Injury-induced myosin-specific tissue-resident memory T cells drive immune checkpoint inhibitor myocarditis

Significance

We determined that cardiac MyHC T_RM cells are present in naive hearts and accumulate in the pericardial area in mice and humans after recovering from cardiac injury. T_RM cells express high levels of PD-1, making them susceptible to activation by anti-PD-1 blocking antibodies. ICI-myocarditis is more prevalent after previous cardiac injury, demonstrating that T_RM cells perpetuate the disease. Thus, our results provide unique insights into risk factors and the disease pathogenesis of ICI-myocarditis.

Abstract

Cardiac myosin-specific (MyHC) T cells drive the disease pathogenesis of immune checkpoint inhibitor–associated myocarditis (ICI-myocarditis). To determine whether MyHC T cells are tissue-resident memory T (T_RM) cells, we characterized cardiac T_RM cells in naive mice and established that they have a distinct phenotypic and transcriptional profile that can be defined by their upregulation of CD69, PD-1, and CXCR6. We then investigated the effects of cardiac injury through a modified experimental autoimmune myocarditis mouse model and an ischemia–reperfusion injury mouse model and determined that cardiac inflammation induces the recruitment of autoreactive MyHC T_RM cells, which coexpress PD-1 and CD69. To investigate whether the recruited MyHC T_RM cells could increase susceptibility to ICI-myocarditis, we developed a two-hit ICI-myocarditis mouse model where cardiac injury was induced, mice were allowed to recover, and then were treated with anti-PD-1 antibodies. We determined that mice who recover from cardiac injury are more susceptible to ICI-myocarditis development. We found that murine and human T_RM cells share a similar location in the heart and aggregate along the perimyocardium. We phenotyped cells obtained from pericardial fluid from patients diagnosed with dilated cardiomyopathy and ischemic cardiomyopathy and established that pericardial T cells are predominantly CD69⁺ T_RM cells that up-regulate PD-1. Finally, we determined that human pericardial macrophages produce IL-15, which supports and maintains pericardial T_RM cells.

Immune checkpoint molecules such as cytotoxic T lymphocyte-associated protein 4 (CTLA-4) and programmed cell death protein 1 (PD-1) are negative regulators of T cell activation (1–4). The development of immune checkpoint inhibitor (ICI) therapies has proven to reinvigorate exhausted tumor-infiltrating lymphocytes (TIL), though clinically efficacious, 40 to 72% of patients go on to develop immune-related adverse events (irAE) (5–8). While most irAEs are mild to moderate in clinical severity, inflammation of the heart muscle, or myocarditis, is the deadliest irAE (7). ICI-associated myocarditis (ICI-myocarditis) predominately affects patients treated with antibodies blocking the PD-1/PD-L1 axis and is a rare irAE that affects less than 1% of patients but has a mortality rate of up to 50% (7–10). Due to the high fatality rate, it is imperative to understand what predisposes patients to develop ICI-myocarditis. Our lab and others recently showed that ICI-myocarditis is mediated by autoreactive T cells specific for the α-heavy chain of cardiac myosin (11, 12). Further characterization of these cardiac myosin-specific (MyHC) T cells revealed an antigen-experienced phenotype and expression of CD69 (11, 12). Upregulation of CD69 has been associated with T cell activation and tissue residency (13–15).

Murine T cells can be separated into functionally distinct subsets with memory T cells distinguished by their expression of the activation marker CD44 (14, 15). In comparison, naive T (T_N) cells are antigen inexperienced and, as such, have yet to be activated. T_N cells and central memory T (T_CM) cells are defined by their ability to enter the secondary lymphoid organs. To do this, they express lymph node homing molecules such as CCR7 and CD62L (14, 15). In contrast, effector memory T (T_EM) cells survey the peripheral tissues and do not express lymph node homing molecules (14, 15). Tissue-resident memory T (T_RM) cells are the most recently discovered subset of memory T cells; they are noncirculating and remain resident in the peripheral tissues (14–16). CD69 and CD103 are classical T_RM cell markers that help maintain tissue residency (14, 15, 17). CD69 is an early activation marker that directly antagonizes the tissue egress protein, sphingosine 1-phosphate receptor 1 (S1PR1) (18, 19). Another classical marker for T_RM cells is CD103, which dimerizes with integrin β7 to bind E-cadherin and is important for retention in epithelial tissues (14, 20). T_RM cells have been shown to accumulate in the tissues after injury or vaccination, where they become long-lived resident immune cells (21–24). T_RM cells act as a first-line defense against invading pathogens in the peripheral tissues by coordinating innate and adaptive immune responses (25–27). Though initially described for their protective qualities, T_RM cells have also been implicated in the pathogenesis of certain autoimmune diseases such as psoriasis and vitiligo (28, 29). In addition, it has been shown that T_RM cells express high levels of PD-1 and use the PD-1/PD-L1 axis to maintain tissue homeostasis (30, 31). Currently, T_RM cells have been characterized in a variety of tissues but have yet to be defined in the heart.

Autoreactive cardiac myosin-specific T cells have been shown to escape negative selection in the thymus and can be found in the periphery (32). Our previous studies have shown that myosin-specific T cells are located in the hearts of naive mice (11). Furthermore, our lab, as well as others, has shown that myosin-specific T cells are recruited to the heart during cardiovascular disease and can play both pathogenic and protective roles (11, 33, 34). Currently, it remains unclear whether these recruited autoreactive MyHC T cells maintain long-term residency in the heart and what function they play. In this manuscript, we describe the contribution of cardiac T_RM cells in the disease pathogenesis of ICI-myocarditis. We demonstrate that murine mouse models of cardiac injury induce the recruitment of autoreactive cardiac myosin-specific CD69⁺PD-1⁺ T_RM cells into the heart, which preferentially locate along the perimyocardium. To investigate the contribution of MyHC T_RM cells to ICI-myocarditis development, we created a two-hit ICI-myocarditis mouse model and showed that mice who recover from previous cardiac injury are more susceptible to ICI-myocarditis development. Finally, we demonstrated that human patients diagnosed with various cardiovascular diseases have pericardial CD69⁺ T_RM cells that up-regulate the expression of PD-1, which match the phenotype and location of murine cardiac T_RM cells.

Results

CD69 and PD-1 Expression Distinguishes Cardiac T_RM cells.

