The solute carrier transporter SLCO5A1 as a novel therapeutic target for the treatment of myocardial infarction-induced heart failure

Abstract

Heart failure (HF) following myocardial infarction (MI) remains a major threat to health worldwide. While transcriptomics has revealed numerous genes whose expression is altered in HF, distinguishing therapeutic targets remains challenging. In this study, we aimed to identify novel therapeutic targets for HF and explore potential pharmacological interventions. We integrated human HF datasets with weighted gene coexpression network analysis (WGCNA) and machine learning (LASSO/SVM-RFE) to screen for candidate genes and applied Mendelian randomization (MR) to assess causality. SLCO5A1 emerged as a prioritized candidate, as it showed a genetically supported protective association with HF and was consistently downregulated in the ischemic failing myocardium. In mice, cardiomyocyte-targeted SLCO5A1 overexpression attenuated post-MI systolic dysfunction and pathological remodeling. Using drug–gene signature mining followed by biophysical and cellular validation, we identified 3-iodothyronamine (T1AM) as a small molecule that directly binds to SLCO5A1 and increases SLCO5A1 protein levels. Pharmacological administration of T1AM increased post-MI survival, improved cardiac function and reduced fibrosis; these benefits were markedly weakened by cardiomyocyte-specific SLCO5A1 knockdown, supporting a functional requirement for SLCO5A1. Mechanistically, SLCO5A1 reduced cardiomyocyte transforming growth factor beta 1 (TGF‑β1) secretion, thereby limiting fibroblast Smad3 activation and myofibroblast marker expression in conditioned-medium assays. In conclusion, our findings demonstrate that SLCO5A1 is a cardioprotective regulator of cardiomyocyte–fibroblast communication in post-MI HF and support a pharmacological increase in SLCO5A1 levels as a potential therapeutic strategy.

SLCO5A1 serves as a novel therapeutic target for heart failure. Enhancing SLCO5A1 expression protects against MI-induced HF by inhibiting TGF-β1/Smad3-mediated cardiomyocyte–fibroblast crosstalk.

Introduction

Heart failure (HF) affects more than 64 million people worldwide and places a high burden on countries [1]. Although guideline-directed therapies have improved outcomes, postinfarction remodeling and progression to HF remain prevalent [1], suggesting that key endogenous protective pathways within the myocardium are incompletely understood. Transcriptomic profiling has identified numerous genes associated with HF; however, the identification of causal factors and potential therapeutic targets remains a major challenge.

Solute carrier (SLC) transporters regulate the cellular bioavailability of hormones, metabolites, and signaling molecules essential for cardiac function [2, 3]. Among SLC transporters, organic anion-transporting polypeptides (OATPs, encoded by SLCO genes) mediate the sodium-independent transmembrane transport of amphipathic substrates and have been widely studied in drug disposition [4, 5]. However, the functions of OATPs in the heart are poorly defined, and whether the dysregulation of cardiac OATPs contributes to pathological remodeling remains largely unknown.

SLCO5A1, which encodes OATP5A1, is relatively enriched in the human myocardium. SLCO5A1 has been classified as an “orphan” transporter for two decades because it does not transport classic OATP substrates, such as estrone-3-sulfate, taurocholate, or bromosulfophthalein, in heterologous expression systems [6–8]. Although their endogenous substrates, physiological functions, and relevance to cardiovascular diseases remain unknown, membrane transporters can intrinsically regulate transmembrane flux and the intracellular availability of bioactive molecules. This regulatory function can influence the development of myocardial hypertrophy, fibrosis, and the progression of HF [5]. Furthermore, in addition to canonical substrate transport, membrane transporters are subject to regulation through protein stability and turnover. Ligand-dependent stabilization, also known as pharmacological chaperoning, has been identified as a therapeutically viable mechanism for several membrane proteins [9]. However, whether OATP5A1 is regulated by such mechanisms in cardiomyocytes and whether this regulation has implications for HF remain to be determined.

Thyroid hormone signaling is essential for normal cardiac metabolism, contractility, and energetics [10, 11]. In addition to the canonical thyroid hormones thyroxine (T4) and triiodothyronine (T3), decarboxylated thyroid hormone derivatives, such as thyronamines, have emerged as bioactive mediators with distinct cardiovascular activities [12]. 3-Iodothyronamine (T1AM), an endogenous thyroid hormone derivative, has been detected in blood and cardiac tissues [13]. T1AM has been reported to regulate cardiac metabolism, mitochondrial function, and contractile responses and to protect against myocardial ischemic injury [14, 15]. However, its pathophysiological significance in HF and the molecular mediators of its cardiac effects remain incompletely understood.

In this study, we investigated whether OATP5A1 is a causal protective factor in HF and whether it can be pharmacologically enhanced to alleviate post-MI remodeling. We first integrated multiple human ischemic cardiomyopathy transcriptomic datasets with network analysis and machine-learning prioritization and then applied Mendelian randomization to identify genes with genetic support for causal association. We subsequently validated SLCO5A1 in a mouse model of MI-induced HF and cardiomyocyte stress. Finally, we identified T1AM as a direct SLCO5A1-binding small molecule that increases SLCO5A1 protein levels and evaluated whether T1AM–SLCO5A1 binding limits pathological cardiomyocyte–fibroblast signaling through the TGF‑β1/Smad3 axis.

Materials and methods

Public datasets

The HF gene expression datasets GSE55296 and GSE57345 were obtained from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo). GSE55296, which is based on the GPL16288 platform, comprises left ventricular tissue samples from 13 patients with ischemic cardiomyopathy and 10 samples from healthy individuals. GSE57345, which is based on the GPL9052 platform, comprises samples from 95 patients with ischemic cardiomyopathy and 139 controls. The expression matrices were subjected to log2 transformation and batch effect correction using the Combat algorithm from the sva package [16]. Principal component analysis (PCA) was conducted [17], and the results revealed low batch variation and clear separation between the HF and control groups.

Identification of differentially expressed genes (DEGs)

Batch effects between datasets were eliminated using the sva package (v3.50.0) in R. Differential expression analysis was performed with the limma package (v3.60.4). DEGs were identified on the basis of an adjusted P value < 0.05 and |log2FC | > 0.585. The ggplot2 (v3.5.1) and pheatmap (v1.0.12) packages were used to visualize the results via volcano plots and heatmaps. Functional enrichment analyses, including Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses, were conducted using the clusterProfiler package (v4.10.1). GO analysis examines the functions of gene products across biological processes (BPs), cellular components (CCs), and molecular functions (MFs). KEGG analysis reveals the involvement of DEGs in diverse signaling pathways. Terms with an FDR < 0.05 were considered significant, and the top 12 terms were visualized using OmicStudio [18].

