转移性去势抵抗性前列腺癌（mCRPC）中的血浆游离 DNA 核小体足迹

Background: Transcription factor (TF) activity can be determined by nucleosome footprints in low-pass whole genome sequencing (lpWGS) of plasma cell-free DNA (cfDNA). cfDNA wrapped around nucleosomes is protected from enzymatic digestion, and TFs induce phased nucleosome positioning around their binding sites, which results in oscillatory sequencing coverage and preferential depletion at open chromatin regions. Herein, we quantify nucleosome occupancy at binding sites for hundreds of TFs, evaluating the clinical utility of these functional readouts in plasma taken from subjects treated on the CARD prospective randomized trial of cabazitaxel (CAB) vs second androgen receptor pathway inhibitor (ARPI). Methods: Plasma cfDNA lpWGS (median coverage ~1.6x) was used to infer nucleosome footprints at TF binding sites (TFBS) of 682 TFs. Overall, 217 CARD trial subjects comprised a Test cohort and 174 patients receiving a taxane on the FIRSTANA and PROSELICA prospective trials comprised a validation set. cfDNA lpWGS of a cohort of 104 healthy participants was used as a control. Hazard ratios (HR) were computed using Cox proportional hazards models. Odds ratios (OR) were computed using logistic regression. Longitudinal changes between baseline and subsequent timepoints were investigated using linear mixed effect models. All analyses were adjusted for tumor fraction. Results: Out of 682 TFs with available binding site annotations, 357 consistently yielded signal-to-noise ratios of at least 2:1 and were further analyzed. Analytical validation utilizing technical replicates (different samples; same time point) and biological replicates (different samples, taken at baseline and screening, weeks apart) was pursued, determining the dynamic range for these TFs at 1000 TFBS each. Assay limits of detection (LOD) and limits of quantification (LOQ) for sequencing coverage and tumor fraction were determined for each TF. Of 357 TFs, 244 were significantly mCRPC associated (Wilcoxon test, adjusted p-value < 0.05) relative to healthy controls, including AR, NKX3-1, E2F1, MYC, MYCN, and MAZ. Predictive analysis in CARD showed accessibility at UBP1 binding sites at baseline was associated with taxane sensitivity over ARSI (OR 1.44, 95% 1.07-1.95, p-value < 0.05). Furthermore, average FOXP1 accessibility scores increased at progression (average increase 1.76, SE 0.49, p-value 0.02), alongside AR (average increase 1.16, SE 0.49, p-value 0.02) and GRHL2 scores (average increase 1.33, SE 0.53, p-value 0.01). These plasma-derived data suggest patients progress with increased AR signaling. Conclusions: Evaluating nucleosome footprints at TFBS using lpWGS is a robust approach that identifies valuable functional biomarkers of drug sensitivity. Such studies offer the opportunity to interrogate disease progression through serial clinical samples, where phenotype profiling is otherwise unfeasible.