Although metabolic reprogramming is a defining feature of cancer, the ways in which metabolites mediate communication between cells remain largely unexplored. To bridge this knowledge gap, we investigated metabolite-driven intercellular interactions using single-cell RNA sequencing (scRNA-seq) data to delineate the metabolic architecture of the tumor microenvironment (TME) in lung adenocarcinoma (LUAD) and uncover novel metabolite signaling pathways. Dimensionality reduction of the scRNA-seq dataset was performed with the Seurat package, and cell types were manually annotated using canonical markers from Cell Marker 2.0 and prior studies. Metabolite abundance and potential cell–cell communication events were inferred with MEBOCOST. In parallel, immune infiltration levels and tumor activity scores in the TCGA-LUAD cohort were evaluated using ESTIMATE and ssGSEA algorithms, and survival analyses were conducted for genes implicated in identified signaling axes. Key findings were validated across two independent GEO datasets. The expression of PTGDS and PTGDR was further verified using RT-qPCR and fluorescence in situ hybridization. Five metabolites—prostaglandin D2 (PGD2), D-Mannose, Choline, L-Cysteine, and Cholesterol—were identified as prominent components of the LUAD TME. PGD2 was a central signaling metabolite, predominantly produced by fibroblasts and plasmacytoid dendritic cells via PTGDS, and by mast cells through HPGDS. PGD2 signaling primarily targeted NK/T cells expressing the receptor PTGDR, and endothelial cells mediated its transport via SLCO2A1. CX3CR1+ NK/T cells participated in autocrine PGD2 signaling, whereas KLRC2+ NK cells, DNAJB1+ NK cells, and CD8+ MAIT cells were engaged in paracrine interactions. PGD2 may additionally facilitate lactate export through SLCO2A1 on endothelial cells. Clinically, this signaling axis correlated with infiltration of cytotoxic immune cells such as DNAJB1+ NK cells and was associated with favorable prognosis. A risk model derived from this axis predicted patient response to immunotherapy across both immunologically “hot” and “cold” tumors, suggesting potential therapeutic options for low-risk patients. RT-qPCR and immunofluorescence further confirmed decreased expression of PTGDS and PTGDR in LUAD tumors compared with normal tissue. Overall, our findings highlight PGD2 and its associated signaling network as critical modulators of tumor-suppressive and anti-inflammatory processes in LUAD, providing promising avenues for prognosis assessment and therapeutic intervention.
