Single-nucleus RNA sequencing reveals the pathological process and key transcription factor of sorafenib-induced cardiotoxicity

Abstract

Background

Sorafenib is a first-line treatment for advanced solid tumors. However, its clinical application is limited by cardiotoxicity. The underlying mechanisms remain unclear. We aimed to identify key mechanisms of sorafenib-induced cardiotoxicity using single-nucleus RNA sequencing (snRNA-seq) and experimental verification.

Methods

SnRNA-seq was performed on cardiac tissue from a mouse model of sorafenib-induced cardiotoxicity. Bioinformatics analyses included weighted gene co-expression network analysis (WGCNA), differentially expressed gene (DEG) analysis, intercellular communication inference, and transcriptional regulatory network analysis. We conducted functional verification both in vivo and in vitro.

Results

SnRNA-seq demonstrated that cell clustering and component analysis confirmed the reduction in cardiomyocytes (CMs) and endothelial cells (ECs), along with an increase in immune cell populations. DEG enrichment analysis indicated that the upregulated DEGs in CMs, ECs, and fibroblasts (FBs) were consistently enriched in pathways associated with cardiomyopathy, whereas the downregulated DEGs were enriched in myocardial remodeling and repair processes. Gene set scoring revealed activation of fibrosis and inflammation programs in CMs and FBs. Cell communication analysis revealed a widespread disruption of cellular crosstalk in sorafenib-induced cardiotoxicity and identified key ligand-receptor pairs contributing to pro-inflammatory and pro-fibrotic phenotypes. Meanwhile we identified the common transcription factors and main effect genes in the three main effector cells during sorafenib-induced cardiotoxicity. Augur analysis and WGCNA consistently identified CMs as the most significantly affected cell type. Transcriptional regulatory network analysis highlighted melanocyte-inducing transcription factor (MITF) as a core transcription factor involved in sorafenib-induced cardiotoxicity. Functional validation showed that MITF expression was significantly increased in cardiac tissues and CMs after sorafenib treatment and that knockdown of MITF attenuated sorafenib-mediated CM hypertrophy in vitro.

Conclusions

By constructing the first snRNA-seq data for sorafenib-induced cardiotoxicity, we identified MITF as a key transcription factor associated with sorafenib-induced cardiotoxicity. Our study provides potential therapeutic targets for mitigating sorafenib-induced cardiotoxicity.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12916-026-05060-y.

Background

Sorafenib, a multikinase inhibitor, is widely used for the treatment of various cancers, including hepatocellular and renal cell carcinomas. However, its clinical application is often limited by adverse effects, particularly cardiotoxicity, which significantly restrict its broader therapeutic utility [1]. The cardiotoxicity of sorafenib is diverse and encompasses myocardial cell injury, myocardial ischemia, arrhythmias, and other cardiovascular pathologies. These complications severely compromise patient prognosis and quality of life, leading to treatment discontinuation and presenting substantial challenges for effective cancer therapy [2].

Previous clinical studies have indicated that sorafenib is associated with an elevated risk of major adverse cardiac and cerebrovascular events (MACE). In a study involving 843 patients who received sorafenib treatment for unresectable hepatocellular carcinoma from 2008 to 2018 (median follow-up: 11.3 months), 4% experienced MACE. According to the four-level risk classification of the European Society of Cardiology, the cumulative risk of MACE progressively increases across risk groups, and is 15.0% in the extremely high-risk group. High-risk patients can be effectively identified using the dichotomous CARDIOSOR scale, with a four-fold increase in MACE risk observed in this subgroup [3]. With the advancement of cancer treatment technologies and improvement in patient survival rates, treatment strategies associated with sorafenib-induced cardiotoxicity have garnered increasing attention. Currently, drugs that protect against sorafenib-induced cardiotoxicity include losartan, captopril, apigenin, and lupeol. These drugs or compounds inhibit cardiotoxicity through their antioxidant and anti-inflammatory effects, whereas sildenafil improves vascular dilation by activating nitric oxide. However, current prevention strategies lack specificity and demonstrate limited efficacy.

Current research on the molecular mechanisms underlying sorafenib-induced cardiotoxicity is mainly divided into on- and off-target effects. Sorafenib exerts on-target cardiotoxicity primarily by inhibiting RAF-1 protein kinase, vascular endothelial growth factor receptor (VEGFR), and platelet-derived growth factor receptor (PDGFR). By targeting the VEGFR-VEGF signaling pathway, sorafenib reduces the expression and activity of nitric oxide synthase, thereby impairing vascular dilation capacity. This compromises the heart’s ability to compensate under pressure overload, ultimately promoting progression from compensatory hypertrophy to heart failure. Additionally, PDGFR-β, as a regulator of the compensatory cardiac response under pressure load, inhibits the pressure load response of the heart through PDGFR, resulting in impaired cardiac function. Off-target effects of sorafenib include autophagy inhibition, activation of multiple regulatory cell death pathways, mitochondrial damage, endoplasmic reticulum stress, inflammatory response, and imbalance in calcium homeostasis.

The heart is a highly complex organ composed of diverse cell types including cardiomyocytes (CMs), cardiac fibroblasts (FBs), and cardiac endothelial cells (ECs). Each cell type plays a distinct role in the pathogenesis of structural heart disease and cardiac remodeling. For instance, CMs are terminally differentiated and their injury directly compromises cardiac contractility and pump function [4]. Studies suggest that sorafenib disrupts mitochondrial calcium homeostasis and increases intracellular reactive oxygen species (ROS) levels in CMs, disrupting normal physiological functions [5, 6]. Concurrently, FB activation may drive myocardial fibrosis, exacerbating structural and functional abnormalities [7], whereas cardiac EC dysfunction can impair myocardial perfusion and metabolic exchange [8]. However, a comprehensive assessment of sorafenib-induced cardiotoxicity is lacking. To address this gap, we propose employing single-nucleus RNA sequencing (snRNA-seq) to construct a high-resolution map of sorafenib-induced cardiotoxicity and elucidate the underlying mechanisms.

We used snRNA-seq to dissect the cellular heterogeneity and molecular mechanisms underlying sorafenib-induced cardiotoxicity at single-cell resolution. By comparing snRNA-seq profiles of cardiac tissues, we aimed to systematically identify differential responses across various cardiac cell types, uncover key cell clusters associated with cardiotoxicity, and delineate specific gene expression signatures. This approach may provide novel insights into sorafenib-induced cardiotoxicity and reveal previously unknown cell-type-specific mechanisms. Further functional characterization of differentially expressed genes (DEGs) within these cell clusters, along with analysis of their intercellular communication networks, may help clarify the molecular pathways driving sorafenib-induced cardiotoxicity. Hence, we aimed to establish a critical theoretical foundation for developing targeted cardioprotective strategies, optimize the clinical utility of sorafenib, and ultimately deliver safer and more effective therapeutic regimens for cancer treatment.

Methods

Animal experiments

Eight-week-old male C57BL/6J mice were purchased from GemPharmatech Co. Ltd. (China). After a one-week acclimation period, all mice were confirmed to be in good health. Inclusion Criteria: (1) Adult animals (8 weeks old); (2) There are no obvious signs of disease or infection; (3) all physiological indicators are normal. Exclusion Criteria: (1) Exclude animals with other diseases that may affect the study of onco-cardiology (e.g., cardiovascular diseases, kidney diseases, liver diseases); (2) Exclude animals with tissue injuries, inflammation and other lesions. To establish a sorafenib-induced cardiotoxicity model, 5 C57BL/6J mice were randomly assigned by drawing lots to the sorafenib treatment group, which was administered 40 mg/kg/day sorafenib (MedChemExpress, USA) by gavage daily for 8 consecutive weeks(5). An additional 5 C57BL/6J mice were randomly assigned by drawing method to the control group, which was administered an equal volume of saline by gavage for 8 weeks. The mice were housed under standard conditions with a 12-h light/12-h dark photoperiod and provided ad libitum access to food and water. All 10 mice survived the entire experimental period and were included in final analyses. All animal studies and data analyses were performed blindly by two independent investigators.

Echocardiography

Cardiac function was evaluated using transthoracic echocardiography with a small animal ultrasound machine (Vevo 2100, Canada) [9]. Cardiac dimensions were measured using M-mode echocardiography after inhalation anesthesia with 2% isoflurane. Left ventricular end-diastolic and end-systolic diameters were measured in the parasternal long-axis view. The left ventricular ejection fraction (LVEF%) and fractional shortening (LVFS%) were calculated using computer algorithms. Global ventricular longitudinal strain (GLS%) was also evaluated. Above parameters served as the primary outcome measure.

Histomorphology staining

Heart sections were dehydrated, embedded in paraffin, and cut into 3-µm-thick sections. Hematoxylin and eosin (H&E) staining was performed according to standard protocols. Stained sections were scanned under a microscope at 1× magnification (Zeiss, Germany). Masson’s trichrome staining was performed to evaluate cardiac fibrosis. Stained heart sections were scanned under a microscope at 40× magnification (Zeiss). To analyze the cross-sectional area of the CMs, cardiac tissues were stained with wheat germ agglutinin (WGA) (Sigma-Aldrich, USA) according to the manufacturer’s instructions. The stained sections were scanned under a microscope at 40× magnification (Zeiss).

Serum analysis of myocardial injury biomarkers

Serum levels of atrial natriuretic peptide (ANP), brain natriuretic peptide (BNP), creatine kinase-MB (CK-MB), and troponin T (cTNT) in the control and sorafenib group were determined according to the manufacturer’s instructions (Nanjing Jiancheng Bioengineering Institute, China).

Isolation of the nuclei from cardiac tissue

Nuclei were isolated from the cardiac tissues according to a previously published approach [10]. To prepare single-cell suspensions, cardiac tissues from three mice were collected for enzymatic digestion. After harvesting, single-cell suspensions of cardiac tissues were washed in ice-cold RPMI1640 and dissociated using Tissue Dissociation Reagent A (Seekone, China) according to the manufacturer’s instructions. DNase I (Sigma-Aldrich) treatment was performed based on the viscosity of the homogenate. Cell count and viability were estimated using a fluorescence Cell Analyzer (Countstar^® Rigel S2) with acridine orange/propidium iodide reagent after removal of erythrocytes (Solarbio, China). Subsequently, the removal of debris and dead cells was decided (Miltenyi, Germany). Finally, fresh cells were washed twice with RPMI1640 and then resuspended at a density of 1 × 10⁶ cells/mL in 1×phosphate buffered saline (PBS) and 0.04% bovine serum albumin.

