Multimodal Profiling of Cell-free DNA for Detection and Characterization of Circulating Tumour DNA in Low Tumour Burden Settings

Circulating tumour DNA (ctDNA) is comprised of cell-free DNA (cfDNA) fragments released into bodily fluids. ctDNA can enable personalized treatment strategies in oncology by providing a non-invasive source of prognostic, predictive, and pharmacodynamic biomarkers. Despite this, in clinical scenarios where disease burden is low, current clinical ctDNA detection methods demonstrate inadequate sensitivity without the use of matched tumour profiling. In recent years a variety of sequencing tools have been developed to improve the sensitivity of detection by increasing the number of fragments analyzed, decreasing sources of error, or enrichment of cancer-derived features. Yet few studies have comprehensively evaluated the clinical utility of tumour-agnostic detection of ctDNA. Therefore, to address this gap in knowledge, I first refined a recently described method for genome-wide profiling of cfDNA, cell-free methylated DNA immunoprecipitation by high-throughput sequencing (cfMeDIP-seq). Specifically, adaptations to the cfMeDIP-seq protocol were conducted to enable analysis of lengths of cfDNA fragments, which are thought to be shorter among tumour-derived fragments (i.e., ctDNA) compared with other sources of cfDNA. Additionally, bioinformatic pipelines were developed to assess quality assurance as well as metrics for robust identification of ctDNA. Next, I evaluated the feasibility of simultaneous profiling of multiple ctDNA-derived metrics for tumour-agnostic detection. I applied these new tools to a cohort of 30 human papillomavirus (HPV) negative head and neck squamous cell carcinoma patients (HNSCC). Specifically, I conducted cfMeDIP-seq as well as cancer personalized profiling by deep sequencing (CAPP-Seq) for detection of ctDNA-derived methylated regions and mutations, respectively. Twenty out of 30 patients had detectable ctDNA by CAPP-Seq, which I utilized to identify 941 hypermethylated regions associated with HNSCC cfDNA when compared to a cohort of 20 healthy donors. These fragments with ctDNA were confirmed by concordance of regions commonly dysregulated within publicly available tumour data, as well as by concordance of reduced fragment length and abundance between cfMeDIP-seq and CAPP-Seq. Finally, I demonstrate that multimodal detection within the same cohort is capable of robust diagnosis, prognostication, and disease surveillance. Altogether, this work describes the development of a novel strategy for ctDNA detection and validation, with promising potential for various research and clinical applications.Ph.D