⁺ (63). Such decreased Ca²⁺ would be expected to have a major impact on GSIS but not on basal hypersecretion that occurs without an increase in cytosolic Ca²⁺ (14). Ca²⁺ storage is important for both Ca²⁺ signaling and ER-associated posttranslational protein modifications. This could explain a common finding under conditions where basal secretion is increased, an increase in incompletely processed insulin and a resulting increase in the percentage of proinsulin that is secreted (3).

Decreased Ca²⁺ stores are likely to impact protein processing and GSIS more than basal secretion. It is also relevant to note that elevated rates of secretion and granule recycling can also contribute to the ER stress characteristic of T1D (5). Adaptive mechanisms, initiated by the unfolded protein response, involve reduced translation of misfolded proteins, enhanced translation of ER chaperones to increase the folding capacity of the ER, and degradation of misfolded proteins through ER-assisted degradation (64,65). ER stress is an important mechanism of apoptosis, since low [Ca²⁺]_ER impairs proinsulin processing and transport (66). Notably, lipids and in particular increased FFA in β-cells have been shown to potentiate inflammation and ER stress (66,67).

The sequence of islet β-cell autoantigenicity is not yet well-defined. Autoantigens could be the trigger of autoimmune attack or the collateral damage as part of an ongoing immune destruction of a β-cell. The β-cell homicide versus suicide debate remains unresolved (64). Except for GAD65, it should be noted that the other major T1D autoantigens, (pro)insulin, ZnT8, IA2, and IA2β, are coordinately regulated at the biosynthetic level. When metabolic demand is increased there will be increased production of these antigenic proteins, which in turn increases the chances of misfolding or aberrant processing with production of antigenic peptides. There are likely multiple pathways through which antigenic peptides produced by the β-cell can interact with the immune system. For example, lipotoxicity in cultured β-cells has been linked with impaired ER-to-Golgi trafficking (68), which could divert potential neoantigens through the endolysosomal pathway and release via extracellular vesicles (65). Along these lines, extracellular vesicles from cytokine-stressed human islets are enriched with T1D autoantigens like GAD65 and proinsulin (66). In mouse models, autoantigens can be released from the β-cell via lysosome-generated crinophagic bodies, which can interact with islet resident macrophages. Finally, stress pathways in the β-cells, such as ER stress, have been linked with HLA class I overexpression (67) and thus may contribute to increased presentation of neoepitopes on the surface of the β-cell via MHC class I molecules. So, at least, during the development of T1D, there would be more active β-cells trying to produce and secrete more insulin to meet the demand, but this could also trigger β-cell demise. Thus, the β-cell, in conjunction with autoimmune attack, may be undergoing an assisted suicide (7).

Our findings lead us to generate the testable hypothesize that the induction of autoimmunity is a consequence of one or more major inflammatory events in individuals with susceptible HLA phenotypes plus elevated sensitivity to cytokines and FFA. Illnesses or environmental agents that dramatically increase cytokine production and/or elevate FFA initiate autoimmune destruction in individuals with specific genetic features. Thus, early prevention should be aimed at decreasing elevated lipids and diminishing excessive simultaneous elevation of cytokines or cytokine- and lipid-induced immune cell proliferation.

Variation in lipid handling is normally induced in response to need and is based on dietary intake. A high-fat diet is accompanied by induction of increased mitochondrial lipid-oxidizing capacity, whereas a high-carbohydrate diet decreases lipid-oxidizing capacity. Thus, the absence or inhibition of mitochondrial lipid-oxidizing capacity will decrease lipid metabolism. This suggests that it might be worthwhile to test whether improved lipid handling can be induced, in genetically susceptible individuals with a single autoantibody, by treatment with PPAR agonists that stimulate lipid metabolism, such as fibrates (bezafibrate or fenofibrate) or thiazolidinediones, which are agonists of PPARγ, that have been linked with a number of anti-inflammatory effects as well as modulation of Treg differentiation and Treg responses (69). Fibrates are PPARα agonists that exert anti-inflammatory and antioxidant effects, and protection from mitochondrial dysfunction, that have been shown to improve outcomes in patients with severe sepsis (70). They promote expression of fat-oxidizing enzymes and are widely used to reduce excessive TG and cholesterol levels. Interestingly, fenofibrate reverses and prevents autoimmune diabetes in nonobese diabetic (NOD) mice (71). PPARα is expressed in pancreatic islets and β-cell lines. PPARα expression depends on glucose level (72), with high glucose repressing expression in isolated rat islets and INS-1E cells. The glucose-dependent upregulation of insulin expression might also rely on PPARα, as glucose did not increase insulin expression in islets from PPARα knockout mice. PPARα has likewise been found to upregulate Pdx-1 in INS-1 cells and isolated rat islets (73). On a whole-body level, PPARα knockout mice are normoglycemic in the fed state but hyperglycemic when fasted. This was associated with a 55% higher plasma insulin; however, they had improved glucose tolerance and increased insulin secretion from isolated islets.

Anecdotal support for benefit of inducing increasing lipid-handling capacity was reported in a single case (74,75). A 19-year-old girl was positive for two autoantibodies, diagnosed with T1D, and immediately started on insulin. Fenofibrate treatment (160 mg daily) was initiated 7 days later. The need for insulin quickly declined, and she took her last dose of insulin 19 days after the first dose of fenofibrate, having regained endogenous control of blood glucose concentrations. She was reported to be insulin independent for 1 year and 9 months, and her C-peptide levels increased by 51% and IA-2 autoantibody level decreased by 65%. Importantly, there is an ongoing trial of fenofibrate to improve insulin responsiveness in individuals with recent-onset T1D (https://www.sdcc.dk/english/research/projects/Pages/Fenofibrate-as-an-early-treatment-option-in-newly-onset-type-1-diabetes.aspx). Finally, results of newly published studies of a combination GLP-1–GIP agonist in individuals with type 2 diabetes indicate a dramatic decrease in circulating lipids and lipid stores (76,77) that could also prove beneficial in T1D. Such treatments are predicted to lead to greater use of fat by all cells including β-cells and immune cells and appropriate storage in adipocytes and, if imposed early enough, could decrease or prevent development of T1D if aberrant lipid accumulation plays a causative role.

Another testable implication of the model, in genetically susceptible individuals with one autoantibody, is that it would be beneficial to treat illnesses that cause loss of appetite or vomiting aggressively to maintain blood glucose, prevent excessive stimulation of lipolysis and accompanying increases in circulating FFA, and diminish hyperinsulinemia and insulin resistance.

Other testable and potentially useful approaches based on our model (Fig. 2) include inhibiting granule recycling and lowering ROS while maintaining blood glucose levels using exogenous insulin. Blockade of β-cell cytokine receptors could also be useful in combination with the other approaches in pre-T1D. Finally, aggressive antioxidant treatment, postinfection, if FFA remain elevated, may be protective. These approaches would not preclude the continuation of effective lifestyle interventions including a low-carbohydrate diet and increased exercise (78). It is also possible that the proposed approaches may prove useful for only a subset of individuals or specific disease-related endotypes (79). Future work may allow such a subset to be identified and selectively treated.

Acknowledgments. The authors thank Richard Kibbey, Yale University, for critical reading and suggested clarifications of the manuscript.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.