Integrated multi-omics analysis identifies and validates endoplasmic reticulum stress and mitophagy-related biomarkers in MASLD

Abstract

Metabolic dysfunction-associated steatotic liver disease (MASLD) is one of the most prevalent chronic liver diseases worldwide. Growing evidence indicates that endoplasmic reticulum stress (ERS) and mitophagy play critical roles in MASLD progression. However, the specific mechanism by which ERS and mitophagy participate in MASLD is not clear, and there is still a lack of treatment strategies for these pathways. Therefore, this study aims to integrate GEO data mining, machine learning, and scRNA-seq analysis to elucidate potential mechanisms by which ERS and mitophagy mediate MASLD progression, and to identify candidate therapeutic agents targeting these pathways through high-throughput virtual screening. Nuclear receptor subfamily 4 group A member 1 (NR4A1) was identified as a key target involved in the progression of MASLD mediated by ERS and mitophagy. Functional enrichment and immune infiltration analyses suggested that NR4A1 participates in immunoregulatory and adaptive responses under metabolic stress. scRNA-seq analysis confirmed that the expression of NR4A1 decreases gradually during the progression of MASLD, particularly throughout macrophage activation and differentiation. Experimental validation further demonstrated that NR4A1 was downregulated in FFA-treated hepatocytes, MASLD mice, and human liver tissues. A ceRNA network centered on NR4A1 was constructed, and potential therapeutic compounds were identified via molecular docking. Taken together, this study highlights NR4A1 as a key target in ERS- and mitophagy-mediated MASLD progression and provides novel insights into therapeutic strategies for alleviating MASLD by targeting ERS and mitophagy pathways.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-43311-3.

Introduction

Metabolic dysfunction-associated steatotic liver disease (MASLD) is a metabolic liver disorder recently redefined to replace nonalcoholic fatty liver disease (NAFLD). It is characterized by hepatic steatosis driven by metabolic dysfunction^1,2. The global prevalence of MASLD has increased markedly and is now recognized as one of the most common chronic liver diseases. MASLD is closely associated with metabolic syndrome, including obesity, type 2 diabetes mellitus, and dyslipidemia³. Without effective intervention, MASLD can progress to steatohepatitis, liver fibrosis, cirrhosis, and eventually hepatocellular carcinoma (HCC), and is also linked to an increased risk of cardiovascular and extrahepatic complications⁴. Current therapeutic strategies for MASLD mainly rely on lifestyle adjustments and dietary intervention; however, their clinical efficacy remains limited. Despite the growing research interest in MASLD, the lack of a clear understanding of its underlying mechanisms has hindered the development of effective clinical treatments and therapeutic agents^5,6.

An increasing number of studies have noticed the critical roles of endoplasmic reticulum stress (ERS) and mitophagy in the progression of MASLD. The endoplasmic reticulum (ER) is a key organelle responsible for protein modification and trafficking. Under stress conditions such as hypoxia, calcium homeostasis imbalance, and sustained metabolic overload, the ER becomes impaired, leading to the accumulation of misfolded proteins and activation of the unfolded protein response (UPR)⁷. In the early stage, ERS is initiated as a response to pathological conditions. However, persistent ERS can cause overactivation of the UPR, which in turn induces hepatic inflammation, hepatocyte injury, and MASLD progression⁸. Studies show that caveolin-1 (CAV1) can alleviate ERS by regulating farnesoid X receptor (FXR/NR1H4) signaling axis and cholesterol transporters ABCG5 and ABCG8, thus reducing hepatocyte death and improving the pathological features of MASLD⁹. These findings suggest that ERS dysregulation is not merely a passive consequence of hepatocellular injury but an active driver of liver damage and MASLD progression.

In addition to ERS, dysregulation of mitochondrial quality control, particularly impaired mitophagy, also plays a critical role in the progression of MASLD. Mitophagy is a selective autophagy process that limits oxidative stress, maintains mitochondrial homeostasis, and prevents excessive accumulation of lipids by removing damaged mitochondria and is essential¹⁰. As a core link of mitochondrial quality control, mitophagy is prone to dysfunction under chronic metabolic stress, which may weaken the adaptive capacity of hepatocytes. Emerging evidence indicates a bidirectional interplay between ERS and mitophagy. Prolonged ERS activation can impair mitochondrial structure and function, thereby inhibiting mitophagy and exacerbating metabolic dysregulation¹¹. Conversely, impaired mitophagy may aggravate ERS by promoting the accumulation of dysfunctional mitochondria and reactive oxygen species (ROS). This process results in sustained UPR activation and cellular stress, forming a vicious cycle. Therefore, the coordinated imbalance between ERS and mitophagy is increasingly recognized as a potential driver of MASLD development. However, the key regulatory molecules linking ERS and mitophagy, as well as the underlying mechanisms in MASLD, remain poorly understood.

