To perform the systematic review of the immune mechanisms associated with ICIs-induced irAEs, for CD4⁺ and CD8⁺ T cells recruitment in ICIs therapy, we performed a PubMed search (until December 2020) with the next set of keywords: “*X* and CD4 and CD8, not review, not clinical trial,” where *X* represents anti-CTLA-4, or ipilimumab, or tremelimumab, or anti-PD-1, or nivolumab, or pembrolizumab, or cemiplimab, or libtayo, or anti-PD-L1, or atezolizumab, or durvalumab, or avelumab.

Developing the following complete keywords: “anti-CTLA-4 and CD4 and CD8, not review, not clinical trial,” “ipilimumab and CD4 and CD8, not review, not clinical trial,” “Tremelimumab and CD4 and CD8, not review, not clinical trial,” “anti-PD-1 and CD4 and CD8, not review, not clinical trial,” “Nivolumab and CD4 and CD8 not review, not clinical trial”, “Pembrolizumab and CD4 and CD8, not review, not clinical trial,” “Cemiplimab and CD4 and CD8, not review, not clinical trial,” “Libtayo and CD4 and CD8, not review, not clinical trial,” “anti-PD-L1 and CD4 and CD8, not review, not clinical trial,” “Atezolimumab and CD4 and CD8 not review, not clinical trial,” “Durvalumab and CD4 and CD8 not review, not clinical trial,” “Avelumab and CD4 and CD8, not review, not clinical trial”. Additionally, the keywords to complete the search were added: ((*X* and CD4 and cancer [TIAB] OR *X* and CD8 and cancer [TIAB]) OR *X* and CD4 and CD8 [TIAB] and cancer NOT review [Publication type]), *X* are the same terms as described before.

To achieve the systematic review of mechanisms including the development of autoantibodies in ICI-mediated toxicity, the following set of the keywords were used: “*X* and autoantibody, not review, not clinical trial” or “((*X* and autoantibody and cancer [TIAB] NOT review [Publication type]),” where *X* represents the same terms as described before. To perform the systematic review of cytokine-mediated ICIs toxicity, the following set of the keywords were used: “*X* AND cytokines NOT review NOT clinical trial” or “((*X* and Cytokine and cancer [TIAB] NOT review [Publication type]),” where *X* represents the same terms as described before.

Titles and abstracts were screened, if the study complied with the inclusion criteria, then the full text, study details, and results were screened. The exclusion of studies from the systematic search was carried out if any of the following terms contained concomitant immunotherapies (such as dendritic cells (DC) cells, cytokines, oncolytic viruses, immunogenic cell death (ICD) inducers, chemotherapies, radiotherapies, resection, adjuvants (as pattern recognition receptors PRR agonist), vaccines, additional infections, additional antibody-based therapies (not related with CTLA-4, PD-1 or PD-L1 blockage), natural compounds, peptides, and articles not written in English. Additional exclusions included clinical trials, review articles, as well as in vitro or in silico studies, or without information of CD4⁺, CD8⁺, cytokines, or antibodies in the abstract.

After duplicate elimination, PubMed articles were analyzed, identifying 340 scientific studies containing T-cell recruitment, 135 studies containing autoantibodies development, and 662 studies containing cytokine production after ICI therapies. The systematic procedure is schematized in Supplementary Figure S1. A total of 160 studies describing irAEs were found, and the results are summarized in the following sections.

Several studies have described CD4⁺ and CD8⁺ T-cell recruitment after ICIs therapy. After the systematic review, we concluded that the anti-tumor effect of ICIs is related to the infiltration of both CD4⁺ and CD8⁺ T cells in tumors (Supplementary Table S1). The irAEs induced by anti-CTLA-4 in monotherapy are driven by both T cells and may be triggered by the increased ratio of CD8⁺: CD4⁺ T cells (Khan and Gerber, 2020). It seems that anti-CTLA-4 antibodies do not promote the depletion of CD4⁺ T cells Foxp3⁺ (Tregs) (Supplementary Table S1).

