Six-week-old male C57BL/6 mice were individually housed (up to four per cage) in ventilated racks under controlled conditions: The mice were maintained at 21°C ± 1°C with a 12-h light/dark cycle and 30%–50% humidity, and were randomly assigned to two groups (n = 3). Sorafenib (HY-10201, MedChemExpress LLC, New Jersey, United States) was administered intraperitoneally to the treatment group at 30 mg/kg per day, while the control group received a vehicle solution of 10% DMSO, 40% PEG300, 5% Tween-80, and 45% saline. Myocardial function was assessed 2 weeks after the final injection, followed by humane euthanasia under isoflurane anesthesia for heart collection. All procedures complied with the ARRIVE guidelines.

Two weeks post daily injections, echocardiographic evaluations were conducted on lightly anesthetized animals (1%–2% isoflurane) using a GE Vivid E9 ML6-15-D Linear Probe (United States). M-mode imaging assessed left ventricular fractional shortening (FS), ejection fraction (EF), and the thickness of the left ventricular posterior wall during diastole (LVPWd) and systole (LVPWs).

Heart tissues were rinsed with phosphate-buffered saline (PBS) and fixed in 4% paraformaldehyde (PFA) for 24 h. The tissues were then embedded in paraffin, sectioned into 2 μm slices, and stained with hematoxylin and eosin (HE) following standard procedures (Solarbio, G1121, China). Images were captured using a Leica^® DMi8 light microscope from Germany.

The H9C2 rat cardiomyocyte cell line (CRL-1446) was obtained from the American Type Culture Collection (ATCC, United States) and cultured in a CO₂ incubator with 5% CO₂ using Dulbecco’s Modified Eagle’s Medium (DMEM, GIBCO, United States) supplemented with 10% fetal bovine serum (FBS, GIBCO, United States). Cells were routinely passaged at approximately 80% confluence. Prior to experiments, the medium was replaced with DMEM containing 0.5% FBS for 12 h. Experimental treatments involved 5 μM sorafenib in dimethyl sulfoxide (DMSO) for 48 h, with 0.1 nM DMSO as the control. Sorafenib was also combined with 1 μM GSK2606414 (MedChemExpress, United States). All experiments were performed independently three times, each in triplicate.

Cell viability was assessed using the calcein AM/PI Double Staining Kit (E-CK-A354, Elabscience). Following PBS rinsing, cells were incubated with calcein AM and propidium iodide for 20 min. Fluorescence microscopy revealed green fluorescence from calcein AM (Ex/Em = 494nm/517 nm) and red fluorescence from PI (Ex/Em = 535nm/617 nm).

Cell viability was evaluated using the Cell Counting Kit-8 (CCK-8, BS350A, Biosharp, China). A 100 μL sample from each culture medium was added to a 96-well plate and incubated at 37°C. After 48 h of treatment, 10 µL of CCK-8 reagent was added to each well, and the absorbance at 450 nm was measured with a Tecan Infinite 200 PRO Microplate Reader (Tecan Group Ltd., Switzerland).

Cell lysates were obtained from cell lines using RIPA buffer with protease inhibitors (Beyotime, P0013B, China), and protein concentrations were determined via the BCA assay (Beyotime, P0012, China). Protein samples (10–20 µg) underwent separation by SDS-PAGE. Initially, electrophoresis was conducted at 70V for 30 min, then at 110V for 90 min (Sevenbio, Beijing, China). Proteins were subsequently transferred to a PVDF membrane at 300 mA for 70 min (Millipore, China). Membranes were blocked with 5% nonfat milk for 1 h at room temperature, followed by an overnight incubation with primary antibodies at 4°C. The study employed the following antibodies. Cleaved caspase-9 p35 (D315) at 1:1000 dilution (Immunoway, YC0014), P-PERK at 1:1000 (Wanleibio, WL05295), PERK at 1:1000 (Wanleibio, WL03378), EIF2S1 (P-eIF2α) at 1:2000 (Abcam, ab32157), eIF2α at 1:500 (Wanleibio, WL01909), GRP78/BiP at 1:1000 (Wanleibio, WL03157), ATF-4 at 1:500 (Wanleibio, WL02330), GAPDH at 1:10000 (Abcam, ab181602), ATF-3 at 1:1000 (Immunoway, YT0387), cleaved caspase-3 at 1:500 (Wanleibio, WL01992), and NDUFS1 at 1:The membrane was incubated with an HRP-conjugated secondary antibody following treatment with the primary antibody (Immunoway, YT3017). Protein bands were detected and quantified using an ECL system (Tanon 5100, China) and analyzed with ImageJ software (NIH v1.8.0).