Library construction of snRNA-seq and sequencing

SnRNA-seq libraries were prepared using the SeekOne^® Digital Droplet Single Cell 3 Library Preparation Kit (SeekGene, China). Briefly, appropriate numbers of cell nuclei were mixed with reverse transcription reagent and then added to the sample well in SeekOne^® Digital Droplet Chip S3. Subsequently, barcoded hydrogel beads and partitioning oil were separately dispensed into the corresponding wells in chip S3. After emulsion droplet generation, reverse transcription was performed at 42°C for 90 min and inactivated at 80°C for 15 min. Next, cDNA was purified from the broken droplets and amplified using polymerase chain reaction (PCR). The amplified cDNA product was cleaned, fragmented, end-repaired, A-tailed, and ligated to a sequencing adaptor. Finally, indexed PCR were performed to amplify the DNA with the 3’ poly A tail expressing genes which also contained a Cell Barcode and Unique Molecular Index. The indexed sequencing libraries were cleaned with SPRI beads, quantified by quantitative PCR (KAPA Biosystems, China), and then sequenced on an Illumina NovaSeq 6000 with a PE150 read length or DNBSEQ-T7 platform with a PE150 read length.

Quality filter of snRNA-seq data

To obtain high-quality true cells, we set a threshold for all samples and used cells with a mitochondrial gene read percentage of less than 15%. Consequently, a total of 155,236 cell nuclei passed through a quality control filter.

Dimensionality reduction and clustering analysis of snRNA-seq data

Using Seurat (v4.3.0), high-quality cells from six snRNA-seq datasets were analyzed for data standardization, highly variable feature selection, linear dimensionality reduction, cell clustering, and population identification. Data were normalized using a logarithmic method with global scaling. Specifically, the gene expression measurements of individual cells were normalized to the total gene expression counts per cell, multiplied by a scaling factor (default = 10,000), and subsequently subjected to logarithmic transformation to obtain the final standardized gene expression values. Based on the standardized dataset, a subset of highly variable characteristic genes (i.e., a high-variable gene set; these genes were highly expressed in some cells and expressed at low levels or absent in others) was calculated. Using variance stabilizing transformation values, the top 2,000 highly variable characteristic genes were selected for downstream analysis, thereby highlighting biologically relevant signals in the single-cell dataset.

The expression values of characteristic genes were normalized to ensure that the mean expression value of each gene across all cells was 0 and the variance of the expression value across all cells was 1. Principal component analysis (PCA) dimensionality reduction was performed on the characteristic genes, and the top 20 PCA were selected for subsequent cell clustering. Briefly, a K-nearest neighbor graph was first constructed in the PCA space using Euclidean distance-based clustering. The edge weights between any two cells were then refined by calculating their local neighborhood overlap (Jaccard similarity). The Louvain modularity optimization algorithm was subsequently applied to iteratively group the cells while optimizing the modularity function. This process incorporated a resolution parameter (set to 0.4) to control cluster granularity, where higher values yielded more distinct cell populations. Nonlinear dimensionality reduction was performed using uniform manifold approximation and projection (UMAP) to visualize cell-clustering relationships. Similar cells were placed together in a low-dimensional space by learning the basic popularity of the dataset.

Cell type annotation of snRNA-seq data

We performed clustering analysis on snRNA-seq data and conducted manual cell-type annotation based on classical marker genes reported in the literature.

Cluster-specific DEG analysis

Following cell cluster identification, marker genes were determined through DEG analysis by sequentially comparing each cell cluster against the aggregate of all the remaining clusters (one-vs.-rest approach). This analysis identified both positive (cluster-specific upregulated genes) and negative markers (cluster-specific downregulated genes) for each population. The Wilcoxon rank-sum test served as the default statistical method for DEG analysis. The criteria for DEG analysis were: the genes exhibited an average absolute log2 fold-change of 0.25 or more between the two groups and the genes showed detectable expression (> 10% cellular detection rate) in at least one population.

Differential gene function enrichment analysis of over representation analysis

To confirm group differences, we conducted differential expression analysis on several cell types in the single-cell data at different processing levels. We used the default parameters of the FindMarkers function in the Seurat package, with |logFC| greater than 0.25 and min.pct greater than 0.1 as the threshold for determining differential genes.

Over representation analysis is a commonly used method for gene function annotation and analysis. This method mainly considers the overlap degree between gene sets and biological function sets, and is used to explain and analyze the biological significance of gene sets. The basic idea is that all genes in the target gene set are functionally annotated using Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG), and then these genes are compared with the background gene set. The proportion of each functional entry appearing in the target and background gene set is calculated, and then statistical methods such as Fisher’s exact test or hypergeometric distribution test are used to calculate the enrichment degree score and identify the significantly enriched functional entries, thereby inferring the biological role played by the gene set and helping researchers better understand the biological significance and interrelationships of the gene set, providing important guidance and reference for subsequent experimental design and data analysis. Over representation analysis enrichment analysis usually has different enrichment analysis methods such as GO and KEGG depending on the reference database.

GO and KEGG pathway enrichment analysis of the DEGs

GO and KEGG pathway enrichment analyses of DEGs identified significantly enriched cellular processes and pathways in the DEGs compared with the whole genome background. Pathways with a raw p-value of less than 0.05 were considered significantly enriched and are presented in the main figures. To account for multiple testing, adjusted p- and q-values were also calculated. Raw p-values and adjusted p- and q-values for all enriched pathways are provided in Additional file 2 for full transparency. In the main text and figures, we report raw p-values to maintain sensitivity for exploratory analyses, while the supplementary tables allow readers to assess the robustness of findings using adjusted values.

Cell communication analysis

To study cell-cell interactions across distinct cell types (including subtypes), we used the CellChat software package (v1.6.1) to systematically evaluate the significance of ligand-receptor pairs among different cell clusters. Ligand-receptor interactions with a p-value of less than 0.05 were considered statistically significant for intercellular communication between the corresponding cell types. The ability of the CellChat software package (v1.6.1) to identify significant changes, namely the increase or decrease in the interaction between the sorafenib and control groups, was achieved by identifying and sequencing differentially expressed ligands and/or receptors.

Construction of transcriptional regulatory networks using pySCENIC

Through analysis using pySCENIC (v0.12.1), a single-cell transcriptional regulatory network was constructed. This approach integrates the GRNBoost2 co-expression network and cisTarget motif enrichment analysis to define direct target gene sets (regulons), and employs AUCell to quantify regulon activity in individual cells. Ultimately, this framework enabled the identification of key transcription factors specifically activated in distinct cell types, along with their corresponding regulatory networks.

Gene set scoring

The gene set score was calculated using R package AUCell (v1.16.0). Activity levels of specific gene sets in each snRNA-seq cell line were calculated. This method revealed pathway heterogeneity among different cell types, and the average expression levels in each cell type were visualized.

Augur perturbation analysis

We used the R package Augur to identify cell types that were most responsive to biological perturbations. Using a machine-learning framework, Augur prioritized perturbation-sensitive cell populations by quantifying the separability of perturbed and unperturbed cells in high-dimensional space.

Transcription factor activity prediction

To identify transcription factors (TFs) that may exert critical regulatory roles in specific cell subpopulations, we performed TF activity analysis on single-cell transcriptome data via the AUCell algorithm, leveraging a TF regulatory network defined based on prior knowledge. This process generates a TF activity scores (activity scores) which reflects the potential functional activity of the TF in a specific cell state. The resulting TF activity scores were then compared between the sorafenib and control groups across different cell types.

Gene module analysis

The hotspot tool (v1.1.1) was used to identify informative genes (and gene modules) in the snRNA-seq dataset. The counts for each gene were fitted into the depth-adjusted negative binomial model, followed by normalization of gene values, which were then smoothed based on the neighborhood of each cell. Finally, PCA was performed to reduce the dimensionality of these values across all gene features, categorizing the genes in the snRNA-seq data into distinct modules. For each functional module, the module score was calculated across all cells, and cells with scores exceeding one standard deviation above the mean of the module were identified as “high-scoring cells”.

Gene set enrichment analysis based on public databases

To comprehensively identify cell types most significantly affected by sorafenib, we curated myocardial injury-related gene sets from multiple public resources. Specifically, we queried the GeneCards database (https://www.genecards.org) to retrieve genes associated with key cardiomyopathy-related keywords, including “dilated cardiomyopathy (DCM)”, “heart failure”, “myocardial damage”, “cardiac hypertrophy”, “myocardium inflammation”, and “cardiac fibrosis” [11]. Additionally, we extracted genes involved in the calcium signaling pathway from the KEGG pathway database (https://www.genome.jp/kegg) [12]. These curated gene sets were then intersected with the list of sorafenib-affected genes identified from our snRNA-seq data. This intersection analysis allowed us to prioritize the cell types most relevant to sorafenib-induced myocardial injury.

Cell culture and transfection

Primary neonatal rat cardiomyocytes (NRCMs) were subjected to enzyme digestion. Briefly, we selected 1-3-day-old Sprague-Dawley rats (purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd., China) and dissected their hearts. The hearts were then placed in ice-cold PBS to remove blood. Subsequently, scissors were used to cut the hearts into uniform tissue fragments, which were then resuspended in PBS. Then, an appropriate amount of digestive fluid (dissolved in PBS), with 0.1% trypsin (Gibco Company, USA) and 0.05% collagenase (Gibco Company), was used to digest the heart fragments at 37 °C for 15 min. The supernatant was collected in a 15 ml centrifuge tube containing 10% fetal bovine serum (FBS) to terminate the digestion process. The cell suspension was centrifuged at 1000 rpm for 5 min. The supernatant was discarded, and the cells were resuspended in a culture medium containing 10% FBS. After 2 h, the unadhered cells were collected and resuspended in one or two adherent cell culture dishes, with the culture medium being Dulbecco’s modified Eagle’s medium (DMEM) containing 20% FBS. After 48 h, the culture medium was replaced with DMEM supplemented with 10% FBS.

