Epigenetics: How Your Environment Changes the Way Your DNA Is Read

Every cell in your body carries essentially the same genome: about 3.05 billion base pairs of DNA, including roughly 19,400 protein-coding genes. Yet a liver cell, a neuron, and a skin cell behave in completely different ways. They are reading the same instructions and producing different outputs.

Epigenetics is the field that explains the gap. It studies the chemical layer on top of DNA that controls which genes are read, when, and how strongly, without altering the sequence itself. Unlike the sequence, this layer can shift with age, diet, smoking, stress, and experience. Below is how it works, which experiments established it, what the evidence does and does not show, and why one of its biggest questions, inheritance across generations, is still argued over.

Key takeaways

• Epigenetics refers to chemical marks on DNA and its packaging proteins, mainly DNA methylation and histone modifications, that change gene activity without changing the DNA sequence.
• These marks are not permanent. Cells add, remove, and rewrite them during development and throughout life, in response to signals that include nutrition and chemical exposures.
• Studies of famine, maternal care, smoking, and exercise have found measurable methylation differences, though most human findings are associations, not proof of cause.
• Whether epigenetic changes pass from parents to children is documented in plants, worms, and flies, but remains disputed in humans.

Why a liver cell and a neuron can share one genome

The scale of the genome makes the puzzle sharper. The most complete human sequence to date, the Telomere-to-Telomere assembly published in Science in 2022, spans about 3.055 billion base pairs (Nurk et al., 2022). The current GENCODE annotation (release 50) lists 19,442 protein-coding genes (GENCODE statistics). Every normal body cell carries that same set, with the notable exceptions of mature red blood cells and the sperm and egg cells.

Think of the genome as a very large cookbook shared by every restaurant in a city. A pastry shop and a steakhouse own the same book, but each one has its own bookmarks, highlighted recipes, and pages taped shut. The text is identical; what differs is which pages are open. The epigenome plays the role of the bookmarks, highlights, and tape. How these marks are laid down and maintained is covered in our explainer on how genes actually work, which is a good companion if the vocabulary of promoters and transcription is new.

The National Human Genome Research Institute defines epigenetics as "a field of study focused on changes in DNA that do not involve alterations to the underlying sequence." Its glossary adds that the DNA letters and the proteins that interact with DNA "can have chemical modifications that change the degrees to which genes are turned on and off" (NHGRI glossary).

The term is easily confused with plain gene expression, which is any change in how active a gene is, including brief responses to a hormone. Epigenetics usually refers to the more stable layer of marks that persists through cell division, so that a dividing liver cell makes two liver cells and not a liver cell and a neuron. Every epigenetic change affects expression, but not every change in expression is epigenetic.

How DNA methylation silences a gene

DNA methylation is the best-studied epigenetic mark. A review in Neuropsychopharmacology describes it as the addition of a methyl group (a carbon atom with three hydrogens) to the fifth carbon of the cytosine base, forming 5-methylcytosine. The mark regulates gene activity in two ways: by recruiting repressive proteins to the DNA, or by physically blocking the transcription factors that would otherwise switch the gene on (Moore, Le, and Fan, 2013).

In mammals the mark mostly occurs where a cytosine is followed by a guanine, a pairing called a CpG site. When many CpG sites in a gene's promoter (the stretch of DNA just upstream that starts transcription) are methylated, the gene is generally quiet. Think of a thick piece of tape over a light switch: the wiring is intact, but the switch is hard to flip.

Methylation patterns change during development through both de novo methylation (laying down new marks) and demethylation (removing them), producing stable, cell-type-specific patterns. That is how a cell "remembers" what it is. Even neurons, which no longer divide, keep the enzymes for adding and removing methyl groups, and the same review notes that neuronal activity and environmental stimuli can alter methylation patterns there.

Smoking leaves a readable trace. The clearest example of the environment writing on the methylome is cigarette smoking. A 2016 meta-analysis in Circulation: Cardiovascular Genetics pooled 15,907 blood samples from 16 cohorts: 2,433 current smokers, 6,518 former smokers, and 6,956 people who had never smoked. Comparing current with never smokers, it found 2,623 CpG sites in 1,405 genes with strongly significant methylation differences. Of those sites, 185 were still significantly different in former smokers than in never smokers (Joehanes et al., 2016, Circulation: Cardiovascular Genetics). The marks are a record of exposure, and a partly erasable one.