H9C2 cells were cultured on sterile coverslips in 24-well plates. After treatment, cells underwent three PBS washes and were fixed with 4% paraformaldehyde for 20 min. Permeabilization was achieved using 0.2% Triton X-100 for 15 min, followed by blocking with 10% goat serum in PBS for 1 h. Cells were then incubated overnight at 4°C with an anti-P-eIF2α antibody. After PBS washing, cells were incubated with a CY3-conjugated secondary antibody for 1 h at room temperature in the dark. Nuclei were stained with DAPI for 10 min at room temperature, protected from light. Fluorescent images were captured using an Olympus microscope (Japan).

The apoptotic ratio of cardiomyocytes was evaluated using flow cytometry with an Annexin V-FITC apoptosis detection kit (Seven Biotech, SC123, China). Cells were collected, washed twice with cold PBS, resuspended in binding buffer, and incubated with 5 µL Annexin V and 10 µL PI for 15 min at room temperature in the dark. Data analysis was performed using FACSDiva (v6.1.3, BD Biosciences) and FlowJo V10 (TreeStar).

Cells were cultured in a 6-well plate at 37°C. After 48 h of treatment, the medium was discarded, and the cells were washed twice with cold PBS. They were then incubated with 5 µL of Annexin V and 10 µL of PI at room temperature in the dark for 15 min.

Total RNA was extracted from cell lines using TRIzol reagent (Invitrogen, United States). cDNA synthesis was conducted with the Transcriptor First Strand cDNA Synthesis Kit (Roche, Germany) according to the manufacturer’s protocol. Quantitative PCR (qPCR) utilized SYBR Green I Master Mix (Roche, Germany) on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad, United States). Ct values were automatically calculated and normalized to GAPDH, serving as the internal control. Gene expression levels were determined using the 2^−ΔΔCT method. Primer sequences are listed in Supplementary Table S1.

Transmission electron microscopy (TEM) was employed to examine the ultrastructural morphology of mitochondria and ER in cell lines. Following treatment, cells were collected, washed three times with precooled sterile 1×PBS, and fixed in 2.5% glutaraldehyde in 0.1M phosphate buffer (pH 7.4) at 4°C for 24 h. TEM was then used to visualize and capture the mitochondrial and ER structures.

Post-treatment, cells were incubated with MitoTracker Green and ER-Tracker Red (Beyotime Institute of Biotechnology) at 37°C for 30 min, then analyzed via a Carl Zeiss confocal microscope.

Chromatin immunoprecipitation (ChIP) assays were conducted using a ChIP kit (#Bes5001, BersinBio, China) according to the manufacturer’s instructions. H9C2 cells underwent pretreatment, trypsinization, collection, and cross-linking with 1% formaldehyde to capture protein-DNA interactions. Glycine was used to terminate cross-linking, followed by a 10-min ice incubation and chromatin fragmentation via sonication. The fragmented chromatin was immunoprecipitated using an ATF3 antibody (ab207434, Abcam) to isolate protein-DNA complexes. After multiple washes, the immunoprecipitated DNA was eluted and subjected to qPCR analysis. The primers used for the ChIP assay are listed in Supplementary Table S2.

ATF3-targeting siRNAs were designed and synthesized by Gene Pharma (China), and transfected using Lipofectamine 3,000 (Invitrogen, United States) according to the manufacturer’s instructions. The siRNA sequences are listed in Supplementary Table S3.

The Mitochondrial Complex I Activity Assay Kit (Solarbio) was utilized to evaluate mitochondrial complex I activity as per the manufacturer’s guidelines. Enzymatic activity was measured by isolating complex I from 5 × 10⁶ cells and observing NADH oxidation at 340 nm over 2 min.

