Profiling treatment-associated evolutionary dynamics in advanced urothelial cancer via deep whole exome sequencing of cell-free DNA.

Abstract

4572 Background: Monitoring treatment response and identifying resistance mechanisms remain critical challenges in managing advanced bladder cancer. We investigated whether circulating tumor DNA (ctDNA) dynamics and deep whole exome sequencing (D-WES) could provide insights into treatment response and resistance patterns. Methods: We prospectively collected blood samples from 70 patients with advanced bladder cancer during treatment. Ultra-low pass whole genome sequencing (ULP-WGS) was performed on 213 plasma samples to quantify tumor fraction, with subsequent D-WES (200X) performed on 47 high-purity (i.e., at least 3% tumor fraction by ichorCNA) samples from 25 patients. Longitudinal analysis of ctDNA levels was correlated with clinical response. Phylogenetic reconstruction for samples in the D-WES cohort was used to identify candidate drivers of therapeutic response and resistance. Results: Changes in ctDNA tumor fraction correlated significantly with clinical trajectories: progression (median +1.3%), stable disease (-0.4%, p = 0.04 vs progression), and regression (-1.2%, p = 0.002 vs progression). Phylogenetic analysis revealed dynamic subclonal competition during treatment response. Mutation burden was significantly elevated in subclones associated with PD-(L)1 response (p = 0.04), including within a single patient. A number of novel candidate drivers of therapeutic response were identified across different therapeutic classes, with five genes meeting FDR-adjusted significance (q < 0.05): NLGN2, SHANK1 (chemotherapy resistance), GPR56 (taxane resistance), FOSL1, and ORC1L (chemotherapy sensitivity). Conclusions: Analysis of ctDNA can effectively monitor treatment response in advanced bladder cancer. D-WES of longitudinal samples revealed novel candidate drivers of therapeutic response and resistance, suggesting distinct molecular mechanisms may govern primary oncogenesis versus treatment response. These findings warrant further investigation in larger cohorts to validate their potential as predictive biomarkers.