The Gene Expression Omnibus (GEO) database¹ was used to download the AMI related microarray datasets GSE66360 (mRNA), GSE61144 (mRNA), GSE60993 (mRNA), and GSE775 (mRNA). GSE66360 comes from the GPL570 platform ([HG-U133_Plus_2] Affymetrix Human Genome U133 Plus 2.0 Array). GSE61144 comes from the GPL6106 platform (Sentrix Human-6 v2 Expression BeadChip). GSE60993 comes from the GPL6884 platform (Illumina HumanWG-6 v3.0 expression beadchip). GSE775 comes from the GPL81 platform ([MG_U74Av2] Affymetrix Murine Genome U74A Version 2 Array). The samples in GSE66360, GSE61144 and GSE60993 were all derived from human blood and grouped into healthy control and AMI groups. The samples in GSE775 were derived from myocardial tissue of myocardial infarction mice. Groups in GSE775 were divided according to different time of infarction (1, 4, 24, 48 h, 1, and 8 week), and each group contained three sham operation mice and three infarction mice. Table 1 shows the information of these 4 datasets obtained from the GEO database.

Raw sequencing data of GSE66360 and GSE61144 were downloaded from GEO database. GEO2R² was used to perform the preliminary differential expression analysis. Benjamini-Hochberg method was used to correct adj. P for potential false positive results and adj. P-value < 0.05 and | logFC| > 0.8 was set as the cut-off criteria of DEGs. Venn diagram web tool³ was used to identify the CoDEGs in GSE66360 and GSE61144.

FerrDb V2⁴ is the world’s first database that dedicates to ferroptosis regulators and ferroptosis-disease associations. A total of 431 regulatory factors including drivers, suppressors and markers were downloaded from FerrDB V2 database (Supplementary File 1). The DEGs in GSE66360 and GSE61144 were intersected with the genes obtained from FerrDB V2 to obtain DEGs associated with Ferroptosis. Based on the raw sequencing data in GSE66360 and GSE61144, the heatmap of these differentially expressed FRGs were produced using the gg plots package of R software (version: x64 3.2.1) (17).

The online biological tool, Database for Annotation, Visualization, and Integrated Discovery (DAVID; version6.8)⁵ was used to perform Gene Ontology (GO) analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of CoDEGs in GSE66360 and GSE61144. Pathways with P < 0.05 were considered statistically significant.

In order to understand the mechanism of interaction between CoDEGs in AMI, a protein-protein interaction (PPI) network was constructed by string database (version10.0)⁶. This obtained PPI network was further visualized using Cytoscape software (version 3.7.1). The Cytohubba plug-in of Cytoscape software was used to screen out the top 10 key genes with high connectivity in this PPI network using the Maximal Clique Centrality (MCC) algorithm (18, 19). Multiple algorithms (Degree, EPC, Betweenness) continued to be used for prediction, and genes predicted by more than three methods at the same time are considered to be Hub Genes for AMI.

GPX4 is significantly downregulated during ferroptosis occurrence and is a well-recognized marker of the ferroptosis pathway. The expression of GPX4 was validated in multiple datasets (GSE60993, GSE61144, and GSE775) to confirm whether ferroptosis was associated with AMI or not.

GSE60993 was an AMI-related sequencing dataset comprising 7 healthy subjects, 9 patients with unstable angina (UA), 10 patients with non-ST segment elevation myocardial infarction (NSTEMI), and 7 patients with ST segment elevation myocardial infarction (STEMI). Based on the relative gene expression data in GSE60993, the abovementioned differentially expressed FRGs in GSE66360 and GSE61144 were further verified to study their expression differences in coronary artery disease (CAD).

Differentially expressed FRGs in validation dataset GSE60993 were considered as potential biomarkers for AMI. Based on the expression data of these FRGs in GSE66360, Graphpad Prism (version:8.0) software was used to compare the control and AMI groups. Independent sample T-test was used to compare the two groups and P < 0.05 was considered statistically significant. ROC analysis was used to evaluate the diagnostic value of these FRGs for AMI. FRGs with the area under the curve (AUC) ≥ 7.0 and P < 0.05 were screened as potential biomarkers.