Cells in the exponential growth phase were collected by centrifugation at 14,000 × g for 2 min at 4°C and rinsed with ice-cold PBS. Lysis was performed using the Enhanced ATP Assay Kit (S0027, Beyotime Biotechnology, Shanghai, China). ATP levels were quantified according to the manufacturer’s protocol, with concentrations calculated from a standard curve and reported as nmol/OD730.

The JC-1 Kit (Beyotime, C2006, China) was employed to evaluate mitochondrial membrane potential (ΔΨm) following the manufacturer’s guidelines. H9C2 cells were exposed to the JC-1 probe for 20 min at 37°C, with CCCP (10 μM) serving as a positive control. Mitochondrial depolarization was assessed by calculating the ratio of JC-1 aggregates to monomers. Imaging was conducted using a cell auto-imaging system (DMi8, Leica^®, Germany), capturing five images per well.

Mitochondrial reactive oxygen species (ROS) production was evaluated using MitoSOX Red (MCE, 1003197–00-9, China) following the manufacturer’s instructions. Cells were cultured in 6-well plates at 37°C under static conditions. After 48 h of treatment, the medium was removed, and 100 μL of 5 μM MitoSOX solution was added to each well. The plate was gently agitated for uniform distribution and incubated at room temperature for 30 min. Cells were then washed twice with PBS for 5 min each and subjected to fluorescence imaging using a microscope.

Intracellular ROS levels were assessed with a commercial detection kit (Beyotime, S0033S, China). Post-treatment, cells were incubated with 10 μmol/L DCFH-DA at 37°C for 20 min and subsequently washed thoroughly. Fluorescence signals were imaged under a fluorescence microscope in a blinded manner.

Intracellular superoxide levels were measured using dihydroethidium (DHE, Beyotime Institute of Biotechnology). After 48 h of treatment, H9C2 cells were exposed to 5 μM DHE for 30 min at 37°C in a dark, humidified environment. The cells were then rinsed with PBS, and fluorescence was observed under a microscope by an investigator unaware of the group assignments.

The GSE146096 dataset was sourced from the Gene Expression Omnibus (GEO) database, as detailed in Supplementary Table S4. The dataset, accessible at//www.ncbi.nlm.nih.gov/geo/, comprises transcriptomic data from human cardiac cells treated with various FDA-approved kinase inhibitors. Generated via RNA sequencing on the Illumina HiSeq 2,500 platform, the dataset was last updated on 13 October 2020. Human heart tissue-derived primary cardiomyocytes were divided into four experimental groups (A, B, D, and E), each exposed to sorafenib for 48 h at different concentrations. 2.5 μM in one subgroup of Line A, 0.5 μM in the remaining A subgroups, and 1 μM for all subgroups in Lines B, D, and E. Group A included four sorafenib-treated samples and 13 controls; Group B had four treated and eight controls; Group D comprised three treated and seven controls; and Group E contained three treated and 11 controls (Supplementary Table S5). Data on transcription factor binding sites (TFBS) for NDUFS1 was sourced from the GeneCards database (https://www.genecards.org/).

The mRNA expression profiles were processed using R software (v4.2.1). Differential expression analysis was conducted with the ‘DESeq2’ package, calculating log2 fold change (logFC) values and adjusted P-values. The Benjamini–Hochberg method was applied to control false discovery rates. Genes with adjusted P-values below 0.05 and absolute logFC greater than 1.0 were considered significantly differentially expressed. Functional enrichment analysis, including gene ontology (GO), kyoto encyclopedia of genes and genomes (KEGG) and gene set enrichment analysis (GSEA), was performed using the ‘clusterProfiler’ package. Data visualization was performed using ‘ggplot2’ for differentially expressed gene distribution plots, principal component analysis (PCA), Venn diagrams, GSEA enrichment maps, GO term charts, and KEGG pathway maps, while ‘ComplexHeatmap’ was used for heatmap generation. Potential binding sites for ATF3 and NDUFS1 were identified via the JASPAR database.

