Prediction of Gemcitabine sensitivity in resectable Pancreatic Cancer using a Glycation Stress Transcriptomic Signature
BACKGROUND & AIMSPancreatic ductal adenocarcinoma (PDAC) is a highly lethal cancer with limited response to systemic therapy. Gemcitabine (GEM) benefits only a subset of patients. Methylglyoxal (MG), a glycolysis byproduct, has been linked to tumor behavior and therapy response in PDAC, suggesting potential as a stratification marker. METHODSWe developed a metabolically informed gene signature (MG-GEM) integrating MG- related glycolytic stress with clinical outcomes. Using the Puleo cohort (n=309), differential expression analysis between tumors with high and low MG stress identified 365 genes. LASSO Cox regression selected 16 prognostic genes, combined into a weighted risk score for patient stratification. MG GEM was validated in internal and external cohorts, including PRODIGE 24/CCTG PA6. Molecular, transcriptomic, and immune features were compared between high and low MG GEM groups. Finally, predictive performance was evaluated against the GemPred signature. RESULTSMG GEM divided PDAC patients into distinct risk groups with marked differences in overall and disease-free survival among GEM treated patients (OS 11.7 vs 27.2 months; DFS 7.6 vs 17.8 months, both p<0.0001). High MG-GEM tumors showed enrichment for KRAS G12D and SMAD4 mutations, basal and activated stroma subtypes, glycolytic metabolism, and reduced immune infiltration. Low MG-GEM tumors showed KRAS G12V, classical and immune subtypes, cholesterogenic metabolism, and adaptive immune favourable signatures. MG-GEM independently predicted GEM-specific clinical outcomes, irrespective of GemPred signature, and significantly enhanced patient stratification when combined with it. Within the PRODIGE-24/TGCC PA6 cohort, MG-GEM exhibited prognostic relevance and selectively identified patients who derived a survival benefit from adjuvant GEM, but not from FOLFIRINOX. CONCLUSIONSThe 16 gene MG GEM signature predicts prognosis in resected PDAC, reflects glycolytic stress driven chemoresistance, surpassing conventional molecular classifications. As a metabolically informed signature, MG-GEM holds promise for guiding chemotherapy selection and informing KRAS-targeted combination strategies, meriting further prospective clinical validation.