Epigenetics: histone modification, DNA methylation, X-inactivation, and imprinting

Intuition Beginner

Every cell in your body carries the same DNA — the same set of roughly 20,000 genes. Yet a neuron looks and behaves nothing like a liver cell or a skin cell. The difference is not in which genes are present, but in which genes are turned on or off. Epigenetics is the study of heritable changes in gene expression that do not alter the DNA sequence itself.

The genome is wrapped around spool-like proteins called histones. Together, DNA and histones form chromatin. Chemical tags attached to DNA and to histones control how tightly the chromatin is packed. Tightly packed chromatin (heterochromatin) silences genes because the transcription machinery cannot access them. Loosely packed chromatin (euchromatin) leaves genes accessible for transcription.

Two major types of chemical tags drive epigenetic regulation. DNA methylation attaches a methyl group directly to cytosine bases, generally silencing nearby genes. Histone modifications attach chemical groups (acetyl, methyl, phosphate, ubiquitin) to histone tails, either loosening or tightening chromatin depending on the specific modification and position.

These marks are copied when a cell divides, so daughter cells inherit the same gene expression pattern. This is how a liver cell divides into two liver cells — the epigenetic marks that keep liver-specific genes on and brain-specific genes off are maintained through cell division.

Visual Beginner

Think of your genome as a vast library. Every cell contains the same books (genes). Epigenetic marks are like colour-coded stickers on the shelves and book covers. A red sticker (DNA methylation) means "do not read this book." A green sticker (histone acetylation) means "this book is open for reading." When the library is copied to build a new cell, all the stickers are duplicated too, so the new librarian knows which books to read and which to ignore.

Worked example Beginner

Consider the Igf2/H19 locus on human chromosome 11. This region contains two genes: Igf2 (a growth factor) and H19 (a non-coding RNA). A short DNA segment called the imprinting control region (ICR) sits between them and is methylated on the chromosome inherited from the father but unmethylated on the chromosome inherited from the mother.

When the ICR is methylated (paternal chromosome), a boundary element called CTCF cannot bind. Without CTCF blocking the path, the downstream enhancer activates Igf2 transcription, and H19 is silenced. When the ICR is unmethylated (maternal chromosome), CTCF binds and blocks the enhancer from reaching Igf2, so Igf2 is off and H19 is transcribed instead.

The result: only the paternal copy of Igf2 is expressed, and only the maternal copy of H19 is expressed. This is genomic imprinting — the parent of origin determines which gene is active, even though both copies have identical DNA sequences.

Check your understanding Beginner

Formal definition Intermediate+

Epigenetic regulation encompasses heritable modifications to chromatin that alter gene expression without changing the underlying DNA sequence. The two principal mechanisms are covalent modification of histone proteins and methylation of cytosine bases in DNA.

Histone modifications

The core histone proteins (H2A, H2B, H3, H4) form an octamer around which 147 bp of DNA is wrapped to form a nucleosome. Each histone has an unstructured N-terminal tail (20-35 residues) that protrudes from the nucleosome core and is subject to diverse post-translational modifications (PTMs).

The major histone tail modifications include:

Acetylation of lysine residues (Kac) is catalysed by histone acetyltransferases (HATs) and removed by histone deacetylases (HDACs). Lysine acetylation neutralises the positive charge, weakening histone-DNA electrostatic interactions and promoting an open chromatin conformation. HATs are classified into three major families: GNAT (GCN5, PCAF), MYST (MOF, TIP60), and p300/CBP. HDACs comprise four classes: class I (HDAC1, 2, 3, 8; ubiquitously expressed, nuclear), class II (HDAC4, 5, 6, 7, 9, 10; tissue-specific, shuttle between nucleus and cytoplasm), class III (sirtuins SIRT1-7; NAD+-dependent), and class IV (HDAC11).

Methylation of lysine and arginine residues is catalysed by histone methyltransferases (HMTs) and removed by histone demethylases. Lysine can be mono-, di-, or tri-methylated, each with distinct functional consequences. The SET-domain HMTs (SUV39H1, EZH2, SET7/9, G9a) deposit methyl marks. Demethylases include LSD1 (FAD-dependent, removes mono- and di-methylation from H3K4 and H3K9) and JmjC-domain enzymes (JMJD2, JARID1; $\text{Fe(II)}$ and $\alpha$-ketoglutarate dependent, can remove all three methylation states).