Results are expressed as mean ± standard error of the mean (SEM) from a minimum of three independent replicates. Statistical analyses and figure creation were performed using GraphPad Prism (version 8.0.2; GraphPad Software, La Jolla, CA, United States). An unpaired Student’s t-test was used to assess differences between two groups, with F-tests confirming variance homogeneity. A p-value of less than 0.05 was considered statistically significant.

To establish a mouse model of sorafenib-induced cardiac injury, C57BL/6 mice were administered intraperitoneal injections of sorafenib following a dosing regimen that mimics clinical application (Figure 1A) (Wilhelm et al., 2004). The results demonstrated that sorafenib treatment impaired cardiac function, as shown by reduced left ventricular systolic performance (Figures 1B–F). Furthermore, hematoxylin-eosin (HE) staining provided histological evidence of myocardial injury. As shown in Figure 1G, sorafenib exposure led to myocardial fiber disarray and cardiomyocyte atrophy, along with cytoplasmic condensation, cell rounding, nuclear pyknosis, fragmentation, and dissolution.

Our study evaluated sorafenib’s effects on rat cardiomyocyte viability using calcein AM/PI dual staining and CCK-8 assays, revealing reduced cell viability compared to controls (Figures 2A–C). Western blot analysis showed increased cleaved caspase-9 expression in sorafenib-treated cells, indicating apoptosis contributes to sorafenib-induced cardiomyocyte toxicity (Figures 2D,E). Apoptosis in rat cardiac H9C2 cells was also assessed by in situ fluorescence after treatment with Annexin V and PI. Sorafenib treatment significantly increased apoptosis compared to controls (Figure 2F).

Apoptosis is crucial in sorafenib-induced cardiotoxicity. This is supported by decreased cell viability, upregulation of cleaved caspase-9, and in situ fluorescence detection. Typical apoptotic morphology, such as nuclear shrinkage and fragmentation, was also observed in cardiomyocytes. These findings were consistently observed in both in vivo mouse models and in vitro rat cardiomyocyte experiments.

The PERK-eIF2α-ATF4 pathway plays a crucial role in cellular responses to endoplasmic reticulum stress (ERS) and is strongly linked to cardiotoxicity. This study developed a cardiomyocyte model to examine sorafenib-induced ER stress and cardiotoxicity, using sorafenib alone or with the PERK inhibitor GSK2606414 to evaluate the pathway’s involvement.

Western blotting and immunofluorescence staining were employed on cultured cardiomyocytes to investigate the PERK-eIF2α-ATF4 signaling pathway’s role in sorafenib-induced myocardial cell injury, as shown in Figures 3A,B. Sorafenib significantly increased PERK and eIF2α phosphorylation, evidenced by elevated p-PERK/PERK and p-eIF2α/eIF2α ratios, and upregulated ER stress markers GRP78 and ATF4. Pretreatment with the PERK inhibitor GSK2606414 reduced the pathway’s activation by sorafenib (Figures 3A,B), indicating its involvement in myocardial damage.

Furthermore, GSK2606414-mediated PERK inhibition mitigated sorafenib-induced cellular injury. To investigate endoplasmic reticulum stress (ERS) in sorafenib-induced cardiac injury, we performed the CCK-8 assay and calcein AM/PI double staining. Reduced cell viability (expressed as%) indicated a cardiomyocyte toxicity in the sorafenib-treated group versus the control group. Remarkably, the cardiotoxicity was attenuated by GSK2606414 treatment (Figures 3C,D). These findings suggest that inhibiting ERS mitigates sorafenib-induced cardiotoxicity in cardiomyocytes, highlighting the importance of the PERK-eIF2α-ATF4 signaling axis in myocardial cell injury.

Apoptosis in rat cardiac H9C2 cells was assessed using flow cytometry and in situ fluorescence after treatment with Annexin V and PI. Sorafenib treatment significantly increased apoptosis compared to controls, an effect reduced by GSK26064 14 co-treatment (Figures 4A,B). Western blot analysis showed similar trends, with elevated levels of cleaved caspase-9 and cleaved caspase-3 in the sorafenib group, which were decreased by GSK2606414 (Figure 4C). The findings suggest that sorafenib induces cardiomyocyte apoptosis through ERS.

