• Citer

Identification, characterization and predictive validation of a methylglyoxal signature in resectable pancreatic ductal adenocarcinoma treated with adjuvant gemcitabine.

par Azurmendi Senar, Oier
Président du jury Deplus, Rachel
Promoteur Van Laethem, Jean-Luc
Co-Promoteur Detours, Vincent
Publication Non publié, 2026-03-03

Thèse de doctorat

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

-Pancreatic ductal adenocarcinoma (PDAC) remains one of the most lethal human malignancies, with a five-year survival rate below 12% and limited therapeutic progress over recent decades. Gemcitabine (GEM) continues to be a cornerstone of systemic treatment, yet only a subset of patients derives meaningful clinical benefit. Increasing evidence has highlighted the role of metabolic reprogramming and particularly enhanced glycolysis and the accumulation of the reactive metabolite methylglyoxal (MG) in promoting chemoresistance and tumor progression. Building on these insights, this study aimed to develop and validate a prognostic and predictive gene signature integrating MG-stress related biology with clinical outcomes in patients with resected PDAC treated with GEM. Three independent patient cohorts were analyzed (Puleo et al., n=309; Erasme University Hospital, n=60; and the PRODIGE24/CCTG PA6 phase III trial, n=350). Differential expression analysis between tumors with high and low MG stress identified 365 genes, from which a Least Absolute Shrinkage and Selection Operator (LASSO) Cox regression model was applied to derive a 16-gene signature termed MG-GEM. Using a weighted gene expression-based risk score, patients were classified into high- and low-risk groups, revealing strong stratification for both overall survival (OS) and disease-free survival (DFS). High MG-GEM tumors exhibited markedly shorter OS and DFS, independently of clinicopathological variables (HR=3.5, p<0.001), and were associated with poor-prognosis transcriptomic subtypes such as “Basal-like” and “Stromaactivated.” These tumors also displayed enrichment of KRAS^G12D and SMAD4 mutations, increased glycolytic activity, a fibroblast-rich and immunosuppressive microenvironment, and upregulation of metabolic resistance mediators including CDA and RRM1. Conversely, low MG-GEM tumors corresponded to “Classical” and “Immune” subtypes, characterized by stronger adaptive immune infiltration, higher expression of the GEM transporter CNT1, and molecular profiles consistent with chemosensitivity. Validation across independent internal and external cohorts confirmed that 6 MG-GEM predicted survival exclusively in GEM-treated patients, but not in those receiving FFX, indicating its treatment-specific predictive value. Comparison with the established GemPred classifier demonstrated that MGGEM provides independent and complementary prognostic information, improving the identification of both GEM-sensitive and GEM-resistant patient subsets. Furthermore, molecular correlation analyses linked MG-GEM stratification to distinct metabolic programs, with high MG-GEM tumors displaying enrichment in glycolytic, hypoxia, and extracellular matrix pathways, while low MG-GEM tumors favoured lipid metabolism and immune activation signatures. Altogether, these findings position MG-GEM as a metabolically informed biomarker that captures the interplay between glycolytic stress, KRAS-driven metabolic rewiring, and chemoresistance in PDAC. By providing an independent and concise 16-gene signature with validated prognostic and predictive value, MG-GEM refines patient stratification beyond existing molecular classifiers and offers a potential clinical tool for guiding adjuvant therapy selection. This work underscores the relevance of metabolic stress in shaping GEM response and opens avenues for personalized treatment strategies targeting metabolic vulnerabilities and KRAS-associated pathways in pancreatic cancer.