Unlocking Resveratrol's Potential: Targeting Ferroptosis in Atherosclerosis Through MAPK1

Abstract

Atherosclerosis (AS) is a chronic inflammatory metabolic disorder and a leading cause of cardiovascular diseases. Resveratrol (RSV), a natural polyphenolic phytoestrogen, exhibits anti-atherosclerotic effects by modulating oxidative stress and ferroptosis, yet its key therapeutic targets remain unclear. Using network pharmacology, bioinformatics, machine learning, and molecular docking, we identified core targets and mechanisms of RSV in ferroptosis and anti-atherosclerosis. Experimental validation was performed using ApoE^-/- mouse fed a high-fat diet (HFD) for 12 weeks to establish AS model. We assessed aortic and aortic root plaque formation, serum oxidative stress, and iron levels. By mining online databases, we identified 31 shared targets at the intersection of RSV-AS-Ferroptosis. A PPI network was generated using STRING, and GeneMANIA, GO and KEGG analyses revealed key biological processes and pathways (such as oxidative stress). Employing eight machine learning algorithms, we pinpointed six key targets: MAPK1, IL1B, RELA, HIF1A, SRC, and PTEN. Differential gene docking and molecular docking analyses showed that MAPK1 (-8.8 kcal/mol binding energy) had relatively good affinity. In vivo, RSV treatment reduced aortic lipid plaques, reduced serum GSSG/GSH, SOD, MDA, and iron levels, and significantly downregulated MAPK1 expression in the aortic root. RSV could modulate the ferroptosis pathway through targeting the MAPK1 gene, providing a new theoretical framework for AS prevention and treatment.

Keywords: MAPK1; atherosclerosis; ferroptosis; machine learning; molecular docking; resveratrol.

FIGURE 1

Graphical abstract.

FIGURE 2

Predicted targets of RSV in relation to AS. (A) AS‐related targets in various disease databases (OMIM, and GeneCards). (B) Drug‐related databases (ETCM, TCMSP, PharmMapper and SwissTarget Prediction) explored the targets of RSV. Volcano (C) and heat map (D) of differentially expressed genes in GEO samples. The results of PCA analysis of the GEO dataset remove batch effects before (E) and after (F) rectification. WGCNA R package was used to construct a co‐expression network (G–I) in GEO samples. When the soft threshold was set to 3, the scale‐free topological fitting index (R2) was 0.85. A heat map of the relationship between the control and the AS module, with the modules represented in different colors. Red indicates a positive correlation, while blue indicates a negative correlation. (J) AS‐related genes obtained from the intersection of WGCNA and DEG in the GEO dataset (K) AS‐related genes obtained from the intersection of AS and RSV.

FIGURE 3

RSV improves atherosclerosis by reducing oxidative stress and iron metabolism. (A) Resveratrol molecular formula; (B) Experimental design; (C) Mouse aortic arch [scale bar = 1 mm]; (D) Oil red o staining the surface aorta of mice [scale bar = 2.5 mm]; (E) HE staining of aortic roots [scale bar = 250 μm]; Serum GSSG (F) and GSSG/GSH content (G) in mice; SOD inhibition rate (H) and enzyme activity (I); MDA content (J) and serum iron (K), n = 5. The data were expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001.

FIGURE 4

Predicted targets and mechanisms of RSV's anti‐AS effect by regulating ferroptosis. (A, B) KEGG enrichment pathway analysis of shared genes of RSV and AS; (C) Common genes in Venn diagram of RSV, AS, and ferroptosis genes; (D, E) KEGG enrichment pathway analysis of 31 related genes enrichment pathway analysis; (F) analysis of functional associations of targets using GeneMANIA; (G–I) GO enrichment map of 31 related genes, including BP, CC, and MF; (J) PPI network based on STRING database; (K) Cytoscape visualized PPI network of 31 genes; (L) Cytohubba plug‐in screened the Cytoscape visualized PPI network of the top 20 genes.

FIGURE 5

Machine learning screens six core targets. (A) Eight machine learning importance distribution plots of 20 hub FRGs calculated by Cytohubba, (B, D) residual plots of eight machine learning results, (C) ROC curves of eight machine learning methods. (E) LASSO regression of 20 hub FRGs calculated with Cytohubba, (F) cross‐validated LASSO regression parameter selection, (G, H) in RF analysis, random forest plots as well as significant gene visualization (score > 1.35). (I) SVM‐RFE algorithm was used to obtain FRGs with minimum error and highest accuracy, which were considered to be the most suitable candidate biomarkers. (J) Comparison between Venn diagram LASSO, RF, NNET, and SVM‐RFE algorithms showing shared/unique genes.

FIGURE 6

Expression of six core targets in differential gene analysis. (A–F) The expression of MAPK1, RELA, HIF1A, IL1B, SRC, and PTEN genes in the test set was analyzed by using R language DEGs. (G–L) DEGs analysis was used to verify the gene expression of MAPK1, RELA, HIF1A, IL1B, SRC, and PTEN in the AS group compared with the control group. Data are presented as mean ± SEM. *p < 0.05, **p < 0.01,***p < 0.001.

FIGURE 7

RSV against AS by reducing MAPK1 expression. Molecular docking of RSV with MAPK1 (A), SRC (B), HIF1A (C), IL1B (D), RELA (E), and PTEN (F). Immunohistochemical staining [scale bar = 250 μm] (G) and quantification (n = 3) (H) of MAPK1 protein expression level in mouse aortic roots. Data are presented as mean ± SEM. **p < 0.01.