Identification of key genes related to macrophages and metabolic reprogramming in myocardial infarct based on single-cell and bulk transcriptomics data and experimental validation

Abstract

Background

Myocardial infarction (MI), one of the most severe cardiovascular diseases, is influenced by altered macrophage status and metabolic reprogramming (MR), yet the mechanisms of their crosstalk remain unclear. This study used bioinformatics to explore key genes associated with macrophages and MR function in MI.

Methods

Single-cell and bulk transcriptomics data of MI were obtained from public databases, and MR-related genes (MRGs) were downloaded from literature. Macrophage-associated differentially expressed genes (DEGs1) were identified from the single-cell data. Candidate genes were determined by intersecting MI-related DEGs (DEGs2), DEGs1, and MRGs. Machine learning, gene expression, and receiver operating characteristic (ROC) curve analyses were used to identify key genes. Drug prediction analysis was performed. Single-cell data were further analyzed to explore the underlying biological mechanisms of MI, followed by cell communication and pseudo-temporal analyses. Finally, RT-qPCR was used to validate the mRNA expression of key genes.

Results

A total of 304 DEGs1, 1,258 DEGs2, and 1,804 MRGs were intersected, identifying 6 candidate genes. The key genes (ABCG1, GNG11, and RPL24) were successfully identified. Pathways co-enriched by key genes included Huntington’s disease, olfactory transduction, oxidative phosphorylation, proteasome, and spliceosome. Two drugs targeting ABCG1, APT and vitamin E, were identified. Macrophages displayed significant interactions with B/plasma cells and dendritic cells (DC) in MI samples. The expression of ABCG1 was significantly increased in MI, while GNG11 and RPL24 were decreased.

Conclusion

ABCG1, GNG11, and RPL24 were identified as key regulators of macrophage function and MR in MI, providing valuable insights for the development of targeted therapies.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12872-026-05822-9.

Introduction

Myocardial infarction (MI) is a leading cause of sudden cardiac death and a major contributor to high morbidity and mortality globally, posing a significant threat to patient health and survival [1]. The American Heart Association’s 2020 report on heart disease statistics reveals that 3% of adults over the age of 20 are affected by MI, with one person experiencing an MI every 40 s [2]. MI is a complex pathological condition that can cause severe damage to the myocardium, affecting tissue structure, energy metabolism, and cardiac function [3]. The pathogenesis of MI is multifaceted, involving mechanisms such as calcium overload, mitochondrial dysfunction, apoptosis, pyroptosis, and ferroptosis [4]. Despite significant advancements in percutaneous coronary intervention (PCI) technology, which has improved the treatment of acute MI (AMI), it remains a leading cause of death worldwide. Although numerous studies have identified biomarkers associated with MI, many of the underlying molecular mechanisms remain unclear [5]. Therefore, identifying new and reliable biomarkers that influence the onset and progression of MI, as well as understanding their roles in key biological processes and signaling pathways, is crucial for advancing the understanding and treatment of MI.

Macrophages were first identified by Elie Metchnikoff in the 19th century as a crucial component of the body’s immune defense, playing a significant role in both innate and adaptive immunity [6]. These essential immune cells are involved in maintaining organism development and homeostasis, clearing host defenses during pathogen infection, and promoting tissue repair following injury. Macrophages reside in specific tissues and exhibit diverse functions, displaying two distinct phenotypes in response to microenvironmental signals: classically activated (M1) macrophages and alternatively activated (M2) macrophages [7]. M1 macrophages are pro-inflammatory and primarily involved in initiating and sustaining the inflammatory response, while M2 macrophages exhibit anti-inflammatory properties, contributing to tissue homeostasis and repair [8]. Macrophages play critical roles in the injury and repair processes during MI, influencing its progression [9]. As heterogeneous cells, macrophages are integral to biological repair and the inflammatory response, though the exact mechanisms by which they contribute to MI remain unclear [10].

Metabolic reprogramming (MR) refers to the alterations in cellular metabolism in response to various environmental cues, encompassing pathways such as glucose, amino acid, and lipid metabolism [11]. This process is tightly linked to the onset and progression of numerous diseases [12]. MR has become a hallmark of cancer, but it also plays a pivotal role in the development of common diseases, particularly in maintaining immune homeostasis and regulating immune responses [13]. Both macrophages and MR contribute to MI repair; however, the precise mechanisms underlying their regulation of MI remain poorly understood [14].

Single-cell RNA sequencing (scRNA-seq) is a powerful technique for investigating gene expression at the single-cell level, widely utilized in biomedicine and cell biology [15]. Compared to traditional RNA sequencing, scRNA-seq offers higher resolution, providing deeper insights into biological processes, developmental dynamics, and disease mechanisms [16].

This study aims to investigate the specific regulatory mechanisms of macrophages and MR-related biomarkers in MI through single-cell transcriptomic analysis. It will further examine the expression profiles and trajectories of key genes in relevant cells and validate these genes through Polymerase Chain Reaction (PCR). The findings are expected to offer new perspectives on the pathogenesis and treatment of MI.