Weighted Gene Coexpression Network Analysis (WGCNA)

We employed the WGCNA package (v1.73) to investigate the relationships between gene expression patterns and traits. For network construction, we retained the top 25% of genes that exhibited the greatest variation in expression. A soft-thresholding power of 11 was determined to be optimal, resulting in a scale-free network with an R² value ≥ 0.8. We subsequently constructed an unsigned weighted adjacency matrix, which was subsequently transformed into a topological overlap matrix (TOM) to assess network connectivity. Modules were identified through hierarchical clustering using a TOM-based dissimilarity measure (1-TOM) with a minimum module size of 60 genes. This process yielded five distinct gene modules, each of which was assigned to a unique color.

Candidate gene screening via machine learning

We employed a machine learning (ML)-based feature selection strategy for 133 overlapping genes to identify diagnostic signature genes for HF. These genes were individually evaluated using the least absolute shrinkage and selection operator (LASSO) and support vector machine-recursive feature elimination (SVM-RFE) algorithms, yielding seven candidate feature genes. Model performance was assessed using the area under the curve (AUC) of the receiver operating characteristic curve.

Nomogram model

A nomogram in which a score was assigned to each feature gene was developed using the “rms” and “rmsa” R packages. This cumulative score was subsequently used to assess the likelihood of an individual having the disease. Additionally, the concordance between the predicted probabilities and actual outcomes was evaluated using calibration curves, and the model was validated using the batch-corrected merged dataset.

Mendelian randomization (MR) analysis

Genetic information for the six identified signature genes was obtained from a genome-wide association study (GWAS) of 4907 aptamer-measured plasma protein levels (pQTLs) in 35,559 Icelandic plasma samples [19]. For the HF association data, we used GWAS summary statistics from the FinnGen database (https://www.finngen.fi/en), which included 37,653 HF cases and 462,695 controls of European ancestry. For the MR analysis, we selected genetic instrumental variables (IVs) using stringent criteria. Single-nucleotide polymorphisms (SNPs) significantly associated with pQTLs were identified using a genome-wide significance threshold of P < 5 × 10⁻⁸. We performed linkage disequilibrium (LD) pruning with thresholds of r² < 0.1 and a physical distance > 10,000 kb and selected IVs with an F-statistic > 10. Two-sample MR analysis was conducted using the TwoSampleMR package (v0.6.8) with the random-effects inverse variance weighted (IVW) method as the primary analysis. Causal effects are expressed as odds ratios (ORs) with 95% confidence intervals (CIs). Heterogeneity among the instrumental variables was evaluated using Cochran’s Q test, with P > 0.05 indicating no significant heterogeneity. Horizontal pleiotropy was assessed using the MR‒Egger regression intercept test and the MR-PRESSO global test, with P > 0.05 suggesting no significant pleiotropy. A leave-one-out analysis was conducted to evaluate the influence of individual SNPs on causal estimates.

Drug identification and molecular docking

The DSigDB database (https://dsigdb.tanlab.org/DSigDBv1.0/geneSearch.html) was used to identify drugs/compounds associated with the identified proteins. The selected drugs/compounds were subsequently subjected to molecular docking with the protein. Drug/compound and protein structures were retrieved from PubChem (https://pubchem.ncbi.nlm.nih.gov/) and the RCSB PDB database (https://www.rcsb.org/), respectively. The protein receptor structure was prepared by removing heteroatoms and water molecules using PyMOL (version 2.5). Molecular docking was performed using AutoDock Vina (version 1.5.6) with the default parameters. The structures were visualized using PyMOL.

Human samples

A cohort of 12 participants from the Second Affiliated Hospital of Guangdong Medical University, China, consisting of eight individuals diagnosed with HF and four non-HF controls, was recruited for this study (YJKT2022-003-2). Blood samples were collected from all participants. The inclusion criteria for patients with HF were as follows: (1) a diagnosis of heart failure in accordance with the Chinese Guidelines for the Diagnosis and Treatment of Heart Failure 2024 [20], (2) classification within the New York Heart Association (NYHA) functional classes II–IV, (3) good compliance and a normal mental status, and (4) normal coagulation function. The exclusion criteria were as follows: (1) incomplete clinical data; (2) severe hepatic or renal dysfunction, as evidenced by persistent symptoms of liver disease; (3) severe anemia; (4) pregnancy or lactation; (5) a diagnosis of malignant tumors; (6) severe infectious diseases; (7) immunodeficiency; (8) rheumatic or autoimmune disease; (9) congenital heart disease; and (10) cardiogenic shock. Individuals with normal ECG results and no history of cardiovascular disease were included in this study as non-HF controls. All participants provided written informed consent after the study protocols were explained to them.

Animals

Male C57BL/6 mice (6–8 weeks old, 20–25 g) were purchased from Guangzhou Tengke Biomedical Technology Co., Ltd. (Guangzhou, China). The animals were housed under specific pathogen-free (SPF) conditions in plastic cages with shaved wood bedding at 25 ± 2 °C on a 12 h light/dark cycle with free access to food and water. All procedures were approved by the Animal Care and Use Committee of Guangdong Medical University (Approval No. GDMU-2025-000166) and followed the ARRIVE 2.0 guidelines. The mice were randomly assigned to groups using computer-generated randomization. The investigators who conducted the outcome assessment and data analysis were blinded to the group assignments.

AMI mouse model

A mouse AMI model was established as previously described [21]. Briefly, the mice were anesthetized with 3% sevoflurane and maintained under controlled ventilation (stroke volume: 200 µL; respiratory rate: 150 breaths/min) with 2% sevoflurane. The left anterior descending (LAD) coronary artery was ligated approximately 2 mm below the left atrial appendage using an 8–0 suture, after which the chest wall and skin were closed in layers. Heart tissues were collected 28 days post-AMI. The animals were humanely euthanized by CO₂ asphyxiation when significant distress was observed.

Echocardiography

Cardiac structure and function in mice were evaluated via echocardiography four weeks after sham or myocardial infarction (MI) surgery utilizing a high-resolution small-animal ultrasound system (Vevo 3100 LT, VisualSonics, Toronto, Canada) equipped with a 30-MHz transducer. The mice were anesthetized with 2% isoflurane and positioned in the supine position on an imaging platform. B/M-mode imaging was performed in the parasternal long- and short-axis views. M-mode tracings of the left ventricle (LV) at the level of the papillary muscle were used to determine the LV end-diastolic diameter (LVEDd) and LV end-systolic diameter (LVESd), from which the LV fractional shortening (FS) and ejection fraction (EF) were calculated.

T1AM administration

T1AM was purchased from Abmole BioScience (Cat# M8734, Houston, USA) and dissolved in normal saline (NS) to make a stock solution of 10 mg/mL for in vivo experiments or in PBS (stock solution at 6 mM) for in vitro experiments. Previous studies have reported the safety and dose of T1AM administered to cardiomyocytes and mice with myocardial injury [22–24]. Accordingly, for the animal experiments, T1AM was administered at low (L), medium (M), and high (H) doses of 12.5 (L), 25 (M), and 50 (H) mg·kg⁻¹ (ip), respectively, every three days for 28 days, with an equivalent amount of NS administered as the control. For cell treatments, T1AM at concentrations of 1.5 μM (L), 3 μM (M), and 6 μM (H) were used.