In this context, increasing attention has been directed toward transcriptional regulators that integrate metabolic stress, cellular homeostasis, and immune responses. NR4A1 is an orphan nuclear receptor that participates in multiple biological processes, including lipid metabolism, inflammatory regulation, and mitochondrial function¹². Previous studies have shown that dysregulation of NR4A1 is closely associated with metabolic imbalance and chronic inflammation, which are key pathological features of the development of MASLD^13,14. Recent evidence also suggests that NR4A1 may be involved in the regulation of ERS and mitochondrial homeostasis^15,16. Under conditions of chronic metabolic stress of MASLD, the decrease in NR4A1 expression may reflect the impairment of cellular stress adaptation. However, whether NR4A1 is directly involved in the synergistic regulation of ERS and mitochondrial autophagy, and its specific mechanism of action in MASLD, remains to be further clarified.

Recent studies emphasize the important roles of ERS and mitophagy in MASLD progression. However, the specific mechanisms underlying their interplay remain unclear, and effective therapeutic strategies targeting these pathways are still lacking. Therefore, this study systematically investigates ERS and mitophagy as two key stress response pathways involved in MASLD pathogenesis. By integrating GEO data mining, machine learning approaches, and scRNA-seq analysis, we aimed to elucidate the potential mechanisms of ERS and mitophagy to mediate the progression of MASLD. Given the emerging druggability of NR4A1, this study identified candidate therapeutic compounds targeting ERS and mitophagy through high-throughput virtual screening. Our findings provide new insights into the development of therapeutic strategies for MASLD.

Methods

Data collection

A total of four microarray datasets (GSE89632, GSE63067, GSE48452, and GSE66676) from the Gene Expression Omnibus (GEO) were analyzed, including 115 MASLD patients and 79 healthy control samples. In addition, GSE135251 was obtained as external validation cohort. This dataset included 206 MASLD samples and 10 normal controls. All raw expression matrices underwent background correction, batch effect removal, and normalization prior to integration and analysis. HCC data were obtained from The Cancer Genome Atlas (TCGA) database. Gene lists associated with ERS and mitophagy were retrieved from the GeneCards database (https://www.genecards.org/) and the Molecular Signatures Database (MSigDB, https://www.gsea-msigdb.org/gsea/msigdb/index.jsp). These gene sets were intersected with DEGs identified from the integrated datasets to derive MASLD-ERS-M. Dataset details are summarized in Tables S1, S2.

Note: Although this study adopts the updated term MASLD, the GEO datasets analyzed (e.g., GSE89632, GSE63067, GSE48452, and GSE66676) were originally annotated under the previous nomenclature “NAFLD”, which is consistent with the diagnostic criteria at the time of data acquisition.

Functional enrichment analysis

The biological functions and signaling pathways associated with the candidate genes were assessed using Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses^17–19. For enrichment analysis, curated gene sets from MSigDB (c5 GO and c2 KEGG collections) were utilized. Pathway activity differences between high and low groups were evaluated via GSVA²⁰.

Machine learning analysis

Three machine learning algorithms, including LASSO, SVM-RFE, and RF, were applied to identify robust candidate genes associated with MASLD²¹ (Table S3). SVM-RFE uses R package “e1071” to distinguish the best features of different groups, and uses tenfold cross-validation as a resampling method to reorder features in each recursive feature elimination process to evaluate model performance. Additionally, RF algorithms were used to assess gene importance and predict their classification ability. By using the “randomForest” package, the classification variables were iteratively scored to find the features with high classification accuracy.

We evaluated the diagnostic value of the candidate genes using receiver operating characteristic (ROC) curves and area under the curve (AUC)²². Logistic regression and nomogram models were used to assess independent predictors and clinical applicability.

Immune infiltration correlation analysis

Normalized transcriptomic profiles were analyzed using the CIBERSORT algorithm to estimate the relative abundances of 22 immune cell types in the MASLD group and the normal group²³. In addition, correlations between the expression levels of the core genes and the degree of immune cell infiltration were calculated to assess the potential roles of these genes within the immune microenvironment.

scRNA

The scRNA-seq dataset GSE174748 was used the “Seurat” package for standard preprocessing of high-throughput sequencing data. Principal component analysis (PCA) was applied for initial dimensionality reduction, followed by Uniform Manifold Approximation and Projection (UMAP) for nonlinear dimensionality reduction and visualization. Cell type annotation was conducted, with reference to canonical marker genes obtained from the CellMarker database. The corresponding cell type annotations for each cluster are summarized in Table S4. We further analyzed the expression distribution and patterns of the key gene across different cell subsets. Intercellular communication in the MASLD microenvironment was inferred using the “CellChat” R package. Additionally, the “Monocle2” package was employed to construct pseudotime trajectories of macrophages and investigate dynamic expression changes of key genes during cellular transitions.

ceRNA network construction and molecular docking analysis

To identify potential upstream regulators of signature genes in MASLD, three online databases were used to predict miRNA targets: miRanda, miRDB and TargetScan. Interactions among mRNAs, miRNAs, and lncRNAs were further explored using the spongeScan database (Table S5). The resulting ceRNA regulatory network was visualized using Cytoscape²⁴.