Materials and methods

Data collection

The GSE61145, GSE61144, and GSE267644 datasets were retrieved from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) (Additional file 1). The GSE61145 dataset (sequencing platforms: GPL6884 and GPL6106) was used as the training set, comprising 14 MI whole blood samples and 17 control whole blood samples. The GSE61144 dataset (sequencing platform: GPL6106) served as the validation set, containing 7 MI whole blood samples and 10 control whole blood samples. Additionally, 8 MI peripheral blood mononuclear cell (PBMC) samples and 3 control PBMC samples were included in the scRNA-seq GSE267644 dataset (sequencing platform: GPL24676). Furthermore, a total of 1,804 MR-related genes (MRGs) were sourced from the literature [17] (Additional file 2).

Treatment of the scRNA-seq data

In the GSE267644 dataset, quality control was performed using the Seurat package (v 4.3.0) [18]. Given that the study subjects were myocardial infarction patients, whose peripheral blood cells were under stress, inflammation, and early apoptosis, the expression of mitochondrial genes might be physiologically elevated. To preserve cellular heterogeneity and avoid losing critical cell subsets due to overly stringent thresholds, a relatively lenient cutoff of “percent.mt < 20%” was applied for mitochondrial gene content in the quality control process[PMID:32840568, 36480284, 39094968]. The quality control criteria were set as 400 < nFeature_RNA < 3,000, percent.mt < 20%, and nCount_RNA < 8,000. To verify whether the threshold of percent.mt < 20% in single-cell data quality control introduced systematic bias in the detection of metabolic reprogramming-related genes (MRGs), genes were first extracted from the metabolic reprogramming-related gene list and deduplicated to obtain MRGs. Then, using the quality-controlled and normalized Seurat object, expression values of MRGs in each cell were extracted and averaged via the FetchData function of the Seurat R package, and percent.mt of each cell was extracted simultaneously. The correlation between the two indicators was assessed using Spearman’s rank correlation coefficient, and the p-value was calculated. Scatter plots were generated with ggplot2, and the distributions of nFeature_RNA, nCount_RNA and percent.mt in each sample after quality control were visualized by VlnPlot.

The SCTransform function in Seurat (v 4.3.0) was employed to select the top 2,000 genes with the highest coefficients of variation across cells, termed highly variable genes (HVGs). Among these, the top 10 HVGs with the greatest variation were identified and labeled. Principal component analysis (PCA) was then performed using the ScaleData function of Seurat (v 4.3.0) based on the 2,000 HVGs to select principal components (PCs) for further analysis (P < 0.05). The resolution was set to 0.3, and unsupervised clustering was performed using the FindNeighbors and FindClusters functions of Seurat (v 4.3.0). Cluster visualization was accomplished using the uniform manifold approximation and projection (UMAP) algorithm. Marker genes from the Human Cell Atlas (HCA) database (https://www.heartcellatlas.org/index.html) and the CellMarker database (http://xteam.xbio.top/CellMarker/) were used for cell type annotation. The annotation results were visualized in a UMAP plot, showing the expression of marker genes across different cell types.

Differential expression analysis

To identify macrophage-associated differentially expressed genes (DEGs1), differential expression analysis was conducted in macrophages between MI and control groups in the GSE267644 dataset using the FindMarkers function of Seurat (v 4.3.0) (|average log₂ fold change (FC)| > 1, P < 0.05).ScRNA-seq data are characterized by high noise and zero inflation, and are easily affected by technical variations and cellular heterogeneity, where low fold changes often represent technical artifacts. In this study, a strict threshold of |log₂FC| > 1 was applied to eliminate noise and identify differentially expressed genes with reliable biological significance, focusing on key genes related to macrophage function. The top 5 genes with significant up- or down-regulation were labeled and sorted by |log₂ FC|. For the GSE61145 dataset, DEGs2 were identified using the limma package (v 3.54.0) [19], with differential expression analysis performed between the MI and control groups (|log₂ FC| > 0.5, P < 0.05). In contrast, bulk RNA-seq reflects the average expression level of tissue samples with lower background noise. A relatively lenient threshold of |log₂FC| > 0.5 was used to comprehensively capture potential relevant genes, avoiding the omission of key molecules associated with macrophage function and metabolic reprogramming, thereby providing abundant targets for subsequent screening and validation.Volcano plots of DEGs2 were created using the ggplot2 package (v 3.4.1) [20], with the top 10 genes (sorted by |log₂ FC|) exhibiting the most significant up- and down-regulation labeled. A heatmap of DEGs2 was generated using the pheatmap package (v 0.7.7) (https://CRAN.R-project.org/package=pheatmap).

Identification, functional enrichment, and protein-protein interaction (PPI) network of candidate genes

To identify genes related to macrophages and MR that are differentially expressed in MI, the VennDiagram package (v 1.7.1) [21] was used to intersect DEGs1, DEGs2, and MRGs, with the resulting genes considered as candidate genes. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were performed on these candidate genes using the clusterProfiler package (v 4.2.2) [22] (adj.P < 0.05). To explore the protein-level interactions of these genes, candidate genes were inputted into the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (https://string-db.org/) with an interaction score threshold > 0.4. These interactions were visualized using Cytoscape software (v 3.8.2) [23]. Additionally, the RCircos package (v 1.2.2) [24] was employed to determine the chromosomal localization of the candidate genes.