Adeno-associated viral (AAV) vector preparation and injection

Adeno-associated virus serotype 9 (AAV9) vectors were used as gene delivery tools to manipulate target gene expression. The following AAV9 vectors were constructed: AAV9-cardiac troponin T (cTnT)-Control (vector) and AAV9-cTNT-SLCO5A1 (OE) for SLCO5A1 overexpression. AAV9-cTNT-miR30-shRNA_SLCO5A1 (sh) was used to knockdown endogenous SLCO5A1, and AAV9-cTNT-miR30-shRNA_scramble was used as a control (sequence information in Table 1). AAV9 vectors were manufactured by PackGene Biotech (China). Six-week-old male mice were injected with AAV9 vectors at a density of 1 × 10¹¹ viral genomes/mouse via intravenous tail injection. The efficiency of SLCO5A1 overexpression or knockdown was verified by immunoblot analysis of tissues two weeks after injection.

Table 1.

Sequences of shRNAs used in this study.

shRNA	Target Sequences (5’-3’)
scrambled ShRNA (Vivo)	ACTACCGTTGTTATAGGTG
shSLCO5A1 (Vivo)	GCTGTCATCTTCATCAACAAA
scrambled shRNA (Vitro)	AAGGTTAGTCGCCCTCGCT
shSLCO5A1 (Vitro)	TTTGTTGATGAAGATGACAG

Isolation of neonatal mouse cardiomyocytes (NMCMs) and cardiac fibroblasts (NMCFs) and cell treatments

NMCMs and NMCFs were isolated from neonatal C57BL/6 mouse pups (Day 1 postbirth) using a previously established protocol [25]. Briefly, heart tissues were digested with 0.125% trypsin and 0.5 mg/mL collagenase II (Cat# C2-Bio C; Sigma‒Aldrich, USA). The cell suspension was plated on uncoated culture dishes and incubated for 90 min to allow differential adhesion. Cardiomyocyte-enriched supernatants were replated on 0.1% gelatin-coated dishes in DMEM/F12 (Cat# 11320033; Thermo Fisher Scientific, Pittsburgh, PA, USA) supplemented with 10% fetal bovine serum (FBS; Cat# 26140079; Thermo Fisher Scientific, Pittsburgh, PA, USA). To obtain normal NMCFs, adherent fibroblasts were cultured in DMEM (Cat# 11965092; Thermo Fisher Scientific, Pittsburgh, PA, USA) supplemented with 10% FBS and passaged twice before further experiments were performed.

For cardiomyocyte hypertrophy experiments, isoprenaline and phenylephrine were used in vitro to mimic the neurohumoral stress characteristic of post-MI heart failure, a condition associated with early sympathetic activation and adverse remodeling [26, 27]. NMCMs were pretreated with T1AM or a plasmid (pcDNA3.1) to mediate the overexpression/knockdown of SLCO5A1, followed by isoprenaline (ISO, 10 μM; Cat# HY-B0468; MedChemExpress, Monmouth Junction, NJ, USA) and phenylephrine (PE, 100 μM; Cat# HY-B0769; MedChemExpress, Monmouth Junction, NJ, USA) treatment for 36 h. For fibrosis experiments, cardiac fibroblasts were cultured for 48 h in a 1:1 mixture of hypertrophied cardiomyocyte medium and complete DMEM-F12 medium supplemented with 10% FBS before analysis.

The DNA sequence of the mouse SLCO5A1 gene or the target sequence for SLCO5A1 knockdown was subcloned and inserted into the pCDNA3.1 vector to construct the overexpression (OE)-SLCO5A1 or shRNA (sh)-SLCO5A1 plasmid, respectively. Transfection was conducted in accordance with the manufacturer’s protocol using a Lipofectamine 3000 Kit (Cat# L3000150; Thermo Fisher Scientific, Pittsburgh, PA, USA). Specifically, 1 μg of plasmid DNA (with 2 μL of P300 enhancer reagent) and 2.5 μL of Lipofectamine 3000 were each prepared in separate tubes containing 125 μL of Opti-MEM (Cat# 31985070; Thermo Fisher Scientific, Pittsburgh, PA, USA). The two mixtures were subsequently combined, gently mixed, and incubated at room temperature for 10 min before being introduced to the cells. The cells were incubated with the resulting liposome‒DNA complexes for 48–72 h prior to harvest.

Cell culture

Cardiac microvascular endothelial cells (cMECs; GN-M122; GAINING, Shanghai, China) were cultured in DMEM (Cat# 11965092; Thermo Fisher Scientific, PA, USA) supplemented with 10% FBS and 2% penicillin‒streptomycin (Cat# C0222; Beyotime, Beijing, China) at 37 °C with 5% CO₂. The medium was replaced every three days.

Molecular dynamics (MD)

The three-dimensional structures of the SLCO5A1 protein and T1AM ligand were retrieved from the UniProt (ID: Q9H2Y9) and PubChem databases. Molecular docking between SLCO5A1 and T1AM was performed using AutoDock Vina, with poses visualized in PyMOL (v2.3.0). Molecular dynamics simulations were conducted using the docked complexes in GROMACS 2024.4.

Cellular thermal shift assay (CETSA)

Prior to protease digestion, sample handling was performed according to the procedure for the drug affinity responsive target stability (DARTS) assay. The samples were subjected to a temperature gradient (40, 44, 48, 52, 56, 60, and 64 °C) for 3 min to induce protein denaturation while keeping proteases inactive. After heating and centrifugation, the target proteins in the supernatant were analyzed by Western blotting.

Free energy landscape analysis

Free energy landscape analysis was employed to evaluate the conformational stability of the SLCO5A1–T1AM complex during molecular dynamics simulations. This analysis was conducted using the built-in g_sham module of GROMACS (2024.4), examining the root mean square deviation (RMSD), root mean square fluctuation (RMSF), hydrogen bonds, and radius of gyration (Rg). A three-dimensional free energy landscape plot with RMSD, Rg, and Gibbs free energy represented on the X, Y, and Z axes, respectively, was generated. This plot reflects the conformational space sampled during the simulation and identifies the lowest-energy conformations. A single, narrow, and smooth energy basin indicates stable protein‒ligand interactions with well-defined conformations, whereas multiple basins or a rugged energy basin suggest conformational heterogeneity or weak binding. In the color-coded plot, deep purple/blue areas represent energy minima, whereas red/yellow areas correspond to higher energies with lower stability.

Surface plasmon resonance (SPR)

Interactions between SLCO5A1 and T1AM were analyzed by SPR using a Biacore Insight system (Cytiva, V6.0; Marlborough, MA, USA). SLCO5A1 was immobilized on a CM5 sensor chip via amine coupling. T1AM (0.03125 nM to 1 nM in PBS-P buffer, pH 7.4) was allowed to flow over the chip at 10 μL/min, and the binding responses were measured in resonance units (RU). After double-reference subtraction, the sensorgrams were fitted to a 1:1 binding model to determine the kinetic rate constant (k_d) and equilibrium dissociation constant (K_D).

