Sven Erik Matzen

Software Architect | Cloud & Security Expert | AI-enabled Solutions

The Writing Above the Genes: Epigenetics, DNA Methylation, and the Memory of Cells

🎧 Listen to this article

Biology · 2026-09-04

EU label: fully AI-generated content Fully AI-generated article (no prior review).

The Hook: Why a Cat Can Be Two Colors

Anyone who strokes a three-colored calico cat or a mottled tortoiseshell is holding a living genetic puzzle. Almost all of these animals are female, and their coat tells a story that classical inheritance alone cannot explain. The gene for orange or black fur sits on the X chromosome. A female cat carries two X chromosomes, one perhaps with the disposition for orange, the other for black. Why, then, is the fur not a uniform blend but broken into cleanly separated orange and black patches?

The answer has nothing to do with the DNA sequence. Every cell in the cat's body carries exactly the same genes in its nucleus – the same two X chromosomes with the same dispositions. And yet each skin cell, early in embryonic development, silences one of its two X chromosomes, randomly the one or the other, permanently and for all of its daughter cells. Where the orange X was switched off, black fur grows; where the black X is silent, orange grows. The blueprint is identical everywhere – what differs is the decision about which parts of it are read and which are locked away.

The same principle explains an even more everyday marvel: each of the roughly thirty trillion cells in our body – the neuron in the brain, the liver cell, the white blood cell – carries the same genome, the same roughly twenty thousand genes. Yet they are as different as a turbine and an accountant. They differ not in which genes they possess, but in which ones they have switched on and off – and, crucially, in the fact that they pass this decision faithfully down through hundreds of cell divisions. There must therefore be, above the DNA sequence, a second layer of information: a memory that is written not in the letter-text of the genes, but in the annotations in their margins. This second layer is the subject of epigenetics.

The Central Concept: Above the Genes, Not Within Them

The word reveals its meaning: the prefix epi is Greek for "above," "upon," or "in addition to." Epigenetics denotes changes in gene activity that do not rest on a change in the DNA sequence and that are nevertheless stable – often across many cell divisions, and in special cases possibly even across generations. If the genome is the sheet music, then epigenetics is the interpretation: which passages are played loudly, which are muted, which are left out entirely. The text remains, yet the music changes.

The term is older than its modern molecular understanding. The British developmental biologist Conrad Waddington coined it in the 1940s, long before anyone knew of the DNA double helix. Waddington sought a bridge between the genotype – the hereditary disposition – and the phenotype, the visible form of an organism. His famous image is the epigenetic landscape: a ball rolling down a hill, choosing one valley or another at each branch point, and finally coming to rest in a particular hollow. The ball is the developing cell, the valleys are the possible fates – nerve, muscle, skin. Once settled into a valley, the cell does not readily reverse course. Waddington sensed that gene expression is channeled, that development sets switches which then hold fixed. He did not know the mechanism. We know it today in outline, and it consists of several interacting tools.

It is scientifically established today that at least four classes of mechanism form this second layer: DNA methylation, the modification of histones around which the DNA is wound, the active remodeling of chromatin by molecular machines, and regulation by non-coding RNA. They interlock, and together they decide which part of the genome is accessible to the reading machinery and which remains sealed. Let us go through them in turn.

Part 1: The Chemical Bookmark – DNA Methylation

The best-understood epigenetic tool is also the simplest. In DNA methylation, an enzyme attaches a tiny chemical group – a methyl group, one carbon atom with three hydrogen atoms – to one of the four letters of the genetic code: to cytosine (C). Cytosine becomes 5-methylcytosine, occasionally called the "fifth letter" of the genome. The sequence itself does not change; the cytosine still pairs with guanine. But the small flag now protruding from the major groove of the double helix changes who is allowed to settle at this spot.

In mammals this methylation occurs almost exclusively where a cytosine in the DNA strand is immediately followed by a guanine – this dinucleotide is written CpG, the p standing for the phosphate bond between the two. Such CpG sites cluster in certain regions, the CpG islands, which strikingly often lie at the start (the promoter) of genes. As a rule of thumb: if the CpG island in a gene's promoter is heavily methylated, the gene is usually silent. The methyl groups serve as docking sites for proteins that pack the region tightly, and at the same time they interfere with the binding of reading factors. Methylation in the promoter usually means: no entry.

Responsible for this marking is a family of enzymes, the DNA methyltransferases (DNMTs). Two tasks are distinguished. The de novo methyltransferases DNMT3A and DNMT3B place new marks, for instance during early embryonic development, when the pattern is first laid down at all. The maintenance methyltransferase DNMT1, by contrast, ensures that the pattern survives every cell division. And here lies the true elegance: after the DNA is copied, the new double strand is at first methylated only on the old half – hemimethylated, that is. DNMT1 recognizes this half-methylation and completes the missing marks on the new strand. In this way the epigenetic pattern is copied like a text and passed to both daughter cells. This is precisely what makes methylation a memory: it is stable, it is heritable from cell to cell, and it explains why a liver cell gives rise to liver cells and not suddenly to neurons.

