We performed a case–control epigenome-wide association study (EWAS) to characterize asymptomatic versus symptomatic SARS-CoV-2 infection. This study was a sub-analysis of a larger study that described methylation patterns associated with COVID-19 severity, and the methods outlined here are as described previously [5].

All patients aged 18 years or older who presented to the University of Colorado Hospital from 1 March 2020 to 31 July 2020 and received a SARS-CoV-2 test by RT-PCR for any reason were eligible for this study (e.g., prior to hospitalization for COVID-19, surgical procedures, routine hospitalization, return to work clearance). Cases were defined as those who tested positive for SARS-CoV-2 by RT-PCR but had no symptoms; symptomatic patients with a positive test result served as controls. We tested peripheral blood DNA samples with the Illumina Infinium MethylationEPIC BeadChip and abstracted clinical data from the associated clinical visit from the electronic health record (EHR). All patients were followed across the University of Colorado Health (UCHealth) System over time for a minimum of 3 months to determine presence of symptoms at the time of the blood draw and to determine development and severity of symptoms and clinical outcomes (e.g., ED discharge, hospital admission, ICU, death).

DNA was extracted on the bead-based, automated extraction Maxwell® RSC System (Promega) and quantified using absorbance (NanoDrop 2000; Thermo Fisher Scientific, Waltham, MA) and fluorescence-based methods (Qubit; Thermo fisher Scientific, Waltham, MA). DNA quality was assessed with Agilent TapeStation (Agilent, Santa Clara, CA). Biospecimens were then uploaded to the Colorado Anschutz Research Genetics Organization (CARGO) laboratory information management system (LIMS).

Purified DNA samples were processed with Zymo EZ-96 DNA Methylation bisulfite conversion kits (Zymo, Irvine, CA). Random hexamer priming and Phi29 DNA polymerase were used for whole-genome amplification, and amplification products were enzymatically fragmented, purified from dNTPs, primers, and enzymes, and applied to the EPIC BeadChip.

EWAS analyses were performed on the entire epigenotyped dataset. Preprocessing and association testing was performed with the GLINT package. We used EPISTRUCTURE and ReFACTor to estimate components to adjust for population structure and to account for cell-type proportions, respectively. A linear mixed-effects model was fit to each probe and adjusted for age, sex, chip position, 6 ReFACTor components, 1 EPISTRUCTURE component, and a covariance matrix accounting for possible relatedness. Probes were annotated to CpG islands and genic regions with annotatr.

The study protocol was approved by the Colorado Multiple Institutional Review Board (COMIRB) and the research adheres to the ethical principles of research outlined in the U.S. Federal Policy for the Protection of Human Subjects. Patients were consented for blood collection and electronic health record (EHR) data abstraction through the University of Colorado COVID-19 Biorepository or the University of Colorado Emergency Medicine Specimen Biobank (EMSB). Raw data were generated at Health Data Compass and the Colorado Center for Personalized Medicine.