In anti-PD-1 monotherapy, several clinical and preclinical studies have concluded that infiltrating and circulating CD4⁺ and CD8⁺ T cells are important in the anti-tumor effect of ICIs (Supplementary Table S2), with major anti-tumor responses biased to the CD8⁺ T-cell effects. The irAEs induced by PD-1 monotherapy are elicited by both CD4⁺ and CD8⁺ T-cell populations, as suggested by their infiltration in the focal area of the irAEs. In anti-PD-L1 monotherapy, the irAEs might be linked mainly to CD4⁺ and CD8⁺ T cells since both populations have been found to increase in circulation and intratumoral sites. However, the specificity and activity have not been tested (Supplementary Table S3). Existing data in dual therapy (anti-CTLA-4 + anti-PD-1) also demonstrated the importance of CD4⁺ and CD8⁺ T-cell recruitment in the anti-tumor effect (Supplementary Table S4).

In this context, the preclinical models have highlighted the importance of the PD-1/PD-L1 axis to limit the T-cell response and, therefore, limit heart damage. In the CD8⁺ T-cell-mediated myocarditis model, PD-L1 expression on endothelial cells increases; consequently, the absence of PD-L1 exacerbates inflammation and promotes antibodies against cardiac proteins (Grabie et al., 2019). Moreover, the lack of PD-1 on this model increases the CD8⁺ response and cardiac damage (Tarrio et al., 2012). This phenomenon is also reproduced in a CD4⁺ T-cell-dependent model of autoimmune myocarditis, in which PD-1 absence enhanced cardiac damage (Tarrio et al., 2012). Preclinical models also have shown that anti-CTLA-4 imposes major boundaries on CD4⁺ T-cell phenotypes, whereas PD-1 subtly limits CD8⁺ T-cell phenotypes.

This systematic review highlighted the development of several autoantibodies after ICIs therapy. However, we did not study the association with the anti-tumoral effect (Supplementary Table S4). The most common irAEs in which autoantibodies have been described are found in 1) anti-CTLA-4 treatment (thyroid dysfunction), 2) anti-PD-1 treatment (myasthenia gravis/myopathy), and 3) anti-PD-L1 therapies (diabetic ketoacidosis), although with fewer reports than the former therapies.

Currently, the possible role of autoantibodies in cardiac irAEs remains unclear. One report describes that myositis related to the use of anti-PD-1 and anti-PD-L1 therapies could be a marker of subsequent myocarditis induced by these ICIs (Supplementary Table S4). Indeed, it is well-known that patients with myasthenia gravis could further develop myocarditis and myositis related to autoantibody cross-reactivity (Suzuki et al., 2017). In these reports, the autoantibodies induced by ICI therapy that could be associated with myositis and myocarditis are anti-acetylcholine receptors, anti-striated muscle antibodies, anti-mitochondrial antibodies, anti-alanyl-tRNA synthetase, anti-signal recognition particle (SRP) antibodies, and anti-3-hydroxy-3-methylglutaryl-coenzyme A reductase (Supplementary Table S4).

Anti-striated muscle antibodies may cross-react with both skeletal and cardiac striated muscle antigens during myositis or myocarditis, inducing antibody-dependent cellular cytotoxicity (ADCC) (Fazal et al., 2020) through the effector functions of the Fc region of the antibodies, which is recognized by specific receptors, principally on natural killer cells, monocytes, macrophages, neutrophils, eosinophils, and dendritic cells (Wang et al., 2015). These effector immune cells induce lysis of the target cell via cytotoxic granule release, Fas signaling, and the initiation of reactive oxygen species. On the other hand, autoantibodies may induce complement-dependent cytotoxicity (CDC) activation (Duensing and Watson, 2018), ending with the target cell lysis and increasing the damage to the heart (Figure 1).