What histone tags do to the winding of DNA

The second major mechanism works on the packaging. DNA does not float loose inside the nucleus. It is wound around protein spools called histones, and the tightness of the winding affects how easily the cell's machinery can reach a gene.

Cells attach small chemical tags to the tails of histone proteins. A review of the so-called histone code lists acetylation, methylation, phosphorylation, ubiquitination, sumoylation, and ADP-ribosylation, and explains that these modifications alter chromatin structure and serve as binding sites for transcription factors and other enzymes (Linggi et al., 2005, Journal of Cellular Biochemistry). Acetylation is generally associated with looser, more active chromatin. Methylation of histones can either activate or repress, depending on which amino acid is tagged.

Picture thread on a spool. Loosely wound thread can be pulled off easily; tightly wound thread cannot. A cell can change how tightly each region is wound without cutting the thread. DNA methylation and histone modification also work together, with each influencing where the other is placed, which is why scientists often speak of a single epigenome made of several layers.

Feature	DNA methylation	Histone modification
What is modified	The cytosine base in DNA	Tails of the histone proteins that package DNA
Typical chemical tags	A methyl group (5-methylcytosine)	Acetyl, methyl, phosphate, and other groups
Main effect	Recruits repressive proteins or blocks activators	Changes how tightly DNA is wound and who can bind
Common measurement	Methylation arrays and sequencing	Antibody-based assays and sequencing
Reversible?	Yes, by active and passive demethylation	Yes, by enzymes that add and remove tags

Four experiments that showed the environment can leave a mark

The idea that nutrition or care could write on the genome was a hypothesis for decades. These studies turned it into data, and each comes with limits worth knowing.

Yellow mice and their mothers' diet

In 2003, Robert Waterland and Randy Jirtle studied viable yellow agouti mice, which carry a mobile piece of DNA (a transposable element) next to the agouti gene. They supplemented pregnant mothers with extra folic acid, vitamin B12, choline, and betaine, nutrients that feed the chemistry supplying methyl groups. The offspring's physical traits changed, and the change tracked with increased methylation at that locus (Waterland and Jirtle, 2003). In this strain the visible effect is a shift in coat color and body weight, which made it a favorite teaching example.

The authors themselves drew a cautious conclusion: dietary supplements long considered purely beneficial could have unintended effects on epigenetic regulation. That is a hypothesis about mice with an unusual, metastable gene. It does not translate into advice for people.

Rat pups and the licking-and-grooming effect

In 2004, Michael Meaney's group reported that rat mothers differ in how much they lick and groom their pups and how often they nurse in an arched-back posture. Those behaviors changed the offspring's methylation at a promoter of the glucocorticoid receptor gene in the hippocampus, a brain region involved in the stress response. The differences emerged during the first week of life, were reversed when pups were swapped to another mother, and persisted into adulthood. Pups of low-nurturing mothers carried more methylation at that promoter. Infusing a drug that blocks histone deacetylase erased the group differences and altered stress responses, which supports a causal link (Weaver et al., 2004). To see why the stress hormone matters here, read our explainer on cortisol and the stress response.

A small 2018 study in Pediatrics looked for a human echo. In 42 healthy five-month-old infants, breastfeeding (used as a rough proxy for maternal care) was associated with lower methylation at a glucocorticoid receptor promoter and with lower cortisol reactivity (Lester et al., 2018). It is suggestive, but 42 infants and a proxy for care leave plenty of room for other explanations.

The Dutch Hunger Winter

During the winter of 1944 to 1945, famine struck the western Netherlands. Researchers later followed people who had been in the womb at the time. About six decades later, those prenatally exposed to famine had less methylation of the imprinted gene IGF2 than their unexposed same-sex siblings, and the effect was specific to exposure around conception (Heijmans et al., 2008). It was the first human evidence that early-life conditions can leave a lasting epigenetic mark.

The health findings are broader. In a cohort of 2,414 people born around the famine, exposure at any stage of pregnancy was associated with glucose intolerance, and exposure in early gestation was associated with coronary heart disease, a more atherogenic lipid profile, obesity, and, in women, breast cancer (Roseboom, de Rooij, and Painter, 2006). These are associations from a natural experiment. Epigenetics is one candidate mechanism, not a proven one.

Six months of exercise

In 2013, Tina Rönn and colleagues took fat tissue samples from 23 healthy, previously inactive men before and after six months of supervised exercise. About 17,975 CpG sites in 7,663 genes changed their methylation. Roughly a third of the regions with altered methylation also showed differences in gene activity, including genes tied to obesity and type 2 diabetes risk (Rönn et al., 2013, PLoS Genetics). The finding fits what researchers suspect about how exercise works on the body, but the study had no control group and tested one tissue in one small cohort.