TEM analysis revealed mitochondrial damage in cardiomyocytes of the sorafenib-treated group compared to the control group, with significant variations in swelling and increased interaction with the ER. Notably, co-treatment with GSK2606414 mitigated the sorafenib-induced disruption (Supplementary Figure S1A).

Confocal imaging revealed that sorafenib treatment led to a marked increase in mitochondria–ER contact and co-localization, with obvious heterogeneity observed across the treated cells. This effect was notably suppressed by GSK2606414 (Supplementary Figure S1B). These findings suggest that sorafenib induces alterations in both the mitochondria and the ER. Notably, the application of GSK2606414 effectively delays these changes.

Supplementary Figure S2 illustrates the analytical workflow. We examined mRNA expression data from the GSE146096 dataset, which includes four sorafenib-treated adult human cardiomyocyte samples, to investigate the transcriptomic response to sorafenib-induced cardiotoxicity. Data from the GEO database (Supplementary Table S4) was analyzed using PCA, revealing a distinct separation between the sorafenib-treated and control groups (Figures 5A–D), indicating a significant impact of the drug on the transcriptomic profile. DEGs were identified with thresholds of P < 0.05 and |logFC| > 1.0, and visualized through Volcano Plots and Heatmaps (Figures 5E–L). Further functional enrichment using GSEA, GO, and KEGG demonstrated significant clustering of DEGs. GSEA was performed on DEGs (Figures 6A–D), with detailed results in Supplementary Table S6, showing associations with pathways like NABA Matrisome, Oxidative Phosphorylation, and others. GO analysis identified key terms across biological process (BP), cellular component (CC), and molecular function (MF), including apoptotic signaling regulation, DNA-binding transcription factor interaction, and response to oxygen levels (Figures 6E–H; Supplementary Figure S3). KEGG analysis linked DEGs to pathways including apoptosis, TGF-β signaling, and protein processing in the ER (Figures 6I–L).

Among the significantly enriched pathways identified by GSEA, the TGF-β signaling pathway and Matrisome components were of particular interest due to their known relevance to cardiac pathology. TGF-β signaling plays a central role in cardiac fibrosis, hypertrophic remodeling, and pro-apoptotic responses under chronic stress conditions. Likewise, the Matrisome gene set, which includes structural extracellular matrix (ECM) proteins and remodeling enzymes, reflects early ECM disorganization—a hallmark of maladaptive cardiac remodeling. These changes may represent compensatory or maladaptive responses to mitochondrial dysfunction and ER stress induced by sorafenib.

A total of 77 upregulated and three downregulated cardiotoxicity-related DEGs were identified across line A, B, D, and E, as illustrated in the Venn diagrams (Supplementary Figure S4). These genes were then cross-referenced with 1,142 ERS-related genes obtained from the National Center for Biotechnology Information (www.ncbi.nlm.nih.gov.ncbi.nlm.nih.) (Supplementary Table S7). Fourteen differentially expressed ERS-related genes were identified, including ATF3, Growth Differentiation Factor 15 (GDF15), and protein phosphatase 1 (PP1), among others, as detailed in Supplementary Table S8. Among the differentially expressed ERS-related genes, the most significant differential expression was ATF3. Subsequently, to verify that sorafenib induces myocardial cell injury by activating ATF3 through ERS, Western blot and PCR were performed on cultured cells, and ATF3 was significantly upregulated (P < 0.05) (Figures 6M,N).

To understand the pathways involved in sorafenib-induced cardiotoxicity, we focused on ATF3’s role in cell survival and examined downstream proteins. Proteomic analysis of two sorafenib-treated cell lines identified four common differentially expressed proteins (DEPs), with NDUFS1 being the most significantly altered (Figures 7A,B). A Spearman correlation analysis revealed a negative relationship between ATF3 and NDUFS1 (Figure 7C). The regulatory sensitivity of ATF3 was evaluated using a ROC curve, yielding an AUC of 0.996, indicating excellent diagnostic performance (Figure 7D). Analysis of the NDUFS1 promoter region via the JASPAR database identified a strong ATF3 binding motif (Figures 7E,F).