The crystal structure of NR4A1 (PDB ID: 8Z5A) was obtained from the Protein Data Bank (PDB). Protein preparation was performed using PyMOL software. Candidate compounds were predicted using the Connectivity Map (CMap) platform, and their three-dimensional structures were retrieved from the PubChem database²⁵. The ligand position in the protein structure was defined as the binding pocket. Molecular docking was performed using AutoDock Vina²⁶. The docking results were visualized using PyMOL²⁷.

Cell culture

An in vitro MASLD model was established by treating HepG2 and Huh7 cells with 1 mM free fatty acids (FFA), consisting of a 2:1 mixture of oleic acid and palmitic acid, for 24 h. Subsequent cellular experiments were performed as described in the following sections. The HepG2 and Huh7 cell lines were kindly provided by Professor Yingyu Chen from Peking University Health Science Center and were obtained from the American Type Culture Collection (Manassas, VA, USA)²⁸.

Sample collection and IHC

Liver tissue specimens from human subjects were collected at Beijing Electric Power Hospital, State Grid Corporation of China. Informed consent was obtained from all participants, and the research protocol has been approved by the Medical Ethics Committee of Beijing Electric Power Hospital, State Grid Corporation of China (Approval No. 2025073010201). All procedures involving human participants were conducted in accordance with relevant guidelines and regulations and in compliance with the Declaration of Helsinki. A total of 8 normal liver tissue samples and 14 liver tissue samples from MASLD patients were obtained. Tissue microarrays were purchased from Outdo Biotech (Shanghai, China). Additionally, liver tissue sections from high-fat diet (HFD) induced MASLD mice and matched controls were kindly provided by Peking University Health Science Center. All animal experiments were approved by the Animal Ethics Committee of Peking University Health Sciences Center (Approval No. LA2022406) and were performed in accordance with relevant guidelines and regulations. Briefly, male C57BL/6J mice were randomly assigned to receive either a standard chow diet (control group) or a high-fat diet to induce MASLD, following established protocols. Mice were maintained on the respective diets for 12 weeks, after which they were anesthetized with isoflurane and euthanized by cervical dislocation. Liver tissues were then harvested for immunohistochemical analyses. Immunohistochemically stained sections were analyzed using ImageJ software for image processing and semi-quantitative scoring.

Western blotting and RT-qPCR

The following primary antibodies were used: NR4A1 (12235-1-AP, Proteintech) and GAPDH (60,004-1-Ig, Proteintech). HRP-conjugated anti-rabbit or anti-mouse secondary antibodies (Poly-HRP, Proteintech) were applied for detection. The specific primer sequences used for RT-qPCR are listed in Table S6.

Oil Red O (ORO) staining and intracellular triglyceride (TG) and total cholesterol (TC) measurement

ORO staining was performed using a commercial kit (Solarbio, China) according to the manufacturer’s instructions. Intracellular lipid droplet accumulation was visualized and recorded under a light microscope. For intracellular TG and TC detection, cells were rinsed with PBS, detached using trypsin, and collected by centrifugation. TG and TC contents were then quantified using specific assay kits (Solarbio, China).

Statistical analysis

All data analyses were conducted using R software (version 4.2.2) and relevant R packages. Statistical analysis was conducted using GraphPad Prism 10 software.

Results

Identification of DEGs in MASLD

We collected and integrated four MASLD-related transcriptomic datasets (GSE48452, GSE63067, GSE66676, and GSE89632) from the GEO database. Before batch correction, the expression distributions of samples across datasets exhibited substantial batch effects (Fig. S1A,C). After normalization and batch correction, the expression profiles became more consistent, as shown by the boxplot and PCA plot (Figs. 1A, S1B). A total of 506 DEGs were identified between MASLD and control groups based on the merged and normalized dataset (Fig. 1B). The expression patterns of the top 50 DEGs were visualized using a heatmap (Fig. 1C).

Fig. 1.

Identification of DEGs in MASLD. (A) PCA plot after batch correction. (B) Volcano plot of DEGs between MASLD and control groups. (C) Heatmap of the top 50 DEGs. PCA, principal component analysis. All visualizations were generated using R (version 4.2.2; https://www.r-project.org/).

