Gastric cancer-sarcopenia axis driven by tumor-derived GDF15, MIF and MDK signaling: evidence from Mendelian randomization, single-cell transcriptomics and survival analysis

Background Gastric cancer often brings sarcopenia and cancer cachexia. Population evidence, including Mendelian randomization, suggests a tumor-driven, unidirectional direction. The ligand networks behind muscle wasting stay undefined. Few instrumental variables and high heterogeneity keep this at exploratory-hypothesis level. Methods We built a three-layer framework: population causality, cellular mechanism, clinical prognosis. Goal: map the gastric cancer–sarcopenia axis. Mendelian randomization exposure: genetic predisposition to gastric cancer (bbj-a-119, East Asian). Mendelian randomization outcome: appendicular lean mass (GCST90000025). Random-effects model. Single-cell data: GSE163558, 42,968 cells, primary and metastatic gastric cancer. Survival set: TCGA STAD, GEO GSE62254, GEO GSE84437. Results Inverse-variance weighted odds ratio 0.968 (95% confidence interval 0.950–0.987, P = 9.5×10⁻⁴). Six instrumental variables. Gastric cancer predisposition causally lowered appendicular lean mass. Direction matches the clinical causal consensus. Reverse Mendelian randomization was non-significant. The link is unidirectional. Only six instrumental variables were available. Heterogeneity was present (Q P = 5.4×10⁻⁶). This Mendelian randomization evidence is supportive, not definitive. CellChat found 241 significant ligand–receptor pairs. 185 pairs ran from tumor epithelium to stroma. Dominant axes: GDF15→TGFBR2, MIF→CD74/CXCR4/CD44, MDK→SDC1/SDC2. All three axes reproduced in TISCH2 and hECA muscle. TGFB2, FAT3, PGF stayed directionally consistent after stage adjustment. All three had hazard ratio > 1. None appeared in single-cell communication. They are not axis-specific drivers. Conclusions The study proposes a tumor-derived GDF15/MIF/MDK ligand axis driving gastric cancer muscle wasting. The three-layer framework localizes this axis in gastric cancer single-cell architecture. It links population, cellular, and prognostic evidence. The findings give translatable clues for early warning and targeted intervention of sarcopenia.