Identification and validation of brown adipocyte-related key genes in ST-segment elevated myocardial infarction

Abstract

ST-segment elevation myocardial infarction (STEMI), a critical manifestation of coronary artery disease, involves complex molecular mechanisms potentially linked to brown adipocyte (BA) biology. This study aimed to identify BA-related key genes and elucidate their roles in STEMI progression. Transcriptomic data from peripheral blood mononuclear cells (PBMCs) of STEMI patients and stable coronary artery disease (CAD) controls were obtained from the GEO database (GSE59867, GSE62646). A 102-gene brown adipocyte-related gene (BARG) set was curated from published literature. Differentially expressed genes (DEGs) were identified using limma, and weighted gene co-expression network analysis (WGCNA) combined with single-sample gene set enrichment analysis (ssGSEA) was applied to detect BARG-associated module genes. The intersection of DEGs and module genes yielded 20 candidate genes, which were further refined using LASSO regression and SVM-RFE machine learning algorithms, identifying ASGR2, MCEMP1, and PPARG as key genes consistently upregulated in STEMI across both datasets. A nomogram model incorporating these three genes achieved an area under the ROC curve (AUC) of 0.973, demonstrating excellent diagnostic performance. Gene set enrichment analysis (GSEA) revealed convergent enrichment of all three key genes in neutrophil degranulation and Toll-like receptor signaling cascades. Immune infiltration profiling using CIBERSORT identified significant increases in monocytes alongside decreases in resting NK cells and eosinophils in STEMI, with all three key genes showing significant correlations with these immune cell populations. Molecular docking identified shinpterocarpin as a candidate compound targeting PPARG with a suggestive binding affinity (-5.8 kcal/mol), warranting further experimental validation. RT-qPCR in an independent preliminary clinical cohort (n = 5 per group) demonstrated upregulation of all three key genes in STEMI patients (p < 0.05), consistent with bioinformatic predictions. This study identifies ASGR2, MCEMP1, and PPARG as BA-associated key genes in STEMI PBMC profiles and highlights their potential immune-metabolic relevance.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13019-026-04649-x.

Introduction

ST-segment elevation myocardial infarction (STEMI), the most severe manifestation of acute coronary syndromes (ACS), results from acute coronary artery occlusion, leading to myocardial ischemia and subsequent cardiomyocyte necrosis [1]. Over recent decades, the global incidence of cardiovascular diseases, particularly STEMI, has risen significantly [2], posing a serious threat to public health and long-term patient survival. The pathogenesis of STEMI is multifactorial. The predominant mechanism involves the accumulation of cholesterol and other lipids within the arterial wall, leading to luminal narrowing and endothelial injury. Upon plaque rupture or erosion, platelet aggregation and coagulation cascade activation occur, culminating in thrombus formation and coronary blockage. Additional contributing factors include myocardial oxygen supply-demand imbalance arising from conditions such as severe anemia, sepsis, or coronary vasospasm [3].

The clinical diagnosis of STEMI is primarily based on persistent symptoms of myocardial ischemia and electrocardiographic (ECG) findings. The hallmark ECG features include ST-segment elevation with or without pathological Q waves; however, these changes may not be evident in the early stages of the condition. Moreover, some patients exhibit transient ST-segment elevation, which should be distinguished from ST-segment resolution following successful reperfusion therapy [4]. The biomarkers of cardiomyocyte injury are widely used in the clinical diagnosis of STEMI. The release of myocardial enzymes from necrotic cardiomyocytes typically occurs several hours after the onset of AMI. Given the time-critical nature of STEMI, reperfusion therapy must be initiated immediately after diagnosis. Current guidelines recommend emergency percutaneous coronary intervention (PCI) as the first-line treatment for achieving timely reperfusion [5], with anticoagulation and antiplatelet pharmacotherapy serving as crucial adjuncts in guideline-directed management. Despite these advances, there remains a pressing need to identify novel molecular biomarkers to improve the early diagnosis and prognostic evaluation of STEMI, and to explore additional therapeutic targets to optimize individualized treatment strategies.

Brown adipose tissue (BAT) is a specialized thermogenic organ that converts chemical energy into heat through non-shivering thermogenesis, playing a pivotal role in systemic energy homeostasis and adaptive thermoregulation [6]. Distinct from white adipose tissue (WAT) in both structure and function, BAT contains multilocular lipid droplets, abundant mitochondria, and uniquely expresses uncoupling protein 1 (UCP1) [7, 8]. Beyond its metabolic role, accumulating evidence suggests that BAT exerts significant cardiovascular protective effects through multiple mechanisms. Firstly, BAT has been shown to effectively mitigate myocardial ischemia-reperfusion (IR) injury. Animal studies demonstrate that TGF family proteins secreted by BAT following IR injury activate the SMAD1/5 phosphorylation signaling pathway, effectively suppressing caspase-mediated apoptosis in cardiomyocytes. This mechanism contributes to a substantial reduction in infarct size-by approximately 35%-along with significantly lowered levels of plasma cardiac troponin I (cTnI) [9]. Secondly, BAT plays an active role in modulating myocardial metabolism. Specifically, BAT-secreted 12,13-diHOME, an oxidized metabolite of linoleic acid, enhances cardiac contractility by promoting Ca²⁺ cycling through the NOS1-RyR2 pathway [10]. Concurrently, it facilitates fatty acid uptake and restores energy metabolism predominantly reliant on fatty acid β-oxidation, thereby reducing lactate accumulation and improving metabolic efficiency [10]. Moreover, BAT transplantation has been shown to attenuate myocardial infarction-induced left ventricular hypertrophy and inhibit adverse ventricular remodeling, an integral pathological process in heart failure progression, thus preserving cardiac structural integrity and functional performance [11]. These findings collectively suggest that there may be a link between BAT-derived molecular signals and cardiometabolic and inflammatory regulation, providing a theoretical basis for exploring brown adipocyte-related molecular characteristics in STEMI.