Machine learning, expression validation, and receiver operating characteristic (ROC) curve

To further refine the candidate gene list, support vector machine-recursive feature elimination (SVM-RFE) and least absolute shrinkage and selection operator (LASSO) regression were applied to the training set. SVM-RFE was performed using the e1071 package (v 1.7–14) (https://cran.r-project.org/web/packages/e1071/), with k-fold cross-validation (k = 10) to identify the SVM-RFE genes. In parallel, LASSO regression was conducted using the glmnet package (v 4.1.8) [25], with the optimal lambda value determined by 10-fold cross-validation. The genes identified by both methods were defined as core genes by intersecting the SVM-RFE and LASSO results using the VennDiagram package (v 1.7.1).

The expression levels of the core genes were compared between the MI and control groups in both the training and validation sets using the Wilcoxon test (P < 0.05). Genes showing consistent expression trends and significant inter-group differences in both datasets were considered as candidate key genes. These candidate key genes were further assessed using the pROC package (v 1.18.0) [26] to generate Receiver Operating Characteristic (ROC) curves, evaluating their ability to distinguish between MI and control samples. Genes with an area under the curve (AUC) value greater than 0.7 were identified as key genes.

Correlation, GeneMANIA network, and Gene set enrichment analysis (GSEA) of key genes

To further assess the functional similarity among key genes, the GO semantic similarity was calculated using the GOSemSim package (v 2.26.1) [24]. Additionally, a gene co-expression network was constructed via the GeneMANIA website (http://www.genemania.org/) to predict the interactions between the key genes and their associated biological functions.

GSEA was performed to investigate the biological functions of key genes throughout the progression of MI. The first step involved calculating the Spearman correlation coefficients between each key gene and all other genes across the entire sample set from the training cohort, using the psych package (v 2.1.6) [27]. The resulting genes were then ranked in descending order according to their Spearman correlation coefficients, and this ranked list was used as the gene set for subsequent analyses. A reference gene set (c2.kegg.v7.4.symbols) was retrieved from the Molecular Signatures Database (MSigDB, https://www.gsea-msigdb.org/gsea/msigdb). GSEA was then carried out using the clusterProfiler package (v 4.2.2), with the thresholds set to |normalized enrichment score (NES)| > 1, P < 0.05, and false discovery rate (FDR) < 0.25.

To investigate the roles of ABCG1, GNG11, and RPL24 in metabolic reprogramming associated with myocardial infarction, this study focused on glycolysis, tricarboxylic acid (TCA) cycle, and fatty acid oxidation (FAO) pathways. Metabolic genes strongly correlated with the key genes were screened by analyzing expression correlations between key genes and pathway genes. Representative gene sets for the three pathways were downloaded from the MSigDB database, and a core metabolic gene set was obtained after merging and deduplication (400 genes, 351 of which were present in the expression matrix). In the GSE61145 training dataset, Spearman correlation analyses were performed using the stats package(v 4.3.2) (https://stat.ethz.ch/R-manual/R-patched/library/stats/html/00Index.html). Expression correlations between each key gene and core metabolic genes were calculated, and genes with |cor| > 0.3 and P < 0.05 were retained.

Regulatory network construction and potential drug prediction

To predict the potential molecular regulatory networks involving the key genes, the miRWalk database (http://www.mirwalk.umm.uni-heidelberg.de/) was utilized to identify microRNAs (miRNAs) targeting the key genes. Additionally, drug predictions for MI-associated key genes were conducted using the DrugBank database (https://www.drugbank.com/). The mRNA-miRNA regulatory network and the drug-gene network were then visualized using Cytoscape software (v 3.8.2).

Cell communication and trajectory analysis

To analyze cell-cell interactions, the strength and frequency of these interactions in the MI single-cell dataset (GSE267644) were evaluated using the aggregateNet function from the CellChat package (v 1.6.1)^[22]. Input and output signal plots of these interactions were generated for further analysis.

Finally, to investigate macrophage heterogeneity, dimensionality reduction clustering was performed on macrophages in the GSE267644 dataset (refer to Sect. 2.2). Macrophage subpopulations were characterized based on the expression levels of specific marker genes. Pseudotime analysis was then conducted using the Monocle2 package (v 2.26.0) [28], and the DifferentialGeneTest function was employed to display the expression trends of key genes during macrophage differentiation. In the Monocle2 pseudotime analysis, the FindVariableFeatures function was used, and the top 2000 highly variable genes were selected as ordering genes by the “vst” method. Key genes were not used to construct the differentiation trajectory to avoid circular reasoning. To display the expression heterogeneity of ordering genes in macrophage subpopulations, the top 50 highly variable genes were extracted, and dot plots were generated using the DotPlot function.

Reverse transcription quantitative polymerase chain reaction (RT-qPCR)

