Investigation of Cell-Free DNA Methylation as a Biomarker of Malignant vs. Benign Lung Nodules

Jennifer Guo1; Jacob E. Till, MD, PhD1; Thrusha Puttaraju1; Aseel Abdalla1; Zhouyang Wang1; Danielle Shoshany1; Wanding Zhou, PhD2; Jeffrey Thompson, MD3*; Erica L. Carpenter, MBA, PhD1*; 1Division of Hematology-Oncology, Department of Medicine, Perelman School of Medicine, University of Pennsylvania; 2Center for Computational and Genomic Medicine, The Children's Hospital of Philadelphia; 3Division of Pulmonary, Allergy, and Critical Care Medicine, Department of Medicine, Perelman School of Medicine, University of Pennsylvania; *Co-Senior Author.

Lung cancer is the leading cause of cancer-related death worldwide, with most patients (67%) diagnosed with advanced stage disease. However, when lung cancer is diagnosed in stage I instead of stage IV, patients have a 20x better survival likelihood and one-third the treatment costs. Patients can be diagnosed with lung cancer in its early stages if lung nodules, which are discovered incidentally or through screening, are deemed to be malignant. However, many (96.4%) lung nodules detected on CT in the recent National Lung Screening Trial turned out to be false positives, ultimately leading to unnecessary invasive procedures, repeat imaging, and increased health care costs. Thus, a non-invasive biomarker that accurately discriminates malignant from benign nodules would be of significant clinical utility. DNA circulating in the blood, called circulating cell-free DNA (ccfDNA), is found even in healthy individuals. In cancer patients, a portion of this ccfDNA may have been shed by the tumor. Evaluation of ccfDNA and/or circulating tumor DNA (ctDNA) represents a non-invasive method of detecting cancer. To assess the potential of ccfDNA methylation to predict lung nodule status, a discovery set of 10 patients with benign nodules and 14 patients with malignant nodules was consented and the plasma ccfDNA samples were profiled using the Infinium Methylation EPIC array to assay 850,000 methylation sites. 159 differentially methylated regions (DMRs, unadjusted p values>0.05 and Δβ>0.15) were identified. Supervised clustering using these DMRs performed well at separating the malignant from the benign cases. Deconvolution of methylation data to determine the originating cell type composition of ccfDNA was also performed, with no significant differences detected between cases and controls for any tissue type. These data suggest a panel of DMRs that may help discern benign from malignant nodules, although these results will require validation in a larger cohort.