For in vitro modeling of sorafenib cardiotoxicity, H9C2 and HL-1 cells received 2.5 or 5 µM sorafenib, while NRCMs received 5 µM (MedChemExpress, USA), all for 24 h. To clarify the effect of melanocyte inducing transcription factor (MITF) on sorafenib-induced cardiotoxicity, MITF small interfering RNA (siMITF, JTS scientific, China) were transfected into NRCMs, H9C2 cells, or HL-1 cells using lipofectamine RNAi MAX (ThermoFisher Scientific, USA), followed by sorafenib stimulation (MedChemExpress, 5µM) for an additional 24 h. Using the same method, the MITF overexpression plasmid (pc-MITF, OBiO Tech, China) was transfected to verify the role of MITF in the sorafenib-induced cardiotoxicity. The siMITF, pc-MITF and siSEC22B (targeting Sect. 22 homolog B) sequences were listed in Additional file 2: Table. S1.

Immunofluorescence staining

For phalloidin staining, CMs were fixed in 4% paraformaldehyde (15 min) and blocked with goat serum (30 min), and incubated with phalloidin solution (100 nM, 30 min, 25°C), followed by washing. Nuclei were stained with 4’,6-diamidino-2-phenylindole (DAPI) (10 min). Images were captured by confocal microscopy (Zeiss). Anti-α-actinin (1:100, Abcam) was used for immunofluorescence.

Quantitative real-time polymerase chain reaction

Total RNA from CMs and cardiac tissues was extracted using TRIzol reagent (TIANGEN BIOTECH, China). Total RNA was converted to cDNA using the PrimeScript RT Reagent kit (Takara, Japan). Amplification using quantitative real-time polymerase chain reaction (qRT-PCR) was performed using SYBR Green 2×PCR mix (Takara, Japan). The primer sequences were listed in Additional file 2: Table. S1. 18 S rRNA gene expression was used for normalization. Relative quantification of the target genes was analyzed according to the 2^−ΔΔCt method.

Western blot

The total proteins from CMs and cardiac tissues were lysed using radioimmunoprecipitation assay buffer supplemented with a protease inhibitor cocktail (ThermoFisher Scientific). Protein concentrations were quantified using a BCA Protein Assay Kit (ThermoFisher Scientific). Extracted proteins were separated via SDS-PAGE, and the proteins were transferred to polyvinylidene difluoride membranes (Millipore, USA). The membranes were blocked in 5% skim milk and incubated with the respective primary antibody overnight at 4 °C. The primary antibodies used were anti-SEC22B (1:1000, Affinity Biosciences), anti-MITF (1:1000, Cell Signaling Technology) and anti-GAPDH (1:1000, Cell Signaling Technology). After incubation, the membranes were incubated with horseradish peroxidase-conjugated secondary antibodies (Jackson ImmunoResearch) for 1 h at room temperature. The signal was detected and quantified using an Amersham Imager 680 (GE Healthcare, USA). Images were quantified using ImageJ software.

Statistical analysis

All statistical analyses were performed using SPSS v22.0 (IBM Inc., Armonk, NY, USA), and graphical representations were generated with GraphPad Prism (version 9.1.0). Normal distribution of the data was tested by Shapiro-Wilk test. The data in all bar graphs are presented as mean and standard deviation (SD) from at least three independent experiments. Differences between two groups were analyzed using the Student’s two-tailed t-test. Differences between multiple groups were analyzed using the one-way analysis of variance. A p-value of less than 0.05 was considered statistically significant.

Results

Morphological characteristics of sorafenib-induced cardiotoxicity

To investigate the intrinsic mechanism of sorafenib-induced cardiotoxicity, we established an oral gavage model of sorafenib-induced cardiotoxicity by daily administration of 40 mg/kg sorafenib for 8 weeks (Fig. 1A). We found that sorafenib did not affect body weight, ratio of heart weight to body weight, or ratio of heart weight to tibial length (Additional file 1: Fig. S1A, B). However, cardiac function (LVEF% and LVFS%) was significantly decreased in the sorafenib group compared with the control group (Figs. 1B, C). The GLS% is an indicator of early left ventricular dysfunction. Compared with the control group, the GLS% in the sorafenib group was significantly decreased (Figs. 1D, E). The mRNA levels of Anp and Bnp in the hearts of the sorafenib group were higher than those in the control group (Fig. 1F). Subsequently, we measured the serum levels of myocardial injury biomarkers and found that the levels of ANP, BNP, CK-MB, and cTNT were significantly higher in the sorafenib group than in the control group (Fig. 1G). Moreover, cardiac hypertrophy and fibrosis occurred in the hearts of the sorafenib group compared with the control group (Figs. 1H, I). Descriptive statistics for all endpoints are presented in Additional file 2: Table. S2. These results collectively confirmed the successful establishment of a mouse model of sorafenib-induced cardiotoxicity.

Fig. 1.

Establishment of sorafenib-induced cardiotoxicity in C57BL/6J mice. A The schematic diagram outlines the sample, analysis and verification methods of sorafenib-induced cardiotoxicity. B Results of M-mode echocardiography in the control and sorafenib groups. C Changes in LVEF (%) and LVFS (%) in the control and sorafenib groups (n=5). D Changes in GLS (%) in the control and sorafenib groups (n=5). E Vevo Strain analysis of myocardial strain in the control and sorafenib groups. F The mRNA levels of Anp and Bnp in the hearts of mice from the control and sorafenib groups using by qRT-PCR (n=5). G The serum levels of ANP, BNP, CK-MB, and cTNT in the mice of the control and sorafenib groups using ELISA (n=5). H-I Representative images and quantitative results of H&E, Masson, and WGA staining in the hearts of mice from the control and sorafenib groups (n=5). Sor, sorafenib; LVEF, left ventricular ejection fraction; LVFS, left ventricular fractional shortening; GLS, global ventricular longitudinal strain; ANP, atrial natriuretic peptide; BNP, brain natriuretic peptide; CK-MB, creatine kinase-MB; cTNT, cardiac troponin T; ELISA, enzyme-linked immunosorbent assay; H&E, hematoxylin and eosin. Data are presented as means ± SD

SnRNA-seq profiling of sorafenib-induced cardiotoxicity

To comprehensively analyze the effects of sorafenib-induced cardiotoxicity on distinct cells in the heart, we performed snRNA-seq of heart tissues from the control and sorafenib groups. After rigorous quality control, we subjected 155,236 high-quality cells to PCA-based dimensionality reduction analysis (Fig. 2A and Additional file 1: Fig. S1C, D, E, F, G, H). Based on the unique molecular markers in each cell cluster, the cells in the hearts of sorafenib and control groups were classified into 10 distinct types: ECs, myeloid cells, FBs, CMs, epicardium cells, pericytes, B cells, T cells, smooth muscle cells (SMCs), and Schwann cells (Fig. 2B and Additional file 1: Fig. S1I). The UMAP plot shows the expression levels of the marker genes for specific cell types in the heart (Additional file 1: Fig. S1J). Subsequently, enrichment analysis was conducted on different cell types based on the marker genes, revealing the corresponding biological functions and characteristics of the different cell populations (Fig. 2C). For instance, the marker genes of CMs were associated with cardiac muscle contractility, those of ECs were enriched in angiogenesis factors, and those of FBs were highly enriched in extracellular matrix (ECM) organization factors.

Fig. 2.

Single-nucleus RNA sequencing profiling of sorafenib-induced cardiotoxicity. A UMAP plot showing the distribution of different cell types in the heart of control and sorafenib-treated mice. B Dot plots indicating the expression of representative genes for each cell type in the hearts from control and sorafenib-treated mice. C Heatmap showing the expression signatures of top 50 cell-type specific genes (Left). Enriched representatives GO terms for each cell type (Right). Colors indicate various cell types; the length of bar indicates −log10 (P-value). D Sankey plot showing the number and the proportions of each cell type in control and sorafenib groups. The length of the bar shows the number of cells. The pie chart showing the ratios of cell types in hearts of control and sorafenib-treated mice. Statistical comparisons between the sorafenib and control groups were performed using a two-tailed unpaired Student's t-test. For all cell types examined, the differences between groups did not reach statistical significance (p > 0.05). E Immunofluorescence staining of CM marker cTNT (red) in the hearts of control and sorafenib-treated mice (n=3). F-G Immunofluorescence staining for Pecam1 (F, green) and CD68 (G, green) in the hearts of control and sorafenib-treated mice (n=3). Sor, sorafenib; UMAP, uniform manifold approximation and projection; GO, Gene Ontology; CM, cardiomyocyte; EC, endothelial cell; FB, fibroblast; SMC, smooth muscle cell; Per, pericyte. Data are presented as means ± SD

To explore the main effects of sorafenib-induced cardiotoxicity, we analyzed the dynamic changes in cardiac tissue cell composition between the control and sorafenib groups. The results indicated that the proportions of CMs and ECs were reduced, and the degree of decline in ECs was more significant in the sorafenib group (Fig. 2D and Additional file 1: Fig. S1K). Furthermore, the proportions of pericytes, SMCs, and immune cells were increased in the hearts of the sorafenib group (Fig. 2D and Additional file 1: Fig. S1K). However, the proportion of FBs did not significantly differ between the control and sorafenib groups (Fig. 2D and Additional file 1: Fig. S1K). To further evaluate whether sorafenib treatment altered the relative abundance of various cell types in the heart tissue, we quantified the proportion of each cell type in the sorafenib and control groups. No statistically significant differences were detected for any of the cell types analyzed (Fig. 2D, Additional file 1: Fig. S1K and Additional file 2: Table. S3). Based on the observed changes in these cells after sorafenib treatment, we performed further validation via immunofluorescence. We observed a significant reduction in the CM-to-total nuclei ratio in the sorafenib group compared with the control group. This finding indicates a shift in myocardial cellular composition, characterized by a lower fractional representation of CMs (Fig. 2E); moreover, the proportions of platelet-and endothelial cell adhesion molecule 1 (Pecam1)-positive ECs in the heart tissues of the sorafenib group were lower (Fig. 2F). This is consistent with the pharmacological effects of sorafenib, which inhibit receptor tyrosine kinases including VEGFR2 and PDGFR [13, 14]. Additionally, immune cell infiltration, including myeloid and B cells, significantly increased after sorafenib treatment (Fig. 2G). Consistently, the proportion of FBs in the sorafenib group did not differ significantly from that in the control group (Additional file 1: Fig. S1K, L).