Identification of ERS and mitophagy related DEGs and function enrichment analysis

Given the reported involvement of ERS and mitophagy in metabolic dysfunction and liver injury, genes related to these processes were further analyzed in MASLD. Gene sets related to ERS (n = 2600) and mitophagy (n = 1802) were obtained from GeneCards and MSigDB databases. Intersecting these with the identified DEGs yielded 15 overlapping genes implicated in MASLD, ERS, and mitophagy (MASLD-ERS-M) (Fig. 2A). Expression analysis revealed marked differences in these genes between MASLD and control groups (Fig. S1D). A PPI network construction identified NR4A1, P4HA1, and several others as central nodes within a highly interconnected network (Fig. 2B). Correlation heatmap analysis demonstrated strong positive and negative associations among several MASLD-ERS-M genes (Fig. 2C). In addition, a chord diagram was used to visualize co-expression patterns and potential synergistic relationships among these genes (Fig. 2D). GO and KEGG enrichment analyses showed that MASLD-ERS-M genes are primarily involved in metabolic and stress response pathways, including fatty acid metabolism, AMPK, and mTOR signaling (Fig. S2A–D). Both AMPK and mTOR pathways are central regulators of cellular energy homeostasis and autophagy, and have been implicated in the modulation of ERS and mitophagy. These findings suggest that MASLD-ERS-M genes are mainly involved in metabolic regulation and stress response signaling, indicating their potential involvement in MASLD.

Fig. 2.

Identification of MASLD-ERS-M genes. (A) Venn diagram showing the overlap among DEGs, ERS-related genes, and mitophagy-related genes. (B) PPI network of MASLD-ERS-M genes. (C) Correlation heatmap of MASLD-ERS-M genes. (D) Chord diagram showing gene co-expression patterns. PPI, protein–protein interaction. All visualizations were generated using R (version 4.2.2; https://www.r-project.org/).

Selection of key biomarkers of MASLD by machine learning

To identify robust gene features with predictive value, three machine learning methods were applied. LASSO regression selected eight potential biomarkers: NR4A1, P4HA1, FADS2, TUBB6, FLNC, RPS6KA1, PRKAA2, and IGF1 (Fig. 3A,B). SVM-RFE algorithm identified three candidate genes based on the minimum cross-validation error: P4HA1, NR4A1, and TUBB6 (Fig. 3C,D). The top five genes ranked by feature importance in the RF model were P4HA1, NR4A1, IGF1, PTGS2, and FADS2 (Fig. 3E,F). The markers selected by LASSO, SVM-RFE, and RF, together with their corresponding importance metrics, are summarized in Table S8. Comparison of the three machine learning models identified NR4A1 and P4HA1 candidate biomarkers with strong diagnostic performance (Fig. 3G). The chromosomal locations of NR4A1 and P4HA1 were mapped to chromosomes (Fig. 3H).

Fig. 3.

Machine learning analysis. (A) Ten-fold cross-validation curve for the LASSO model. (B) Coefficient profiles of the LASSO model. (C) Ten-fold cross-validation accuracy for SVM-RFE with different feature numbers. (D) Cross-validation error curve for SVM-RFE. (E) Relationship between the number of trees and the error rate in the RF model. (F) Feature importance ranking of MASLD-ERS-M genes in the RF model. (G) Venn diagram of genes identified by LASSO, SVM-RFE, and RF. (H) Circos plot of the chromosomal locations of NR4A1 and P4HA1. LASSO, least absolute shrinkage and selection operator; SVM-RFE, support vector machine recursive feature elimination; RF, random forest.

Analysis of key genes and diagnostic performance

The expression of NR4A1 and P4HA1 was significantly downregulated in MASLD patients compared to healthy controls (Fig. 4A). This expression pattern was consistently observed in the independent GSE135251 validation cohort (Fig. 4B). The combined diagnostic model demonstrated good performance, with AUC values of 0.836 in the training cohort and 0.887 in the validation cohorts (Fig. 4C,D). ROC analysis indicated that NR4A1 showed higher diagnostic accuracy than P4HA1 in both datasets, with AUC = 0.775 and 0.708 for NR4A1, compared to 0.769 and 0.702 for P4HA1 (Figs. 4E and S3A). A nomogram incorporating NR4A1 and P4HA1 was constructed as a visual tool for individualized MASLD risk prediction (Figs. 4F and S3B). Calibration curves indicated good calibration of the model (Figs. 4G and S3C). Decision curve analysis (DCA) showed demonstrated a net benefit (Figs. 4H and S3D). Logistic regression analysis demonstrated that NR4A1 and P4HA1 were independent predictors of MASLD. NR4A1 was identified as a significant protective factor in both univariate and multivariate models (OR < 1, P < 0.001) (Fig. S3E,F).

Fig. 4.

Expression and diagnostic performance of NR4A1 and P4HA1 in MASLD. (A,B) Expression levels of NR4A1 and P4HA1 in the training (A) and validation (B) cohorts. (C,D) ROC curves and AUC values for the model in the training (C) and validation (D) cohorts. (E) ROC curves of NR4A1 and P4HA1 in the training cohort. (F) Nomogram for MASLD risk prediction. (G) Calibration curve of the nomogram. (H) DCA of the nomogram. ROC, receiver operating characteristic; AUC, area under the curve; DCA, decision curve analysis.

Immune infiltration analysis and functional characterization of NR4A1

Immune cell infiltration patterns between MASLD and control groups were analyzed using the CIBERSORT algorithm based on liver transcriptome data from the GEO database (Fig. 5A,B). MASLD samples exhibited significantly reduced proportions of naïve B cells, plasma cells, activated dendritic cells, and activated mast cells, whereas resting NK cells, M2 macrophages, and resting mast cells were significantly increased (Fig. 5C).

