ESPE Yearbook of Paediatric Endocrinology

In Brief: The authors harness large-scale genetic and DNA methylation datasets in an epigenome-wide Mendelian randomization approach, to identify CpG sites that appear to be causally related to aging-related traits. These sites are collated to produce 2 epigenetic clocks: DamAge has adverse impacts on age-related outcomes, including mortality. AdaptAge appears to confer beneficial adaptations.

Steve Horvath and others originally described various epigenetic clocks based on robust age-related changes in DNA methylation that correspond closely to chronological age in a highly reproducible pattern. For example, such tools can be used in forensics to estimate the age of a person by their crime-scene DNA sample with an error margin as low as 2–5 years, or to identify individuals who exhibit an accelerated ageing trajectory. The ambition has now moved beyond prediction to causation – to identify epigenetic changes that have a causal impact on ageing. With that aim, Horvath and others established Altos Labs, whose funders include Jeff Bezos the founder of Amazon, with the goal of developing therapies to reverse aging and extend human lifespan.

In contrast to previous phenotypic observation-based approaches, here they used Mendelian randomization (MR), a genetic causal inference analysis approach that reduces the possibilities of confounding and reverse causality. Hence MR produces more robust evidence than that from typical phenotypic observations. DamAge showed remarkably strong inverse correlation with time of cell reprogramming ( R =−0.93, P =4×10−12) when mature fibroblasts were artificially converted to induced pluripotent stem cells (iPSC), indicating that these methylation changes have closely synergy with cell ‘stemness’ factors. Advanced DamAge was also seen in the rare ageing disorders Hutchinson–Gilford progeria and Werner syndrome, and in children conceived by IVF with fresh or cryopreserved embryos, but was surprisingly reduced in children born with low birth weight (SGA).

In currently available human genetic and DNA methylation datasets, only ~10% of DNA methylation sites (CpG sites) measured by arrays have a robust genetic signal (meSNP) that can be used in MR studies to model the causal impacts of changes in methylation. As such source datasets grow, this type of approach will increase substantially in power and coverage.