Although these findings provide important evidence supporting the protective role of BAT in STEMI, the precise molecular mechanisms underlying this relationship and their potential for clinical translation require further elucidation. In this study, we integrated PBMC transcriptomic datasets with machine learning algorithms to identify key genes associated with brown adipocyte-related molecular features in STEMI. We further elucidated the biological role of these candidate genes and characterized their interactions with the immune microenvironment, constructed molecular regulatory docking and RT-qPCR validation to verify gene expression profiles. A nomogram diagnostic model incorporating the identified key genes was also developed and evaluated. This research aims to provide novel insights into STEMI pathogenesis and to lay a foundation for developing improved diagnostic biomarkers and targeted therapeutic strategies.

Materials and methods

Data acquisition

Gene expression profiles for the training cohort were obtained from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo) using the GSE59867 dataset (GPL6244). This dataset contained transcriptomic data from peripheral blood mononuclear cells (PBMCs) of 8 ST-segment elevation myocardial infarction (STEMI) samples and 46 stable coronary artery disease (CAD) controls. The validation dataset GSE62646 (GPL6244), also retrieved from GEO database, comprised PBMCs transcriptomic data from 28 STEMI samples and 14 stable CAD controls. A curated list of 102 brown adipocyte-related genes (BARGs) was assembled based on published literature [12] (Supplementary Table S1).

Identification of differentially expressed genes (DEGs)

Using the “limma” R package (v 3.56.2) [13], we screened for DEGs between STEMI and stable CAD groups in the GSE59867 dataset, considering genes with |log2FC| > 0.5 and adjusted p < 0.05 as significant. A volcano plot was then plotted using the “ggplot2” package (v 3.5.1) [14] to display the results. Expression patterns of DEGs were illustrated using the R package “ComplexHeatmap” (v 2.14.0) [15].

Comparative analysis of BARGs scores

To calculate BARGs scores for all samples in GSE59867 dataset using the ssGSEA algorithm implemented in the R package “GSVA” (v 1.53.28) [16]. Differences in BARGs scores between STEMI samples and stable CAD controls were statistically assessed employing the Wilcoxon test (p < 0.05).

Weighted gene co-expression network analysis (WGCNA)

To detect BARGs score-associated module genes, WGCNA was conducted on GSE59867 dataset using the R package “WGCNA” (v 1.73) [17]. Initially, the “GoodSamplesGenes” function was applied to cluster all samples in the GSE59867 dataset, allowing the detection and removal of outlier samples. Subsequently, the optimal soft threshold (power) was selected by analyzing the scale-free topology fit index and mean connectivity. A fit index of R² = 0.9 was targeted across the tested power values ranging from 1 to 20. The selected power value ensured the constructed network adhered to scale-free topology criteria. The adjacency matrix was converted into a topological overlap matrix (TOM) to assess gene-gene connectivity. Subsequently, genes were clustered hierarchically with dynamic tree cutting, setting a minimum module size of 100. Finally, highly similar modules (height cutoff = 0.3) were merged. Module-trait associations were assessed by calculating Spearman correlation coefficients (cor) between module eigengenes and BARGs scores, modules showing significant correlations were obtained (p < 0.05, |cor| > 0.3). Significant modules were graphicalized through a correlation heatmap created using the R package “pheatmap” (v 1.0.12) [18]. To visualize the association between module membership (MM, representing gene-module correlations) and gene significance (GS, reflecting gene-trait correlations), scatter plots were generated for genes within key modules using the R package “ggplot2” (v 3.5.1). BA-related module genes (BARMGs) were identified based on MM > 0.6 and GS > 0.4.

Identification of candidate genes

The intersection between DEGs and BARMGs was visualized using the R package “ggvenn” (v 0.1.9) [19]. Overlapping genes were defined as candidate genes.

Functional enrichment analysis and protein-protein interaction (PPI) network

To investigate the biological functions of candidate genes in STEMI progression, we performed Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses using the “clusterProfiler” package (v 4.15.0) [20]. GO analysis was conducted across three functional categories, comprising biological process (BP), cellular component (CC), and molecular function (MF). Enrichment terms with adjusted p values < 0.05 were considered statistically significant. The PPI network for candidate gene-encoded proteins was constructed using the STRING database (http://string-db.org) with a confidence score ≧ 0.15.

Machine learning

To refine candidate genes, least absolute shrinkage and selection operator (LASSO) regression and support vector machine-recursive feature elimination (SVM-RFE) algorithms were applied. For LASSO regression analysis, R package “glmnet” (v 4.1–8) [21] was utilized on the GSE59867 dataset. The LASSO algorithm compressed genes with weaker disease relevance to zero while retaining those with stronger associations by imposing a penalty parameter. A 10-fold cross-validation was employed to determine the optimal penalty parameter (lambda) for the LASSO regression. Two values were selected: lambda.min, which gives the minimum cross-validation error, and lambda.1se, which provides the most regularized model within one standard error of the minimum. The lambda.min value was selected to construct the final LASSO model, yielding a subset of feature genes. For SVM-RFE analysis, the R package “e1071” (v 1.7–16) [22] was employed. This backward elimination algorithm iteratively removed features with the lowest SVM-derived importance scores while retraining the model with the remaining features. The process continued until the feature subset achieving maximum classification accuracy and minimum classification error rate was identified. Overlapping genes derived from both LASSO regression and SVM-RFE analyses were extracted as feature genes.

Expression validation of feature genes

To validate the expression patterns of the feature genes, we analyzed both the GSE59867 and GSE62646 datasets, comparing expression levels between STEMI samples and stable CAD controls using the Wilcoxon test (p < 0.05). Feature genes demonstrating congruent expression trends and statistically significant differences in both datasets were selected as key genes for subsequent analyses. Visualization of expression patterns facilitated by the “ggplot2” package (v 3.5.1).

