Fibrotic Rewiring of CD8⁺ T Cells by Doxorubicin Drives Pathological Remodeling of Cardiac Extracellular Matrix

Abstract

Background

Anthracycline chemotherapy can lead to cardiotoxicity and chronic heart failure, and specific preventative therapies are nonexistent to date. As such, mechanistic investigation is needed to develop cardioprotective strategies. We previously showed that the anthracycline doxorubicin elicits a cardiotropic CD8⁺ T-cell response that contributes to cardiac fibroblast transformation, fibrosis, and chronic cardiotoxicity. However, the exact molecular mechanisms of T-cell–induced fibrosis, and whether T cells can directly induce cardiac remodeling, remain unknown.

Objectives

We sought to investigate T-cell transcriptomes from anthracycline-treated patients, mice, and cultured T cells to better understand mechanisms by which T cells influence adverse cardiac remodeling upon doxorubicin treatment across species.

Methods

We combined bulk RNA sequencing of human circulating CD8⁺ T cells from patients before or 3 months after anthracycline treatment with single-cell RNA sequencing of murine doxorubicin- or PBS-treated cardiac CD8⁺ T cells. Primary murine T cells were treated with doxorubicin in vitro for molecular mechanistic studies, and ex vivo mechanical characterization was performed on cardiac extracellular matrix (ECM).

Results

Anthracycline treatment induces novel profibrotic signature characterized by higher expression of ECM and ECM-crosslinking proteins in human and murine CD8⁺ T cells, which we term “matrix remodeling T cells” (mxT-cells). Functional mechanical testing of cardiac ECM revealed that mxT-cells stiffen cardiac ECM through T-cell lysyl oxidase (LOX) activity. Mechanistically, doxorubicin activates HIF1α-P53 signaling in mxT-cells, inducing autocrine TGFβ1 that enhances LOX and results in ECM crosslinking and elevated cardiac stiffness.

Conclusions

Our data demonstrate that doxorubicin imprints a novel intrinsic profibrotic function in CD8⁺ T cells responsible for ECM crosslinking, fibrosis, and cardiac remodeling.

Central Illustration

Anthracycline treatment, including doxorubicin, is a widely effective chemotherapeutic strategy for cancer patients that is limited by dose-dependent cardiac toxicity, often leading to chronic heart failure (HF).^1,2 Patients develop similar hallmarks of HF seen in other etiologies, including systolic dysfunction, inflammation, and cardiac fibrosis. However, the underlying mechanisms that may explain the limited efficacy of standard HF treatments to prevent cardiotoxicity in these patients remain poorly understood.3, 4, 5 Alongside an innate immune response, including neutrophil and macrophage activation,5, 6, 7 our group has shown that doxorubicin induces a cardiac CD8⁺ T cell response conserved across humans, dogs, and mice, which is required for systolic dysfunction, fibrosis, and chronic HF.⁸

We previously reported that cytotoxic CD8⁺ T cells treated with doxorubicin in vivo and in vitro upregulate IFNγ expression and cytotoxic granzyme degranulation, which contributes to profibrotic fibroblast transformation and cardiac fibrosis.^8,9 Doxorubicin directly acts on fibroblasts and induces activation, priming fibroblasts to interact with cytotoxic T cells,^8,10 a cellular communication axis central to doxorubicin-induced adverse remodeling. Cardiac fibrosis and ventricular remodeling are regulated not just by extracellular matrix (ECM) deposition, but also by enzymes that crosslink collagen and thereby enhance its tensile strength, including lysyl oxidases.^11,12 CD8⁺ T-cell–derived lysyl oxidation contributes to metastatic niche formation in the lung through ECM remodeling in paclitaxel chemotherapy.¹³ This suggests that CD8⁺ T cells may contribute to more complex fibrotic mechanisms than previously known. We tested the hypothesis that doxorubicin triggers proinflammatory and profibrotic reprogramming of CD8⁺ T cells, resulting in enhanced cardiac stiffness.

Here, we use bulk RNA sequencing of CD8⁺ T cells derived from anthracycline-treated patients, single-cell transcriptomics of doxorubicin-treated mouse hearts, and both in vivo and in vitro doxorubicin treatment to show that doxorubicin upregulates ECM remodeling enzymes including the lysyl oxidase family member LOX in CD8⁺ T cells, which we term “matrix remodeling T cells” (mxT-cells). We find that doxorubicin directly induces LOX expression through a HIF1α-P53-TGFβ signaling axis, leading to cardiac ECM stiffening. These findings reveal a novel immune regulatory mechanism underlying doxorubicin-induced fibrosis and introduce a new paradigm in the immunoregulation of ECM remodeling resulting from chemotherapy.

Methods

Please see the Supplemental Methods for additional methods.

Study design and patient recruitment

In this multicenter observational study, adult patients were prospectively enrolled at Beth Israel Deaconess Medical Center (BIDMC) and Tufts Medical Center from 2024 to 2025 and divided into discovery and validation cohorts. Eligible participants were individuals aged ≥18 years with malignancies who were scheduled to receive treatment with anthracyclines. Participants additionally had at least 1 risk factor for the development of cardiotoxicity, including age >65 years, hypertension, treatment for aggressive non-Hodgkin's lymphoma, prior chest irradiation, or a planned cumulative doxorubicin dose ≥300 mg/m². For patients receiving anthracyclines other than doxorubicin, conversions were made using an equivalent doxorubicin dose based on prior publications. All study participants provided their consent before enrollment. The study procedures and informed consent forms were approved by the Dana Farber/Harvard Cancer Center Institutional Review Board (IRB, protocol #18 to 126) with local oversight from the BIDMC Cancer Clinical Trials Office and by the Tufts Health Sciences IRB (protocol #13457). Participants underwent a baseline transthoracic echocardiogram before receiving the initial dose of anthracycline chemotherapy. After providing informed consent, a baseline blood draw was obtained along with answers to a questionnaire assessing cardiovascular risk factors. A follow-up blood draw was performed at 3 months post-treatment, and an echocardiogram was performed approximately 6 months later. Left ventricular ejection fraction was measured by an experienced cardiologist blinded to patient identification and study time points. Additional patient information is provided in Table 1.

Table 1.

