Ferroptosis and nutrient-metabolic pathway genes in ovarian cancer: integrated bulk and single-cell analysis of Iron-metabolic programs, immune-marker enrichment, and prognosis

Abstract

Ovarian cancer (OC), particularly high-grade serous ovarian carcinoma (HGSOC), remains a lethal gynecological malignancy characterized by frequent recurrence and treatment resistance. Ferroptosis is governed by nutrient-related metabolic processes involving iron handling, cystine–glutathione antioxidant defense, lipid peroxidation, amino-acid metabolism, and cellular energy metabolism. However, the organization of these programs across bulk tumors and individual tumor-microenvironment compartments remains incompletely understood. Bulk RNA-sequencing data from The Cancer Genome Atlas ovarian cancer cohort (TCGA-OV) were analyzed in 421 primary tumors, with 419 patients evaluable for overall survival, using a curated 40-gene ferroptosis–nutrient-metabolic panel. Gene correlations, module scores, survival associations, a 10-fold cross-validated LASSO-Cox ferroptosis risk score (FRS), immune-marker enrichment, and immune-checkpoint correlations were evaluated. Single-cell RNA-sequencing data from GSE146026 were analyzed in a 10x cohort of 9,609 cells from 6 patients, with a Smart-seq2 cohort of 1,297 cells used for consistency analyses. Selected genes were further validated by RT-qPCR in hTERT FNE1, NIH:OVCAR-3, and Caov-3 cells. FTL, RPL8, FTH1, LDHA, and GPX4 showed the highest mean expression in TCGA-OV. Lower SLC7A11 expression was associated with shorter survival (HR = 0.83 per SD, 95% CI 0.73–0.94, p = 0.004), whereas higher LPCAT3 expression was associated with shorter survival (HR = 1.19, p = 0.010), although neither survived false-discovery-rate correction. The two-gene FRS separated survival groups (log-rank p = 0.019) but showed modest discrimination (C-index = 0.56; held-out concordance = 0.55). The ferroptosis–nutrition score correlated most strongly with macrophage marker enrichment (rho = 0.46). At single-cell resolution, the overall score was highest in fibroblasts, epithelial tumor cells, and macrophages, whereas the iron-storage module was highest in macrophages. Cross-platform concordance was observed between 10x and Smart-seq2 (rho = 0.63) and between epithelial pseudobulk and TCGA-OV bulk expression (rho = 0.65). RT-qPCR confirmed higher expression of FTL, FTH1, GPX4, SLC7A11, and LPCAT3 in both ovarian-cancer cell lines than in hTERT FNE1. Ferroptosis-related nutrient-metabolic genes display coordinated but cell-type-specific organization in ovarian cancer and are linked to immune-marker enrichment. RT-qPCR provides orthogonal expression-level validation of selected candidates. The FRS remains exploratory, and the transcriptomic scores should not be interpreted as direct measures of ferroptotic activity or clinical nutritional status.