CD69 and CD103 are well-characterized tissue retention markers that distinguish T_RM cells in various tissues (14, 15, 17). To investigate whether these classical markers differentiate cardiac T_RM cells from circulating T cells, we performed multiparameter flow cytometry on hearts of naive A/J mice. T cells were stratified into T_N, T_CM, T_EM, and T_RM populations based on the expression of CD62L and CD44 (SI Appendix, Fig. S1 A and B). CD69 was differentially expressed by cardiac T_RM cells, while CD103 was not (SI Appendix, Fig. S1 C and D). To further characterize cardiac T_RM cells, we performed intravascular labeling on naive A/J mice, which allowed us to distinguish between infiltrating immune cells trapped in the vasculature and cardiac resident leukocytes (Fig. 1 A and B) (35). Resident cardiac T cells up-regulated CD69 in comparison to infiltrating T cells (Fig. 1C and SI Appendix, Fig. S1E). To further characterize resident cardiac lymphocytes, we assessed the expression of PD-1, which has been reported to be up-regulated by resident memory T cells in other organs (30, 31). First, we investigated the expression of PD-1 on infiltrating and resident T cells and found that cardiac resident T cells had significantly higher expression of PD-1 compared to infiltrating T cells (SI Appendix, Fig. S1 F–H). To determine whether resident T_RM cells up-regulate the expression of PD-1, we compared the expression of PD-1 on resident CD69⁺ and CD69⁻ T cells and found that PD-1 expression was significantly higher on resident CD69⁺ T cells (Fig. 1 D and E). Together, these findings suggest that PD-1 and CD69 are coexpressed by cardiac T_RM cells. Autoreactive MyHC T cells were shown to escape negative selection in the thymus (32). To determine whether these autoreactive T cells are cardiac resident, we stained cells with a tetramer specific for the TCR, which recognizes the α-chain of cardiac myosin (MyHCα_334−352). Autoreactive CD4⁺ and CD8⁺ MyHC T cells were significantly increased in the resident lymphocyte population (Fig. 1 F and G). This finding indicates that autoreactive MyHC T cells maintain residency in the naive mouse heart but are also found in circulation, consistent with our previous findings (11). Autoreactive effector MyHC T cells recruited to the heart after ischemic injury have been shown to up-regulate a regulatory T cell (Treg) signature (33, 34). Tregs are defined by their expression of the transcription factor FOXP3 and the high affinity IL-2 receptor alpha (CD25) (36). To determine whether resident MyHC T cells in a naive mouse heart are Tregs, we costained for FOXP3, CD25, and the cardiac myosin-specific tetramer. We found that a majority of resident autoreactive MyHC T cells were conventional T cells (Tconv) and not Tregs (Fig. 1 H and I). Thus, cardiac T_RM cells, including MyHC T cells, reside in the heart during naive states and are characterized by their expression of CD69 and upregulation of PD-1.

Fig. 1.

CD69 and PD-1 expression distinguishes cardiac T_RM cells from circulating T cells. (A) Graphical depiction of the intravascular labeling technique. (B) Gating strategy for determining infiltrating and resident T cells. (C–G) Comparing the expression of CD69, PD-1, and cardiac myosin tetramer in infiltrating (CD45IV⁺) and resident (CD45IV⁻) cardiac T cells. (H) Determining the proportion of regulatory T cells (Tregs) compared to all resident myosin-specific T cells. (I) Determining the proportion of Tregs and conventional T cells (Tconv) from all resident MyHC⁺ T cells. Naive A/J male, female, young (10 wk), and aged (40 wk) mice were included in this analysis. Group size was n = 17. Statistical significance was determined through a paired t test.

Cardiac T_RM Cells are Transcriptionally Distinct from Peripheral T Cells.

To determine whether cardiac T_RM cells have a distinct transcriptional profile compared to peripheral T cells, we used fluorescence-activated cell sorting (FACS) to isolate circulating T_N, T_CM, T_EM, and cardiac CD69⁺ T_RM cells from naive A/J mice and then performed RT-qPCR (cell sorting gating strategy in SI Appendix, Fig. S2 A and B). We then plotted the generated CT values as a ratio over the housekeeping gene GAPDH, therefore, transcripts with higher expression have a lower ratio. We investigated whether cardiac T_RM cells down-regulate the tissue egress protein S1PR1, which is known to be suppressed by CD69 as a mechanism for cells to maintain residency in peripheral tissues (18, 19). We found that cardiac T_RM cells did not alter the expression of S1pr1 but had a low, or nondetectable expression, of S1pr5 (SI Appendix, Fig. S2C). We then examined the gene expression of the activation marker PD-1 (Pdcd1) and the chemokine receptor Cxcr6 and found that cardiac T_RM cells up-regulated both compared to circulating T cells (Fig. 2A). Finally, T_RM cells are known to have altered metabolism and express various isoforms of the fatty acid binding proteins (FABPs), which are important for the formation and maintenance of T_RM cells in peripheral tissues such as the skin and the liver (37, 38). Of the nine isoforms, we found that cardiac T_RM cells up-regulated the expression of three isoforms: Fabp3, Fabp4, and Pmp2 (Fig. 2A).

Fig. 2.

Cardiac CD69⁺ T_RM cells are transcriptionally distinct from circulating T cells. (A) Peripheral and resident cardiac T cells were isolated from naive A/J mice and FACS sorted into distinct T_N, T_CM, T_EM, and T_RM cell populations. Graphs depict the Ct values as a ratio over the housekeeping gene Gapdh, meaning that lower ratios depict higher expression levels. n = 3 or 4 with each n representing 4 to 5 mice that were pooled together. Aged (46 wk) retired male and female breeders were used for pooling. Statistical significance was determined through a one-way Brown–Forsythe and Welch ANOVA. A t test was used to determine the significance of Pmp2 expression. (B) Uniform Manifold Approximation and Projection (UMAP) clustering of cardiac T cells. (C) Feature plots showing gene expression of Cd69, Pdcd1, Cxcr6, and S1pr1 in the various T cell clusters. (D) Bubble plot showing genes differentially expressed by distinct cardiac T cell populations. Statistical significance was assessed by a Wilcoxon signed-rank test. (E) Graphical depiction of the significantly up-regulated genes by each cardiac T cell subcluster.

To validate our RT-qPCR findings, we analyzed a publicly available single-cell RNA sequencing dataset of murine cardiac leukocytes (39). In this dataset, mice were induced with experimental autoimmune myocarditis (EAM), and at various stages of the disease, hearts were harvested, and leukocytes were sorted and run on a 10X Genomics Platform (39). Our analysis of this dataset demonstrated that cardiac T cells subclustered into T_N, T_CM, T_EM, and T_RM populations based on distinct gene expression profiles (Fig. 2B). We investigated the expression of T_RM cell markers such as CD69, PD-1, and CXCR6 and found that the expression of all these genes together were up-regulated in the cardiac T_RM cell subcluster (Fig. 2C). We then performed differential gene analysis and found that cardiac T_N cells exhibited the highest expression of S1pr1 and Klf2, which are responsible for tissue egress (Fig. 2 D and E and SI Appendix, Table S1). We found that cardiac T_RM cells displayed up-regulated expression of Runx3, a well-defined tissue-resident transcription factor that is important for T_RM cell development (40, 41) (Fig. 2 D and E). We then analyzed the expression of other immune checkpoint molecules, such as CTLA-4, TIGIT, and LAG3. We found that, unlike PD-1, the gene expression of the other checkpoint molecules was up-regulated by T_CM cells and not T_RM cells (Fig. 2 D and E). Subsequently, we examined effector cytokines and found that cardiac T_RM cells showed up-regulated gene expression of both Tnf and Il17a (Fig. 2 D and E). IL-17A has been implicated in negative cardiac remodeling during autoimmune myocarditis by our lab and others (42, 43). We then investigated the gene expression of cytokine receptors and found that cardiac T_RM cells up-regulated the gene expression of IL-2Rβ, a shared receptor for both IL-2 and IL-15 signaling (Fig. 2 D and E) (23, 44). IL-15 is part of the common γ-chain family of cytokines and is necessary for T_RM cell maintenance in the skin (23). Finally, to demonstrate that cardiac T_RM cells have a similar gene signature to other T_RM cells, we integrated our reanalysis of cardiac T_RM cells with a publicly available bulk RNA sequencing dataset of skin T_RM cells and spleen T_EM cells (17). We then developed a model that allowed us to ascribe a cardiac T_RM score to both the skin T_RM cells and the spleen T_EM cells. For this model, the higher the score, the more transcriptionally similar the sample is to the cardiac T_RM cells. We found that skin T_RM cells were more similar to cardiac T_RM cells than spleen T_EM cells (SI Appendix, Fig. S2D), indicating that cardiac T_RM cells share a gene signature that is more similar to other resident memory T cells than circulating T cells. Together, these findings indicate cardiac T_RM cells are transcriptionally distinct and up-regulate a T_RM-like signature.