Study	Subject	Exposure	Epigenetic finding	Main limit
Waterland and Jirtle, 2003	Viable yellow agouti mice	Maternal methyl-donor diet	Increased methylation at the agouti locus	Unusual mouse strain; not a human result
Weaver et al., 2004	Rats	High vs low maternal licking and grooming	Different hippocampal GR promoter methylation	Animal model; one brain region
Heijmans et al., 2008	Humans	Prenatal famine, 1944 to 1945	Lower IGF2 methylation six decades later	Observational; one gene studied
Rönn et al., 2013	23 healthy men	Six months of exercise	Methylation changes at about 17,975 CpG sites	Small, no control group
Joehanes et al., 2016	15,907 blood samples	Smoking	2,623 CpG sites differ; 185 persist after quitting	Association, blood only

Is nutrition a methylation switch?

The methyl groups attached to DNA come from the diet, which is why the agouti experiment drew so much attention. Folate, vitamin B12, choline, and betaine all feed the cellular chemistry that supplies methyl groups. That does not mean that eating more of them tunes your genome in a useful way.

Consider folic acid. The US Centers for Disease Control and Prevention says that two large trials in the 1990s showed folic acid taken before and during early pregnancy helped lower the chance of a baby having a neural tube defect, and that the US Public Health Service recommends 400 micrograms daily for people who could become pregnant, starting at least one month before conception (CDC). The trial evidence concerns outcomes. Whether DNA methylation explains the benefit is a separate question, and this article does not claim that it does.

The honest summary is that nutrients are inputs to methylation chemistry, and extreme conditions such as famine have measurable epigenetic correlates. Between those poles, the evidence for specific diets rewriting specific marks in ways that improve health is thin. Anyone with questions about supplements should ask a clinician.

Resetting the genome: what happens in embryos and germ cells

If the environment can write on the epigenome, why aren't all those notes passed down? Because most of them are erased.

A 2001 review in Science described genome-wide reprogramming of DNA methylation in germ cells and in the embryo before implantation, a process that generates cells with broad developmental potential and links to genomic imprinting and cloning (Reik, Dean, and Walter, 2001). In mice, the cells that will become sperm and eggs lose most of their methylation between days 10.5 and 12.5 of gestation, though some repetitive sequences are only partly erased (Hajkova et al., 2002, Mechanisms of Development).

Think of a classroom whiteboard wiped clean between lessons, with a few stubborn marker stains. The wipe is the reason a fertilized egg can build every cell type, and it explains why a single skin cell can be coaxed back into a stem cell. Our articles on how stem cells know what to become and the Yamanaka iPSC experiments cover that reprogramming, and the Gurdon frog cloning experiment showed decades earlier that a mature cell's nucleus can be reset.

A small set of genes, called imprinted genes, escape the wipe. For them, the parent of origin leaves a methylation mark that the embryo keeps, so only one copy is active. Errors at imprinted regions cause disorders such as Angelman and Beckwith-Wiedemann syndromes. Imprinting is the clearest proof that some epigenetic information does cross generations. The question is how much else does.

Epigenetic clocks: what they measure and what they do not

One practical use of methylation data is estimating age. In 2013, Steve Horvath built a predictor from about 8,000 samples spanning 51 healthy tissues and cell types, using 353 CpG sites. The estimated methylation age is close to zero for embryonic and induced pluripotent stem cells, and all 20 cancer types he analyzed showed significant age acceleration, averaging about 36 years (Horvath, 2013).

Later clocks are tuned to predict health rather than calendar age. DunedinPACE, published in eLife in 2022, was built from 19 indicators of organ-system integrity tracked over two decades in the Dunedin Study. It condenses them into one blood test of how fast a person is aging, and the authors report links to illness, disability, and mortality, plus faster aging in young adults who experienced childhood adversity (Belsky et al., 2022). For how this connects to the wider biology of aging, see why we grow old and what telomeres do.

Can a clock be turned back? A 2021 pilot randomized trial in Aging enrolled 43 healthy men aged 50 to 72 and compared an eight-week program of diet, sleep, exercise, and relaxation guidance plus probiotic and phytonutrient supplements against no intervention. Epigenetic age measured by the Horvath clock was 3.23 years lower in the treatment group than in controls (p = 0.018). The within-group drop was 1.96 years, which did not reach conventional significance (p = 0.066) (Fitzgerald et al., 2021). The study was small, limited to men, used saliva samples, and the authors say larger and longer trials are needed. A lower clock reading is not proof of a longer or healthier life. Our broader review of longevity interventions sorts the better-supported claims from the weaker ones.

