Abstract PS06-09: Fragmentomic analysis of a circulating tumor DNA targeted cancer gene panel discriminates ER status in metastatic breast cancer liquid biopsies

Abstract

Abstract Background: Circulating tumor DNA (ctDNA) isolated from the peripheral blood of patients with cancer is now routinely used in the clinic to detect tumor somatic mutations, with multiple ctDNA-targeted sequencing panels commercially available for Food and Drug Administration (FDA)-approved biomarker indications to guide targeted therapy options. More recently, it has been appreciated that ctDNA fragmentation patterns are influenced by the chromatin structure of the cell of origin, with greater diversity in fragment size in areas with open chromatin and less diversity in fragment size in areas with closed chromatin. In consequence, multiple studies have found that ctDNA fragmentation patterns can be used to accurately infer epigenomic and transcriptomic phenotypes from their tumor cells of origin. However, these analyses have previously relied on whole-genome sequencing (WGS), and it is not feasible to identify FDA-approved biomarker indications in a cost-effective manner with WGS. Methods: Peripheral blood was collected from a cohort of 90 patients with metastatic HER2-negative breast cancer (ER-positive n=71, ER-negative n=19). Cell-free DNA was extracted from plasma and sequenced using a custom 822 gene pan-cancer targeted sequencing panel (IDT). Shannon entropy of first coding exon fragment sizes (E1SE) was used to quantify fragmentation patterns. Higher diversity of fragment sizes, leading to a higher E1SE, are expected in ctDNA derived from areas of the genome with more open chromatin. A machine learning model utilizing the E1SE metric across all genes in the targeted panel was developed to distinguish between ER-positive and ER-negative disease. Results: Training cross-validated accuracy of the machine learning model was 87.8% with AUC of 0.91 to distinguish between ER-positive and ER-negative disease, despite a median ctDNA fraction of only 0.02. E1SE metrics for ESR1 and CCND1 were significantly higher in ER-positive than ER-negative samples, while E1SE metrics for EGFR, c-Kit and members of the MAP kinase family were significantly higher in ER-negative than ER-positive samples, consistent with anticipated differences in gene expression between ER-positive and ER-negative triple negative breast cancers. In contrast, HER2 E1SE was similar between groups, as expected in this clinically HER2-negative cohort. Conclusions: We have shown for the first time that sequencing from standard targeted ctDNA panels can be utilized to infer phenotypic information through analysis of ctDNA fragmentation patterns in patients with metastatic breast cancer. This expands the depth of potential biologic data that can be extracted from targeted ctDNA sequencing panels in addition to DNA mutation status in a cost-effective manner. In the future, this approach can be exploited to expand ctDNA-based biomarker development beyond DNA mutation status. Citation Format: Kyle Helzer, Jamie Sperger, Yue Shi, Viridiana Carreno, Hannah Krause, Katherine Kaufmann, Leilani Mora-Rodriguez, Matthew Bootsma, Mark Burkard, Ruth O'Regan, Kari Wisinski, Joshua Lang, Shuang (George) Zhao, Malinda West, Marina Sharifi. Fragmentomic analysis of a circulating tumor DNA targeted cancer gene panel discriminates ER status in metastatic breast cancer liquid biopsies [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PS06-09.