This study systematically elucidated the regulatory network of differentially expressed ferroptosis-related genes (DE-FRGs) in AMI. Firstly, based on the Gene Expression Omnibus (GEO) database GSE61144 expression profile, differentially expressed mRNAs (DE-mRNAs) in AMI were screened and intersected with the ferroptosis-related genes to obtain DE-FRGs, followed by GO/KEGG (Gene Ontology/Kyoto Encyclopedia of Genes and Genomes) enrichment analysis, PPI (protein-protein interaction) network construction, expression heatmap visualization, and tissue-expression landscape mapping with FUMA-GENE2FUNC to examine their biological functions. Subsequently, integrating cis-eQTLs data from the genotype-tissue expression (GTEx) project v8 (54 tissues) with the AMI genome-wide association study (GWAS) summary statistics from FinnGen R12 (11622 cases and 207170 controls), we applied summary-data-based Mendelian randomization (SMR) coupled with heterogeneity in dependent instruments (HEIDI) and Bayesian colocalization analysis. These methods identified SP1 (Specificity Protein 1) as the sole DE-FRG whose expression changes show a causal relationship with AMI susceptibility. SP1, a ubiquitously expressed transcription factor that regulates genes involved in inflammation, oxidative stress, and cell survival, has been increasingly implicated in cardiovascular pathophysiology. In this study, we investigated its potential causal role in AMI through multi-omics integration. Upstream miRNAs targeting SP1 were predicted by integrating data from starBase, TargetScan and miRWalk. Downstream targets of SP1 were identified using KnockTF and TRRUST, thereby constructing a comprehensive miRNA-TF-mRNA regulatory network. Further, using the GSE76604 and GSE4648 datasets, differentially expressed miRNAs and mRNAs were screened to pinpoint the miR-133bi/SP1/PLAUR axis. Ultimately, through prediction of transcription factor binding sites, GSEA and GeneMANIA interaction analysis, prediction of gene-diseases and gene-drugs associations, phenome-wide association study(PheWAS), ROC curve and RT-qPCR validation, the regulatory mechanism, diagnostic value and therapeutic potential of this axis in AMI were clarified. Figure 1 summarized and illustrated the overall workflow of this study.

This study utilized publicly available datasets, no human subjects or tissues were directly involved.

The mRNA and miRNA expression profiles of AMI (GSE61144, GSE4648 and GSE76604) were obtained from the GEO database (https://www.ncbi.nlm.nih.gov/geo/) (12–14). The number of samples included in each dataset and the corresponding platform information are described in detail in Table 1. A total of 2177 ferroptosis-related genes were obtained from the FerrDb (http://www.zhounan.org/ferrdb/current/) (15) and GeneCards (https://www.genecards.org/) (16) databases. 61 ferroptosis-related miRNAs were obtained from the ncFO (http://www.jianglab.cn/ncFO/) (17) database.

Differentially expressed mRNAs (DE-mRNAs) and miRNAs (DE-miRNAs) were identified using GEO2R (https://www.ncbi.nlm.nih.gov/geo/geo2r/), with a significant threshold of P<0.05 and |log₂ Fold change (FC)| > 0.263 between AMI and control samples. GEO2R is a free online tool developed by National Center for Biotechnology Information (NCBI) specifically for performing differential expression analysis on gene expression data stored in the GEO database. Visualization of DE-mRNAs was achieved through volcano plot. Then, by intersecting DE-mRNAs with ferroptosis-related genes (FRGs) and DE-miRNAs with ferroptosis-related miRNAs, and visualizing them with Venn diagrams, we identified differentially expressed ferroptosis-related genes (DE-FRGs) and miRNAs (DE-FRmiRNAs), respectively.

To explore the specific mechanism by which DE-FRG drives AMI pathology, we performed Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway and Gene Ontology (GO) enrichment analyses to identify the significant pathways with Metascape (https://metascape.org) (18). Results were visualized as bar and bubble charts.

The PPI network can cluster gene sets into tightly connected modules based on protein interaction strength and identify potential hub proteins through network topology metrics. Therefore, the STRING (https://cn.string-db.org/) database was used to construct the PPI network to find key DE-FRGs.

Based on the GSE61144 dataset, the top 50 DE-FRGs with the highest fold changes were selected for heatmap visualization to compare their expression differences between the control and MI groups. Furthermore, their expression distribution characteristics in 54 GTEx tissues were evaluated using FUMA-GENE2FUNC (https://fuma.ctglab.nl/gene2func).