Nomogram construction and validation

To establish a predictive model for STEMI risk assessment, a nomogram was assembled based on the identified key genes using the R package “rms” (v 6.8-1) [23]. The nomogram integrated the contributions of each key gene to STEMI risk through multivariate regression analysis, assigning points proportional to their regression coefficients. Total points were calculated by summing individual gene points, enabling the visualization of the relationship between genetic predictors and disease probability. We evaluated the nomogram’s predictive performance using calibration curves generated with the “rms” package (v 6.8-1), which graphically compared predicted probabilities with observed outcomes. The goodness-of-fit (GoF) residual test was implemented to quantify the discrepancy between predicted and actual values (p > 0.05). Clinical applicability of the nomogram was validated through clinical impact curves generated by the R package “rmda” (v 1.6) [24]. The diagnostic performance of the nomogram was further evaluated using receiver operating characteristic (ROC) curves generated with the “pROC” package (v 1.18.5) [25]. The area under the curve (AUC) was estimated to quantify discrimination accuracy (AUC > 0.7), where higher AUC values (closer to 1) reflected superior predictive capability.

Gene set enrichment analysis (GSEA)

To elucidate the biological functions and pathways associated with the key genes, we performed Spearman correlation analyses between each key gene and all other genes in the GSE59867 dataset using the “psych” package (v 2.4.6) [26]. Cor values were calculated for every key gene-gene pair, followed by ranking genes based on their absolute cor values for each key gene. GSEA was subsequently conducted using “clusterProfiler” package (v 4.15.0). Gene set enrichment analysis employed the “c2.cp.kegg.v7.4.symbols.gmt” collection from Molecular Signatures Database (MSigDB) (https://www.gsea-msigdb.org) as the reference, with significantly enriched pathways defined by |NES| > 1, false discovery rate (FDR) < 0.25, and p < 0.05.

Immune infiltration analysis

Immune cell infiltration in the GSE59867 dataset was profiled using CIBERSORT to estimate the relative proportions of 22 immune cell types in STEMI samples and stable CAD controls. The resulting distributions were visualized using stacked bar plots created with the “ggplot2” package (v 3.5.1). To identify differentially infiltrated immune cells between two groups, Wilcoxon tests were performed (p < 0.05). Correlations among differentially infiltrated immune cells were assessed via Spearman correlation analysis using “psych” package (v 2.4.6) (|cor| > 0.3, p < 0.05). Correlation matrices were graphically represented as heatmaps through the R package “corrplot” (v 0.95) [27]. To explore associations between key genes and differentially infiltrated immune cells, we calculated correlation coefficients (|r| > 0.3, p < 0.05).

Chromosomal localization, friends’ analysis, and correlation analysis

The chromosomal positions of key genes were mapped using the R package “RCircos” (v 1.2.2) [28] based on annotation information extracted from the Homo sapiens.GRCh37.87.gtf reference file. Gene functional similarity among key genes was evaluated using the R package “GOSemSim” (v 2.33.0) [29] through Friends analysis. Cor values between key genes were computed using the “psych” package (v 2.4.6) across STEMI samples and stable CAD controls in the GSE59867 dataset (|cor| > 0.3, p < 0.05).

Gene interaction network and miRNA-lncRNA prediction

Functional interaction networks were first assembled based on GeneMANIA (http://genemania.org/), which integrates genomic and proteomic data to identify functionally similar genes. For upstream regulatory analysis, microRNAs (miRNAs) interacting with key genes were obtained by applying StarBase (http://starbase.sysu.edu.cn/). The top three miRNAs based on TDMDScore were selected for further investigation. In addition, long non-coding RNAs (lncRNAs) interacting with these miRNAs were detected, retaining the top 20 lncRNAs per miRNA. The predicted interaction network encompassing key genes, miRNAs, and lncRNAs was visualized using Cytoscape software (v 3.7.2) [30].

Therapeutic compounds prediction and molecular docking

To identify potential therapeutic compounds targeting key genes in the treatment of STEMI, compounds were detected using the Drug-Gene Interaction Database (DGIdb, https://dgidb.genome.wustl.edu/). For molecular docking, 3D structures of predicted compounds were gathered from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/). Corresponding protein structures of key genes were downloaded from the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB PDB, https://www.rcsb.org/). Structural preprocessing was performed using AutoDockTools 1.5.6 (https://autodock.scripps.edu/), including removal of water molecules, addition of polar hydrogens. Processed structures were converted into PDBQT format. A rigid-box docking protocol was implemented using AutoDock software (v 1.2.0) [31]. Molecular docking was performed using the Lamarckian Genetic Algorithm (LGA) to identify the optimal binding pose. Following grid preparation with autogrid4, the results were visualized using PyMol (v 2.5) [32].

Reverse transcription-quantitative PCR (RT-qPCR)

Peripheral blood samples from STEMI patients and stable CAD control individuals were collected at the Second Hospital of Tianjin Medical University. The study received ethical approval from medical Ethics Committee of The Second Hospital of Tianjin Medical University. Total RNA was isolated from samples (n = 5 per group) using TRIzol reagent (Vazyme, R401-01, China) following the protocol of the manufacturer. Following the assessment of RNA concentration and purity with a NanoPhotometer N50 (Implen), cDNA was synthesized using the HiScriptIII 1 st Strand cDNA Synthesis SuperMix Kit (Yeasen Biotechnology, 11141ES60). cDNA products were diluted 5–20-fold with DNase/RNase-free water before qPCR. Quantitative PCR was performed using Servicebio^® 2×Universal Blue SYBR Green Master Mix (G3326-05) on a CFX Connect Real-Time PCR System (BIO-RAD). Each 10 µL reaction contained 3 µL cDNA, 5 µL Master Mix, and 1 µL each of forward and reverse primers (10 µM; sequences in Supplementary Table S2). The thermal protocol was 95 °C for 1 min; 40 cycles of 95 °C for 20 s, 55 °C for 20 s, and 72 °C for 30 s. Reaction specificity was confirmed by melt curve analysis. All reactions were performed in technical triplicates, and Ct values were standardized against GAPDH using the 2^−ΔΔCt method [33]. Statistical analyses were performed using GraphPad Prism (v 8.0).