The functional outcome of histone methylation depends on the specific residue and degree of methylation:

• H3K4me3 (trimethylation of H3 lysine 4): marks active gene promoters.
• H3K36me3: marks actively transcribed gene bodies.
• H3K27me3 (deposited by Polycomb Repressive Complex 2, PRC2): marks developmentally silenced genes.
• H3K9me3 (deposited by SUV39H1): marks constitutive heterochromatin (pericentromeric regions, transposable elements).

Phosphorylation of serine and threonine residues (H3S10ph, H3S28ph) is associated with chromosome condensation during mitosis and with rapid transcriptional activation at immediate-early genes. Aurora B kinase phosphorylates H3S10 during mitosis, while MSK1/2 phosphorylate H3S10 at inducible promoters.

Ubiquitination of histone H2A at K119 (H2AK119ub1, deposited by PRC1) and H2B at K120 (H2BK120ub1, deposited by RNF20/40) has distinct regulatory roles. H2AK119ub1 is associated with transcriptional repression by Polycomb, while H2BK120ub1 facilitates H3K4 methylation and H3K79 methylation during active transcription (trans-histone cross-talk).

The histone code hypothesis

The histone code hypothesis (Strahl and Allis, 2000; Jenuwein and Allis, 2001) proposes that specific combinations of histone modifications — rather than individual marks — constitute a code read by effector proteins that dictate downstream chromatin states and transcriptional outcomes.

Effector proteins recognise specific modifications through dedicated domains:

• Bromodomains bind acetylated lysines. BET family proteins (BRD2, BRD3, BRD4) contain tandem bromodomains and recruit transcriptional co-activators (P-TEFb) to acetylated promoters.
• Chromodomains bind methylated lysines. HP1 binds H3K9me2/3 via its chromodomain, propagating heterochromatin. Polycomb (CBX) proteins bind H3K27me3.
• PHD fingers bind methylated H3K4. ING proteins and BPTF read H3K4me3 at active promoters.
• Tudor domains bind methylated arginines and lysines (e.g., JMJD2A Tudor domain binds H4K20me2/3).
• WD40 repeat domains (WDR5) bind H3K4me2 and participate in HMT complex assembly.

The combinatorial nature of the code means that different modification patterns on the same histone tail can recruit distinct reader proteins, producing opposing transcriptional outcomes. A single nucleosome can carry dozens of modifications simultaneously, and the information content scales with the number of modifiable residues.

Writers, erasers, and readers

Epigenetic regulation can be conceptualised through three functional classes of proteins:

Writers deposit marks: HATs (acetylation), HMTs (methylation), kinases (phosphorylation), ubiquitin ligases (ubiquitination), DNMTs (DNA methylation).

Erasers remove marks: HDACs (deacetylation), histone demethylases (LSD1, JmjC), phosphatases (dephosphorylation), deubiquitinases (DUBs), TET enzymes (DNA demethylation).

Readers recognise and bind marks, translating the modification into a functional outcome: bromodomains (acetylation), chromodomains (methylation), PHD fingers (methylation), MBD proteins (methylated DNA), HP1 (H3K9me3).

This writer-eraser-reader framework allows dynamic, reversible regulation of chromatin state. The same genomic locus can transition between active and repressed states as the balance of writer and eraser activity shifts during development or in response to signalling.

DNA methylation

In vertebrates, DNA methylation occurs predominantly at CpG dinucleotides, where a methyl group is attached to the 5-carbon position of cytosine, producing 5-methylcytosine (5mC). Approximately 70-80% of CpG sites in the human genome are methylated, with the unmethylated minority concentrated in CpG islands — regions of at least 200 bp with a GC content above 50% and an observed-to-expected CpG ratio above 0.6.

CpG islands are found at the promoters of approximately 70% of human genes (particularly housekeeping and developmental regulator genes) and are generally maintained in an unmethylated state, allowing gene expression. Aberrant methylation of CpG island promoters silences the associated gene and is a hallmark of cancer.

