Monoclonal antibodies that block specific immune checkpoints enable durable anti-cancer response in a subset of patients with previously futile prognoses. Drugs targeting Programmed Cell Death Protein-1 (PD-1), its ligand (PD-L1) and Cytotoxic T-lymphocyte Antigen-4 (CTLA-4) are now considered part of standard of care for treatment of multiple cancers. Both targets represent immune checkpoints that contribute to physiological immune tolerance (as summarized here (1)). Despite remarkable response to PD-1/PD-L1 and/or CTLA-4 blockade in some patients, however, a substantial proportion do not gain benefit and mechanisms underlying individual variation in response are incompletely understood.

Immune-related adverse events (irAEs) are autoimmune sequelae of ICI therapy that can affect any organ in the body. Multiple endocrine irAEs including thyroiditis, hypophysitis, diabetes mellitus and adrenalitis, have been recognized with the use of ICI. Endocrinopathies are amongst the most common irAEs. Partly due to the large reserve capacity of many endocrine organs, they are distinct in their tendency to be diagnosed at the point of organ failure. In contrast, most non-endocrine irAEs are detected based on early clinical indications of inflammation, such as dyspnea or diarrhea. As such, the management of endocrine irAEs typically includes life-long hormone replacement (2). While non-endocrine irAEs are often treated with glucocorticoids, there is no clear evidence that immunosuppression is beneficial in preventing progression of endocrinopathy (2).

Successful checkpoint inhibition in cancer treatment requires a fundamental shift in immune homeostasis to a state of reduced immune tolerance. ICI therapy is thus inherently linked to irAEs, because lowering immune tolerance to cancer inextricably lowers tolerance to self. Through this lens, manifestation of irAEs may be a marker of successful immune activation. Accordingly, the occurrence and severity of irAEs, particularly those of endocrine origin, have been linked to improved cancer outcomes in studies that have corrected for immortal time bias in several tumors (3–6).

Significantly, many irAEs are clinically and pathologically indistinguishable from autoimmune disease associated with inherited variation in relevant checkpoint genes. Penetrance of irAEs following pharmacologic blockade of a checkpoint, however, is not equivalent to that of primary genetic blockade of the same gene (7, 8). Intriguingly, pharmacologic blockade in some cases causes more frequent endocrinopathy than genetic blockade, and in some cases less ( Table 1 ). Clinical autoimmunity may manifest when acquired mutations layer on inherited mutations, stochastically bypassing tolerance to precipitate disease (19–22). Pharmacologic blockade of a checkpoint (such as PD-1) may be equivalent to an acquired functional genetic mutation, serving as an additional ‘switch’ weighing in favor of autoimmunity. Improved understanding of the mechanism for drug-induced irAEs and their similarities and differences to genetic autoimmune disease, therefore, has potential to improve patient selection and cancer outcomes.

In this review, we explore physiological mechanisms of immune tolerance that are leveraged in cancer immunotherapy and examine ICI-induced toxicity and analogous genetic disease, with a focus on endocrinopathies and host contribution to irAE response.

Understanding tolerance in different cell types and physiological settings is an important foundation to understanding ICI response and toxicity. The immune system balances a challenging equilibrium of mounting appropriate response to pathogens whilst tolerating self and non-self commensals. Differing mechanisms of tolerance induction and maintenance for T and B cells, in central (bone marrow, thymus) and peripheral (spleen, lymph nodes) lymphoid organs, are summarized below. The majority of cancer immunotherapy research to date has focused on the role of T-cells, however B-cells play an important role in analogous primary autoimmune disease. Furthermore, auto-antibodies are a feature of many endocrine-related irAEs and may be key to understanding pathogenesis.

“Successful” cancer cells achieve unchecked proliferation by exploiting physiological mechanisms of immune tolerance to evade detection and/or destruction by the host immune system. These include downregulation of antigen-presenting machinery including MHC-1 (51), overexpression of immune-inhibitory molecules such as PD-L1 (52), induction of an immunosuppressive tumor microenvironment enriched for regulatory T cells (Tregs) and myeloid-derived suppressor cells (53–55) and secretion of directly immunosuppressive factors such as IL-10, Transforming Growth Factor-β (TGF-β), gangliosides, prostaglandin E2, and vascular endothelial growth factor (VEGF) (56–60).

Tregs play a critical role in tolerance to cancer cells through suppression of cytotoxic CD8+ T-cell proliferation, contributing to immune escape. Tregs express high levels of multiple checkpoint receptor molecules including CTLA-4, PD-1, T cell immunoreceptor with Ig and ITIM domains (TIGIT), Lymphocyte-Activation Gene 3 (LAG-3) and T-cell immunoglobulin domain and mucin domain 3 (TIM3) (61, 62).