Statistical analysis

Bioinformatics analyses were executed utilizing the R programming language (v 4.3.3). Wilcoxon tests were utilized to compare the differences between two groups, and the difference between different groups in the RT-qPCR was gained using the t-test (p < 0.05).

Results

Identification of BA-related candidate genes in STEMI progression

Comparative transcriptomic analysis between STEMI samples and stable CAD controls revealed 60 DEGs (adj. p < 0.05, |log₂FC| > 0.5), including 28 upregulated and 32 downregulated genes in STEMI samples (Fig. 1A-B). Notably, the BARGs score exhibited significant differences between two groups in GSE59867 dataset (Wilcoxon test, p < 0.05) (Fig. 1C). Cluster analysis of all samples identified two outliers, which were excluded from subsequent analyses to ensure robustness (Supplementary Fig. 1 A, B). Constructing a biologically meaningful co-expression network, a soft threshold (power = 5) was selected (Supplementary Fig. 1 C). Initial module detection using dynamic tree cutting yielded 17 distinct gene modules (Supplementary Fig. 1D). Merging highly correlated modules at a cut height of 0.3 refined the network into 12 modules, with the grey module excluded as unclassified genes (Supplementary Fig. 1E, F). Among these, the purple module showed the strongest positive correlation with BARGs scores (r = 0.75, p = 2 × 10^− 10) and was prioritized for further investigation, encompassing 829 genes (Fig. 1D). Intramodular analysis revealed a strong correlation between module membership (MM, representing gene-module correlations) and gene significance (GS, reflecting gene-trait correlations) for BARGs (cor = 0.86). Genes meeting stringent thresholds (MM > 0.6, GS > 0.4) were designated as BA-related module genes (BARMGs), including 419 genes (Fig. 1E). Intersection of BARMGs with DEGs identified 20 candidate genes (Fig. 1F). These results delineated 20 candidate genes associated with brown adipocyte-related transcriptional scores and provided preliminary clues regarding immune-metabolic features in STEMI progression.

Fig. 1.

Identification of BA-related candidate genes in STEMI progression. (A) Volcano plot of differentially expressed genes (DEGs) between STEMI samples and stable CAD controls. Red dots represent upregulated genes, blue dots represent downregulated genes, and gray dots represent genes with no significant difference. (B) Heatmap of DEGs. Higher expression levels are shown in red, and lower expression levels are shown in blue. (C) Boxplot of BA-related gene scores (BARGs) in the GSE59867 dataset. (D) Heatmap of module-trait relationships. The vertical axis represents module names, and the heatmap colors represent the correlation coefficient with BARGs scores. Red indicates positive correlation, and blue indicates negative correlation. (E) Scatter plot of gene significance (GS) vs. module membership (MM) for the purple module. The horizontal axis represents the correlation between genes and modules (MM), and the vertical axis represents the correlation between genes and traits (GS). (F) Venn diagram of candidate genes

Functional enrichment and protein interaction analysis of candidate genes

To elucidate their biological significance, the roles of the 20 candidate genes in STEMI progression were analyzed. GO enrichment analysis identified 156 significant terms, including 142 biological process (BP) terms, 7 cellular components (CC) terms, and 7 molecular function (MF) terms. The top BP terms were predominantly mononuclear cell differentiation, type II interferon production, and T cell differentiation. For CCs, candidate genes were enriched in secretory granule membrane and specific granule. MF terms highlighted immune receptor activity and fatty acid binding (Fig. 2A-C). KEGG pathway analysis further identified 8 pathways enriched for candidate genes. The most significantly enriched pathways encompassed hematopoietic cell lineage, osteoclast differentiation, and B cell receptor signaling (Fig. 2D). The protein-protein interaction (PPI) network revealed 16 candidate genes that exhibited interconnected relationships (interaction score ≥ 0.15) (Fig. 2E). These results further suggested that the candidate genes may be involved in immune and metabolic processes associated with brown adipocyte-related molecular signatures in STEMI.

Fig. 2.

Functional analysis and protein-protein interaction analysis of candidate genes. (A) GO enrichment analysis of biological processes (BP). The left side shows the names of candidate genes, and the right side shows the BP terms. (B) GO enrichment analysis of cellular components (CC). The left side shows the names of candidate genes, and the right side shows the CC terms. (C) GO enrichment analysis of molecular functions (MF). The left side shows the names of candidate genes, and the right side shows the MF terms. (D) Bar chart of KEGG pathway enrichment analysis. The color of the bars represents the enrichment p-value, and the length of the bars represents the number of genes enriched in the pathway. (E) Protein-protein interaction (PPI) network. Each node and connection line represent the interaction between proteins, with red nodes indicating upregulated genes and blue nodes indicating downregulated genes

Identification of key genes in STEMI

To detect robust key genes for STEMI, least absolute shrinkage and selection operator (LASSO) regression analysis was applied to the candidate genes, yielding 3 feature genes (ASGR2, MCEMP1, and PPARG) with optimal lambda values (lambda.min = − 4.1314) (Supplementary Fig. 2A-B). Concurrently, support vector machine-recursive feature elimination (SVM-RFE) analysis of the 20 candidate genes revealed that a subset of 19 genes reached the highest predictive accuracy (Supplementary Fig. 2C-D). Intersection of LASSO and SVM-RFE results identified ASGR2, MCEMP1, and PPARG as feature genes (Supplementary Fig. 2E). Wilcoxon rank-sum analysis confirmed consistent upregulation of these 3 feature genes in STEMI samples compared to stable CAD controls across both GSE59867 and GSE62646 datasets (p < 0.01) (Supplementary Fig. 2F-G). These findings indicated that ASGR2, MCEMP1, and PPARG may represent key genes associated with brown adipocyte-related molecular signatures in STEMI PBMC transcriptomic profiles.