Three catalytically active DNA methyltransferases operate in mammals:

• DNMT3A and DNMT3B are de novo methyltransferases that establish new methylation patterns during embryonic development and gametogenesis. They methylate previously unmethylated CpG sites without requiring a hemi-methylated template. DNMT3L is catalytically inactive but stimulates DNMT3A/3B activity and is essential for establishing methylation imprints in oocytes.
• DNMT1 is the maintenance methyltransferase that copies pre-existing methylation patterns to the newly synthesised strand during DNA replication. DNMT1 is recruited to replication foci by UHRF1 (also called NP95 or ICBP90), which specifically recognises hemi-methylated CpG sites (methylated on the parental strand, unmethylated on the daughter strand) via its SRA domain. This ensures faithful propagation of methylation through cell divisions.

Active DNA demethylation: TET enzymes

DNA demethylation occurs through both passive (replication-dependent dilution when DNMT1 is excluded) and active (enzyme-catalysed) mechanisms. The TET (Ten-Eleven Translocation) family enzymes (TET1, TET2, TET3) catalyse iterative oxidation of 5mC:

1. 5mC is oxidised to 5-hydroxymethylcytosine (5hmC).
2. 5hmC is oxidised to 5-formylcytosine (5fC).
3. 5fC is oxidised to 5-carboxylcytosine (5caC).

5fC and 5caC are recognised and excised by thymine DNA glycosylase (TDG), and the resulting abasic site is repaired by the base excision repair (BER) pathway, restoring an unmodified cytosine. Additionally, 5hmC is not recognised by DNMT1 as a substrate for maintenance methylation, so passive dilution of 5hmC through replication rounds also contributes to demethylation.

5hmC is particularly abundant in the brain (0.5-1% of all cytosines in cortical neurons) and in embryonic stem cells. Loss of 5hmC is associated with cancer: TET2 is among the most frequently mutated genes in myeloid malignancies (acute myeloid leukaemia, myelodysplastic syndrome), and TET2 mutations produce a genome-wide hypermethylation phenotype that silences differentiation-promoting genes.

Chromatin remodelling complexes

ATP-dependent chromatin remodelling complexes use the energy of ATP hydrolysis to slide, eject, or restructure nucleosomes, regulating DNA accessibility. Four major families are defined by their ATPase subunit:

SWI/SNF family (BAF/PBAF in mammals): BRG1 or BRM ATPase. Slides nucleosomes and ejects histones to open chromatin at promoters and enhancers. Frequently mutated in cancer (SMARCB1 loss in rhabdoid tumours; ARID1A mutations in ovarian clear cell carcinoma).

ISWI family (SNF2H, SNF2L): promotes regular nucleosome spacing and phasing. The RSF complex establishes nucleosome arrays at newly replicated DNA; the NuRF complex slides nucleosomes to regulate transcription.

CHD family (CHD1-9): contains chromodomains in addition to the ATPase. CHD1 deposits H3.3-containing nucleosomes at gene bodies. CHD4 (Mi-2) is the ATPase of the NuRD repressive complex.

INO80 family (INO80, SWR1): exchanges histone variants. SWR1/SRCAP replaces canonical H2A with H2A.Z in nucleosomes flanking transcription start sites. INO80 slides nucleosomes and facilitates replication fork progression.

Key mechanism Intermediate+

Maintenance of DNA methylation through cell division and the interplay with histone marks.

The faithful propagation of DNA methylation patterns through cell division is the molecular basis of epigenetic memory. When the replication fork passes a methylated CpG site, the parental strand retains its methyl group but the newly synthesised daughter strand is unmethylated, creating a hemi-methylated CpG duplex. The maintenance methyltransferase DNMT1, guided by UHRF1, recognises these hemi-methylated sites and methylates the cytosine on the daughter strand, restoring the fully methylated state.

UHRF1 (Ubiquitin-like containing PHD and RING finger domains 1) is the central coordinator of maintenance methylation. It contains multiple domains that read both DNA methylation and histone modification states:

• The SRA domain flips out the 5mC base from the parental strand and recognises the hemi-methylated CpG.
• The TTD (Tudor domain) binds H3K9me3.
• The PHD finger binds unmethylated H3K4.
• The RING finger has E3 ubiquitin ligase activity that ubiquitinates histone H3 at lysines 18 and 23.