Patient Characteristics

	Full Cohort (N = 9)	RNA Sequencing (n = 5)
Demographic characteristics
Age, y	59 ± 18.7	53 ± 22.8
Male	7 (77.8)	5 (100.0)
Female	2 (22.2)	0 (0.0)
White	8 (88.9)	5 (100.0)
African American	1 (11.1)	0 (0.0)
Malignancy
DLBCL	5 (55.6)	2 (40.0)
Hodgkin's lymphoma	3 (33.3)	3 (60.0)
Invasive ductal carcinoma	1 (11.1)	0 (0.0)
Cardiovascular treatments
Beta-blocker	3 (33.3)	2 (40.0)
ACE inhibitor	0 (0.0)	0 (0.0)
ARB	2 (22.2)	1 (20.0)
Cardiovascular risk factors
Hypertension	7 (77.8)	2 (40.0)
Diabetes mellitus	2 (22.2)	0 (0.0)
Hyperlipidemia	2 (22.2)	1 (20.0)
Smoking	2 (22.2)	2 (40.0)
BMI, kg/m2	28.8 ± 6.8	28.1 ± 6.0
eGFR, mL/min/1.73 m2	84 ± 26.3	90 ± 27.6
Average doxorubicin dose, mg/m2	216.6 ± 53.6	183.8 ± 36.1
Systolic BP, mm Hg	130 ± 11.4	126 ± 10.4
Diastolic BP, mm Hg	76 ± 11.7	72 ± 14.7
Baseline LVEF, %	60 ± 8.1	58 ± 10.7
Follow-up LVEF, %	56 ± 12.0	52 ± 15.6

Values are mean ± SD or n (%).

ACE = angiotensin-converting enzyme; ARB = angiotensin receptor blocker; BMI = body mass index; BP = blood pressure; DLBCL = diffuse large B-cell lymphoma; eGFR = estimated glomerular filtration rate; LVEF = left ventricular ejection fraction.

Isolation of human CD8⁺ T cells

Peripheral blood mononuclear cells were isolated from whole blood using a Ficoll gradient (Sigma GE17-1440-03). CD8⁺ T cells were then isolated by magnetic activated cell sorting using negative selection (Miltenyi 130-096-495). Cells were pelleted and stored at −80 °C before use for experiments.

Bulk RNA sequencing

RNA was extracted from sorted cells using the Qiagen QIASymphony RNA Kit on a QIASymphony SP/AS instrument. The isolated RNA was assessed for quality using an Agilent Fragment Analyzer and used as input for sequencing library preparation with the Illumina Stranded Total RNA with Ribo-Zero Plus kit, following the manufacturer's instructions. Completed libraries were analyzed and quantified on an Agilent Fragment Analyzer and pooled at equimolar concentrations for sequencing. Sequencing was performed on an Illumina NovaSeq X Plus using 1.5B 100 cycle chemistry in a single-read 100-base format. Basecalling and demultiplexing of the raw data were carried out using bclconvert on a DRAGEN server. Raw data in FASTQ format were processed for quality control using FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/), then mapped to the human genome using STAR (http://code.google.com/p/rna-star/) and quantified by featureCounts (http://subread.sourceforge.net/). Using QluCore software, a heat map of differentially expressed genes using hierarchical clustering with a P value of <0.05 and |log2(fold change [FC]) | of >1 for significance, which can be found in Supplemental Table 2. Gene set enrichment analysis was performed against Human MSigDB collections (https://www.gsea-msigdb.org/gsea/msigdb/human/collections.jsp), with P < 0.05 indicating significance. Gene ontology (GO) analysis for pathways and transcription factor enrichment was performed using ShinyGO 0.82.¹⁴ Full pathway analyses can be found in Supplemental Table 3. Raw data have been deposited at NCBI GEO under the accession number GSE346562.

Doxorubicin treatment in mice

All animal studies were approved by the Institutional Animal Care and Use Committees of Tufts University. Male C57BL/6J mice at 8 to 12 weeks of age were treated with 5 mg/kg doxorubicin (Selleckchem S1208) or phosphate-buffered saline (PBS) by intraperitoneal injection weekly for 4 or 8 weeks, resulting in cumulative doses of 20 or 40 mg/kg, respectively, or were euthanized and tissues collected at humane terminal endpoints including weight loss of >20%. A minimum sample size of 6 mice per group was calculated to detect a mean change in left ventricular ejection fraction of ∼12% with a power of >80% and α = 0.05. All analyses were performed in a blinded fashion using an identification number assigned by a third party.

Statistical analysis

Statistical analysis was performed using GraphPad Prism version 9.4.1 (GraphPad Software). All data are presented as mean ± standard error of the mean (SEM) unless otherwise indicated. Individual data points represent data from individual mice (for in vivo studies) or independent replicates (for in vitro studies). Normally distributed data were compared using Student's or Welch's t-test for 2 groups, while >2 groups were compared using 1-way analysis of variance (ANOVA) with Tukey's post hoc test for multiple pairwise comparisons. Non-normally distributed data were compared using the Mann-Whitney U test for between-group or the Wilcoxon signed-rank test for within-group comparisons. Pearson's correlation coefficient (r) was used to evaluate the association of 2 continuous variables. Differences were considered statistically significant at P < 0.05.

Results

Anthracycline treatment induces a profibrotic transcriptional profile in human cytotoxic T cells

We previously showed that cytotoxic CD8⁺ T cells are increased in the circulation of human and canine cancer patients treated with anthracyclines and are required for chronic cardiac remodeling mice.⁸ To evaluate transcriptional changes in CD8⁺ T cells in patients after chemotherapy in an unbiased manner, we collected blood from patients before and 3 months after the onset of anthracycline treatment, a timepoint in which we previously showed circulating T cell expansion and increased cytokine expression⁹ (Figure 1A), and isolated CD8⁺ T cells, yielding more than 90% CD8⁺ T cells before bulk RNA sequencing (Figure 1B). Our patients had an average age of 59 ± 18.7 years, were majority male, and had either a hematologic malignancy or breast cancer (Table 1). Overall, we identified 1,416 differentially expressed genes (DEGs), including 1,106 genes upregulated and 310 genes downregulated after anthracycline treatment (|log₂FC|>1; P < 0.05; Figures 1C and 1D, Supplemental Table 2). Gene set enrichment analysis revealed that genes involved in collagen formation (Figure 1E), ECM organization (Figure 1F) and integrin surface interactions (Figure 1G) were significantly enriched among the upregulated genes in CD8⁺ T cells from anthracycline-treated patients. An in-depth analysis of genes encoding collagen and collagen-modifying enzymes demonstrated the upregulation of multiple collagen isoforms, lysyl oxidases (Lox) and matrix metalloproteinases (Mmp) in response to anthracycline treatment (Figure 1H). To further validate these findings, we isolated CD8⁺ T cells from the blood of additional patients before and after 3 months of anthracycline treatment and measured several transcriptional hits by reverse transcription polymerase chain reaction (RT-qPCR). We found increased expression of LOX (Figure 1I), MMP9 (Figure 1J) and COL1A1 (Figure 1K) genes in patient-derived CD8⁺ T cells after anthracycline treatment. Collectively, these data show that anthracycline treatment induces a profibrotic transcriptional profile in human circulating CD8⁺ T cells and suggest their ability to remodel ECM.

Figure 1.