Fig. 5.

Immune infiltration and functional enrichment analysis. (A) Immune cell composition estimated by CIBERSORT. (B) Correlation heatmap of immune cell populations. (C) Differential analysis of immune cell components in the MASLD datasets. (D) Correlation between NR4A1 expression and immune cell infiltration. (E,F) GO and KEGG analysis based on GSVA scores of NR4A1 high and low expression groups. GSVA, Gene Set Variation Analysis. All visualizations were generated using R (version 4.2.2; https://www.r-project.org/).

Correlation analysis further revealed that NR4A1 expression was positively correlated with activated mast cells, activated dendritic cells, and neutrophils, and negatively correlated with CD8⁺ T cells, M1 macrophages, and resting mast cells (Fig. 5D). Hallmark pathway analysis showed that NR4A1 expression was inversely correlated with lipid and bile acid metabolism, while positively correlated with ERS, oxidative stress, and inflammatory signaling (Fig. S5A,B).

In order to further explore the biological pathways related to NR4A1, we conducted GSVA. In the NR4A1 high-expression group, multiple molecular functions related to ubiquitin-dependent protein degradation and mitochondrial activity were significantly enriched, such as “regulation of ubiquitin protein ligase activity” and “mitochondrial fusion” (Fig. 5E). Moreover, multiple pathways related to metabolic and stress response were also enriched, including the tricarboxylic acid (TCA) cycle, homologous recombination, mismatch repair, peroxisome, and pyrimidine metabolism, suggesting that NR4A1 may exert a protective role in maintaining cellular homeostasis and adaptive stress responses (Fig. 5F).

Experimental validation of NR4A1

To validate the bioinformatic findings, a MASLD cell model was established by treating HepG2 and Huh7 cells with FFA. After 24 h, ORO staining revealed substantial lipid droplet accumulation in both cell lines (Fig. 6A). Intracellular TG and TC levels were significantly elevated (Fig. 6B,C), confirming successful model induction. Western blotting and RT-qPCR consistently demonstrated a significant reduction of NR4A1 expression at both protein and mRNA levels after FFA treatment (Fig. 6D,E).

Fig. 6.

Experimental validation of NR4A1. (A) ORO staining of lipid accumulation in HepG2 and Huh7 cells. (B) Quantification of intracellular TG levels in HepG2 and Huh7 cells. (C) Measurement of TC levels in both cell lines. (D) Western blot analysis of NR4A1 protein expression. (E) RT-qPCR analysis of NR4A1 mRNA expression in HepG2 and Huh7 cells. Scale bar = 200 μm. ORO, Oil Red O; TG, triglyceride; TC, total cholesterol; FFA, free fatty acids.

In vivo, an HFD-induced MASLD mouse model was employed. ORO staining showed marked lipid accumulation in liver, and IHC staining showed that the expression of NR4A1 in liver tissues of MASLD mice was significantly reduced compared with normal controls (Fig. 7A).

Fig. 7.

NR4A1 expression in mouse models, patient samples, and commercial tissue microarrays. (A) ORO staining and IHC in control and MASLD mice. (B) NR4A1 mRNA expression in liver tissues from MASLD patients measured by RT-qPCR. (C) NR4A1 IHC in liver tissues from MASLD patients. (D) NR4A1 IHC in a liver tissue microarray, including normal liver, MASLD, cirrhosis, and HCC tissues. (E) NR4A1 expression in normal and tumor tissues. (F) Kaplan–Meier survival curves stratified by NR4A1 expression (P = 0.017). Scale bar = 200 μm. HFD, high-fat diet.

In addition, we validated NR4A1 expression in liver tissue samples of clinical patients collected by our hospital. RT-qPCR showed a significant decrease in NR4A1 mRNA levels in patient samples (Fig. 7B), and IHC analysis also confirmed a significant downward regulation of NR4A1 protein expression (Fig. 7C). Then we analyzed a liver tissue microarray containing normal liver tissue, MASLD, cirrhosis, and HCC. The results showed that NR4A1 expression progressively decreased from normal liver to HCC (Fig. 7D).

The TIMER database was used for pan-cancer analysis. The results showed that NR4A1 was expressed differently in a variety of tumor types, and the expression level in normal liver tissues was significantly higher than that in LIHC samples (Fig. S5C). This trend was validated in the TCGA-LIHC dataset, where NR4A1 expression was consistently lower in tumor tissues, including in matched tumor-normal paired samples (Figs. 7E, S5D). Kaplan–Meier survival analysis showed that patients with higher NR4A1 expression had significantly poorer overall survival (Fig. 7F), implying a context dependent, regulatory role of NR4A1 in hepatocarcinogenesis.

scRNA-seq reveals potential mechanisms of NR4A1-mediated MASLD progression

To further elucidate the cell-type-specific roles of NR4A1 in MASLD, we performed scRNA-seq analysis using the GSE171748 dataset. After quality control and dimensionality reduction by PCA and UMAP, 16 clusters were identified and subsequently annotated into nine major cell types based on SingleR algorithm and canonical marker genes (Figs. 8A, S4A–E). UMAP visualization revealed differences in the overall cell distribution between MASLD and control samples (Fig. 8B). The number of NR4A1-positive cells was higher in the control group (Fig. 8C). Among all cell types, NR4A1 expression was primarily enriched in endothelial cells, macrophages, and fibroblasts (Fig. 8D), and was globally downregulated in the MASLD group compared with controls (Fig. 8E).