Cardiac Injury and Aging Induce Autoreactive CD69⁺ T_RM Cells.

It has been shown that pathogen exposure or localized inflammation leads to the accumulation of T_RM cells in various tissues (22, 23, 45). This led us to hypothesize that cardiac injury would cause the induction of T_RM cells into the heart. To study this, we employed a modified EAM (mEAM) mouse model, which induces a low level of cardiac inflammation that mostly resolves after 8 wk (Fig. 3A). To characterize the phenotype of cardiac T cells throughout the model, we performed a time course experiment looking at T cells at weeks 0, 2, 3, and 8. To determine the phenotype of cardiac CD4⁺ T cells during mEAM, we first performed unbiased clustering through t-distributed stochastic neighbor embedding (t-SNE) (Fig. 3B). From the unsupervised clustering, we distinguished 9 clusters from which we assigned conventional T cell classifications (gating strategy in SI Appendix, Fig. S3A). We found that the cardiac T_RM cell cluster coincided with the highest expression of CD69, CD44, PD-1, and CXCR6 (Fig. 3B). We then stratified the CD4⁺ T cells based on the time course and determined that at day 0, the predominant populations were infiltrating T_N cells and T_EM cells (Fig. 3C). After week 8, we found that a majority of inflammation had abated and that there was a substantial increase in cardiac T_RM cells (Fig. 3C). These findings mirror our analysis of the publicly available conventional EAM single-cell RNA sequencing dataset, where cardiac T_RM cells are enriched on day 60 after EAM (SI Appendix, Fig. S3B) (39). We then quantified the proportion of CD69⁺PD-1⁺ T cells in the heart before and 8 wk after mEAM induction and found a significant increase in both CD4⁺ and CD8⁺ T_RM cells in mice who received mEAM (Fig. 3D and SI Appendix, Fig. S3C). We next investigated the phenotype of resident MyHC T cells in naive mice and mice who recovered from mEAM and found that all MyHC T cells expressed PD-1 and a significant portion expressed CD69 (Fig. 3E). Next, we inspected whether recovery from mEAM led to an increase in MyHC CD69⁺PD-1⁺ T_RM cells compared to naive mice and found a trending increase in CD4⁺ T_RM cells but not CD8⁺ T_RM cells (Fig. 3F and SI Appendix, Fig. S3D). To investigate the phenotype of the recruited MyHC T_RM cells, we induced mEAM and allowed mice to recover for 8 wk before staining for FoxP3 and CD25. Similar to our findings in naive A/J mice (Fig. 1 H and I), we found that a majority of induced MyHC T_RM cells were Tconv cells and not Tregs (SI Appendix, Fig. S3E). To determine whether other models of cardiac injury would lead to the induction of CD69⁺PD-1⁺ T_RM cells, we employed an ischemia–reperfusion (I/R) mouse model, where ischemia was induced for 60 min followed by a 7-wk recovery period (Fig. 3G). We found that mice who recovered from I/R injury had no difference in the number of CD69⁺PD-1⁺CD4⁺ T_RM cells compared to mice who received sham surgery but had a significant increase in CD69⁺ PD-1⁺CD8⁺ T_RM cells (Fig. 3H and SI Appendix, Fig. S3F). This indicates that the type of cardiac injury may influence whether CD4⁺ or CD8⁺ T cells are recruited to the heart and eventually mature into cardiac T_RM cells. It has been shown that T_RM cells accumulate in the peripheral organs, such as the skin, during aging (46, 47). To determine the effect of aging on the recruitment of T_RM cells into the heart, we performed intravascular labeling on young (15 wk) and aged (53 wk) naive A/J mice. We found that aged mice had a significant increase in CD8⁺ T_RM cells but not CD4⁺ T_RM cells compared to young A/J mice (Fig. 3I and SI Appendix, Fig. S3G). Collectively, we determined that cardiac injury and aging lead to the accumulation of CD69⁺ PD-1⁺ T_RM cells in the heart.

Fig. 3.

Cardiac injury induces autoreactive cardiac CD69⁺PD-1⁺ T_RM cells. (A) Graphical description of mEAM mouse model timeline. (B) A/J mice from all groups were concatenated into a single file before unsupervised t-SNE clustering was performed on the cardiac CD4⁺ T cells. T-SNE clusters were assigned conventional T cell populations based on the expression of CD62L, CD44, KLRG1, and CD69. Infiltrating T cells (CD45IV) and resident immune T cells (CD45R) were labeled. Heat map illustrating the expression of the core T_RM cell genes CD69, CD44, PD-1, and CXCR6. The experiment has a total of n = 12. (C) Time course of mEAM showing the distribution of CD4⁺ T cells on weeks 0, 2, 3, and 8. (D) mEAM was induced in A/J mice, and after 8 wk, the number of T_RM cells was assessed by flow cytometry. Quantifying the proportion of CD4⁺ and CD8⁺ T_RM cells after mEAM compared to naive A/J mice. Group size was n = 9 or 10 mice. Statistical significance was determined through a Welch t test. (E) Representative flow graph showing expression of PD-1 and TCR specificity for cardiac myosin on resident CD69⁻ and CD69⁺ CD4⁺ T cells. Both naive mice and mEAM mice were included in this analysis. Group size was n = 19. Statistical significance was determined through a paired t test. (F) Graph quantifying the proportion of autoreactive myosin-specific CD69⁺PD-1⁺CD4⁺ T_RM cells after mEAM. Group size was n = 9 or 10 mice. Statistical significance was determined through a Welch t test. (G) Graphical description of I/R mouse model timeline. (H) I/R injury was induced in male A/J mice; after 8 wk, the number of T_RM cells was assessed by flow cytometry. Quantifying the proportion of CD4⁺ and CD8⁺ T_RM cells after I/R injury compared to A/J mice who received sham surgery. Group size was n = 9. Statistical significance was determined through a Welch t test. (I) Quantifying the proportion of CD4⁺ and CD8⁺ T_RM cells in young mice (15 wk) to aged mice (53 wk). Group size was n = 13 or 11. Statistical significance was determined through a Welch t test.