Patient-Derived Cytotoxic T Cells Express a Profibrotic Transcriptome After Anthracycline Treatment

(A) CD8⁺ T cells were isolated from patients before and 3 months after receiving anthracycline treatment. (B) Samples used for bulk RNA sequencing contained >90% CD8⁺ T cells by flow cytometry. (C) Heat map of differentially expressed genes (DEGs) in patients before and after anthracycline treatment shows a clear phenotypic separation between CD8⁺ T cells before and after anthracycline treatment. (D) Volcano plot showing DEG from (C); the x-axis is log₂(fold change) of means and the y-axis is log₁₀(P value). (E to G) Selected enriched gene sets connected to extracellular matrix (ECM) remodeling. (H) Heat map of selected DEGs connected to ECM remodeling. (I to K) Selected DEGs from bulk RNA sequencing were verified in additional samples by RT-qPCR. Data are presented as mean ± SEM and were compared using the Wilcoxon signed-rank test. ∗P < 0.05.

Anthracycline treatment in patients and mice induces a specific subset of matrix remodeling T cells (mxT-cells) distinct from traditional activated Tc1 cells

To assess if the expression of profibrotic transcripts was also detectable in cardiac CD8⁺ T cells of anthracycline-treated patients, we used a publicly available single-nucleus RNA sequencing (snRNA-seq) dataset of cardiac tissue from healthy donors or patients with anthracycline-induced cardiotoxicity (ie, doxorubicin cardiomyopathy [DR-CM]).¹⁵ Unbiased clustering identified all major cardiac cell types: cardiomyocytes, vascular, stromal, and immune populations (Figure 2A). We found a higher number of T cells expressing LOXL1 and COL1A1 at the transcript level in DR-CM patients compared to healthy donors, albeit the number of T cells expressing matrix remodeling–related genes was low (Figures 2B and 2C). As such, we computed a module score combining the expression levels of LOX, LOXL1, LOXL2, MMP2, and COL1A1, which we termed “mx score”. In line with the expression of LOXL1 and COL1A1, the mx score was higher in cardiac T cells from DR-CM patients compared to healthy donors (Figure 2D). Across cardiac cell types, we found the highest mx score in cardiac fibroblasts (CFBs). As expected in these fibrotic conditions, the mx score was also increased in CFBs from DR-CM patients compared to healthy donors (Supplemental Figures 1A and 1B).

Figure 2.

Mouse and Human Cardiac CD8⁺ T Cells Upregulate Collagen Modifying Genes After Anthracycline Treatment

(A) Unbiased clustering and cluster annotation based on snRNA-seq of cardiac samples from patients with doxorubicin cardiomyopathy (DR-CM; n = 11) or healthy donors (n = 10). (B to D) Violin plots of Loxl1, Col1a1, or the mx score in CD3e⁺CD8a⁺ T cells from snRNA-seq comparing donors to DR-CM patients. (E) Unbiased clustering and cluster annotation based on CITE-seq of CD45⁺ cardiac cells from PBS- or doxorubicin (DR)-treated mice (n = 6 per group). (F to I) Violin plots of Lox, Loxl1, Mmp9, or mx score in CD3e⁺CD8a⁺ T cells from CITE-seq comparing PBS to DR treatment. (J) Unbiased clustering of CD3e⁺CD8a⁺ T cells from the same CITE-seq data as in (E) to (I). (K) Violin plot of mx score in reclustered CD3e⁺CD8a⁺ T cells. (L) Feature plot of mx score in CD3e⁺CD8a⁺ T cells. (M) Selected pathway analysis from differentially expressed genes between T cells with a positive vs negative mx score. (N and O) Violin plots of Cd69 or Ifng expression comparing T cells with a positive vs negative mx score.

To assess fibrotic gene expression in cardiac T cells in greater mechanistic depth, we sourced a previously generated cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq) dataset from our group on isolated CD45⁺ cardiac non-myocytes after cumulative doxorubicin treatment.¹⁶ Unbiased clustering revealed that all major leukocyte populations were represented in the dataset (Figure 2E). We then subclustered cytotoxic T cells expressing CD3e and CD8a, and in line with our results in patients, we found that doxorubicin treatment induced the transcription of the matrix remodeling–related genes Lox, Loxl1, and Mmp9 in cardiac CD8⁺ T cells (Figures 2F to 2H). As in the human snRNA-seq data, we generated the mx score by combining the expression of Lox, Loxl1, Loxl2, Mmp9, and Col1a1, and, likewise, the mx score was higher in cardiac CD8⁺ T cells from doxorubicin-treated mice compared to PBS (Figure 2I). We then identified 9 distinct subclusters of CD8⁺ T cells in the heart, of which clusters 0, 3, and 4 were characterized by the highest mx score (Figures 2J to 2L). From here on, we refer to CD8⁺ T cells with a positive mx score as “matrix remodeling T cells” (mxT-cells). To evaluate if mxT-cells were distinct from the activated, IFNγ-expressing T cells that we previously showed contribute to doxorubicin induced cardiotoxicity,⁸ we performed GO analysis using genes with differential expression between mxT-cells and other CD8⁺ T cells. GO analysis revealed that mxT-cells expressed higher levels of genes associated with collagen and ECM, as expected, while expressing lower levels of genes associated with T cell activation and protein translation (Figure 2M, Supplemental Table 3). Importantly, mxT-cells from both mouse and human cardiac datasets expressed CD3e and CD8a, albeit at lower levels compared to the respective healthy controls, which is consistent with a less activated T cell phenotype (Supplemental Figures 1C to 1F). In line with this, mxT-cells from mice and humans expressed lower levels of Cd69 and Ifng compared to other cardiac CD8⁺ T cells (Figures 2N and 2O, Supplemental Figures 1G and 1H). Additionally, Cd69 and Ifng expression was highest in cardiac CD8⁺ T cell subclusters 1, 5, and 6 and thus distinct from those with the highest mx score (Supplemental Figure 1I). Together, these data provide evidence that anthracycline treatment induces a novel subset of CD8⁺ T cells (ie, mxT-cells) in the hearts of humans and mice, characterized by increased expression of matrix-remodeling genes and decreased traditional activation markers.

