Identification of biomarkers in myocardial hypertrophy and fibrosis via integrated transcriptomic and phosphoproteomic profiling Authors: Kang Luo, Yangguang Ma, Yayong Zhang, Yunfeng Zi, Guangyuan Guo, Xingyuan Chen, Wenjuan Feng, Qifan Jing, Lin Li, Shen Han, Bingrong Zheng Published: 2026 ABSTRACT AbstractMyocardial hypertrophy and fibrosis (MHF) are key structural changes in adverse cardiac remodeling. They are commonly observed in hypertension, metabolic disorders, and cardiomyopathies. These processes contribute to heart failure and sudden death, yet their molecular mechanisms remain unclear. We established in vivo and in vitro models of MHF and performed transcriptomic and label-free phosphoproteomic sequencing. These data were integrated with public single-cell RNA sequencing datasets. This multi-tier strategy enabled us to move beyond bulk-tissue signals and prioritize candidates based on cross-omics concordance and cell-type-specific expression. We identified five candidate biomarkers, CLIP2, PLA2G4A, PRRX1, SLAIN2, and XPO1. These candidates showed coordinated upregulation or downregulation at both the transcriptomic and phosphoproteomic levels. They were also specifically enriched in disease-relevant cardiac cell populations, including cardiomyocytes and fibroblasts. PRRX1 and PLA2G4A were enriched in fibrogenic fibroblast subsets and increased along activation trajectories. XPO1 and SLAIN2 were downregulated in cardiomyocytes and fibroblasts, suggesting impaired nuclear export and cytoskeletal regulation. Molecular docking revealed flurandrenolide and selinexor as potential modulators of PLA2G4A and XPO1 respectively. However, their cardiovascular safety and efficiency remain uncertain. Notably, the therapeutic implications of XPO1 modulation appear context-dependent, warranting careful evaluation of its functional role in specific MHF etiologies. Overall, these findings highlight novel targets and provide mechanistic insights into MHF progression.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-47797-9.Keywords: Hypertrophic and fibrosis cardiomyopathy, Biomarkers, Transcriptome, Label-free quantitative phosphoproteome, Single-cell RNA sequencingSubject terms: Cardiology, Cell biology, Diseases, Molecular biology Introduction Cardiovascular diseases remain the leading cause of morbidity and mortality worldwide1. Among the structural alterations that drive adverse cardiac remodeling, myocardial hypertrophy and fibrosis (MHF) are critical pathological hallmarks. They directly contribute to heart failure, arrhythmia, and sudden cardiac death2. Myocardial hypertrophy typically develops in response to chronic pressure overload, neurohormonal activation, or metabolic stress. It is frequently observed in hypertension, valvular heart disease, and metabolic syndrome3. Although hypertrophy is initially compensatory and helps preserve cardiac output, sustained hypertrophy leads to maladaptive remodeling and progressive dysfunction4. At the same time, myocardial fibrosis is characterized by excessive extracellular matrix (ECM) deposition, primarily collagen I and III. This reduces myocardial compliance, impairs diastolic function, and promotes arrhythmogenesis5,6. The progression of MHF is regulated by complex and interconnected molecular networks, including the TGF-β and Smad axis, the renin-angiotensin-aldosterone system (RAAS), and PI3K/Akt pathways, as well as inflammatory cascades7,8. Despite advances in mechanistic understanding, current therapeutic strategies mainly focus on reducing hemodynamic load and neurohormonal activation. They have limited efficacy in reversing established hypertrophy and fibrosis9. In addition, early-stage myocardial fibrosis is often asymptomatic and may not be detected by conventional imaging, such as echocardiography or even cardiac magnetic resonance. Cardiac magnetic resonance can also be limited by accessibility and cost10,11. The lack of reliable molecular biomarkers and the limited capacity for individualized risk stratification hinders early diagnosis and targeted interventions, particularly in patients with metabolic disorders or genetic predispositions. To comprehensively dissect the molecular landscape of MHF, we integrated data from three complementary experimental systems. These included an isoproterenol (ISO)-induced rat model that represents acute neurohormonal stress, a transverse aortic constriction (TAC)-induced mouse model that mimics chronic hemodynamic overload, and an in vitro rat H9c2 cardiomyoblast system for mechanistic analyses under controlled conditions. Although these models differ in species, etiology, and biological context, integrating them supports the identification of convergent molecular nodes that are consistently dysregulated across distinct triggers of MHF. Such cross-model convergence may reflect core regulatory mechanisms extend beyond specific upstream stimuli. It may also increase the likelihood of identifying evolutionarily conserved and functionally relevant biomarkers. Integrative analyses across heterogeneous disease models have been effective in prior cardiovascular studies, including cross species single cell mapping of conserved cardiac cell states12 and multi-model validation of IL-11 as a central driver of fibrosis13. These observations highlight the need for a deeper molecular understanding of MHF pathogenesis and the discovery of actionable biomarkers to support precision medicine strategies14,15. Multi-omics technologies, including transcriptomics, proteomics, and phosphoproteomics, provide an opportunity to systematically dissect disease mechanisms from gene expression to protein function and signaling dynamics16. Compared with single-omics approaches, integrative multi-omics offers a more comprehensive view of molecular alterations. It also links transcriptional regulation with post-translational modifications that are crucial for pathological remodeling. Furthermore, single-cell RNA sequencing (scRNA-seq) reveals cell-type-specific expression profiles and dynamic cellular states, which helps address the limited resolution of bulk tissue analysis17. In this study, we applied an integrative multi-omics framework that combines transcriptomics, proteomics, and label-free quantitative (LFQ) phosphoproteomics, with single-cell transcriptomic analysis to investigate the molecular landscape of MHF. By identifying differentially expressed genes and phosphoproteins and mapping them to cellular contexts, we aimed to discover key biomarkers, clarify underlying signaling pathways, and propose potential therapeutic targets. This systems biology approach advances mechanistic understanding of MHF and supports the development of early diagnostic tools and precision therapies for cardiac remodeling. Materials and methods Animal and cell models for myocardial hypertrophy and fibrosis To investigate the molecular mechanisms underlying MHF, we established in vivo and in vitro models. Eight-week-old male Sprague–Dawley rats (250–300 g; SPF Biotechnology Co., Ltd., Beijing, China) were randomly assigned to a control group or an MHF group (n = 3 per group). The MHF model was induced by subcutaneous injection of ISO (1.5 mg/kg/day) for 7 consecutive days. Control rats received an equal volume of saline18. At the endpoint, rats were euthanized by inhalation of an overdose of isoflurane and hearts were excised. Heart tissues were used for morphological assessment and histological validation by hematoxylin-eosin (H&E) staining (Supplementary Fig. 5B). They were also used for molecular analyses, including quantitative polymerase chain reaction (qPCR) of ANP and BNP expression (Supplementary Fig. 5E) and transcriptomic sequencing. All animal procedures complied with institutional ethical guidelines and were approved by the Ethics Committee of Yan’an Hospital, Kunming City (Approval No. 2022035). For the in vitro model, rat cardiomyoblast H9c2 cells (Procell Life Science, Wuhan, China) were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS). Cells were treated with 25 µM ISO for 48 h to induce hypertrophic and fibrotic phenotypes19. After treatment, some cells was subjected to rhodamine-phalloidin staining to visualize cytoskeletal alterations (Supplementary Fig. 5A). The remaining cells were collected for qPCR of ANP and BNP expression (Supplementary Fig. 5D), as well as proteomic and phosphoproteomic analyses. Transcriptome sequencing and data pre-processing Total RNA was extracted from rat myocardium. RNA quality was assessed using a NanoPhotometer (IMPLEN) and an Agilent 2100 Bioanalyzer. Libraries were sequenced on an Illumina NovaSeq platform to generate 150 bp paired-end reads. Clean reads were aligned to the Rattus norvegicus reference genome using HISAT2. Gene expression was quantified as fragments per kilobase of transcript per million mapped reads (FPKM) using featureCounts. Differentially expressed genes (DEGs) were identified using DESeq2, with thresholds of |log₂(fold change)| > 0.5 and p < 0.05. Principal component analysis (PCA) confirmed clear separation between MHF and control groups20,21. Proteome sequencing and data pre-processing Proteins were extracted from H9c2 cells and quantified using a bicinchoninic acid (BCA) assay. Samples (500 µg protein per sample) were processed using the filter-aided sample preparation (FASP) method22, followed by trypsin digestion and C18-based desalting. Phosphopeptides were enriched using Fe-NTA kits (Thermo Scientific)23. Liquid chromatography tandem mass spectrometry (LC-MS/MS) was performed on an Easy-nLC 1200 system coupled to a timsTOF Pro mass spectrometer (Bruker) in parallel accumulation serial fragmentation (PASEF) mode. Raw data were searched against the UniProt Rattus norvegicus proteome (release 2022.01) using MSFragger24. Differentially expressed phosphoproteins were defined as those with |log₂(fold change)| > 0.5 and a false discovery rate (FDR) < 0.05. Additional parameters are provided in Supplementary Tables 4 − 3. LFQ phosphoproteome sequencing and data pre-processing Protein extraction and digestion Cell pellets were lysed in SDT buffer (4% SDS, 100 mM Tris-HCl, 100 mM dithiothreitol (DTT), pH 7.6). Lysates were heated at 100 °C for 3 min, sonicated, and centrifuged at 16,000×g for 20 min at 4 °C. Protein concentrations were measured using a BCA assay. For each sample, 500 µg of protein was processed using FASP. This procedure included reduction with DTT, alkylation with iodoacetamide (IAA), and trypsin digestion at a 1:100 enzyme-to-substrate ratio at 37 °C overnight. Peptides were desalted, vacuum-dried, and reconstituted in 0.1% formic acid (FA). Phosphopeptide enrichment Phosphopeptides were enriched using an Fe-NTA Phosphopeptide Enrichment Kit (Thermo Scientific, A32992) according to the manufacturer’s instructions. Samples were then dried and resuspended in 0.1% FA. Phosphoproteomic data acquisition and processing Phosphopeptide-enriched samples were analyzed by LC-MS/MS using an Easy-nLC 1200 system coupled online to a timsTOF Pro mass spectrometer (Bruker) operating in PASEF mode. Peptides were separated on a C18 reversed-phase column (250 mm × 75 μm, 1.6 μm particle size; IonOpticks). A linear gradient of 2% to 80% acetonitrile in 0.1% FA over 60 min at 300 nL/min. The instrument was operated in data-dependent acquisition mode. The scan range was m/z 100 to 1700, and the ion mobility range wad 0.6 to 1.6 Vs/cm². Dynamic exclusion was set to 0.4 min. Each cycle consisted of one MS scan followed by 10 PASEF MS/MS scans, with a total cycle time of 1.1 s. Raw files were processed using MSFragger (v3.4) within the FragPipe platform (v18.0) against the UniProt Rattus norvegicus reference proteome (UP000002494; downloaded January 2022; 36,261 entries). Search parameters were as follows: trypsin digestion with up to two missed cleavages; carbamidomethylation (C) as a fixed modification; oxidation (M), protein N terminal acetylation, and phosphorylation (S/T/Y) as variable modifications; precursor mass tolerance of 20 ppm; and fragment mass tolerance of 0.02 Da. Peptide-spectrum matches (PSMs) and protein identifications were filtered at 1% FDR using the target-decoy strategy implemented in Percolator. Phosphorylation site localization was performed using PTM-Shepherd, and only sites with localization probability ≥ 0.75 were retained for quantification25. LFQ intensities for phosphopeptides were extracted and log₂-transformed. Missing values were imputed Perseus (v2.0) from a normal distribution (width = 0.3 × SD; downshift = 1.8 × SD) restricted to low intensity regions26. Data were normalized by median centering across samples. Differentially phosphorylated sites were identified using limma (v3.54.0) with empirical Bayes moderation. P-values were adjusted using the Benjamini–Hochberg procedure. Sites with adjusted p < 0.05 and |log₂(fold change)| > 0.58, which corresponds to a 1.5 fold change, were considered significantly altered27. Differential expression analysis Differential expression analysis of transcriptome data was performed using DESeq2 (v1.34.0). DEGs were identified between MHF and control groups using thresholds of |log2 fold change (FC)| > 0.5 and p < 0.0528. For phosphoproteome data, differentially expressed phosphorylated proteins (DEPPs) were identified using limma (v3.54.0) with the same criteria27. Volcano plots for both DEGs and DEPPs were generated using ggplot2 (v3.4.1)29. The plots highlighted the top 10 upregulated and downregulated genes or phosphoproteins based on |log2FC|. Heatmaps were constructed using ComplexHeatmap (v2.14.0)30 to visualize expression patterns across samples. In addition, protein-level differential expression was assessed using limma. Log2FC values were calculated for all proteins to support subsequent ratio analysis between the phosphoproteome and proteome datasets. Identification and enrichment analysis of biomarkers To correct for baseline protein expression, ratio values were calculated by dividing the log2FC of DEPPs from the phosphoproteome by the corresponding log2FC of the matched proteins from the proteome. DEPPs with a ratio > 1.0 were considered phosphorylation-specific and selected as candidate DEPPs. Venn diagrams were generated using VennDiagram (v1.7.1)31 to identify overlapping genes between upregulated DEGs and upregulated candidate DEPPs. Overlaps were also identified between downregulated DEGs and downregulated candidate DEPPs. The intersecting genes were mapped to their human homologs using msigdbr (v7.5.1)32. The top five candidates, ranked by ratio values, were designated as key biomarkers. These biomarkers were annotated within the nine-quadrant plot of DEGs and DEPPs. Functional enrichment analysis was performed using clusterProfiler (v4.2.2)33. Analyses focused on Gene Ontology (GO) terms, including biological process (BP), cellular component (CC), and molecular function (MF), and on Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, using a cutoff of p < 0.0534. The top five GO terms in each category and the top ten KEGG pathways were visualized using enrichplot (v1.18.3) and GOplot (v1.0.2)35. Comprehensive functional characterization analysis of biomarkers To further characterize the biomarkers, functional similarity scores were calculated using GOSemSim (v2.24.0)36. Scores > 0.50 were considered indicative of significant functional relationships. Chromosomal locations of the biomarkers were visualized via RCircos (v1.22)37. A gene-gene interaction (GGI) network was constructed using the GeneMANIA database (http://genemania.org/). The network highlighted co-expression, co-localization, and pathway interactions associated with these biomarkers. Subcellular localization was predicted using mRNALocater (http://bio-bigdata.cn/mRNALocater/) based on FASTA sequences retrieved from NCBI (https://www.ncbi.nlm.nih.gov/). Gene set enrichment analysis (GSEA) To explore the biological pathways associated with the biomarkers, GSEA was performed using the pre-processed transcriptome data. For each biomarker, Spearman correlation analysis was conducted between the biomarker and all other genes using the psych package (v2.1.6)38. Genes were then ranked by correlation coefficients to generate the input gene lists. GSEA was performed using clusterProfiler (v4.2.2) with the MSigDB gene set c2.cp.kegg.v2023.1.Hs.symbols.gmt as the reference. Significant enrichment was defined as FDR < 0.05 and |normalized enrichment score (NES)| > 1. For each biomarker, the top 10 significantly enriched pathways were visualized using the enrichplot (v1.18.3) and ranked by absolute NES. Construction of regulatory network To investigate the upstream regulatory mechanisms, transcription factors (TFs) were predicted using the TRRUST database via the miRNet platform. TF-mRNA regulatory networks were then constructed. miRNA-mRNA interactions were retrieved from the NetworkAnalyst database. In addition, lncRNA-miRNA interactions were predicted using both miRNet and StarBase. Overlapping lncRNAs were retained to improve prediction reliability. These interactions were integrated to construct TF–miRNA–lncRNA–mRNA regulatory networks. All networks were visualized using Cytoscape (v3.8.2)39. Drug prediction and molecular docking analyzes Potential therapeutic compounds targeting the biomarkers were predicted using the Drug–Gene Interaction Database (DGIdb) (https://dgidb.org). Drug–gene interaction networks were visualized in Cytoscape (v3.8.2). Candidate drugs with high interaction scores and approved clinical status were prioritized for molecular docking. The 3D structures of the selected drugs were obtained from the PubChem. Protein structures of the biomarkers were retrieved from the Protein Data Bank (PDB). Docking simulations were performed using the CB-Dock web server, which performs blind docking based on cavity detection. Binding affinities were evaluated using binding energy values. A threshold of less than − 5 kcal/mol was used to indicate strong binding. Docking conformations and interactions were visualized using PyMOL (v2.5)22. The scRNA-seq analyzes To dissect the cellular landscape of MHF, we reanalyzed the publicly available scRNA-seq dataset GSE12006440. This dataset includes mouse heart tissues collected at 0 weeks (control, n = 2), 2 weeks, and 5 weeks after pressure overload (n = 9 total). Data were processed using Seurat (v4.1.0)23 with standard workflows for quality control, normalization, and clustering. Cell types were annotated based on canonical marker genes from reported in the literature. DEGs between clusters were identified using thresholds of |log₂(fold change)| > 1.0 and adjusted p < 0.05. Cell-type-specific expression validation in cardiomyocytes To address concerns regarding the integration of bulk heart tissue transcriptomics with H9c2 cardiomyoblast phosphoproteomics, we performed cell-type-specific expression validation using the scRNA-seq dataset GSE120064. Cardiomyocytes were extracted from the annotated cell types and subclustered to obtain a homogeneous cardiomyocyte population. Differential expression analysis of the five candidate biomarkers (CLIP2, PLA2G4A, PRRX1, SLAIN2, and XPO1) between MHF and control groups was conducted using the Wilcoxon rank-sum test implemented in the stats package (v4.3.1) in R. Expression patterns were visualized using violin