Tissue immunofluorescence

Mouse hearts were fixed in 4% paraformaldehyde with 20% sucrose for 18 h at 4 °C. The tissues were embedded in OCT compound (Cat# 4583; SAKURA, Tokyo, Japan) and frozen on dry ice. The cryosections were incubated with an α-SMA antibody (1:100 dilution; Cat# 124964; Abcam, London, UK), followed by incubation with secondary antibodies (Table 2). Nuclei were counterstained with DAPI (Cat# ab228549; Abcam, London, UK).

Table 2.

Antibody information.

Primary antibodies	Clone/Catalog	Supplier
Collagen I	ab260043	Abcam, UK
Fibronectin	ab45688	Abcam, UK
α-SMA	ab124964	Abcam, UK
ANP	27426-1-AP	Proteintech, China
BNP	13299-1-AP	Proteintech, China
β-MHC	A7564	ABclonal, China
Smad3	9523 T	CST, USA
p-Smad3	9520 T	CST, USA
SLCO5A1	ab191412	Abcam, UK
SLCO5A1	25744-1-AP	Proteintech, China
Tubulin	11224-1-AP	Proteintech, China
GAPDH	GB15004	Servicebio, China
HRP-Goat anti Rabbit IgG (H + L)	SA00001-4	Proteintech, China
HRP-Goat anti-Mouse IgG (H + L)	SA00001-1	Proteintech, China
Anti-Mouse IgG (Alexa Fluor 647)	ab150115	Abcam, UK

Histological staining

For histomorphometry analysis, the hearts were fixed in 10% neutral buffered formalin overnight and then embedded in paraffin. Sections (4–5 μm thick) were stained with wheat germ agglutinin (WGA; Cat# L4895; Sigma–Aldrich, St. Louis, MO, USA), hematoxylin and eosin (H&E), Masson’s trichrome (Cat# G1340; Solarbio, Beijing, China), and Picrosirius Red (PSR) as previously described [28, 29].

Western blotting

Protein lysates were separated by SDS‒PAGE (7.5%–12% gels), and the protein bands were transferred to PVDF membranes (Cat# IPVH00010; Millipore, USA). After blocking with 5% nonfat milk in TBST, the membranes were incubated overnight at 4 °C with specific primary antibodies (Table 2) and horseradish peroxidase (HRP)-conjugated secondary antibodies. Blots were visualized using an enhanced chemiluminescence substrate (Cat# BL520A; Biosharp, Beijing, China) and imaged using a Gel View 6000Plus (Biolight, Guangzhou, China) or Tanon 5200 (Tanon, Shanghai, China) instrument. Band intensity was quantified using ImageJ software (NIH, USA).

Enzyme-linked immunosorbent assay (ELISA)

Circulating SLCO5A1 concentrations in the plasma of HF patients and controls were measured using ELISA. A primary antibody against SLCO5A1 (Cat# 25744-1-AP; Proteintech, Wuhan, China) was diluted 1:2000 and used to coat a 96-well microplate (Cat# M0661; Sigma‒Aldrich, St. Louis, MO, USA) that was incubated overnight at 4 °C. After the microplate washed, the plasma samples were added to the microplate and incubated at 37 °C for 1 h. The secondary antibody (Cat# SA00001-1; Proteintech, Wuhan, China) was then added and incubated at 37 °C for 1 h. The chromogenic substrate TMB (Cat# T0440; Sigma‒Aldrich, st. Louis, MO, USA) was added, followed by a stop solution. The TGF-β1 concentrations in the supernatants and cardiac tissues were determined using a commercial ELISA kit (Cat# E-EL-0162; Elabscience, China) following the manufacturer’s instructions. The absorbance at 450 nm was measured using a microplate reader. The SLCO5A1 protein concentration was calculated using a standard curve constructed with purified SLCO5A1 protein (Cat# 512859; Novopro, Shanghai, China) at 0, 2.5, 5, 10, and 20 ng/ml.

Statistical analysis

Statistical results are presented according to the type of variable as either medians with interquartile ranges (IQRs, 25%–75%) or means ± standard deviations (SDs). Categorical variables are presented as absolute numbers and corresponding percentages. Group comparisons for continuous variables were conducted using t tests or Mann‒Whitney U tests, while Fisher’s exact test was employed for categorical data. For comparisons involving multiple groups, one-way ANOVA or the Kruskal‒Wallis test was used depending on the normality of the data distribution, with post hoc analysis performed using the Bonferroni method. Interaction effects in two-factor designs were assessed via two-way ANOVA. All analyses were conducted using GraphPad Prism 9.0 (GraphPad Software, Inc., USA), and statistical significance was set at P < 0.05.

Results

Integrated transcriptomics identifies candidate genes associated with HF

First, we integrated two gene expression datasets pertaining to human ischemic cardiomyopathy (GSE55296 and GSE57345) to identify novel candidate genes involved in HF. After batch correction, the merged dataset comprised 257 left ventricular samples (149 control and 108 HF samples), with good separation between groups indicated by PCA (Fig. 1a, b). Differential expression analysis revealed 270 differentially expressed genes (DEGs) in the HF samples compared with control samples (143 upregulated and 127 downregulated; Fig. 1c and Supplementary Table S1). GO enrichment analysis indicated that these DEGs are involved primarily in extracellular matrix (ECM) organization, cell–matrix adhesion, or collagen trimer formation or act as ECM structural constituents (Supplementary Fig. S1a–1c). Similarly, KEGG pathway analysis revealed enrichment in ECM–receptor interactions, cytokine‒cytokine receptor interactions, and the PI3K-Akt signaling pathway (Supplementary Fig. S1d).

Fig. 1. Screening of candidate genes based on DEG analysis and WGCNA.

a PCA of GSE55296 and GSE57345 before batch correction. b PCA after batch correction. c Volcano map of DEGs (HF vs. control). Upregulated genes are shown in red, and downregulated genes are shown in green. d Module‒trait correlation heatmap from WGCNA. e Scatter plot showing the relationship between genetic significance for HF and module membership in the blue module. f Venn diagram showing overlap between DEGs and blue-module genes. g KEGG enrichment analysis of overlapping genes. h GO enrichment analysis of overlapping genes. DEG, differentially expressed gene; GO, Gene Ontology; HF, heart failure; KEGG, Kyoto Encyclopedia of Genes and Genomes; PCA, principal component analysis; WGCNA, weighted gene coexpression network analysis.

We subsequently utilized WGCNA to construct a gene coexpression network. The soft-thresholding power was optimized to β = 11 to ensure a scale-free topology (R² ≥ 0.8) while preserving suitable mean connectivity (Supplementary Fig. S2a–S2c). Five distinct gene modules were identified (Figs. S2d and 1d). Notably, the “blue” module demonstrated the most robust positive correlation with the HF trait (Supplementary Table S2). Within this module, significant correlations were observed between gene significance and clinical traits (cor = 0.71, P = 1e-55) (Fig. 1e). Intersecting the blue-module genes with DEGs yielded 133 overlapping genes (Fig. 1f and Supplementary Table S3), which remained strongly enriched in ECM-related biological processes and pathways (Fig. 1g, h). These analyses defined a HF-associated signature for subsequent feature selection.