Doxorubicin treatment in mice leads to increased fibrotic gene expression in splenic and circulating cytotoxic T cells and elevated cardiac ECM stiffness

We then sought to assess the presence of mxT-cells across time in a chronic model of doxorubicin cardiotoxicity. We treated mice with doxorubicin or PBS weekly for 4 or 8 weeks, then assessed cardiac function by echocardiography (Figure 3A).⁸ As expected, doxorubicin treatment led to a progressive reduction of left ventricular ejection fraction over 8 weeks (Figure 3B). We did not see significant differences in global longitudinal strain or global circumferential strain between PBS- and doxorubicin-treated mice (Table 2). To better assess the stiffness of cardiac ECM directly, we performed mechanical testing of decellularized left ventricular tissue. We found a progressive increase in left ventricular ECM stiffness, evidence of fibrosis and ECM remodeling, in doxorubicin-treated mice compared to PBS controls (Figure 3C). Strikingly, higher ECM stiffness correlated with decreased left ventricular ejection fraction across time points, highlighting the importance of ECM mechanics for cardiac function in doxorubicin-induced HF (Figure 3D). To further interrogate the profibrotic effects of doxorubicin on CD8⁺ T cells, we isolated CD8⁺ T cells from the spleens and blood of PBS- or doxorubicin-treated mice. We found that the mRNA expression levels of Lox (Figures 3E and 3J), Loxl1 (Figures 3F and 3K), and Mmp9 (Figures 3G and 3L) were increased in both splenic and circulating CD8⁺ T cells by doxorubicin treatment after 8 weeks compared to PBS-treated controls. To validate this finding at the protein level, we quantified intracellular LOX by flow cytometry and found higher LOX expression in CD8⁺ T cells in both the spleens and circulation of doxorubicin-treated mice compared to PBS controls (Figures 3I and 3N, Supplemental Figures 2A and 2B). However, we did not find significant differences in the number of LOX⁺ cardiac CD8⁺ T cells or average LOX expression per cell (Supplemental Figures 2C to 2E). Considering the robust increase in transcript in both humans and mice, and how rapid LOX can be secreted, this may suggest that LOX is actively secreted by CD8⁺ T cells during the cardiac sample processing and, therefore, undetectable intracellularly. We also measured LOX expression in CFBs, as the main fibrotic cell type in the heart, and found that doxorubicin treatment increased the number of LOX expressing CFBs but not average expression per cell, supporting the idea that cells secrete LOX in the heart (Figures 3O and 3P, Supplemental Figure 2F). Importantly, the level of splenic CD8⁺ T-cell Lox transcript correlated significantly with ECM stiffness across time points, suggesting that mxT-cells are associated with increased ECM stiffness (Figure 3Q). Taken together, these data demonstrate that doxorubicin induces progressive ECM stiffening that correlates with a decline in cardiac function. We demonstrate a novel profibrotic signature in cytotoxic T cells, which is characterized by the expression of collagen-modifying enzymes and may contribute to ECM stiffening in anthracycline-induced HF.

Figure 3.

CD8⁺ T Cells From Mice Treated With Doxorubicin Increase LOX Expression Systemically

(A) Male C57B/6J mice were treated with 5.0 mg/kg doxorubicin (DR) or PBS for 4 or 8 weeks (n = 6 per group). (B) Left ventricular ejection fraction (EF) was measured by echocardiography. (C) Uniaxile tensile testing was used to determine extracellular matrix (ECM) stiffness from left ventricular samples. (D) Correlation between EF and ECM stiffness. (E to G) CD8⁺ T cells were isolated from spleens, then RT-qPCR was used to measure Lox, Loxl1, or Mmp9 expression. (H and I) Flow cytometry was used to measure CD8⁺ T cell LOX protein levels in spleens of mice treated with PBS (n = 8) or DR for 8 weeks (n = 4). Shown are representative plots gated on CD45⁺CD8⁺ cells (H) with quantification of mean fluorescence intensity (MFI; I). (J to L) CD8⁺ T cells were isolated from the blood, then RT-qPCR was used to measure Lox, Loxl1, or Mmp9. (M and N) Flow cytometry was used to measure circulating CD8⁺ T-cell LOX protein levels of mice treated with PBS (n = 8) or DR for 8 weeks (n = 4); shown are representative plots gated on CD45⁺CD8⁺ cells (M) with quantification of MFI (N). (O and P) Heart tissue from mice treated with PBS (n = 8) or DR for 8 weeks (n = 4) was enzymatically digested, then LOX expression in cardiac fibroblasts was measured by flow cytometry. Shown are representative plots gated on CD31⁻CD45⁻PDGFRa⁺ cells (O) with quantification of MFI (P). (Q) Correlation of ECM stiffness (G) with splenic CD8⁺ T-cell Lox (I). Data are presented as mean ± SEM and were compared using 1-way ANOVA with Tukey's post hoc test (B-C, E-G, J-L), Mann-Whitney test (I, N, P) or Pearson's correlation (D, Q); ns = no significance, ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001.

Table 2.

Echocardiographic Analysis

	PBS-Treated (n = 8)	DR-Treated (n = 6)	P Value
LVAWd, mm	0.90 ± 0.04	0.78 ± 0.04	0.068
LVAWs, mm	1.37 ± 0.06	1.18 ± 0.05	0.047
LVPWd, mm	0.87 ± 0.04	0.69 ± 0.07	0.070
LVPWs, mm	1.40 ± 0.09	0.93 ± 0.09	0.005
Ejection fraction, %	64.2 ± 4.0	49.9 ± 3.3	0.026
GLS, %	−11.7 ± 1.2	−11.4 ± 0.7	0.85
GCS, %	−22.6 ± 2.1	−18.7 ± 1.0	0.16

Echocardiography was performed on mice treated with 4 weeks of doxorubicin (DR) or PBS treatment. Values are mean ± SEM and were compared using the Welch's t-test, with P values shown. P values < 0.05 are highlighted in bold.

GCS = global circumferential strain; GLS = global longitudinal strain; LVAWd = left ventricular anterior width in diastole; LVAWs = left ventricular anterior width in systole; LVPWd = left ventricular posterior width in diastole; LVPWs = left ventricular posterior width in systole.

Doxorubicin treatment of murine cytotoxic T cells upregulates collagen-modifying enzymes and induces LOX-dependent ECM stiffening

We next investigated if doxorubicin directly induced profibrotic gene expression in CD8⁺ T cells. Our previous work demonstrated that splenic CD8⁺ T cells are responsive to doxorubicin in vitro by upregulating IFNγ expression and that IFNγ⁺ CD8⁺ T cells (Tc1 cells) are more abundant in the hearts of doxorubicin-treated mice,⁹ however thus far these appear distinct from mxT-cells. To test if this effect occurred alongside the profibrotic profile we have observed so far, we isolated murine splenic CD8⁺ T cells, polarized them toward a Tc1 subtype, and treated them with doxorubicin for 24 hours. We found that doxorubicin increased the expression of Lox, Loxl1, Mmp9, Mmp13, and Col1a1 in Tc1 cells, compared to vehicle-treated Tc1 (Figures 4A and 4E). Interestingly, inhibition of IFNγ alongside doxorubicin treatment did not affect transcription of any of these genes (Supplemental Figure 3). In contrast to Tc1 cells, doxorubicin did not induce Lox or Loxl1 expression in CFBs treated in vitro, suggesting the increase in LOX⁺ CFBs observed in vivo may not be directly triggered by doxorubicin actions on CFBs (Supplemental Figure 4). We further assessed protein levels of LOX in cultured Tc1 cells by Western blot and found that doxorubicin treatment increased the abundance of a ∼35 kDa form of LOX (Figure 4F), which is an enzymatically active fragment with high collagen affinity.¹⁷

Figure 4.