Subsequent fundamental experiments validated the findings of the preceding proteomic analysis. ATF3 expression was notably lower in sorafenib-treated cells transfected with si-ATF3 compared to those with si-NC (negative control) (Figures 8A,B). Additionally, we demonstrated that sorafenib failed to downregulate NDUFS1 levels and enhance apoptosis-associated protein levels in ATF3-silenced cardiomyocytes (Figures 8A,B). In situ fluorescence staining revealed that ATF3 silencing did not significantly increase apoptosis in H9C2 cardiomyocytes treated with sorafenib (Figure 8C). Furthermore, based on the JASPAR database prediction, a chromatin immunoprecipitation-qPCR (ChIP-qPCR) analysis was performed, revealing enhanced ATF3 binding to the NDUFS1 promoter region following sorafenib treatment (Figure 8D; Supplementary Figure S5).

Cardiomyocyte function and viability are closely tied to mitochondrial energy production, with NDUFS1, a crucial subunit of mitochondrial complex I, playing a vital role in ATP synthesis and ROS generation. It is pivotal in regulating metabolic reprogramming, oxidative stress, and apoptosis in various diseases. Considering that ATF3-mediated downregulation of NDUFS1 aggravated sorafenib-induced cardiotoxicity, and TEM analysis revealed mitochondrial damage along with increased mitochondria-associated endoplasmic reticulum in sorafenib-treated H9C2 cells compared to controls (Figure 3A), we propose that sorafenib-induced cardiotoxicity may be linked to alterations in mitochondrial morphology and function. Mitochondrial complex I activity and ATP levels were significantly higher in the ATF3 knockdown group than in the sorafenib-treated group (Figures 8A,B). JC-1 staining indicated sorafenib-induced mitochondrial dysfunction in H9C2 cardiomyocytes, which was prevented by ATF3 knockdown (Figures 8C,D). Mitochondrial complex I is a major source of ROS. Compared to the sorafenib-treated group, ATF3 knockdown resulted in a significant reduction in mitochondrial ROS levels, as shown by MitoSOX staining (Supplementary Figure S6A,B). We assessed the impact of NDUFS1 depletion on oxidative stress by measuring ROS levels, finding a significant increase in ROS in sorafenib-treated cells, which was reduced in ATF3 knockdown cells, suggesting that ATF3 knockdown alleviates oxidative stress (Supplementary Figure S6C,D). Superoxide anion levels showed similar trends (Supplementary Figure S6E,F).

Sorafenib has shown efficacy in treating various cancers, including HCC and RCC, but its use is limited by significant cardiotoxic side effects (Uraizee et al., 2011; Wu and Shemisa, 2017; Wang et al., 2023). Previous studies have demonstrated that sorafenib induces cardiovascular toxicity, primarily through mechanisms such as oxidative stress, myocardial apoptosis, endoplasmic reticulum stress and mitochondrial dysfunction (Ma et al., 2020; Bouitbir et al., 2022; Chen et al., 2023; Wang et al., 2023). However, there has been limited exploration of the specific molecular pathways involved, particularly those associated with endoplasmic reticulum stress and mitochondrial function regulation. Our findings address this gap by elucidating the involvement of ER stress signaling, specifically the ATF3-mediated pathway, in sorafenib-induced cardiotoxicity.

Although this study did not utilize a tumor-bearing model, previous reports have shown that sorafenib can induce cardiac dysfunction in both cancer patients and preclinical models independent of tumor burden (Jue et al., 2024). Investigating its cardiotoxic mechanisms in non-tumor models allows clearer assessment of drug-specific effects on cardiomyocytes, avoiding confounding factors associated with cancer biology. Nonetheless, future studies incorporating tumor models would help validate whether these mechanisms are preserved in more complex oncologic settings.

The in vitro concentration of 5 µM sorafenib used in this study was selected based on our group’s previous findings showing potent inhibitory effects on H9C2 cardiomyocytes (Li Y. et al., 2022; Yilan et al., 2024). While this dose may appear higher than physiological free drug levels, it remains below the reported total plasma C_max of 20–35 µM observed in patients receiving standard oral dosing (Dirk et al., 2004; Wilhelm et al., 2004). Given sorafenib’s high protein binding (>99.5%), the actual unbound fraction is significantly lower. Furthermore, sorafenib is known to accumulate in metabolically active tissues such as the heart, making localized exposure at or above 5 µM plausible. Therefore, this concentration is appropriate for modeling early cardiotoxic mechanisms under clinically relevant exposure conditions.