Cell type-specific alterations in transcriptional regulatory networks of sorafenib-induced cardiotoxicity

To distinguish between the common and unique gene expression changes associated with sorafenib-induced cardiotoxicity across different cell clusters, we characterized the DEGs in each cluster. A total of 4,069 DEGs were identified under the criteria of |log2FoldChange|> 0.25, and adjusted for a p-value of less than 0.01 (Figs. 3A, B). The cell types most significantly affected by sorafenib were CMs, myeloid cells, and epicardial cells, with 514, 609, and 846 DEGs, respectively (Figs. 3A, B).

Fig. 3.

Cell type-specific alterations in transcriptional regulatory programs of sorafenib-induced cardiotoxicity. A-B The circos plots illustrate the situation of upregulated DEGs (A) or downregulated DEGs (B) in different cell types of the control and sorafenib groups, respectively. Each boundary line indicates a DEG that exists simultaneously in both cell types. C-D Dot plot shows the enriched KEGG pathways of DEGs in the nine cell types with the highest number of DEGs in the hearts of control or sorafenib-treated mice (C: upregulated, D: downregulated). The color scale represents the statistical significance of enrichment (raw p-value). E-H Violin plot showing the AUC score of calcium signaling pathway gene set, cardiac muscle contraction gene set, myocardium inflammation gene set, and focal adhesion gene set in CMs. I-J Violin plot showing the AUC score of focal adhesion gene set and myocardium inflammation gene set in FBs. Sor, sorafenib; DEGs, differentially expressed genes; KEGG, Kyoto Encyclopedia of Genes and Genomes; CMs, cardiomyocytes; FBs, fibroblasts; AUC, area under the curve

To further investigate the pathways associated with sorafenib-induced cardiotoxicity across the different cell clusters, GO term annotation analysis was performed. In CMs, the significantly enriched GO terms (p < 0.05) were primarily associated with cardiac muscle contraction, fiber development, and cell differentiation (Additional file 1: Fig. S2A, B and Additional file 2: Table. S4). GO annotations of ECs were significantly enriched for angiogenesis, cardiac muscle fiber development, and ECM organization (Additional file 1: Fig. S2A, B and Additional file 2: Table. S4). GO annotations of FBs were significantly enriched for ECM organization, cell population proliferation, and cellular responses to oxidative stress (Additional file 1: Fig. S2A, B and Additional file 2: Table. S4). All three terms were highly significant (Additional file 2: Table. S4). Myeloid cells were significantly enriched in endocytosis, immune system processes, and phagocytosis (Additional file 1: Fig. S2A, B and Additional file 2: Table. S4). Complete results, including exact adjusted p-values for all enriched terms, were provided in Additional file 2: Table. S4. The upregulated DEGs were significantly associated with pro-inflammatory signals, including Dnm2, Rab1a, Vav family members, and Dock family members(15–17], as well as other myocardial disease-related signals, such as genes related to calcium ion metabolism disorders (Ryr2, Tnnt2, and Ank2) (Additional file 1: Fig. S2A and Additional file 2: Table. S4)(18–20]. The downregulated DEGs were significantly enriched in pathways, including the collagen-activated tyrosine kinase receptor signaling pathway (Ddr2, Col4a1, and Col4a2) [21, 22], angiogenesis (Cfh, Clic4, Rhoj, and Tek) [23–25], and ERK1/ERK2 cascade (Zfp36l1, Smad4, and Kdr) (Additional file 1: Fig. S2B and Additional file 2: Table. S4) [26, 27]. These findings are consistent with the known pharmacological effects of sorafenib on the inhibition of tyrosine enzymes.

KEGG pathway analysis was conducted for each cell cluster associated with sorafenib-induced cardiotoxicity. CMs exhibited the most significant enrichment, with pathways primarily related to DCM, calcium signaling, cardiac muscle contraction, and hypertrophic cardiomyopathy (HCM) (Figs. 3C, D and Additional file 2: Table. S5). EC pathways were observably enriched in adherens junctions, DCM, focal junctions, and HCM (Figs. 3C, D and Additional file 2: Table. S5). The FB pathways were markedly enriched in the ECM-receptor interaction, HCM, and Wnt signaling pathways (Figs. 3C, D and Additional file 2: Table. S5). Myeloid cells were significantly enriched in the B-cell receptor signaling pathway, phagocytosis, and phagosomes (Figs. 3C, D and Additional file 2: Table. S5). Subsequent enrichment analysis of SMCs and pericytes also revealed a disruption in the ECM-receptor interaction in the sorafenib group (Figs. 3C, D and Additional file 2: Table. S5). Complete results, including exact adjusted p-values for all enriched terms, are provided in Additional file 2: Table. S5. Despite cell-specific DEG patterns, common KEGG pathways and GO term annotations were identified across multiple clusters. The upregulated DEGs in various cell types, such as CMs, ECs, FBs, and pericytes, indicated cardiomyopathy-related pathways, including HCM and DCM (Fig. 3C). In contrast, most downregulated DEGs were associated with the PI3K-Akt signaling pathway, which promotes cell metabolism, proliferation, survival, and growth, and is involved in cardiac remodeling, regeneration, and post-ischemic repair after myocardial injury (Fig. 3D). In addition, both upregulated and downregulated DEGs were enriched in pro-fibrotic pathways, including focal adhesion and ECM-interactions, which is consistent with the previously reported effect of sorafenib in disrupting the stromal microenvironment of the cardiac tissue (Figs. 3C, D).

Furthermore, we conducted gene set scoring for CMs and FBs. The AUCell scores of myocardial contraction and calcium ion signaling pathways in CMs were abnormally activated after sorafenib treatment (p < 0.01 for all comparisons, Figs. 3E, F). The area under the curve (AUC) scores for focal adhesions and myocardial inflammation in the CMs and FBs were significantly higher in the sorafenib group than in the control group (p < 0.01 for all comparisons, Figs. 3G, H, I, J). Combined with the results of the DEG enrichment analysis and AUCell scoring, we concluded that sorafenib-induced cardiotoxicity involved multiple biological processes, including pro-inflammatory and pro-fibrotic processes.

Analysis of cellular communication patterns in sorafenib-induced cardiotoxicity

We used the Cell Chat tool (v1.6.1) to investigate the communication of different cell types in cardiac tissues. To analyze the interaction patterns between the sorafenib and control groups, we visualized the quantity and intensity of cellular interactions using heat maps and circular plots (Figs. 4A, B and Additional file 1: Fig. S3A, B). When SMCs were designated as “Targets,” the number and intensity of cell communications with CMs, FBs, pericytes, and Schwann cells was considerably enhanced. We observed disordered ligand-receptor pairs of Lama5-(Itga1 + Itgb1), Lama5-(Itga7 + Itgb1), and Lama2-(Itga7 + Itgb1) between CMs and SMCs, which were closely associated with vascular dilation and constriction dysfunction (Fig. 4C and Additional file 1: Fig. S3C). Sorafenib may disrupt these pathways, potentially resulting in microvascular destabilization, fibrosis, and impaired myocardial repair [28, 29]. Furthermore, our analysis revealed a reduction in intercellular communication among CMs, SMCs, and ECs, indicating myocardial structural disorganization during sorafenib-induced cardiotoxicity. Reduced crosstalk appeared to be mediated by the direct toxicity of sorafenib to ECs via the ligand-receptor pairs Vegfa-Vegfr2 and Ptprm-Ptprm (Fig. 4C and Additional file 1: Fig. S3D).

Fig. 4.

The transcriptional landscape and intercellular communication underlying sorafenib-induced cardiotoxicity. A-B Heatmap showing changes in the number and strength of ligand-receptor interactions detected between each two cell types and/or within a single cell type in the sorafenib vs. control group, respectively. The changes in the number (A) or strength (B) of ligand-receptor interactions are showed by color block colors. The red color block indicates an enhancement in the number or strength of ligand-receptor interactions in the sorafenib group, and the blue color block indicates a reduction. C Identification of signaling by comparing the communication probabilities mediated by ligand-receptor pairs from CMs to other cell types in the control and sorafenib groups. The genetic types of ligands and receptors are indicated below. Color dots from violet to red indicate the ligand-receptor interaction events from low to high. D-E. Plots showing significantly downregulated (D) and upregulated (E) DEGs (raw p-value < 0.05) shared by three cell types (CMs, FBs, and ECs). Color keys from light to dark indicate the expression of target genes from low to high. F Dot plots showing that DEGs overlapped with genes from heart disease-associated and cardiotoxicity pathological process-associated gene sets. Red nodes represent the upregulated DEGs. Blue nodes represent the downregulated genes. The node size shows the frequency of DEGs. Sor, sorafenib; DEGs, differentially expressed genes; CMs, cardiomyocytes; FBs, fibroblasts; ECs, endothelial cells

We observed a pronounced enhancement in the interaction intensity between ECs and FBs compared with other cell types, with the ligand-receptor pair Ptprm-Ptprm playing a central role in this process (Additional file 1: Fig. S3D). We also found that the interaction intensity between FBs and various cells, such as CMs, ECs, and pericytes, was significantly increased (Fig. 4B and Additional file 1: Fig. S3C, D). Notably, the increased cellular communication between CMs and FBs aligned with their respective AUCell scores, suggesting functional overactivation of FBs in the sorafenib group.

