Abstract PD9-08: Modeling clonal structure over narrow time frames via circulating tumor DNA in metastatic breast cancer

Abstract

Abstract Introduction: Circulating tumor DNA (ctDNA) offers the ability to repeatedly interrogate tumor genomic information, providing an opportunity for real-time monitoring of tumor genomic dynamics. In this study, we deeply analyzed multiple ctDNA samples collected over narrow time frames (days-to-weeks) from seven patients with metastatic triple-negative breast cancer (mTNBC), a cancer type known to have high ctDNA content. Methods: Patients with mTNBC were enrolled in a clinical trial of multi-kinase inhibitor cabozantinib, providing uniform and targeted treatment, and samples were collected day 1, day 8, then every 21 to 42 days on study. ctDNA was extracted from each plasma sample and underwent ultra-low pass whole genome sequencing (ULP-WGS; average depth 0.1x; n=42 samples), deep targeted panel sequencing (TPS) of 402 cancer-related genes with unique molecular identifier indexing (depth 10,000x; n=42 samples), and samples with tumor fraction (TFx) >10% underwent whole exome sequencing (WES; depth 200x; n=31 samples), with whole blood germline sequencing of both TPS and WES for subsequent analyses. Somatic copy number alterations (SCNAs) were identified from ULP-WGS and WES. PyClone with TPS was employed for clonal dynamic analyses. Predicted neoantigens were determined from WES using HLAthena. Results: A total of 42 total plasma samples from 7 patients (range 4-8 samples per patient) were collected at narrow time intervals, median 21 days (range 6 to 42 days) between samples. The median TFx across all samples was 18.1% (range 2.5% to 44.3%). TFx estimates were concordant when comparing orthogonal sequencing approaches (ULP-WGS, WES) and tumor fraction estimation algorithms (ichorCNA, FACETS). Despite all seven patients having ‘stable disease’ as best objective response, TFx dynamics were widely variable with TFx declining to lower limit of detection in three of seven patients. Of all samples, 31/42 (73.8%) had tumor fraction >10% and underwent WES; each patient had at least 3 samples that underwent both WES and TPS. There was strong agreement between TPS and WES: across all 31 shared samples, mutation recall in TPS versus WES (gold standard) was 95.5%. Variant allele frequency across all mutations detected in both TPS and WES was highly concordant (Pearson’s r=0.949). Clonal mutations were consistently detected across multiple samples within patients. When comparing genome-wide copy number from last to first available sample within each patient, copy number log ratios were largely stable within patients (union Pearson’s r=0.924) and there were not recurrent shifts in SCNAs across patients. Through statistical modeling of TPS data, we tracked distinct clonal populations for each patient over their sampling windows. Modeled clonal architecture in most patients revealed stable, polyclonal profiles, with important breast cancer driver alterations (e.g. TP53 and PIK3CA) recurrently presenting at high prevalence. Infrequently, we also detected emergence and expansion of clones over narrow time frames (weeks) containing acquired alterations poorly annotated in the breast cancer literature. We successfully predicted neoantigens from ctDNA WES at multiple time points in each patient, with evidence that patients acquired new mutations predicted to be ‘strong binder’ neoantigens over time on therapy. Conclusions: Analysis of serial ctDNA samples collected at narrow time intervals (days-to-weeks) provides unique insight into the dynamics of ctDNA. We demonstrate strong concordance across ctDNA sequencing appraoches. Evolving genomic features of tumor populations can be identified via ctDNA while on treatment, potentially providing real time insight for clinical decision-making. Citation Format: Daniel G Stover, Katharine A Collier, David Tallman, Juliet Forman, Sachet Shukla, Sarah Asad, Justin Rhoades, Samuel Freeman, Mathew Cherian, Sagar Sardesai, Romualdo Barroso-Sousa, Carrie Cibulskis, Niall Lennon, Gavin Ha, Sara M Tolaney, Viktor A Adalsteinsson, Zachary Weber. Modeling clonal structure over narrow time frames via circulating tumor DNA in metastatic breast cancer [abstract]. In: Proceedings of the 2020 San Antonio Breast Cancer Virtual Symposium; 2020 Dec 8-11; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2021;81(4 Suppl):Abstract nr PD9-08.