Background: Colorectal cancer (CRC) progression is associated with tumor metabolic reprogramming and an immunosuppressive tumor microenvironment, yet coordinated metabolic interactions between malignant epithelial and immune cells remain unclear. This study aimed to characterize metabolic crosstalk in CRC, validate spatial organization, and develop a metabolism-based prognostic model. Methods: Six CRC single-cell RNA sequencing datasets were integrated to identify cell populations, evaluate metabolic pathway activity, and infer cell–cell communication. Spatial transcriptomics was used to assess regional co-enrichment of key cell-subset signatures and metabolic activities. Bulk transcriptomic cohorts and targeted metabolomics data were analyzed for pathway-level support. Patients were stratified using metabolic features of selected subsets, followed by protein–protein interaction analysis and elastic net modeling. Results: Across six scRNA-seq datasets comprising 431,217 cells from 173 samples (107 tumor, 60 normal, and 6 border), we identified a metabolically reprogrammed malignant epithelial subset (SLC6A20+ epithelial cells) and an immunosuppressive SPP1+ tumor-associated macrophage (TAM) subset. Both exhibited elevated glycolysis, vitamin B6 metabolism, and aromatic amino acid metabolism. Spatial transcriptomics supported regional co-enrichment of their signatures and shared metabolic activities within the same tumor regions. Independent bulk transcriptomic cohorts and targeted metabolomics further supported these pathway alterations. Cell–cell communication analysis revealed extensive bidirectional ligand-receptor interactions. Based on metabolic features of these subsets, patients were stratified into two prognostic groups. A 14-gene elastic net signature predicted the high-risk subtype with consistent performance across independent cohorts. Conclusions: SLC6A20+ epithelial cells and SPP1+ TAMs showed coordinated, transcriptome-inferred metabolic programs and predicted bidirectional communication in CRC. These features provide candidate biologically interpretable biomarkers and a metabolism-based prognostic model for patient stratification.
Xue et al. (Wed,) studied this question.