The research secured ethical approval under reference (2024)-LL-183, and the study was conducted in accordance with the principles stipulated in the 2024 revised Helsinki Declaration, with patients having provided informed consent. Blood samples were collected from 5 patients with STEMI and 5 control subjects at People’s Hospital of Ningxia Hui Autonomous Region for RT-qPCR analysis.All STEMI patients received primary percutaneous coronary intervention(PPCI). The average age of the STEMI group was 69.60 ± 10.69 years, and that of the control group was 37.00 ± 4.47 years. The percentage of males in both groups was 60%. In the STEMI group, 20% of the patients had hypertension, and there were no diabetic patients in either group. In the STEMI group, the time from symptom onset to PCI was 106 ± 71.27 min; the time elapsed from PCI to blood collection was 3.60 ± 1.52 h; and the door-to-balloon time was 70 ± 7.31 min. The main culprit vessels in the STEMI group were the left anterior descending artery (60%) and the right coronary artery(40%). Total RNA was extracted from the 10 blood samples using TRIzol reagent (Ambion, USA), and RNA concentration was determined using a NanoPhotometer N50. cDNA synthesis was performed via reverse transcription with the SureScript First-Strand cDNA Synthesis Kit and the S1000™ Thermal Cycler (Bio-Rad, USA). Primer sequences for qPCR are provided in Additional file 3. Quantitative PCR was conducted using the CFX Connect Real-Time Fluorescence Quantitative PCR Instrument (Bio-Rad, USA). Relative quantification of key genes was calculated using the 2^−ΔΔCT method. RT-qPCR data were organized in Excel and analyzed statistically, and graphs were plotted using GraphPad Prism 10 software (P < 0.05).

Statistical analysis

Statistical analyses were performed in R software (v 4.2.2), with differences between groups assessed using the Wilcoxon test (P < 0.05). In this study, **** indicates P < 0.0001, *** indicates P < 0.001, ** indicates P < 0.01, * indicates P < 0.05, and ns indicates P > 0.05.

Results

The 304 DEGs1 were associated with macrophages

Quality control was first performed on the raw data of the GSE267644 dataset to prepare for subsequent analysis. Initially, the dataset contained 36,361 cells and 103,708 genes. After quality control, the gene count was reduced to 98,200, while the cell count remained unchanged at 36,361 (Fig. 1a). No significant differences in quality control indices including nFeature_RNA, nCount_RNA, and percent.mt were observed between control and MI groups after quality control. Spearman correlation analysis revealed no significant correlation between percent.mt and the mean expression level of MRGs (r = -0.023), indicating that the proportion of mitochondrial genes exerted no systematic effect on the mean expression of MRGs (Additional file 4). Collectively, the threshold of percent.mt < 20% in single-cell data quality control was proven reasonable and acceptable. No substantial interference with the core conclusions was detected, and the detection of MRGs was independent of mitochondrial contamination.Next, the top 2,000 HVGs were identified, with the top 10 genes showing the greatest variability, including ICHAIN, IGHA1, and HBB (Fig. 1b). PCA indicated that after the 30th PC, the significance decreased, and the curve in the PC scree plot flattened. Consequently, the top 30 PCs were selected for further analysis (Fig. 1c). Cells were then clustered into 18 groups using the UMAP clustering method (Fig. 1d). Based on the expression of marker genes (Fig. 1e), these 18 clusters were annotated into five cell types: macrophages, T cells, natural killer (NK) cells, B/plasma cells, and dendritic cells (DC) (Fig. 1f). Subsequently, 304 DEGs1 in macrophages were identified between the MI and control groups, with 175 genes upregulated and 128 downregulated in the MI group (Fig. 1g).

Fig. 1.

Treatment of the scRNA-seq data. (a) Quality Control metrics for single-cell RNA sequencing data form MI and control samples. (left) Distributions of before quality control (Right) Distribution of after quality control (b) Selection of highly variable genes. Red: Top 2000 high variable genes. (c) Data normalization analysis Top 50 PCs. (d) Uniform manifold approximation and projection cell clustering analysis. Different colors are different clusters of Cells. (e) Expression status of marker genes. The larger the dot, The better the expression. (f) SingleR cell annotated. Annotated 5 cell types. (g) Identification of DGEs in macrophages of control and MI groups. Red: upregulate. Blue: downregulate

The 6 candidate genes were associated with macrophages and MR in MI

Differential expression analysis revealed 1,258 DEGs2 between the MI and control groups. Of these, 871 genes were upregulated, while 387 were downregulated in the MI group (Fig. 2a-b). A total of 6 shared genes were identified across 304 DEGs1, 1,258 DEGs2, and 1,804 MRGs and selected as candidate genes (Fig. 2c). GO analysis showed significant enrichment in 186 different categories, including cytoplasmic translation, polysomal ribosome, and structural constituent of ribosome (Additional file 5, Fig. 2d). KEGG pathway analysis revealed significant enrichment in 24 pathways, including the ribosome, coronavirus disease-COVID-19, and ABC transporters (Fig. 2e, Additional file 6). These results suggest that the candidate genes may play a role in the progression of MI through these pathways. The candidate genes RPL39, ABCG1, RPS28, RPL24, GNG11, and ABCA1 were located on chromosomes X, 21, 19, 3, 7, and 9, respectively (Fig. 2f). Additionally, a PPI network was constructed with a confidence score of 0.4, revealing strong interactions among RPS28, RPL39, and RPL24 (Fig. 2g). The PPI network highlighted the complex interactions among these genes, emphasizing their potential significance in MI progression.

Fig. 2.