Fig. 8.

ScRNA-seq analysis of NR4A1 in MASLD. (A) UMAP visualization of annotated cell types. (B) UMAP distribution of cells from control and MASLD samples. (C) UMAP feature plot of NR4A1 expression, split by control and MASLD groups. (D) NR4A1 expression across major cell types. (E) Comparison of NR4A1 expression between groups (P < 0.0001).

Cell–cell communication patterns were analyzed using the CellChat framework. The overall communication network showed widespread signaling interactions among hepatocytes, macrophages, monocytes, epithelial cells, endothelial cells, and fibroblasts (Fig. 9A). Focusing on macrophages, we found that they transmitted strong outward signals to endothelial cells, epithelial cells, fibroblasts, and monocytes (Fig. 9B). At the ligand-receptor level, key signaling pathways were identified, including VEGFA-VEGFR1/2, NAMPT-INSR, and TNF-related pathways (Fig. 9C). These pathways are known to regulate ERS and mitochondrial function.

Fig. 9.

Cell–cell communication and pseudotime trajectory analysis of NR4A1 in MASLD. (A) Overall intercellular communication networks inferred by CellChat in the control and MASLD groups. The left panel shows the number of interactions (count), and the right panel represents the interaction strength (weight). (B) Outgoing communication patterns of macrophages in the MASLD group. (C) Bubble plot of macrophage-related ligand-receptor pairs differentially enriched between groups, including TNF, VEGF, NAMPT, and CCL pathways. (D) Pseudotime trajectory of macrophages. (E) Distribution of macrophage subpopulations along the trajectory colored by pseudotime (left) and cluster identity (right). (F) NR4A1 expression along pseudotime in macrophages. (G) NR4A1 expression trends along pseudotime.

We then utilized the Monocle algorithm to construct pseudotime trajectories. The resulting trajectories showed a clear branching pattern, with multiple cell states and transition checkpoints distributed along the pseudotime axis (Figs. 9D, S4F). When stratified by NR4A1 expression, cells with high NR4A1 levels were mainly located at early stages, whereas those with low expression were more prevalent in later phases (Fig. 9E,F). Furthermore, overall NR4A1 expression progressively declined along the pseudotime axis (Fig. 9G), indicating that NR4A1 may participate in early macrophage activation and become transcriptionally silenced during later differentiation, potentially contributing to MASLD progression.

ceRNA network and candidate therapeutic drug

To investigate the upstream regulatory mechanisms of the key gene NR4A1, we established a ceRNA regulatory network (Fig. 10A). Based on integrative predictions from miRanda, miRDB, and TargetScan databases, two microRNAs, hsa-miR-665 and hsa-miR-342-5p, were identified as potential regulators of NR4A1. Furthermore, 17 upstream lncRNAs were predicted to interact with these miRNAs. Together, these components constitute a classical lncRNA-miRNA-mRNA regulatory axis.

Fig. 10.

ceRNA network and candidate therapeutic compounds targeting NR4A1. (A) ceRNA network centered on NR4A1. (B–E) Molecular docking simulation results showing the binding interactions between NR4A1 and four candidate compounds: (B) Cinobufagin, (C) Digitoxin, (D) Ouabain, and (E) Proscillaridin. All compounds exhibited favorable binding affinities with NR4A1 (binding energy < − 5.0 kcal/mol).

Candidate compounds targeting NR4A1 were predicted using the CMap database (Table S7). Based on connectivity scores, four representative compounds, cinobufagin, digitoxin, ouabain, and proscillaridin, were selected for further analysis. Molecular docking results indicated that all four compounds exhibited favorable binding affinities with NR4A1, with binding energies of -7.2, -8.6, -7.6, and -8.2 kcal/mol. All four compounds interacted with NR4A1 near the ARG-123 and TYR-122 residues, suggesting that these regions may be involved in ligand binding (Fig. 10B–E).

Based on these findings, we constructed a schematic diagram to illustrate the proposed mechanism (Fig. 11).

Fig. 11.

Schematic diagram of the proposed mechanism.