plots generated with ggplot2 (v3.4.1). Statistical significance was defined as p < 0.05. Additionally, to validate the biological relevance of PLA2G4A phosphorylation in cardiac contexts, we queried the PhosphoSitePlus database (https://www.phosphosite.org/) for experimentally verified phosphorylation sites of PLA2G4A across multiple tissues, including heart tissue41. Literature evidence supporting PLA2G4A phosphorylation in myocardial ischemia models was also reviewed42. Functional characterization of cell functional characterization and identification of key cells To characterize biological functions across annotated cell types, pathway enrichment analysis was performed using the ReactomeGSA (v1.12.0) package (p < 0.05)43. The top 10 most variable pathways were selected based on the difference between their maximum and minimum enrichment scores across cell types. These pathways were visualized using heatmaps. We also examined the abundance and distribution of each cell type in control and MHF samples. This analysis revealed clear shifts in cellular composition between healthy and diseased states. Next, we mapped the expression of the five biomarkers across cell types. Cell types with consistently higher biomarker expression, particularly cardiomyocytes and fibroblasts, were considered major contributors to MHF progression. This interpretation was supported by marker gene expression profiles and prior studies linking these cell populations to myocardial remodeling. Cell communication and pseudo-time analysis of key cells To investigate intercellular communication dynamics in MHF, ligand-receptor interaction s were inferred using CellChat (v1.5.0)44. This analysis identified signaling pathways mediating communication among cardiomyocytes, fibroblasts, and other cell types. Results were visualized as bubble plots that display the strength and direction of incoming and outgoing signaling. To further explore cellular heterogeneity within key populations, cardiomyocytes and fibroblasts were subclustered (resolution = 0.4) to identify functional subpopulations. Developmental trajectories were inferred using Monocle (v2.22.0)45. Pseudotime analysis modeled transitions among cellular states and revealed dynamic changes in gene expression, including biomarker expression, along inferred trajectories. These analyses provide insight into temporal changes in cell states that may contribute to MHF pathogenesis. Stress response gene association analysis To address concerns regarding the potential identification of generalized stress-response genes rather than MHF-specific biomarkers, we performed a stress response gene filtering analysis. The GOBP_INTEGRATED_STRESS_RESPONSE_SIGNALING gene set (32 genes) was downloaded from MSigDB (https://www.gsea-msigdb.org/gsea/msigdb/mouse/geneset/GOBP_INTEGRATED_STRESS_RESPONSE_SIGNALING.html). Spearman correlation analysis was conducted between each of the five candidate biomarkers (CLIP2, PLA2G4A, PRRX1, SLAIN2, and XPO1) and the stress response gene set using the stats package (v4.3.1) in R. Correlation coefficients and corresponding p-values were calculated to quantify the degree of overlap and association between the candidate biomarkers and generalized stress response genes. Results were visualized using lollipop plots to display the Spearman correlation analysis between biomarkers and stress response genes. Statistical analysis All analyzes were performed in R (v 4.2.2). A p-value < 0.05 was considered statistically significant. Results Identification of 1779 DEGs and 1909 DEPPs PCA showed clear separation between the MHF and control groups along the first principal component (PC1). PC1 accounted for 42.7% of the total variance in transcriptomic data (Supplementary Fig. 1A) and 28.9% in phosphoproteomic data (Supplementary Fig. 1B). These results indicate robust molecular differences between conditions. Using the pre-processed transcriptome dataset, we identified 1779 DEGs between MHF and control samples. Among them, 1177 genes were upregulated and 602 genes were downregulated (|log2 FC| > 0.5, p < 0.05) (Fig. 1A). Using the pre-processed LFQ phosphoproteome dataset, we identified 1909 DEPPs between MHF and control samples. This set included 834 upregulated proteins and 1,075 downregulated proteins (Fig. 1B). Differential expression results for the proteome dataset are provided in Supplementary Table 1. Identification of omics identifies a preliminary candidate set for MHF, refined by cell-type-specific expression analysis. Transcriptomic and phosphoproteomic profiling of MHF versus control samples reveals widespread molecular alterations. (A) Volcano plot displaying 1779 differentially expressed genes (DEGs) from rat whole-heart RNA-seq (|log₂FC| > 0.5, p < 0.05). Points represent individual genes with significant expression changes; red dots indicate upregulated genes and blue dots downregulated ones (P < 0.05). (B) Volcano plot of 1,909 differentially expressed phosphorylated proteins (DEPPs) from ISO-treated H9c2 cells. (C, D) Venn diagram showing overlap between 1779 DEGs and 592 candidate DEPPs, yielding 22 preliminary candidate molecules. Principal component analysis (PCA) confirmed a clear separation between MHF and control samples (Supplementary Fig. 1A–B). Proteomic differential results are detailed in Supplementary Table 1. (E) Nine-quadrant plot integrating transcript and phosphoprotein regulation; five core biomarkers (CLIP2, PLA2G4A, PRRX1, SLAIN2, XPO1) were subsequently prioritized based on their enrichment in disease-relevant cardiac cell types, as validated by single-cell RNA sequencing (see Figs. 6 and 8). (F) Gene Ontology (GO) enrichment for biological processes (BP), cellular components (CC), and molecular functions (MF) (p < 0.05). (G) KEGG pathway enrichment reveals significant involvement in lipid metabolism and cellular signaling pathways. KEGG database imagery and information used in this figure are with permission from Kanehisa Laboratories. Supplementary Tables 2–4 provide detailed statistics. Selection and enrichment analysis of 5 biomarkers and functional roles A total of 592 candidate DEPPs were identified after correcting for proteomic effects (Supplementary Table 2). It is important to note that the subsequent integration of bulk heart tissue transcriptomics (reflecting a mixture of cardiomyocytes, fibroblasts, endothelial cells, and immune cells) with H9c2 cardiomyoblast phosphoproteomics represents a methodological limitation, as the cellular contexts differ significantly. The resulting overlapping candidates should therefore be considered as a preliminary list requiring further cell-type-specific validation. Integrating 1779 DEGs with these DEPPs yielded 22 candidate genes (Fig. 1C and D), of which 20 were mapped to human homologs (excluding LOC100359600 and RGD1560065, which are rat-specific; Supplementary Table 3). Based on the highest ratio values, five key biomarkers were prioritized: CLIP2, PLA2G4A, PRRX1, SLAIN2, and XPO1, and were visualized in the nine-quadrant plot (Fig. 1E). Stress response gene association analysis To evaluate whether the identified biomarkers represent generalized stress-response genes or MHF-specific molecular signatures, we assessed the correlation between the five candidate biomarkers and the GOBP_INTEGRATED_STRESS_RESPONSE_SIGNALING gene set (32 genes) from MSigDB. Spearman correlation analysis revealed varying degrees of association between the biomarkers and stress response genes (Figure S2). Among the five biomarkers, CLIP2 showed the lowest overall correlation with stress response genes, with only one significant correlation observed (DDIT3: r = − 0.886, p = 0.033), suggesting limited overlap with generalized stress response pathways. In contrast, PLA2G4A demonstrated multiple significant correlations with stress response genes, including AKT2 (r = − 0.886, p = 0.033), EIF2S1 (r = − 0.886, p = 0.033), HSPA5 (r = − 0.943, p = 0.017), DELE1 (r = − 0.886, p = 0.033), and TMEM33 (r = − 0.886, p = 0.033), indicating partial involvement in stress response mechanisms. PRRX1 exhibited one highly significant correlation with TMEM33 (r = − 1.000, p = 0.0028), while SLAIN2 showed significant positive correlations with PTPN2 (r = 0.886, p = 0.033) and TMEM33 (r = 0.943, p = 0.017). XPO1 displayed significant positive correlations with EIF2S1 (r = 0.943, p = 0.017), HSPA5 (r = 0.886, p = 0.033), and IMPACT (r = 1.000, p = 0.0028). Despite these individual correlations, the overall pattern indicated that no single biomarker was uniformly correlated with the majority of stress response genes. These results suggest that while some biomarkers may partially overlap with stress response pathways, the overall signature is not solely driven by generalized cellular stress, supporting their potential specificity for MHF pathogenesis. Functional enrichment and subcellular localization GO enrichment analysis indicated that the biomarkers were mainly associated with biological processes such as ether lipid biosynthetic process, cellular components such as microtubule, and molecular functions such as nuclear export signal receptor activity (Fig. 1F, Supplementary Table 4). KEGG analysis identified 24 significantly enriched pathways. The top pathways included alpha linolenic acid metabolism, linoleic acid metabolism, and ether lipid metabolism, suggesting roles in lipid metabolism and intracellular structural dynamics (Fig. 1G). Functional similarity analysis showed that CLIP2 had the highest similarity score among the biomarkers (Supplementary Fig. 3A), which suggests a central role in shared biological functions. Chromosomal mapping located CLIP2 on chr7, PLA2G4A