Clinical validation of the nomogram model

The nomogram showed promising discriminative performance for distinguishing STEMI from stable CAD in the analyzed datasets. As the total points increased, the probability of STEMI occurrence rose proportionally, with a total point of 195 corresponding to a 98.2% predicted STEMI prevalence (Fig. 3A). Calibration analysis revealed a slope close to 1 (p = 0.208), indicating a high level of concordance between predictions and observations (Fig. 3B). Clinical impact curve analysis showed near overlap between the number of high-risk and number of high-risk with event curves, underscoring the model’s precision in identifying true STEMI cases (Fig. 3C). Furthermore, the model achieved an area under the curve (AUC) of 0.973 in receiver operating characteristic (ROC) curve analysis, demonstrating high discriminative accuracy for distinguishing STEMI from stable CAD (Fig. 3D). Collectively, these analyses suggested that the nomogram may have potential value for distinguishing STEMI from stable CAD, although further validation in larger independent cohorts is required.

Fig. 3.

Construction and evaluation of the nomogram model. (A) Nomogram. Points represent the score for each gene. The total points for the three key genes are calculated and mapped to the linear score and outcome probability. (B) Calibration curve of the nomogram. (C) Clinical impact curve of the nomogram. The red curve represents the number of individuals classified as positive (high risk) by the simple model at various threshold probabilities; the blue curve represents the number of true positives at different threshold probabilities. The higher the overlap between the two curves, the stronger the predictive ability of the nomogram model. (D) ROC curve of the nomogram. The area under the curve (AUC) closer to 1 indicates better predictive performance

Association of key genes with pathways in STEMI

To delineate the biological roles of key genes (ASGR2, MCEMP1, and PPARG) in STEMI pathogenesis, gene set enrichment analysis (GSEA) was performed. ASGR2 exhibited significant enrichment in 732 pathways (|NES|) > 1, FDR < 0.25, and p < 0.05), with prominent involvement in reactome neutrophil degranulation (NES = 2.84), wp tyrobp causal network in microglia (NES = 2.81), and lysosome (NES = 2.64) (Fig. 4A, Supplementary Table S3). For MCEMP1, 459 pathways were significantly enriched, such as reactome neutrophil degranulation (NES = 2.97), oxidative phosphorylation (NES = 2.43), and reactome degradation of the extracellular matrix organization (NES = 2.34) (Fig. 4B, Supplementary Table S4). PPARG demonstrated enrichment in 560 pathways, with the top-ranked terms being reactome neutrophil degranulation (NES = 2.99), FC γ mediated phagocytosis (NES = 2.43), and reactome Toll-like receptor cascades (NES = 2.34) (Fig. 4C, Supplementary Table S5). Strikingly, all key genes converged on reactome neutrophil degranulation and Toll-like receptor cascades, underscoring their collective contribution in STEMI. The unique enrichment of ASGR2 in microglial signaling, MCEMP1 in oxidative phosphorylation, and PPARG in phagocytosis further highlighted their distinct yet complementary roles in STEMI. These results further suggested that these key genes may act as immune-metabolic nodes associated with brown adipocyte-related molecular signatures in STEMI progression.

Fig. 4.

GSEA pathway enrichment analysis of key genes. (A) ASGR2. Different colors represent different pathways. A positive ES indicates that the functional gene set is enriched at the front of the ranked list, suggesting a positive correlation with the gene, while a negative ES indicates that the functional gene set is enriched at the end of the ranked list, suggesting a negative correlation with the gene. (B) MCEMP1. (C) PPARG

Immune infiltration landscape in STEMI

To characterize alterations in the immune microenvironment of STEMI, we employed the CIBERSORT algorithm to profile immune cell infiltration. Wilcoxon tests disclosed significant differences in three immune cells between STEMI samples and stable CAD controls, including resting natural killer (NK) cells (p < 0.001), monocytes (p < 0.05), and eosinophils (p < 0.01) (Fig. 5A-B). Spearman correlation revealed a significant inverse correlation between resting NK cells and monocytes (r > −0.5, p < 0.001) (Fig. 5C). Notably, there were significant correlations between the three key genes (ASGR2, MCEMP1, PPARG) and differentially infiltrated immune cells. All key genes showed positive correlations with monocytes (r > 0.3, p < 0.01) but negative correlations with resting NK cells (r < −0.3, p < 0.01). Conversely, PPARG and ASGR2 were uniquely negatively correlated with eosinophils (r < −0.3, p < 0.05) (Fig. 5D). These results revealed associations between the key genes and the immune microenvironment, providing exploratory insights into immune alterations potentially linked to brown adipocyte-related molecular signatures during STEMI progression.

Fig. 5.

Immune microenvironment analysis in STEMI. (A) Stacked bar chart of the infiltration proportions of 22 immune cell types in STEMI samples and stable CAD controls. Different colors represent different immune cell types. (B) Boxplot of immune cell infiltration differences between STEMI samples and stable CAD controls. Green represents STEMI samples, and yellow represents stable CAD samples. ^*p < 0.05, ^**p < 0.01, ^***p < 0.001. (C) Correlation heatmap of differentially infiltrated immune cells. Both the horizontal and vertical axes represent different immune cells. Red indicates positive correlation, blue indicates negative correlation, and deeper colors indicate stronger correlations between immune cells. ^***p < 0.001. (D) Correlation heatmap of the three key genes (ASGR2, MCEMP1, PPARG) with differentially infiltrated immune cells. Red represents positive correlation, blue represents negative correlation, and deeper colors indicate stronger correlations. ^*p < 0.05, ^**p < 0.01, ^***p < 0.001

Genomic localization, interaction networks, and therapeutic exploration of key genes

