This retrospective multicenter study includes a total of 51 patients from 12 centers (Cologne, Erlangen, Essen, Freiburg, Goettingen, Hannover, Luebeck, Mainz, Minden, Munich, Nashville, Zurich) with confirmed cardiac irAE drawn from our SERIO-registry and was approved by the ethics committee of the LMU University Hospital Munich (No. 20-1122). The diagnosis of irMyocarditis was confirmed by cardiologists at the corresponding center, who assessed symptoms, ECG alterations, troponin levels, and functional or structural changes in cardiac imaging, alongside a confirmed temporal association with the administration of ICIs. Cardiac MRIs were conducted on 33 patients, revealing findings consistent with irMyocarditis, such as late gadolinium enhancement, myocardial edema, wall motility disorders, and/or ischemia. Furthermore, endomyocardial biopsy results from 18 patients showed evidence of toxic damage to myocardial cells, including interstitial fibrosis, degenerative changes, and increased infiltration of lymphocytes, especially T cells, and CD68-positive macrophages, further supporting the diagnosis. If available, we collected creatine kinase (CK), creatine phosphokinase-MB (CK-MB), troponin T, and N-terminal prohormone of brain natriuretic peptide (NT-proBNP) serum levels at peak, as well as examination results of electrocardiogram (ECG), echocardiography (Echo) with left ventricular ejection fraction (LVEF), cardiac magnetic resonance imaging (MRI), and coronary angiography.

Clinical data was gathered using the international web-based Side Effect Registry Immuno-Oncology (SERIO; www.serio-registry.org). It was initiated in cooperation with the Paul-Ehrlich-Institute, to document rare, complex, or therapy-refractory side effects induced by immunotherapies. Within SERIO patients’ demographics, tumor entity, ICI therapy applied, tumor outcome, as well as type of irAE, irAE onset, grade, treatment and outcome are documented. The grading of irAEs followed the Common Terminology Criteria for Adverse Events (CTCAE) version 5.0. IrAE outcome was classified as resolved, improved, resolved with sequelae, or ongoing. The classification of completely “resolved” cardiac irAEs refers to cases where all cardiac symptoms and abnormalities associated with the irAE have disappeared. “Improved” defines irAE cases where there is a noticeable enhancement in cardiac irAE symptoms and/or abnormalities but which haven't entirely disappeared. “Resolved with sequelae” indicates that while the symptoms have subsided, there are residual effects or complications of the cardiac irAE, e.g., elevated troponin levels, remaining. “Ongoing” signifies cardiac irAE cases where symptoms and abnormalities persist without any significant improvement or resolution. Ethical approval for analyses from the SERIO registry was granted (Erlangen Nr. 2_20 B, Erlangen Nr. 17_16 Bc). SERIO was queried for cases of cardiac irAEs in February 2024.

For lab investigations participants gave their informed written consent prior to analysis. In total, peripheral blood mononuclear cells (PBMCs) from 19 patients were analyzed longitudinally for immunophenotyping at baseline and during ICI-therapy or during irAE. PBMCs from irMyocarditis patients (n = 4 baseline and n = 8 during irAE) were compared to patients without toxicity under ICI-therapy (n = 4 baseline and n = 7 during ICI-therapy). Heart muscle biopsies included irMyocarditis patients from Erlangen, Luebeck, Munich (D), and Zuerich (CH) (n = 11 available for lab investigations). Healthy heart muscle biopsies were obtained from pediatric cardiac surgery (n = 5). The healthy muscle was from a biobank that captured heart muscle from patients with valve reconstruction.

PBMCs were isolated from lithium heparin blood samples by density gradient centrifugation with Ficoll-Paque density-gradient centrifugation followed by a cryopreservation with X-VIVO medium (Lonza) enriched with 20% fetal bovine serum (FBS, Pan Biotech) as well as with 10% DMSO (Thermo Fisher) and stored in liquid nitrogen. PBMCs were obtained from 8 patients with irMyocarditis and 11 ICI-patients without development of toxicities.