Discussion

MASLD has become a major global health concern, with increasing prevalence and limited effective therapeutic options^28,29. ERS and mitophagy play essential roles in maintaining hepatic metabolic homeostasis and adapting to metabolic overload, both of which are related to the development of MASLD^30,31. However, the current systematic research on the synergistic disorder of ERS and mitophagy in MASLD is still insufficient, and the key regulatory molecules connecting these two stress response pathways have not been clarified. In this study, we integrated multiple GEO transcriptomic datasets and applied machine learning approaches to systematically identify MASLD-associated genes related to ERS and mitophagy (MASLD-ERS-M). Among these genes, NR4A1 exhibited a relatively consistent expression pattern across independent cohorts and showed a significant relevance to MASLD in multiple predictive models. Further validation at multiple levels demonstrated a consistent downregulation of NR4A1 in MASLD experimental models and liver tissue samples. These findings suggest that NR4A1 may be involved in metabolic stress imbalance during MASLD progression and may participate in the coordinated regulation of ERS and mitophagy.

NR4A1 is an orphan nuclear receptor whose transcriptional regulatory activity does not depend on classical ligand binding. On the contrary, it responds to various endogenous and exogenous signals through its zinc finger and ligand binding domains and regulates downstream gene expression by interacting with transcription factors, signaling modulators, and coregulatory proteins³². NR4A1 has gained attention for its role in metabolic liver diseases. Another research demonstrated that NR4A1 can suppress the expression of lipogenic genes and promote fatty acid oxidation, thereby contributing to lipid homeostasis in hepatocytes^33,34. NR4A1 has also been identified as an important modulator of ERS. By regulating ERS-related signaling pathways, NR4A1 can partially alleviate cellular stress and reduce the risk of apoptosis³⁵. Growing evidence has highlighted the involvement of NR4A1 in organelle homeostasis. At the mitochondrial level, NR4A1 has been reported to influence the assembly of p62 and SQSTM1-associated complexes through phase separation, thereby mediating drug-induced mitophagy and suggesting a noncanonical role in mitochondrial quality control³⁶. In addition, NR4A1 can promote mitochondrial fission via Fis1 and suppress Parkin-mediated mitophagy, contributing to the regulation of mitochondrial homeostasis under stress conditions of myocardial ischemia reperfusion injury³⁷. These findings indicate that NR4A1 does not regulate mitophagy in a unidirectional manner, but rather acts as a context-dependent regulator of mitochondrial quality control³⁷. Nevertheless, its precise role in ERS- and mitophagy-mediated MASLD progression remains to be elucidated. In this study, we observed a significant downregulation of NR4A1 at both the mRNA and protein levels in FFA-induced steatosis hepatocyte models, accompanied by increased intracellular TG and TC levels. In liver tissues from MASLD mice and patients, NR4A1 expression was also markedly reduced in MASLD samples. Furthermore, analysis of a tissue microarray containing samples from MASLD, hepatitis and HCC showed a progressive decrease in NR4A1 expression during liver disease progression. These findings suggest that, as a key regulator of metabolic homeostasis, stress response, and immune regulation, sustained downregulation of NR4A1 may contribute to a shift of hepatocytes from adaptive metabolism toward persistent stress, and may be involved in fibrotic remodeling and even malignant transformation.

Cell type-specific remodeling of the immune microenvironment is considered a key feature of MASLD progression. Accumulating evidence indicates that macrophages and CD8⁺ T cells infiltrate extensively during MASLD progression and contribute to the maintenance of chronic inflammation and disease advancement³⁸. Consistent with these findings, our immune infiltration analysis showed that low NR4A1 expression was significantly associated with increased proportions of macrophages and CD8⁺ T cells, suggesting a close relationship between NR4A1 expression and immune microenvironment alterations in MASLD. scRNA-seq analysis further found that NR4A1 expression was markedly reduced in macrophages, fibroblasts, and endothelial cells in MASLD samples. Notably, sustained downregulation of NR4A1 in macrophages suggests a potential link with changes in macrophage functional states and a possible role in regulating macrophage immune phenotypes. Pseudotime analysis further revealed a gradual decrease in NR4A1 expression during macrophage differentiation and state transitions. This finding suggests that NR4A1 may primarily function during early immune activation or the adaptive response phase and becomes progressively suppressed as disease advances. In addition, fibroblasts and endothelial cells are known to influence tissue repair and intercellular signaling, thereby indirectly promoting fibrotic remodeling and disease progression³⁹. Taken together, sustained downregulation of NR4A1 may be associated with immune imbalance and enhanced cellular stress during MASLD progression. It may further affect ERS and mitochondrial quality control processes, including PINK1- and Parkin-related mitophagy pathways. These changes may aggravate oxidative stress and mitochondrial dysfunction, thereby collectively promoting the transition of MASLD from simple steatosis to MASH and fibrotic stages. However, the precise molecular mechanisms underlying these processes require further investigation.

