Characterising developmental dynamics of adult epigenetic clock sites

Abstract

Background: DNA methylation (DNAm), an epigenetic mechanism that regulates gene activity in response to genetic and environmental influences, changes as we age. DNAm at specific sites on the genome can be used to calculate 'epigenetic clocks', which are powerful biomarkers of age, as well as of ageing. However, little is known about how these clock sites 'behave' during development and what factors influence their variability in early life. This knowledge could be used to optimise healthy ageing well before the onset of age-related conditions.

Methods: We leveraged results from two longitudinal population-based cohorts (N = 5019 samples from 2348 individuals) to characterise trajectories of adult clock sites from birth to early adulthood. To explore what factors may drive early individual differences at these clock sites, we also tested for enrichment of genetic factors and prenatal exposures based on existing epigenome-wide association meta-analyses.

Findings: We find that clock sites (i) diverge widely in their developmental trajectories, often showing non-linear change over time; (ii) are substantially more likely than non-clock sites to vary between individuals already from birth, differences that are predictive of DNAm variation at later ages; and (iii) show enrichment for genetic influences and prenatal environmental exposures, including prenatal smoking, diet and maternal physical health conditions.

Interpretation: These results suggests that age(ing)-related epigenetic processes might originate-and differ between individuals-already very early in development. Understanding what drives these differences may in future help us to devise better strategies to promote healthy ageing.

Funding: This research was conducted while C.A.M.C. was a Hevolution/AFAR New Investigator Awardee in Aging Biology and Geroscience Research. Full personal funding details, as well as cohort funding details, can be found in the Acknowledgements.

Keywords: ALSPAC; DNA methylation; Development; Early origins; Epigenetic clocks; The generation R study.

Fig. 1

Overview of comparisons made between clock sites (first-generation [n = 967] and second/third-generation [n = 821]) versus non-clock sites on the 450 K array, on the presence of a) DNAm change from birth to 18 years; b) nonlinear DNAm change at 9 years; c) inter-individual DNAm differences at birth; d) inter-individual differences in DNAm change starting from birth; e) inter-individual differences in DNAm change starting from age 6 years (mid-childhood); f) inter-individual differences in DNAm change starting from 9 years (late childhood); g) Pearson correlations between inter-individual differences at birth and at 17 years; h) Pearson correlations between inter-individual differences at birth and in DNAm change from birth; i) genetic associations with DNAm at birth; and j) prenatal environmental association with DNAm at birth. The error bars indicate the range of percentages found across clocks; significant (p < 0.05; Fisher's exact test) differences between clock sites and non-clock sites on the array based on enrichment analyses are depicted with an asterisk (∗).

Fig. 2

Examples of (a) a site from Zhang's and Hannum's clock for which DNAm is predicted to increase across development and (b) a site from Zhang's clock for which DNAm is predicted to decrease across development. Each line represents an individual's predicted level of DNA methylation over time (n = 5019 samples 2348 individuals) and the black line represents the group-level DNAm value over time, as predicted by Model 2 (M2).

Fig. 3

Example of a site from (a) Horvath's clock, for which DNAm is predicted to increase until age 6, after which no significant change happens up to age 18 (slope change p > 1 × 10⁻⁰⁷ in Model 2), and a site from the (b) the PhenoAge clock, for which DNAm is predicted to increase until age 9, after which it decreases (n = 5019 samples 2348 individuals).

Fig. 4

Example of a site from the PhenoAge clock. The arrow indicates the range of predicted inter-individual differences in level of DNA methylation at birth (n = 5019 samples 2348 individuals).

Fig. 5

Examples of a site from Zhang's (a) and Hannum's (b) and the PhenoAge (c) clock, for which inter-individual differences in DNA methylation are predicted to appear from birth (a), mid-childhood (b), and late childhood (c) onwards. The arrows indicate the range of predicted inter-individual differences in rate of change (n = 5019 samples 2348 individuals).

Fig. 6

An example of a site from the DunedinPACE clock (a) and the PhenoAge clock (b) showing predicted individual DNAm differences at birth related to a genetic polymorphism, which appear maintained (a) or exacerbated (b) across development (n = 5019 samples 2348 individuals).