The ubiquitinated H3 tail is then recognised by DNMT1, which is recruited to the replication fork. This dual readout — DNA methylation on the parental strand and H3K9me3 on the local nucleosome — ensures that DNMT1 acts only at appropriate genomic locations. The coupling of DNA methylation maintenance to H3K9me3 reading creates a self-reinforcing loop: H3K9me3 recruits DNMT3A/3B (via their ADD domains interacting with HP1), which deposit DNA methylation; DNA methylation recruits MBD proteins and the NuRD complex, which includes the HDAC1/2 that remove acetyl groups, maintaining a substrate for SUV39H1 to methylate H3K9, perpetuating the heterochromatic state.

This positive feedback loop explains the remarkable stability of epigenetic silencing. Once a locus is packaged into heterochromatin with both H3K9me3 and DNA methylation, the two marks maintain each other across cell divisions. Breaking this loop requires simultaneous removal of both marks — for example, during epigenetic reprogramming in the early embryo or during induced pluripotent stem cell generation.

The SWI/SNF complex opposes heterochromatin formation by ejecting nucleosomes at promoters, allowing transcription factor binding. Mutations in SWI/SNF subunits (ARID1A, SMARCB1, SMARCA4) cause cancer by permitting aberrant heterochromatin spreading and silencing of tumour suppressors, or by preventing the opening of chromatin at differentiation genes.

Exercises Intermediate+

X-inactivation, imprinting, reprogramming, and disease Master

X-chromosome inactivation

In placental mammals, females (XX) silence one of their two X chromosomes to equalise X-linked gene dosage with males (XY). This process, called X-chromosome inactivation (XCI), converts one X into a transcriptionally inert Barr body — a densely staining heterochromatic mass visible at the nuclear periphery.

The X-inactivation centre (Xic). XCI is initiated from a locus on the X chromosome called the X-inactivation centre, which produces the long non-coding RNA XIST (X-inactive specific transcript). XIST RNA is expressed exclusively from the X chromosome destined for inactivation (the Xi) and coats it in cis, spreading along the chromosome from the Xic in both directions.

Counting mechanism. The cell counts X chromosomes relative to autosomal sets. In a diploid cell (two autosome sets), one X remains active and all additional X chromosomes are inactivated. The counting mechanism involves the Rnf12 gene, which encodes an X-linked RING finger protein that acts as a dose-dependent XIST activator. With two X chromosomes, RNF12 protein exceeds a threshold concentration that triggers XIST upregulation on one X. The second X then produces a blocking factor (potentially involving Tsix, an antisense transcript that overlaps XIST and represses it) that prevents its own inactivation. The precise molecular identity of the blocking factor remains debated.

Spreading of inactivation. Once XIST RNA coats the X chromosome in cis, it recruits silencing factors in a defined sequence:

1. SPEN (SHARP): recruited directly by XIST RNA via its RNA-binding domains. SPEN is a transcriptional repressor that recruits HDAC3, rapidly removing histone acetylation from the coated X.

2. PRC2 (Polycomb Repressive Complex 2): recruited by XIST RNA, deposits H3K27me3 across the inactive X. PRC1 is subsequently recruited by H3K27me3 and deposits H2AK119ub1.

3. SMCHD1: a noncanonical SMC protein that contributes to DNA methylation of the inactive X, providing long-term epigenetic memory.

4. MacroH2A deposition: the histone variant macroH2A1 is enriched on the Xi, contributing to its compact structure.

5. DNA methylation: CpG islands on the Xi become hypermethylated, providing the most stable layer of silencing. DNA methylation locks in XCI after the initial XIST-dependent silencing is established.

Escape genes. Not all genes on the Xi are silenced. Approximately 15-25% of X-linked genes "escape" inactivation and are expressed from both X chromosomes. Escape genes are enriched in the pseudoautosomal regions (PAR1 and PAR2) but are also scattered across the X chromosome. Escape genes tend to be located in regions with open chromatin marks (H3K4me3, H3K27ac) and reduced XIST coating. The evolutionary implication is that some genes benefit from a double dose in females, and the inactivation machinery has been selectively permeable at these loci.

Timing and randomness. In mouse embryos, XCI occurs in two waves. The first wave (imprinted XCI) silences the paternal X in the pre-implantation embryo (4-8 cell stage). This imprinted XCI is then reversed in the inner cell mass, and a second wave (random XCI) silences either the maternal or paternal X stochastically in each cell. Once established, the choice is heritable: all descendants of that cell inactivate the same X. In humans, random XCI is established around the blastocyst stage, and the existence of a prior imprinted wave is debated.