Thawed PBMCs were rested for one hour in RPMI medium + 10% FCS + 1% Pen/Strep + 1% L-Glutamine. Prior to staining, cells were incubated with human TrueStain Fc blocking reagent (Biolegend). Cells were then incubated (30 min, 4 °C) with eBioscience Fixable Viability Dye eFluor506 (Invitrogen) and the surface antibody mix (Table 1) in PBS. After fixation and permeabilization with eBioscience fixation/permeabilization reagent (Invitrogen; 30 min, 4 °C) and washing, cells were incubated (30 min, 4 °C) with the intracellular/intranuclear antibody mix (Table 1) and then measured the same day on an LSR Fortessa Cell Analyzer (BD Biosciences). FlowJo software (10.10.1) (BioSciences) was used for analysis.

Total RNA was isolated from heart muscle specimens using TRIzol™ Reagent (Thermo Fisher Scientific, Germany) according to the manufacturer's instructions.

Gene expression was analyzed using the nCounter PanCancer Immune Profiling Panel™ (human) (Nanostring, XT-CSO-HIP1–12). Sample detection and analysis were completed on a nCounter® Digital Analyzer. RNA gene expression analysis was performed using the nCounter SPRINT Profiler and the nCounter MAX Analysis System (NanoString Technologies, Inc.). Statistical analysis of gene expression results was implemented with the nSolver Analysis Software (NanoString Technologies, Inc.) and ROSALIND® (ROSALIND, Inc., San Diego, CA) (https://rosalind.bio/). Pathway analysis was conducted using EnrichR and EnrichR-KG, applying the databases KEGG 2021 Human and GO Biological Process 2021.

Raw data processing, quality control and normalization were performed using the NanoStringQCPro package for the R (version 4.1.0) environment. Quality control (QC) was performed with an imaging QC of >80% field of view registration, binding density QC within 0.05–2.25 range, and positive control scaling factors within a range of 0.3–3. In the differential expression analysis, a false discovery rate (FDR, Benjamini and Hochberg) adjusted p-value of ≤0.05 was applied as cutoff. One sample two-tailed unpaired Student's t-test, one sample Wilcoxon test and Mann–Whiney U-test were used to test statistical significance. *: p ≤ 0.05, **: p ≤ 0.01, and ***: p ≤ 0.001.

The median time from initiation of ICI therapy to onset of cardiac irAEs was 37 days with a wide range (4–1,074 days). The time to onset of irMyocarditis significantly differed from other cardiac irAEs. In total, irMyocarditis occurred early after ICI initiation with a median time to onset of 36 days (range 4–1,074 days), whereas other cardiac irAEs had a median time to onset of 101 days (range 28–119 days). More than half of all patients (58.8%; 30/51) initially presented with symptoms like dyspnea (63.3%; 19/30) or chest pain (23.3%; 7/30) while 41.2% were asymptomatic. In over 50.0% of irAE cases, which ultimately resolved completely, elevated laboratory parameters (Troponin T, CK, CK-MB, and NT-proBNP) were initially observed (Table 3). In those patients that fully recovered, echocardiographic findings were predominantly within the normal range (92.9%; 13/14) and coronary angiographies presented with no signs of stenosis (75.0%; 9/12). The results of ECG and cardiac MRI in this cohort were nonspecific, showing a distribution of 50.0% abnormal and 50.0% normal results. Notably, biopsies of the myocardium of patients who fully recovered were indicative of myocarditis in 7 out of 8 cases (87.5%). Interestingly, in the cohort that ultimately led to fatality, cardiac diagnostic assessments were not performed in more than 50.0% of cases. Upon conducting the investigations in this cohort, abnormalities were observed in ECG (100.0%; 3/3) and coronary angiography (100.0%; 2/2), whereas the echocardiogram (66.7%; 2/3), LVEF (100.0%; 3/3), and MRI (100.0%; 1/1) findings were predominantly within the normal range (Table 3).

In total, 92.5% (49/53) of patients with cardiac irAE were treated with systemic steroids, 5.7% (3/53) were solely treated with symptomatic therapies such as beta blockers and antihypertensive drugs, and 17.0% (9/53) required second-line therapy for steroid-refractory or steroid-dependent cardiac irAEs. As second-line immunosuppressants, intravenous immunoglobulins (IVIG), infliximab, anti-thymocyte globulin (ATG), mycophenolate mofetil (MMF) were applied. Optimal response rates of 90.0% (4/5) were attained through the administration of IVIG. In total, 72.5% (37/51) of patients required hospitalization.