Cardiac Injury Causes Accumulation of T_RM Cells Along the Perimyocardium.

To determine the location of CD69⁺ T_RM cells in the heart, we utilized the mEAM mouse model to induce CD69⁺ PD-1⁺ T_RM cells (Fig. 3A). After induction of the disease, we allowed mice to recover for 8 wk before assessing residual inflammation via histology. We observed that 40% of the mice that received mEAM developed modest perimyocarditis, especially around the right ventricle (Fig. 4A). We found no observable infiltration in other areas of the heart, such as the atria and the ventricular myocardium (Fig. 4A). To identify the location of T cells in the heart, we performed immunohistochemistry (IHC) and stained for CD3. We found that mEAM induced accumulation of T cells predominantly along the perimyocardium but also in the atria (Fig. 4B). To confirm that these T cells were cardiac T_RM cells, we performed IHC for PD-1, Runx3, and CXCR6 and found that a majority of the cells along the atria and the perimyocardium were stained for all the T_RM cell markers (Fig. 4C). Thus, we determined that cardiac injury induces the accumulation of cardiac T_RM cells within the atria and along the ventricular perimyocardium.

Fig. 4.

Cardiac injury causes accumulation of T_RM cells along the perimyocardium. (A) Representative H&E histology of a mouse heart induced with mEAM. (B) Immunohistochemistry (IHC) showing the distribution of CD3⁺ T cells on representative histology of murine atria, myocardium, and pericardium in naive mice or mice who have recovered from mEAM. (C) IHC showing staining of CD3, PD-1, Runx3, and CXCR6 on representative histology from mice who recovered from mEAM.

Previous Cardiac Injury Increases the Susceptibility of Mice to Develop ICI-Myocarditis.

ICI-myocarditis is a rare but fatal adverse event caused by treatment with antibodies binding to PD-1 or PD-L1 (7, 9). We show that cardiac injury induces autoreactive MyHC T_RM cells that express high levels of PD-1 (Fig. 3 E and F). These findings led us to hypothesize that autoreactive MyHC T_RM cells are susceptible to activation by αPD-1 antibodies and go on to mediate ICI-myocarditis. To test this hypothesis, we developed a double-hit ICI-myocarditis mouse model (Fig. 5A). In this model, we first induce autoreactive MyHC T_RM cells through mEAM and then allow mice to recover for 8 wk to resolve cardiac inflammation before starting treatment with αPD-1 antibodies (Fig. 5A). To assess the severity of myocarditis, we evaluated heart tissue samples stained with H&E and found that mice who received mEAM + αPD-1 antibodies had a higher severity of cardiac inflammation compared to mice treated with αPD-1 antibodies alone (Fig. 5 B and C). Next, we assessed the incidence of atrial and ventricular myocarditis separately. We found that mice who received mEAM + αPD-1 antibodies had an atrial myocarditis incidence rate of 62.5% compared to 20% in the control groups (Fig. 5D). When assessing the incidence of ventricular myocarditis, we found that mice who received mEAM + αPD-1 antibodies had an incidence rate of 67.3% (Fig. 5D). In comparison, none of the mice who received αPD-1 antibodies alone went on to develop ventricular myocarditis, while 50% of the mice who received mEAM+Isotype antibodies had low levels of remaining ventricular myocarditis (Fig. 5D). Next, we characterized the immune profile of cardiac inflammation by flow cytometry. We found that mice who received mEAM + αPD-1 antibodies had a trending increase in cardiac leukocytes, NK cells, and γδT cells (Fig. 5E and SI Appendix, Fig. S4A). Our lab and others have shown that the disease pathogenesis of ICI-myocarditis is driven by CD8⁺ T cells (11, 12). We found that mice who received mEAM followed by αPD-1 antibodies had a significant increase in CD3⁺ T cells, specifically CD8⁺ T cells (Fig. 5 F and G and SI Appendix, Fig. S4B). To investigate whether other types of cardiac injury could increase susceptibility to ICI-myocarditis development, we induced the recruitment of T_RM cells through our I/R mouse model before starting αPD-1 antibodies (Fig. 5H). We then assessed the severity of cardiac inflammation through histology and found that mice receiving I/R + αPD-1 antibodies had increased immune infiltration compared to those receiving αPD-1 alone (Fig. 5I). Next, we assessed the incidence of ICI-myocarditis. To do this, we separately examined the atria and nonischemic regions of the ventricular myocardium. We found that mice receiving I/R + αPD-1 antibodies had a higher incidence of atrial myocarditis compared to control groups (Fig. 5J). Finally, we assessed ventricular myocarditis; to do this, we scored the incidence of myocarditis outside the zone of infarction. We found that mice who received I/R + αPD-1 antibodies had a myocarditis incidence of 41.7%. In contrast, mice who received sham surgery followed by αPD-1 antibodies had an incidence rate of 25%, while mice who received I/R+ Isotype antibodies had an incidence rate of 18.2% (Fig. 5J). Our findings indicate that previous cardiac injury induces autoreactive PD-1⁺ T_RM cells in the heart, which increases the susceptibility of mice to develop ICI-myocarditis after PD-1/PD-L1 blockade.

Fig. 5.

Previous cardiac injury increases the susceptibility of mice to develop ICI-myocarditis. (A) Graphical description of ICI-myocarditis mouse model where cardiac T_RM cells were induced in A/J mice through mEAM. (B) Representative H&E histology of mice hearts, where mice received αPD-1, EAM+Isotype, or EAM + αPD-1. (C) Histological assessment was used to determine the severity of myocarditis in mice. Group size was n = 8 or 9. Statistical significance was determined by the Kruskal–Wallis test. (D) Pie charts showing the incidence of atrial and ventricular myocarditis between. (E–G) Bar graphs showing the proportion of cardiac leukocytes, CD3⁺ T cells, and CD8⁺ T cells as determined by flow cytometry. Group size was n = 8 or 9. Statistical significance was determined by the Brown–Forsythe and Welch ANOVA test. (H) Graphical description of ICI-myocarditis mouse model where cardiac T_RM cells are induced by I/R injury in male A/J mice. (I) Histological assessment was used to determine the severity of myocarditis in mice. Group size was n = 9, 11, or 12. Statistical significance was determined by the Kruskal–Wallis test. (J) Pie charts showing the incidence of atrial and ventricular myocarditis.

Human Pericardial Fluid Is Enriched with CD69⁺PD-1⁺ T_RM Cells.