The clonal heritability of random XCI produces the mosaic pattern visible in female heterozygotes for X-linked traits. Calico cats, for example, display orange and black patches because different pigment-producing cell lineages have inactivated different X chromosomes carrying different pigment alleles.

Genomic imprinting

Genomic imprinting is parent-of-origin-specific gene expression: certain genes are expressed exclusively from the maternal or paternal allele, while the other allele is epigenetically silenced. Approximately 100-200 imprinted genes are known in the human genome, many of them involved in growth, development, and metabolism.

Imprinting is established during gametogenesis through differential DNA methylation at imprinting control regions (ICRs). In oogenesis, DNMT3A (assisted by DNMT3L) deposits methylation at maternal ICRs during the growing oocyte stage. In spermatogenesis, DNMT3A and DNMT3L deposit methylation at paternal ICRs. After fertilisation, the genome undergoes global demethylation, but ICR methylation is protected from erasure by ZFP57 and its cofactor KAP1/TRIM28, which recruit SETDB1 and maintain H3K9me3 at ICRs, protecting them from TET-mediated demethylation.

The Igf2/H19 locus. The best-characterised imprinted locus sits on chromosome 11p15.5 in humans. The ICR between Igf2 and H19 is methylated on the paternal allele and unmethylated on the maternal allele. On the unmethylated maternal allele, the CTCF protein binds the ICR and forms a boundary that blocks downstream enhancers from activating Igf2, directing them instead to activate H19. On the methylated paternal allele, CTCF cannot bind (methylation at its recognition site abolishes binding), and the enhancers activate Igf2. The result: Igf2 is expressed exclusively from the paternal allele, and H19 from the maternal allele.

Prader-Willi and Angelman syndromes. The chromosome 15q11-q13 region contains a cluster of imprinted genes whose misregulation produces two distinct neurodevelopmental disorders:

• Prader-Willi syndrome (PWS): caused by loss of paternally expressed genes (SNURF-SNRPN, necdin, MAGEL2, and multiple snoRNAs). This occurs through paternal deletion (70%), maternal uniparental disomy (UPD, 25%), or ICR methylation defects (5%). Features include neonatal hypotonia, hyperphagia, obesity, intellectual disability, and hypogonadism.

• Angelman syndrome (AS): caused by loss of the maternally expressed UBE3A gene (a ubiquitin ligase). This occurs through maternal deletion (70%), paternal UPD (5%), ICR methylation defects (5%), or direct UBE3A mutation (10%). Features include severe intellectual disability, ataxia, seizures, and a characteristic happy demeanour with frequent laughter.

The same chromosomal region produces two entirely different syndromes depending on which parental contribution is lost, because different genes in the cluster are imprinted in opposite directions.

Epigenetic reprogramming

Two waves of genome-wide epigenetic reprogramming occur during mammalian development:

Pre-implantation reprogramming. After fertilisation, the paternal genome is rapidly and actively demethylated (by TET3 oxidation) while the maternal genome is passively demethylated through replication in the absence of maintenance. By the blastocyst stage, global methylation has dropped from ~80% to ~20%. ICR methylation is protected by ZFP57/KAP1. This reprogramming creates a nearly blank epigenetic slate that allows the early embryo to establish new cell-type-specific methylation patterns during differentiation.

Primordial germ cell (PGC) reprogramming. PGCs undergo a second wave of global demethylation as they migrate to the genital ridges (embryonic days 10.5-13.5 in mouse). This erasure removes somatic epigenetic marks, preparing the germ cell to establish sex-specific imprints. PGC demethylation involves both TET-mediated active demethylation and replication-dependent passive dilution. Transposable elements resist complete demethylation through piRNA-directed silencing (see 17.06.03), preventing TE reactivation during this vulnerable window.

Somatic cell nuclear transfer (SCNT). When a differentiated somatic nucleus is transplanted into an enucleated oocyte, the oocyte cytoplasm reprograms the somatic epigenome to a totipotent state. SCNT efficiency is low (1-5% live births) because residual somatic epigenetic marks (particularly DNA methylation and H3K9me3) resist erasure. The histone demethylase KDM4D (which removes H3K9me3) improves SCNT efficiency when injected into the reconstructed embryo, demonstrating that H3K9me3 is a major barrier to epigenetic reprogramming.