ER stress, resulting from misfolded proteins, is associated with various diseases and can lead to apoptotic cell death if chronic or excessive (Almanza et al., 2019). It plays a significant role in heart failure and ischemic injury in cardiac tissues (Ren et al., 2021; Li M. et al., 2022; Preetha Rani et al., 2022). Our study builds on this by focusing on the activation crucial stress response proteins, in sorafenib-induced cardiotoxicity and its involvement in subsequent pathways. Specifically, our findings indicate that sorafenib activates the PERK-eIF2α-ATF4 signaling pathway in cardiomyocytes, leading to enhanced ER stress signaling and upregulation of stress-related markers. Pharmacological inhibition of PERK significantly attenuated this activation, suggesting that PERK functions as a key upstream mediator of sorafenib-induced ER stress. Furthermore, blocking this pathway alleviated cardiomyocyte injury, supporting the causal role of ER stress in sorafenib-induced cytotoxicity. These results not only reinforce the importance of ER stress in drug-induced cardiac damage but also identify the PERK-eIF2α-ATF4 axis as a potential therapeutic target for cardioprotection in patients receiving sorafenib.

Our omics data further supported this mechanistic link by revealing enrichment of TGF-β signaling and Matrisome-related pathways, both of which are intricately involved in cardiac stress remodeling and fibrosis. The activation of TGF-β signaling may represent a downstream consequence of ROS accumulation and apoptotic signaling triggered by ATF3-mediated mitochondrial dysfunction. Likewise, altered Matrisome gene expression may reflect early disruption of ECM homeostasis, which could sensitize cardiomyocytes to mechanical stress and cell death. These pathways may interact with the ATF3-NDUFS1 axis to amplify sorafenib-induced cardiac injury. Further studies are needed to dissect the temporal and causal relationship between these transcriptional signatures and the core mitochondrial damage axis.

Our study reveals a previously uncharacterized molecular axis by which sorafenib induces cardiotoxicity through ER stress and mitochondrial dysfunction. Sorafenib triggers the PERK-eIF2α-ATF4 signaling cascade, significantly increasing ATF3 expression, a stress-responsive transcription factor involved in apoptosis, inflammation, and metabolic regulation (Ge et al., 2023; Zhang et al., 2024; Gao et al., 2025). We demonstrate that ATF3 directly binds to the promoter of NDUFS1, a key component of mitochondrial complex I critical for electron transport and ATP synthesis (Bingchao et al., 2022), thereby suppressing its transcription.

The downregulation of NDUFS1 disrupts mitochondrial oxidative phosphorylation, reduces ATP synthesis, elevates ROS, and ultimately triggers cardiomyocyte apoptosis (Lothar and Judy, 2006; Irene et al., 2016; Zhuang et al., 2020; Kim et al., 2022). Notably, ATF3 knockdown restored NDUFS1 expression, improved mitochondrial membrane potential, reduced oxidative stress, and preserved mitochondrial bioenergetics. These results not only establish ATF3 as a critical modulator of mitochondrial function in drug-induced cardiac injury but also provide the first direct evidence linking the ATF3-NDUFS1 axis to sorafenib cardiotoxicity.

Previous studies have implicated ATF3 in stress-induced apoptosis across multiple tissues (Li et al., 2021; Yoo et al., 2021; Zhang et al., 2021; Gao et al., 2023), but its role in regulating mitochondrial function under pharmacologic stress was poorly understood. By integrating transcriptomics, proteomics, ChIP-qPCR, and functional assays, our findings position ATF3 as a dual effector—transducing ER stress signals and impairing mitochondrial function via NDUFS1 repression. Considering the pivotal function of mitochondria in cardiac physiology and the growing evidence linking mitochondrial dysfunction to cardiotoxicity (Jin et al., 2024; Junjun et al., 2025), this newly defined pathway offers both mechanistic insight and therapeutic targets. Targeting ATF3 or restoring NDUFS1 could serve as a viable approach to alleviate sorafenib-induced cardiac toxicity while preserving its anticancer efficacy.