This hyperactive FB phenotype may be attributed to dysregulated ligand-receptor pairs, including Lamc1-Dag1 and Hspg2-Dag1, which mainly exert pathological effects, such as myocardial hypertrophy, calcium signal disorders, and fibrosis (Additional file 1: Fig. S3D). Furthermore, although FBs exhibited an increased number of interactions with SMCs, their interaction intensity with pericytes was the strongest. Among them, key ligand-receptor pairs, such as Col6a2-(Itga1 + Itgb1) and Col4a2-(Itga1 + Itgb1), were aberrantly activated between FBs and pericytes, which directly drove sorafenib-induced myocardial fibrosis (Additional file 1: Fig. S3D). These ligand-receptor pairs may also promote myocardial fibrosis by transdifferentiating pericytes into myofibroblasts.

To further visually demonstrate whether FBs were in an activated state, we conducted sub-clustering analyses on FBs using established FB activation marker genes. First, we subclustered FBs using Seurat and identified six distinct FB subpopulations (Additional file 1: Fig. S4A, B). Then, we compiled an FB activation gene signature based on well-established markers from the literature, including myofibroblast (periostin [Postn], actin alpha 2, smooth muscle [Acta2], collagen triple helix repeat containing 1 [Cthrc1], fibronectin 1 [Fn1]) and ECM-related markers (collagen type I alpha 1 [Col1a1], collagen type III alpha 1 [Col3a1], collagen type V alpha 1 [Col5a1], collagen type VI alpha 1 [Col6a1]). Using this gene set, we calculated an activation score for each FB using the AUCell algorithm and compared scores between the sorafenib and control groups across and within individual subclusters of FBs. The results showed that an increase was observed across multiple FB subclusters (Additional file 1: Fig. S4C), suggesting that activation was not restricted to a rare subpopulation but represented a widespread phenotypic shift from FB inactivation to activation. Additionally, we found that individual activation markers, particularly ECM-related markers, showed consistent upregulation in sorafenib-treated FBs (Additional file 1: Fig. S4C), consistent with the histological fibrosis observed in our study. These findings demonstrate that sorafenib treatment induced FB activation without altering overall FB abundance.

Collectively, these findings reveal that sorafenib induces a pro-inflammatory and pro-fibrotic state in cardiac tissues through the dysregulation of specific molecular mechanisms.

Hotspot genes in sorafenib-induced cardiotoxicity

Previous studies have reported that CMs and ECs are the primary effectors of sorafenib-induced cardiotoxicity [2]. By integrating these established findings with our multi-dimensional analyses, including shifts in cellular composition, DEG-based pathways, functional enrichment, and cell-cell communication networks, we focused our investigation on CMs, ECs, and FBs in the subsequent analysis.

We focused on common DEGs among CMs, ECs, and FBs during sorafenib-induced cardiotoxicity (Figs. 4D, E and Additional file 2: Table. S6). Among the observably down-regulated DEGs were several genes implicated in cardioprotection, including solute carrier family 25 member 3 (Slc25a3), KLF transcription factor 9 (Klf9), and ferritin heavy chain 1 (Fth1). These genes have been linked to antioxidant stress, inhibition of fibrosis, and inflammation regulation in the cardiovascular system (Fig. 4D). Slc25a3 is a mitochondrial membrane-localized transporter protein that plays a critical role in regulating energy and amino acid metabolic homeostasis after myocardial injury. Slc25a3 knockdown exacerbates the progression of pathological cardiac hypertrophy progression [30, 31]. Klf9 exerts dual cardioprotective effects through distinct molecular mechanisms: it enhances the expression of antioxidant enzymes via activation of the NFE2 like BZIP transcription factor 2 (Nrf2)-antioxidant response element (ARE) pathway, thereby mitigating CM apoptosis, while simultaneously inhibiting the transforming growth factor beta 1 (Tgf-β1)-SMAD family member 3 (Smad3) pathway to suppress FB activation and collagen deposition, ultimately attenuating myocardial fibrosis [32]. Meanwhile, Fth1, the heavy chain of ferritin, counteracts hypoxia-induced stress in CMs by inhibiting lipid peroxidation and chelating iron ions to resist the hypoxia stress suffered by CMs [33].

Among the upregulated DEGs, circadian rhythm-related proteins (period circadian regulator 3 [Per3] and thyrotroph embryonic factor [Tef]) were consistently upregulated in CMs, ECs, and FBs under sorafenib-induced cardiotoxicity. Per3 activation effectively inhibits myocardial remodeling and regulates blood pressure [34]. Upregulated Rora exerts key cardioprotective effects through anti-inflammatory, antioxidative, and anti-apoptotic mechanisms [35, 36]. WW domain-containing oxidoreductase (Wwox) modulates stress and apoptotic responses by regulating p53 and mitogen-activated protein kinase 8 (JNK1) activation and may exacerbate cardiac remodeling by promoting apoptosis and inflammation [37].

Subsequently, we identified a subset of DEGs that overlapped with genes associated with cardiovascular diseases, including cardiac hypertrophy, heart failure, and DCM (Fig. 4F). Overlapping DEGs in CMs, CFs, and FBs were associated with myocardial inflammation, cardiac hypertrophy, cardiac fibrosis, myocardial damage, heart failure, and DCM. Notably, most of these DEGs were enriched in CMs. The upregulated DEGs in CMs included Mybpc3, Myh6, Myl2, Ryr2, Scn5a, Tnnc1, Tnnt2, and Ttn, whereas the downregulated genes included Adra1b, Angpt1, Cp, Csrp3, Ctsl, Gsk3b, Hand2, Itgb1, Pln, and Ppargc1a (Additional file 1: Fig. S5A). The upregulated DEGs in FBs included Actb and Tbx20, whereas the downregulated genes included Adam17 and Cp (Additional file 1: Fig. S5B). In addition, the upregulated DEGs in ECs were Actb, Calr, Myh6, and Ttn, while those in ECs were Cp and Nos3 (Additional file 1: Fig. S5C). These overlapping DEGs revealed the cell type-specific mechanisms of sorafenib-induced cardiotoxicity. Collectively, these findings highlight key hotspot genes and provide new insights into sorafenib-induced cardiotoxicity.

CMs were the most sensitive cell type for sorafenib-induced cardiotoxicity

To further characterize the transcriptional characteristics of sorafenib-induced cardiotoxicity, we performed Augur analysis on six major effector cell types in cardiac tissues (CMs, ECs, FBs, myeloid cells, epicardial cells, and pericytes). We found that CMs were most significantly affected by sorafenib treatment (Fig. 5A), which was closely associated with fibrosis repair, the inflammatory microenvironment, and microcirculation dysfunction induced by sorafenib, collectively contributing to CM dysfunction. To comprehensively identify the cell types most significantly affected by sorafenib, we curated myocardial injury-related gene sets from multiple public databases and performed an intersection analysis with the genes identified from our snRNA-seq data. Compared with other cell types, the gene hotspots related to heart diseases (DCM and heart failure) and pathological cardiotoxicity (myocardial damage, cardiac hypertrophy, calcium signaling pathway disorders, myocardial inflammation, and cardiac fibrosis) were highly enriched in CMs (Fig. 5B).

Fig. 5.

MITF is a core transcription factor of sorafenib-induced cardiotoxicity. A Bar plot showing the prioritization of the most responsive cell types during sorafenib-induced cardiotoxicity by Augur analysis. B Dot plots showing the overlapping gene number between the DEGs in each cell types and genes from the datasets of heart diseases and pathological process of cardiotoxicity. The size of the circle represents the number of overlapping genes. Color keys from green to red represent the number of overlapping genes. C Hotspot modules in all cells. The heatmap comprises 491 highly variable genes with significant autocorrelation (FDR < 0.01). D Gene modules are visualized with their summary, per bar code, module scores. E The bubble chart shows KEGG-enriched items of module 1 analyzed by WGCNA. The color scale represents the statistical significance of enrichment (raw p-value), with red indicating higher significance (smaller p-values). F Plot showing regulon activity and number of target genes of 20 regulators regulating sorafenib-induced cardiotoxicity in CMs. The size of the dots is positively correlated with the number of target genes. Sor, sorafenib; KEGG, Kyoto Encyclopedia of Genes and Genomes; DEGs, differentially expressed genes; CM, cardiomyocyte; FDR, false discovery rate; ARVC, arrhythmogenic right ventricular cardiomyopathy; DCM, dilated cardiomyopathy; HCM, hypertrophic cardiomyopathy. *Ra > 0.1, **Ra > 0.17

To further identify the direct and core cell types involved in sorafenib-induced cardiotoxicity, we quantified the associations between gene co-expression modules and cardiomyopathy-related phenotypes. WGCNA was performed on the integrated snRNA-seq dataset using the Hotspot tool (v1.1.1) to identify functionally relevant gene modules. Combined WGCNA and KEGG enrichment analyses revealed that module 1 was significantly enriched in pathways related to cardiomyopathy, cardiac muscle contraction, and calcium ion channels (Figs. 5C, D, E and Additional file 1: Fig. S6). Therefore, we focused on the analysis of the main effect cells within module 1. We then calculated a module score for each functional module across all cells and defined those with scores exceeding one standard deviation above the mean as “high-scoring cells”. The results demonstrated that CMs were the predominant high-scoring cells in module 1 (Fig. 5D). Furthermore, DEGs in CMs showed marked enrichment and compensatory upregulation of calcium signaling pathways and cardiac muscle contraction.

Given the significant increase in DEGs in CMs, along with the KEGG enrichment results showing that DEGs associated with HCM, DCM, and right ventricular cardiomyopathy (ARVC) were predominantly concentrated in the CM population, we selected CMs as the primary focus for sorafenib-induced cardiotoxicity (Fig. 5E). Therefore, we conducted a gene enrichment analysis of CMs and observed that DEGs were significantly enriched in CM contractions (Cacna1c, Myh6, Tnnc1, Tnnt2, Ryr2, Myl2, and Tnni3). Beyond these shared contraction-related genes, the upregulated DEGs involved in HCM and DCM (Mybpc3, Ttn, Itgb6, and Itga9) were significantly enriched. Correspondingly, adrenergic signaling in CMs (Scn5a, Camk2a, Rapgef4, and Calm2) was significantly upregulated (Additional file 2: Table. S7). Conversely, the downregulated DEGs were significantly enriched in the cGMP-PKG signaling pathway (Pln, Prkg1, Pde3a, Irs1, Ppp1cb, Rgs2, Ednra, Adra1b, Plcb4, and Ppp3ca), which exerts cardioprotective effects through anti-hypertrophic/fibrotic actions, reduces oxidative stress/inflammatory factor expression, and maintains endothelial barrier integrity (Additional file 2: Table. S8). The downregulated DEGs were enriched in ECM-receptor interactions (Col4a1, Cd47, Col1a2, and Lamc1), indicating dysregulation of the microenvironment after myocardial injury (Additional file 2: Table. S8). Collectively, these results demonstrate that structural and functional impairments in CMs play a central role in sorafenib-induced cardiotoxicity.