Machine-learning feature screening yields seven candidate genes with diagnostic performance

To refine the key features of the 133 identified genes, we employed two machine learning methods (LASSO and SVM-RFE). In the LASSO model, 5-fold cross-validation was utilized to determine the optimal regularization parameter, resulting in the retention of 21 high-weight features (Supplementary Fig. S3a, S3b). The SVM-RFE algorithm, optimized through 10-fold cross-validation iterations using a radial basis function kernel, identified 16 features with maximal classification information entropy (Supplementary Fig. S3c, S3d). The intersection of these two distinct algorithms revealed seven potential candidate genes, namely, ficolin (FCN3), immunoglobulin superfamily containing leucine-rich repeat (ISLR), KIAA0040, ladinin 1 (LAD1), secreted frizzled-related protein 4 (SFRP4), SLCO5A1, and XIAP-associated factor 1 (XAF1) (Supplementary Fig. S3e). Subsequently, ROC curves were constructed to validate the diagnostic performance of these candidate genes, and all AUC values exceeded 0.75 (Supplementary Fig. S3f). To illustrate the clinical utility of these genes, a nomogram to predict HF risk on the basis of these genes was constructed (Supplementary Fig. S3g). The calibration curve demonstrated excellent concordance between the probabilities predicted by the model and actual observations (Supplementary Fig. S3h). These results support the prioritization of the seven-gene set for causal inference testing.

MR supports a protective association between SLCO5A1 and HF risk

We next tested whether genetically predicted levels of the candidate gene products were associated with HF risk using two-sample MR. XAF1 was excluded from the analysis because cis-pQTL instruments were unavailable. MR highlighted SLCO5A1 among the remaining candidates as the only gene associated with a significant protective effect against HF (Fig. 2a and Supplementary Tables S4, S5). Specifically, a genetically predicted increase in circulating SLCO5A1 was associated with a decreased risk of HF (OR = 0.862, 95% CI = 0.749–0.993; P = 0.039). Sensitivity analyses confirmed the robustness of this association, revealing no evidence of heterogeneity (Cochran’s Q test) or horizontal pleiotropy (MR‒Egger intercept and MR-PRESSO tests) (Supplementary Table S6). Reverse-direction MR did not support HF as a causal determinant of SLCO5A1 (Fig. 2a, Supplementary Tables S7–S9), which is consistent with the hypothesized directionality of this association. On the basis of these results, we focused our subsequent experiments on SLCO5A1.

Fig. 2. SLCO5A1 is downregulated in the failing myocardium and cardiomyocytes and exerts a protective effect on HF.

a Bidirectional two-sample MR results for candidate genes and HF risk (forward MR: SLCO5A1 toward HF; reverse MR: HF toward SLCO5A1). b, c Representative immunoblotting and quantification of SLCO5A1 protein levels in mouse hearts 4 weeks after MI or sham surgery (n = 6). d, e Representative immunoblotting and quantification of SLCO5A1 protein expression in NMCMs treated with ISO-PE or vehicle (n = 6). f, g Immunoblotting and quantification of SLCO5A1 protein expression in CFs, CMs and cMECs (n = 6). h, i Representative immunofluorescence images and quantification of SLCO5A1 expression in hearts after MI or sham surgery (scale bar = 50 μm). **P < 0.01, ***P < 0.001 vs. the indicated control. CF, cardiac fibroblast; CM, cardiomyocyte; cMEC, cardiac microvascular endothelial cell; ISO, isoprenaline; MI, myocardial infarction; NMCM, neonatal mouse cardiomyocyte; PE, phenylephrine.

SLCO5A1 is downregulated in ischemic failing hearts and in stressed cardiomyocytes

To validate the in silico findings, we investigated SLCO5A1 expression in experimental models. SLCO5A1 expression in the left ventricular tissue was reduced at 4 weeks post-MI compared with that in the sham group (Fig. 2b, c). Consistent with the in vivo results, SLCO5A1 expression was also diminished in a primary cardiomyocyte hypertrophy model induced by ISO and PE (ISO-PE) (Fig. 2d, e). Furthermore, in a small clinical cohort (Table 3), plasma SLCO5A1 levels were lower in patients with HF than in controls (Fig. S4). Across major cardiac cell types, SLCO5A1 protein expression was markedly higher in cardiomyocytes (CMs) than in cardiac fibroblasts (CFs) and endothelial cells (cMECs), as determined by immunoblotting of isolated cells (Fig. 2f, g), in alignment with single-cell RNA sequencing annotations from the Human Protein Atlas database (https://www.proteinatlas.org). Immunofluorescence staining further confirmed that SLCO5A1 expression was reduced in cardiomyocytes from mouse hearts after MI (Fig. 2h, i). These results support a cardiomyocyte-centered role for SLCO5A1 during remodeling.

Table 3.

Clinical characteristics of patients with HF and non-failing controls (Control).

Variables	Control (n = 4)	HF (n = 8)	P (HF vs Control)
Male, n (%)	2 (50%)	3 (37.5%)	0.045
Age (year)	67.25 ± 6.00	73.25 ± 6.59	0.384
LVEF (%)	61.50 ± 6.683	41.50 ± 6.895	0.016
LVFS (%)	33.00 ± 4.328	22.00 ± 5.492	0.029
NT-proBNP (pg/ml)	37.27 (17.8-71.8)	4470.5 (3012.0-33508.6)	<0.001
hs-cTnT (ng/L)	5.21 (0.03-6.17)	33.54(21.33-1021)	<0.001

LVEF left ventricular ejection fractions, LVFS left ventricular fractional shortening, NT-proBNP N-terminal pro-brain natriuretic peptide, hs-cTNT high-sensitivity cardiac troponin T.

Cardiomyocyte-specific SLCO5A1 overexpression mitigates MI-induced cardiac remodeling and dysfunction

Next, we induced cardiomyocyte-specific overexpression of the gene encoding SLCO5A1 using AAV9-cTNT-SLCO5A1 through tail-vein injection to assess the therapeutic efficacy of SLCO5A1 (Supplementary Fig. S5a). Immunoblotting confirmed significant SLCO5A1 overexpression in cardiac tissue (Fig. S5b, S5c). Echocardiographic analysis revealed that SLCO5A1 overexpression significantly ameliorated left ventricular systolic dysfunction, as evidenced by increased EF and FS compared to those in the vector control group (Figs. 3a, b). Histological examination revealed that SLCO5A1 overexpression reduced cardiac fibrosis and collagen deposition in HF mice (Fig. 3c–f). Western blot analysis confirmed that hearts overexpressing SLCO5A1 exhibited significantly lower levels of the fibrosis markers fibronectin, collagen I, and α-SMA (Fig. 3g, h). Consistently, SLCO5A1 overexpression markedly suppressed the protein expression of β-MHC, ANP, and BNP in HF tissues (Fig. 3i, j) and reduced the myocardial size (Fig. 3k, l). Together, these data indicate that increasing cardiomyocyte SLCO5A1 levels blunted structural remodeling and improved function after MI.

