Circulating tumor DNA reveals complex biological features with clinical relevance in metastatic breast cancer

Abstract

Liquid biopsy has proven valuable in identifying individual genetic alterations; however, the ability of plasma ctDNA to capture complex tumor phenotypes with clinical value is unknown. To address this question, we have performed 0.5X shallow whole-genome sequencing in plasma from 459 patients with metastatic breast cancer, including 245 patients treated with endocrine therapy and a CDK4/6 inhibitor (ET + CDK4/6i) from 2 independent cohorts. We demonstrate that machine learning multi-gene signatures, obtained from ctDNA, identify complex biological features, including measures of tumor proliferation and estrogen receptor signaling, similar to what is accomplished using direct tumor tissue DNA or RNA profiling. More importantly, 4 DNA-based subtypes, and a ctDNA-based genomic signature tracking retinoblastoma loss-of-heterozygosity, are significantly associated with poor response and survival outcome following ET + CDK4/6i, independently of plasma tumor fraction. Our approach opens opportunities for the discovery of additional multi-feature genomic predictors coming from ctDNA in breast cancer and other cancer-types.

Plasma ctDNA is a promising method to determine patient outcome in multiple cancer types. Here, the authors use shallow WGS to create machine learning signatures to identify tumor phenotypes and predict therapy response in patients with metastatic breast cancer.