In Murine Tc1 Cells, Doxorubicin Upregulates Collagen Modifying Enzymes and Induces LOX-Dependent ECM Stiffening

Murine splenic Tc1 cells were treated with doxorubicin (DR; 0.1 μg/mL) for 24 hours (n = 5-6 biologic replicates/group). (A to E) mRNA expression levels of Lox, Loxl1, Mmp9, Mmp13, or Col1a1 were assessed by RT-qPCR. (F) Protein levels of LOX were assessed by Western blot; shown is a representative blot from n = 3 biologic replicates with quantification of n = 6/group. (G) Conditioned media from Tc1 cells (control [C] or DR-treated, 24 hours) was used to treat decellularized extracellular matrix (dECM) preparations for 16 to 20 hours in the presence or absence of BAPN (500 μM). Stiffness of dECM was tested using uniaxial tensile testing before and after treatment. (H) Representative stress/strain curves for each treatment group with elastic modulus quantified. (I) Fold change of elastic modulus in response to treatment as indicated from n = 4-6/group. Data are presented as mean ± SEM and were compared using Student's t-tests (A-F) or 1-way ANOVA with Tukey's post hoc test (I); ns = no significance, ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001. BAPN = β-aminopropionitrile (pan-lysyl oxidase inhibitor).

To evaluate the functional relevance of T cell LOX upregulation for ECM modification after doxorubicin treatment, we asked if soluble, Tc1-cell–derived mediators from doxorubicin-treated T cells could alter the stiffness of native cardiac ECM. To do this, we established an assay in which we first measured the initial stiffness of decellularized cardiac ECM (dECM) from healthy mice. We then treated dECM with conditioned media collected from Tc1 cells treated with doxorubicin or vehicle and lastly tested the stiffness of the same dECM preparation after treatment. As such, any changes in dECM stiffness can be attributed to Tc1-derived, secreted factors and their differential regulation in the response of Tc1 cells to doxorubicin (Figure 4G). We found that conditioned media from control Tc1 cells slightly reduced the stiffness of the dECM compared to before the treatment (0.6-fold change). In contrast, when the Tc1 cells were treated with doxorubicin, the T-cell–conditioned media caused a 2.2-fold increase in dECM stiffness compared to before conditioned media treatment. This stiffening fully depended on lysyl oxidase activity, as it was completely abrogated by the presence of the lysyl oxidase inhibitor BAPN (β-amino-propionitrile) (Figures 4H and 4I). Together, these findings demonstrate that doxorubicin is sufficient to induce matrix remodeling gene expression in murine Tc1 cells and that doxorubicin-dependent, Tc1-derived lysyl oxidases efficiently increase the stiffness of cardiac ECM in vitro.

Doxorubicin induces expression of LOX in CD8⁺ T cells through an autocrine TGFβ1-dependent mechanism

To gain deeper insight into the molecular mechanism underlying doxorubicin-induced profibrotic gene expression in CD8⁺ T cells, we analyzed our human and murine sequencing data sets and performed predictive analysis of upregulated transcription factor activity based on DEGs in response to doxorubicin. In our human bulk CD8⁺ T cell sequencing data from anthracycline-treated patients, we identified HIF1A, SMAD3, E2F1, TP53, and JUN as enhanced transcription factors after anthracycline treatment (Figure 5A, Supplemental Table 3). In parallel, the murine CITE-seq data revealed decreased SMAD7 and increased ATF4, SMAD3, SMAD4, and TP53 as predicted transcription factors in circulating CD8⁺ T cells from 8-week doxorubicin-treated mice vs control (Figure 5C, Supplemental Table 3). The main shared pathway between these 2 datasets centers around the profibrotic growth factor TGFβ, as increased SMAD3/SMAD4 are the canonical transcription factors downstream of TGFβ and SMAD7 is an inhibitor of TGFβ signaling. As such, we first focused on the involvement of TGFβ signaling in doxorubicin-induced LOX expression and ECM remodeling by CD8⁺ T cells. We found increased TGFB1 transcript levels in human circulating CD8⁺ T cells after 3 months of anthracycline treatment (Figure 5B) as well as in murine circulating CD8⁺ T cells after 8 weeks of doxorubicin (Figure 5D). Using flow cytometry, we assessed the phosphorylation of SMAD2 (p-SMAD2), which interacts with SMAD3 and SMAD4 downstream of TGFβR ligation and indicates active canonical TGFβ signaling, in cardiac CD8⁺ T cells (see Supplemental Figure 2 for gating strategy). We found higher numbers of p-SMAD2⁺ CD8⁺ T cells and higher average levels of p-SMAD2 in cardiac CD8⁺ T cells from doxorubicin-treated mice compared to PBS-treated mice (Figures 5E to 5G). We then tested TGFβ signaling in splenic CD8⁺ T cells after 8 weeks of doxorubicin treatment and found increased Tgfb transcript as well as elevated levels of p-SMAD2 by Western blot (Figures 5H and 5I). In vitro, doxorubicin treatment of Tc1 cells was sufficient to cause Tgfb1 upregulation, leading to TGFβ pathway activation in doxorubicin-treated Tc1 cells as assessed by p-SMAD2 (Figures 5J and 5K). Treatment with recombinant TGFβ1 was sufficient to upregulate Lox, Loxl1, Mmp9, and Mmp13 transcripts in Tc1 cells in vitro (Supplemental Figure 5). In both mouse and human transcriptomics datasets, we found evidence that cardiac CD8⁺ T cells increase Tgfb1 expression, similar to our findings in splenic CD8⁺ T cells in vivo and Tc1 cells in vitro (Supplemental Figures 6A and 6D). Interestingly, we did not see consistent changes of Tgfbr subunits in mouse or human cardiac CD8⁺ T cells (Supplemental Figures 6B and 6C, 6E and 6F). We then analyzed other cell types in the human heart, and, as expected, found that TGFB1 was broadly expressed in immune and stromal cell populations, with T cells and NK cells showing the most striking increase of TGFB1 expression in DR-CM patients compared to healthy donors (Supplemental Figure 7G). Together, this suggests that doxorubicin induces Tc1 TGFβ1 expression and release, which could in turn signal in an autocrine loop to activate canonical TGFβ signaling (Figure 5L). TGFβ, used to induce Foxp3, or in combination with IL6 to induce Rorc, driving Treg and Th17 cell polarization, respectively, did not result in significant changes in the expression of Foxp3, Tbx21, or Rorc in fully differentiated Tc1 cells (Supplemental Figure 7), suggesting that autocrine TGFβ1 activity does not affect Tc1 polarization.