MITF was a core TF for sorafenib-induced cardiotoxicity

To further identify the core TFs associated with sorafenib-induced cardiotoxicity within these major effector cell types, we performed single-cell regulatory network inference and clustering (SCENIC) of CMs, ECs, and FBs (Additional file 1: Fig. S7A). Among the co-upregulated TFs shared by CMs, ECs, and FBs, zinc finger and BTB domain containing 20 (Zbtb20), nuclear factor I A (Nfia), nuclear factor I B (Nfib), and PBX homeobox 1 (Pbx1) are all involved in cardiac development and are expressed at low levels in adult hearts. However, these genes were reactivated following sorafenib treatment. Nfia and Nfib cooperatively amplify pro-inflammatory signaling and are involved in the inflammatory response and apoptosis of CMs [38, 39]. Moreover, Pbx1 and Zbtb20 exhibit aberrant activation that disrupts myocardial regeneration and accelerates maladaptive remodeling under pathological conditions [40–42]. Other commonly upregulated TFs, including Klf12, BCL2 associated transcription factor 1 (Bclaf1), and nuclear receptor subfamily 3 group C member 1 (Nr3c1), were also activated after sorafenib treatment (Additional file 1: Fig. S7A). These TFs accelerate myocardial injury by promoting inflammation, fibrosis, and apoptosis [43–45].

To further investigate cell type-specific TFs, we examined the core TFs in ECs, CMs, and FBs. In CMs, the transcriptional activity of MITF increased in the sorafenib group (Fig. 5F and Additional file 1: Fig. S7B). Highly expressed MITF may serve as a metabolic-inflammatory crosstalk node, and modulation of its activity could improve energy supply and oxidative balance in the hypertrophic myocardium [46]. Moreover, the transcriptional activity of forkhead box P1 (FOXP1) in ECs was lower in the sorafenib group than in the control group (Additional file 1: Fig. S7C, D). The dysregulation of FOXP1 is associated with cardiovascular diseases, including cardiac hypertrophy, heart failure, and DCM [47, 48]. The transcriptional activity of transcription factor 12 (Tcf12) in FBs increased in the sorafenib group (Additional file 1: Fig. S7E, F). As a transcriptional promoter of Tgf-β1, the highly expressed Tcf12 contributes to collagen distribution imbalance in the heart [49].

In the TF regulatory network analysis, the transcriptional activity of MITF in CMs was most significantly upregulated in the sorafenib group (Fig. 5F). MITF is a DNA-binding protein with a basic helix-loop-helix structure and rich leucine content, and is recognized as a dominant TF. It is expressed in various cell types, including melanocytes and mast cells [50, 51]. The heart-specific isoform MITF-H is highly expressed in CMs and plays an important role in cardiac hypertrophy. MITF transcriptionally activates the actin light chain 1a, leading to enhanced contractility and increased calcium sensitivity [52]. The functional role of MITF observed in the current study aligned with both the snRNA-seq results and the pathological phenotypes observed in sorafenib-induced cardiotoxicity. Together, these findings identify MITF as a core regulatory TF in CMs during sorafenib-induced cardiotoxicity.

Subclustering analysis of CMs revealed distinct subpopulations with differential MITF expression

To investigate the heterogeneity of MITF expression among CMs, we performed subclustering analysis on the CM population identified from our snRNA-seq data (Additional file 1: Fig. S8A). Unsupervised clustering of CMs resolved eight distinct subclusters based on their transcriptional profiles (Additional file 1: Fig. S8A). The top marker genes for each subcluster were shown in Additional file 1: Fig. S8B, confirming their distinct molecular identities. Next, we examined the expression pattern of MITF across these CM subclusters. As shown in the UMAP feature plot, MITF expression was detectable in all CM subclusters but exhibited significant enrichment in specific subpopulations (Additional file 1: Fig. S8C). Quantitative analysis revealed that MITF was significantly upregulated in subcluster 0 (p = 6.59E-25, logFC = 0.34) and subcluster 7 (p = 0.004, logFC = 0.47) compared to other CM subpopulations (Additional file 1: Fig. S8D). Given that subcluster 0 represents the largest CM subpopulation, we focused on this cluster for further functional analysis. The modest but highly significant logFC (0.34, p = 6.59E-25) in subcluster 0 indicates that MITF dysregulation was a consistent and widespread phenomenon across the majority of CMs, rather than a dramatic change confined to a rare subpopulation. This global shift in transcriptional regulation is particularly relevant for a transcription factor like MITF, as it suggests a broad functional impact on CM biology.

When we compared subcluster 0 against the combined remaining subclusters (subclusters 1–7), MITF expression was significantly higher in subcluster 0 (Additional file 1: Fig. S8E). To gain insights into the functional role of subcluster 0, we performed KEGG pathway analysis on its marker genes. Interestingly, subcluster 0 was significantly enriched in pathways related to ARVC, DCM, and HCM (Additional file 1: Fig. S8F), linking this major CM subpopulation to key pathological processes in cardiac disease. Based on these findings, we selected subcluster 0 as the primary CM subpopulation for subsequent investigation of MITF’s role in pro-hypertrophic signaling.

MITF expression was increased in vivo and in vitro

To explore the role of MITF in sorafenib-induced cardiotoxicity, we conducted an in-depth analysis of the snRNA-seq data. MITF participated extensively in the upstream regulation of 178 sorafenib-induced cardiotoxicity-related DEGs in the CMs (Fig. 6A). Consequently, increased MITF expression led to dysregulation of these target genes following sorafenib treatment. Specifically, the upregulated MITF target genes were associated with ARVC (Ryr2, Dsg2, and Dmd), cardiac muscle contraction (Atp1a3 and Ryr2), HCM and DCM (Ryr2 and Dmd) (Fig. 6B and Additional file 2: Table. S9). These data indicate that upregulation of MITF and hyperactivation of its downstream transcriptional network may play an important role in sorafenib-induced cardiotoxicity. According to the existing literature, both MITF and Brahma-related gene 1 are activated in CMs under hypertrophic pathological stimulation and synergistically activate the expression of GATA Binding Protein 4 (GATA4), leading to the promotion of cardiac hypertrophy [53]. GATA4, a key hypertrophic TF in the heart, participates in the transcriptional regulation of BNP and ANP as well as in the re-expression of embryonic heart genes. These findings are consistent with the results of our analysis of TFs in CMs (Fig. 6F). These findings suggest that MITF may be a key target for alleviating sorafenib-induced cardiotoxicity.

Fig. 6.

MITF expression is increased in the hearts and cardiomyocytes after sorafenib treatment. A UMAP plot showing activity score of the MITF target genes in CMs. B The bubble chart shows representative terms of MITF target gene enrichment in CMs. The size of the circle shows the number of genes in each KEGG term. Color keys from blue to red indicate the expression of -log10 (p-value). C The mRNA level of Mitf in the hearts of control and sorafenib group mice is examined using by qRT-PCR (n=5). D The protein expression of MITF in the hearts of control and sorafenib group mice is detected using western blot (n=5). E Immunohistochemical staining of MITF in the hearts from control and sorafenib-treated mice (n=5). F Immunofluorescence staining of MITF and cTNT in the hearts of control and sorafenib-treated mice (n=5). G The mRNA level of Mitf in the NRCMs of the control and sorafenib groups is detected by qRT-PCR (n=5). H The protein expression of MITF in the NRCMs of control and sorafenib group mice is detected using western blot (n=5). I The protein expression of MITF in the cytoplasm (up) and nucleus (down) from sorafenib-treated NRCMs is performed using western blot (n=5). Sor, sorafenib; UMAP, uniform manifold approximation and projection; MITF, melanocyte inducing transcription factor; KEGG, Kyoto Encyclopedia of Genes and Genomes; CMs, cardiomyocytes; NRCMs, neonatal rat cardiomyocytes; qRT-PCR, quantitative real-time polymerase chain reaction. Data are presented as means ± SD

To verify the regulation of MITF in sorafenib-induced cardiotoxicity, we first examined MITF expression in the heart tissues of the sorafenib and control groups. Compared with the control group, the levels of MITF in the heart tissues of the sorafenib group were significantly increased at both the mRNA and protein levels (Figs. 6C, D). Immunohistochemical analysis and immunofluorescence staining confirmed that MITF expression in the myocardial tissue increased after sorafenib treatment, with MITF expression in both the cytoplasm and nucleus (Figs. 6E, F).

In vitro, sorafenib increased the expression of MITF in H9C2 cells and HL-1 cells in a dose-dependent manner (Additional file 1: Fig. S9A, B, C, D). Consistent with these findings, the mRNA and protein levels of MITF were also increased in NRCMs following sorafenib treatment (Figs. 6G, H). To delineate the subcellular distribution of MITF, we examined its expression in the cytoplasmic and nuclear fractions of NRCMs. The results indicated increased MITF expression in NRCMs, with enhanced nuclear localization after sorafenib treatment (Fig. 6I). Immunofluorescence staining also indicated that MITF was significantly expressed in the nucleus following sorafenib treatment of the H9C2 cell line (Additional file 1: Fig. S9E, F). These results demonstrated that sorafenib treatment increased MITF levels in both the cytoplasm and nucleus.