Data evaluation of cardiac irAE outcomes indicated that 34.0% (18/53) of cases completely resolved, 15.1% (8/53) significantly improved, while 22.6% (12/53) resolved with relevant sequelae and 11.3% (6/53) were still ongoing, with an overall irAE response rate of 71.7% (Figure 1). In 11.3% (6/53) of cases, cardiac irAEs led to death (Figure 1). In our study, 5 patients received ICI rechallenge after cardiac irAE. In 80.0% (4/5) of cases ICI was successfully applied without recurrence of cardiac toxicities, while in one case the patient died early after ICI re-initiation due to exacerbated cardiac symptoms. In one case, ICI rechallenge was successfully administered with prophylactic steroid treatment. Tumor response in patients with cutaneous melanoma was assessed at the time of irAE: Complete response was reported in 5.9% (2/34), partial response in 8.8% (3/34), stable disease in 20.6% (7/34) and progressive disease in 32.4% (11/34) of patients (RECIST 1.1). Patients showed a median PFS of 8 months (range 0–60 months) and a median overall survival (OS) of 17 months (range 0–130 months). In total, 11.8% (4/34) of patients were treated in an adjuvant setting with recurrence-free survival between 1 and 13 months.

To elucidate the gene expression signatures of irMyocarditis, heart tissue of 12 irMyocarditis patients was compared to heart tissue of 5 healthy controls using nCounter® Digital Analyzer. After normalizing the raw data, with a ratio >1.5 or <1.5 and p < 0.05 as the screening criteria for differential expression genes (DEG), the DEGs analysis identified a set of 113 differentially regulated genes in irMyocarditis compared to healthy controls, of which 99 genes were upregulated and 14 were downregulated (Figure 2A). The top five upregulated genes in irMyocarditis were IL6R (logFC 2.1, adjusted p-value 3.34 × 10⁻⁵), HLA-C (logFC 3.62, adjusted p-value 6.31 × 10⁻⁴), HLA-E (logFC 3.58, adjusted p-value 7.19 × 10⁻⁴), STAT2 (logFC 2.64, adjusted p-value 8.25 × 10⁻⁴) and IKBKB (logFC 2.41, adjusted p-value 9.22 × 10⁻⁴). The top five downregulated genes in irMyocarditis were DDX43 (logFC −10.04, adjusted p-value 3.01 × 10−4), HMGB1 (logFC −2.75, adjusted p-value 1.90 × 10⁻³), SYT17 (logFC −5.17, adjusted p-value 3.64 × 10⁻³), S100A12 (logFC −7.79, adjusted p-value 5.43 × 10⁻³) and IL13RA2 (logFC −9.45, adjusted p-value 9.85 × 10⁻³). The 113 DEGs were investigated using gene set enrichment and pathway analysis (Figure 2B), demonstrating the most significant pathways of the DEGs: (1) Cytokine-mediated signaling pathway (p = 6.6445 × 10⁻³⁷), (2) Cellular response to cytokine production (p = 5.5129 × 10⁻³²), (3) Positive regulation of cytokine production (p = 1.4357 × 10⁻³⁰), (4) Hematopoietic cell lineage (p = 1.4458 × 10⁻²²), (5) Epstein-Barr virus infection (p = 2.5618 × 10⁻²²), (6) Cell adhesion molecules (p = 2.858 × 10⁻²²), (7) Th17 cell differentiation (p = 2.6598 × 10⁻²⁰), (8) Human T-cell leukemia virus 1 infection (p = 3.6892 × 10⁻²⁰), (9) Regulation of interleukin-6 production (p = 1.724 × 10⁻¹⁵), (10) Innate immune response (p = 6.7096 × 10⁻¹⁵). Thus, the regulation of cytokines, especially type I IFNs and NFkB signaling was differently shaped in irMyocarditis compared to healthy control, indicating might also be relevant—IL-6 can be therapeutically inhibited.