Figure 5.

Doxorubicin Induces Expression of LOX in CD8+ T Cells Through an Autocrine, TGFβ-Dependent Mechanism

(A) Predicted transcription factor enrichment analysis after anthracycline treatment from differentially expressed genes in bulk sequencing of patient CD8⁺ T cells (n = 5). (B) RT-qPCR was used to measure TGFB1 mRNA in patient CD8⁺ T cells before anthracycline treatment (Pre) or 3 months after anthracycline (3m; n = 9). (C) Predicted transcription factor enrichment analysis after anthracycline treatment from differentially expressed genes in CD8⁺ T cells from mouse cardiac CITE-seq. (D) RT-qPCR was used to measure Tgfb1 mRNA in circulating CD8⁺ T cells. (E to G) Heart tissue from mice treated with PBS (n = 8) or doxorubicin (DR) for 8 weeks (n = 4) was enzymatically digested then pSMAD2 expression in CD8⁺ T cells was measured by flow cytometry. Shown are representative plots gated on CD45⁺CD8⁺ T cells (F) with quantification of mean fluorescence intensity (MFI; G). (H) RT-qPCR was used to measure Tgfb1 mRNA in splenic CD8⁺ T cells from PBS- or DR-treated mice (n = 6). (I) Representative Western blot and quantification of pSMAD2 in splenic CD8⁺ T cells from PBS to 8-week DR-treated mice (n = 3). (J) RT-qPCR was used to measure Tgfb1 mRNA from Tc1 cells treated with 0.1 μg/mL DR for 24 hours (n = 6 biologic replicates). (K) Representative Western blot (n = 3) and quantification (n = 6) of pSMAD2 in Tc1 cells treated with DR. (L) Schematic of proposed autocrine TGFβ signaling in DR-treated CD8⁺ T cells. (M) RT-qPCR was used to measure Lox from Tc1 cells treated with DR in the presence or absence of a TGFβ-blocking antibody (αTGFβ 5 μg/mL; n = 4). (N) Representative Western blot for LOX and pSMAD2 protein from Tc1 cells treated with DR in the presence or absence of αTGFβ. (O) Fold change of elastic modulus of ECM treated with supernatant collected from Tc1 cells (DR or DR + αTGFβ; n = 6-8). Data are presented as mean ± SEM and were compared using Student's t-test (B, F-G, I, K, O) or 1-way ANOVA with Tukey's post hoc test (D, H, M); ns = no significance, ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001.

To assess if doxorubicin-induced autocrine TGFβ1 was required for LOX expression and ECM stiffening, we then performed doxorubicin treatment of Tc1 cells in vitro in the presence of a TGFβ-neutralizing antibody, as previously utilized.¹⁸ We found that the doxorubicin induced upregulation of Lox in Tc1 cells at the RNA and protein level were diminished in the presence of anti-TGFβ (Figures 5M and 5N). Consequently, anti-TGFβ also abolished the ability of doxorubicin-treated Tc1-derived conditioned media to stiffen the dECM (Figure 5O). Collectively, these data suggest the critical requirement of autocrine TGFβ1 signaling in doxorubicin-induced LOX expression, ECM stiffening and, thus, fibrotic remodeling by mxT-cells.

Doxorubicin induces TGFβ1 expression and LOX-dependent ECM stiffening through HIF1α-P53 signaling

We next aimed at understanding the cellular mechanism by which doxorubicin induces TGFβ1 in CD8⁺ T cells. We performed pathway analysis of DEGs from both the human and mouse sequencing datasets, and both revealed HIF1α pathway enrichment, as well as P53 and transcriptional misregulation in cancer (Figures 6A and 6B). This was further supported by our previous transcription factor analysis, which indicated P53 and HIF1α as important transcription factors driving the doxorubicin-induced transcriptomic signature (Figure 5). Thus, we hypothesized that doxorubicin triggers HIF1α activation and P53 phosphorylation, both of which have been previously reported to interact with TGFβ1 to drive profibrotic responses in other cell types.19, 20, 21 To test for activation and temporal regulation of these pathways in Tc1 cells, we performed a time course of doxorubicin treatment. We found that HIF1α protein levels increased between 1 and 4 hours post-treatment and then sharply decreased. This was followed by P53 phosphorylation at 4 hours, which was sustained at 8 hours, and a steady increase in SMAD2 phosphorylation occurring at 8 and 24 hours (Supplemental Figure 8). This provides evidence that doxorubicin sequentially activates HIF1α, P53, and canonical TGFβ signaling through SMAD2 (Figure 6C). We then used flow cytometry to measure phosphorylated P53 (p-P53) in cardiac CD8⁺ T cells from doxorubicin-treated mice and found an increased number of p-P53⁺ CD8⁺ T cells in doxorubicin-treated compared to PBS-treated mice (Figures 6D and 6E). To further evaluate the full pathway, we assessed protein levels of HIF1α and p-P53 in murine splenic CD8⁺ T cells after 8 weeks of doxorubicin treatment in vivo, as well as in Tc1 cells after 24 hours of doxorubicin treatment in vitro. In both cases, we found decreased HIF1α in response to doxorubicin (in line with our time course showing HIF1α downregulation after activation), and increased p-P53 at a time point past the HIF1α peak expression (Figures 6F and 6G). These data show a sequential signaling cascade HIF1α-P53-SMAD2 in T cells upon doxorubicin treatment in vitro and in vivo. Inhibition of HIF1α by echinomycin during doxorubicin treatment in vitro diminished the doxorubicin-induced upregulation of Tgfb1 (Figure 6H) and Mmp9 (Figure 6I). The upregulation of LOX protein by doxorubicin was also abolished by HIF1α inhibition (Figures 6K and 6L). Moreover, the conditioned media from Tc1 cells treated with the HIF1α inhibitor echinomycin showed decreased ECM stiffening as compared with the conditioned media from doxorubicin-treated Tc1 cells with intact HIF1α and LOX expression (Figure 6M).

Figure 6.

