Mechanisms of response and resistance to KRAS inhibition in pancreatic cancer

Abstract

Pancreatic ductal adenocarcinoma (PDAC) is a highly lethal and treatment-refractory malignancy. Despite its clinical heterogeneity and complexity, PDAC displays a strikingly uniform mutational landscape dominated by alterations in the tumor suppressor genes TP53, CDKN2A, SMAD4, and by oncogenic mutations in KRAS, found in ~90% of patients, where it acts as the primary oncogenic driver. Mutations in KRAS have been associated with a more aggressive disease and poorer outcomes, yet the protein had remained “undruggable” for over three decades. Recent advances in medicinal chemistry have enabled the development of direct inhibitors of KRAS, which have demonstrated promising efficacy across cancer types; however, resistance emerges in most patients, highlighting the need for an improved understanding of the determinants of response and resistance. Here, we systematically characterize mechanisms of response and resistance to oncogenic KRAS inhibition in PDAC across preclinical models and patient specimens. Circulating tumor DNA profiling of patients with PDAC treated with the KRASG12C inhibitors sotorasib and adagrasib identified mutations in PIK3CA, and amplifications of KRASG12C, MYC, MET, EGFR, and CDK6 emerging at acquired resistance, captured in approximately half of evaluable patients. In preclinical in vitro models of PDAC, epithelial-to-mesenchymal transition and PI3K-AKT-mTOR signaling were associated with decreased response to the KRASG12D inhibitor MRTX1133. Treatment of the KrasLSL-G12D/+; Trp53LSL-R172H/+; p48-Cre (KPC), and KrasLSL-G12D/+; p48-Cre (KC) genetically-engineered mouse models of PDAC with MRTX1133 drove deep tumor regressions in most animals, followed by acquired resistance marked by amplification of Kras, Yap1, Myc, and Cdk6, alongside co-evolution of drug-resistance transcriptional programs. While resistance was marked by strong genetic and transcriptional heterogeneity, we captured a recurrent polarization of PDAC cells into a well-differentiated, epithelial classical cell state after short-term treatment through maximal response and stable disease. Probabilistic inference of temporal cell state transitions during this period revealed a high sensitivity of mesenchymal and partial epithelial-to-mesenchymal transition cell states, which transitioned towards the classical cell state, demonstrating increased fitness early during treatment through stable disease. Finally, leveraging single-cell lineage tracing with joint CRISPR-based gene perturbation, we functionally interrogated the effects of perturbing epithelial- and mesenchymal-associated transcription factors on cell state dynamics upon pan-RAS inhibition, demonstrating the ability to prevent treatment-driven cell state plasticity and "lock" cells into a responsive mesenchymal state. Together, this work defines both genetic and nongenetic mechanisms of KRAS inhibitor resistance and identifies cell state dynamics that shape therapeutic response, providing a framework for developing more effective combination strategies in PDAC. Finally, leveraging single-cell lineage tracing with joint CRISPR-based gene perturbation, we functionally characterized the effects of classical- and mesenchymal-associated transcription factors on cell state dynamics upon RAS-GTP inhibition, demonstrating the ability to prevent treatment-driven cell state plasticity and "lock" cells into the responsive mesenchymal state. Together, this work defines both genetic and nongenetic mechanisms of KRAS inhibitor resistance and identifies cell state dynamics that shape therapeutic response, providing a framework for developing more effective combination strategies in PDAC.