MITF knockdown alleviated sorafenib-induced cardiotoxicity in vitro

To further study the effects of MITF on CM hypertrophy, MITF was knocked down in H9C2 cells. The most significant siRNAs were selected for subsequent experiments (Additional file 1: Fig. S10A, B). MITF knockdown alleviated hypertrophy of H9C2 cells (Additional file 1: Fig. S10C). Similar results were obtained for HL-1 cells (Additional file 1: Fig. S10D, E, F, G). Moreover, the effect of MITF on H9C2 cells hypertrophy following sorafenib treatment was investigated. Compared with the corresponding negative control (siNC), the mRNA and protein expression of MITF were significantly decreased in siMITF-treated H9C2 cells (Additional file 1: Fig. S11A, B, C). MITF knockdown reduced the transcriptional levels of Anp and Bnp in H9C2 cells, with or without sorafenib treatment (Additional file 1: Fig. S11D, E, F). Phalloidin staining revealed that MITF knockdown inhibited CM hypertrophy after sorafenib treatment (Additional file 1: Fig. S11G).

Considering the limitations of the myocardial cell lines in hypertrophy research, we introduced NRCMs with terminal differentiation and non-proliferative characteristics to jointly verify the myocardial hypertrophy-promoting effect of MITF under physiological and pathological conditions. Under physiological conditions, after low-expression treatment of MITF on NRCMs, we observed that the transcriptional levels of Anp and Bnp in the siMITF treatment group were lower than those in the control group (Figs. 7A, B, C). Under pathological conditions, we observed that the transcription and translation levels of MITF in NRCMs treated with only sorafenib were significantly higher than those in the control group. After sorafenib treatment combined with siMITF treatment, the elevated transcriptional levels of Anp and Bnp induced by sorafenib returned to normal (Figs. 7D, E, F). In addition to detecting the expression levels of CM hypertrophy markers in NRCMs, we more intuitively demonstrated the effect of MITF on the morphological hypertrophy of NRCMs through immunofluorescence. As shown in Fig. 7G, we detected the cell surface area of NRCMs by α-actinin staining, and observed that sorafenib induced an increase in the cell area, while low expression of MITF rescued the CM hypertrophy induced by sorafenib treatment.

Fig. 7.

MITF knockdown alleviates sorafenib-induced cardiotoxicity and MITF overexpression aggravates sorafenib-induced cardiotoxicity in vitro. A-B The MITF protein levels in NRCMs is examined by western blot (n=5). C The mRNA expressions of Mitf, Anp and Bnp in NRCMs are detected by qRT-PCR (n=5). D-E Effect of MITF knockdown on the MITF protein level in NRCMs is examined by western blot (n=5). F Effect of MITF knockdown on the mRNA levels of Mitf, Anp and Bnp in NRCMs is examined by qRT-PCR (n=5). G Effect of MITF knockdown on the cell hypertrophy of NRCMs is examined by α-actinin staining (n=5). H-I The MITF protein levels in NRCMs is examined by western blot (n=5). J The mRNA expressions of Mitf, Anp and Bnp in NRCMs are detected by qRT-PCR (n=5). K-L Effect of MITF overexpression on the MITF protein level in NRCMs is examined by western blot (n=5). M Effect of MITF overexpression on the mRNA levels of Mitf, Anp and Bnp in NRCMs is examined qRT-PCR (n=5). N Effect of MITF overexpression on the cell hypertrophy of NRCMs is examined by α-actinin staining (n=5). Sor, sorafenib; MITF, melanocyte inducing transcription factor; NRCMs, neonatal rat cardiomyocytes; qRT-PCR, quantitative real-time polymerase chain reaction; ANP, atrial natriuretic peptide; BNP, brain natriuretic peptide. Data are presented as means ± SD

MITF overexpression aggravated sorafenib-induced cardiotoxicity in vitro

To further study the effect of MITF on CM hypertrophy, MITF was overexpressed in NRCMs and H9C2 cells respectively. Compared with the corresponding negative control (pc-NC), the mRNA and protein expression of MITF were significantly increased in MITF-overexpressing NRCMs (Figs. 7H, I, J). MITF overexpression increased the transcriptional and translational levels of Anp and Bnp in NRCMs, regardless of sorafenib treatment (Figs. 7K, L, M). While α-actinin staining also revealed that MITF overexpression aggravated NRCMs hypertrophy, with or without sorafenib treatment (Fig. 7N). The same results were also observed in the H9C2 cells (Additional file 1: Fig. S11H, I, J, K, L, M, N). These results indicate that MITF overexpression aggravates CM hypertrophy.

SEC22B was identified as a downstream effector positively regulated by MITF

To identify potential downstream targets of MITF, we assessed the strength of association between MITF and candidate target genes using bioinformatic analysis. Among the candidates, Sect. 22 homolog B (SEC22B) exhibited the highest association score with MITF (Additional file 1: Fig. S12A and Additional file 2: Table. S9), suggesting it as a potential direct target. Under sorafenib treatment conditions, SEC22B was significantly upregulated at both the protein and transcriptional levels (Additional file 1: Fig. S12B, C, D). To validate MITF-dependent regulation, we performed MITF knockdown experiments. As shown in Additional file 1: Fig. S12E, F, G, H, sorafenib treatment induced a significant increase in SEC22B expression, and this effect was rescued by MITF knockdown, confirming that SEC22B was positively regulated by MITF. To investigate whether SEC22B regulates CM hypertrophy, we knocked down SEC22B expression in CMs and examined the expression changes of the CM hypertrophy markers ANP and BNP. The results showed that, compared with the siNC-treated group, the siSEC22B-treated group significantly suppressed sorafenib-induced upregulation of Anp and Bnp transcript levels (Additional file 1: Fig. S12I, J, K, L). Therefore, MITF participates in sorafenib-induced CM hypertrophy through SEC22B.

Discussion

To the best of our knowledge, this is the first study to use snRNA-seq to characterize the transcriptional profile of cardiac tissue following sorafenib-induced cardiotoxicity. We identified dynamic changes in cellular composition across distinct cell populations and delineated the DEG profiles associated with sorafenib-induced cardiotoxicity in each cell type. The most prominent injury-related alterations were observed in CMs, with MITF being identified as a key regulatory TF in sorafenib-treated CMs. MITF expression was significantly increased in sorafenib-treated CMs, and MITF knockdown alleviated sorafenib-induced CM hypertrophy (Fig. 8). Our findings not only expand the current transcriptomic landscape of sorafenib-induced cardiotoxicity, but also provide an understanding of the specific molecular changes underlying this phenomenon.

Fig. 8.

Graphical abstract

Sorafenib is currently a first-line targeted anti-tumor drug. However, sorafenib-induced cardiotoxicity limits its clinical application [30, 54]. Although previous studies have identified multiple mechanisms involved in sorafenib-induced cardiotoxicity, including CM apoptosis [55], necrosis [56], mitochondrial dysfunction [57], and oxidative stress [58], no effective therapeutic targets have been established for clinical use. To elucidate the cellular composition changes and transcriptomic alterations across different cell types underlying sorafenib-induced cardiotoxicity, we employed snRNA-seq for the first time to characterize the transcriptional landscape of sorafenib-induced cardiotoxicity.

Through comprehensive bioinformatics analysis comparing the sorafenib and control groups, we identified dynamic changes in cellular composition and intercellular communication, and characterized DEGs associated with sorafenib-induced cardiotoxicity across distinct cell types. Among all the cell populations, CMs were the most significantly affected by sorafenib. In the present study, we found that the proportion of CMs and ECs in the sorafenib group was decreased, consistent with the known pharmacological effect of sorafenib as a tyrosine kinase inhibitor and the observed reduction in cardiac function and elevation of myocardial injury markers in sorafenib-treated mice. Furthermore, sorafenib-induced cardiotoxicity is closely related to the inflammatory microenvironment [59]. Notably, inconsistencies in the quantity and proportional distribution of CMs were observed in the snRNA-seq dataset, suggesting that sorafenib treatment may induce global cellular alterations affecting multiple cell lineages. This results in the preservation of cellular compositional consistency despite an overall reduction in cell numbers. Integrating the experimental findings of sorafenib-induced CM hypertrophy in this study, sorafenib emerged as a critical driver of myocardial remodeling, characterized by compensatory hypertrophy of residual CMs alongside numerical changes in diverse cell populations within the myocardial tissue. Our results also indicated that the proportion of immune cells increased after sorafenib treatment, along with the significant activation of immune cells upon functional enrichment analysis. Interestingly, the proportion of FBs remained unchanged after sorafenib treatment, suggesting that cardiac fibrosis cannot be explained solely by alterations in cellular composition. Therefore, we analyzed the functional enrichment of DEGs in FBs and the changes in cell-to-cell communication. The DEGs in FBs were enriched in fibrosis-related biological processes, including ECM organization, cell population proliferation, cellular responses to oxidative stress, HCM, and the Wnt signaling pathway. Moreover, the interaction between the CMs and FBs was enhanced by sorafenib treatment. Further analysis using AUCell revealed that the scores of fibrosis-related gene sets in both CMs and FBs in the sorafenib group were upregulated. Meanwhile, our subclustering and gene set enrichment analysis revealed that although the total number of FBs remained stable, sorafenib treatment significantly increased the activation status of multiple FB subpopulations, characterized by the upregulation of gene sets related to myofibroblast markers and ECM. Hence, although the proportion of FBs remained unchanged, more FBs were activated. FBs can transform into an activated state that produces ECM while maintaining their total number. The FB activation state that we observed further provides a potential cellular mechanism for the observed histological fibrosis. Additionally, SMCs and pericytes exhibited altered cell proportions and communication, suggesting their potential involvement in cardiac fibrosis. Nevertheless, the precise contributions of SMCs and pericytes to sorafenib-induced cardiac fibrosis remain unresolved in current preclinical studies, warranting systematic investigations in future research.

To identify the cell types most affected by sorafenib-induced cardiotoxicity, we employed a multimodal analytical approach comprising Augur analysis, quantitative comparison of DEGs across populations, functional pathway enrichment profiling, cross-referencing with established cardiotoxicity gene signatures, and WGCNA. The integrated results revealed that CMs were the most significant cell type for sorafenib-induced cardiotoxicity. Therefore, we focused on the pathological changes in CMs during sorafenib-induced cardiotoxicity. DEGs in CMs were associated with heart disease (DCM and heart failure), cardiotoxicity pathological processes (myocardial damage, myocardial hypertrophy, calcium signaling pathway, myocardial inflammation, and myocardial fibrosis), and cardiac muscle contraction.