Monoclonal antibodies targeting CTLA-4 and PD-1/PD-L1 are now considered part of standard of care in multiple cancer types, including melanoma, Non-Small Cell lung cancer (NSCLC), renal cell carcinoma, head and neck cancer, urothelial cancer, Merkel cell carcinoma, mesothelioma, and Hodgkin lymphoma.

CTLA-4 is a negative co-inhibitory receptor of T-cell activation and proliferation, playing a critical role in the maintenance of self-tolerance (63). It is constitutively upregulated in Tregs and transiently upregulated on the cell surface of conventional T cells following the engagement of co-stimulatory molecule CD28 with its ligands, CD80 and CD86, in order to curtail T cell responses (64). CTLA-4 has several mechanisms of action including competitively binding CD80 and CD86 and reducing CD28-dependent co-stimulation (65–67). Additionally, CTLA-4 is thought to strip the CD80 and CD86 ligands from APCs through a process of trans-endocytosis, further reducing potential for CD28 co-stimulation (65). It has been postulated that CTLA-4 may also directly restrain T-lymphocytes through down-regulating IL2 expression, as well as repression of T-cell proliferation (68).

PD-1 is a co-inhibitory receptor expressed on the surface of activated T-cells (including tumor infiltrating lymphocytes), B cells, NK cells, monocytes and dendritic cells (64, 69, 70). Ligation of PD-1 with its ligands, PD-L1 and PD-L2, activates inhibitory signaling cascades associated with exhaustion, resulting in the downregulation of T-cell response (69). PD-1 suppression of T cell activity occurs primarily within the peripheral tissues and tumor microenvironment (71).

The incidence and pattern of endocrine irAEs differs significantly between the major classes of checkpoint inhibitors, reflective of differing mechanisms of action ( Table 1 ). ICI-induced thyroiditis is more frequent after PD-1 compared with CTLA-4 blockade (10-20.4% versus 1.5-9%). Whereas ICI-induced hypophysitis and colitis are more common following CTLA-4 compared with PD-1 inhibition (2.3-18% versus 1.2-2.2% and 29-45% versus 3.9-25% respectively). Combination checkpoint blockade, targeting both CTLA-4 and PD-1, increases rate and severity of irAEs, with tendency to earlier events and multi-organ autoimmunity (72, 73).

Accurate predictive models for type and severity of irAEs are lacking, although several potential biomarkers are being studied involving immune cells (74, 75), cytokines (76), autoantibodies (77), immunogenetics (78, 79) and microbiome (80, 81). Time of onset is also unpredictable; most irAEs develop within the first few weeks following commencement of an ICI, however they can manifest years after treatment initiation (82). IrAEs occur in up to 80% of patients receiving combination treatment (83), and are predominantly mild (CTCAE grade 1-2) (84), with less than 1% resulting in death (85).

Endocrinopathies are a subset of irAEs that affect endocrine organs, including thyroiditis, hypophysitis, diabetes mellitus and adrenalitis (86). Distinct to non-endocrine irAEs, endocrine-related irAEs are frequently irreversible, do not usually prompt treatment discontinuation and often require lifelong hormone replacement therapy (87). While numerous studies have examined the relationship between irAEs and treatment outcomes, most do not account for immortal time bias; an adjustment for shorter time on drug and less time available to develop irAEs for non-responders. Landmark analyses avert this bias and also support a correlation of irAEs with outcomes, particularly those involving endocrine organs and skin for patients with NSCLC and melanoma (3–6). The correlation of irAEs with survival in many settings suggests a possible shared driver of systemic and anti-cancer immune activation.

Pharmacological inhibition of CTLA-4 and/or PD-1/PD-L1 has significant overlap, but also important differences, to autoimmune manifestations seen in humans with inborn errors of the corresponding genes, CTLA4 and PDCD1 ( Table 1 ).