Immune infiltration analysis was performed to examine immune microenvironment changes in AMI. Specifically, the CIBERSORT tool was used to analyze the immune cell infiltration abundance of 22 immune cells in the GSE61144 expression profile dataset (19). Additionally, Using the three major R packages “ggpubr, corrplot, and reshape2”, we visualized the results of the immune infiltration analysis: stacked bar charts showed the proportion of different immune cell subsets in each sample, correlation heatmaps systematically analyzed the interaction network among immune cells, and boxplots displayed the differences in various immune cells between the control and MI groups to identify immune cells with significantly different expression in the MI group.

Mendelian randomization (MR) is a statistical method that uses genetic variants as instrumental variables to infer potential causal relationships between exposures and outcomes (20). Summary-data-based Mendelian randomization (SMR) extends the concept of MR, serving as an efficient genetic analysis method that integrates summary statistics from GWAS and cis-expression quantitative trait loci (cis-eQTL) information, significantly enhancing research efficacy (21). SMR similarly adheres strictly to the three core assumptions of Mendelian randomization: (1) Relevance Assumption: genetic variation is robustly and significantly associated with the exposure factor to avoid “weak instrument variable” bias; (2) Independence assumption: genetic variation is independent of any confounding factors; (3) Exclusivity assumption: genetic variation affects the outcome only through the exposure pathway, and there is no horizontal pleiotropy pathway independent of the exposure (22). Compared with traditional MR, SMR utilizes cis-QTL instrumental variables and summary data, providing higher statistical power in independent large samples. The accompanying HEIDI test distinguishes linkage from pleiotropy, enabling causal inference across multi-omics layers such as gene expression, methylation, and protein abundance. In the SMR analysis, we corrected for multiple testing using the Benjamini–Hochberg (BH) procedure to control the false discovery rate (FDR). SMR p-values for all exposure–outcome associations were adjusted using the p.adjust function in R with the BH method. Associations with an FDR-adjusted p-value < 0.05 were considered statistically significant. SMR is better suited for publicly available GWAS/QTL data and has become an important method for efficiently dissecting gene-phenotype mechanisms (21).

The databases for exposures and outcomes were obtained from different sources, which can reduce the possibility of population overlap and minimize potential bias in MR estimates (23). For exposure, we selected cis-eQTLs data from four types of tissues (coronary artery, atrial appendage, left ventricle and whole blood) in the GTEx database (https://www.gtexportal.org/) (24), which is a large-scale genomics project dedicated to exploring how genetic variation influences gene expression across human tissues. We selected these four tissues based on their key role in the pathogenesis of AMI, as well as the strong evidence from previous studies in the field of cardiovascular genomics. In the GTEx database, coronary arteries are the main sites for the formation and rupture of atherosclerotic plaques (25, 26); heart left ventricle is the main myocardial region affected by ischemic injury, with gene expression reflecting cardiac remodeling and dysfunction (27); heart atrial appendage, while not directly injured in AMI, is closely linked to post-AMI complications such as atrial fibrillation (28); whole blood captures systemic inflammatory processes that drive plaque instability and thrombosis, and is frequently used in cardiovascular disease diagnosis (29). Together, these tissues comprehensively represent the vascular lesion site, affected myocardium, and systemic immune environment, providing a robust pathophysiological foundation for our SMR analysis. For outcoms, the four GWAS datasets related to AMI were acquired from the FinnGen database (https://www.finngen.fi/en/access_results) (30) and IEU Open GWAS (https://opengwas.io/), detailed characteristics of outcome datasets are shown in Table 2. In this study, we used the FinnGen AMI dataset as the discovery cohort and validated our findings in three additional independent AMI datasets.

In this study, we used the SMR method to explore the potential causal relationship between DE-FRGs and AMI using SMR software (version 3.1.1 https://yanglab.westlake.edu.cn/software/smr/) (21). We retained significant SMR probes based on an FDR-corrected P < 0.05 and required the HEDI P > 0.05 to indicate the absence of heterogeneity. Key parameter settings were as follows: Gene region definition: The cis-region was defined as ±1 Mb around the transcription start site (TSS) of each gene, consistent with default and commonly used practice in eQTL-based MR studies (31). eQTL screening threshold: Only independent cis-eQTLs with p < 5×10^-8 in the exposure (eQTL) dataset were used as instrumental variables. This stringent threshold ensures high-confidence eQTL instruments and minimizes inclusion of weak or spurious associations. Additional default settings were applied, including LD reference from the 1000 Genomes Project (Phase 3, EUR population) and HEIDI test with default window size (1000 kb) and LD r² threshold of 0.9 for clumping. We selected four types of tissues (left ventricle, atrial appendage, coronary artery and whole blood) from the GTEx database to perform four SMR analyses. The positive results of the four SMRs were presented using a forest plot. Subsequently, we took the intersection of the positive results of the four SMRs to identify key gene that have a significant causal relationship with AMI in all four tissues.