on chr1, PRRX1 on chr1, SLAIN2 on chr4, and XPO1 on chr2 (Supplementary Fig. 3B). A GGI network constructed using GeneMANIA identified functional links with 20 associated genes, including SLAIN1, RANGAP1, and MKNK1. These genes were enriched in processes related to microtubule end dynamics, depolymerization, and regulation of microtubule polymerization (Supplementary Fig. 3C). Subcellular localization analysis suggested that all five biomarkers mainly localized in the cytoplasm (Supplementary Fig. 3D). This finding supports their potential involvement in cytoskeletal organization, intracellular signaling, and transport mechanisms. Signaling pathways enriched for biomarkers To explore mechanisms through which the biomarkers may contribute to MHF, we performed GSEA using transcriptome data from the MHF group. CLIP2, PLA2G4A, PRRX1, SLAIN2, and XPO1 were significantly enriched in 12, 21, 26, 25, and 12 pathways, respectively (Supplementary Table 5). All five biomarkers were co-enriched in the ribosome pathway (Fig. 2A–E). In addition, CLIP2, PLA2G4A, PRRX1, and SLAIN2 were commonly enriched in the aminoacyl tRNA biosynthesis. PLA2G4A, PRRX1, SLAIN2, and XPO1 shared enrichment in the proteasome pathway. PLA2G4A, PRRX1, and SLAIN2 were also enriched in glycosaminoglycan degradation, regulation of autophagy, propanoate metabolism, and the citrate cycle (TCA cycle). These results suggest that the biomarkers may act in interconnected pathways related to protein synthesis, protein degradation, and metabolic regulation. They also support a broader role in cellular homeostasis and disease associated processes. Gene set enrichment analysis (GSEA) reveals shared and distinct pathway enrichment for core biomarkers in MHF. GSEA illustrates core biological pathways associated with biomarker expression. (A–E) Gene set enrichment plots for CLIP2, PLA2G4A, PRRX1, SLAIN2, and XPO1 show consistent enrichment in the ribosome pathway. Additional co-enriched pathways include aminoacyl-tRNA biosynthesis, proteasome, glycosaminoglycan degradation, regulation of autophagy, and the citrate cycle (TCA). Full enrichment results are provided in Supplementary Table 5. Revelation of regulatory networks for biomarkers TRRUST was used to predict TFs that may regulate the biomarkers. In total, 56 TFs were identified and used to construct a TF mRNA network (Fig. 3A, Supplementary Table 6). Within this network, several TFs, including NFKB1 and RFX1, were linked to three biomarkers. This suggests that they may act as upstream regulators by modulating multiple biomarkers simultaneously. In addition, 141 miRNAs targeting the biomarkers were predicted using NetworkAnalyst (Supplementary Table 7). A total of 698 lncRNAs were predicted in the miRNet and 750 lncRNAs were predicted in the Starbase. Among them, 297 lncRNAs overlapped between the two databases (Supplementary Table 8). Based on these interaction pairs, we constructed an lncRNA-miRNA-mRNA regulatory network. This network included putative regulators such as XIST (lncRNA) and has-miR 485 5p (miRNA)(Fig. 3B). These results support further investigation of regulatory interactions that may contribute to MHF. Regulatory networks of biomarkers at transcriptional and post-transcriptional levels. Biomarkers are regulated by distinct transcriptional and post-transcriptional mechanisms. (A) TF–mRNA interaction network generated using TRRUST, identifying 56 transcription factors (TFs); key regulators such as NFKB1 and RFX1 co-regulate multiple biomarkers. (B) lncRNA–miRNA–mRNA (ceRNA) network constructed by integrating miRNA predictions from NetworkAnalyst and lncRNAs from miRNet and StarBase. XIST and hsa-miR-485-5p are highlighted as central nodes. Full details in Supplementary Tables 6–8. Predicted drug interactions and molecular docking Using DSigDB, we predicted drugs potentially targeting the biomarkers. We identified 12 candidate drugs for PLA2G4A and six drugs for XPO1.No drugs were predicted for CLIP2, PRRX1, and SLAIN2. A drug biomarker interaction network was constructed for each biomarker (Fig. 4A). Drug–biomarker interaction network and docking analysis. Predicted drug–target interactions suggest therapeutic potential. (A) DSigDB-predicted drug interactions identify 12 compounds for PLA2G4A and 6 for XPO1. No drugs were identified for CLIP2, PRRX1, or SLAIN2. (B) Docking simulation of flurandrenolide with PLA2G4A yields a binding energy of − 6.8 kcal/mol; hydrogen bonds involve residues D43 and L39. (C) XPO1–selinexor docking reveals a binding energy of − 7.3 kcal/mol with T931 and T933 contributing to hydrogen bonding. Full docking parameters are listed in Supplementary Table 9. For PLA2G4A, flurandrenolide had the highest combined score and was selected for docking (Supplementary Table 9). For XPO1, selinexor had the highest combined score and was selected for docking (Supplementary Table 9). Docking results suggested strong binding affinities for both target drug pairs (Fig. 4B, C) The strongest predicted binding was observed for XPO1 and selinexor. PLA2G4A showed a binding energy of − 6.8 kcal/mol with flurandrenolide. Hydrogen bonds involved residues such as D43 and L39 (Fig. 4B). XPO1 showed a binding energy of − 7.3 kcal/mol with selinexor. Hydrogen bonds involved residues such as T931 and T933 (Fig. 4C). Overall, these results support potential interactions between the biomarkers and candidate drugs and suggest possible therapeutic relevance. Annotation of six cell types After bulk RNA-seq analysis, we reanalyzed the scRNA-seq dataset GSE120064 to explore MHF at the cellular level. We first visualized nFeature RNA, nCount RNA, and percent mitochondrial content before and after quality control (QC) (Supplementary Fig. 4A, B). After standard processing, 2000 highly variable genes (HVGs) were selected for downstream analysis. The top 10 most variable genes were labeled (Supplementary Fig. 4C). PCA was then performed, and the top 20 principal components (PCs) were selected for downstream analyses (Fig. 5A and B). Clustering identified 20 distinct clusters (Fig. 5C and D). These clusters were annotated into six major cell types, including cardiomyocytes, endothelial cells, fibroblasts, macrophages, T cells, and granulocytes (Fig. 5E, F). A bubble plot showed marker genes with high expression specificity across these cell types (Fig. 5G). scRNA-seq identifies six cardiac cell types in MHF samples. Single-cell transcriptomics defines six cardiac cell types and marker gene expression. (A, B) PCA and variance plots based on 2,000 highly variable genes (HVGs). (C, D) UMAP dimensionality reduction identifies 20 initial clusters. (E) Expression levels of marker genes across distinct cell clusters. (F) Annotation results reveal six major cardiac cell types: cardiomyocytes, endothelial cells, fibroblasts, macrophages, T cells, and granulocytes. (G) Bubble plot illustrates representative marker gene expression across the six cell populations. Cellular functions and identification of cardiomyocytes and fibroblasts as key cell types Pathway enrichment analysis across the six cell types identified 1733 significantly enriched pathways (p < 0.05) (Supplementary Table 10). Top pathways included ficolins bind to repetitive carbohydrate structures on the target cell surface, ALKBH2 mediated reversal of alkylation damage, and mitochondrial transcription termination. These results indicate that diverse biological processes involved in MHF (Fig. 6A). Single-cell resolution validates the cellular origin of core MHF biomarkers and addresses tissue heterogeneity concerns. Single-cell resolution mapping of preliminary omics candidates in pressure-overloaded mouse hearts. (A) Functional enrichment of all six cell types yields 1733 significantly enriched pathways, including immune response and mitochondrial function. (B) Cell count distribution indicates cardiomyocytes are the most abundant. (C) Cell proportion analysis shows increased cardiomyocytes in MHF, whereas T cells and endothelial cells are relatively enriched in control samples. (D) UMAP visualization of expression for the five core biomarkers across all cell populations. (E) Quantitative analysis confirms compartment-specific enrichment: CLIP2 and SLAIN2 are predominantly expressed in cardiomyocytes, whereas PLA2G4A, PRRX1, and XPO1 show strong upregulation in activated fibroblasts—the two principal effector lineages in MHF pathogenesis. This cell-type anchoring substantiates the pathological relevance of these candidates despite their initial identification from bulk tissue. Cardiomyocytes were the most abundant cell type (Fig. 6B). Cell proportion analysis demonstrated that cardiomyocytes were enriched in MHF samples. In contrast, T cells and endothelial cells were more prevalent in controls, which suggest disease-associated remodeling of cellular composition (Fig. 6C). We next examined biomarker expression across cell types (Fig. 6D). SLAIN2 and PRRX1 showed higher expression in cardiomyocytes. PLA2G4A, PRRX1, SLAIN2, and XPO1 showed higher expression in fibroblasts (Fig. 6E). This single-cell resolution mapping confirms that our candidate biomarkers are not artifacts of mixed-cell signals but are genuinely associated with the two principal cell types—cardiomyocytes and fibroblasts—known to be central drivers of myocardial hypertrophy and fibrosis. This interpretation is consistent with previous studies that highlight the roles of these cell populations in MHF46. Cell-type-specific expression validation in cardiomyocytes To validate that the identified biomarkers reflect cardiomyocyte-specific changes rather than artifacts of cellular composition differences between bulk