Supporting preclinical data have shown that co-inhibitory molecules are essential to prevent cardiac diseases (Grabie et al., 2019) in which antibodies are related (Nishimura et al., 2001; Okazaki et al., 2003). PD-1 deficiency causes dilated cardiomyopathy mediated by antibodies deposition on cardiac cells in mice with a Th2 bias response (Nishimura et al., 2001; Okazaki et al., 2003), which are more prompt to develop a humoral response (Busch et al., 2016). In contrast, mice with a Th1 bias (cellular immune response) developed atrial fibrillation (Fu et al., 2013). Moreover, exacerbated inflammation in preexistent inflammation stimuli, induced by the absence of co-inhibitory molecules, promotes autoantibodies related to cardiac self-antigens (Grabie et al., 2007). Furthermore, antibodies against cardiac antigens exogenously administrated in wild-type mice induced heart dysfunction (Okazaki et al., 2003). Altogether, these data highpoint the role of cardiac antibodies in promoting ICI-mediated heart disease.

According to this systematic review, there are few reports regarding the inflammatory cytokine production in ICI toxicity (Supplementary Table S5). IFN-ƴ was the main cytokine identified in ICI toxicity. Other cytokines and chemokines such as interleukin 1β (IL-1β) and the chemokine (C-X-C motif) ligand 10 (CXCL10) were also identified. However, the evidence is not enough to determine their role in ICI cardiotoxicities (Supplementary Table S5).

During heart damage, proinflammatory cytokines and chemokines are released by cardiomyocytes, epithelial cells, and fibroblasts (Hanna and Frangogiannis, 2020; Dick and Epelman, 2016; Shirazi et al., 2017), inducing the recruitment of neutrophils and macrophages to the heart. They are key players in perpetuating inflammatory and chemoattracting factors, such as CXCL10, which induces T-cell recruitment. Binding between self-antigens and T cells induces anergy and T-cell apoptosis under normal conditions. However, when ICI therapy is administered, T-cell activity increases and worsens in the heart, inducing further injury, damage-associated molecular patterns (DAMPs) release, self-antigen liberation, and increased proinflammatory cytokines in the myocardium.

On the other hand, proinflammatory molecules may cause direct cardiac dysfunction. In this context, inflammatory molecules, such as IL-6 and C-reactive protein, increase the risk of supraventricular and ventricular cardiac arrhythmias and their complications (Kanda and Takahashi, 2004; Dick and Epelman, 2016). Moreover, immune-related myocardial inflammation could cause sick sinus and inter-atrial block (Darnell et al., 2020; Tajiri et al., 2018). Regarding the cardiac pro-arrhythmic effects caused by ICI therapy, Johnson et al. recently reported a case of complete heart block and cardiac arrest following initial doses of nivolumab and ipilimumab (Johnson et al., 2016).

Experimental evidence shows that cytokines directly affect cardiomyocytes altering the excitation–contraction coupling by affecting calcium handling, enabling spontaneous calcium release (Fernández-Sada et al., 2017), and facilitating the substrate to ventricular arrhythmias (Colman, 2019), (Chen et al., 2020), (Kouvas et al., 2018). For example, it has been shown that IL-1β increases calcium sparks in cardiomyocytes favoring arrhythmias associated with CaMKII oxidation and phosphorylation (Monnerat et al., 2016). On the other hand, TNF-α can modulate sarcoplasmic reticulum Ca-ATPases (SERCA2a) methylation, altering calcium handling and promoting cell dysfunction (Kao et al., 2010).

The mechanisms involved in cardiac irAEs after ICI treatment remain unclear; however, here, we propose that concurrent injury or stress is required to induce cardiac irAEs (Figure 1A). Under normal conditions, T-cell responses are in homeostasis due to the actions of co-inhibitory molecules (CTLA-4, PD-1, and PD-L1) promoting inflammatory resolution. However, tumor cells often overexploit this mechanism, preventing them from being eliminated (Figure 1A2). Indeed, when ICIs are administered to patients with cancer, these cells are removed. Therefore, we hypothesize that in some cases, when a concomitant heart injury exists, the release of self-antigens promotes cardiac inflammation and damage, and due to ICI treatment, this response avoids cardiac inflammation resolution (Figures 1A3). Furthermore, we propose that self-antigens released due to chemotherapy, stress, or elimination from cancer cells can cross-react with cardiac antigens, enhancing the inflammatory signals (Figure 1B). However, these hypotheses need to be demonstrated.