Epigenetic drugs and the link to cancer

Cancer shows what happens when the epigenome goes wrong. Reviews of the histone code describe how faulty histone modifications and DNA methylation contribute to multistep cancer development (Hake, Xiao, and Allis, 2004, British Journal of Cancer). Genes that should be silent can switch on, and protective genes can be switched off, without any change to the DNA letters. Our article on what causes cancer puts this in context with mutations and other drivers.

The first drugs aimed at the epigenome are already in use. Azacitidine was approved by the US Food and Drug Administration in May 2004 for myelodysplastic syndromes, a group of bone marrow disorders, and was the first drug approved for them; the approval rested on a randomized trial of 191 patients against supportive care. Decitabine followed in May 2006. Both are DNA methyltransferase inhibitors: they are built into dividing cells' DNA and trap or deplete the enzyme DNMT1 so that methylation marks are not copied. They show that changing the epigenome can be a treatment strategy, not just a research topic.

Can a parent's experience reach their children?

This is the question that draws the most attention and the most disagreement. It helps to get the vocabulary straight first. When a pregnant animal is exposed to something, the embryo and its developing germ cells are exposed too. An environmental health review explains the convention: effects that extend to the second generation (F2) are called multigenerational, and only effects that reach the third generation (F3) count as transgenerational, because F3 is "the first generation to be totally unexposed to the original agent" (Schmidt, 2013).

What animals show. In 2005, Matthew Anway and Michael Skinner reported that briefly exposing pregnant rats to the fungicide vinclozolin or the pesticide methoxychlor lowered sperm counts and viability in male offspring, and the effect persisted through the male line from F1 to F4, alongside altered germ-line methylation (Anway et al., 2005, Science). The result has been challenged. A 2009 Pacific Standard report described replication attempts at two chemical companies that did not find the transgenerational effect, differences in dosing route and rat strain that Skinner thought could explain the gap, and the reporter's caution that the companies' conflicts of interest could not be dismissed (Brown, 2009).

A 2014 mouse study by Brian Dias and Kerry Ressler is easier to follow. Male mice learned to fear the smell of acetophenone. Their offspring and grand-offspring showed heightened sensitivity to that odor but not to others, had an enlarged neural representation for the receptor that detects it, and the fathers' sperm showed lower methylation in the receptor's gene. In vitro fertilization and cross-fostering experiments pointed to transmission through the gametes (Dias and Ressler, 2014, Nature Neuroscience).

What humans show. The Dutch famine data discussed above concern people who were exposed in the womb, which is direct exposure rather than inheritance. For true transgenerational effects, the bar is higher. Bernhard Horsthemke wrote in Nature Communications in 2018 that transgenerational epigenetic inheritance is documented in plants, nematodes, and fruit flies, but is disputed in mammals and especially humans, and that the main obstacle is that genetic, ecological, and cultural inheritance can confound the observations (Horsthemke, 2018). Edith Heard and Robert Martienssen reached a similar view in Cell in 2014: epigenetic inheritance clearly occurs in some organisms, especially plants, and how far it applies to humans is an open question (Heard and Martienssen, 2014).

A useful picture is pencil notes in the margins of a library book. Some notes may survive a rushed erasing, but if a child's copy has similar notes, it could also be because both copies were annotated by the same teacher. Separating inheritance from shared environment is the hard part.

It also helps to compare this with classic evolution. Charles Darwin's mechanism, described in our article on evolution and natural selection, works through variation in DNA sequence. Epigenetic inheritance, where it exists, is a different channel, usually shorter-lived, and nothing in the current evidence overturns the sequence-based account.

What people get wrong about epigenetics

"Epigenetics means genes don't matter." The sequence sets the range of possible outcomes, and epigenetic marks operate inside it. A person who lacks a disease-linked variant cannot gain that disease risk through methylation alone. For how sequence variation is copied and sometimes changed, see DNA replication and genetics explained.

"Marks you gain from a bad year are permanent." Some are, such as certain imprints. Many are not: in the smoking data, most of the sites that differed in current smokers no longer differed significantly in former smokers. Reversibility depends on the mark, the tissue, and the timing.