tissue and H9c2 cells, we examined the expression of the five candidate biomarkers in the cardiomyocyte subpopulation from GSE120064. Wilcoxon rank-sum test analysis revealed that three of the five biomarkers showed significant differential expression between MHF and control groups in cardiomyocytes (Supplementary Fig. 6). Specifically, XPO1 was significantly downregulated in MHF cardiomyocytes (p = 8.0 × 10–19), corroborating the downregulation pattern observed across multiple omics layers. PRRX1 also showed significant downregulation in MHF cardiomyocytes (p = 1.7 × 10 − 6), consistent with its expression trend in bulk transcriptomics. SLAIN2 exhibited significant downregulation in MHF cardiomyocytes (p = 5.9 × 10–14), further supporting its role in MHF pathogenesis. In contrast, CLIP2 (p = 0.75) and PLA2G4A (p = 0.16) did not reach statistical significance in cardiomyocytes alone, though their expression trends were consistent with bulk tissue patterns. Notably, the lack of significant transcriptional changes for PLA2G4A in cardiomyocytes, despite its phosphorylation alterations in H9c2 cells, suggests that PLA2G4A may be primarily regulated at the post-translational level rather than the transcriptional level in cardiac stress responses. These results demonstrate that at least three biomarkers (XPO1, PRRX1, and SLAIN2) exhibit concordant dysregulation in purified cardiomyocytes, supporting their biological relevance beyond mixed-cell signals. To further substantiate the functional significance of PLA2G4A phosphorylation in cardiac contexts, we queried the PhosphoSitePlus database and identified experimentally verified phosphorylation sites of PLA2G4A in 14 tissues, including heart tissue. This confirms that PLA2G4A undergoes phosphorylation modifications in cardiac tissues under physiological and pathological conditions. Moreover, previous studies in mouse myocardial ischemia-reperfusion models have demonstrated that PLA2G4A (cPLA2α) phosphorylation is induced by TNF-α signaling and contributes to cardiomyocyte injury. Specifically, Saito et al. showed that TNF-α stimulation triggers cPLA2α phosphorylation in cultured cardiomyocytes, and pharmacological inhibition of TNF-α signaling suppresses this phosphorylation event42. These findings provide independent evidence that PLA2G4A phosphorylation is a biologically relevant modification in cardiomyocytes under stress conditions, strengthening the validity of integrating phosphoproteomic data from H9c2 cells with transcriptomic data from whole-heart tissue. Together, these cell-type-specific validations and literature evidence partially mitigate concerns regarding the integration of heterogeneous sample types. The concordant expression patterns in purified cardiomyocytes for three biomarkers (XPO1, PRRX1, and SLAIN2), combined with established evidence of PLA2G4A phosphorylation in cardiac contexts, support the biological plausibility of our multi-omics integration strategy. Specific communications of cell types Intercellular communication analysis showed significant differences in signaling patterns between MHF and control samples. Overall, MHF samples had fewer cell–cell interactions and lower interaction strength. This suggests attenuated intercellular communication under pathological conditions (Fig. 7A and B). Intercellular communication networks among cardiac cell types in MHF and control samples. Analysis of ligand–receptor interactions reveals network remodeling. (A, B) Cell–cell interaction networks in MHF show a decrease in the number and strength of communication events. Cardiomyocytes strongly interact with endothelial cells; fibroblasts engage granulocytes. (C, D) Bubble plots of ligand–receptor pairs show CXCL2–CXCR2 as the dominant axis between macrophages and granulocytes in both groups. Despite this overall reduction, cardiomyocytes maintained strong interactions with other cell types in both groups. In the MHF samples, cardiomyocytes showed stronger communication with endothelial cells. Fibroblasts also showed stronger interactions with granulocytes, which may reflect pathological remodeling (Fig. 7A). In contrast, the control group showed more balanced and extensive interactions among cardiomyocytes, fibroblasts, and other cell types (Fig. 7B). Ligand receptor interaction analysis further indicated that several key signaling pairs were reduced in MHF compared with controls (Fig. 7C and D). The strongest interaction in both conditions was observed between macrophages and granulocytes. This interaction was mainly mediated by the Cxcl2–Cxcr2 signaling axis. This persistent signal may contribute to immune homeostasis or inflammation in MHF and warrants further functional validation. It should be noted that the drug–target prediction and molecular docking analyses in this study are exploratory and hypothesis-generating. They are intended to prioritize candidate compounds for subsequent in vitro and in vivo validation. They do not establish therapeutic efficacy or clinical utility. Specifically regarding selinexor and XPO1, we acknowledge that XPO1 was downregulated in our MHF models. The rationale for proposing XPO1 inhibition is based on the hypothesis that pharmacological modulation may restore nucleocytoplasmic transport homeostasis by altering the subcellular localization of specific signaling molecules (e.g., NF-κB, STAT3, YAP), rather than compensating for XPO1 expression levels. This context-dependent therapeutic strategy requires rigorous experimental validation. Pseudo-time trajectory inference in cardiomyocytes and fibroblasts To further explore the dynamic states in key cell types, we performed pseudotime trajectory analysis in cardiomyocytes and fibroblasts. Dimensionality reduction and clustering divided cardiomyocytes into eight subclusters (Fig. 8A) and fibroblasts into four subclusters (Fig. 8B). Trajectory inference identified nine distinct states for cardiomyocytes (Fig. 8C) and five states for fibroblasts (Fig. 8D). These states describe a progression from earlier to later cellular states. The color gradient indicates pseudo-time, with darker colors representing earlier stages. Pseudotime trajectory analysis reveals dynamic, cell-type-specific regulation of core biomarkers during MHF progression. Subclustering and trajectory inference were performed independently for each lineage. (A, B) Subclustering divides cardiomyocytes and fibroblasts into 8 and 4 subgroups, respectively. (C, D) Monocle-based trajectory inference identifies 9 cardiomyocyte and 5 fibroblast developmental states, indicating temporal heterogeneity. (E, F) Expression dynamics along pseudotime demonstrate that PLA2G4A and PRRX1 are progressively upregulated in activated fibroblasts, while SLAIN2 and XPO1 are downregulated in both cardiomyocytes and fibroblasts during disease progression. These patterns reinforce the functional involvement of the core biomarkers in key pathological processes within their respective cellular compartments. Along these trajectories, the five biomarkers showed dynamic and stage-specific expression changes. In cardiomyocytes, PLA2G4A, PRRX1, SLAIN2, and XPO1 showed complex and non-linear patterns across pseudo-time, which suggests associations with specific transitional states. In contrast, CLIP2 remained relatively stable, which may indicate a constitutive role in cardiomyocytes (Fig. 8E). In fibroblasts, all five biomarkers showed pronounced changes across pseudo-time, consistent with roles in fibroblast maturation and function (Fig. 8F). Together, these results suggest that the biomarkers are dynamically regulated during cardiomyocyte and fibroblast state transitions and may contribute to cell type specific processes in MHF. Discussion MHF are key pathological processes that drive adverse cardiac remodeling and are closely linked to poor clinical outcomes, including heart failure and arrhythmias47. These structural abnormalities arise from diverse etiologies, such as chronic pressure overload, ischemic injury, and metabolic dysfunction. Despite substantial progress in understanding MHF pathophysiology, the molecular mechanisms that govern initiation and progression remain incompletely defined. Recent advances in multi-omics technologies, including transcriptomics, proteomics, phosphoproteomics, and scRNA-seq, enable multi-dimensional exploration of disease biology at both bulk and cellular resolution48,49. Using this integrative approach, we identified five candidate molecules, CLIP2, PLA2G4A, PRRX1, SLAIN2, and XPO1. They showed consistent dysregulation across multiple experimental models of MHF. These molecules represent convergent hubs in the MHF regulatory networks. They also warrant further evaluation as potential biomarkers or therapeutic targets in human disease. A range of experimental models has been developed to recapitulate key aspects of MHF, including TAC and neurohormonal stimulation with agents such as angiotensin II or ISO. Each model reflects distinct features of cardiac remodeling50,51. No single model fully captures the complexity of human MHF, but these approaches remain essential for mechanistic studies. In this study, we used in vivo and in vitro ISO induced models to simulate MHF. This design allowed us to examine both myocardial tissue and cardiomyocytes under pathological stress. We performed transcriptomic and phosphoproteomic profiling to capture changes in gene expression and post-translational modifications. We also integrated scRNA-seq data from GSE120064, which is based on aortic banding induced pressure overload remodeling. This integrative strategy improves biological interpretability and supports a more comprehensive view of