We performed Bayesian colocalization analysis using the “coloc” R package to determine whether GWAS and eQTL signals share a common causal variant (32). This analysis computes posterior probabilities for five hypotheses (PPH): H0: no association with either trait, H1: association with only trait 1, H2: association with only trait 2, H3:association with both traits due to different causal variants and H4: association with both traits due to the same causal variant. Following previous literature, we adopted a posterior probability of colocalization (PPH4) > 0.8 as stringent evidence that the expression-associated and disease-associated signals share the same causal variant (33). We used the key gene obtained from the above SMR for colocalization analysis, and visualized the results using the “locuscomparer” R package.

To further explore the regulatory mechanism of the key transcription factor in AMI, we constructed the miRNA-TF-mRNA network. Firstly, upstream miRNAs targeting the transcription factor were predicted with starBase (https://rnasysu.com/encori/index.php) (34), TargetScan (https://www.targetscan.org/vert_80/) (35) and miRWalk (http://mirwalk.umm.uni-heidelberg.de/) (36), while its downstream target genes were identified using knockTF (http://www.licpathway.net/KnockTF/) (37) and TRRUST (https://www.grnpedia.org/trrust/) (38). Next, we performed differential expression analysis using datasets GSE4648 and GSE76604 to obtain DE-mRNAs and DE-miRNAs, respectively, and then intersected them with the predicted mRNAs and miRNAs to further pinpoint the ferroptosis-mediated miRNA-TF-mRNA regulatory pathways in AMI. Finally, we visualized the miRNA-TF-mRNA network using Cytoscape software.

To comprehensively analyze the regulatory interplay among miRNA, TF and mRNA, and to pinpoint their potential binding sites for subsequent experimental validation, this study adopted the following bioinformatics workflow. First, the StarBase (https://rnasysu.com/encori/index.php) database was employed to predict the binding sites between selected miRNA and TF. Next, the JASPAR (https://jaspar.elixir.no/) database was used to predict the promoter regions of target mRNA for putative TF-binding motifs. Finally, the predictions from both databases were integrated to construct a systematic miRNA-TF-mRNA regulatory network.

Hub genes were inputted to GeneMANIA (https://genemania.org/) to retrieve functionally related genes and build a PPI network, along with data on co-expression, co-localization, shared pathways, and protein-DNA or protein-protein interactions (39). In this study, we explored the top 20 functionally similar genes of hub genes and performed functional enrichment analysis. Use the “ClusterProfiler” package to perform Gene Set Enrichment Analysis (GSEA) on each core gene to obtain the important regulatory pathways involved by the genes. The subset of KEGG pathways in the “C2: curated gene sets” database of MSigDB was used as a reference gene set, and the screening threshold was set at an adj. P-value < 0.05. Finally, the analysis results were visualized with ridge plots based on normalized enrichment scores (NES) using the “enrichplot” package (40).

Gene-drug and gene-disease interactions for the two core genes were independently obtained from the CTD (https://ctdbase.org/) database. Entries were ranked by descending inference score, and the top 20 were retained for each gene. The intersections yielded six common drugs and seventeen common diseases.

Through the AstraZeneca PheWAS Portal (https://azphewas.com/) database, the associations between candidate genes and various phenotypes can be systematically assessed, thereby inferring their phenotypic pleiotropy and potential adverse reaction risks as drug targets (41). This database hosts summary statistics for approximately 15,500 binary and 1,500 quantitative traits.

ROC curves were constructed using SPSS software based on the normalized expression matrix data from the GSE61144 and GSE76604 datasets to assess the diagnostic performance of key genes in AMI. The AUC (area under the curve) represents the area under the ROC curve and quantifies the overall performance of the diagnostic model. Higher AUC values (closer to 1) suggest better diagnostic performance of DE-FRGs or DE-FRmiRNA for AMI.

