Hallmark 3 of 12

Epigenetic alterations

All cells in your body carry the same DNA yet behave completely differently. Epigenetics makes the difference: chemical marks that determine which genes get read. These marks shift with age in a strikingly regular pattern. That regularity gave rise to the epigenetic clocks, currently the best available methods for estimating biological age.

Scientific term: Epigenetic alterations
Group: Primary causes
Hallmarks of Aging: Described in 2013

The hallmarks of aging are a research framework, not a diagnostic tool. They describe processes that occur during ageing, but they do not prove causation and are not a basis for medical decisions.

What lies behind it

The most important mark is DNA methylation, in which a methyl group is attached to a cytosine, usually at so-called CpG sites. Heavily methylated gene promoters are generally read less. In addition, histones, the packaging proteins of DNA, carry their own pattern of acetylation and methylation that regulates the accessibility of entire regions. Non-coding RNA molecules intervene in gene regulation as well.

With age this pattern shifts in two directions at once. Globally, methylation is lost, which loosens densely packed regions and can reactivate silenced sequences. Locally, methylation increases at individual promoters, frequently at genes governing cell identity and development. This is called epigenetic drift: the cell does not lose its genetic information but the clarity about which part of it should currently apply.

A second phenomenon is loss of heterochromatin organisation. Regions that should remain permanently silenced loosen up. As a consequence retrotransposons can become active, and their transcripts can in turn trigger an inflammatory response.

The regularity of this drift makes it possible to build clocks. An algorithm selects a few hundred CpG sites whose combined methylation levels correlate well with calendar age or with health characteristics, and computes an estimate from them.

How robust is the evidence

Robust evidence

Consistent human data and mechanistic animal models point in the same direction.

The first broadly usable clock came from Steve Horvath and estimates calendar age from 353 CpG sites across many tissues with a mean deviation of a few years. That accuracy was remarkable for a biological measure, but initially it only said how old someone is, not how they are doing.

The second generation was therefore aligned to health data. Levine PhenoAge was trained against a phenotypic age derived from nine blood values and is validated against ten-year mortality. GrimAge was trained directly against mortality and smoking behaviour and predicts mortality and disease onset better than calendar age alone in several cohorts.

The third generation measures pace rather than age. DunedinPACE was developed in the Dunedin cohort, in which numerous organ functions were measured repeatedly over two decades. The value indicates how many biological years pass per calendar year. Conceptually this is closer to what matters in a preventive consultation.

On the causality question, work from David Sinclair’s laboratory is relevant, in which double-strand breaks were deliberately introduced in mice and then repaired. The DNA sequence remained largely intact, the epigenetic pattern did not, and the animals showed features of ageing. That is a strong argument that epigenetic information loss does not merely accompany ageing but drives it. It is, however, an artificial mouse model.

Findings in humans

The Horvath clock estimates calendar age from 353 CpG sites across many tissues to within a few years.
Levine PhenoAge is validated against ten-year mortality and uses nine routine blood values as its training target.
GrimAge predicts mortality and disease onset better than calendar age in several cohorts.
DunedinPACE quantifies pace of ageing and draws on two decades of repeated organ measurements.
In the CALERIE trial, two years of caloric restriction measurably slowed the pace of ageing as measured by DunedinPACE.

What of this is measurable at YEARS

Partly, via adjacent markers

Of all twelve hallmarks this is the one that can most sensibly be measured. YEARS uses a methylation analysis for this and evaluates it through several established algorithms.

Methylation analysis with epigenetic clocks

Horvath, PhenoAge, GrimAge and DunedinPACE are evaluated. The practically most useful value is usually DunedinPACE, because it reports pace rather than an absolute figure and therefore reveals a direction across repeated measurements.

The nine PhenoAge blood values

Albumin, creatinine, glucose, CRP, lymphocyte percentage, mean cell volume, red cell distribution width, alkaline phosphatase and white blood cell count are part of the lab panel anyway. They allow a phenotypic age estimate even without sequencing.

