Plasma cell-free DNA hydroxymethylomes define disease state in EGFR-mutant non-small cell lung cancer

The epigenome of plasma cell-free DNA (cfDNA) has demonstrated promise as both a prognostic and predictive cancer biomarker in liquid biopsies. Currently, it remains unknown whether and why cfDNA 5-hydroxymethylcytosine (5hmC) can identify disease state for non-small cell lung cancer (NSCLC). We profiled 5-hydroxymethylomes using the plasma cfDNA of 302 EGFR-mutant NSCLC patients with different disease states. We found NSCLCs were epigenetically heterogeneous among individuals, especially for cfDNA 5hmC on gene EGFR. The diversity of age, sex, race, smoking status, EGFR-mutation of patients increased the epigenetic heterogeneity of NSCLC, however, only smoking status shaped disease state-associated cfDNA 5hmC. More importantly, disease state-dependent and patients' characteristics-independent cfDNA 5hmC can be linked to lung function and regulatory elements in human lung cells. Interestingly, although 5-hydroxymethylome heterogeneity of plasma cfDNA among NSCLC patients were detected substantially, cfDNA-5hmC-based machine learning model can accurately predict different disease state in NSCLC. Overall design: We performed 5hmC-Seal for plasma cfDNA of 302 EGFR-mutant NSCLC samples which included 240 CONTROLLED and 62 UNCONTROLLED samples. “Controlled Disease” was defined as: no change of therapy within 30 days of sample collection. “Uncontrolled Disease” was defined as either untreated disease or requiring a change in therapy within 30 days of sample collection.