H9C2 cardiomyocytes were purchased from the American Type Culture Collection (ATCC) and maintained at 37°C under 5% CO₂ in DMEM medium with low bicarbonate containing 10% serum. To model AMI in vitro, cells were subjected to oxygen–glucose deprivation (OGD) by replacing the medium with glucose-free, serum-free DMEM and incubating them in a modular hypoxia chamber flushed with 95% CO₂/5% N₂ for 4 h at 37°C.

To validate the biomarkers, reverse transcription-quantitative polymerase chain reaction (RT-qPCR) was performed on H9C2 cardiomyocytes. Total RNA was extracted using the TransZol Up Plus RNA Kit, and cDNA synthesis and qPCR was carried out using the TransScript^® Uni Cells to CT 1-Step Probe Kit. β-actin was used as the endogenous reference gene, and the 2^−ΔΔCt method was employed to calculate relative expression levels, with statistical significance defined as P < 0.05. PCR primers were designed by Sengon Biotech and primer sequences are provided in Table 3.

In the GSE61144 dataset, 2,292 significantly differentially expressed genes were identified between the AMI and the healthy group, of which 1,063 genes were upregulated in the myocardial infarction group and 1,229 genes were downregulated (Figure 2A). By intersecting 2,177 ferroptosis-related genes with 2,292 DE-mRNAs, 269 DE-FRGs were obtained (Figure 2B), providing key clues for further exploration of the ferroptosis mechanism.

GO and KEGG enrichment analyses were performed on 269 DE-FRGs to gain insights into their functional significance. GO enrichment analysis showed that in the biological process category, the DE-FRGs were mainly involved in “response to oxidative stress” (P = 3.5E-2) and “cellular response to chemical stress” (P = 2.0E-6); in terms of molecular function, “ubiquitin-like protein ligase binding” (P = 2.7E-10) and “antioxidant activity” (P = 3.4E-10) were significantly enriched; cellular component analysis indicated that the gene products were predominantly located in the “cytosolic ribosome” (P = 9.7E-12) and “ficolin-1-rich granule” (P = 1.1E-11) (Figure 2C). KEGG pathway enrichment analysis demonstrated that the DE-FRGs were significantly enriched in multiple signaling pathways, including lipid metabolism and atherosclerosis (P = 9.8E-14), FoxO (P = 1.7E-11), HIF-1 (P = 2.0E-11), mitophagy (P = 2.7E-9) and ferroptosis (P = 8.5E-8) signaling pathways (Figure 2D). Together, these findings implicate the DE-FRGs in oxidative injury, hypoxia adaptation, mitochondrial quality control and lipid-driven ferroptosis as potential drivers of AMI progression.

A PPI network was constructed based on 269 DE-FRGs using the STRING database. Only interactions with a confidence score > 0.9 were retained to ensure high reliability. The resulting network was visualized using cytoscape software (Figure 2E). The Molecular Complex Detection (MCODE) plugin was applied to identify functionally significant gene clusters. Three tightly connected gene modules were ultimately identified (Figures 2F–H). The key gene SP1 (specificity protein 1) we studied is in the third cluster.

The heatmap in Figure 3A illustrated the expression profiles of the top 50 DE-FRGs (ranked by fold-change) across control and AMI samples, clearly separated the two groups and provided an intuitive foundation for further gene-expression analyses. From the heatmap, it can be seen that the expression of the SP1 and PLAUR (urokinase-type plasminogen activator receptor gene) is increased in AMI samples. Using FUMA-GENE2FUNC, we analyzed 50 DE-FRGs expression patterns across GTEx v8 tissues (Figure 3B). As shown in Figure 3C, these 50 genes exhibited differential expression across the four examined tissues: coronary artery, atrial appendage, left ventricle, and whole blood.

The immune cell composition of the myocardium has been recognized as a critical determinant of infarct expansion and post-myocardial infarction remodeling. Through the box plot (Figure 3D), we gained valuable insights into the differential expression of various immune cells between the control and AMI samples. Figure 3E depicted the relative abundance of 22 immune cell subsets in each sample from GSE60993. A robust positive relationship was observed between CD8⁺ T cells and resting mast cells (r = 0.70), and a pronounced inverse association was found between neutrophils and regulatory T cells (r = -0.82) (Figure 3F), providing a comprehensive overview of the immune-cell interaction landscape following MI.