The GSE775 dataset included myocardial tissue transcriptome data at 1, 4, 24, 48 h, 1 and 8 week after myocardial infarction in mice. Based on their raw expression data in GSE775, the expression of FRGs were further verified to study the change of expression of FRGs in myocardial infarction tissues over time. The comparison between Sham operation group and AMI group was performed by independent sample T test. P < 0.05 was considered statistically significant.

The human cardiomyocyte AC16 cell line purchased from BeNa Culture Collection (Beijing, China) was cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco) with 10% fetal bovine serum (FBS, Gibco) and 1% penicillin-streptomycin (Sigma, St. Louis, MO, USA) at 37°C with 5% CO₂. The three-gas incubator was used to establish an anoxic environment with 1% O₂, 5% CO₂, and 94% N₂. The AC16 cardiomyocytes were exposed to 24 h of this condition to generate a hypoxia model. AC16 cells in the control group were under normoxia condition (21% O₂, 5% CO₂, and 74% N₂) all the time.

The expression of these FRGs in AC16 hypoxia cell model was further verified by qPCR. According to the manufacturer’s protocol, total RNA was isolated using FastPure^® Cell/Tissue Total RNA Isolation Kit V2 (Vazyme, Nanjing, China), and then synthesized with Hifair^® III 1st Strand cDNA Synthesis SuperMix for qPCR (Ye Sen, Shanghai, China). The Bio-Rad CFX96 Real-time PCR Detection System was performed to carry out the qRT-PCR, using Hieff UNICON^® Universal Blue qPCR SYBR Green Master Mix (Ye Sen, Shanghai, China). β-actin was used as the reference gene and the relative fold change was calculated by the method of 2^–ΔΔCt. Details of all primers were presented in Supplementary File 2. Data were presented as mean values ± standard error of mean (SEM) from at least three independent experiments. Independent sample T test was used to compare the normoxia group and hypoxia group. FRGs with P < 0.05 was considered statistically significant.

The AMI related datasets GSE66360 and GSE61144 were downloaded from the Gene Expression Omnibus (GEO) database. Based on the screening criteria mentioned above (P value < 0.05 and | logFC| > 0.8), DEGs in each dataset were screened out. There were 1232 DEGs, including 688 high-expressed, and 544 low-expressed, in GSE66360. There were 468 DEGs, including 325 high-expressed, and 143 low-expressed, in GSE61144. Supplementary File 3 presents detailed results of differential expression analysis. The top 5 upregulated DEGs and the top 5 downregulated DEGs in each dataset are shown in Table 2. The volcano plots of these DEGs in GSE66360 and GSE61144 are shown in Figures 2A,B. Based on the obtained DEGs and the ferroptosis related genes in FerrDB V2, the online venn tool was used to identify CoDEGs and differentially expressed FRGs (Figure 2C). A total of 131 CoDEGs were identified in GSE66360 and GSE61144. There are 45 and 14 differentially expressed FRGs in these two datasets respectively (Supplementary File 4). Seven ferroptosis related genes ACSL1, CREB5, GABARAPL1, LAMP2, PANX2, PGD, and TLR4 were significantly differentially expressed in both datasets. Figure 2D shows the heatmap of differentially expressed FRGs in GSE66360. Figure 2E shows the heatmap of differentially expressed FRGs in GSE61144.

