Borderline acceleration in blood epigenetic age of unclear biological importance in knee osteoarthritis progression: Specimens and data from the Osteoarthritis Initiative and Johnston County Osteoarthritis Project

OBJECTIVE: Aging is a strong risk factor for osteoarthritis (OA). Previous studies have not found accelerated blood epigenetic aging in OA patients generally. The goal of this study was to evaluate peripheral blood epigenetic aging rates among OA patient subgroups. METHODS: Genome-wide baseline blood DNA methylation data from the Osteoarthritis Initiative (n = 554) and Johnston County Osteoarthritis Project (n = 128) were generated using Illumina MethylationEPICv1 arrays on baseline samples of OA progressors (radiographic and/or pain progression in 2-5 years) vs. non-progressors. Epigenetic age was calculated using Horvath, GrimAge, PhenoAge, Horvath-IEAA, Hannum, Horvath-SkinBlood, and DunedinPACE calculators and epigenetic age acceleration was compared. Age-associated proteins were imputed using GrimAge and generalized logistic models developed to differentiate progressor groups. RESULTS: Borderline epigenetic age acceleration was found among radiographic progressors using the SkinBlood calculator (difference in epigenetic age acceleration, ΔEAA=1.5 years, q = 0.02), dual progressors (pain+radiographic) (ΔEAA = 1.2y, q = 0.05) and any progressor (pain+radiographic+dual) (ΔEAA = 1.1y, q = 0.02), although this was less than the reported error of the calculator (2.5 years). Radiographic progressors also exhibited accelerated epigenetic aging using the PhenoAge calculator (ΔEAA = 2.1y, q = 0.02). Imputed age protein-based models showed marginal performance for radiographic progression (area under the curve (AUC) = 0.63) and poor performance for other progressor types (pain AUC=0.54, dual AUC=0.52, any AUC = 0.57). Proteins chosen most frequently during modeling had prior links to OA (i.e., GDF8, MATN3, and interleukin 18). CONCLUSIONS: No meaningful difference in epigenetic age was seen among radiographic OA using 5 of 7 aging calculators, with minimal and likely not-clinically-relevant epigenetic age acceleration using the SkinBlood and PhenoAge calculators.