Like mice, human cardiac T_RM cells have yet to be characterized. Our investigation of murine cardiac T_RM cells has demonstrated that cardiac injury leads to the recruitment of CD69⁺PD-1⁺ MyHC T_RM cells (Fig. 3F), defined by their location along the perimyocardium (Fig. 4 B and C). To test whether chronic cardiac inflammation leads to the accumulation of human T_RM cells in the pericardial cavity, we examined blood and pericardial fluid samples from patients with ischemic cardiomyopathy (iCMP) and with dilated cardiomyopathy (DCM), where genetic causes were not identified. We chose to research DCM patients since one-third of patients diagnosed with myocarditis develop DCM (42, 48, 49). The development of ischemic cardiomyopathy (iCMP) is a chronic condition caused by ischemic injury and accompanying cardiac inflammation (50). Thus, DCM and iCMP are similar to the two types of cardiac injury we investigated with our murine models. We received pericardial fluid and blood samples from patients undergoing cardiac catheterization at the Institute of Clinical and Experimental Medicine (IKEM) in the Czech Republic (Fig. 6A and SI Appendix, Table S2). First, we assessed the distribution of CD4⁺ and CD8⁺ T cells in the pericardial fluid and PBMCs and found no difference in the CD4⁺:CD8⁺ ratio based on T cell location or disease diagnosis (SI Appendix, Fig. S5 A and B). To assess the phenotype of pericardial T cells, we performed unbiased clustering through t-SNE and assigned T cell subsets based on conventional markers such as CD45RA, CCR7, and CD69 (SI Appendix, Fig. S5C). We found that CD4⁺ T cells subclustered into distinct biologically relevant populations and that pericardial T_RM cells grouped separately from the other T cell subtypes (Fig. 6B). We determined that T_RM cells were absent in PBMCs but were the predominant population in pericardial fluid (Fig. 6B). Next, we analyzed the expression of CD69 and PD-1 and found that they were predominately expressed by pericardial T_RM cells compared to circulating T cells (Fig. 6 B–D). Similar to our previous finding (SI Appendix, Fig. S5B), we determined that the patient’s cardiovascular diagnosis did not affect the proportion of CD69⁺ T_RM cells in pericardial fluid (SI Appendix, Fig. S5D). We found that both DCM and iCMP patients had on average, greater than 80% of their CCR7^-CD45RA⁻ T cells expressing CD69 in both CD4⁺ and CD8⁺ T cells (SI Appendix, Fig. S5D). We showed that murine T_RM cells up-regulated PD-1 compared to other cardiac T cell populations (Figs. 1E and 2D), therefore, we assessed the expression of PD-1 on human pericardial T cells and found that, on average, 74% of CD4⁺ pericardial T_RM cells expressed PD-1, compared to only 11% of T_N cells, 38% of T_CM cells and 47% of T_EM cells (Fig. 6 E and F). When accessing the expression of PD-1 on pericardial CD8⁺ T_RM cells, we found that, on average, 91% of CD8⁺ T_RM cells expressed PD-1, which was higher when compared to other CD8⁺ pericardial T cell populations (Fig. 6F). Next, we investigated the cytokine profile of DCM and iCMP patient pericardial fluid through a 48-cytokine multiplex (Millipore Sigma). We investigated the concentration of IL-2, IL-10, IL-15, TNF, IFN-γ, IL-17A, and IL-17F and found that only IL-10 and IL-15 were up-regulated in pericardial fluid (Fig. 6G and SI Appendix, Fig. S5E). IL-15 is necessary for the maintenance of T_RM cells in the kidney and skin (23, 44), thus, the presence of IL-15 may be required for pericardial T_RM cell survival. We ran a linear regression analysis and found that the concentration of pericardial IL-15 positively correlated with the number of pericardial T cells (Fig. 6H). IL-15 is produced by a variety of nonhemopoietic and hemopoietic cells, including epithelial cells, dendritic cells, and macrophages (51–53). It has been shown that resident myeloid cells can support T_RM cell survival in the peripheral tissues through local production of chemokines and cytokines (22, 44). The pericardial cavity is known to be a reservoir for resident cardiac GATA6⁺ macrophages (54–56), which may be supporting resident T_RM cell population through the local production of IL-15 and IL-10. To assess the source of pericardial IL-15 and IL-10, we performed single-nucleus RNA sequencing on pericardial cells from patients diagnosed with iCMP (57). We then evaluated the gene transcript for both IL-15 and IL-10 and found that they were up-regulated by pericardial macrophages compared to pericardial NK and T cells (Fig. 6I). Therefore, pericardial IL-15 and IL-10 are most likely secreted by pericardial macrophages. Together, these findings demonstrate that DCM and iCMP patients have pericardial T_RM cells that express CD69 and PD-1, which are supported by pericardial macrophages through the production of IL-15. Collectively, our findings indicate that human and murine cardiac T_RM cells can be found in the pericardial cavity and share a similar CD69⁺PD-1⁺ phenotype.

Fig. 6.

Human pericardial fluid is enriched with CD69⁺PD-1⁺ T_RM cells. (A) Graphical depiction of deidentified human patient samples received from the biorepository at the Institute for Clinical and Experimental Medicine (IKEM). Samples were collected from patients diagnosed with either dilated cardiomyopathy (DCM) or ischemic cardiomyopathy (iCMP). (B) Patient PBMCs and pericardial cells concatenated into a single file before unsupervised t-SNE clustering was performed on the CD4⁺ T cells. Heat mapping illustrating CD69 and PD-1 expression on CD4⁺ T cells from patient PBMCs and pericardial cells. (C and D) Expression of CD69 and PD-1 on CD4⁺ and CD8⁺ T cells on patient PBMCs and pericardial cells. Group size was n = 14. Statistical significance was determined by the Wilcoxon test. (E) Representative flow plots showing the expression of PD-1 on T_N, T_CM, T_EM, and T_RM pericardial CD4⁺ T cells. (F) Expression of PD-1 on different pericardial T cell populations. Group size was n = 14. Statistical significance was determined by the RM one-way ANOVA. (G) Comparing the concentrations of cytokines in human pericardial fluid compared to matched patient plasma. Group size was n = 10. Statistical significance was determined by the Wilcoxon test. (H) Simple linear regression analysis comparing the pericardial fluid concentration of IL-15 to the proportion of CD3⁺ T cells (from lineage⁻ gate). Group size was n = 14. (I) Bubble plot showing gene expression of IL-15 and IL-10 by pericardial macrophages and NK and T cells.

Discussion

T_RM cells are noncirculating lymphocytes that coordinate rapid effector functions upon stimulation (14, 15, 25). T_RM cells are phenotypically and transcriptionally distinct cell types, which can be identified through classical markers such as CD69 and CD103 (14, 15, 17). We first set out to characterize T_RM cells in the hearts of naive mice. Compared to circulating lymphocytes, we found that cardiac memory T cells up-regulated CD69 but not CD103. CD69 has also been shown to distinguish resident memory T cells across various peripheral tissues and is expressed by T_RM cells in barrier organs such as the skin and nonbarrier organs such as the salivary gland, pancreas, and liver (17, 22, 28, 30, 44). CD69 is an important marker for defining T_RM cells since it facilitates tissue retention by antagonizing egress proteins (18, 19). CD103 is an integrin that binds E-cadherin and is predominately expressed by T_RM cells in epithelial tissues such as the lung and the skin (58–60). To further characterize cardiac T_RM cells, we transcriptionally profiled cardiac T cells from naive mice and mice who received EAM. We found that cardiac T_RM cells up-regulated the crucial transcription factor Runx3, which is important for the upregulation of tissue retention genes in both murine and human T_RM cells (40, 41). Transcriptional reprogramming by Runx3 is important for T_RM cell recruitment and maintenance in various tissues, such as the intestines and lungs (40, 41). Transcriptional reprogramming of memory T_RM cells is associated with altered metabolism, specifically the upregulation of fatty acid oxidation (37, 38). Furthermore, T_RM cells have been shown to up-regulate tissue-specific isoforms of the FABPs, which facilitate the uptake of lipids by resident immune cells (37, 38). We found the cardiac CD69⁺ T_RM cells up-regulated three of the nine FABP isoforms, FAPB3, FABP4, and Pmp2. Upregulation of specific FABP isoforms has been shown to be crucial for T_RM cell function and survival in the peripheral tissues; for example, FABP4 and FAPB5 are important for skin T_RM cells, while FABP1 is necessary for liver T_RM cells (37, 38). Thus, we were able to conclusively demonstrate that cardiac T_RM cells are present in naive mice and can be distinguished from circulating T cells by their expression of the classical T_RM marker CD69, the transcription factor Runx3, and the metabolic proteins FABP3, FABP4, and Pmp2.

