What Is Epigenetics? Gene Expression Beyond DNA

The Marks Themselves

DNA methylation. A methyl group is added to the fifth carbon of a cytosine base, almost always where a cytosine is followed by a guanine, a site called a CpG. Methylation in a gene promoter region usually reduces transcription, and the mark is copied to the daughter strand during replication, which is what makes it heritable across cell divisions. It is also the mark used by every epigenetic clock, because it is stable, quantifiable at hundreds of thousands of sites simultaneously, and cheap to assay on an array.

Histone modification. DNA is wound around histone octamers, and the histone tails carry acetylation, methylation, phosphorylation, ubiquitination and other marks. Acetylation generally loosens the wrapping and permits transcription; the effect of histone methylation depends entirely on which residue is modified. This is where sirtuins act, as histone deacetylases.

Chromatin architecture. Above individual marks, the genome is organised into compacted heterochromatin and accessible euchromatin, and into topological domains that determine which regulatory elements can reach which genes.

Non-coding RNA. MicroRNAs and long non-coding RNAs regulate transcript stability and translation, and are often grouped with epigenetics though they operate by a different logic.

What Changes with Age

Age-related epigenetic change is not random drift, and this is the finding that made the field central to aging biology.

Directional methylation change at specific sites. Some CpG sites gain methylation with age and others lose it, consistently enough across individuals that a few hundred sites predict chronological age to within a few years. That reproducibility is the entire basis of epigenetic clocks.

Global hypomethylation with focal hypermethylation. Overall methylation declines while CpG islands in some promoter regions gain it. The combination is characteristic.

Heterochromatin loss. Compacted regions loosen, allowing expression of sequences that should be silent, including retrotransposable elements. Their activation contributes to inflammatory signalling through cytoplasmic DNA sensing.

Loss of epigenetic information. The strongest current framing, associated with work from David Sinclair's laboratory and others, holds that aging involves degradation of the cell's record of its own identity rather than accumulated damage to the sequence. The supporting experiments show that resetting expression of certain factors restores youthful expression patterns and function in mouse tissues, which is a striking result whose scope and durability are still being established.

Epigenetic Clocks in One Paragraph Each

First generation. Horvath and Hannum clocks, trained to predict chronological age from methylation. They do that well, which is a slightly odd objective: a perfect predictor of calendar age tells you nothing calendar age did not.

Second generation. PhenoAge and GrimAge, trained against clinical markers and mortality rather than against calendar age. They predict outcomes considerably better, and GrimAge incorporates methylation-based estimates of smoking history and plasma proteins.

Third generation. DunedinPACE and similar, trained on longitudinal data to estimate the current rate of biological aging rather than accumulated state. Conceptually the most interesting for intervention assessment, because a rate can change in months where an accumulated state cannot.

Tissue-specific and cause-specific clocks. Skin, blood, brain and other tissue clocks exist, as do clocks trained on specific exposures. Their disagreement with one another is informative rather than embarrassing.

All of them carry test-retest variation that can exceed a year, which sets a floor below which reported changes are not changes.

What Is Modifiable, and How Much

The honest answer separates three claims that are routinely merged.

Marks respond to environment: well established. Smoking leaves a distinctive and long-lasting methylation signature. Diet, exercise, sleep, psychological stress and pollutant exposure all associate with methylation differences. Folate, B12, B6, choline and betaine supply the one-carbon metabolism that provides methyl donors, which is a direct biochemical link between nutrition and the mark.

Estimators can be moved: partially established. Several small trials report reductions in epigenetic age estimates following multi-component lifestyle or supplement interventions. The studies are small, often without adequate controls, sometimes with the estimator chosen after the fact, and the reported changes are frequently within the noise range of the assay.

Moving the estimator changes the trajectory: not established. This is the claim that matters and the one with the least support. An algorithm trained to predict mortality from methylation can be moved by altering its inputs without the mortality relationship following. Establishing that it does follow requires long outcome studies that do not yet exist.

What the Word Does Not License

Epigenetics has become a rhetorical device, and three usages should be recognised as such.

"Epigenetics means genes are not destiny." True in a narrow sense and usually deployed to imply far more control than exists. Most epigenetic variation is a consequence of cell type, developmental history and age, not of choices made last year.

"You can reprogramme your epigenetics." Reprogramming has a specific technical meaning involving forced expression of pluripotency factors, demonstrated in mice and in cell culture. It is not what a supplement or a diet does, and borrowing the term for lifestyle change is a category error.

"Trauma is inherited epigenetically." There is animal evidence for transgenerational effects and human evidence is thin, confounded by shared environment, and considerably weaker than popular accounts suggest.

The genuinely interesting statement is more modest and more useful: the epigenome is a readable record of biological state that changes with age in reproducible ways, and it is the best current substrate for measuring aging even though what to do about the measurement remains unsettled.