SMR was employed to estimate the causal effects of the 269 DE-FRGs on AMI. For the finn-b-I9_MI outcome, the forest plot (Figure 4A) vividly displays the positive outcomes from the SMR analysis in four tissues [coronary artery (pSMR = 0.001, pHEIDI = 0.268, OR = 1.13, 95%CI = 1.05-1.22), atrial appendage (pSMR = 0.001, pHEIDI = 0.218, OR = 1.09, 95%CI = 1.03-1.16), left ventricle (pSMR = 0.007, pHEIDI = 0.255, OR = 1.19, 95%CI = 1.05-1.34) and whole blood (pSMR = 0.008, pHEIDI = 0.061, OR = 1.40, 95%CI = 1.09-1.80)]. Within the plot, red dots symbolize a positive correlation between the genes and AMI, potentially indicating genes that may promote the development of AMI. Green dots, on the other hand, represent a negative correlation, suggesting genes that could have a protective effect against AMI. Blue dots highlight genes whose SMR associations remained significant across four tissues, indicating robust, cross-tissue reproducible evidence for putative causality.

Subsequently, we employed a Venn diagram (Figure 4B) to visually represent the positive results obtained from the four SMR analyses. Through this graphical approach, we identified a common intersecting gene SP1. This finding suggests that SP1 may exhibit a causal relationship with AMI across ventricular, atrial appendage, coronary artery and peripheral blood tissues.

Bayesian colocalization analysis provided strong evidence for shared genetic causality between the SP1 locus and AMI. Specifically, the LocusZoom plot (Figure 4C) revealed a robust colocalization signal at rs10876447 (chr12), with a posterior probability of hypothesis 4 (PPH4 = 0.81), implicating SP1 as a plausible mediator of genetic susceptibility to AMI.

The ebi-a-GCST011364 dataset is used for replication validation, the forest plot (Figure 5A) vividly displays the significant outcomes from the SMR analysis in four tissues [coronary artery (pSMR = 0.028, pHEIDI = 0.459, OR = 1.091, 95%CI = 1.010-1.178), atrial appendage (pSMR = 0.021, pHEIDI = 0.719, OR = 1.09, 95%CI = 1.014-1.185), left ventricle (pSMR = 0.046, pHEIDI = 0.652, OR = 1.126, 95%CI = 1.002-1.265) and whole blood (pSMR = 0.039, pHEIDI = 0.652, OR = 1.336, 95%CI = 1.014-1.760)]. The replication analysis using the ebi-a-GCST011364 dataset confirms the association between SP1 expression and AMI risk across all four tissues, with statistically significant SMR effects and non-significant HEIDI tests, indicating consistent evidence of a potential causal relationship. In this myocardial infarction dataset, we also identified another potential key gene—Macrophage Migration Inhibitory Factor (MIF). MIF exhibited OR values consistently below 1 across all four tissues, suggesting a potential causal relationship with AMI that is protective in nature. This finding provides a novel foundation for our future integrative research.

For the ebi-a-GCST011365 outcome, we can observe that SP1 has a potential causal relationship with AMI in left ventricular tissue (pSMR = 0.038, pHEIDI = 0.843, OR = 1.094, 95%CI = 1.005-1.191) from the forest plot (Figure 5B), which is consistent with the results we expected from our study. Additionally, we identified another potential key factor—Mitogen-Activated Protein Kinase Kinase Kinase 11 (MAP3K11). The SMR results for MAP3K11 were consistent across three tissues, suggesting that it may play an important role in the pathogenesis and progression of AMI.

For the ukb-e-I21_CSA outcome, forest plot visualization (Figure 5C) revealed a putative causal association between SP1 and AMI in whole blood (pSMR = 0.046, pHEIDI = 0.128, OR = 8.308, 95%CI = 1.040-66.342), corroborating our expected findings.

We focused on SP1 as the core gene and used two or more online tools to jointly predict the upstream and downstream targets of it. The Venn diagram in Figure 6A illustrates the intersection of miRNAs predicted from three independent databases: miRWalk, TargetScan, and starBase. This integrative analysis identified 132 candidate miRNAs that were consistently predicted across three platforms. Furthermore, as demonstrated in Figure 6B, we conducted an integrative prediction of mRNAs regulated by SP1, leveraging data from two separate datasets: KnockTF and TargetScan. We identified 358 candidate mRNAs that were consistently projected as targets of SP1 in both cohorts. Finally, we constructed the miRNA-TF-mRNA network as shown in the Figure 6C.