Common DEGs in GSE66360 and GSE61144 were input into DAVID database for GO annotation and KEGG pathway enrichment analysis. Figures 3A–D shows the top 10 enriched GO annotation terms and KEGG pathways. For GO biological process (BP) analysis, the results showed that these CoDEGs are mainly enriched in the term of “innate immune response,” “inflammatory response,” “immune response,” “response to lipopolysaccharide,” and “defense response to bacterium” (Figure 3A). For GO cellular component (CC) analysis, the top 5 significantly enriched terms are “plasma membrane,” “integral component of membrane,” “extracellular exosome,” “integral component of plasma membrane,” and “extracellular region” (Figure 3B). For GO molecular function (MF) analysis, the top 5 significantly enriched terms are “receptor activity,” “protein homodimerization activity,” “calcium ion binding,” “carbohydrate binding,” “RAGE receptor binding” (Figure 3C). Furthermore, “Hematopoietic cell lineage,” “Phagosome,” “Cytokine-cytokine receptor interaction,” “Toll-like receptor signaling pathway,” and “TNF signaling pathway” are significant enrichment pathways in KEGG analysis (Figure 3D).

The CoDEGs were introduced into the String database. Based on the String database’s default execution criteria, a PPI network with 101 nodes and 370 edges was constructed and then visualized using Cytoscape software (Figure 3E). The 91 orange red nodes in this network represent high-expressed genes and the 10 blue nodes in this network represent low-expressed genes. The depth of the color represents the degree of upward or downward movement in the corresponding direction. Next, based on MCC algorithm, the top 10 genes (TLR2, TREM1, TLR8, TLR4, ITGAM, S100A9, S100A12, S100A8, NCF2, and MMP9) were screened out using the plug-in Cytohubba in Cytoscape (Figure 3F). KOBAS (version 3.0)⁷ online tool was used for further path enrichment analysis of these genes. The result shows that TLR2, TLR4, NCF2, and ITGAM are involved in “Phagosome” pathway, S100A8, S100A9, and MMP9 are involved in “IL-17 signaling pathway” pathway, TLR2, TLR4, and TLR8 are involved in “Toll-like receptor signaling pathway” pathway, NCF2, ITGAM, and MMP9 are involved in “Leukocyte transendothelial migration” pathway. In addition, the other three algorithms (Degree, EPC, Betweenness) were continued to be used for prediction, and eventually the genes predicted by at least three algorithms simultaneously were considered as Hub genes (Supplementary File 5 and Supplementary Figure 1). Finally, the top 7 hub genes ITGAM, S100A12, S100A9, TLR2, TLR4, TLR8, and TREM1 were considered as Hub Genes of AMI.

Based on their raw expression data in GSE66360 and GSE61144, FRGs in healthy and AMI patients were compared and analyzed. The results show that there are 45 and 14 FRGS in the dataset GSE66360 and GSE61144, respectively. The 45 FRGs in GSE66360 were all significantly differentially expressed, with high expression of all genes except for the low expression of nine genes, ACADSB, CIRBP, DCAF7, FZD7, MDM2, METTL14, PARP11, PDSS2, and TGFBR1 (Figure 4A). All 14 FRGs in GSE61144 were significantly increased except for the low expression of PRDX1 (Figure 4B). Figure 4C shows the Venn results of FRGs in these two datasets. Further PPI analysis and enrichment analysis were performed on FRGs mentionedabove. Figure 4D, consisting of 37 nodes and 121 edges, is the PPI network constructed by 45 FRGS identified in GSE66360. Similarly, Figure 4E, with 13 points and 33 edges, is the PPI network constructed by 14 FRGS identified in GSE61144. KEGG pathway enrichment analysis shows that FRGs in GSE66360 are significantly enriched in the term of “Ferroptosis,” “Necroptosis,” “NOD-like receptor signaling pathway,” “AGE-RAGE signaling pathway in diabetic complications,” and “Autophagy” (Figure 4F). FRGs in GSE61144 are significantly enriched in the term of “Autophagy,” “Adipocytokine signaling pathway,” “Peroxisome,” “AMPK signaling pathway,” “Fatty acid biosynthesis,” “Ferroptosis,” “Fatty acid metabolism,” “PI3K-Akt signaling pathway,” and “PPAR signaling pathway” (Figure 4G).

The expression of GPX4 was verified in datasets GSE61144, GSE60993, and GSE775. As predicted, the expression of GPX4 was significantly down-regulated in all three datasets (Figures 5A–C), suggesting that the biological process of ferroptosis may be involved in AMI.