Macrophages in MI and candidate genes for MR. (a-b) Differential expression analysis between MI and control group. Volcano plot: Red: up, Blue: down. Heatmap: Red: MI group, Blue: control group. (c) Venn analysis between 304 DEGs1, 1,258 DEGs2, and 1,804 MRGs identified 6 intersecting genes. (d) GO analysis of candidate genes.BP: biological process. CC: cell component. MF: molecular function. (e) KEGG analysis of candidate genes. (f) The position of candidate genes on chromosomes. (g) PPI network diagram constructed from candidate genes. Line represent interactions

ABCG1, GNG11, and RPL24 were identified as key genes

The LASSO algorithm, with a minimum lambda value of 0.1771676, identified four genes from the candidate pool: ABCG1, GNG11, RPL24, and RPS28 (Fig. 3a-b). Using the SVM-RFE algorithm, the accuracy peaked at over 0.9, selecting six genes: RPS28, RPL39, RPL24, ABCA1, GNG11, and ABCG1 (Fig. 3c). The intersection of the four LASSO genes and six SVM-RFE genes resulted in four core genes: ABCG1, GNG11, RPL24, and RPS28 (Fig. 3d). Gene expression analysis confirmed that GNG11 and RPL24 were significantly downregulated in MI samples, while ABCG1 was significantly upregulated in both the GSE61145 and GSE61144 datasets (Fig. 3e-f). These consistent expression patterns led to the identification of ABCG1, GNG11, and RPL24 as key genes for further analysis. ROC curve analysis revealed that the AUC values for ABCG1, GNG11, and RPL24 exceeded 0.7 in both the GSE61145 and GSE61144 datasets (Fig. 3g-h), demonstrating their ability to distinguish between patients with MI and control samples, thereby highlighting their potential diagnostic value. Consequently, ABCG1, GNG11, and RPL24 were confirmed as key genes.

Fig. 3.

Identification of Key genes. (a) Lasso coefficient spectrum. If the genes is on the Y-asix at Log(lambda.min) ≠ 0, then the gene is obtained. (b) The result of Lasso algorithm. At the value of log(lambda. min), the model error is minimized. (c) The of SVM-RFE algorithm. The X-axis at peak is 6. (d) Venn of 4 LASSO genes and 6 SVM-RFE genes. Obtain 4 core genes. (e) Gene expression analysis of four core genes in MI group and control group in the GSE61145. (f) Gene expression analysis of four core genes in MI group and control group in the GSE61144. (g) Receiver Operating Characteristic curve of three candidate key genes in the GSE61145. AUC > 0.7 was identified as key genes. (h) Receiver Operating Characteristic curve of three candidate key genes in the GSE61144. AUC > 0.7 are identified as key genes

Key gene interactions were significantly enriched in multiple signalling pathways

An analysis of the functional relationships between the key genes revealed higher functional similarity between GNG11 and RPL24 (P > 0.5) (Fig. 4a). GeneMANIA identified 19 genes functionally associated with ABCG1, GNG11, and RPL24 (Fig. 4b). For example, RPL9 was linked to RPL24, while GNB3 was associated with ABCG1. GSEA showed that ABCG1 was significantly enriched in 75 pathways, GNG11 in 83 pathways, and RPL24 in 64 pathways (Additional files 7–9). Both ABCG1 and GNG11 were co-enriched in several common pathways, including Huntington’s disease, olfactory transduction, oxidative phosphorylation, proteasome, and spliceosome (Fig. 4c-e). These results suggest that ABCG1, GNG11, and RPL24 may contribute to cellular degradation and protein synthesis, which could play a role in the pathogenesis of MI.

Fig. 4.

Gene Set Enrichment Analysis of key genes. (a) Association of functionally between the key genes. The heatmap displays. The functional similarity between GNG11 and RPL24 were higher (P > 0.5). (b) Screening of functional genes related to key genes. (c-e) Gene Set Enrichment Analysis of three key genes. The 10 lines in the upper part represent the enriched pathways and their corresponding enrichment scores. The barcode like portion in the lower part corresponds to the pathway enriched in the first part, with each vertical line representing the genes enriched in that pathway. (f) Heatmap of correlations between key genes and metabolic genes. Red indicates positive correlation and blue indicates negative correlation. The values represent correlation coefficients. The vertical axis represents key genes, and the horizontal axis represents metabolic genes

ABCG1 was significantly positively correlated with PYGL (a key enzyme in glycogenolysis), PDK3 (a negative regulator of the TCA cycle), and ACOX1 (the rate‑limiting enzyme of fatty acid β‑oxidation). GNG11 was strongly correlated with CPT2 (the rate‑limiting enzyme of mitochondrial fatty acid oxidation), ACSBG1 (involved in fatty acid β‑oxidation priming), and ABCD1 (a peroxisomal fatty acid transporter), suggesting that energy substrate selection in cardiomyocytes might be regulated via G protein signaling. RPL24 was strongly positively correlated with PFKP (the rate‑limiting enzyme of glycolysis), whose activity directly determined glycolytic flux (Fig. 4f). These results suggested that targeting ABCG1, GNG11, and RPL24 to regulate the balance among glycolysis, fatty acid oxidation, and glycogenolysis might be beneficial for attenuating ischemic myocardial injury, improving myocardial energy status, and reducing the risk of heart failure after myocardial infarction.

