Transcriptome-based classification to predict FOLFIRINOX response in a real-world metastatic pancreatic cancer cohort

Abstract

Pancreatic ductal adenocarcinoma (PDAC) is often diagnosed at metastatic stage and typically treated with fluorouracil, leucovorin, irinotecan and oxaliplatin (FOLFIRINOX). Few patients benefit from this treatment. Molecular subtypes are prognostic in particularly resectable PDAC and might predict treatment response. This study aims to correlate molecular subtypes in metastatic PDAC with FOLFIRINOX responses using real-world data, providing assistance in counselling patients. We collected 131 RNA-sequenced metastatic biopsies and applied a network-based meta-analysis using published PDAC classifiers. Subsequent survival analysis was performed using the most suitable classifier. For validation, we developed an immunohistochemistry (IHC) classifier using GATA6 and keratin-17 (KRT17), and applied it to 86 formalin-fixed paraffin-embedded samples of advanced PDAC. Lastly, GATA6 knockdown models were generated in PDAC organoids and cell lines. We showed that the PurIST classifier was the most suitable classifier. With this classifier, classical tumors had longer PFS and OS than basal-like tumors (PFS: 216 vs. 78 days, p = 0.0002; OS: 251 vs. 195 days, p = 0.049). The validation cohort showed a similar trend. Importantly, IHC GATA6^low patients had significantly shorter survival with FOLFIRINOX (323 vs. 746 days, p = 0.006), but no difference in non-treated patients (61 vs. 54 days, p = 0.925). This suggests that GATA6 H-score predicts therapy response. GATA6 knockdown models did not lead to increased FOLFIRINOX responsiveness. These data suggest a predictive role for subtyping (transcriptomic and GATA6 IHC), though no direct causal relationship was found between GATA6 expression and chemoresistance. GATA6 immunohistochemistry should be seamlessly added to current diagnostics and integrated into upcoming clinical trials.