Although this study demonstrated that ATF3 silencing alleviates mitochondrial dysfunction and restores NDUFS1 expression, further validation using ATF3 overexpression or NDUFS1 rescue approaches will be important to confirm the bidirectional regulatory relationship and to establish causality more robustly.

While our results demonstrate that sorafenib treatment and ATF3 upregulation are associated with marked repression of NDUFS1 and mitochondrial dysfunction, whether NDUFS1 downregulation alone is sufficient to recapitulate the full spectrum of mitochondrial impairment remains to be tested. We did not perform isolated NDUFS1 knockdown experiments in the absence of sorafenib in this study. Nevertheless, the tight temporal correlation between NDUFS1 suppression and key mitochondrial injury markers—including decreased Complex I activity, reduced ATP production, and ROS accumulation—strongly supports its central role. Future studies incorporating NDUFS1 loss-of-function models without sorafenib exposure will be essential to confirm whether NDUFS1 repression is sufficient, and not merely necessary, for driving the cardiotoxic phenotype.

The cardiotoxic effects of sorafenib pose a major limitation to its long-term clinical use, especially in patients without prior cardiovascular risk factors (Balachandran et al., 2024; Mohammed et al., 2024). Our findings reveal that the ATF3–NDUFS1 axis is not only a central mediator of mitochondrial dysfunction and cardiomyocyte apoptosis, but also a promising target for clinical intervention. The strong inverse correlation between ATF3 and NDUFS1, coupled with the ability of ATF3 knockdown to restore mitochondrial function and cell viability, suggests that these molecules may serve as sensitive biomarkers for predicting sorafenib-induced cardiotoxicity.

Therapeutic strategies aimed at inhibiting ATF3 or restoring NDUFS1 expression could potentially protect cardiac tissues from TKI-induced injury without compromising the anticancer efficacy of sorafenib. The development of selective ATF3 inhibitors, mitochondrial-protective agents, or gene therapy approaches may thus represent viable adjunct treatments to reduce cardiovascular side effects. Importantly, because ER stress is a shared mechanism across multiple chemotherapeutic and targeted agents, the ATF3-NDUFS1 axis may have broad applicability beyond sorafenib. Targeting this conserved stress-response pathway could form the basis for a generalizable cardio-oncology strategy aimed at preserving cardiac function during cancer therapy. Future research should confirm these results in clinical cohorts and evaluate the effectiveness of ATF3-targeted therapies in preclinical and human studies.

Although H9C2 cells provide a useful platform for mechanistic studies, their immortalized nature and embryonic rat origin may not fully represent the physiological responses of mature human cardiomyocytes. While our in vivo mouse model partially compensates for this limitation, future studies using more physiologically relevant systems—such as primary cardiomyocytes or human induced pluripotent stem cell–derived cardiomyocytes (hiPSC-CMs)—will be essential to validate the translational relevance of the ATF3–NDUFS1 signaling axis in sorafenib-induced cardiotoxicity. Nevertheless, H9C2 cells have been widely employed in early-phase cardiotoxicity research, particularly for studying endoplasmic reticulum stress, mitochondrial dysfunction, and apoptosis, due to their reproducibility and sensitivity to stress stimuli (Ma et al., 2020; Hou et al., 2021). These characteristics justify their use in this study to explore fundamental mechanisms before transitioning to higher-fidelity models.

In clinical settings, early detection of sorafenib-induced cardiac injury typically involves monitoring for symptoms such as fatigue, dyspnea, and chest discomfort, in addition to periodic evaluation of cardiac biomarkers (e.g., troponins, BNP/NT-proBNP) and imaging studies like echocardiography to assess left ventricular function. Electrocardiographic changes may also provide early warning signs. If cardiotoxicity is suspected, dose adjustment or temporary discontinuation of sorafenib is often recommended. Management may include supportive therapies such as β-blockers, ACE inhibitors, or mineralocorticoid receptor antagonists, which are commonly used in heart failure treatment. These measures can mitigate the progression of cardiac dysfunction while allowing continuation of anticancer therapy where feasible.