molecular alterations associated with MHF. While our initial biomarker discovery pipeline combined data from heterogeneous bulk tissue and a homogeneous cell line—a recognized limitation—the subsequent single-cell RNA sequencing analysis provided critical validation. By mapping the expression of CLIP2, PLA2G4A, PRRX1, SLAIN2, and XPO1 onto specific cardiac cell types, we confirmed their enrichment in cardiomyocytes and/or fibroblasts, the very cells orchestrating pathological remodeling. This cell-type-specific validation significantly strengthens the biological plausibility of our findings and mitigates concerns regarding the initial integrative approach. A notable observation in our study is the consistent downregulation of XPO1, a master regulator of nuclear export, in both cardiomyocytes and fibroblasts across multiple MHF models. We explicitly acknowledge that proposing an XPO1 inhibitor (selinexor) for a target identified as downregulated in our data appears paradoxical at first glance. At first glance, this finding appears to conflict with a prior report showing that XPO1 is upregulated in ischemic cardiomyopathy. These studies also reported cardioprotective effects of XPO1 inhibition, including pharmacological inhibition with selinexor and gene silencing via AAV9 shXPO1 delivery in myocardial infarction models. We propose that this discrepancy reflects the context-dependent role of XPO1 across distinct cardiac disease etiologies. Ischemic injury is dominated by acute oxidative stress, inflammation, and cardiomyocyte apoptosis. In this setting, increased XPO1 may enhance nuclear export and cytoplasmic sequestration of cardioprotective transcription factors, such as p53, FOXO1, or IκB, thereby aggravating cell death52. In contrast, the neurohormonal stress model (ISO) and the hemodynamic stress model (TAC) used in our study reflect chronic maladaptive remodeling. This process is characterized by hypertrophy, interstitial fibrosis, and metabolic reprogramming. These features are regulated by signaling axes such as TGF-β and SMAD, calcineurin and NFAT, and β-adrenergic pathways. In this context, reduced XPO1 may not reflect a simple loss-of-function. Instead, it may indicate a maladaptive disruption of nucleocytoplasmic transport homeostasis. This disruption could promote aberrant nuclear retention of profibrotic regulators, such as SMAD2 and SMAD3, or alter the activity of transcriptional cofactors. Importantly, we hypothesize that the therapeutic potential of selinexor does not depend on further inhibiting an already downregulated target. Instead, it may act by pharmacologically resetting a globally dysregulated nuclear export system. By inducing a controlled and transient blockade of XPO1-mediated cargo export, selinexor may change the subcellular localization of specific signaling molecules, such as NF-κB, STAT3, or YAP. This could reshape downstream transcriptional programs involved in fibroblast activation and cardiomyocyte growth53,54. This concept is consistent with systems pharmacology models in which drugs can restore dynamic equilibrium in perturbed networks rather than normalize static expression levels55. Therefore, our proposal to explore selinexor is based on the hypothesis of restoring transport fidelity rather than compensating for XPO1 abundance. Our data suggest that XPO1 dysregulation is a feature of non-ischemic MHF, but its functional role is highly dependent on disease context. The contrasting expression patterns between ischemic and pressure-overload or hormonal models highlight the need for etiology specific targeting strategies. Future studies using conditional manipulation of XPO1 in specific cardiac cell types will be essential to define its causal role. These studies will also help determine whether activation or inhibition is optimal in each setting. Our integrative multi-omics framework combined bulk transcriptomics from an in vivo ISO-induced rat model, phosphoproteomics from stressed H9c2 cardiomyocytes, and orthogonal validation via scRNA-seq from a pressure-overload mouse model (GSE120064). This approach enabled identification of a coordinated core module of five candidate biomarkers, CLIP2, PLA2G4A, PRRX1, SLAIN2, and XPO1. These candidates were not selected based on prior literature. They emerged from a stringent consensus filter. They showed concordant directional changes across transcriptomic and phosphoproteomic layers and were enriched in disease relevant cardiac cell types, particularly cardiomyocytes and activated fibroblasts, as supported by single cell mapping. This coordinated dysregulation suggests a functional axis linking cytoskeletal dynamics, inflammatory signaling, fibroblast activation, and nucleocytoplasmic transport in MHF. Three markers, CLIP2, PLA2G4A, PRRX1, were consistently upregulated, whereas SLAIN2 and XPO1 were downregulated. This pattern suggests a bimodal imbalance. It may reflect hyperactivation of pro-fibrotic and stress-response programs together with reduced homeostatic regulation. The convergence of transcriptional and post-translational changes in the same molecules highlights the valuer of integrated analyses for identifying high confidence candidates that may be missed by single modality approaches. For example, XPO1 has been reported with variable patterns across cardiac conditions. Our data integrate bulk omics, cell-type-specific expression, and pseudotime trajectory analysis. Together, they indicate progressive and coordinated downregulation of XPO1 in both cardiomyocytes and fibroblasts during pathological remodeling. This consistency across models and modalities supports a context-dependent role in maladaptive MHF that differs from its reported upregulation in ischemic injury52. Similarly, PRRX1 was enriched in activated fibroblasts and was upregulated in our mammalian model. This contrasts with zebrafish regeneration studies56 and supports a pathological, rather than reparative function in chronic MHF. These examples illustrate how cross-model and multi-layer analyses can help resolve apparent inconsistencies in the literature by linking observations to specific disease contexts and cellular compartments. Functional and pathway enrichment analyzes further supported the biological relevance of the identified biomarkers. Functional similarity scoring highlighted CLIP2 as a central node, which suggests a key role in regulating cytoskeletal organization during cardiac remodeling. GeneMANIA network analysis showed that these biomarkers were enriched in processes related to microtubule end organization and regulation of microtubule polymerization, which are essential for preserving cardiomyocyte structural integrity and mechanotransduction57. GSEA demonstrated significant enrichment of the biomarkers in pathways including the ribosome, aminoacyl-tRNA biosynthesis, and the proteasome. Disruption of these pathways compromises protein synthesis fidelity and proteostasis. It can also increase endoplasmic reticulum (ER) stress and active maladaptive stress responses58,59. This dysfunction may promote myocardial hypertrophy, fibroblast activation, and ECM accumulation. Collectively, these findings suggest that alterations in core biosynthetic and degradative processes are closely linked to the pathogenesis and progression of MHF. In addition to molecular characterization, regulatory network analysis provided insight into upstream control mechanisms for the identified biomarkers. TF analysis using the TRRUST database identified 56 TFs, with NFKB1 and RFX1 emerging as central hubs. NFKB1 is a well-established master regulator of inflammation and fibrosis and plays a pivotal role in cardiac remodeling60. RFX1 is implicated in immune homeostasis and transcriptional regulation relevant to fibrotic responses61. At the post-transcriptional level, 141 miRNAs were predicted to target the biomarkers, including hsa-miR-485-5p, which has been previously associated with the progression of cardiomyopathy62. Integration of miRNA-lncRNA interactions from the miRNet and StarBase databases identified 297 common lncRNAs. Notably, lncRNAs such as XIST and NEAT1, which are recognized drivers of myocardial fibrosis and hypertrophy, were prominent components of the constructed lncRNA-miRNA-mRNA regulatory network63. These findings suggest that the expression and function of the biomarkers are orchestrated by a coordinated and multilayered regulatory framework. This framework spans transcriptional and post transcriptional levels and provides multiple potential nodes for therapeutic intervention. Drug prediction combined with molecular docking provided preliminary insights into potential therapeutic strategies for MHF. Among the identified candidates, flurandrenolide, a synthetic corticosteroid, and selinexor, a selective inhibitor of nuclear export (SINE), showed high predicted binding affinities with PLA2G4A and XPO1, respectively. Flurandrenolide is known for anti-inflammatory and anti-fibrotic effects, largely mediated by glucocorticoid receptor signaling. It can suppress pro-inflammatory cytokines, ECM production, and fibroblast activation, thereby attenuating fibrotic remodeling64,65. Selinexor has shown anti-proliferative and anti-fibrotic effects in oncological and fibrotic disease models by promoting nuclear retention of tumor suppressors and dampening inflammatory signaling pathways66,67. Our docking analysis suggests a potential interaction of selinexor as a tool compound to probe the role of nuclear export in maladaptive remodeling. However, both compounds have been associated with cardiovascular adverse effects, including hypertension, arrhythmias, and fluid retention, particularly