Ketone body metabolism is associated with macrophage heterogeneity and a protumor phenotype in gastric cancer through single cell and bulk transcriptomic analysis

Abstract Background The tumor microenvironment (TME) in gastric cancer (GC) is characterized by significant heterogeneity, particularly among tumor-associated macrophages (TAMs). While metabolic reprogramming is a hallmark of cancer, the role of ketone body metabolism in shaping macrophage heterogeneity and its functional impact in GC remains largely unexplored. Methods We analyzed the single-cell RNA-seq dataset GSE206785 (24 tumor and 24 paired normal tissues from treatment-naïve GC patients). After quality control, 110,187 cells were annotated using Seurat. Myeloid subsets were subclustered, and their functional states were assessed via GSVA, Monocle2 pseudotime trajectory analysis, and CellChat intercellular communication. The FOXM1_Mac subset was independently validated using an additional scRNA-seq cohort (GSE150290). Validation was performed using TCGA-STAD bulk RNA-seq data (n = 407). A ketone body metabolism-related prognostic signature was constructed using univariate Cox regression, LASSO Cox regression, and multivariate Cox regression, and was externally validated in the GSE66229 (ACRG) and GSE84437 cohorts. Multiplex immunofluorescence staining of 12 GC tissue samples was performed to validate the FOXM1_Mac subset and its metabolic association. Results Nine distinct macrophage/monocyte subsets were identified. The FOXM1_Mac showed the highest ketone body metabolism activity, with elevated expression of OXCT1, ACAT1, ACAT2, and HMGCL. Pseudotime analysis showed gradual upregulation of these enzymes along differentiation, peaking in the terminal FOXM1_Mac branch. Heatmap and spatial expression analyses revealed that inflammation/stress genes (e.g., FOS, JUN) were enriched in transitional segments, whereas ketone body metabolism genes were highly expressed in a separate terminal cluster. Cell–cell communication uncovered enhanced tumor-specific signaling (e.g., MIF-CD74/CXCR4, GAS6-AXL) targeting FOXM1_Mac and other subsets. These findings were reproducible in an independent scRNA-seq cohort (GSE150290). In TCGA-STAD, ketone body response pathways were significantly upregulated in tumors. Immune infiltration analysis indicated increased FOXM1_Mac abundance in tumor tissues. A prognostic model based on ketone body-related genes derived from LASSO and multivariate Cox regression effectively stratified patients into high- and low-risk groups with significantly different overall survival. The model achieved a 3-year AUC of 0.693 in the training cohort and demonstrated cross-platform and cross-population generalizability in two external validation cohorts (GSE66229 and GSE84437), suggesting potential clinical relevance. Immunofluorescence confirmed co-localization of FOXM1 and OXCT1 in CD11b⁺ myeloid cells, and a strong positive correlation between FOXM1 and OXCT1 expression was observed in 12 GC tissue samples. Conclusion Our study identifies the heterogeneity of gastric cancer macrophages and suggests that ketone body metabolism is associated with the FOXM1_Mac, which exhibits a terminal, pro-tumor-like phenotype. The prognostic model indicates the potential clinical relevance of ketone body metabolism, providing correlative insights into GC pathogenesis and potential therapeutic targets. Experimental immunofluorescence data further support the association between FOXM1 expression and ketone body metabolism in GC CD11b⁺ myeloid cells (including macrophages).