To further characterize cardiac T_RM cells, we used a comprehensive approach that included transcriptional profiling and intravascular labeling to determine the residency of CD69⁺ cardiac T_RM cells. To determine whether cardiac T_RM cells featured transcriptional changes that would support residency in the peripheral tissues, we profiled cardiac T cells from naive mice and mice who received EAM. We found cardiac T_RM cells down-regulated the tissue egress protein S1PR1 and the transcription factor KLF2. Vertebrate animals contain five distinct receptors for S1P, which aid in embryogenesis, cell signaling, and lymphocyte egress (19, 61, 62). Downregulation of S1PR1 is necessary for the formation and maintenance of both human and murine T_RM cells (17, 18). KLF2 is a zinc finger transcription factor that stimulates lymphocyte egress by promoting the transcription of S1PR1 (63). Downregulation of KLF2 promotes T_RM formation within peripheral tissues and has been documented in a variety of organs, such as the kidney, brain, and salivary gland (14). To further investigate the residency of cardiac CD69⁺ T_RM cells, we employed IV labeling to distinguish cells trapped within the vasculature and those occupying the tissue. We confirmed that the expression of CD69 was exclusive to resident T cells compared to lymphocytes trapped within the cardiac vasculature. It has been shown that the perfusion of organs after killing of animals is insufficient to remove circulating immune cells, which can confound the identification and characterization of resident leukocytes (35). Intravascular labeling is an effective method to discriminate the location and residency of murine T_RM cells in peripheral tissues such as the lung, kidney, and skin (31, 35, 64). Therefore, we were able to demonstrate, through transcriptional analysis and intravascular labeling, that cardiac CD69⁺ T cells are conclusively resident memory T cells.

Autoimmune MyHC T cells escape negative selection in the thymus and can be found in the periphery of naive mice and humans (32). Our previous studies have characterized MyHC T cells in naive murine hearts and found that they demonstrate an antigen-experienced phenotype as well as up-regulate CD69 (11). To determine whether MyHC T cells are cardiac resident, we combined intravascular labeling with tetramer staining. We found that a significant proportion of MyHC T cells were cardiac resident and up-regulated CD69. Autoimmune T cells have been identified within healthy-appearing tissues and have been proven to exhibit a T_RM-like signature (15, 65). Furthermore, T_RM cells have been implicated in various autoimmune diseases such as vitiligo and psoriasis (15, 28, 65). During inflammation, activated T cells are primed in the secondary lymphoid organs before being recruited into the peripheral tissues (66, 67). After the resolution of inflammation, there is a significant contraction, and a small population of mature T_RM cells remain (66, 67). In the heart, it has been shown that MyHC Tregs are recruited during the acute phase of ischemic injury and play a protective role during cardiac remodeling (33, 34). To determine how myocardial damage would affect the autoimmune MyHC T_RM cells population, we employed the mEAM and I/R injury mouse models to induce cardiac damage. We found that cardiac injury induced an increase in autoimmune CD69⁺ MyHC T_RM cells. Phenotyping of these T_RM cells demonstrated that they appear predominantly to be conventional T cells and not Tregs. However, due to the limitations of flow cytometry in stratifying rare populations, we can’t exclude the possibility that some MyHC T_RM cells are Tregs. Recruitment of antigen-specific T_RM cells has been documented in a variety of murine tissues following an infection, vaccination, or injury (22, 23, 31, 64). It has been demonstrated that antigen-specific T_RM cells maintain rapid effector function upon restimulation, which can be protective against subsequent viral challenges or pathogenic in the case of allergen-induced asthma (22, 68). Aging is associated with reduced thymopoiesis and increased recruitment of T_RM cells into the peripheral tissues (46). To determine the effect of aging on MyHC T_RM cells, we performed intravascular labeling on naive young and aged mice. We found that aged mice had an increased proportion of CD8⁺CD69⁺ T_RM cells. It has been shown that aged mice accumulate CD8⁺ T_RM cells in the lung, which induces chronic inflammation, leading to tissue fibrosis (69). In addition, aging is associated with a reduction of cardiac function, which has been linked to T cell accumulation in the mediastinal lymph nodes (70, 71). Overall, we were able to establish that MyHC T cells are cardiac T_RM cells, which expand after cardiac injury and accumulate in the heart during aging.

T_RM cells are highly activated resident immune cells that up-regulate PD-1 in a variety of tissues, such as the lung, skin, and pancreas (14, 30, 31). It has been shown that the PD-1/PD-L1 axis is a critical negative regulator of T_RM cells, which helps maintain immune homeostasis and prevents tissue fibrosis (30, 31). To further characterize MyHC T_RM cells, we examined the expression of PD-1 and found that MyHC T cells significantly up-regulated both CD69 and PD-1. Although PD-1 expression is typically associated with loss of effector function, a hallmark of T cell exhaustion, T_RM cells maintain effector polyfunctionality despite high levels of PD-1 expression (30, 31). ICI-myocarditis is a highly fatal irAE, which occurs rapidly after beginning immune checkpoint blocking therapies and predominantly affects patients treated with immunotherapies targeting the PD-1/PD-L1 axis (7, 8). The majority of patients develop ICI-myocarditis after only 1 to 2 treatments with a median onset time of 34 d (7, 9, 10, 72). The rapid development of the disease indicates that certain patients may be predisposed to develop ICI-myocarditis. Risk factors for ICI-myocarditis development include the combination of ICI therapies, age, previous cardiovascular disease, and diagnosis of an autoimmune disease (7, 72). To determine whether resident autoimmune CD69⁺ MyHC T_RM cells are responsible for ICI-myocarditis, we created mouse models where we induced MyHC T_RM cells either through mEAM or I/R injury and started treatment with αPD-1 antibodies after recovery from these cardiac injuries. We determined that mice with more CD69⁺ MyHC T_RM cells had a greater incidence of ICI-myocarditis development, as well as significant infiltration of CD8⁺ T cells. T_RM-like TILs have been described across various cancers and have been shown to highly up-regulate PD-1 as well as other immune checkpoint molecules (73, 74). Treatment with immune checkpoint therapies directly activate and expand T_RM-like TILs, which is clinically beneficial and is associated with a robust antitumor response in certain cancers like melanoma (73, 74). Thus, we demonstrated that ICI-myocarditis is mediated by MyHC T_RM cells that are enriched in the heart after cardiac injury.