To refine our list of candidate genes, we initially conducted a differential expression analysis utilizing the GSE4648 dataset. Subsequently, we intersected the DE-mRNAs with ferroptosis-related genes as well as the 358 candidate genes. This intersection process ultimately yielded 7 candidate mRNAs (EGR1, IL6, MYC, NR4A1, P4HA1, PLAUR and VEGFA). In a parallel approach, employing the GSE76604 dataset for differential expression analysis, we similarly intersected the DE-miRNAs with ferroptosis-related miRNAs and the 132 candidate miRNAs. This strategy enabled us to pinpoint 2 miRNAs (miR-133b and miR-326) (Figure 6D). Given that miRNAs typically exert an inhibitory influence on their downstream target genes and SP1 was up-regulated in AMI samples, we ultimately select miR-133b as the upstream regulatory factor. Finally, we successfully established the miR-133b/SP1/PLAUR/ferroptosis signaling pathway (Figure 6E).

Following the previous steps, we conducted a weighted gene co-expression network analysis (WGCNA) using the GSE4648 dataset. Sample clustering revealed no outliers and supported reliable group separation (Figure 7A). A scale-free topology was obtained by choosing a soft-thresholding power of β = 10, which yielded a scale-free fit index R² > 0.8 while keeping the mean connectivity low (Figure 7B). After applying dynamic tree cut (MEDissThres = 0.25), closely related modules were merged, giving seven final modules (Figure 7C). A trait–module correlation heatmap showed that the brown module (MEbrown) had the strongest positive association with AMI (cor = 0.29, P < 0.05) (Figure 7D). This module was therefore designated as the AMI-related module and contained 803 genes for further analysis.

Subsequently, we intersected the 7 candidate mRNAs with the 803 key module genes to obtain 4 key genes (Figure 7E). Then, we performed a colocalization analysis of 4 candidate mRNAs with MI GWAS data from the Finnish database. Through this analysis, we identified that only PLAUR (plasminogen activator urokinase receptor) demonstrated strong colocalization evidence with acute myocardial infarction (PPH4 = 0.99). It was the same data as the outcome data in the SMR analysis, which added the reliability of our findings. The visual evidence for this finding was provided in the (Figure 7F).

Ultimately, through a series of rigorous and systematic analyses, including differential expression analysis, SMR, colocalization validation and regulatory relationship determination, we successfully established the miR-133b/SP1/PLAUR/ferroptosis signaling pathway. The construction of this pathway is based on robust evidence from various bioinformatics approaches, providing a comprehensive framework for understanding the molecular mechanisms of ferroptosis regulation in the context of acute myocardial infarction.

The regulatory mechanism of the miR-133b/SP1/PLAUR axis within cells under physiological conditions is shown in Figure 8A. Under physiological conditions, miR-133b constrains the transcription factor SP1, thereby silencing PLAUR transcription. However, this regulatory axis is disrupted during AMI. Based on this, our study first proposes the following mechanistic hypothesis: in the ischemic and hypoxic microenvironment of the myocardium, miR-133b is initially downregulated, relieving its inhibition on the transcription factor SP1. The enhanced activity of SP1 subsequently upregulates PLAUR expression. High expression of PLAUR mediates lipid peroxidation damage in cardiomyocytes by activating the ferroptosis pathway, ultimately exacerbating myocardial tissue necrosis and deterioration of cardiac function.

To validate the authenticity of the miR-133b/SP1/PLAUR axis, we first systematically reviewed the literature and confirmed that there are conserved binding sequences between miR-133b and the SP1 3’UTR, as well as between the SP1 promoter and the PLAUR promoter region (42–46). Subsequently, we used online tools for interaction prediction and the results were fully consistent with the literature, providing a reliable sequence basis for subsequent functional experiments. First, using the starBase database, the complementary binding sites between miR-133b and the SP1 were predicted, and the resulting binding sequence is shown in Figure 8B. Pearson correlation analysis was conducted using the GEPIA2 database (GTEx left ventricular tissue), and the results showed a significant positive correlation between SP1 and PLAUR expression levels (r = 0.45, P = 2.8e-11, Figure 8C), suggesting that SP1 may be involved in the transcriptional regulation of PLAUR. Subsequently, the SP1 position-weight matrix (MA0079.1) was downloaded from JASPAR (Figure 8D), and motif scanning was performed within 2000 bp upstream and 100 bp downstream of the PLAUR transcription start site (TSS) with a relative profile score threshold ≥ 0.8. A total of 9 potential binding sites were identified, among which the site at chr19: 43670069–43672169 had the highest score (0.85), suggesting that SP1 may directly bind to the PLAUR promoter region and up-regulate its expression (Figure 8E).

