BACKGROUND: Prostate cancer is one of the most common malignancies in men and exhibits substantial clinical heterogeneity. A considerable proportion of patients experience biochemical recurrence after primary treatment, and metabolic reprogramming has emerged as an important driver of tumor progression and tumor microenvironment remodeling. METHODS: Metabolism-related genes (MRGs) were systematically analyzed using TCGA as the training cohort, with DKFZ and three independent GEO microarray datasets (GSE46602, GSE70768, and GSE70769) used for external validation. Expression matrices were preprocessed in a platform-specific manner, normalized, and batch-corrected using ComBat. BCRFS-associated MRGs were identified through differential expression analysis (tumor vs. normal) and univariate Cox regression. Model construction employed 10-fold cross-validation across 101 machine-learning algorithm combinations. Immune characteristics were assessed by correlating the risk score with activities of the cancer-immunity cycle steps. Single-cell RNA-seq data (GSE264573) were used to evaluate hexosamine biosynthesis (HB) pathway activity and cell-cell communication via CellChat. Functional validation was performed by shRNA-mediated TK1 knockdown in PC-3 and DU-145 prostate cancer cell lines followed by colony formation assays. RESULTS: Twenty-four BCRFS-associated MRGs were identified. The StepCoxbackward + RSF algorithm combination achieved the highest average concordance index and was selected for the final metabolism-related prognostic model (MRPM). The derived risk score robustly stratified patients into high- and low-risk groups across all five cohorts, with high-risk patients showing significantly shorter biochemical recurrence-free survival (BCRFS). Model performance was further supported by time-dependent ROC analysis and PCA-based separation of the risk groups. The risk score was positively associated with multiple malignant biological programs, including cell cycle, DNA replication, and DNA repair pathways, suggesting enhanced proliferative and repair activity in high-risk tumors. In contrast, its associations with the cancer-immunity cycle were heterogeneous, indicating complex immune remodeling rather than a uniformly activated antitumor response. Single-cell analysis demonstrated broad activation of the hexosamine biosynthesis program across multiple cell types, including epithelial, fibroblast, basal, endothelial, and immune cells, together with markedly increased intercellular communication strength and complexity in the HB-high state, particularly among epithelial, fibroblast, and basal populations. Among the five signature genes (TK1, SUV39H1, NME3, MCCC1, and DEGS1), TK1 showed the strongest positive correlation with the risk score and was associated with shorter BCRFS; its knockdown significantly suppressed the long-term proliferative capacity of prostate cancer cells. CONCLUSIONS: The MRPM provides a robust, metabolism-informed tool for BCRFS risk stratification in prostate cancer. It highlights a close link between metabolic reprogramming, particularly TK1-related nucleotide metabolism and hexosamine biosynthesis, and multicellular communication within the tumor microenvironment, providing insight into metabolism-associated mechanisms in prostate cancer.
Zhu et al. (Mon,) studied this question.