Resident cardiac macrophages have been shown to be distributed throughout the heart (70, 75). In certain incidences, the cellular location has been linked to a specific function; for example, cardiac macrophages cluster around the atrioventricular bundle and provide support for electrical conduction (76). We determined that injury-induced cardiac T_RM cells were preferentially located within the atria and along the perimyocardium. To investigate the location of human cardiac T_RM cells, we received pericardial fluid samples from a biorepository at the Institute for Clinical and Experimental Medicine (IKEM) in the Czech Republic. For our study, we included patients diagnosed with dilated cardiomyopathy (DCM) and ischemic cardiomyopathy (iCMP). DCM is characterized by the dilation of the left ventricle, which causes impaired ventricular contraction and leads to reduced cardiac output (77, 78). Approximately one-third of patients diagnosed with myocarditis go on to develop DCM (77–79). iCMP is a chronic condition initiated through myocardial damage caused by ischemia, leading to impaired cardiac function (50, 80). We found that a majority of pericardial T cells express a resident memory phenotype and up-regulate CD69 and PD-1. In humans, it is known that T cells are key components of the inflammatory milieu recruited to the heart during myocardial inflammation (43, 81, 82). In mice, we found that cardiac injury induces autoreactive MyHC CD69⁺PD-1⁺ T_RM cells that preferentially locate along the perimyocardium. We determined CD69⁺PD-1⁺ T_RM cells are the predominant T cell population in the pericardial cavity in patients diagnosed with DCM and iCMP, which supports our finding that cardiac injury induced by mEAM and I/R injury induces autoreactive MyHC CD69⁺PD-1⁺ T_RM cells in mice. Our findings indicate that regardless of disease etiology, myocardial inflammation leads to the recruitment of pericardial PD-1⁺ T_RM cells. Finally, we determined that human T_RM cells have the highest expression of PD-1 compared to other pericardial T cell populations. Our finding that a majority of pericardial T_RM cells express PD-1 parallels our investigation of murine cardiac T_RM cells and suggests that human pericardial T_RM cells may also be susceptible to activation by αPD-1 antibodies.

The pericardial cavity hosts a variety of different leukocytes, such as GATA6⁺ resident macrophages and B cells (54, 55). In mice, T_RM cells in the female genital tract have been shown to form lymphocyte clusters with resident macrophages. This has been determined to be important for T_RM cell retention since macrophages provide support through chemokine production (22). To determine whether the microenvironment of the pericardial cavity could support T_RM cells, we performed a multiplex on pericardial fluid and plasma isolated from human patients. We found that human pericardial fluid was enriched for IL-15, which shares the same common γ-chain receptor as IL-2 (83). In our reanalysis of the publicly available single-cell RNA sequencing dataset, we found that murine cardiac T_RM cells differentially up-regulated the gene expression of the IL-2 receptor β-chain, indicating that cardiac T_RM cells are dependent on IL-15 singling. IL-15 has been shown to be necessary for T_RM survival and maintenance in a variety of tissues, such as the skin and kidney (23, 44, 53). To determine the source of human pericardial IL-15, we performed single-nucleus RNA sequencing on pericardial immune cells from iCMP patients and found that pericardial macrophages were the main source of IL-15. This indicates that resident pericardial macrophages may be supporting the survival of pericardial T_RM cells, which may explain why murine and human CD69⁺PD-1⁺ T_RM cells preferentially locate along the perimyocardium. Collectively, our findings indicate that human and murine cardiac T_RM cells share a prominent CD69⁺PD-1⁺ phenotype and reside in the same location within the heart.

In summary, we were able to determine that CD69 is differentially expressed by cardiac tissue resident memory T cells, which are phenotypically and transcriptionally distinct from circulating lymphocytes. We confirmed that murine autoreactive MyHC T cells are present in naive mouse hearts and demonstrated that MyHC T cells express CD69 and are tissue resident memory T cells. It is poorly understood why MyHC T_RM cells are recruited to the heart and what function they play during immune homeostasis, which should be the focus of future studies. MyHC T_RM cells are highly activated cells that up-regulate PD-1 and are recruited to the heart after cardiac injury or during aging. We were able to show that cardiac T_RM cells can be found within the atria and along the ventricular perimyocaridum. However, how T_RM cells localize to the pericardial membrane remains uncertain. Finally, we demonstrated with our double-hit ICI-myocarditis mouse models that mice with more MyHC T_RM cells developed ICI-myocarditis more frequently and acquired more severe disease. Finally, we demonstrate the high translatability of our characterization of murine T_RM cells because we found CD69⁺PD-1⁺ T_RM cells are the predominant T cell population in pericardial fluid isolated from patients with DCM or iCMP. This study provides evidence that ICI-myocarditis is directly mediated by autoreactive MyHC T_RM cells and provides a mechanism for why age and previous cardiac injury are risk factors for ICI-myocarditis development.

Materials and Methods

Study Design.

This study aimed to describe cardiac resident memory T cells and their contribution to the disease pathogenesis of ICI-myocarditis. In this study, we assessed the impact of aging and cardiac injury on the recruitment and retention of cardiac T_RM cells. Cardiac injury was induced either through a modified experimental autoimmune myocarditis model or an ischemia–reperfusion injury model. A detailed description of the experimental design, sample size, and statistical test can be found in the figure legends. All experiments were repeated as independent experiments two to three times unless otherwise stated in the figure legend.

Mice.

Wild-type A/J breeders (strain#000646) were purchased from the Jackson Laboratory. Mice included in studies were female unless otherwise stated. All animals were housed in specific pathogen-free conditions at a Johns Hopkins University animal facility. Experiments were conducted on animals 8 to 12 wk old unless otherwise stated in the figure legend. All animal experiments were conducted under protocols reviewed and approved by the Animal Care and Use Committee at Johns Hopkins University.

Cell Isolation and Flow Cytometry.

Hearts were perfused through the aorta with PBS for 3 min before being removed and chopped into small pieces. Hearts were digested with Collagenase II and Dnase I for 30 min at 37 °C on a shaker. Tissue was dissociated and placed through a 70 μm cell strainer. Cells were washed and transferred to a 96-well plate for staining. Cells were blocked with anti-CD16/32 and sera for 30 min at 4 °C. Cells were stained with primary antibodies at 4 °C for 30 min. Cells were fixed overnight with BD Cytofix before being acquired on an LSRFortessa flow cytometer (BD), and data were analyzed using FlowJo software (BD). Detailed instructions including heart tissue processing, tetramer staining, and intravascular labeling can be found in SI Appendix, Supplemental Methods, and a list of antibodies and concentrations can be found in (SI Appendix, Table S3).

qRT-PCR.