Potential regulatory mechanisms and targeted drugs for key genes were identified

Furthermore, 542 miRNAs were identified through overlapping predictions from two databases (Fig. 5a). Specifically, 335 miRNAs (e.g., hsa-miR-1914-3p, hsa-miR-4494) were found to target GNG11, while 248 miRNAs (e.g., hsa-miR-1224-3p, hsa-miR-205-5p) targeted ABCG1. Notably, 41 miRNAs, including hsa-miR-1914-3p, hsa-miR-6881-5p, and hsa-miR-6529-5p, simultaneously regulated both GNG11 and ABCG1, suggesting their pivotal role in regulating these genes and influencing their downstream functions in MI. However, no miRNAs were predicted for RPL24. Potential therapeutic interventions were explored by identifying drugs targeting ABCG1, GNG11, and RPL24. Two drugs targeting ABCG1, APT and vitamin E, were identified. However, GNG11 and RPL24 were not predicted to be relevant targets for these drugs (Fig. 5b). These results suggest that APT and vitamin E may offer therapeutic potential for MI by targeting ABCG1.

Fig. 5.

Potential regulatory mechanisms of key genes and confirmation of targeted drugs.(a) The prediction results of two databases overlap. 542 miRNAs was identified. (b) Network diagram of genes and drugs. Line indicate corresponding Targets

The interactions and differentiation of macrophages were explored

Cell communication analysis revealed significant interactions between annotated cell types (Fig. 6a-b). Macrophages displayed the strongest interactions with B/plasma cells and DCs in MI samples, with a diminished interaction observed in the control samples. Both afferent and efferent signaling pathways showed robust signaling in MI and control samples, including MHC-II, ANNEXIN, GALECTIN, ADGRE5, and SELPLG (Fig. 6c-d).

Fig. 6.

Cell communication and trajectory analysis of macrophages. (a-b) Network for Cell communication analysis of among annotated cell types.The thicker the line, the closer the connection. (c-d) The interaction between incoming and outgoing information in MI samples and control samples. The darker the green, the stronger the relative strength. (e) UMAP clustering analysis of macrophages. Revealed 8 distinct subclusters. (f) Marker genes expression. The bigger of bot. the bigger percent expressed. (g) UMAP clustering analysis of macrophages. Revealed 7 distinct subclusters. (h) Analysis of cellular components and pseudotime in macrophage subtypes. (i) Pseudotime-dependent gene expression dynamics during macrophage differentiation in myocardial infarction

Further investigation of macrophage subpopulations through secondary dimensionality reduction and clustering identified 8 distinct subclusters (Fig. 6e). The top 50 highly variable genes showed significant expression heterogeneity among macrophage subpopulations, and their expression proportions and average expression levels could effectively distinguish different cell states (Additional file 10), indicating that the differentiation trajectory constructed based on highly variable genes could accurately reflect the transcriptomic dynamic changes of macrophages during myocardial infarction. Trajectory construction was independent of subsequent biomarker screening, which effectively avoided circular reasoning.Based on marker gene expression (Fig. 6f), these subclusters were annotated into seven cell types: activated macrophages, CD9 + macrophages, LYPD2 + macrophages, CCR7 + CAMK4+ macrophages, FCER1A+ macrophages, JCHAIN+ macrophages, and CCL3L1 + macrophages (Fig. 6g). Pseudo-temporal analysis of macrophage differentiation revealed a progressive trajectory from early (dark blue) to more mature (light blue) states. The subclusters could be divided into five differentiation stages, with CD9 + macrophages, LYPD2 + macrophages, and FCER1A+ macrophages representing the earliest stages (Fig. 6h). The expression levels of ABCG1, GNG11, and RPL24 were found to be elevated during the mid-stage of macrophage differentiation, suggesting their potential role in macrophage development at this stage (Fig. 6i).

The expression trends of ABCG1, GNG11, and RPL24 were opposing expression patterns in MI

To validate the expression of key genes, RT-qPCR experiments were conducted using clinical samples. The results, shown in Fig. 7a-c, indicated a significant increase in ABCG1 expression in the MI group compared to the control group (P < 0.001). Conversely, GNG11 and RPL24 levels were significantly decreased in MI samples relative to control samples (P < 0.05). These findings corroborated the bioinformatics analysis, supporting the hypothesis that macrophages and MR play a significant role in the progression of MI.

Fig. 7.

Validation of macrophage differentiation-related gene expression in clinical myocardial infarction (MI) samples by RT-qPCR. (a)Validation of macrophage ABCG1 expression in clinical myocardial infarction (MI) samples by RT-qPCR. *** represented P < 0.001. (b)Validation of macrophage GNG11 expression in clinical myocardial infarction (MI) samples by RT-qPCR. * represented P < 0.05. (c)Validation of macrophage RPL24 expression in clinical myocardial infarction (MI) samples by RT-qPCR.* represented P < 0.05

Summary: The analysis began with quality control, screening high-variance genes, dimensionality reduction, cluster analysis, and cell group annotation using the single-cell transcriptomic dataset, identifying 304 macrophage-related DEGs1. Differential expression analysis in MI samples from the training set yielded 1,258 DEGs2. The intersection of DEGs1, DEGs2, and MRGs led to the identification of six candidate genes associated with macrophages and MR in MI. GO, KEGG enrichment analysis, and PPI network construction were performed on these six genes to explore their biological functions. LASSO and SVM-RFE algorithms were then used to analyze the candidate genes, ultimately selecting four characteristic genes (ABCG1, GNG11, RPL24, and RPS28). Expression and ROC verification identified three key genes—ABCG1, GNG11, and RPL24. To further investigate the functions of these key genes, GeneMANIA and GSEA analyses were conducted, followed by the creation of miRNA regulatory networks and drug predictions. Finally, single-cell analysis provided insights into the expression of key genes during macrophage differentiation through cell communication and pseudotime analysis, offering new perspectives for MI treatment.