in oncologic clinical settings68,69. This raises critical concerns about their cardiac safety profiles, especially in patients with pre-existing cardiovascular dysfunction. While these results nominate flurandrenolide and selinexor as computational hits for preclinical validation, their suitability for therapeutic development in MHF remains to be determined. Particularly for selinexor, future experimental studies must directly address the apparent paradox of targeting a downregulated protein to determine whether XPO1 modulation (inhibition vs. activation) yields beneficial outcomes in non-ischemic MHF contexts. Future studies should evaluate whether modified derivatives or targeted delivery strategies could mitigate off-target cardiotoxicity while preserving efficacy. Complementing the molecular findings, our scRNA-seq analysis provided critical insights into the cellular landscape and dynamic remodeling processes underlying MHF. Six major cardiac cell types were identified, and cardiomyocytes and fibroblasts emerged as predominant contributors to pathological remodeling. Cardiomyocytes showed transcriptomic features consistent with mitochondrial dysfunction, oxidative stress, and cytoskeletal disorganization, which align with a hypertrophic and maladaptive phenotype. Cardiomyocytes can also secrete pro inflammatory cytokines, including interleukin 6 (IL 6), which may promote fibroblast proliferation and amplify inflammatory signaling. In addition, hypertrophic cardiomyocytes can release TGF-β, a key driver of fibroblast to myofibroblast transition and ECM deposition70. Fibroblasts are the main source of collagen and other ECM proteins. Under pathological conditions, they can exhibit β catenin pathway activation, which amplifies profibrotic gene programs and may also promote cardiomyocyte hypertrophy through paracrine signaling71,72. Pseudotime trajectory analysis further revealed a coordinated progression of cardiomyocytes toward a hypertrophic fate and fibroblasts toward activated myofibroblast phenotypes. These transitions suggest dynamic interdependence between cellular states and may contribute to progressive remodeling and heart failure73. Among the five biomarkers, PLA2G4A showed elevated expression in fibroblasts in MHF, particularly in fibrotic subpopulations. Pseudo-time analysis indicated a continuous increase in PLA2G4A expression along the fibroblast trajectory, which supports a role in myofibroblast activation and ECM deposition. In contrast, XPO1 was consistently downregulated in both cardiomyocytes and fibroblasts, with progressive decreases along disease associated trajectories, which suggests impaired nuclear-cytoplasmic transport during MHF progression. These single-cell patterns validate our bulk transcriptomic and phosphoproteomic results. They also support PLA2G4A as a profibrotic driver and XPO1 as a regulator whose dysfunction may exacerbate maladaptive remodeling. We also observed strong inferred interactions between cardiomyocytes and endothelial cells in MHF tissues, which suggests active crosstalk during disease progression. Cardiomyocyte-derived VEGF-A is important for maintaining coronary microvascular integrity, and its deficiency can lead to microvascular rarefaction and cardiac dysfunction74. In addition, endothelial VEGFR2 and cardiomyocyte ErbB signaling may help coordinate hypertrophy with angiogenesis to preserve myocardial structure and function75. Together, these findings highlight the potential importance of the cardiomyocyte–fibroblast–endothelial network in pathological remodeling. We acknowledge that our integrative approach combines data from models that differ in species (rat vs. mouse), pathological drivers (acute β-adrenergic overstimulation vs. chronic pressure overload), and biological complexity (in vivo vs. in vitro). These differences inherently limit direct comparability at the level of individual gene expression. In particular, the integration of bulk heart tissue transcriptomics (containing mixed cell populations) with H9c2 cardiomyoblast phosphoproteomics represents a methodological limitation, as the cellular contexts differ significantly. However, rather than aiming to reproduce identical transcriptional profiles, our strategy focused on identifying recurrently altered pathways and hub molecules that emerge despite divergent etiologies. The rationale is that molecules consistently perturbed across such heterogeneous contexts may represent fundamental effectors of the MHF phenotype, rather than model-specific stress artifacts. To address concerns regarding the potential identification of generalized stress-response genes, we performed additional stress response gene association analysis using the GOBP_INTEGRATED_STRESS_RESPONSE_SIGNALING gene set from MSigDB. Our results demonstrated that while some biomarkers showed moderate correlations with stress response genes, the overall signature was not solely driven by generalized cellular stress, partially mitigating this concern. To further address the concern regarding cellular context differences, we performed cell-type-specific expression validation using scRNA-seq data from GSE120064, which confirmed that three of the five biomarkers (XPO1, PRRX1, and SLAIN2) exhibited significant concordant dysregulation in purified cardiomyocytes (all p < 0.001). Notably, PLA2G4A did not show significant transcriptional changes in cardiomyocytes (p = 0.16), but its phosphorylation alterations were supported by PhosphoSitePlus database evidence75 and literature from myocardial ischemia models, suggesting post-translational regulation. Furthermore, database queries and literature evidence confirmed that PLA2G4A undergoes phosphorylation modifications in cardiac tissues, with TNF-α signaling identified as an upstream regulator in myocardial ischemia models. Nevertheless, we recognize that some of the identified signatures may still reflect generalized cellular stress responses or cell-type composition differences76. Future validation in human cardiac tissues and functional studies in genetically engineered models will be essential to confirm the specificity and causality of these candidates in human-relevant MHF pathogenesis. In conclusion, this study integrated transcriptomic, proteomic, phosphoproteomic, and scRNA-seq analyses and identified CLIP2, PLA2G4A, PRRX1, SLAIN2, and XPO1 as candidate biomarkers associated with MHF. These genes are linked to dysregulated protein synthesis, metabolism, cytoskeletal integrity, and nuclear transport and may contribute maladaptive remodeling. Multi-layered regulatory networks involving TFs, miRNAs, and lncRNAs further suggested potential upstream regulators of these candidates. Future work should prioritize functional validation, mechanistic studies, and clinical translation to evaluate their therapeutic potential. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Supplementary Material 4. Supplementary Material 5. Supplementary Material 6. Supplementary Material 7. Supplementary Material 8. Supplementary Material 9. Supplementary Material 10. Supplementary Material 11. Supplementary Fig. 1. Principal component analysis (PCA) of transcriptome and phosphoproteome data in MHF and control samples. PCA analyzes assess inter-group variability and sample clustering consistency, confirming the reliability of omics data for downstream comparisons. (A) PCA of pre-processed transcriptome sequencing data illustrates the distribution and variance among MHF and control samples. (B) PCA of pre-processed label-free quantitative (LFQ) phosphoproteome data reveals similar inter-group variance and sample clustering. In both omics layers, PCA demonstrated clear group separation while reflecting the heterogeneity within each group, thereby supporting the discriminatory power and data robustness for downstream analyzes. Supplementary Material 12. Supplementary Fig. 2. Spearman correlation analysis between five candidate biomarkers and stress response genes. Lollipop plot showing the Spearman correlation coefficients between each biomarker (CLIP2, PLA2G4A, PRRX1, SLAIN2, and XPO1) and the GOBP_INTEGRATED_STRESS_RESPONSE_SIGNALING gene set from MSigDB. Each lollipop represents the correlation coefficient (r value) between a biomarker and a stress response gene. Asterisks indicate statistical significance (*p < 0.05, **p < 0.01). The x-axis represents the Spearman correlation coefficient ranging from − 1.0 to 1.0. Supplementary Material 13. Supplementary Fig. 3. Functional annotation and interaction landscape of five core biomarkers. Functional characterization and molecular context of the identified biomarkers. (A) Functional similarity analysis based on GO terms ranks CLIP2 highest among the five biomarkers. (B) Chromosomal localization shows CLIP2 (Chr7), PLA2G4A and PRRX1 (Chr1), SLAIN2 (Chr4), and XPO1 (Chr2). (C) Gene–gene interaction (GGI) network constructed via GeneMANIA identifies 20 functionally linked genes associated with microtubule dynamics. (D) Subcellular localization analysis indicates that all five biomarkers are predominantly expressed in the cytoplasm. Supplementary Material 14. Supplementary Fig. 4. Quality control and identification of highly variable genes (HVGs) in scRNA-seq data from GSE120064. The summarizing of data filtering and feature selection steps ensures robust single-cell transcriptomic analysis. (A, B) Quality control plots showing the distribution of total detected genes per cell (nFeature RNA), total transcript counts per cell (nCount RNA), and mitochondrial gene content (% mitochondrial) before (A) and after (B) quality filtering. These metrics were used to eliminate low-quality or stressed