Moreover, TF regulatory network analysis of CMs revealed that MITF was the most significantly altered TF in CMs during sorafenib-induced cardiotoxicity. The subclustering analysis of CMs revealed that MITF was consistently upregulated in the predominant CM subpopulation (Cluster 0), which showed enrichment in cardiomyopathy-related pathways (ARVC, DCM, HCM). This finding suggests that MITF dysregulation is not confined to a rare cell subset but represents a broad transcriptional shift in the majority of CMs, consistent with its proposed role as a key regulator of pro-hypertrophic signaling. MITF is expressed in various cells including melanocytes, mast cells, osteoclasts, and CMs [60–62]. In a preclinical model of DOX-mediated cardiotoxicity, transcription factor EB, a member of the MITF transcription factor family, mediated autophagy dysregulation, mitochondrial dysfunction, and cell death, leading to heart failure [63]. MITF mutations can inhibit isoproterenol-induced increases in BNP levels and cardiac hypertrophy induced by isoproterenol [46, 64]. MITF induces cardiac hypertrophy by negatively regulating the expression of miR541 in CMs [65]. However, the role of MITF in sorafenib-induced cardiotoxicity remains unclear. In this study, we respectively verified the expression changes and roles of MITF during the pathological process of sorafenib-induced myocardial toxicity in NRCMs and cardiac cell lines in order to enhance the consistency and reliability of our findings. Our study indicated that MITF expression was significantly increased in the heart and CMs after sorafenib treatment. MITF overexpression aggravated sorafenib-induced CM hypertrophy, with MITF knockdown alleviating sorafenib-induced CM hypertrophy. Hence, our integrated analyses revealed that MITF is associated with CM hypertrophy following sorafenib treatment.

To further explore the regulatory mechanism of MITF in sorafenib-induced cardiotoxicity, we screened out the candidate target gene SEC22B with high correlation to MITF through bioinformatics. As one of the members of soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE), SEC22B is involved in biological processes such as vesicle transport membrane fusion between the endoplasmic reticulum and the Golgi apparatus, antigen cross-presentation, and secretory autophagy [66]. However, there are no related reports in cardiovascular diseases. In this study, we observed that MITF could rescue the sorafenib-induced increase in SEC22B expression. Furthermore, knockdown of SEC22B significantly suppressed sorafenib-induced upregulation of ANP and BNP transcript levels. This finding demonstrates that MITF indirectly regulates CM hypertrophy via SEC22B. Of note, previous studies have established that MITF directly regulates the promoter activity of the CM hypertrophy marker BNP [46]. Collectively, we propose that MITF contributes to sorafenib-induced CM hypertrophy through both direct and indirect regulatory mechanisms. However, under sorafenib treatment conditions, as a target molecule positively regulated by MITF, SEC22B still needs to be further clarified on the pathways through which it promotes sorafenib-mediated cardiomyocyte hypertrophy and the specific regulatory mechanism of MITF on SEC22B.

Notably, MITF is well characterized as an oncogene that drives the progression of malignant tumors, such as melanoma. MITF has emerged as a promising therapeutic target owing to its crucial role in melanocyte proliferation and survival. However, MITF expression fluctuates throughout melanoma progression and drug resistance. MITF plasticity influences the heterogeneity and malignancy of melanomas. Maintaining high MITF expression levels is considered a potential strategy for inhibiting drug resistance. Drug screening revealed that the pan-receptor tyrosine kinase inhibitor tivozanib enhances the transcriptional activity of MITF in melanocytes, thereby enhancing the dependence of melanocytes on MITF and their sensitivity to targeted inhibitors [50]. Targeting MITF may be a promising strategy to simultaneously address drug resistance in melanoma and manage sorafenib-induced cardiotoxicity, though further preclinical validation is required. To the best of our knowledge, our study is the first to provide evidence that MITF is a key regulator of sorafenib-induced cardiotoxicity and reveal its pathogenic role in this process, which parallels its established function in tumor progression.

This study has several limitations. First, although the transcriptional and protein levels of MITF in CMs were significantly elevated after sorafenib treatment, the precise mechanism underlying MITF upregulation remains unclear. Second, although we have preliminarily explored the potential downstream molecule of MITF, the mechanism by which MITF regulates CM hypertrophy induced by sorafenib remains to be fully clarified. Third, the functional role of MITF in sorafenib-induced cardiotoxicity requires further validation in animal experiments. Fourth, despite sorafenib being an anticancer drug, our in vivo model did not include a tumor. This limitation raises questions regarding how closely our model reflects clinical sorafenib-induced cardiotoxicity and whether the dosing regimen used aligns with therapeutically relevant conditions.

Conclusions

We identified MITF as a novel regulatory factor underlying sorafenib-induced cardiotoxicity by constructing the first snRNA-seq dataset associated with sorafenib-induced cardiotoxicity. MITF was found to have the highest correlation with sorafenib-induced cardiotoxicity among genes and cell types, respectively. MITF expression was significantly increased in sorafenib-treated CMs and cardiac tissue. MITF is regarded as a transcription factor associated with CM hypertrophy due to sorafenib. Our findings not only expand the current transcriptomic landscape of sorafenib-induced cardiotoxicity, but also provide important insights into the underlying mechanisms and potential therapeutic pathways for the alleviation of sorafenib-induced cardiotoxicity.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ANP

atrial natriuretic peptide

AR

androgen receptor

ARE

anti-oxidant response element

ARVC

arrhythmogenic right ventricular cardiomyopathy

Acta2

actin alpha 2, smooth muscle

Bclaf1

BCL2 associated transcription factor 1

BNP

brain natriuretic peptide

BW

body weight

CM

cardiomyocyte

CK-MB

creatine kinase-MB

Cthrc1

collagen triple helix repeat containing 1

cTNT

troponin T

Col1a1

collagen type I alpha 1

Col3a1

collagen type III alpha 1

Col5a1

collagen type V alpha 1

Col6a1

collagen type VI alpha 1

DAPI

4’,6-diamidino-2-phenylindole

DEG

differentially expressed gene

DCM

dilated cardiomyopathy

DMEM

Dulbecco’s modified Eagle’s medium

EC

endothelial cell

ECM

extracellular matrix

FDR

false discovery rate

Fth1

ferritin heavy chain 1

FBS

fetal bovine serum

FB

fibroblast

Fn1

fibronectin 1

FOXP1

forkhead box P1

GATA4

GATA Binding Protein 4

GEO

Gene Expression Omnibus

GO

gene ontology

GLS%

global ventricular longitudinal strain

H&E staining

hematoxylin and eosin staining

HO-1

heme oxygenase 1

HCM

hypertrophic cardiomyopathy

HW

heart weight

Klf9

KLF transcription factor 9

KEGG

Kyoto Encyclopedia of Genes and Genomes

LVEF%

left ventricular ejection fraction

LVFS%

left ventricular fractional shortening

MACE

major adverse cardiac and cerebrovascular events

MITF

melanocyte inducing transcription factor

JNK1

mitogen-activated protein kinase 8

NRCMs

primary neonatal rat cardiomyocytes

Nrf2

NFE2 like BZIP transcription factor 2

Nfia

nuclear factor I A

Nfib

nuclear factor I B

Nr3c1

nuclear receptor subfamily 3 group C member 1

Pbx1

PBX homeobox 1

Per3

period circadian regulator 3

PBS

phosphate buffered saline PBS

Pecam1

platelet-and endothelial cell adhesion molecule 1

PDGFR

platelet-derived growth factor receptor

PCA

principal component analysis

Postn

periostin

qRT-PCR

quantitative real-time polymerase chain reaction

ROS

reactive oxygen species

SCENIC

single-cell regulatory network inference and clustering

SD

standard deviation

SEC22B

Section 22 homolog B

snRNA-seq

single-nucleus RNA-sequencing

Smad3

SMAD family member 3

SMC

smooth muscle cell

Slc25a3

solute carrier family 25 member 3

SOD2

superoxide dismutase 2

Sor

sorafenib

SRA

sequence read archive

Tcf12

transcription factor 12

Tef

thyrotroph embryonic factor

TFs

transcription factors

Tgf-β1

transforming growth factor beta 1

TL

tibia length

UMAP

uniform manifold approximation and projection

VEGFR

vascular endothelial growth factor receptor

WGCNA

weighted gene co-expression network analysis

WGA staining

wheat germ agglutinin staining

Wwox

WW domain containing oxidoreductase

Zbtb20

zinc finger and BTB domain containing 20

Author contributions

YNW: Conceptualization, Data curation, Validation, Methodology, Writing-original draft. ZMY: Validation, Methodology, Formal analysis. JW: Software, Methodology, Data curation. ZQL and YZ: Methodology, Validation, Data curation. XLZ: Formal analysis, Validation. HXS: Data curation, Validation. CHY: Conceptualization, Software, Methodology. DL and YLH: Project administration, Conceptualization, Writing-review & editing, Supervision, Project administration. All authors read and approved the final manuscript. All authors agree to be fully accountable for ensuring the integrity and accuracy of the work.

Funding

This work was supported by the National Natural Science Foundation of China (82570354 and 82470303) and the Natural Science Foundation of Liaoning Province (2026‑BS‑0900).

Data availability

The raw snRNA-seq data supporting the findings of this study have been deposited in the NCBI Sequence Read Archive (SRA) under BioProject accession number PRJNA1482921 and are publicly available at: https://www.ncbi.nlm.nih.gov/sra/PRJNA1482921. The processed gene expression data have been deposited in the NCBI Gene Expression Omnibus (GEO) under accession number GSE336870 and are accessible at: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi? acc=GSE336870.

Declarations

Ethics approval and consent to participate

All the animal care and experimental procedures conducted in this study were reviewed and approved by the General Hospital of the Northern Theater Command (Approval no.2022-19). The protocols strictly adhered to the guidelines provided by the National Institutes of Health for the ethical use of laboratory animals or equivalent standards.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.