Mice homozygous for ctla-4 loss of function develop lymphoproliferative symptoms in multiple organs and die within weeks, but mice with ctla4 haploinsufficiency (one functional allele retained) remain healthy (117–119). In humans, inherited genetic heterozygosity for CTLA4 inactivating mutations results in immunodeficiency and autoimmunity (9, 10, 117, 120). In the largest study of CTLA4 haploinsufficiency in humans, 133 patients were characterized (10). The median age of onset of disease was 11 years and there was clinical penetrance (manifestation of disease) in 67% (10). The predominant clinical phenotypes included hypogammaglobulinemia (84%), lymphoproliferation (73%) and organ-specific autoimmunity such as hematological autoimmune cytopenia (62%), respiratory (68%), and gastrointestinal (59%) or neurological manifestations (29%) (10) ( Table 1 ). Some auto-immune sequelae of germline CTLA-4 deficiency are almost indistinguishable from toxicity following its pharmacologic blockade, such as gastrointestinal inflammation (10, 12, 13). One factor contributing to observed differences may be distinct temporal dynamics between in-born genetic loss of CTLA-4 compared with acquired pharmacologic blockade, usually later in life. Mouse models of conditional CTLA-4 inhibition later in life, however, do not more closely resemble anti-CLTA-4 induced irAEs, suggesting other differences at play (121).

Key discrepancies between genetically and pharmacologically driven disease highlight the need to further understand both pathways. Firstly, clinical apparent autoimmunity, in general, appears to be more prevalent in humans with CTLA-4 haploinsufficiency compared with patients receiving CTLA-4 pharmacological blockade (10, 12, 13). This supports the notion that irAEs associated with CTLA-4 blockade are dose-dependent, and the monoclonal antibodies may only partially inhibit CTLA4 function leading to incomplete constellation of symptoms. Indeed, severe, life-threatening autoimmune toxicities are more frequent with higher doses of ipilimumab supporting this dose-dependent effect (122). Secondly, and somewhat counter to this, certain organ toxicities are significantly more common with CTLA-4 drug blockade compared with CTLA-4 germline haploinsufficiency. For example, severe hypophysitis is reported in up to 5% of patients treated with ipilimumab but is extremely rare (<1%) in genetic CTLA4 haploinsufficiency (10, 13) ( Table 1 ). Finally, certain manifestations of primary CTLA4 deficiency are not seen in the drug-induced setting, significantly immunodeficiency, which is contributed to by autoantibody mediated cytopenia, lymphadenopathy and splenomegaly (9, 120, 123). The reason for this discord is not well understood.

In contrast to CTLA4, germline deficiencies in PD-1 more frequently lead to single organ effects, most commonly involving endocrine organs. Here, exogenous blockade appears to more closely match germline inactivation (124–126). Mice deficient in PD-1 initially develop normally but go on to manifest strain-dependent autoimmunity and heightened inflammation during infections (127–129). PD-1 deficiency also accelerates the onset and frequency of organ-specific endocrine autoimmunity such as T1DM in predisposed mouse models (130).

In humans, homozygous loss of function mutations in the gene encoding PD-1, PDCD1, is exceedingly rare (15). In one case study, homozygous frameshift (loss of function) mutation resulted in lymphoproliferative autoimmunity, autoimmune thyroiditis, T1DM, juvenile idiopathic arthritis, and abdominal tuberculosis. The patient died of pneumonitis at age 10 years (15). Though complete loss of PDCD1 is rare in humans and difficult to study, partial deficiency shows significant overlap with drug-induced irAEs ( Table 1 ).

Further investigation is needed to explore similarities and discrepancies between genetic and drug-induced autoimmunity. As discussed above, timing and duration of exposure to pathway blockade is likely to play a role. Inherited genetic abnormalities may only manifest as autoimmunity when accumulation of acquired mutations leads to a break in tolerance (20). Treatment with anti-CTLA-4 and/or anti-PD-1 can be considered akin to an acquired mutation in terms of onset later in life, however is distinct from an acquired mutation in that blockade is maintained for a fixed period only. Comparison with analogous genetic disease is valuable because it can help elucidate mechanism, however genetic predisposition is likely to be just one of many factors contributing to outcome. Environmental factors including the microbiome are likely to play a major role (80, 81), but are beyond the scope of this review.

Recent genetic association studies have revealed that polymorphisms in CTLA4 may modulate both autoimmunity and response and toxicity to ICIs (131, 132). Specific gene associations with autoimmune disease are outlined in Table 2 . Many of these Genome-wide association studies (GWAS) single nucleotide polymorphisms (SNPs) highlight the dialectic nature of T-cell immune regulation in cancer and autoimmunity - the allele for cancer risk is often the opposite to that for autoimmune disease (132, 134, 157, 158, 166–170). Systemic testing for common autoimmune risk loci could provide a pathway for personalized biomarkers for ICI response and toxicity (171).

