At the Heart of Immune Checkpoint Inhibitor–Induced Immune Toxicity

Immune checkpoint inhibitor (CPI) use in patients with cancer has been met with tremendous excitement due to their ability to disrupt immune-suppressive mechanisms active within the tumor microenvironment, leading to potent antitumor immune responses. However, at homeostasis, immune checkpoints, including PD-1 and CTLA4, are essential in protecting against self-reactivity by restricting potentially autoreactive CD8⁺ and CD4⁺ effector T cells while enhancing the suppressive function of regulatory T cells. As a consequence, CPIs can break immune tolerance, causing autoimmune and autoinflammatory immune-related adverse events (irAE) that can precipitate tissue damage in multiple organs. This represents a major clinical challenge in the utility of CPIs, which is only magnified through the increasing use of combination immunotherapies.

Translational studies of irAEs in humans are hampered by limitations in accessibility to tissue as well as difficulty in obtaining samples at meaningful time points due to the variability and rapidity in timing of these wide-ranging events. As an example, CPI-induced myocarditis is a rare occurrence in CPI-treated patients with cancer and thus extremely difficult to study. Preclinical mouse models have been essential in dissecting the genetics and pathogenic mechanisms of immune-mediated diseases and may enable understanding of therapeutic modalities that may moderate irAEs while maintaining antitumor immunity (Fig. 1; ref. 1). In this issue, Wei and colleagues report on the establishment of a genetic model of CPI-induced myocarditis by disrupting PD-1 and CTLA4 expression (2). The lymphocytic infiltrate in the heart as well as the clinical pathology observed in these mice mimics features identified in CPI-induced myocarditis that develops in patients with cancer. This animal model, particularly if it can be combined with appropriate tumor models, may provide an important opportunity to (i) predict which patients with cancer are susceptible to myocardial irAEs, (ii) identify the mechanism(s) by which individual myocardial irAEs are initiated, and (iii) determine targeted strategies that mitigate immune toxicity without affecting antitumor immune responses.

In the present study, the investigators used a genetic approach to investigate the role of immune checkpoints in irAEs. Although mice with PD-1 deficiency survive, the degree of autoimmune manifestations is limited and often mild depending on the background of each individual mouse strain. In sharp contrast, CTLA4 disruption results in massive systemic autoimmunity and is uniformly lethal within weeks of birth in all mouse strains. Thus, to mimic the appearance of clinical irAEs in the combinatorial CPI setting, the investigators tested the effect of CTLA4 haploinsufficiency, which is known to impair CTLA4 expression and function. The studies identified that a single-allelic loss of CTLA4 in C57BL/6 PD-1–deficient mice resulted in immune-mediated myocarditis (2). This is the first description of a pathologic role for CTLA4 haploinsufficiency in mice and mimics observations in humans where increased rates of CPI-induced myocarditis are observed in patients with cancer receiving a combination of CTLA4 and PD-1/PD-L1 inhibitors.

The requirement for both PD-1 deletion and CTLA4 haploinsufficiency highlights the nonredundant functionality of these immune checkpoint pathways. CTLA4 inhibitors have been suggested to expand the breadth of antigen-reactive T cells by lowering the threshold for T-cell activation, whereas PD-1/PD-L1 antagonists reverse the dysfunction of inactive T cells residing in both tumor and tissues. In combination, targeting these pathways has been shown to promote both improved tumor responses and the initiation of more frequent and severe irAEs. In fact, in two cases of CPI-induced myocarditis, a clonal expansion of T cells with shared T-cell receptors was identified in both tumor and cardiac muscle following ipilimumab (anti-CTLA4) and nivolumab (anti–PD-1) treatment (3). In addition, PD-L1 expression was enriched in diseased myocardium infiltrated with T cells, which suggests that a tissue-resident suppressive mechanism was unleashed by anti–PD-1 treatment (3). These mechanistic insights into CPI-induced myocarditis are consistent with a synergistic role of CTLA4 and PD-1 in tissue protection.

It is important to note that not all CTLA4-haploinsufficient, PD-1–deficient mice succumbed to cardiac disease (2). This incomplete disease penetrance is similar to the clinical experience and suggests that additional triggers are involved in the transition from subclinical to clinical disease. In this regard, both the animal model and patients with cancer developing CPI-induced myocarditis exhibited increased penetrance in females, a trait that is not observed in conventional myocardial autoimmune disease or other irAEs. These penetrance differences suggest that this model may be useful in modeling sex biases and other environmental differences observed in both CPI-induced myocarditis and other autoimmune settings. Finally, a small number of CPI-induced myocarditis cases develop in response to single-agent PD-1/PD-L1 treatment in humans. It is unclear whether these patients display a lower threshold of CTLA4 expression that may be equivalent to CTLA4 haploinsufficiency or blockade.