What has been written can also be erased. An enzyme family called TET oxidizes 5-methylcytosine step by step and thereby initiates its removal, so that unmarked cytosine is restored. There are thus writers (DNMTs), erasers (TETs), and, as we shall see shortly, readers as well. Epigenetics is not a rigid label but a dynamic system of placing, reading, and removing.

Part 2: The Packaging Decides – Histones and the Histone Code

DNA is astonishingly long. If the genome of a single human cell were stretched out, it would form a thread about two meters long – housed in a nucleus a few thousandths of a millimeter across. So that this thread fits inside without matting into a hopeless tangle, it is intricately coiled. The DNA winds around small protein spools, the histones. Roughly 147 base pairs wrap around a core of eight histone proteins; this unit is called a nucleosome and resembles a bead on a string. The whole of DNA plus packaging proteins is called chromatin.

This packaging is not passive storage. How tightly or loosely the DNA is wound around the histones determines whether the reading machinery can reach it at all. Densely packed chromatin – heterochromatin – is effectively locked; loosely packed euchromatin is open and active. And the cell regulates this density deliberately by chemically altering the ends of the histone proteins that protrude from the nucleosome (the "histone tails").

The variety of these alterations is large. An acetylation – attaching an acetyl group to certain lysines of the histone tail – neutralizes their positive charge, loosens their grip on the negatively charged DNA, and opens the chromatin; it therefore usually acts activating. It is placed by histone acetyltransferases (HATs) and removed by histone deacetylases (HDACs). A methylation of histones, by contrast, is ambiguous: depending on its location it can activate or repress. The mark H3K4me3 (methylation at lysine 4 of histone H3) is found at active gene starts, whereas the marks H3K9me3 or H3K27me3 characterize silenced regions. Add to this phosphorylation, ubiquitination, and further modifications.

From this abundance grew an influential idea around the turn of the millennium, the histone code hypothesis, formulated among others by Brian Strahl and C. David Allis (2000) and by Thomas Jenuwein and Allis (2001). It holds that the combination of histone marks forms a kind of code that is read by specialized proteins and translated into an activity state. The actors involved are often sorted into three roles: writers, which place marks; erasers, which remove them; and readers, which recognize a particular mark and thereupon recruit further machinery. Whether "code" is quite the right word is debated, since the marks often act more like probability shifts than strict determinism. But the basic picture – that the chemical annotation of the packaging steers gene activity – is well supported today.

Part 3: Remodeling and Silent Conductors – Chromatin Remodeling and RNA

Two further mechanisms complete the toolkit. The first is the active remodeling of chromatin. Molecular machines that consume energy in the form of ATP can slide nucleosomes along the DNA, remove them entirely, or exchange their components. In doing so they expose or hide regions – a bouncer rearranging nucleosomes like furniture to make a spot accessible or to block it.

The second is regulation by non-coding RNA. Not every RNA molecule is translated into a protein; many perform a task themselves. Short microRNAs can intercept the messenger RNA of other genes and throttle their translation. Long non-coding RNAs can organize entire chromosome segments. The most striking example takes us back to the calico cat: the silencing of one X chromosome is initiated by a long RNA called Xist, which is read from the very X to be silenced, drapes itself like a cloak over the chromosome, and there recruits the repressive packaging. The British geneticist Mary Lyon postulated this X-inactivation as early as 1961 – decades before Xist was known. The result, a densely packed and silenced X, is visible under the microscope as a compact body at the edge of the nucleus, named the Barr body after its discoverer. The coat patches of the cat are the visible map of a decision that each of her precursor cells once made in the embryo.

Part 4: Whose Copy Counts – Genomic Imprinting

We inherit one copy of nearly every gene from each parent, and normally both copies carry equal weight. For a small but important group of genes – in humans estimated at a few hundred – this is different. Here only the copy from one particular parent is read, the other being epigenetically silenced from the outset. Whether the maternal or the paternal copy is muted depends solely on which germline it came from. This phenomenon is called genomic imprinting, and the imprint is set by DNA methylation in the egg or sperm cell respectively.