LOX Expression CD8⁺ T Cells in Response to Doxorubicin is p53-HIF1α Dependent

(A) Predicted pathway analysis after anthracycline treatment from differentially expressed genes in bulk sequencing of patient CD8⁺ T cells (n = 5). (B) Predicted transcription factor enrichment analysis after anthracycline treatment from differentially expressed genes in CD8⁺ T cells from mouse cardiac CITE-seq. (C) Schematic of hypothesized signaling pathway activated by doxorubicin (DR) in Tc1 cells with in vitro interventions (Echin = echinomycin). (D and E) Heart tissue from mice treated with PBS (n = 8) or DR for 8 weeks (n = 4) was enzymatically digested, then p-P53 expression in CD8⁺ T cells was measured by flow cytometry. Shown are representative plots gated on CD45⁺CD8⁺ T cells (D) with quantification of mean fluorescence intensity (MFI; E). (F) Representative Western blot and quantification of HIF1α and p-P53 in splenic CD8⁺ T cells from PBS or 8-week DR-treated mice (n = 3). (G) Representative Western blot (n = 3) and quantification of HIF1α and p-P53 (n = 6) in Tc1 cells treated with DR. (H and I) RT-qPCR was used to measure Tgfb1 and Mmp9 in Tc1 cells treated with 0.1 μg/mL DR in the presence or absence of Echin (20 ng/mL) for 24 hours. (J) Tc1 cells were treated with the indicated dose of nutlin-3 for 24 hours, then Western blot was used to measure p-P53. (K and L) Representative Western blot (n = 3) and quantification (n = 6) of LOX in Tc1 cells treated with DR with or without Echin or nutlin-3 as indicated. (M) Fold change of elastic modulus of ECM treated with supernatant collected from Tc1 cells treated as in (K) and (L) (n = 6-7). Data are presented as mean ± SEM and were compared using the Mann-Whitney test (E-G) or 1-way ANOVA with Tukey's post hoc test (H-M); ns = no significance, ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001.

To further test whether P53 signaling followed HIF1α activation, one mechanism by which doxorubicin induces LOX expression and T-cell–induced ECM remodeling, we performed a rescue experiment using nutlin-3, a small molecule that hyperactivates P53.¹⁷ We first confirmed that nutlin-3 treatment increased total and phosphorylated P53 levels in a dose-dependent manner in Tc1 cells (Figure 6J). We then tested if the addition of nutlin-3 to doxorubicin/echinomycin-treated Tc1 cells rescued the expression of LOX. As expected, compared to doxorubicin/echinomycin treatment alone, addition of nutlin-3 increased the expression of LOX (Figures 6K and 6L). Importantly, nutlin-3 significantly enhanced dECM stiffening by Tc1-conditioned media despite HIF1α inhibition by echinomycin (Figure 6M). Together, these data demonstrate a comprehensive cellular signaling cascade through which doxorubicin reprograms CD8⁺ T cells to become mxT-cells. doxorubicin-induced reprogramming involves the HIF1α-P53–dependent production of TGFβ1, resulting in profibrotic LOX expression and ECM remodeling, as summarized in the Central Illustration.

Central Illustration.

Anthracycline-Induced mxT-Cells Contribute to Cardiac Extracellular Matrix Stiffening

CD8⁺ T cells acquire a novel matrix-remodeling T-cell phenotype (mxT-cells) after exposure to doxorubicin in vitro, in vivo in mice, and in anthracycline-treated patients. This phenotype originates from a doxorubicin-triggered T-cell–intrinsic autocrine TGFβ-p53-HIF1α–dependent signaling mechanism that results in lysyl oxidase (LOX) expression and subsequent extracellular matrix (ECM) stiffening.

Discussion

While it is well-appreciated that a subset of patients treated with anthracyclines experience chronic HF, our understanding of the underlying cellular and molecular mechanisms remains incomplete, leading to a paucity of specific cardioprotective options. Cardiac inflammation and fibrosis have emerged as new contributors to cardiotoxicity in these patients,^7,8,22,23 but mechanisms by which this leads to chronic cardiac dysfunction are not well understood. Here, we show for the first time that cytotoxic CD8⁺ T cells adopt a profibrotic transcriptomic signature in response to anthracycline/doxorubicin treatment in patients and mice. We term these “matrix-remodeling T cells” (mxT-cells) and demonstrate the ability of mxT-cells to stiffen native cardiac ECM when treated with doxorubicin in vitro. We discover a molecular mechanism that results in upregulation of ECM crosslinking enzymes downstream of HIF1α and autocrine TGFβ signals. Our findings demonstrate that in the onset of doxorubicin, cytotoxic T cells directly contribute to cardiac ECM stiffening via a novel T-cell–intrinsic mechanism of ECM remodeling that functions independently of CFB and complements their previously established role in activating CFB in vivo.⁸

Classically, CFB are recognized as the main producers and remodelers of ECM at homeostasis and in pathologic fibrotic conditions such as HF. They can both respond to inflammatory triggers and help recruit immune cells to the inflamed heart, leading to fibrosis.^24,25 In line with this, we previously demonstrated that fibroblasts help recruit cytotoxic T cells after doxorubicin treatment, and that T cell contact with CFB and subsequent degranulation in the heart triggers myofibroblast transformation. The resulting increase in collagen production contributes to doxorubicin-induced HF.^8,9 Fibrosis, however, is a complex process that involves not only the production of new collagens and other ECM proteins, but also altering their crosslinking to convey mechanical strength.²⁶ While the de novo production of ECM is largely attributed to fibroblasts, as supported by the increased expression of LOX and matrix-remodeling genes we found in CFB, non-fibroblast cell types can also steer fibrotic transformation of fibroblasts (through direct contact or secreted factors) or can directly change mechanical ECM properties. For example, ECM-producing neutrophils have been identified in the skin, where they play a crucial role in tuning the mechanical properties during wound healing. Interestingly, ECM production in these neutrophils was dependent on TGFβ/SMAD signaling.²⁷ Cytotoxic T cells in the heart favor adverse CFB remodeling through the release of granzyme B in response to doxorubicin, as well as in myocardial infarction.^8,28 We show that cytotoxic T cells themselves produce LOX in response to doxorubicin treatment and that this is sufficient to stiffen cardiac ECM in vitro in the absence of CFB. Interestingly, we find that doxorubicin treatment does not directly induce Lox mRNA expression in CFB in vitro. This suggests that LOX induction in vivo occurs through a different mechanism, perhaps through T cell–CFB interactions and granzyme B signaling. This represents a novel mechanism of anthracycline-induced cardiac fibrosis in which T cells are directly implicated in ECM remodeling. Additional studies are needed to delineate the relative contribution of T-cell LOX induced by doxorubicin, T-cell–mediated CFB activation, and other fibrotic factors to ECM stiffening in vivo.

