Twenty-year longitudinal evaluation of blood-based DNA methylation biomarkers for early prediction of Alzheimer’s disease

Abstract Background DNA methylation (DNAm) is an epigenetic mechanism reflecting both inherited and environmental influences, and is a promising biomarker of multifactorial aging-related disorders like Alzheimer’s disease (AD). Early prediction of AD is critical, but little is known about the time-course of DNAm biomarkers long before symptom onset. Methods The long-term predictive ability of four existing DNAm-based epigenetic age acceleration clocks was tested in a longitudinal case-control sample (50 late-onset AD cases; 51 age- and sex-matched controls) with prospective data up to 16 years prior to clinical onset (mean: 8 years), and a post-onset follow-up. In addition, novel blood-based DNAm biomarkers for AD prediction were generated with epigenome-wide longitudinal linear mixed effects models, as well as sparse partial least squares discriminant analysis applied at time-points 10–16 years pre-onset and 0–7 years post-onset. Results Epigenetic age acceleration clocks did not differentiate cases from controls at any point during the 20-year follow up time (ps > 0.05). Our new DNA biomarkers, comprising 73, 7, and 27 CpG sites respectively, had excellent in-sample discriminatory and predictive accuracy on average 8 years prior to clinical onset (AUCs = 71.1–98.2% including age, sex, and white blood cell proportions). The longitudinal panel of CpGs replicated nominally (p = 0.012) in an external cohort (n = 146 cases, 324 controls). However, compared with the established genetic marker APOE ε4 our panel had a limited effect size (OR = 1.38 per 1 SD panel score increase vs. OR = 13.58 for ε4-allele carriage) and discriminatory accuracy in the external cohort (AUC = 77.2% vs. 87.0% for models with age, sex, and white blood cell proportions). A literature review showed low overlap (n = 4) across 3275 CpGs previously reported to be AD-associated in 8 published studies, and no overlap with our currently identified CpGs. Conclusions The results extend prior studies showing a limited predictive and prognostic value of epigenetic age acceleration in AD by considering a longer pre-onset follow-up time, and with appropriate control for age, sex, APOE, and white blood cell proportions. The findings further highlight challenges with replicating discriminatory or predictive CpGs across studies.