The textbook example is the gene pair IGF2 and H19, which hang on a shared control region. IGF2 encodes a growth factor that makes the fetus grow; it is read only from the paternal copy. The neighboring H19 acts rather as a growth brake and is read only from the maternal copy. Why this strange asymmetry? The most influential explanation comes from the evolutionary biologist David Haig and is called conflict theory or kinship theory. From the standpoint of the paternal genes, a large, strong offspring is advantageous, even if it demands much of the mother – the father does not bear the costs of pregnancy. From the standpoint of the maternal genes, restraint is wiser, for the mother must divide her resources among several possible offspring. The imprint would then be the fossilized outcome of an evolutionary tug-of-war between the parents' interests, fought out in the child's genome. I am of the opinion that this theory is elegant and well supported, but honesty requires saying that it does not explain every imprinted locus and remains under discussion.

How consequential imprinting is becomes clear from two diseases that can arise from the same missing piece of chromosome. If a certain segment on the paternal chromosome 15 is missing, Prader-Willi syndrome results, marked among other things by insatiable hunger and muscle weakness. If exactly the same segment on the maternal chromosome 15 is missing, the entirely different Angelman syndrome arises, with severe developmental disorders. The same gap, two wholly different fates – the only difference being from which parent the intact counterpart should have come and whether it is imprinted, that is, silenced, there. No example makes it clearer that in our heredity it counts not only what is written, but who wrote it.

Part 5: The Environment Writes Too – Nutrition and the Hunger Winter

Up to here one might take epigenetics for a purely internal program: neatly laid down during development, faithfully copied. But the exciting – and for the reader perhaps unsettling – part is that the environment can write into this second layer. Two famous studies made this popular.

The first comes from the Duke researchers Robert Waterland and Randy Jirtle in 2003 and uses a special mouse model, the agouti mouse. These animals carry, upstream of the coat-color gene, a built-in jumping DNA element whose degree of methylation varies randomly. If the element is lightly methylated, the mouse is yellow, obese, and disease-prone; if it is heavily methylated, the mouse is lean and brown – with genetically identical heredity. Waterland and Jirtle fed pregnant mice a diet rich in so-called methyl-group donors (substances such as folic acid, vitamin B12, choline, and betaine, which supply the body with the material for methylation). The result was unambiguous: the offspring of the mothers fed this way were more often brown and lean, their jumping element more heavily methylated – although nothing had changed in the genes themselves. The mother's food had shifted the epigenetic setting of the children.

The second study concerns humans and a tragic natural experiment. In the winter of 1944/45 the German occupiers imposed a blockade on the west of the occupied Netherlands; the result was a severe famine, the Dutch Hunger Winter. Because the Netherlands kept good birth records before and after, decades later it was possible to identify people who were conceived or carried during exactly this period. Bastiaan Heijmans and colleagues compared such individuals, in a much-cited 2008 paper, with their unexposed siblings. Those exposed to the hunger in the first roughly ten weeks after conception showed, even six decades later, lower methylation at the imprinted IGF2 gene than their own sibling. Those affected only at the end of pregnancy did not show this difference. This suggests that early development is a particularly sensitive window in which the environment leaves lasting epigenetic traces – traces that fit the unfavorable metabolic profile found in this group.

Here scientific caution is in order. These studies demonstrate that early environmental conditions are associated with epigenetic differences within the same individual. They do not, by themselves, demonstrate that such traces are passed on across generations – that is an entirely different, more contested claim, to which we now turn.

Part 6: The Great Erasure – Why Inheritance Across Generations Is Hard

The tempting notion that the experiences of our grandparents live on through epigenetic marks in our cells must be measured against a hard biological fact: reprogramming. Between two generations, the epigenetic pattern in mammals is not preserved but largely erased and rewritten twice.

The first great wave of erasure runs directly after fertilization: the methylation pattern of egg and sperm is almost entirely cleared away, so that the fertilized egg becomes a totipotent cell able to give rise to any tissue. The patterns are then set anew in a tissue-specific manner. The second wave of erasure strikes the primordial germ cells, from which egg and sperm later arise: here too the epigenome is radically reset, so that the next generation begins with a clean, correctly imprinted slate. This double erasure is the reason why an epigenetic change acquired in the body normally does not reach the grandchildren. It is, in a sense, the biological safeguard against the Lamarckian idea that acquired characteristics are inherited directly.

That cells can reset their epigenome so deeply at all is itself one of the great discoveries of modern biology. Shinya Yamanaka showed in 2006 that already-specialized body cells can be returned to an embryonic state with only four factors – to induced pluripotent stem cells (iPS cells). Waddington's ball can therefore, under laboratory conditions, roll back up the hill. Yamanaka received the Nobel Prize for this in 2012. Reprogramming is real; it is only, in natural inheritance, an enemy of the transmission of epigenetic experience.