Chromosomal localization analysis revealed distinct genomic positions for the key genes, PPARG on chromosome 3, ASGR2 on chromosome 17, and MCEMP1 on chromosome 19 (Supplementary Fig. 3 A). Friends’ analysis identified MCEMP1 as the most interconnected node among the key genes, exhibiting the strongest correlations (Supplementary Fig. 3B). Pairwise correlation analysis further confirmed robust positive associations between all key genes (cor > 0.5, p < 0.001) (Supplementary Fig. 3 C), indicating coordinated regulation in STEMI pathophysiology. GeneMANIA-based interaction network prediction indicated that physical interactions were the most robust among these genes, and these genes participated in various functions, such as monocarboxylate transport, lipid transport, and monocarboxylate binding (Fig. 6A). Subsequently, 63 miRNAs targeting PPARG were identified based on StarBase, with hsa-miR-301b-3p, hsa-miR-130a-3p, and hsa-miR-130b-3p ranking as top regulators (Fig. 6B). No significant miRNAs were predicted for ASGR2 or MCEMP1. In addition, compounds targeting key genes via the Drug-Gene Interaction Database (DGIdb, https://dgidb.genome.wustl.edu/) identified 4 ion-based inhibitors targeting ASGR2 (CS+, BA2+, H+, MG2+) and 125 compounds interacting with PPARG. The top compounds with highest interaction scores targeting PPARG included SHINPTEROCARPIN, CLX-0921, and ECHINATIN. Molecular docking simulations demonstrated strong binding affinities between PPARG and these 3 compounds (binding affinity < −5 kcal/mol) (Table 1; Fig. 6C-H). These results mapped the genomic localization, interaction networks, and candidate compound interactions of the key genes, providing exploratory clues for subsequent functional validation.

Fig. 6.

Predicted interaction networks and molecular docking of key genes. (A) GeneMANIA interaction network of the key genes. Different colors represent different biological functions and types of interactions. (B) Predicted PPARG-miRNA-lncRNA interaction network. Pink represents PPARG, green represents miRNAs, and blue represents lncRNAs. (C) Molecular docking overview of PPARG with CLX-0921, with red representing the PPARG protein structure and green representing the compound structure. (D) Close-up of PPARG–CLX-0921 binding site with key residues labeled. (E) Molecular docking overview of PPARG with ECHINATIN. (F) Close-up of PPARG–ECHINATIN binding site with key residues labeled. (G) Molecular docking overview of PPARG with SHINPTEROCARPIN. (H) Close-up of PPARG–SHINPTEROCARPIN binding site with key residues labeled

Table 1.

Molecular docking

Protein	Compound	Binding Score (kcal/mol)
PPARG	CLX 0921	−5.0
PPARG	ECHINATIN	−5.4
PPARG	SHINPTEROCARPIN	−5.8

Expression levels of key genes

To validate the bioinformatically predicted expression of key STEMI genes, we performed RT-qPCR to quantify mRNA levels of ASGR2, MCEMP1, and PPARG. This experimental validation aimed to confirm the differential expression patterns identified in prior computational analyses. The mRNA expression of ASGR2, MCEMP1, and PPARG was analyzed in a cohort comprising 5 STEMI samples and 5 stable CAD controls. Transcriptional levels differed significantly between the groups, with ASGR2, MCEMP1, and PPARG showing markedly higher expression in STEMI samples than in stable CAD controls (p < 0.05) (Fig. 7). This upregulation highlighted distinct molecular dysregulation associated with STEMI pathogenesis. These findings were consistent with the earlier bioinformatics predictions, suggesting that ASGR2, MCEMP1, and PPARG may serve as potential biomarkers associated with STEMI.

Fig. 7.

RT-qPCR experimental validation of key genes. (A) ASGR2. (B) MCEMP1. (C) PPARG. ^*p < 0.05, ^**p < 0.01

Discussion

STEMI is a severe cardiovascular disease characterized by high morbidity and mortality, with pathogenesis involving complex interactions among lipid metabolism, immune activation, inflammation, endothelial dysfunction, and myocardial injury [34, 35]. BAT, as a metabolically active endocrine organ, has garnered growing interest for its potential role in cardiovascular disease, especially regarding its influence on systemic lipid metabolism, energy homeostasis, and inflammatory regulation [36, 37]. Against this backdrop, we used PBMC transcriptomic analysis to explore brown adipocyte-related molecular features in STEMI and identified ASGR2, MCEMP1, and PPARG as key genes associated with this molecular signature. A diagnostic model constructed from these three genes demonstrated favorable discriminatory performance in the analyzed datasets, although larger independent cohorts are required for clinical validation. Subsequent enrichment and immune infiltration analyses linked these genes to immunometabolic alterations, including neutrophil degranulation, Toll-like receptor signaling activation, and monocyte-related immune responses, supporting the hypothesis that brown adipocyte-related molecular features may be associated with peripheral immunometabolic remodeling in STEMI and providing a foundation for further mechanistic studies.

The ASGR2 gene encodes the minor subunit of the asialoglycoprotein receptor (ASGPR), a liver-specific type II transmembrane protein that forms functional complexes with its major subunit ASGR1 on hepatocyte surfaces [38, 39]. ASGPR mediates clathrin-dependent endocytosis and subsequent lysosomal degradation of senescent or damaged serum glycoproteins, thereby regulating systemic glycoprotein and iron homeostasis [40, 41]. Emerging evidence has expanded our understanding of ASGPR’s physiological role beyond hepatic clearance functions. ASGPR could systemically modulate cardiovascular pathophysiology through three interconnected mechanisms: (1) clearance of prothrombotic desialylated plasma glycoproteins [42], (2) regulation of atherogenic lipoprotein particles (LDL/VLDL) [38], and (3) removal of activated platelets exposing surface N-acetylgalactosamine (GalNAc) residues [42]. These clearance functions play a vital role in cardiovascular protection, exerting beneficial effects through multiple mechanistic pathways: a reduction in prothrombotic factors lowers the propensity for coronary thrombosis; modulation of lipoprotein metabolism slows the progression of atherosclerotic lesions; and the clearance of activated platelets helps mitigate the risk of acute vascular occlusion, thereby contributing to the maintenance of vascular homeostasis. The findings of this study demonstrate a significant upregulation of ASGR2 gene expression in STEMI patients compared to control subjects. Therefore, the upregulation of ASGR2 in STEMI may reflect an adaptive peripheral response associated with glycoprotein clearance, lipid metabolism, and thrombo-inflammatory activation. Its association with brown adipocyte-related molecular features further suggests that ASGR2 may be involved in immunometabolic interactions during STEMI, which warrants further validation in experimental models involving adipose and myocardial tissues.