Fig. 3. Cardiomyocyte-targeted SLCO5A1 overexpression attenuates post-MI cardiac dysfunction and remodeling.

a Representative echocardiographic images from sham or HF mice injected with AAV9-cTnT-vector (Vec) or AAV9-cTnT-SLCO5A1 (OE) via tail-vein injection. b Quantitative analysis of EF and FS (n = 6). c Representative images of Masson’s trichrome staining of LV sections (scale bar = 1 mm). d Representative PSR staining images (scale bar = 1 mm). e, f Quantitative analysis of fibrosis and collagen areas from (c, d) (n = 6). g, h Western blotting and quantitative analysis of cardiac fibronectin, collagen I, and α-SMA expression. i, j Western blotting and quantitative analysis of cardiac β-MHC, ANP, and BNP expression. k, l WGA staining and quantitative analysis. ^**P < 0.01, ^***P < 0.001, ^##P < 0.01, and ^### P < 0.001. Fibronectin, FN; Collagen I, Col.

T1AM directly binds to SLCO5A1 and increases its protein expression and thermal stability

To identify a pharmacological agent that can increase SLCO5A1 expression, we queried the DSigDB database and identified two candidates: T1AM and amiodarone. Given the lack of a survival benefit of amiodarone across various HF phenotypes [30, 31], we prioritized T1AM for mechanistic and functional testing (chemical structure shown in Fig. 4a).

Fig. 4. T1AM directly interacts with SLCO5A1 and increases its protein expression and thermal stability.

a Chemical structure of T1AM. b Representative Western blotting images and quantitative analysis of SLCO5A1 expression in ISO-PE-induced NMCMs treated with T1AM at the indicated concentrations (n = 6). d Representative Western blotting images and quantitative analysis of SLCO5A1 expression in the hearts of mice with HF treated with vehicle or T1AM at the indicated doses (n = 6). c, e Quantitative analysis of the Western blotting results (b, d). f Representative molecular docking pose between SLCO5A1 and T1AM. g MM-GBSA analysis of energy catabolism showing the top 10 amino acids that contribute to the binding of T1AM and SLCO5A1. h RMSD of the SLCO5A1–T1AM complex during MD simulation. i RMSF of residues during MD simulation. j Number of hydrogen bonds between T1AM and SLCO5A1 during MD simulation. k Free energy landscape of the SLCO5A1–T1AM complex. l–m CETSA immunoblot showing SLCO5A1 levels after heating in vehicle-treated vs. T1AM-treated samples. n Quantitative analysis of the CETSA results (l, m). o SPR sensorgram showing the binding of T1AM to immobilize SLCO5A1 and model fitting. ^**P < 0.01, ^***P < 0.001, ^#P < 0.05, ^##P < 0.01, and ^###P < 0.001.

Treatment with T1AM resulted in the dose-dependent upregulation of SLCO5A1 expression in post-MI hearts and ISO-PE-induced NMCMs (Fig. 4b–e). Molecular docking analysis suggested a plausible SLCO5A1-T1AM binding pose (docking score of −6.490 kcal/mol) (Fig. 4f), and MD simulations supported a stable interaction during the simulation window. MM-GBSA decomposition identified residues (TYR-306 and SER-278) that make prominent contributions to binding energy as the primary contributors to the interaction between SLCO5A1 and T1AM (Fig. 4g). RMSD and RMSF analyses indicated that the T1AM–SLCO5A1 complex remained stable throughout the simulation (Fig. 4h, i), as evidenced by the formation of hydrogen bonds (Fig. 4j) and the free energy landscape (FEL) profile (Fig. 4k). To experimentally validate the interaction between T1AM and SLCO5A1, CETSA was performed. Compared with vehicle treatment, T1AM treatment significantly shifted the thermal denaturation profile of SLCO5A1 (Fig. 4l–n). SPR further confirmed the direct interaction between T1AM and SLCO5A1 (Fig. 4o). Collectively, these results suggest that T1AM directly binds to SLCO5A1 and increases SLCO5A1 protein abundance in cells and tissues.

Pharmacological T1AM treatment alleviates cardiac damage in MI-induced HF

Next, we assessed the therapeutic effects of T1AM in vivo. The administration of T1AM at doses of 12.5, 25, and 50 mg·kg⁻¹ increased the survival rates of HF mice in a dose-dependent manner (Fig. 5a). T1AM treatment significantly increasedleft ventricular EF and FS (Fig. 5b, c) and preserved the myocardial tissue structure (Fig. 5d). Additionally, T1AM reduced cardiac fibrosis (Fig. 5e, i), the collagen content (Fig. 5f, j), the cardiomyocyte cross-sectional area (Fig. 5g, k), and in situ α-smooth muscle actin (α-SMA) expression in a dose-dependent manner (Fig. 5h, l). These findings were further confirmed by Western blot analysis, which revealed significant reductions in the levels of fibrosis-related proteins (Fibronectin, Collagen I, and α-SMA) and HF biomarkers (β-MHC, ANP, and BNP) (Fig. 5m–p). These results suggest that T1AM significantly attenuates post-MI remodeling in vivo.

Fig. 5. T1AM increases survival and improves cardiac function and cardiac remodeling after MI.

a Twenty-eight-day Kaplan‒Meier survival curves for sham or HF mice treated with NS or T1AM at the indicated doses (n = 30 mice per group). b Representative echocardiographic images at 28 days post-MI (n = 6). c Quantitative analysis of EF and FS. d Representative images of H&E-stained LV sections (scale bar = 50 μm). e Representative images following Masson’s trichrome staining (scale bar = 1 mm). f Representative images following PSR staining (scale bar = 1 mm). g Representative images following WGA staining of cross-sectional cardiac tissues (scale bar = 50 μm). h Representative images following immunofluorescence staining of α-SMA in cardiac sections (scale bar = 100 μm). i, j Quantitative analysis of fibrosis and collagen areas from (e, f) (n = 6). k Quantitative analysis of the cardiomyocyte cross-sectional area from (g) (n = 6). l Quantitative analysis of ɑ-SMA immunofluorescence staining data (n = 6). m Representative Western blotting images showing cardiac Collagen I, Fibronectin, and α-SMA expression in mice 28 days post-MI. n Representative Western blotting images showing β-MHC, ANP, and BNP expression. o, p Quantitative analysis of the Western blotting results (m, n) (n = 6). ^*P < 0.05. ^**P < 0.01. ^***P < 0.001, ^#P < 0.05. ^##P < 0.01, and ^###P < 0.001. Fibronectin, FN; Collagen I, Col.