At present, MASLD lacks clearly defined molecular targets and approved disease-specific pharmacological therapies, and targeted interventions against key pathogenic pathways are under active investigation. Given the regulatory role of NR4A1 in ERS and mitophagy, its feasibility as a potential therapeutic target has attracted increasing attention. For example, Bruceine A has been reported to enhance NR4A1 function by stabilizing the NR4A1 protein⁴⁰, whereas the indoquinazoline derivative R17 suppresses triglyceride accumulation in adipocytes through activation of NR4A1⁴¹. However, there are still no clinically available NR4A1-targeted stabilizers. The regulatory function of NR4A1 in ERS and mitophagy makes it a potential therapeutic target for intervening in MASLD progression. Therefore, we further performed high-throughput virtual screening to identify potential NR4A1 stabilizers. Four candidate compounds were identified from the CMap, including cinobufagin, digitoxin, ouabain, and proscillaridin. Interestingly, all four compounds bound to NR4A1 at the ARG-123 and TYR-122 residues, suggesting that these sites may be critical for compound-protein interactions. These candidates belong to the cardiac glycoside family and possess a conserved steroid core structure. Both residues capable of interacting with the shared steroidal backbone of the compounds, which may help explain their similar binding patterns with NR4A1. Despite these structural similarities, they can be further classified into two subtypes: cardenolides (digitoxin and ouabain) and bufadienolides (cinobufagin and proscillaridin), which primarily differ in their lactone ring structures. Cardiac glycosides can modulate Na⁺/K⁺ ATPase activity, disturb cellular ion homeostasis, and induce oxidative stress and metabolic remodeling^42,43. By identifying four potential compounds for NR4A1, our study provides lead compounds for the development of NR4A1 stabilizers, laying a strong foundation for subsequent NR4A1-targeted drug development.

This study expands current knowledge of the biological functions of NR4A1 through multiple analytical levels. Early studies mainly focused on the immunoregulatory and metabolic roles of NR4A1 in cardiovascular and autoimmune diseases, whereas the study of its participation mechanism in metabolic liver disease was relatively insufficient. We reveal a dynamic pattern of NR4A1 expression that is higher in the early stage of macrophage development and gradually decreases. These results broaden our understanding of NR4A1’s involvement in ERS- and mitophagy-mediated MASLD progression. Although NR4A1 was mainly evaluated in liver tissues in this study, it may be used as a molecular indicator related to the progression of the disease and could supplement histopathological assessment rather than replace liver biopsy. Future studies are warranted to determine whether NR4A1 or its downstream signatures can be detected in peripheral blood, which may help develop minimally invasive diagnostic strategies. There are still some limitations in this study. (1) Although multiple MASLD datasets were integrated, a larger sample size is still required to further evaluate the wide application of our findings; (2) while the crucial role of NR4A1 in the progression of MASLD mediated by ERS and mitophagy has been identified, further studies are still needed to clarify its deeper regulatory mechanisms through transcriptomic profiling, siRNA mediated knockdown, overexpression assays, and functional in-depth research including rescue experiments; (3) Although four compounds with potential binding capacity to NR4A1 were identified and the putative binding residues were preliminarily characterized, further studies involving site-directed mutagenesis are required to determine the functional importance of these residues.

Overall, our findings suggest that NR4A1 contributes to MASLD progression through regulation of ERS, mitophagy-related processes, lipid metabolic reprogramming, and immune microenvironment remodeling. These results support NR4A1 as a MASLD-related regulatory molecule and provide a foundation for future studies exploring NR4A1-centered therapeutic strategies and drug development targeting ERS and mitophagy pathways.

Conclusion

By integrating multi-omics analyses and experimental validation, this study preliminarily identified NR4A1 as a key regulatory molecule associated with MASLD. Changes in NR4A1 expression were closely related to ERS and mitophagy-related processes. These findings provide a theoretical basis for further investigations into NR4A1-centered regulatory mechanisms and the development of potential therapeutic strategies.

Author contributions

Zhang Z.M. conceptualized and designed the study, acquired funding, and supervised the research. Chen Q.R., Liu L.M. and Zhang Z.M. collected the experimental data and drafted the manuscript. Chen Q.R. and Feng J.Q. performed data analysis and prepared the figures. Liu L.M. and Zhang C. interpreted the results. All authors contributed to the revision of the manuscript and approved the final submitted version.

Funding

This work was supported by the Beijing Municipal Science & Technology Commission (No. Z171100000417056) and the Key Support Project of Guo Zhong Health Care of China General Technology Group (GZKJ-KJXX-QTHT-20230626, 20240429).

Data availability

The data sets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Compliance with ethical Standards

The data used in this study were obtained from the GEO database, and all data usage strictly adhered to the terms and ethical guidelines provided by the data sources. Mouse liver tissue sections used in this study were obtained from previously established MASLD mouse models provided by Peking University Health Science Center. The original animal experiments were approved by the Animal Ethics Committee of Peking University Health Sciences Center (Approval No. LA2022406). Clinical samples were collected under the approval of the Medical Ethics Committee of Beijing Electric Power Hospital, State Grid Corporation of China (Approval No. 2025073010201). All procedures were conducted in accordance with relevant guidelines and regulations.

Competing interests

The authors declare no competing interests.