Mast cell-expressed membrane protein 1 (MCEMP1) is a type II transmembrane glycoprotein characterized by its predominant expression in myeloid immune cells, with particularly high specificity for mast cells [43]. In the context of immunoregulation, MCEMP1 exerts multiple critical functions. On one hand, it regulates monocyte/macrophage chemotaxis, cellular adhesion, and migratory capacity through TGFβ-dependent mechanisms, thereby affecting the distribution of the inflammatory response [44]. On the other hand, MCEMP1 serves as both a binding partner and functional activator of KIT, establishing a novel regulatory mechanism for SCF-dependent mast cell proliferation leading to a modulation of mast cell activation and subsequent effector functions [45]. These immunoregulatory functions hold significant implications in cardiovascular diseases, as activated mast cells and macrophages directly contribute to pathological processes such as atherosclerotic plaque formation, vascular inflammation, and myocardial IR injury through the release of bioactive mediators including histamine, proteases, pro-inflammatory cytokines, and chemokines [46]. Clinical investigations have also identified aberrant MCEMP1 expression in immune cells from patients with various inflammatory disorders, such as idiopathic pulmonary fibrosis, further supporting its central role in inflammatory regulation [22]. Our PCR results revealed a significant upregulation of MCEMP1 in STEMI patients, a finding of considerable pathophysiological significance. Elevated MCEMP1 expression may reflect immune-inflammatory activation triggered by acute myocardial infarction, particularly involving monocyte/macrophage- and mast cell-related responses. The observed correlation between MCEMP1 and brown adipocyte-related molecular features implies that MCEMP1 could serve as a peripheral immune-inflammatory factor involved in metabolic regulation, thereby providing a promising avenue for future investigations into immunometabolic crosstalk in STEMI.

Peroxisome proliferator-activated receptor gamma (PPARγ) is a ligand-activated transcription factor belonging to the nuclear receptor superfamily [47]. This receptor is highly expressed in adipose tissue, macrophages, and the large intestine, playing a critical role in lipid metabolism and inflammation [47]. PPARγ interacts with multiple protein partners, not only modulating lipid metabolism and glucose homeostasis but also suppressing the expression of pro-inflammatory genes, thereby exerting anti-inflammatory and protective functions [48]. However, dysregulation of PPARγ has been implicated in the pathogenesis of atherosclerosis and related cardiovascular diseases. Mechanistically, PPARγ dysfunction disrupts lipid homeostasis and exacerbates chronic inflammation, thus contributing to endothelial dysfunction, accelerated plaque formation, and increased plaque instability. These pathological changes ultimately elevate the risk of adverse cardiovascular events [49, 50]. Our PCR results demonstrated a significant upregulation of PPARγ expression in STEMI patients compared to control subjects. In view of the well-established roles of PPARG in adipose tissue biology, lipid metabolism, and inflammatory modulation, this gene offers a biologically rational connection between brown adipocyte-associated molecular characteristics and STEMI-related immunometabolic changes. Accordingly, the upregulation of PPARG observed in this study may represent a compensatory regulatory mechanism in response to acute ischemic injury and further underscores its promise as a candidate biomarker that merits additional verification.

The ROC curve analysis demonstrated that the nomogram model integrating ASGR2, MCEMP1, and PPARG achieved an AUC of 0.973, indicating outstanding diagnostic performance and excellent discriminatory power. These results underscore the model’s significant potential for clinical application in the early diagnosis of STEMI.

GSEA results indicated that the three pivotal genes jointly modulate neutrophil degranulation and the Toll-like receptor (TLR) signaling cascade. These findings underscore the presence of synergistic interactions among these genes across diverse cellular processes. Neutrophil degranulation emerges as a critical innate-immune pathway of STEMI by contributing to both acute inflammatory responses and subsequent tissue injury. In the event of coronary occlusion, ischemia-reperfusion injury (IRI) triggers rapid neutrophil recruitment to the infarcted myocardium, where they release cytotoxic contents, exacerbating myocardial damage and promoting adverse remodeling, further amplifying oxidative stress and cardiomyocyte apoptosis [51, 52]. TLRs are pivotal pattern recognition receptors (PRRs) that mediate innate immune responses by detecting damage-associated molecular patterns (DAMPs) released during myocardial IRI in STEMI. Activation of TLR signaling exacerbates inflammatory cascades, contributing to cardiomyocyte apoptosis, adverse ventricular remodeling, and infarct expansion [53, 54]. Notably, the present study found that ASGR2, MCEMP1, and PPARG were all closely associated with neutrophil degranulation and TLR signaling pathways, suggesting their potential involvement in the inflammatory and immunometabolic processes of STEMI. These findings provide a reasonable biological context for understanding the participation of brown adipocyte-related molecular features in STEMI-associated immune responses, and point toward directions for future mechanistic validation.

