Epigenetic clocks to measure age emerged from the ability to cost-effectively obtain the moment to moment epigenome of an individual, the distribution of epigenetic marks on nuclear DNA that control gene expression. Cells react to their environment, and some of those reactions are characteristic of the ways in which the cellular environment changes with age. Given this data and ample computational power, it is possible to find weighted combinations of, for example, DNA methylation status at specific CpG sites that fairly accurately correlate with age. More interestingly, this appears to be a measure of biological age rather than chronological age, in that people with a higher epigenetic age than chronological age tend to have a higher incidence and later risk of age-related disease and dysfunction – and vice versa.
It remains unclear is what exactly it is that is being measured by an epigenetic clock. Which processes of aging, the accumulation of damage and downstream change, actually cause these characteristic epigenetic changes across all individuals? Is it all of them? Or only some of them? Researchers have produced clocks based on patterns of transcription and protein levels in addition to epigenetic marks, and some of these later clocks use only a handful of transcripts, proteins, or marks. It seems unlikely that the more abbreviated clocks measure more than a fraction of the causative processes of aging. Since these processes interact, and all of the facets of aging proceed at much the same pace in most people, then a clock that measures, say, only chronic inflammation, might be just as good today as a clock that is affected by all mechanisms of aging.
This is true, at least, until we start being able to repair specific forms of underlying cell and tissue damage, such as the presence of senescent cells. Some clocks will stop working usefully, and we don’t really know which ones are vulnerable to the deployment of any given approach to rejuvenation. Which is a challenge, because assessing the results of therapies that repair specific forms of underlying cell and tissue damage is exactly how we’d like to use these clocks. As things stand, no clock, epigenetic or otherwise, can be trusted for such a task until it is fairly well calibrated against a class of rejuvenation therapy via multiple life span studies.
Individuals of the same chronological age display different rates of biological ageing. A number of measures of biological age have been proposed which harness age-related changes in DNA methylation profiles. These measures include five ‘epigenetic clocks’ which provide an index of how much an individual’s biological age differs from their chronological age at the time of measurement. The five clocks encompass methylation-based predictors of chronological age (HorvathAge, HannumAge), all-cause mortality (DNAm PhenoAge, DNAm GrimAge) and telomere length (DNAm Telomere Length). A sixth epigenetic measure of ageing differs from these clocks in that it acts as a speedometer providing a single time-point measurement of the pace of an individual’s biological ageing. This measure of ageing is termed DunedinPoAm.
In this study, we examined associations between six major epigenetic measures of ageing and the prevalence and incidence of the leading causes of mortality and disease burden in high-income countries. DNAm GrimAge, a predictor of mortality, associated with the prevalence of COPD and incidence of various disease states, including COPD, type 2 diabetes, and cardiovascular disease. It was associated with death due to all-cause mortality and outperformed competitor epigenetic measures of ageing in capturing variability across clinically associated continuous traits. Higher values for DunedinPoAm, which captures faster rates of biological ageing, associated with the incidence of COPD and lung cancer. Higher-than-expected DNAm PhenoAge predicted the incidence of type 2 diabetes in the present study. Age-adjusted measures of DNAm Telomere Length associated with the incidence of ischemic heart disease. Our results replicate previous cross-sectional findings between DNAm PhenoAge and body mass index, diabetes, and socioeconomic position (in a basic model). We also replicated associations between DNAm GrimAge and heart disease.
In conclusion, using a large cohort with rich health and DNA methylation data, we provide the first comparison of six major epigenetic measures of biological ageing with respect to their associations with leading causes of mortality and disease burden. DNAm GrimAge outperformed the other measures in its associations with disease data and associated clinical traits. This may suggest that predicting mortality, rather than age or homeostatic characteristics, may be more informative for common disease prediction. Thus, proteomic-based methods (as utilised by DNAm GrimAge) using large, physiologically diverse protein sets for predicting ageing and health may be of particular interest in future studies. Our results may help to refine the future use and development of biological age estimators, particularly in studies which aim to comprehensively examine their ability to predict stringent clinically defined outcomes. Our analyses suggest that epigenetic measures of ageing can predict the incidence of common disease states, even after accounting for major confounding risk factors. This may have significant implications for their potential utility in clinical settings to complement gold-standard methods of clinical disease assessment and management.