Our mechanistic studies delineate a signaling pathway induced by doxorubicin in T cells that involves HIF1α and p53, identified as top hits in T cells from mouse and humans treated with doxorubicin. We demonstrate that HIF1α, TGFβ1, and LOX transcription and secretion are together responsible and required for ECM crosslinking in vitro. While we show increased transcription of Lox and Loxl1, the inhibitor we used in vitro inhibits the activity of all LOX family members, and as such the respective contributions of each enzyme remain unknown. Interestingly, while CD4⁺ T cells typically respond to TGFβ by differentiation into regulatory T cells, doxorubicin-induced TGFβ did not alter expression of Foxp3 or Tbx21 in fully differentiated Tc1 cells, suggesting there is no alteration of Tc1 polarization during CD8⁺ T-cell–mediated ECM stiffening. However, we demonstrate that mxT-cells express lower levels of T cell activation markers Cd69 and Ifng in anthracycline-treated patients and mice. This suggests mxT-cells form a population distinct from the highly activated, IFNγ-expressing T cells that we previously characterized.⁸ In ischemic heart failure, the loss of CD69 in regulatory T cells has been associated with worse outcomes in patients and mice.²⁹ Foxp3 expression was very low in these datasets; as such, whether mxT-cells overlap with T regs or whether the reduction of Cd69 expression is causally involved in the development of mxT-cells remains elusive. We hypothesize that mxT-cells coexist with other classic T cell populations in the inflamed heart and contribute to fibrosis through distinct mechanisms. Future studies will further characterize how mxT-cells compare to and differ from classic T cell populations beyond their ability to remodel ECM. Additionally, doxorubicin treatment is known to increase production of reactive oxygen species (ROS), which we hypothesize is the mechanism for activation of HIF1α observed in T cells. In cardiomyocytes, doxorubicin-induced ROS activate hypoxia signaling through HIF1α, resulting in apoptotic cell death,³⁰ however we only observed minor signs of cell death in Tc1 cells. Future work will aim to elucidate the exact molecular link between doxorubicin and HIF1α activation in these cells and test if ROS scavenging can ameliorate this.

In the context of the heterocellular heart, where multiple cell types interact at any given time, TGFβ may be released by CFB as well as by T cells, and therefore act on a cell in autocrine manner, as we demonstrate herein in isolated T cells treated with doxorubicin in vitro, or in neighboring cells in a paracrine manner. These include CFB, in which TGFβ treatment not only induces a general profibrotic transformation, but also upregulates LOX expression.³¹ However, the apparent lack of doxorubicin-induced Lox in CFB in vitro suggests that anthracycline treatment causes CFB activation at least partially through a mechanism that is distinct from the axis we identified in mxT-cells. Additionally, ECM stiffness itself is sensed by CFB and is a potent driver of profibrotic fibroblast transformation.^32,33 The relative contribution of mxT-cell–derived LOX and TGFβ to the adverse mechanical remodeling of the heart compared to fibroblast-derived TGFβ, ECM, and ECM-modifying proteins will be investigated in future studies. Similarly, future studies will answer whether LOX inhibition protects from myocardial ECM stiffening and improves cardiac function in doxorubicin-treated mice.

Activation of LOX by HIF1α is known in the context of the hypoxic environment of tumors.³⁴ LOX expression in response to hypoxia stiffens the tumor ECM, which limits diffusion and increases resistance to chemotherapy in triple-negative breast cancer.^35,36 The cellular source of LOX in the context of cancer is often unclear. Interestingly, paclitaxel treatment has been shown to increase T cell LOX expression in the context of breast cancer, which promotes metastasis by altering pulmonary ECM.¹³ Other examples of LOX-mediated ECM remodeling in non-tumor tissues favoring metastasis formation have been shown,^37,38 however whether doxorubicin promotes this process through alterations to CD8⁺ T cells remains unknown.

Study limitations

Our current study was performed in non–cancer-bearing mice, representing a limitation of our preclinical work. However, as we find an increase in T cell LOX expression also in CD8⁺ T cells from cancer patients treated with doxorubicin, we hypothesize that this finding is independent of the cancer. This suggests that preventing LOX expression in T cells might not only reduce anthracycline-induced cardiac remodeling, but also prevent metastasis, potentially increasing the effectiveness of anthracycline chemotherapy with beneficial oncological implications. Our study is also limited by a small patient sample size, and larger patient populations are needed to understand whether this phenomenon is influenced by factors such as malignancy, patient age or sex, exact chemotherapeutic regimen, and more. We further only isolated patient samples at a single time point; as such, future studies will aim to perform a longer time course to understand how T cell ECM remodeling might influence chronic HF. Finally, we acknowledge that while we show increased cardiac and circulating CD8⁺ T cell LOX expression and the ability of mxT-cell–derived LOX to stiffen cardiac ECM in vitro, additional studies are needed to isolate the specific functional contribution of mxT-cell–derived LOX to ECM stiffening and ultimately HF progression. Further, while LOX inhibition has been tolerated well without toxicity in phase 1 trials thus far,³⁹ additional study of the chronic effects of LOX inhibition more broadly beyond the heart will be necessary before translation. Despite these limitations, this study sheds new light on immune mechanisms and molecular elements contributing to cardiac stiffening in doxorubicin cardiotoxicity.

Conclusions

Widely effective and specific treatment strategies for anthracycline-induced HF are lacking, in part because of our incomplete understanding of the molecular mechanisms and contributions of newly appreciated cardiac cell types in this disease. Our present work shows a conserved mechanism in both human anthracycline-treated cancer patients and doxorubicin-treated mice by which cytotoxic T cells are reprogrammed into mxT-cells through HIF1α-P53-TGFβ1 signaling. These mxT-cells are characterized by a profibrotic transcriptional program involving high expression of the ECM crosslinking enzyme LOX. Importantly, we provide evidence that mxT-cell–derived LOX is sufficient to increase ECM stiffness in vitro and correlates with ECM stiffness and systolic dysfunction in vivo. Together, these data highlight the importance of mxT-cells for adverse cardiac remodeling in anthracycline-treated patients and evidence a novel fibrotic mechanism with potential implications for therapeutic intervention.

Data Availability

The data, methods, and materials will be available to other researchers for the purpose of reproducing the results or replicating procedures upon reasonable request to the corresponding author.

Funding Support and Author Disclosures

This work was supported by National Institute of Health (NIH) grants R01 1HL144477 and HL165725 (to Dr Alcaide), R01 HL163172 (to Dr Asnani), Tufts Springboard Tier 1 Grant (to Dr Alcaide), NIH F30 grant HL162200 (to Dr Bayer), German Research Foundation (DFG) within the Walter Benjamin program 539486371 (to Dr Emig), and NIH F31 grants F31HL175911 (to Ms Zambrano) and K08 HL146969 (to Dr Upshaw). The authors have reported that they have no relationships relevant to the contents of this paper to disclose.

Perspectives.

COMPETENCY IN MEDICAL KNOWLEDGE: Mechanical ECM remodeling and fibrosis lead to impaired cardiac function. Profibrotic, cytotoxic T cells that induce cardiac remodeling in anthracycline cardiotoxicity and can be identified in patient blood are an attractive biomarker as well as a potential novel therapeutic target for cardiotoxicity.

TRANSLATIONAL OUTLOOK: Further research will evaluate the potential of CD8⁺ T cell LOX as antifibrotic therapeutic target in anthracycline cardiotoxicity, as well as circulating T cell factors as future biomarkers.