A detailed description of how cells were prepared and sorted can be found in SI Appendix, Supplemental Methods. RNA was extracted from sorted cells using TRIzol reagent (Life Technologies, Cat#15596018). RNA was normalized before cDNA was synthesized using iScript Reverse Transcript Supermix (Bio-Rad, Cat#1708841). Genes of interest were amplified using PowerUp SYBER Green Master Mix (ThermoFisher, Cat#10029284). The samples were run on a MyiQ2 thermal cycler (Bio-Rad). The data were analyzed by comparing the Ct value of the gene of interest to the Ct value of the housekeeping gene Gapdh. Primers used in the study can be found in (SI Appendix, Table S4).

Murine Single-Cell RNA Sequencing Analysis.

The publicly available experimental autoimmune myocarditis single-cell RNA sequencing dataset was retrieved using the accession code: GSE142564 (39). Cells from different days were combined using the merge command. Following regularized negative binomial regression normalization, cells were clustered via UMAP nonlinear dimensional reduction to allow biological cluster annotation. Clusters identified to be T cells were isolated and subclustered. T_EF were identified but were excluded from subsequent differential gene analysis. A detailed description of our quality control and subsequent analysis can be found in SI Appendix, Supplemental Methods.

Human Pericardial Single-Nucleus Isolation, Library Generation, Sequencing, and Analysis.

Frozen cells were thawed and washed with culture media. Cells were treated with DNase for 5 min on ice. Pericardial cells were incubated in a lysis buffer on ice for 10 min. Nuclei were washed with a wash buffer and spun down. Nuclei were resuspended in chilled nuclei buffer and stained with TotalSeq™-B anti-human Hashtag antibodies for 20 min on ice. A detailed description of lysis, wash, and nuclei buffers is available in (SI Appendix, Table S5). 3’ HT single-nucleus RNA sequencing with barcoded cell hashing was performed using the 10x Genomics platform. Data were analyzed using the Seurat (v4.3.0) workflow in R (v4.3.0). Cells with nFeature_RNA > 250, nFeature_RNA < 7,000, nCount_RNA < 80,000, and percent.mt < 27) were examined. Following QC, 9367 cells were retained. Cells were normalized using the SCTransform method and then reduced via PCA and UMAP. Cells were clustered and separated into biologically meaningful populations. Gene expression was visualized using DotPlot() using the log-normalized gene expression count matrix.

Modified Experimental Autoimmune Myocarditis (mEAM) Induction.

To induce mEAM, A/J mice were immunized with two 100 μg doses of the heavy α-chain of cardiac myosin peptide (MyHCα_334−352 DSAFDVLSFTAEEKAGVYK), which was purchased from GenScript. For the first immunization, the cardiac myosin peptide was emulsified in Complete Freud’s Adjuvant (Sigma-Aldrich, Cat#F5881), which had been supplemented with 40 mg of heat-killed Mycobacterium tuberculosis (Difco, Cat#DF3114-33-8). During the first immunization, mice were also administered 500 ng of pertussis toxin (List Biological Laboratories, Cat#180) through an IP injection. For the second immunization, the cardiac myosin peptide was emulsified in Incomplete Freud’s Adjuvant (Sigma-Aldrich, Cat#F5506). Immunizations with myosin peptide occurred on Day 0 and Day 7, and both immunizations were administered SC.

Ischemia–Reperfusion (I/R) Mouse Model.

Ischemia–reperfusion injury was induced in a two-step manner as previously published (84). A detailed description of the procedure can be found in SI Appendix, Supplemental Methods. Mice were intubated and connected to a respirator (Mini Vent 845) and anesthetized with isoflurane. Hair over the chest cavity was removed, the chest opened, a lateral thoracotomy was performed, and the rib cage opened. Using an 8-0 monofilament suture, an occlusion device (PE-10 tubing) was placed at 2 mm from the auricle. The ribcage was then sutured closed, and the remaining string from the 8-0 suture was stored under the skin. Mice were allowed to recover for 2 wk between surgeries. Mice were anesthetized with Ketamine and further anesthetized with 1 to 3% isoflurane. The skin was opened, and the string attached to the occluding device was taped to a 5 g lead weight. The lead weight was then hung on a bearing, allowing compression of the left anterior descending artery for 60 min. Afterward, the weight was removed, and the skin was closed with an 8-0 suture.

Cardiac Histology Assessment.

Cardiac H&E slides were evaluated for immune infiltration with the following criteria. Score 0, no inflammation; score 1, <10%; score 2, 10 to 30%; score 3, 30 to 50%; score 4, 50 to 90%; and score 5, >90%. The atria, ventricular myocardium, and pericardium were evaluated separately. Slides were scored independently by two blinded investigators.

Immunohistochemistry.

Unstained slides were baked at 60 °C for 20 min. Slides were placed in Trilogy (Cell Marque, Cat#920P-06) and cooked in a pressure cooker for 20 min. Slides were cooled and rinsed in distilled water. Slides were incubated for 5 min with the dual endogenous enzyme block (Dako, Cat#S2003) followed by a 1-h incubation in the primary antibody. Slides were rinsed before being incubated with the 2° antibodies (Leica, Cat#PV6119 or Cat#PV6114) for 30 min. Slides were rinsed before incubating in DAB chromogen (Vector Laboratories, Cat#SK-4100) for 4 min. Slides were rinsed before being counterstained with Hematoxylin 1 (ThermoFisher, Cat#72221). Slides were rinsed and then placed in the Bluing reagent (Epredia, Cat#6769001) for 1 min before a sequential dehydration in ethanol. Slides were mounted with OPTICMOUNT 1 (Mercedes Scientific, Cat#MER7720) and left to cure overnight. A detailed list of antibodies and concentrations can be found in SI Appendix, Supplemental Methods.

Two-Hit ICI-Myocarditis Induction.

T_RM cells were recruited to the heart either through mEAM or I/R injury. After the cardiac injury, mice were allowed to recover, and inflammation was allowed to resolve before beginning immune checkpoint inhibitor blocking therapies. Mice were treated IP with 150 μg of either αPD-1 antibodies (BioXCell, Cat#BE0146) or isotype control antibodies (BioXCell, Cat#BE0089) as previously published (11). Mice were treated biweekly for three weeks.

Human Samples.

We obtained human pericardial fluid samples and PBMC from a biorepository at the Institute of Clinical and Experimental Medicine (IKEM) in the Czech Republic. Samples were collected at the institute, deidentified, and shipped to our lab at Johns Hopkins University. Upon receipt, the frozen samples were stored at −80 °C. We determined the cytokine composition of patient plasma and pericardial fluid through a 48-cytokine Multiplex (Milipore, Cat#HCYTA-60 K-PX48).

Statistical Analysis

Statistical significance analysis was performed by GraphPad Prism 10 software. Data were analyzed using a variety of tests such as the paired t test, Welch t test, paired one-way ANOVA with Geisser-Greenhouse corrections, Wilcoxon signed-rank test, simple linear regression, Brown–Forsythe and Welch ANOVA, and Kruskal–Wallis test. The statistical test used for each experiment has been included in all the figure legends. Numerical P values were included in the figures or labeled as nonsignificant (ns). Data are presented as the mean ± SD, and symbols represent individual samples, unless otherwise stated.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Original single-nucleus data have been deposited in figshare (https://figshare.com/s/a3c29c74f375c9ae6025) (57). All study data are included in the article and/or SI Appendix.