"You can optimize your epigenome with the right diet." Studies show correlations and a handful of small trials. They do not yet show that tuning specific marks improves health in the general population.

"Epigenetic age tests tell you how long you will live." Clocks are statistical estimates. They are more robust across large groups than for an individual, and different clocks can give different answers for the same person.

"Epigenetics revives Lamarck." Jean-Baptiste Lamarck proposed that traits acquired during life are inherited as a general rule. The evidence here concerns narrow, specific marks, mostly in model organisms, whose effects fade within a few generations. For humans, the case is unproven.

The takeaway

Epigenetics solves an old puzzle: how one genome gives rise to hundreds of cell types and why identical twins drift apart. A 2005 study of monozygotic twins found that they were epigenetically similar early in life, while older twins showed notable differences in methylation and histone acetylation that affected gene expression (Fraga et al., 2005, PNAS). The marks are real, measurable, and responsive to the environment, from smoking to famine. They are also reversible in many cases, which is both the hope and the reason for caution. The strongest claims, about inherited trauma or diets that rewrite your genome, run well ahead of the evidence. The field is best read as a map of how experience can leave a molecular record, with the question of how far that record travels still open.

FAQ

Q: Does epigenetics mean your genes don't matter? A: No. DNA sequence sets the range of what is possible. Epigenetic marks work within that range by modulating how strongly genes are used. A person without the genetic variants for a certain condition cannot develop it through epigenetic changes alone, while a person with a strong predisposition may see their risk raised or lowered by epigenetic factors.

Q: Can you reverse epigenetic changes? A: Often, to some degree. Cells carry enzymes that add and remove methyl groups and histone tags, and smoking-related methylation differences fade for many sites after quitting. A pilot trial reported a lower epigenetic age after an eight-week lifestyle program, but it was small, included only men, and needs replication. Reversibility varies by mark, tissue, and timing.

Q: Is epigenetics the same as gene expression? A: Related but not identical. Gene expression is whether a gene is being transcribed into RNA and translated into protein at a given moment. Epigenetics usually refers to chemical marks that can persist through cell division and help set expression levels. All epigenetic changes affect expression, but many expression changes are transient and not epigenetic.

Q: Can stress really change your children's biology? A: In rodents, yes in specific experiments: maternal care changed offspring methylation at a stress-receptor gene, and parental odor conditioning in mice was linked to changes in offspring. In humans, exposure during pregnancy can affect a fetus directly, and the Dutch famine data show lasting associations. Evidence for epigenetic transmission across several generations in humans is limited and debated.

Q: What does an epigenetic clock actually measure? A: A clock uses methylation levels at a chosen set of CpG sites, such as the 353 sites in Horvath's 2013 model, to estimate biological age. Newer ones, such as DunedinPACE, estimate the pace of aging. They are statistical tools, most reliable for groups, and a single reading for one person has wide uncertainty.

Q: Are consumer epigenetic tests worth buying? A: The measurements can be technically valid, but the interpretation is the weak point. A score is not a diagnosis, and no large trial shows that acting on a consumer clock result improves health outcomes. A clinician can say whether any test result changes your care.

Sources

• Nurk et al., The complete sequence of a human genome, Science, 2022; GENCODE human statistics; NHGRI genetics glossary: epigenetics.
• Moore, Le, and Fan, DNA methylation and its basic function, Neuropsychopharmacology, 2013.
• Waterland and Jirtle, Molecular and Cellular Biology, 2003; Weaver et al., Nature Neuroscience, 2004 (DOI 10.1038/nn1276); Heijmans et al., PNAS, 2008 (DOI 10.1073/pnas.0806560105); Rönn et al., PLoS Genetics, 2013; Joehanes et al., Circulation: Cardiovascular Genetics, 2016.
• Horvath, Genome Biology, 2013 (DOI 10.1186/gb-2013-14-10-r115); Belsky et al., eLife, 2022 (DOI 10.7554/eLife.73420); Fitzgerald et al., Aging, 2021 (DOI 10.18632/aging.202913).
• Horsthemke, Nature Communications, 2018; Heard and Martienssen, Cell, 2014; Schmidt, Environmental Health Perspectives, 2013; Anway et al., Science, 2005; Dias and Ressler, Nature Neuroscience, 2014.
• CDC, Folic acid.

---

Read next:

• DNA Replication and Genetics Explained
• Epigenetics: How Genes Are Turned On and Off Explained
• Cloning Explained: How Dolly the Sheep Was Made