Integrated genomic and transcriptomic analysis to predict FOLFIRINOX response in metastatic pancreatic ductal adenocarcinoma.

4153 Background: Given the lack of benefit of first-line chemotherapy (FOLFIRINOX) for a large proportion of metastatic pancreatic ductal adenocarcinoma (mPDAC) patients, predictive markers for treatment outcome are urgently needed. Previous studies showed that transcriptome-based classifiers and homologous recombination deficiency (HRD) associate with outcome in this patient group. Here, we aimed to validate the benefit of RNA and whole genome sequencing (WGS) in predicting treatment outcome in a Dutch mPDAC FOLFIRINOX-treated cohort. Methods: Patients with mPDAC who intended to start FOLFIRINOX were included in the CPCT-02 cohort. 108 patients started FOLFIRINOX and had successful WGS and RNA sequencing of their pretreatment biopsies. 51 mPDAC samples of non-FOLFIRINOX treated patients were also included in the cohort in order to explore the relation between transcriptome-based subtype and somatic mutational status with more power. Transcriptomic subtypes were established according to literature. HRD labels were derived from the previously published CHORD algorithm. Results: 88 percent of the included samples had a KRAS mutation and 82 percent had mutated TP53. The KRAS wild-type patients showed MYC amplifications, gene fusions in NTRK1, NTRK3, BRAF, NRG1, ALK, FGFR2 and RAF1, and mutations in mismatch repair proteins. Patients with a KRAS wild-type tumor had a better survival following FOLFIRINOX treatment (p = 0.009) compared to KRAS mutated tumors. No other mutations were found to be prognostic. As described before, transcriptome-based classification according to Moffitt revealed a cluster of tumors with a “Moffitt-negative” subtype, showing low expression of classical and basal-like signature genes. We found that KRAS wild-type tumors were enriched for this Moffitt negative subtype (p < 0.0001), while they did not have lower tumor cellularity. The gene expression of these Moffitt negative tumors correlated with neural-like and neuroendocrine gene sets. In contrast to previous studies, we did not observe a better response to FOLFIRINOX for patients with an HRD-tumor (OS: p = 0.21, PFS: p = 0.054). We noted an increased expression for the alternative non-homologous end joining pathways in HRD samples (p = 0.049) that might have countered the advantageous effect of HRD. Conclusions: We found that KRAS wild-type tumors do not fit in the classical – basal-like classification schemes. We could not prove a significant prognostic effect for HRD. Both KRAS status and transcriptome-based subgroup were prognostic for FOLFIRINOX treatment response and should therefore be further validated for incorporation in diagnostic practice.