Discussion

AMI is one of the leading causes of death globally. It is triggered by the sudden partial or complete occlusion of the coronary artery, resulting in a sharp reduction or interruption of blood flow, which leads to severe and persistent myocardial ischemia and necrosis [29]. Peripheral blood monocytes are the main source of macrophages in infarcted myocardium, and their transcriptomic characteristics can reflect the local inflammatory activation, immune cell infiltration, and polarization status in the myocardium to a certain extent [30, 31].Macrophage polarization plays a pivotal role in the inflammatory response following AMI [32]. Macrophage polarization significantly influences cardiac healing after myocardial injury. In the acute phase of injury, M1 macrophages rapidly infiltrate the infarcted region, releasing a large number of pro-inflammatory cytokines, such as interleukin-6 (IL-6) and tumor necrosis factor-α (TNF-α), which promote cardiomyocyte apoptosis. In contrast, M2-like macrophages, which express arginase-1 (Arg1) and anti-inflammatory IL-10, help to reduce inflammation and facilitate tissue healing and remodeling [33, 34]. The polarization of macrophages is closely linked to their energy metabolism pathways. While classical M1 macrophages primarily rely on glycolysis for energy, M2 macrophages tend to utilize oxidative phosphorylation [35, 36]. Therefore, regulating macrophage polarization and inhibiting the pro-inflammatory cascade have become central targets in current MI treatment research. Through bioinformatics analysis, this study identifies macrophage and MRGs as key players in MI. Three key genes, ABCG1, GNG11, and RPL24, were identified. ABCG1 was significantly up-regulated in the MI group, whereas GNG11 and RPL24 were markedly down-regulated. These findings offer new potential avenues for MI treatment.

Among these key genes, ABCG1 is a member of the ATP-binding cassette (ABC) transporter family. Located on the long arm of chromosome 21 [37], ABCG1 gene consists of 23 exons [38]. The protein encoded by ABCG1 is involved in regulating lipid metabolism, particularly the transport of cholesterol and phospholipids [39]. The primary function of ABCG1 is to transport free cholesterol and phospholipids to the precursors of high-density lipoprotein (HDL) particles, specifically β-HDL particles [40, 41], rather than lipid-free ApoA-I [42–44]. This process aids in the removal of excess cholesterol from tissue cells [45]. ABCG1 plays a key role in lipid metabolism, inflammatory regulation, and macrophage polarization, and its abnormal expression is closely associated with inflammatory responses, myocardial injury, and cardiac functional repair after myocardial infarction [46, 47]. Inflammatory and stress responses induced by acute myocardial infarction can trigger metabolic reprogramming in macrophages/monocytes. As a key cholesterol transporter, ABCG1 modulates cholesterol efflux and lipid homeostasis, thereby affecting cell membrane fluidity, mitochondrial function, and inflammatory signaling pathways, and shifting metabolic patterns such as glycolysis/oxidative phosphorylation, which in turn regulates macrophage polarization and inflammatory phenotypes. Therefore, ABCG1 not only participates in lipid transport but also acts as an important molecule governing immune cell metabolic reprogramming, and is closely associated with the myocardial microenvironment and systemic inflammation [48–52].ABCG1 not only reduces oxidative stress but also maintains endothelial integrity, regulates cholesterol levels, and modulates vascular reactivity. It plays a key role in lipid homeostasis and the inflammatory response of endothelial cells, providing protective effects against atherosclerosis [53–55]. Furthermore, ABCG1 assists in removing excess cholesterol from HDL in macrophages [56], preventing foam cell formation, which is a critical step in the development of atherosclerosis. Additionally, ABCG1 helps clear oxidized sterols and other toxic molecules from macrophages [57, 58]. Deficiency of ABCG1 in macrophages leads to the accumulation of lipids within these cells, exacerbating the development of atherosclerotic lesions.

GNG11, a member of the guanine nucleotide-binding protein gamma family, encodes a membrane protein anchored in lipids. This protein plays a pivotal role in transmembrane signaling systems as part of the heterotrimeric G protein complex [59]. GNG11 is involved in immune cell signal transduction and the regulation of monocyte/macrophage function, and its expression level in peripheral blood can indirectly reflect the intensity of local myocardial inflammatory response and the degree of tissue damage [60, 61].SNPs rs4262 and rs180238 in GNG11 are linked to resting heart rate variability (HRV) and anxiety disorders through HRV levels [62]. Specifically, the C allele of these SNPs is associated with lower HRV, which may lead to reduced expression and availability of the γ-11 subunits. This change could affect cardiac activity through the G protein heterotrimer in the GIRK signaling pathway [63]. Additionally, genome-wide association studies (GWAS) have linked GNG11 loci to resting heart rate [33, 34, 63, 64].RPL24 is a ribosomal protein located on the 3q12.3 region of human chromosome 3. It is a protein-coding gene, and its product is ribosomal protein L24. RPL24 is a component of the large ribosomal subunit, belonging to the L24E family of ribosomal proteins, and is localized in the cytoplasm. RPL24 is essential for ribosome assembly, structural stability, and normal ribosome function, all of which are critical for intracellular protein synthesis. RPL24 is expressed in various tissues, including the ovary, endometrium, and heart, with its expression levels varying across different tissues [65, 66]. In cardiomyocytes, RPL24 plays a fundamental role in protein synthesis and the maintenance of normal cell function. Abnormal expression of RPL24 can impair the tolerance and repair capacity of cardiomyocytes to ischemia, hypoxia, and other injuries, indirectly affecting the onset, progression, and prognosis of MI [67].