GeneMANIA (Figure 9A) pinpointed 20 genes tightly linked to SP1 and PLAUR, among which post-GPI attachment to proteins inositol deacylase 1 (PGAP1) exhibited prominent connectivity. Moreover, these nodes exhibited mutual interconnections, thereby giving rise to a highly intricate and complex network. Through GeneMANIA analysis, it was found that the studied genes are interconnected with a series of lipid - associated functions, including the metabolic and biosynthetic processes of GPI anchors, protein lipidation, lipoprotein biosynthesis and metabolism, along with glycolipid and phosphatidylinositol metabolism.

In this study, GSEA was employed to identify key signaling pathways potentially involving SP1 and PLAUR in AMI. Enrichment plots revealed bidirectional pathway activity contingent on SP1 abundance in AMI: high SP1 expression (Figure 9B) was concomitant with the significant activation of five immune- and stress-related signaling pathways (FC-gamma-R-mediated phagocytosis, insulin signaling, leukocyte transendothelial migration, neurotrophin signaling and Toll-like receptor signaling), whereas SP1 down-regulation (Figure 9C) preferentially engaged five metabolic or structural modules (asthma, DNA replication, oxidative phosphorylation, purine metabolism, and ribosome). GSEA disclosed a PLAUR-dependent switch in AMI-related pathway activity: elevated PLAUR (Figure 9D) selectively enriched pro-inflammatory and metabolic signaling circuits (adipocytokine, chemokine, FC-gamma-R-mediated phagocytosis, insulin and Toll-like receptor pathways), whereas PLAUR down-regulation (Figure 9E) shifted the transcriptional landscape toward immune-dysregulation and biosynthetic programs (allograft rejection, asthma, DNA replication, intestinal immune network for IgA production and ribosome).

In the CTD Database, candidate drugs and diseases associated with SP1 and PLAUR were retrieved independently. For each gene, compounds and disease entries were ranked by their CTD Inference Score, and the top 20 drugs and diseases were retained. The two ranked drug lists and the two ranked diseases lists were then intersected, yielding six small-molecule compounds (Figure 10A) and seventeen diseases (Figure 10B) that are jointly targeted to both SP1 and PLAUR, thus prioritizing dual-targeting candidates for further investigation. Through this strategy, we identified the potential dual-target intervention candidates, providing direction for subsequent AMI mechanism research and drug development.

Using the PheWAS analysis, we scanned for phenotypes linked to SP1 or PLAUR. No associations reached genome-wide significance (P < 5 × 10^-8) (Figures 10C, D), indicating that therapeutic targeting of SP1 or PLAUR is expected to carry minimal adverse drug reactions or unintended horizontal pleiotropy.

The receiver operating characteristic (ROC) curves were generated to evaluate the discriminatory power of the genes in distinguishing AMI cases from non - AMI individuals. The area under the curve (AUC) values were calculated to quantify the overall diagnostic accuracy. A higher AUC value indicates a better ability of the biomarker to correctly classify AMI patients and healthy controls. Through this analysis, the ROC curve (Figure 11A) showed that the AUC of miR-133b was 0.811 (95% CI = 0.701–0.922, P = 0.000). The ROC curve (Figures 11B, C) showed that the AUC of SP1 was 0.764 (95% CI = 0.574–0.955, P = 0.030) and the AUC of PLAUR was 0.893 (95% CI = 0.767–1.000, P = 0.001). In summary, the ROC curve analyses of miR-133b, SP1, and PLAUR using data from GSE76604 and GSE61144 datasets, respectively, offer valuable insights into their diagnostic potential as biomarkers for AMI.

Figures 11D–F displayed the expression levels of miR-133b, SP1 and PLAUR in both the GSE76604 and GSE61144 datasets. In AMI group, miR-133b was significantly downregulated, whereas SP1 and PLAUR were upregulated, consistent with the differential expression analysis and SMR findings. These results were further validated by pRT-PCR experiments in H9C2 cardiomyocytes subjected to hypoxia-induced injury, which showed the same expression trends (Figures 11G–I).