Background Epigenetic clocks use DNA methylation (DNAm) levels of specific sets of CpG dinucleotides to accurately predict individual chronological age. A popular application of these clocks is to explore whether the deviation of predicted age from chronological age is associated with disease phenotypes, where this deviation is interpreted as a potential biomarker of biological age. This wide application, however, contrasts with the limited insight in the processes that may drive the running of epigenetic clocks. Results We perform a functional genomics analysis on four epigenetic clocks, including Hannum’s blood predictor and Horvath’s multi-tissue predictor, using blood DNA methylome and transcriptome data from 3132 individuals. The four clocks result in similar predictions of individual chronological age, and their constituting CpGs are correlated in DNAm level and are enriched for similar histone modifications and chromatin states. Interestingly, DNAm levels of CpGs from the clocks are commonly associated with gene expression in trans. The gene sets involved are highly overlapping and enriched for T cell processes. Further analysis of the transcriptome and methylome of sorted blood cell types identifies differences in DNAm between naive and activated T and NK cells as a probable contributor to the clocks. Indeed, within the same donor, the four epigenetic clocks predict naive cells to be up to 40 years younger than activated cells. Conclusions The ability of epigenetic clocks to predict chronological age involves their ability to detect changes in proportions of naive and activated immune blood cells, an established feature of immuno-senescence. This finding may contribute to the interpretation of associations between clock-derived measures and age-related health outcomes.

Background: Loss of epigenetic control is a hallmark of aging. Among the most prominent roles of epigenetic mechanisms is the inactivation of one of two copies of the X chromosome in females through DNA methylation. Hence, age-related disruption of X-chromosome inactivation (XCI) may contribute to the ageing process in women. Methods: We analyzed 9,777 CpGs on the X chromosome in whole blood samples from 2343 females and 1688 males. We replicated findings in duplicate using one whole blood and one purified monocyte data set (in total, 991/924 females/males). We used double generalized linear models (DGLM) to detect age-related differentially methylated CpGs (aDMCs), whose mean methylation level differs with age, and age-related variable methylated CpGs (aVMCs), whose methylation level becomes more variable with age. Results: In females, aDMCs were relatively uncommon (n=33) and preferentially occurred in regions known to escape XCI. In contrast, many CpGs (n=987) were found to display an increased variance with age (aVMCs). Of note, the replication rate of aVMCs was also high in purified monocytes (95%), indicating that their occurrence may be independent of cell composition. aVMCs accumulated in CpG islands and regions subject to XCI. Although few aVMCs were associated with X-linked genes in all females studied, an exploratory analysis suggested that such associations may be more common in old females. In males, aDMCs (n=316) were primarily driven by cell composition, while aVMCs replicated well (94%) but were infrequent (n=37). Conclusions: Age-related DNA methylation differences at the inactive X chromosome are dominated by the accumulation of variability.