The expression of 52 FRGs in GSE66360 and GSE61144 were further verified in external dataset GSE60993 (Figures 6A–T). Compared with healthy controls, a total of 19 FRGs, ACSL1, ACSL4, ALOX5, ATG7, CAMKK2, CREB5, GABARAPL1, KDM6B, LAMP2, PANX2, PGD, PROK2, PTEN, PTGS2, SAT1, STAT3, TLR4, ULK1, and ZFP36 were significantly up-regulated in STEMI patients, but only PRDX1 was down-regulated in STEMI patients. In addition, the expression of PANX2 was significantly different between CON and UA groups, the expression of GABARAPL1, LAMP2, PANX2, PTEN, and ULK1 was significantly different between NSTEMI and STEMI groups.

Based on the raw expression data in GSE66360, ROC analysis was performed on these 20 differentially expressed FRGs (ACSL1, ACSL4, ALOX5, ATG7, CAMKK2, CREB5, GABARAPL1, KDM6B, LAMP2, PANX2, PGD, PRDX1, PROK2, PTEN, PTGS2, SAT1, STAT3, TLR4, ULK1 and ZFP36) and AUC values were calculated. Finally, the AUC of these 20 FRGs were all greater than 0.6, and 13 of them, ACSL1 (AUC = 0.8727), ATG7 (AUC = 0.7310), CAMKK2 (AUC = 0.7124), GABARAPL1 (AUC = 0.9016), KDM6B (AUC = 0.7286), LAMP2 (AUC = 0.7045), PANX2 (AUC = 0.8365), PGD (AUC = 0.7882), PTEN (AUC = 0.7257), SAT1 (AUC = 0.7980), STAT3 (AUC = 0.7792), TLR4 (AUC = 0.8290), and ZFP36 (AUC = 0.8796), with AUC ≥ 0.7 were identified as potential markers of AMI. Figures 7A–L shows the ROC analysis results of the top 12 AUC values. The ROC analysis results for the remaining eight FRGs are shown in Supplementary File 5 and Supplementary Figure 2.

The expression of 20 differentially expressed FRGs (ACSL1, ACSL4, ALOX5, ATG7, CAMKK2, CREB5, GABARAPL1, KDM6B, LAMP2, PANX2, PGD, PRDX1, PROK2, PTEN, PTGS2, SAT1, STAT3, TLR4, ULK1, and ZFP36) were further verified to study the expression changes of these FRGs in myocardial infarction tissues with time. As shown in Figures 8A–I, 9 FRGs (ACSL4, ATG7, LAMP2, PGD, PTEN, PTGS2, SAT1, STAT3, and ZFP36) were significantly differentially expressed at least twice at the six test time points (1, 4, 24, 48 h, 1, and 8 week after myocardial infarction). All these FRGs showed an upward trend in expression. The expression of STAT3 and ZFP36 was different only within 24 h after myocardial infarction. The differences in ATG7 expression and PGD expression did not occur until more than 48 h after myocardial infarction. The expression of ACSL4, LAMP2, PTEN, PTGS2, and SAT1 were different in the early and late stages of myocardial infarction.

The expression of 20 differentially expressed FRGs (ACSL1, ACSL4, ALOX5, ATG7, CAMKK2, CREB5, GABARAPL1, KDM6B, LAMP2, PANX2, PGD, PRDX1, PROK2, PTEN, PTGS2, SAT1, STAT3, TLR4, ULK1, and ZFP36) were further verified to study the effect of hypoxia on FRGs. The results showed 12 of them were differentially expressed (Figures 9A–L). They were ALOX5 (up), CAMKK2 (up), CREB5 (down), GABARAPL1 (down), KDM6B (up), LAMP2 (up), PGD (down), PTEN (up), PTGS2 (up), STAT3 (down), ULK1 (up), and ZFP36 (down). The expression of ALOX5, CAMKK2, KDM6B, LAMP2, PTEN, PTGS2, and ULK1 were consistent with the predicted result.