Germline polymorphisms in CTLA4 have also been shown to influence susceptibility to irAEs. Refae et al. investigated a cohort of 94 patients with advanced cancer treated with anti-PD-1/PD-L1 and demonstrated that SNPs within a range of genes including UNG, IFNW1, CD274 and IFNL4 were predictive of irAEs corresponding to that gene function (78). Whole genome sequencing of 479 patients with bladder cancer treated with anti-PD-L1 therapy showed that high polygenic risk scores for vitiligo and psoriasis were associated with increased risk of skin-related irAEs (7).

A key question is whether the genetic risk for classical T1DM overlaps with ICI-DM. Polymorphisms of multiple genes are known to influence risk of T1DM, with the HLA-DQ, HLA-DR alleles having the largest effect, but also non-HLA genes such as PTPN22 and CTLA-4 (172). HLA polymorphisms linked to classic T1DM class II haplotypes – HLA-DR3-DQ2 and DR4-DQ8 in European populations, and DR4-DQ4 and DR9-DQ9 in Asian populations, have been demonstrated to be overrepresented patients with ICI-DM (106). Only one ICI-DM case has been evaluated for multiple HLA and non-HLA risk genes as part of a T1DM genetic risk score (GRS); this case had a GRS below the 5^th percentile for T1DM (173), suggesting that ICI-DM genetic risk factors may be distinct from T1DM. Overall, HLA susceptibility haplotypes for T1DM may predispose to the development ICI-DM, but further studies are needed to understand the genetic contribution of HLA and non-HLA genes.

There is also known genetic susceptibility to spontaneous AITD, within which certain at-risk genes may also be certain at-risk genes may also be shared with ICI-induced thyroiditis. Data suggests there is an association of AITD with certain HLA alleles, particularly HLA-DR3, and with polymorphisms within CTLA4 and CD40 (174). GWAS have demonstrated key susceptibility genes implicated in both immune regulation and increased risk of ICI-induced thyroiditis, including CD69, CTLA4, PTPN22 and LPP (96). These loci and others were combined to calculate a polygenic risk score that could identify a cohort of patients with more than 6-fold increased risk of ICI-induced thyroiditis. Furthermore, germline susceptibility to AITD disease was found to be associated with improved survival, in a cohort of patients with triple-negative breast cancer treated with combination anti-PD-L1 inhibitor and chemotherapy (96).

Immune checkpoint inhibitors are now established as a pillar of cancer treatment, alongside chemotherapy, radiotherapy and surgery. Understanding mechanism of response and non-response to the drugs is paramount to broadening their potential for clinical benefit. Investigating drivers of ‘off-target’ effects, specifically irAEs, is key to this analysis, because anti-self responses are inextricably linked with anti-cancer responses in many settings due to lowering of immune tolerance thresholds. Furthermore, many irAEs are clinically and biologically analogous to genetic disease, for which we have significant insight into drivers and mitigating factors.

Autoimmunity of endocrine organs are common both as primary disease and as ICI-induced toxicity in patients treated for cancer. Endocrine autoimmunity driven by genetic variation in checkpoint genes can mimic drug-induced irAEs, however key differences exist in the pathophysiology for each. Both primary and drug-induced endocrine disease may involve underlying genetic susceptibility as part of a multifactorial pathogenesis. Further investigation into shared drivers of irAEs with analogous primary disease may lead to predictive models for genetic predisposition to ICI-induced toxicity. Furthermore, the correlation between irAEs and anti-cancer response highlights the potential for these genetic drivers to serve as novel therapeutic targets, with the caveat that ensuing toxicity may need aggressive management.

Current predictive markers (such as tumor and immune cell PD-L1 expression) can enrich for responders to immunotherapy but provide little insight into variation in cancer response or toxicity for the individual patient. There is increasing recognition that host factors, including genetic predisposition to both autoimmunity (for example to thyroid disease) and irAEs (for example to psoriasis and vitiligo), can be predictive of checkpoint inhibitor treatment outcomes. Genetic testing for common autoimmune risk loci could form part of a comprehensive personalized biomarker for ICI response and toxicity, leading to improved treatment selection and toxicity management for individual patients.

There is much to learn from host factors that contribute to maintenance and breakdown of physiological immune tolerance. Shared drivers of clinical autoimmunity and ICI-induced irAEs are pivotal to this analysis and have the potential to progress patient outcomes in cancer treatment significantly.

AC, IA, MB and DB jointly wrote the manuscript. All authors contributed to the conception, design, and interpretation of the manuscript, and reviewed and wrote the final paper.

This work was supported by National Health and Medical Research Council (NHMRC) (Grants 10375), and by Bioplatforms Australia and The Kinghorn Foundation as part of the Australian Exceptional Responders Program (Grants PR10285).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.