Finally, GSEA revealed that these three key genes—ABCG1, GNG11, and RPL24—are enriched in multiple biological pathways, including Huntington’s disease, olfactory transduction, oxidative phosphorylation, and proteasome pathways. ABCG1, GNG11, and RPL24 are jointly enriched in Huntington’s disease-related processes, which overlap with fundamental cellular mechanisms in its pathogenesis, such as mitochondrial dysfunction and protein homeostasis imbalance [68, 69]. These pathways are closely linked to the inflammatory response and fibrosis process in MI. Alterations in oxidative phosphorylation can impact macrophage energy metabolism, potentially influencing the repair processes during the later stages of MI [35, 36]. Proteasome activity also plays a critical role in regulating intracellular protein degradation and the inflammatory response, representing a potential therapeutic target for cardiac remodeling after MI [70]. In terms of drug predictions, ABCG1 was found to be associated with drugs such as aspirin and vitamin E, both of which are important in cardiovascular disease prevention and treatment. Aspirin reduces MI risk by inhibiting platelet aggregation and reducing thrombosis [71]. Vitamin E, a potent antioxidant, regulates various intracellular signaling pathways, which have beneficial effects on heart tissue protection [72].These findings provide novel insights and.

potential directions for the future development of MI treatment strategies.

This study has several limitations. First, the specific mechanisms of the screened key genes in myocardial infarction have not been fully elucidated. Although potential targeted drugs were predicted, their clinical applicability and therapeutic efficacy still require validation by extensive evidence‑based medical evidence. Second, this study only conducted preliminary validation with a small sample size; the insufficient sample size limits the reliability of the identified genes as clinical biomarkers, and the conclusions need to be further confirmed in large-sample clinical cohorts. Meanwhile, the public databases used in this study, including GSE61145, GSE61144, and GSE267644, have inherent deficiencies in clinical data, such as a lack of key information including myocardial infarction subtypes and precise blood sampling time points. These limitations prevent the correction of the dynamic effects of disease progression on the transcriptome and introduce uncertainty into the interpretation of the results. In addition, a relatively lenient quality control threshold for mitochondrial genes was applied to the PBMC samples in the single-cell sequencing analysis, which may affect the accurate identification of metabolism-related genes. Although the core genes were screened by machine learning and validated in multiple independent datasets, which reduced bias to a certain extent, prospective studies are still needed in the future. Samples with complete clinical information should be collected, and stricter quality control standards should be adopted to further dissect the gene functions and therapeutic value, as well as clarify their correlations with cardiac functional parameters, so as to provide a solid basis for clinical translation.

Conclusions

This study identified ABCG1, GNG11, and RPL24 as key genes related to macrophage activity and MR in AMI, suggesting their involvement in signal transduction and the pathogenesis of AMI. These findings offer valuable insights into MI pathogenesis and may contribute to enhancing clinical diagnosis and treatment strategies.

Abbreviations

MI

Myocardial Infarction

MR

Metabolic Reprogramming

MRGs

MR-Related Genes

DEGs1

Differentially Expressed Genes

ROC

Receiver Operating Characteristic

RT-qPCR

Reverse Transcription Quantitative Polymerase Chain Reaction

HVGs

Highly Variable Genes

PCA

Principal Component Analysis

PCs

Principal Components

UMAP

Uniform Manifold Approximation and Projection

HCA

Human Cell Atlas

GO

Gene Ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

SVM-RFE

Support Vector Machine-Recursive Feature Elimination

LASSO

Least Absolute Shrinkage and Selection Operator

AUC

Area Under the Curve

GSEA

Gene Set Enrichment Analysis

miRNAs

MicroRNAs

RT-qPCR

Reverse Transcription Quantitative Polymerase Chain Reaction

Authors’ contributions

TM: visualization, writing-original draft.YPQ: formal analysis.YY and JY: investigation, data curation, software.QZ and PM: methodology, resources.SJX and LJG: Conceptualization, supervision.LGL: Conceptualization, writing-review and editing. All authors read and approved the final manuscript.

Funding

This research was funded by a grant from Ningxia Natural Science Foundation (Project No.2025AAC030354 ).

Data availability

The datasets supporting the conclusions of this article are available in the Gene Expression Omnibus (GEO) repository, at the URL https://www.ncbi.nlm.nih.gov/geo/. The specific datasets are GES267644, GSE6114, and GSE61145.

Declarations

Ethics approval and consent to participate

The study was carried out in compliance with the 2024-revised Helsinki Declaration. The involvement of human participants was approved by the Ethics Committee of the People’s Hospital of Ningxia Hui Autonomous Region (Approval No. 2024-LL-183). Written informed consent was obtained from all participants prior to their inclusion in the study, ensuring their understanding and voluntary participation. All data collected were treated with strict confidentiality and anonymity to protect the privacy of the participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.