How, then, does the much-discussed transgenerational epigenetic inheritance stand? The finding is split. In plants, in the roundworm C. elegans, and in the fruit fly, the stable transmission of epigenetic states across several generations is well documented – these organisms erase their epigenome far less thoroughly. In mammals, by contrast, it remains contested. The obstacle is precisely the reprogramming described: a trait can only be truly transgenerationally inherited if the underlying mark survives both waves of erasure and can moreover be cleanly separated from genetic and maternal-environmental influences. One therefore distinguishes strictly between intergenerational effects – such as the direct imprinting of a fetus, and of its germ cells already formed within the fetus, by the mother's environment – and true transgenerational inheritance, still visible in generations that were never themselves exposed to the original cause. Review articles from 2024 counsel caution: much of what circulates popularly as "the grandparents' trauma in the genes" is not cleanly demonstrated in humans and is often inseparable from genetic factors. At the same time, ingenious experiments are accumulating – for instance with CRISPR-based epigenetic editing that transmits deliberately placed methylation in mice across at least one generation (Takahashi and colleagues) – showing that transmission is possible in principle when a mark escapes erasure. I am of the opinion that the most honest summary is this: in mammals the mechanism is the exception, not the rule, and extraordinary claims here require extraordinary evidence.

Part 7: When the Memory Goes Wrong – Epigenetics in Medicine

Because the epigenetic layer steers gene activity, its going astray does considerable harm – and because, unlike the DNA sequence, it is reversible, it opens new therapeutic routes at the same time.

This shows most clearly in cancer. Tumor cells almost always carry a distorted epigenome. Typical is a double pattern: the genome as a whole loses methylation (global hypomethylation), which makes the chromosomes unstable and awakens normally dormant regions. At the same time, of all things, the promoters of important tumor suppressor genes – the genes that put on the brakes and send damaged cells to their death – are deliberately hypermethylated and thereby silenced. The effect resembles a mutation that destroys the brake gene, except that here no letter has been altered; the gene is merely locked. Cancer is thus also an epigenetic disease. This is good news for therapy, for a locked gene can in principle be reopened. Azacitidine and decitabine inhibit the DNA methyltransferases and can reactivate silenced protective genes; they are approved for certain blood disorders such as myelodysplastic syndrome. HDAC inhibitors, which block the histone deacetylases and open the chromatin, are used in certain lymphomas. These epigenetic drugs are no panacea, but they are the living proof that the second layer can be deliberately written to and erased.

A second, fascinating field of application is the measurement of age. In 2013 Steve Horvath showed that the methylation pattern at a carefully selected handful of sites – his first clock used 353 CpG positions – changes with age so regularly that one can estimate a person's age astonishingly precisely from a tissue sample. This epigenetic clock works across many tissues, from the embryonic cell to the centenarian. Remarkably, its prediction not only hits calendar age but can also deviate from it: if a person's clock shows a higher "biological" age than their identity card, this is statistically linked to elevated risk of disease and death. The epigenetic clock is thereby one of the best measures of biological aging we have – a tangible proof that the second layer changes over a lifetime and in doing so reveals something about our condition. Anyone interested in the related cellular aging will find the other great clock of aging described in The Countdown in the Nucleus: Telomeres, Telomerase, and the Clock of Aging.

The Central Takeaway

If this article is to leave a single idea, let it be this: a blueprint is not the same as what is built from it. The DNA sequence determines what is possible – the complete repertoire of components. Epigenetics determines what actually becomes of it, by deciding in every cell and at every moment which parts of the plan are read and which are locked. This decision is stable enough to be a memory that lets a liver cell give rise to liver cells; and it is flexible enough to be influenced by nutrition, development, and environment.

For the engineer who builds systems, this is a familiar and deep lesson. Between the rigid specification and the running behavior there is always a layer of configuration, of state, of interpretation – and in this layer often dwells the real complexity. Whoever reads only the source code does not understand the system; one must also know which switches are set, which state has accumulated over time, and which environmental influences have shaped it. The cell has been solving this problem for billions of years with a writable, readable, erasable memory above the genes. At the same time, reprogramming counsels humility toward overblown promises: biology has built in safeguards that prevent every acquired imprint from being passed unfiltered to the next generation.

A practical prompt to action: the next time you meet a bold claim – "grandmother's stress is in your genes," "this food reprograms your heredity" – ask yourself whether an intergenerational effect (direct imprinting in the womb) is being confused with true transgenerational inheritance, and whether genetic factors have been cleanly excluded. The same skepticism that marks a good security analysis also makes for good reading of epigenetics.

A Question to Reflect On

If two genetically identical people – identical twins – develop an increasingly different epigenome over the years because their lives take different courses: how much of what we take to be our unchangeable "genetic fate" is in truth a configuration written over time and, in principle, changeable – and what follows from that for our notion of identity, responsibility, and change?

Cross-References in the Vault

Sources

← All articles