Induced pluripotent stem cells (iPSCs). Yamanaka and Takahashi (2006) showed that forced expression of four transcription factors (OCT4, SOX2, KLF4, c-MYC — the "Yamanaka factors") reprograms differentiated somatic cells to pluripotent stem cells. The reprogramming process involves sequential epigenetic changes: initial chromatin opening at pluripotency enhancers (pioneer factor activity of OCT4/SOX2), followed by DNA demethylation at pluripotency gene promoters, and finally establishment of the H3K4me3/H3K27me3 bivalent state at developmental regulators. Reprogramming is stochastic and inefficient (0.01-1%), with H3K9 methylation and DNA methylation at pluripotency loci as major epigenetic barriers.

Cancer epigenetics

Cancer cells display widespread epigenetic dysregulation characterised by two paradoxical features:

Global hypomethylation: DNA methylation is reduced by 20-60% across the cancer genome compared to normal tissue. Hypomethylation activates transposable elements (promoting genomic instability through insertional mutagenesis), activates oncogenes, and causes loss of imprinting (LOI). LOI at the Igf2 locus (biallelic expression) is observed in Wilms tumour, colorectal cancer, and other malignancies, producing excess growth factor signalling.

Focal hypermethylation of CpG island promoters: Despite global hypomethylation, specific CpG island promoters become aberrantly hypermethylated, silencing tumour suppressor genes without mutation. The tumour suppressors most frequently silenced by promoter hypermethylation include CDKN2A (p16, cell cycle arrest), MLH1 (DNA mismatch repair), BRCA1 (DNA repair), RASSF1A (RAS signalling), and APC (Wnt pathway). This "CpG island methylator phenotype" (CIMP) defines distinct molecular subtypes in glioblastoma, colorectal cancer, and endometrial cancer.

Mutations in epigenetic regulators are among the most frequent events in cancer:

• DNMT3A mutations in acute myeloid leukaemia (AML) alter de novo methylation patterns and block differentiation.
• TET2 mutations in myeloid malignancies prevent active demethylation, producing hypermethylation.
• IDH1/IDH2 mutations produce the oncometabolite 2-hydroxyglutarate (2-HG), which competitively inhibits TET enzymes and JmjC demethylases, producing a combined hypermethylation phenotype.
• EZH2 (PRC2 catalytic subunit) gain-of-function mutations in follicular lymphoma and diffuse large B-cell lymphoma increase H3K27me3, silencing differentiation genes.
• ARID1A (SWI/SNF subunit) loss-of-function mutations in ovarian clear cell carcinoma and gastric cancer impair chromatin opening at tumour suppressor promoters.

Epigenetic clocks

Epigenetic clocks are mathematical models that estimate biological age from DNA methylation patterns at specific CpG sites. The Horvath clock (2013) uses methylation at 353 CpG sites to estimate age across multiple tissues, with a median error of approximately 3.6 years. The clock measures "epigenetic age," which can diverge from chronological age in disease states:

• Accelerated epigenetic ageing is observed in cancer tissue, Down syndrome, and obesity.
• Decelerated epigenetic ageing is associated with caloric restriction and supercentenarian status.
• The clock is reset to zero during reprogramming to iPSCs, demonstrating that epigenetic age is reversible.

The Horvath clock was derived by training an elastic net regression model on methylation data from thousands of samples spanning multiple tissues, regressing chronological age on methylation beta values. The existence of a tissue-conserved methylation-based age estimator suggests that a defined subset of CpG sites undergoes programmed, directional methylation change with age, though the mechanistic basis remains debated.

Environmental epigenetics

Environmental exposures can alter epigenetic marks, and some of these changes persist across the lifespan. The evidence for transgenerational epigenetic inheritance in humans remains controversial, but several well-documented examples exist:

The Dutch Hunger Winter (1944-1945) cohort shows that individuals exposed to famine during early gestation have altered DNA methylation at metabolic and growth-related genes (including Igf2) six decades later, with increased rates of obesity, type 2 diabetes, and cardiovascular disease. The methylation changes are consistent with a programmed response to nutritional deprivation that becomes maladaptive in a nutrient-rich environment.