Developing biomarkers that identify preexisting susceptibility or portend the development of CPI-induced irAEs is a growing area of research inquiry. Interrogation of serum-based changes to cytokines, autoantibodies, and modulation of clinical laboratory values have been shown to be associated with both severe tissue-agnostic irAEs and tissue-specific irAEs and may enable increased clinical surveillance. CTLA4 and PD-1 disruption in mice resulted in increased serum concentrations of troponin I associated with intracardiac immune infiltrate (2). Similar observations have been seen in CPI-treated patients developing myocarditis, wherein circulating concentrations of troponin were increased at the time of diagnosis of cardiac malfunction (4). It is unclear whether increased troponin levels precede the initiation of disease as it is not regularly assessed during treatment of patients prior to symptom onset. However, troponin could represent a useful clinical biomarker to determine the prodrome of CPI-induced myocarditis and provide clinicians with greater awareness of organ dysfunction.

Given that CTLA4 disruption was the trigger for disease in this setting, Wei and colleagues tested whether abatacept (CTLA4–Ig) was able to suppress cardiac immune toxicity (2). Abatacept is a soluble fusion protein of the CTLA4 molecule that effectively blocks the interaction of the costimulatory receptor CD28 with its ligands, CD80 and CD86, and is approved or undergoing clinical trials for the treatment of several autoimmune diseases, including in CTLA4 haploinsufficiency to reduce cytopenia (NCT03733067). Abatacept treatment in both CTLA4-haploinsufficient PD-1–deficient mice and CPI-treated patients with cancer with myocarditis prolonged survival and normalized clinical laboratory markers of cardiac inflammation, respectively (2). Due to the fulminant nature of CPI-induced myocarditis, it is critical to have highly efficacious therapeutic options, such as abatacept, to resolve this disease. It is conceivable that resumption of single-agent PD-1/PD-L1 inhibitors may be possible following abatacept without reinitiating CPI-induced myocarditis. In addition, abatacept may not antagonize antitumor immunity in settings in which there is complete tumor remission, where the priming and expansion phase of the T-cell response initiated by anti-CTLA4 treatment has been maximized and could potentially be considered in other rheumatologic irAEs similar to its use in spontaneous autoimmunity.

The ability of abatacept to disarm the immune system remains a concern with regard to the antitumor immune response initiated by combination immunotherapies. This highlights the need for preclinical models that assess both immune-mediated axes, simultaneously. In fact, in terms of next steps, the maximal clinical utility of the animal model described in this study will be realized only when combined with a tumor model. There is increasing evidence suggesting that the presence of irAEs may be connected to improved survival outcomes for CPI-treated patients with cancer (5), although the relationship between CPI-induced myocarditis and clinical efficacy remains undefined. In considering therapeutic options to uncouple the immune response toward irAEs and the tumor, autoimmune-prone preclinical models that have syngeneic tumor models available will be very valuable (Fig. 1; refs. 1, 6–8). Similarly, preclinical models that initiate irAEs in response to immunotherapies or inducible gene-targeted systems rather than genetic ablation may be more representative of clinical conditions, as they will provide greater control over the initiation of irAEs alongside different phases of tumor initiation, growth, and metastasis.

The results from Wei and colleagues' study highlight the connectivity of mouse models with clinical attributes of irAEs and demonstrate their utility to both predict and determine the best treatment course for patients with cancer. There is a range of irAEs that can develop in patients with cancer, which emphasizes the need for a suite of preclinical models to examine individual aspects of both immune toxicity and antitumor immunity in response to genetic and therapeutic manipulation of immunomodulatory pathways (Fig. 1; refs. 2, 6–11). Moreover, animal models of CPI-induced autoimmunity may provide new insights into the mechanism for conventional human autoimmune diseases, enabling the development of new therapeutic modalities poised for clinical success.

J.A. Bluestone reports personal fees and other from Sonoma Biotherapeutics, Gilead Biosciences, Vir Biotechnology, Arcus Biosciences, and Provention Bio, and grants from NIAID and NIDDK outside the submitted work. A. Young was supported by the NCI under Award Number K99CA246061. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. In addition, A. Young was supported by a NHMRC C.J. Martin Fellowship (GNT1143981) and the Parker Institute for Cancer Immunotherapy.