Genetic risk analysis of a patient with fulminant autoimmune type 1 diabetes mellitus secondary to combination ipilimumab and nivolumab immunotherapy

The exact mechanisms by which immune checkpoint inhibitors induce autoimmunity and cause irAEs are not fully understood. Recent research on anti-CTLA-4 associated hypophysitis suggests that local expression of CTLA-4 in the pituitary leads to antibody binding and activation of the complement pathway with subsequent site-specific inflammation and tissue injury [17]. However, this work is observational, and the exact pathophysiology of other irAEs, including the possible role of complement-mediated injury, remains elusive. It is suspected that checkpoint inhibitors interfere with peripheral tolerance and potentiate autoimmune disorders in genetically predisposed individuals [5]. While the occurrence of irAEs in an individual on combination checkpoint inhibitor therapy is expected, the dramatic presentation of fulminant T1D and its potential for other complications is notable and warrants further consideration.

Autoimmune diabetes is characterized by the development of an adaptive immune response against specific ?-cell antigens [18]. Longitudinal studies in patients have shown that certain autoantibodies, such as anti-insulin (IAA), anti-islet cell antigen 512 (ICA512), and anti-glutamic acid decarboxylase (GAD65), define preclinical disease as they are present in the serum for years prior to symptom development [19]. They have been shown to be markers heralding a disease state but do not directly mediate disease progression. This is also true for patients with latent autoimmune diabetes mellitus in adults (LADA), which has a slow progression to insulin dependency that may be predicted by the presence of islet autoantibodies [20]. Certain cases of T1D are characterized as fulminant if the patient presents with diabetic ketoacidosis soon after the onset of hyperglycemic symptoms, has a near normal HbA1c, and has a low C-peptide level [21]. These patients typically do not have autoantibodies, the absence of which is thought to be indicative of the rapid nature of the disease pathogenesis [22]. In the case presented here, the concurrent development of both anti-GAD65 antibodies and insulin dependency is highly atypical. Indeed, to our knowledge, it is the first reported case of documented seroconversion temporally associated with immune checkpoint inhibitor therapy.

The clinical presentation of T1D is due to the lack of endogenous insulin secondary to the destruction of insulin-producing ?-cells in the pancreas. This process is attributed to infiltration of the pancreatic islets by autoreactive T cells, causing ‘insulitis’ [23]. The most widely used model to study this process is the non-obese diabetic (NOD) mouse. In this population, insulitis and overt diabetes spontaneously develop at 20 to 30 weeks of age depending on gender. The NOD mouse model has been used to assess the relationship between disease progression and various alleles related to immune regulation, including the pathways targeted by checkpoint inhibitor therapies, PD-1 and CTLA-4 [24]. Studies in this murine model have implicated a cooperative relationship between these negative regulatory receptors in the pathogenesis of T1D via immune dysregulation [25, 26]. CTLA-4 has a role in modulating the survival and function of the regulatory T cell population, which is responsible for suppressing autoreactive T cell activation. The importance of CTLA-4 in regulating autoimmunity was demonstrated early on in CTLA-4 knockout mice, in which CD4-predominant lymphoproliferation develops and results in T cell infiltration of multiple organs and early mortality at 3–4 weeks [27]. Indeed, studies by Ansari et al. demonstrated that CTLA-4 blockade in NOD mice induces T1D. However, blockade of this pathway was noted to only induce diabetes in neonates and not in adult mice, suggesting that the role of CTLA-4 may be limited to inhibiting the activation of naïve T cells [26]. In contrast, PD-1 blockade precipitates diabetes in NOD mice at any age. This is likely due to PD-1 involvement in inhibiting both naïve T cells as well as the effector function of activated autoreactive T cells [26]. Another role of PD-1 in regulating peripheral tolerance for these autoreactive T cells is to modulate T cell mobility. Fife et al. demonstrated that blockade of PD-1 in pre-diabetic NOD mice inhibits T cell migration, thus promoting the formation of stable conjugates between T cells and antigen presenting dendritic cells by prolonging their interactions. The authors concluded that this results in increased T cell activation and precipitates diabetes in NOD mice [28]. The protective effect of PD-1 has also been suggested in humans as individuals with T1D have significantly lower levels of CD4⁺ T cell PD-1 expression compared to healthy controls [29]. Taken together, these studies indicate that the CTLA-4 and PD-1 pathways function at different stages in the development of T cell tolerance to prevent autoimmune diabetes [30].