Immune cell infiltration analysis revealed significant differences in the infiltration levels of three distinct immune cell types between STEMI patient samples and stable CAD control samples. These included resting NK cells, eosinophils, and monocytes. In-depth analysis of immune infiltration revealed a significant increase in monocyte infiltration, along with decreased levels of resting NK cells and eosinophils, in patients with STEMI compared to controls (all p < 0.05). Increased monocyte infiltration may contribute to vascular inflammation and endothelial dysfunction by promoting the release of pro-inflammatory cytokines and matrix metalloproteinases. Studies have demonstrated that monocytes release pro-inflammatory cytokines and matrix metalloproteinases (MMPs), which exacerbate vascular injury and remodeling [55, 56]. Notably, the non-classical monocyte subset (CD14⁺CD16⁺⁺) plays a critical role in STEMI. Evidence suggests that this cell population serves as an important predictor of adverse clinical outcomes in STEMI patients. The underlying mechanism involves CCR-mediated cellular migration: as a key chemokine receptor, CCR2 guides the trafficking of CD14⁺CD16⁺⁺/CCR2⁺ monocytes from the circulation to the infarcted myocardium, leading to localized accumulation and exacerbation of inflammatory responses and myocardial injury [57]. A direct consequence of this migration is a rapid decline in circulating non-classical monocytes, the extent of which correlates directly with the severity of myocardial injury [58], further underscoring their role as key responders in acute STEMI. In contrast to increased monocyte infiltration, STEMI patients exhibit significant reductions in resting NK cells and eosinophils, alterations with important pathophysiological implications. Under normal cardiac homeostasis, resting NK cells exert protective effects by curbing excessive maturation and trafficking of inflammatory cells [59]. Following myocardial infarction, NK cells mediate cardioprotective effects through the dual mechanisms of inhibiting cardiac fibrosis and stimulating vascularization and angiogenesis [60]. However, clinical studies have shown that NK cell deficiency is closely associated with coronary artery disease, and their sustained reduction is often accompanied by persistent low-grade inflammation [59]. Similarly, eosinophils play a crucial role in myocardial repair: through the secretion of IL-4 and cationic proteins, they reduce cardiomyocyte apoptosis, suppress excessive cardiac fibroblast activation, attenuate fibrosis, and promote beneficial M2 macrophage polarization [61]. The loss of these protective immune cells carries significant prognostic implications. A low eosinophil count(<0.02 × 10⁹/L) or a low eosinophil percentage(≤ 0.3%) has been identified as an independent predictor of in-hospital major adverse cardiovascular events and long-term all-cause mortality in STEMI patients [62]. Thus, the observed elevation in monocytes coupled with reductions in resting natural killer cells and eosinophils may point to a state of peripheral immune dysregulation, marked by augmented pro-inflammatory activation and diminished reparative immunity. These immune perturbations offer a rational interpretation for the link between the identified brown adipocyte-associated hub genes and immunometabolic reprogramming in the context of STEMI.

Further correlation analysis revealed statistically significant associations between all key genes and the differentially infiltrated immune cell populations (|cor| > 0.3). These findings provide a potential immunometabolic context for understanding STEMI progression and may help guide future mechanistic investigation of immune–metabolic crosstalk in this disease. Through virtual screening and molecular docking analyses, the natural compound shinpterocarpin was identified to exhibit favorable binding affinity with PPARG, suggesting its potential role as a promising agonist for PPARG regulation. Shinpterocarpin, a pterocarpan-type flavonoid, is primarily derived from plants such as licorice (Glycyrrhiza species). Although direct studies on shinpterocarpin in the cardiovascular field remain limited, existing research has demonstrated that licorice and its metabolites exert significant cardioprotective effects in rat models, primarily attributed to their antioxidant properties. Continuous 30-day administration of licorice improved cardiac function and preserved the histological structure of cardiomyocytes by enhancing endogenous antioxidant defenses in ischemic heart disease. Given that PPARG plays an important role in lipid metabolism and inflammatory regulation, the predicted interaction between shinpterocarpin and PPARG provides preliminary clues for future studies exploring PPARG-related regulatory mechanisms in STEMI. Further experimental validation is still needed to elucidate its bioactivity, binding specificity, and potential for translational applications.

This study has several limitations. First, the sample size is relatively limited, and future studies with larger cohorts from multiple centers are needed to further validate the diagnostic performance of the key genes and the nomogram model. Second, the BARG gene set used in this study was derived from previously published brown adipocyte-related signatures rather than directly from human brown adipose tissue transcriptomic data. Therefore, ASGR2, MCEMP1, and PPARG are more appropriately interpreted as genes associated with brown adipocyte-related molecular features in the PBMC transcriptome of STEMI patients, rather than as strict brown adipose tissue-specific genes. In addition, this study was primarily based on PBMC transcriptomic data without direct analysis of myocardial tissue or brown adipose tissue samples. Further validation of their roles in adipose–cardiac immunometabolic crosstalk is still required, using authentic human brown adipose tissue samples, myocardial tissue data, and functional experiments. Finally, the sample size for RT-qPCR validation in this study was relatively small, and the molecular docking results remain computational predictions that require further experimental verification. This cautious interpretation was further supported by an additional analysis of the external human BAT versus WAT dataset GSE113764, which showed that ASGR2, MCEMP1, and PPARG were not preferentially upregulated in BAT compared with WAT.

In summary, our study has identified ASGR2, MCEMP1, and PPARG as pivotal genes linked to brown adipocyte-associated molecular signatures within the PBMC transcriptome of STEMI patients and has conducted a preliminary investigation into their immunometabolic correlations. Collectively, these findings offer novel perspectives on the molecular underpinnings of STEMI and may serve as a valuable reference for future explorations into diagnostic biomarkers and immunometabolic modulation.

Author contributions

Hongmin Li contributed equally to Conceptualization, Methodology, Investigation, Formal Analysis, Writing-Original Draft. Yuanyuan Wang contributed Conceptualization, Methodology, Funding Acquisition, Supervision, Writing-Review & Editing. Xiaolin Chen performed the Validation. Changhui Ma performed the Data Curation and Visualization.

Funding

This research was funded by the Scientific Research Program of Tianjin Municipal Education Commission (grant number2021KJ226).

Data availability

The datasets analysed during the current study are available in the Gene Expression Omnibus (GEO) repository [http://www.ncbi.nlm.nih.gov/geo/] with access numbers (GSE59867 and GSE62646). Data processing procedures are available from the Correspondence author (Yuanyuan Wang) on a reasonable request.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki and approved by the Medical Ethics Committee of The Second Hospital of Tianjin Medical University under the protocol number KY2025K171, 26 February 2025. Informed consent has been obtained from the patient.

Competing interests

The authors declare no competing interests.