Abstract B067: Dissection of the in vitro response landscape in PDAC under KRAS inhibition reveals co-occurring conserved and adaptive transcription programs

Abstract

RAS-targeted therapies are expected to become a cornerstone of systemic treatment in pancreatic ductal adenocarcinoma (PDAC). Despite exceptional clinical responses, cancer progression inevitably occurs, highlighting the need for a better understanding of mechanisms of response and resistance. In this study, we used in vitro perturbation to decipher the early transcriptional response programs in patient-derived 2D and 3D cell models of PDAC to KRAS inhibition. To this end, 3D patient-derived organoids (PDO) and 2D patient-derived cell lines (PDC) were treated with one of two (K)RAS inhibitors (daraxonrasib, zoldonrasib) at three dose levels (DMSO, 1x EC50 or 10x EC50). Cells were then harvested at three time points (2, 24 or 72 hours). KRAS mutant fraction and CNV were quantified from DNA at baseline, while RNA sequencing and Jess-based protein quantification (KRAS, phospho-ERK1/2, DUSP6) were carried out for all samples. Four patient-matched PDO/PDC pairs served as the discovery cohort, while an additional five PDO and two PDC served as the internal validation cohort. Across modalities, dose, time and culture effects revealed a complex response landscape. Employing linear modelling, we identified conserved, adaptive and context-sensitive transcriptional programs. Accounting for culture effects while preserving dose- and time-dependent responses, we derived an 18-gene transcriptional core signature that, benchmarked to existing signatures, enabled time-robust, culture-independent KRAS activity scoring while preserving dose sensitivity. In addition, model integration with phospho-ERK1/2 protein expression showed a robust coupling between phospho-ERK1/2 protein levels and our signature-derived KRAS activity score across (K)RAS-targeting compounds. Notably, baseline KRAS activity scores moderately correlated with in vitro sensitivity to daraxonrasib across models (Spearman’s Rho=0.6, p=0.02). Further, the KRAS score was validated in the internal validation cohort and generalized to external data under KRAS inhibition. Beyond this conserved core program, KRAS inhibition elicits a highly dynamic transcriptional remodeling. Whereas MYC/E2F proliferation programs were uniformly suppressed, other canonical Hallmark pathways - including glycolysis, cholesterol homeostasis, the unfolded protein response, and the mitotic spindle - resolved into distinct transcriptional modules with divergent temporal trajectories. Together, these data establish a systems-level framework for resolving conserved and adaptive transcriptional responses to KRAS inhibition and identify a robust RNA-based signature with potential utility for pharmacodynamic monitoring and the study of adaptive resistance.Artificial intelligence tools were used to assist with language refinement and editing of the abstract text; the authors take full responsibility for the content.