The Agouti viable yellow ($A^{vy}$) mouse model demonstrates that maternal diet (methyl donor supplementation with folic acid, vitamin B12, choline, and betaine) shifts the methylation state of a retrotransposon upstream of the Agouti gene, producing a spectrum of coat colours from yellow (hypomethylated, agouti overexpressed) to brown (hypermethylated, agouti silenced). The methylation state is established early in development and is heritable through cell division, linking maternal nutrition to offspring phenotype through epigenetic modification.

Whether environmental epigenetic changes are transmitted across generations through the germline (true transgenerational epigenetic inheritance) remains one of the most debated questions in the field. The two waves of epigenetic reprogramming (pre-implantation and PGC) erase most methylation, and rigorous transgenerational studies in humans are confounded by shared environment and genetics. In C. elegans (which lacks DNA methylation), small RNA-mediated transgenerational silencing is well established. In mammals, the evidence is strongest for intergenerational effects (exposure of the pregnant mother affects the F1 foetus and F2 germ cells directly) rather than true transgenerational effects (F3 and beyond).

Connections Master

• Mutation and repair 17.06.01. Epigenetic silencing is a source of phenotypic variation without DNA sequence change. Aberrant DNA methylation can mimic mutation by silencing genes. The base excision repair pathway participates in active DNA demethylation through TET/TDG-mediated excision of oxidised cytosine derivatives.

• Transposable elements 17.06.03 pending. DNA methylation and H3K9me3 are the primary mechanisms for silencing transposable elements. The piRNA pathway connects small RNA biology to DNA methylation at TE loci. Failure of epigenetic TE silencing during reprogramming leads to TE reactivation and genomic instability.

• Transcription 17.05.02. Chromatin accessibility determines whether transcription factors and RNA polymerase can bind promoters. Histone acetylation opens chromatin; DNA methylation recruits repressive complexes that close it. The histone code integrates signalling inputs to regulate transcription.

• DNA replication 17.05.01. DNMT1 maintenance methylation is coupled to the replication fork through UHRF1. Nucleosome assembly behind the fork re-establishes chromatin structure. Replication-coupled dilution of 5hmC contributes to passive DNA demethylation.

• Cell cycle and cancer 17.08.01. Epigenetic silencing of CDKN2A (p16) removes a cell cycle checkpoint. Global hypomethylation in cancer causes chromosomal instability. Mutations in DNMT3A, TET2, IDH1/2, and EZH2 are among the most frequent driver mutations in haematological malignancies.

• Immunology 17.10.01. AID/APOBEC cytidine deaminases in B cells contribute to DNA demethylation. Epigenetic regulation of immune genes determines tolerogenic versus immunogenic states. RAG1/2-mediated V(D)J recombination is regulated by chromatin accessibility.

• Developmental biology 17.03.01. Cell identity is maintained by epigenetic marks. The bivalent state (H3K4me3 + H3K27me3) at developmental regulators in embryonic stem cells keeps lineage genes poised for activation or repression. XCI and imprinting are developmental epigenetic processes essential for normal embryogenesis.

Historical notes Master

The term "epigenetics" was coined by Conrad Waddington in 1942 to describe the study of "causal interactions between genes and their products which bring the phenotype into being." Waddington's epigenetic landscape — a metaphorical model in which a ball (cell) rolls down a surface of valleys (developmental pathways) shaped by underlying gene interactions — provided the conceptual foundation for the field, though it predated the molecular understanding of epigenetic mechanisms.

X-chromosome inactivation was proposed by Mary Lyon in 1961 based on her observations of coat colour mosaicism in female mice heterozygous for X-linked coat colour mutations. She hypothesised that one X chromosome was randomly inactivated in each somatic cell early in development and that the inactivation was stably inherited. The Barr body (first observed by Murray Barr and E.G. Bertram in 1949 as a densely staining chromatin body in female cat neurons) was subsequently identified as the inactive X.

The XIST gene was cloned in 1991 by Brown, Ballabio, and colleagues, who identified it as an X-linked gene expressed exclusively from the inactive X. The discovery that XIST produced a long non-coding RNA that coated the inactive X in cis (Brown et al., 1992; Clemson et al., 1996) established the paradigm of lncRNA-mediated chromatin regulation.

