Molecular mechanisms of ferroptosis in ulcerative colitis: insights from machine learning, WGCNA, and immune cell infiltration analysis.

BACKGROUND: This study aimed to investigate ferroptosis-related biomarkers and their potential molecular basis in UC. METHODS: UC datasets (GSE87466 and GSE47908) from the Gene Expression Omnibus database were merged as the training set, and batch effects were removed. Ferroptosis-related differentially expressed genes (DE-FRGs) were selected to construct a diagnostic risk model in UC. Machine learning (lasso regression and SVM-RFE), Weighted Gene Co-expression Network Analysis (WGCNA) and PPI were then used to obtain candidate hub genes. After identifying common DE-FRGs, functional enrichment analysis, GSEA and GSVA functional enrichment analysis and immune cell infiltration were performed to explore the pathogenesis of UC. Besides, the correlation of hub gene expression and ferroptosis signature markers (GPX4 and ACSL4) was validated in external validation (GSE92415) and in vitro experiments. Finally, we employed the human intestinal epithelial Caco-2 cell to establish an in vitro inflammatory model by treatment with LPS (1 μg/ml) for 24 hours. This model was used to validate the correlation between the expression levels of ferroptosis-related essential genes (ACSL4 and GPX4) and pro-inflammatory cytokines (TNF-α, IL-6, and IL-1β). Furthermore, to confirm ferroptosis involvement, Caco-2 cells were co-treated with RSL3 (a ferroptosis inducer) or Ferrostatin-1 (Fer-1, an inhibitor), followed by measurement of GSH, MDA as an indicator of lipid peroxidation, and cellular iron load. Mitochondrial ultrastructure was assessed via transmission electron microscopy (TEM) to detect ferroptosis-associated morphological changes. RESULTS: MFN2 and CBS were identified as hub genes after further validation. Functional estimation, gene set enrichment analysis, and immune infiltration signature identification showed notable associations of the hub genes with macrophages, mast cells resting, and follicular helper T cell levels. In vitro, we observed that treatment with LPS/RSL3 obviously activated ferroptosis in Caco-2 cells, as indicated by altered expression of key ferroptosis-related genes (down-regulation of GPX4, CBS, and MFN2; up-regulation of ACSL4) and the levels of surrogate ferroptosis markers (elevated MDA and iron levels, along with reduced GSH). In addition, LPS-induced ferroptosis in Caco-2 cells could be reversed by Fer-1. CONCLUSIONS: MFN2 and CBS may represent potential therapeutic targets and could serve as biomarkers for immune regulation in UC, warranting further investigation.