Based on the analyzed gene expression patterns of immune response mechanisms, we next focused on characterization of the profile of the systemic immune response. We performed longitudinal analysis of PBMCs from irMyocarditis patients (baseline and time of adverse event). In longitudinal analysis, PBMCs of patients during irMyocarditis (ae) demonstrated significantly decreased frequencies of CD103+ activated CD4+ T cells and increased CD103+ activated CD4+ Treg cells compared to baseline (bl) whereas activated T cells as well as Treg cells showed no significant difference between bl and ae. Furthermore, no significant difference in CD103+ activated CD8+ T cells and activated CD8+ Treg cells was observed. Integrin CD103 is related to tissue residency in the context of inflammation and cancer (Figures 3A,B). The peripheral T cell compartment was investigated to determine the relative abundance of various T cell phenotypes, including effector-like T cells (Teff), effector memory-like T cells (Tem), central memory-like T cells (Tcm), and naïve T cells (Tn). This analysis utilized established markers such as CD45RA, CD45RO, CD62l, CCR7, and CD127. The frequency of Tn, Tcm, Tem, and Teff showed no significant difference neither in CD4+ nor in CD8+ T cells (Supplementary Figures S1A,B). Furthermore, markers associated with T cell activation (CD44, CD27, CD28, ICOS, and CD107a) and exhaustion (CTLA4, LAG3, TIM3, and TIGIT) were examined to gain further functional insights into differences in systemic T cell immunity. The CD8+ T cells showed an increased expression of CD44 which was not observed on CD4+ T cells (Figure 3C). Furthermore, we focused on the function and phenotype of antigen-presenting cells (APCs). We determined three different DC subsets in peripheral blood including BDCA1- cDC, BDCA1+ cDC, and monocyte-derived DC (moDC) based on the markers CD11c, CD19, CD56, CD14, BDCA1, and HLA-DR and determined the expression of the activation markers CD80, CD86 and HLA-DR as well as the inhibitory molecule PD-L1. Regarding DC, we found significantly higher frequencies of activated DC during the adverse event compared to baseline in irMyocarditis patient's PBMC as determined by the expression of the costimulatory molecule CD80 (Figure 3D). Besides, for DC subsets, the expression of the coinhibitory molecule PD-L1 on BDCA1- cDC was significantly reduced in ae compared to bl, indicating the effect might be related to the development of irMyocarditis (Figure 3E). The expression of CD80, CD86, and the MHC class II molecule HLA-DR on BDCA1- cDC, BDCA1+ cDC, and moDC showed no significant difference between ae and bl (Supplementary Figures S1C–E). To summarize, the peripheral T cell compartment of irMyocarditis patients showed ICI-driven enhanced levels of activated CD4+ Treg cells, CD44, CD80, PD-L1 expression on DC, and reduced levels of activated CD4+ T cells.