SLCO5A1 upregulation suppresses hypertrophic marker induction in cardiomyocytes in vitro

In ISO-PE-treated NMCMs, T1AM significantly decreased ANP and BNP protein expression in a concentration-dependent manner, which coincided with the progressive upregulation of SLCO5A1 protein expression (Fig. 6a–c). To test whether SLCO5A1 is sufficient to reproduce this effect, we overexpressed SLCO5A1 in NMCMs. SLCO5A1 overexpression suppressed the ISO-PE-induced upregulation of β-MHC, ANP, and BNP (Fig. 6d–f). These in vitro findings further support the hypothesis that the therapeutic effects of T1AM are mediated by SLCO5A1 upregulation.

Fig. 6. SLCO5A1 upregulation suppresses hypertrophic signaling in vitro.

a Experimental T1AM treatment regimen applied to ISO-PE-treated NMCMs. b, c Representative Western blotting images and quantitative analysis of SLCO5A1, β-MHC, ANP, and BNP expression in ISO-PE-induced NMCMs treated with vehicle or T1AM under the indicated conditions (n = 6). d Experimental scheme for SLCO5A1 overexpression in ISO-PE-treated NMCMs. e, f Representative Western blot images and quantitative analysis of SLCO5A1, β-MHC, ANP, and BNP expression under the indicated conditions (n = 6). ^**P < 0.01, ^***P < 0.001, ^#P < 0.05, ^##P < 0.01, and ^###P < 0.001.

SLCO5A1 is required for the cardioprotective effects of T1AM in HF

To confirm that SLCO5A1 serves as the functional target of T1AM, we conducted cardiac-specific knockdown of SLCO5A1 utilizing AAV9-shRNA in T1AM-treated HF mice (Fig. S5d–S5f). Compared with negative control shRNA (Neg), SLCO5A1 knockdown (sh) attenuated the protective effects of T1AM, resulting in significant reductions in left ventricular EF and FS (Fig. 7a, b). Furthermore, SLCO5A1 knockdown reversed the antifibrotic effects of T1AM, leading to increases in cardiac fibrosis (Figs. 7c, e) and the collagen content (Fig. 7d, f). Western blot analysis confirmed that the suppression of fibrosis and cardiac remodeling markers by T1AM was inhibited by SLCO5A1 knockdown (Fig. 7g–j). In vitro, T1AM significantly reduced cardiomyocyte hypertrophy, as assessed by α-actinin staining, in ISO-PE-treated cells; however, this effect was negated by SLCO5A1 knockdown (Fig. 7k–m). Similarly, the T1AM-induced reduction in β-MHC, ANP, and BNP expression was reversed upon silencing SLCO5A1 expression (Fig. 7n, o). Collectively, these data indicate that SLCO5A1 is an important functional mediator of T1AM-mediated cardioprotection in both in vivo and in vitro models.

Fig. 7. SLCO5A1 knockdown weakens the protective effects of T1AM on HF.

a Representative echocardiographic images of HF mice treated with NS (HF-NS), T1AM (HF-T1AM), T1AM with Neg (HF-T1AM+Neg), or SLCO5A1 knockdown shRNA (HF-T1AM + sh). b Quantitative analysis of EF and FS (n = 6). c Representative images following Masson’s trichrome staining (scale bar = 1 mm). d Representative images following PSR staining (scale bar = 1 mm). e, f Quantitative analysis of the results in (c) and (d) (n = 6). g Representative Western blot images showing cardiac Fibronectin, Collagen I, and α-SMA expression. h Representative Western blot images showing cardiac β-MHC, ANP and BNP expression. i, j Quantitative analysis of the results in (g) and (h) (n = 6). k Experimental scheme used to combine T1AM treatment and SLCO5A1 knockdown in ISO-PE-treated NMCMs. l, m Representative images following immunofluorescence staining of α-actinin and quantitative analysis of cardiomyocyte size (n = 6). Scale bar = 25 μm. n, o Representative Western blotting images and quantitative analysis of SLCO5A1, β-MHC, ANP, and BNP expression in NMCMs under the indicated conditions (n = 6). ^*P < 0.05, ^**P < 0.01, ^***P < 0.001, ^#P < 0.05, ^##P < 0.01, ^###P < 0.001, and ^$$P < 0.01. Fibronectin, FN; Collagen I, Col.

SLCO5A1 limits profibrotic cardiomyocyte‒fibroblast crosstalk by reducing TGF-β1 secretion and fibroblast Smad3 activation

Because fibrosis is primarily mediated by fibroblasts, whereas SLCO5A1 is enriched in cardiomyocytes, we tested whether SLCO5A1 modulates cardiomyocyte-derived profibrotic signals. TGF-β signaling is known to be activated in infarcted and failing hearts [32] and thereby contribute to cardiac fibrosis and remodeling. In post-MI hearts, T1AM reduced myocardial TGF-β1 levels in a dose-dependent manner (Fig. 8a). In vitro, SLCO5A1 overexpression significantly decreased TGF-β1 levels in the supernatant of ISO-PE-treated cardiomyocytes (Fig. 8b). Similarly, T1AM reduced TGF-β1 secretion from stressed cardiomyocytes, and this effect was weakened by SLCO5A1 knockdown (Fig. 8c, d).

Fig. 8. SLCO5A1 upregulation reduces TGF-β1 secretion in cardiomyocytes under HF conditions to alleviate NMCF fibrosis through the TGF-β1/Smad3 pathway.

a ELISA of TGF-β1 levels in mouse cardiac tissue after MI and treatment with vehicle or T1AM at the indicated doses (n = 6). b ELISA of TGF-β1 levels in the supernatant from ISO-PE-induced NMCMs with or without SLCO5A1 overexpression (n = 6). c ELISA of TGF-β1 levels in the supernatant from ISO-PE-induced NMCMs treated with vehicle or T1AM (n = 6). d ELISA of TGF-β1 levels in the supernatant from ISO-PE-induced NMCMs treated with T1AM plus Neg or SLCO5A1 shRNA (n = 6). e Schematic diagram of conditioned-medium transfer from NMCMs to NMCFs. f–k Representative Western blotting images and quantitative analysis of Collagen I, Fibronectin, and α-SMA expression in NMCFs treated with conditioned medium from NMCMs under the indicated treatment conditions (n = 6). l–p Representative Western blot images and quantification of p-Smad3 (Ser423/425) and total Smad3 protein expression in NMCFs treated with the indicated conditioned medium (n = 6). ^**P < 0.01, ^***P < 0.001, ^#P < 0.05, ^##P < 0.01, and ^###P < 0.001.

To directly assess intercellular signaling, we treated cardiac fibroblasts with conditioned medium from cardiomyocytes subjected to the indicated interventions (Fig. 8e). Conditioned medium from ISO-PE-treated cardiomyocytes increased the expression of profibrotic markers (α-SMA, Collagen I, and Fibronectin) in NMCFs. However, conditioned medium from SLCO5A1-overexpressing cardiomyocytes (Fig. 8f, g) or T1AM-treated cardiomyocytes (Fig. 8h, i) attenuated fibroblast activation. Conditioned medium from SLCO5A1-silenced cardiomyocytes in the presence of T1AM increased the expression of fibroblast markers (Fig. 8j, k). Furthermore, conditioned medium from SLCO5A1-overexpressing or T1AM-treated cardiomyocytes reduced fibroblast Smad3 phosphorylation (Ser423/425), whereas conditioned medium from SLCO5A1-knockdown cardiomyocytes increased p-Smad3 levels (Figs. 8l–p). These results support a model in which cardiomyocyte SLCO5A1 limits profibrotic signaling in fibroblasts by reducing cardiomyocyte TGF-β1 secretion and downstream fibroblast Smad3 activation.