Context

A single measurement says little. The methods vary between blood draws, and a result two years above or below calendar age often lies within measurement uncertainty. The exercise becomes meaningful only through repeated measurements under comparable conditions. An epigenetic age is also not a diagnosis and on its own does not justify any treatment.

Limits of this hallmark

Epigenetic clocks are statistical constructs, not a measurement of a biological state. The selected CpG sites are those that best predict the target value in the training data set. Why these particular sites is mechanistically unexplained in most cases. Two clocks can give different answers for the same person because they were trained against different targets.

Blood cell composition influences the result considerably, because different immune cell types carry different methylation patterns. An acute infection or a change in the blood count can shift the value without anything having changed in the ageing process. Good analyses correct for this, but not completely.

Most importantly: no study to date shows that deliberately improving epigenetic age leads to less disease or longer life. The clocks predict outcomes but are not validated as a surrogate endpoint. Justifying an intervention solely by lowering epigenetic age overstretches the evidence.

Frequently asked questions

What are epigenetic alterations?→

Epigenetic marks are chemical additions to DNA and its packaging proteins that control which genes are read. The best known is DNA methylation. The genetic text itself remains unchanged; only which passages are active changes. With age this pattern shifts systematically: methylation is lost overall while it increases at individual control sites. Specialists call this epigenetic drift. The cell gradually loses precision about which part of its programme should currently apply. The regularity of this drift is what makes epigenetic clocks computable.

How reliable are epigenetic clocks?→

They are the best-validated ageing markers currently available, and they have clear limits. Second- and third-generation clocks such as PhenoAge, GrimAge and DunedinPACE predict mortality and disease onset in large cohorts better than calendar age. At group level this works well. For an individual the variability is substantial: repeated measurements of the same sample can differ by several years, and blood cell composition shifts the result further. A single value therefore does not amount to a verdict. The measurement becomes informative only over time, with the same laboratory and comparable conditions.

Can you lower your epigenetic age?→

There are indications that measured pace of ageing can change. In the CALERIE trial, a randomised study of caloric restriction in healthy adults, the pace measured with DunedinPACE was slower in the intervention group after two years. That is the best randomised evidence so far. The framing matters: the effect concerned a measurement, not a clinical endpoint. Whether participants consequently fall ill later or live longer was not shown. An improvement in epigenetic age is therefore an interesting signal but not a demonstrated benefit in itself.

How do Horvath, PhenoAge, GrimAge and DunedinPACE differ?→

They were trained against different targets and therefore answer different questions. The Horvath clock estimates calendar age and matters chiefly as a proof of concept. PhenoAge targets a phenotypic age derived from nine blood values and is validated against mortality. GrimAge was trained directly against mortality and performs best at predicting disease onset. DunedinPACE estimates not an age but the speed of ageing in biological years per calendar year. For follow-up the last approach is the most practical, because a pace lends itself to repeated measurement better than an absolute value.

Sources

Horvath S. DNA methylation age of human tissues and cell types. Genome Biology. 2013;14(10):R115.
Levine ME, Lu AT, Quach A, et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging (Albany NY). 2018;10(4):573–591.
Lu AT, Quach A, Wilson JG, et al. DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging (Albany NY). 2019;11(2):303–327.
Belsky DW, Caspi A, Corcoran DL, et al. DunedinPACE, a DNA methylation biomarker of the pace of aging. eLife. 2022;11:e73420.
Waziry R, Ryan CP, Corcoran DL, et al. Effect of long-term caloric restriction on DNA methylation measures of biological aging in healthy adults from the CALERIE trial. Nature Aging. 2023;3:248–257.
Yang JH, Hayano M, Griffin PT, et al. Loss of epigenetic information as a cause of mammalian aging. Cell. 2023;186(2):305–326.

Related hallmarks

Hallmark 1 of 12