This cooperative relationship has been further supported by studies demonstrating that the risk of T1D is directly related to the degree of genetic predisposition. Work by Kochupurakkal et al. in NOD mice demonstrated that specific strains are protected from developing T1D after PD-1 blockade by having functional alleles for Il2 and Ctla4. It was speculated that the functional Ctla4 allele with the help of IL-2 was sufficient to maintain self-tolerance and prevent T1D [25]. While the exact mechanism remains unclear, similar associations between the risk of developing T1D and variations in alleles for CTLA-4 and PD-1 have been identified in multiple human populations [31–35].

Another genetic predisposition for developing T1D, and perhaps the most widely studied, relates to the major histocompatibility complex (MHC) and the associated human leukocyte antigen (HLA) molecules. HLA class I and class II molecules are responsible for presenting endogenous and exogenous antigens, respectively, to T cells and initiating the immune response. Specific variations in HLA-I and HLA-II molecules are associated with increased risk of or protection against the development of T1D [36]. The incidence of T1D in association with checkpoint inhibitor therapy has been previously reported as part of a larger case series, and a few reports have included HLA typing of the affected individuals [10–15, 37]. Of the reported cases of immunotherapy-associated T1D included in Table 1, HLA typing was performed on eight individuals. Six of these individuals expressed the HLA-II DR4 haplotype, which is a well-known risk allele for T1D with an odds ratio (OR) of 5.68 for developing the disease [36, 38]. Our HLA typing of this patient identified the HLA-I A2 and HLA-II DQB1*0602 alleles with no evidence of a previously characterized HLA risk allele for T1D. HLA-II DQB1*0602 is associated with one of the most protective haplotypes for developing T1D with an OR of 0.03, though this protective effect seems to be restricted to childhood [38, 39]. We conclude that the HLA typing for this patient is not consistent with an immunologic predisposition to the development of T1D.

Diabetic autoantibodies referenced include GAD65, ICA5, and insulin (IAA). Normal GAD65 titers 0.5 U/ml, ICA5 1.0 U/ml, IAA 5.0 U/ml

In order to investigate this patient’s genetic risk for developing T1D in more detail, we also conducted a SNP analysis of 26 different loci previously associated with the development of T1D and calculated a genetic risk score (GRS) emulated from Oram et al. [16]. This GRS summarizes risk-associated variation across the genome. In this case, the patient’s GRS was 0.2072, a score that is below the 5^th percentile of the T1D cohort [16]. These findings indicate that this patient did not have a genetic profile consistent with any known predisposing factors for the development of T1D.

This case demonstrates that combination ipilimumab and nivolumab immunotherapy is capable of inducing the development of T1D even in those patients with no discernable risk factors for this disease. These findings highlight the importance of the PD-1 and CTLA-4 negative regulatory T cell receptors in the pathogenesis of T1D and suggest that dual checkpoint blockade may be unleashing the activation of previously existing islet-reactive T cell clones in healthy individuals. Given that all reported cases of checkpoint inhibitor-related T1D have been associated with anti-PD-1 antibody therapy (Table 1), it seems likely these islet-reactive T cell clones reside in or near pancreatic ?-islet tissues or their associated draining lymph node tissues prior to therapy.

In summary, multiple studies have implicated both the CTLA-4 and PD-1 pathways in the pathogenesis of T1D and suggest a synergistic relationship between these two negative regulatory receptors to potentiate autoimmune disorders. Kochupurrakal et al. concluded that the combined blockade of the CTLA-4 and PD-1 pathways poses a risk of disrupting peripheral tolerance and generating T1D [25]. This hypothesis has been supported by studies of concurrent nivolumab and ipilimumab administration for the treatment of advanced melanoma, in which the incidence of grade 3 or 4 irAEs was higher than in trials of either agent alone [40]. While the HLA-II DR4 haplotype has been previously reported in cases of immunotherapy-associated T1D (Table 1), whether these risk alleles predict for the development of irAEs in patients undergoing checkpoint inhibitor immunotherapy remains unclear. Given the lack of association of any known genetic risk allele for T1D in the case presented here and the rarity of this complication in patients undergoing immunotherapy, the use of established genetic and immunologic screening studies for T1D prior to initiating a patient on checkpoint inhibitor immunotherapy may not be indicated. However, this represents a single case report and more definitive studies will be necessary to address this issue. Finally, this study suggests that additional investigation is needed to determine whether T cell receptor clonotypic analysis to identify the presence of either peripheral blood or tissue-resident autoreactive T cell clones may be an effective approach for predicting which patients are at increased risk for developing autoinflammatory toxicities while undergoing checkpoint inhibitor immunotherapy.