cells. (C) Selection of 2,000 highly variable genes (HVGs) after data normalization. The top 10 most variable genes are highlighted, serving as input features for downstream dimensionality reduction and clustering. Supplementary Material 15. Supplementary Fig. 5. Validation of in vitro and in vivo MHF models. The validation of the pathological features of isoproterenol-induced myocardial hypertrophy and fibrosis in both cellular and animal models with 3 replicas. (A) Rhodamine-phalloidin staining of H9c2 cells treated with PBS or isoproterenol (ISO, 25 µM, 48 h), showing increased cell size and cytoskeletal reorganization in the ISO group. (B) Gross images of rat hearts after 7-day ISO (1.5 mg/kg/day) or PBS treatment, with visible cardiac enlargement in the ISO group. (C) H&E staining of left ventricular tissue, revealing cardiomyocyte hypertrophy and fibrotic changes in ISO-treated rats. (D) qPCR analysis of ANP and BNP mRNA levels in H9c2 cells treated with PBS or isoproterenol (ISO, 25 µM, 48 h). Data are presented as mean ± SD (n = 3 independent experiments). *p < 0.05, **p < 0.01 vs. control group by two-tailed unpaired t-test. (E) qPCR quantification of ANP and BNP mRNA in myocardial tissues of rats after 7-day ISO (1.5 mg/kg/day) or PBS treatment. Data are presented as mean ± SD (n = 3 animals per group). *p < 0.05, **p < 0.01 vs. control group by two-tailed unpaired t-test. Supplementary Material 16. Supplementary Fig. 6. Cell-type-specific expression validation of core biomarkers in cardiomyocytes. Violin plots showing log-normalized expression levels of five hub genes (XPO1, PRRX1, SLAIN2, CLIP2, and PLA2G4A) in cardiomyocytes from GSE120064 between Control and MHF groups. Each violin plot represents the distribution of expression values across individual cells, with the inset box plot showing the median and interquartile range. Wilcoxon rank-sum test was used for statistical comparison. Three biomarkers showed significant differential expression in cardiomyocytes: XPO1, PRRX1, and SLAIN2 were all significantly downregulated in MHF (p = 8.0 × 10− 19, p = 1.7 × 10− 6 and p = 5.9 × 10− 14, respectively). CLIP2 (p = 0.75) and PLA2G4A (p = 0.16) did not reach statistical significance (ns). Asterisks indicate statistical significance (***p < 0.001, ns = not significant). Acknowledgements Not applicable. Author contributions B.Z., L.L., and S.H. conceived the project and supervised the research. K.L. and Y.M. contributed to the study design, performed the principal experiments, and analyzed the data. Y. Zhang, Y. Zi, G.G., X.C., W.F., and Q.J. assisted with data collection, additional experiments, and interpretation of results. K.L. and Y.M. drafted the manuscript. All authors provided critical revision of the manuscript for important intellectual content, approved the final version for submission, and agree to be accountable for all aspects of the work. Funding This work was supported by the Chinese National Natural Science Foundation [grant number 82260063 to Shen Han]; and the Science and Technology department of Yunnan Province [grant number 202201AT070282 to Shen Han, 202401AY070001-190 to Yayong Zhang]. Data availability The transcriptomic dataset generated during the current study has been deposited in the NCBI Gene Expression Omnibus (GEO) repository under accession number GSE326552. The mass spectrometry proteomics and phosphoproteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifiers PXD076291 and PXD076293, respectively. In addition, the publicly available single-cell RNA sequencing dataset reanalyzed in this study is available in GEO under accession number GSE120064. Declarations Competing interests Q.J. is an employee of Nanjing Geneseeq Technology Inc. The remaining authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Kang Luo and Yangguang Ma contributed equally to this work. Contributor Information Lin Li, Email: lilin@kmmu.edu.cn. Shen Han, Email: hslsj910@163.com. Bingrong Zheng, Email: zhengbr@ynu.edu.cn. References Associated Data Supplementary Materials Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Supplementary Material 4. Supplementary Material 5. Supplementary Material 6. Supplementary Material 7. Supplementary Material 8. Supplementary Material 9. Supplementary Material 10. Supplementary Material 11. Supplementary Fig. 1. Principal component analysis (PCA) of transcriptome and phosphoproteome data in MHF and control samples. PCA analyzes assess inter-group variability and sample clustering consistency, confirming the reliability of omics data for downstream comparisons. (A) PCA of pre-processed transcriptome sequencing data illustrates the distribution and variance among MHF and control samples. (B) PCA of pre-processed label-free quantitative (LFQ) phosphoproteome data reveals similar inter-group variance and sample clustering. In both omics layers, PCA demonstrated clear group separation while reflecting the heterogeneity within each group, thereby supporting the discriminatory power and data robustness for downstream analyzes. Supplementary Material 12. Supplementary Fig. 2. Spearman correlation analysis between five candidate biomarkers and stress response genes. Lollipop plot showing the Spearman correlation coefficients between each biomarker (CLIP2, PLA2G4A, PRRX1, SLAIN2, and XPO1) and the GOBP_INTEGRATED_STRESS_RESPONSE_SIGNALING gene set from MSigDB. Each lollipop represents the correlation coefficient (r value) between a biomarker and a stress response gene. Asterisks indicate statistical significance (*p < 0.05, **p < 0.01). The x-axis represents the Spearman correlation coefficient ranging from − 1.0 to 1.0. Supplementary Material 13. Supplementary Fig. 3. Functional annotation and interaction landscape of five core biomarkers. Functional characterization and molecular context of the identified biomarkers. (A) Functional similarity analysis based on GO terms ranks CLIP2 highest among the five biomarkers. (B) Chromosomal localization shows CLIP2 (Chr7), PLA2G4A and PRRX1 (Chr1), SLAIN2 (Chr4), and XPO1 (Chr2). (C) Gene–gene interaction (GGI) network constructed via GeneMANIA identifies 20 functionally linked genes associated with microtubule dynamics. (D) Subcellular localization analysis indicates that all five biomarkers are predominantly expressed in the cytoplasm. Supplementary Material 14. Supplementary Fig. 4. Quality control and identification of highly variable genes (HVGs) in scRNA-seq data from GSE120064. The summarizing of data filtering and feature selection steps ensures robust single-cell transcriptomic analysis. (A, B) Quality control plots showing the distribution of total detected genes per cell (nFeature RNA), total transcript counts per cell (nCount RNA), and mitochondrial gene content (% mitochondrial) before (A) and after (B) quality filtering. These metrics were used to eliminate low-quality or stressed cells. (C) Selection of 2,000 highly variable genes (HVGs) after data normalization. The top 10 most variable genes are highlighted, serving as input features for downstream dimensionality reduction and clustering. Supplementary Material 15. Supplementary Fig. 5. Validation of in vitro and in vivo MHF models. The validation of the pathological features of isoproterenol-induced myocardial hypertrophy and fibrosis in both cellular and animal models with 3 replicas. (A) Rhodamine-phalloidin staining of H9c2 cells treated with PBS or isoproterenol (ISO, 25 µM, 48 h), showing increased cell size and cytoskeletal reorganization in the ISO group. (B) Gross images of rat hearts after 7-day ISO (1.5 mg/kg/day) or PBS treatment, with visible cardiac enlargement in the ISO group. (C) H&E staining of left ventricular tissue, revealing cardiomyocyte hypertrophy and fibrotic changes in ISO-treated rats. (D) qPCR analysis of ANP and BNP mRNA levels in H9c2 cells treated with PBS or isoproterenol (ISO, 25 µM, 48 h). Data are presented as mean ± SD (n = 3 independent experiments). *p < 0.05, **p < 0.01 vs. control group by two-tailed unpaired t-test. (E) qPCR quantification of ANP and BNP mRNA in myocardial tissues of rats after 7-day ISO (1.5 mg/kg/day) or PBS treatment. Data are presented as mean ± SD (n = 3 animals per group). *p < 0.05, **p < 0.01 vs. control group by two-tailed unpaired t-test. Supplementary Material 16. Supplementary Fig. 6. Cell-type-specific expression validation of core biomarkers in cardiomyocytes. Violin plots showing log-normalized expression levels of five hub genes (XPO1, PRRX1, SLAIN2, CLIP2, and PLA2G4A) in cardiomyocytes from GSE120064 between Control and MHF groups. Each violin plot represents the distribution of expression values across individual cells, with the inset box plot showing the median and interquartile range. Wilcoxon rank-sum test was used for statistical comparison. Three biomarkers showed significant differential expression in cardiomyocytes: XPO1, PRRX1, and SLAIN2 were all significantly downregulated in MHF (p = 8.0 × 10− 19, p = 1.7 × 10− 6 and p = 5.9 × 10− 14, respectively). CLIP2 (p = 0.75) and PLA2G4A (p = 0.16) did not reach statistical significance (ns). Asterisks indicate statistical significance (***p < 0.001, ns = not significant). Data Availability Statement The transcriptomic dataset generated during the current study has been deposited in the NCBI Gene Expression Omnibus (GEO) repository under accession number GSE326552. The mass spectrometry proteomics and phosphoproteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifiers PXD076291 and PXD076293, respectively. In addition, the publicly available single-cell RNA sequencing dataset reanalyzed in this study is available in GEO under accession number GSE120064.