Discussion

In the present study, we combined human ischemic cardiomyopathy transcriptomics with network analysis, machine learning, and MR to identify SLCO5A1 as a candidate protective gene in HF. We then provided experimental evidence that cardiomyocyte-specific SLCO5A1 overexpression attenuates post-MI dysfunction and remodeling in a mouse MI model and in NMCMs. In addition, through biophysical/cellular assays, we identified T1AM as a small molecule that directly binds to SLCO5A1 and increases SLCO5A1 protein abundance, and we showed that the cardioprotective effect of T1AM in vivo and in vitro is markedly weakened by SLCO5A1 knockdown in cardiomyocytes. Together, these findings suggest that SLCO5A1 protects against pathological cardiac remodeling post-MI, at least in part, by modulating cardiomyocyte‒fibroblast communication through the TGF-β1/Smad3 signaling pathway. Our findings suggest that a pharmacological increase in SLCO5A1 may serve as a therapeutic strategy for post-MI HF.

The SLCO superfamily generally facilitates the sodium-independent transport of endogenous substrates [33]. Although SLCO5A1 belongs to this family, it is structurally atypical, and its physiological substrates remain poorly defined. Our study provides evidence that SLCO5A1 is functionally linked to cardiac protection in MI-induced HF. MR analysis provides additional genetic support for a protective association between high SLCO5A1 expression and low HF risk, helping to overcome a common limitation in omics studies [34]. We used murine MI models and plasma from HF patients to confirm the downregulation of SLCO5A1 in human HF datasets, indicating the clinical translational potential of SLCO5A1. Furthermore, through cardiac-specific overexpression or knockdown of SLCO5A1 combined with T1AM treatment, which increases SLCO5A1 expression, our findings demonstrated that SLCO5A1 protects against cardiac damage and dysfunction in HF.

T1AM is an endogenous thyroid hormone metabolite that acts independently of nuclear thyroid hormone receptors [35]. Previous studies have shown that T1AM is detectable in blood and cardiac tissue, that cardiomyocytes can take up and metabolize T1AM, and that circulating T1AM concentrations are increased in patients with chronic heart failure and are inversely associated with LVEF [13, 15, 36]. However, the endogenous role of T1AM in HF remains unclear. In this study, we did not measure endogenous circulating or myocardial T1AM levels. Therefore, we could not determine whether T1AM was increased in our HF model or whether increased T1AM levels represented a compensatory response to alter thyroid metabolite handling [37]. Previous studies have shown that exogenous T1AM can attenuate ischemic cardiac damage in models of ischemia‒reperfusion or hypoxia/reoxygenation [13, 22, 23]. Consistent with these studies, our experimental evidence demonstrated that exogenous T1AM administration had beneficial effects in both in vivo and in vitro HF models.

Previous studies have often attributed the cardiovascular effects of T1AM trace amine-associated receptor 1 (TAAR1) or mitochondrial mechanisms [14, 15]. Our data elucidates the related regulatory network by identifying SLCO5A1 as a new direct target of T1AM. Molecular docking, MD simulations, CETSAs, and SPR consistently supported direct physical binding between T1AM and SLCO5A1. Moreover, T1AM treatment increased SLCO5A1 protein levels in stressed cardiomyocytes and post-MI hearts. In contrast, the cardioprotective effects of T1AM were significantly weakened by SLCO5A1 knockdown. These findings suggest that SLCO5A1 not only is associated with T1AM treatment but also acts as a functional effector of T1AM in terms of anti-remodeling outcomes.

Cardiac fibrosis is largely mediated by activated fibroblasts [38, 39], but our data and public single-cell data in the Human Protein Atlas database indicate that SLCO5A1 is more enriched in cardiomyocytes relative to fibroblasts and endothelial cells. This cell-type distribution led us to examine cardiomyocyte–fibroblast crosstalk. TGF-β1, a critical regulator of fibrosis [40], is activated in infarcted and remodeling hearts [32]. TGF-β1 typically activates Smads, such as Smad3, by phosphorylation to promote the phenotypic switch of fibroblasts [41]. We found that SLCO5A1 overexpression reduced TGF-β1 levels in cardiomyocyte-conditioned medium, with corresponding changes in fibroblast Smad3 phosphorylation and profibrotic marker expression. In contrast, SLCO5A1 knockdown resulted in the opposite outcome. These results are consistent with emerging findings that cardiomyocyte-derived paracrine signaling is a critical mechanism underlying fibrotic remodeling in HF [42, 43].

The molecular mechanism underlying SLCO5A1-mediated inhibition of cardiomyocyte TGF-β1 secretion remains unclear. Our current data reveal a functional role for SLCO5A1 in regulating TGF-β1 levels in cardiomyocyte-conditioned medium; however, we did not establish a direct molecular interaction between SLCO5A1 and TGF-β1 processing or secretory machinery. As a membrane transporter, SLCO5A1 might influence the intracellular processing, maturation, secretion, reuptake, clearance, or degradation of TGF-β1 [44, 45]. Future investigations, such as coimmunoprecipitation with mass spectrometry to identify SLCO5A1-associated protein complexes involved in TGF-β1 processing or secretion, are warranted to elucidate the underlying molecular mechanisms.

This study has several limitations. First, although we demonstrated the efficacy of T1AM in a murine MI-induced HF model, its anti-HF effects in other types of HF and in humans require further evaluation. Second, although we revealed that the TGF-β/Smad3 axis is an important downstream pathway, the precise intracellular mechanisms through which SLCO5A1 suppresses TGF-β1 secretion from cardiomyocytes require further investigation. Third, endogenous circulating and myocardial T1AM levels were not determined in this study; thus, the pathophysiological importance of endogenous T1AM in HF remains unresolved. Finally, only male mice were used in the MI model; given the known sex differences in HF, future studies should investigate whether this mechanism is conserved in females.

In conclusion, our study identified SLCO5A1 as a novel therapeutic target for ischemic heart failure and demonstrated that pharmacological T1AM administration alleviates cardiac remodeling by binding to SLCO5A1 and interrupting profibrotic cardiomyocyte‒fibroblast crosstalk in HF. These findings suggest a strategy aimed at enhancing SLCO5A1 signaling in the failing heart.

Author contributions

WLC, KYL, and LQZ designed the research; KYL and QHC performed the research; KG and FF analyzed the data; KYL and WLC wrote the paper; WLC, LQZ, and JRX obtained funding, supervised and approved the final submission of the manuscript; all the authors revised and approved the manuscript.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41401-026-01827-4.