Following the longitudinal immunophenotype changes in irMyocarditis patients from baseline to adverse event, we analyzed the irMyocarditis patients during the adverse events in comparison to patients who did not develop toxicities while undergoing ICI-therapy toxicities (Ø tox ICI-patients). PBMCs of irMyocarditis patients showed significantly higher frequencies of CD4+ Treg cells and higher levels of activated CD8+ T cells compared to Ø tox ICI-patients (Figure 4A). Interestingly, different from previous results, irMyocarditis patients showed significantly higher frequencies of CD4+ Tcm cells and CD4+ Teff cells compared to Ø tox ICI-patients. Furthermore, we observed a significant decrease in CD4+ Tem cells and CD8+ Tem cells in irMyocarditis compared to Ø tox ICI-patients (Figure 4B), whereas no significant alterations of CD4+ naïve T cells and CD8+ naïve T cells could be observed (Supplementary Figure S2A). Noticeably, among the tested functional markers only CD28 of CD4+ T cells was elevated in Ø tox ICI-patients (Supplementary Figures S2B,C). Of note, both CD4+ as well as CD8+ T cells of irMyocarditis patients showed significantly increased expression of the T cell exhaustion marker TIGIT, however a decrease of TIM-3 expression on CD8+ T cells (Figure 4C). For other T cell exhaustion markers, the expression of LAG3 was elevated in CD4+ T cells but not in CD8+ T cells in irMyocarditis compared to Ø tox ICI-patients (Supplementary Figure S2D). The expression of CTLA4, however, was not different in irMyocarditis and Ø tox ICI-patients neither in CD4+ nor in CD8+ T cells (Supplementary Figure S2E). Further analysis of APC cells showed significantly higher expression of CD86 on activated DC, BDCA1- cDC, BDCA1+ cDC, and moDC (Figure 4D). Regarding other costimulatory molecules, we found significantly higher frequencies of CD80 on BDCA1- DC as well as the HLA-DR on BDCA1- cDC, BDCA1+ cDC in irMyocarditis patients, pointing towards an increase in cDC activation potentially induced by irMyocarditis (Figures 4E,F). The expression of CD80 markers on activated DC, BDCA1+ DC, and moDC, and the expression of HLA-DR on moDC and activated DC, however, was not statistically different between irMyocarditis and Ø tox ICI-patients (Supplementary Figures S2F,G). Strikingly, the expression of PD-L1 showed a significant increase in all DC subsets in irMyocarditis patients (Figure 4G). Furthermore, in our irMyocarditis cohort, circulating B cells were significantly less abundant than in Ø tox ICI-patients (Supplementary Figure S2H). Lastly, a difference between CD56high/CD16- NK cells and CD56low/CD16+ NK cells could not be detected between irMyocarditis and Ø tox ICI-patients, which indicated the cardiotoxicity might not be related to NK cell-mediated cytotoxicity (Supplementary Figure S2I). Taken together, these data suggest that irMyocarditis is characterized by activation of T cells, particularly Tem cells, additionally the exhaustion markers TIGIT, TIM and LAG3 were identified to be of greater importance in irMyocarditis patients.

Furthermore, we aimed to analyze the baseline immunophenotype of PBMC of irMyocarditis patients in comparison to Ø tox ICI-patients before initiation of ICI-therapy. First, we determined the frequencies of activated T cells in peripheral blood. Here, CD103+ activated CD4+ T cells and CD103+ activated CD4+ Treg cells showed significant expansion in bl-irMyocarditis compared to bl-Ø tox ICI-patients,. whereas the abundance of CD103+ activated CD8+ T cell CD103+ activated CD8+ Treg cell were unaltered (Figures 5A,B). Regarding other phenotypes of T cells, we found no difference in frequencies of Tn, Tcm, Tem, and Teff neither in CD4+ nor in CD8+ T cells between bl-irMyocarditis patients and bl-Ø tox ICI-patients (Supplementary Figures S3A,B). To gain additional functional insights, we analyzed activation and exhaustion markers of the T cells. Interestingly, most activation markers showed significantly higher levels in bl-irMyocarditis, such as CD107a in CD8+ T cells, and ICOS in both CD4+ and CD8+ T cells, however CD28 was significantly higher in CD4+ T cells in bl-Ø tox ICI-patients compared to bl-irMyocarditis (Figure 5C). Expression of CD44, as well as CD27, showed no difference between bl-irMyocarditis and bl-Ø tox ICI-patients (Supplementary Figures S3C,D). Markedly, expression of the exhaustion marker LAG3 was highly elevated in both CD4+ and CD8+ T cells in bl-irMyocarditis. Significantly enhanced expression of CTLA4 in CD4+ T cells and TIGIT in CD8+ T cells, however a significant decrease of TIM-3 in CD4+, respectively, was detected in bl-irMyocarditis compared to bl-Ø tox ICI-patients (Figure 5D). In DC subsets, we observed an increased expression of PD-L1 on activated DC, BDCA1- cDC, BDCA1+ cDC, and moDC in bl-irMyocarditis patients (Figure 5E). Additionally, in our bl-irMyocarditis cohort, circulating B cells were significantly lower abundant than in bl-Ø tox ICI-patients. A significantly higher amount of moDC could be detected in the bl-irMyocarditis cohort compared to bl-Ø the tox ICI cohort (Figure 5F). To summarize, these data suggest that expression of exhaustion markers, particularly LAG3 on T cells and PD-L1 on DCs at baseline could be used as indicators to predict the occurrence of irMyocarditis.