Because several ICIs are administrated after the first-line treatments such as chemotherapy or radiotherapy, the injury may be triggered by these first-line treatments. Typical examples of this phenomenon are anthracyclines, including doxorubicin, daunorubicin, and epirubicin, the best-studied anti-cancer agents with cardiotoxic side effects (Dick and Epelman, 2016). For example, doxorubicin leads to the necrosis of cardiomyocytes and, therefore, the release of DAMPs, which could induce immune activation against cardiac self-antigens. The release of DAMPs may also be provoked by chemotherapy, radiotherapy, or resection of the tumor (Kouvas et al., 2018) due to the damage, necrosis, or stress in cancer (or other) cells, and these DAMPs or self-antigens could cross-react with some heart and skeletal muscle proteins.

Therefore, both preexistent cardiac injury and injury in other tissues may act as inflammation trigger signals. When the administration of ICI therapy occurs, the resolution of inflammation is blocked, promoting cardiac inflammation mediated by T cells (Nirschl and Drake, 2013), the production of cardiac antibodies (Drugs@FDA, 2021), and the impaired cardiac cell function mediated by autoantibodies and cytokines signaling (Dyck and Mills, 2017). Additionally, PD-L1, expressed by cardiac cells, might be directly recognized by ICIs, and promotes cardiac cell death by mechanisms such as the ADCC (Vaddepally et al., 2020). Recent evidence showed that T-cell infiltration is an important mechanism involved in ICI cardiotoxicity. Wei et al., 2021 described a genetic mouse model that recapitulates ICI-associated myocarditis in patients and results in premature mice death, myocardial infiltration (T cells and macrophages), and severe electrocardiogram abnormalities (Wei et al., 2021). Bocksthaler et al. (2020) showed that Troponin I (TnI)-directed autoimmune myocarditis (TnI-AM) is mediated by CD4⁺ T-cells, and the immunoproteasome is a key player in this autoimmunity. These characteristics are also presented in 2 cases of ICI-related myocarditis demonstrating the ICI-induced autoimmunity (Bockstahler et al., 2020). The study published by Michel et al. (2022) reported a murine model on which anti-PD1 therapy promotes myocardial infiltration (CD4⁺ and markedly activated CD8⁺ T cells) and impaired left ventricular (LV) function. The analogous results of impaired LV function were found in patients with metastatic melanoma receiving anti-PD1 therapy. Remarkably, they found that the blockade of TNFα preserved the LV function without attenuating the anti-cancer efficacy of anti-PD1 therapy (Michel et al., 2022). Finally, the study of Finke et al. (2021) found leukocytic infiltration (CD3⁺ and CD8⁺ cells) in cardiac biopsies from patients with ICI-associated myocarditis and the upregulation of IFN-ƴ pathway, which may indicate a possible link to the inflammasome activation and M1 macrophages activity(Fujiwara et al., 2016). Recently, a preprint article highlights the importance of CD8⁺ T cells in a genetic murine model (Pdcd1−/−Ctla4+/−) that recapitulates clinicopathologic characteristics of ICI-myocarditis. They identified the alpha-myosin as a cardiac-specific antigen; peptides from this protein induce the expansion of peripheral blood T cells from two patients with ICI-myocarditis indicating its clinical importance (Balko et al., 2022).

It has been proposed that the PD1-PD-L1 pathway protects the heart and lungs from autoimmune damage (Colman, 2019) since PD-1 deficiency has been shown to induce fatal myocarditis in Murphy Roths Large (MRL) mice (Chen et al., 2020). In this scenario, the direct binding of anti-PD-L1 to the PD-L1 expressed on cardiomyocytes contributes to the failure in maintaining chronic T-cell activation and thus the loss of tolerance.