DNA methylation was first detected in mammalian DNA by Hotchkiss in 1948, who observed a minor base (now known to be 5mC) in hydrolysed DNA from calf thymus. The functional significance was not appreciated until Holliday and Pugh (1975) and Riggs (1975) independently proposed that DNA methylation at CpG sites could serve as a heritable epigenetic mark, with maintenance methylation propagating the pattern through cell division. The first mammalian DNA methyltransferase (DNMT1) was purified by Bestor, Laudano, and Mattaliano in 1988, and the de novo methyltransferases DNMT3A and DNMT3B were cloned by Okano, Bell, and Li in 1999.

Histone modifications were first described by Allfrey, Faulkner, and Mirsky in 1964, who demonstrated that histone acetylation correlated with RNA synthesis. The histone code hypothesis was formally articulated by Strahl and Allis in 2000, proposing that specific combinations of histone marks constitute a readable code. The first histone methyltransferase (SUV39H1) was identified by Rea et al. in 2000, and the first histone demethylase (LSD1) by Shi et al. in 2004, overturning the long-held belief that histone methylation was irreversible.

Genomic imprinting was proposed based on nuclear transplantation experiments by McGrath and Solter (1984) and Surani, Barton, and Norris (1984), who demonstrated that mouse embryos with two paternal or two maternal pronuclei (uniparental embryos) failed to develop normally, indicating that maternal and paternal genomes contributed non-equivalent information. The first imprinted genes (Igf2 and H19) were cloned by DeChiara et al. (1991) and Bartolomei et al. (1991).

The TET enzymes and 5hmC were discovered by Tahiliani et al. in 2009, who showed that the TET1 protein (named for a translocation involving chromosome 10 and 11 in leukaemia) could convert 5mC to 5hmC. Kriaucionis and Heintz simultaneously reported that 5hmC was enriched in Purkinje neurons, establishing it as a biologically significant modification rather than a mere intermediate.

Reprogramming to pluripotency was achieved by Takahashi and Yamanaka in 2006 using four transcription factors (OCT4, SOX2, KLF4, c-MYC) to convert mouse fibroblasts to induced pluripotent stem cells, earning Yamanaka the Nobel Prize in 2012 (shared with John Gurdon, who pioneered SCNT in frogs in 1962).

The Horvath epigenetic clock was published by Steve Horvath in 2013, demonstrating that a multi-tissue DNA methylation-based age estimator could be constructed from 353 CpG sites.

Bibliography Master

1. Waddington, C. H. — The epigenotype. Endeavour 1 (1942) 18-20.

2. Lyon, M. F. — Gene action in the X-chromosome of the mouse. Nature 190 (1961) 372-373.

3. Holliday, R. & Pugh, J. E. — DNA modification mechanisms and gene activity during development. Science 187 (1975) 226-232.

4. Riggs, A. D. — X inactivation, differentiation, and DNA methylation. Cytogenet. Cell Genet. 14 (1975) 9-25.

5. Strahl, B. D. & Allis, C. D. — The language of covalent histone modifications. Nature 403 (2000) 41-45.

6. Tahiliani, M. et al. — Conversion of 5-methylcytosine to 5-hydroxymethylcytosine in mammalian DNA by MLL partner TET1. Science 324 (2009) 930-935.

7. Takahashi, K. & Yamanaka, S. — Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors. Cell 126 (2006) 663-676.

8. Horvath, S. — DNA methylation age of human tissues and cell types. Genome Biol. 14 (2013) R115.

9. Brown, C. J. et al. — A gene from the region of the human X inactivation centre is expressed exclusively from the inactive X chromosome. Nature 349 (1991) 38-44.

10. McGrath, J. & Solter, D. — Completion of mouse embryogenesis requires both the maternal and paternal genomes. Cell 37 (1984) 179-183.

11. Alberts, B. et al. — Molecular Biology of the Cell, 7th ed. (Garland Science, 2022).

12. Lodish, H. et al. — Molecular Cell Biology, 9th ed. (W. H. Freeman, 2021).

13. Allis, C. D., Jenuwein, T. & Reinberg, D. — Epigenetics, 2nd ed. (Cold Spring Harbor Laboratory Press, 2015).

14. Kriaucionis, S. & Heintz, N. — The nuclear DNA base 5-hydroxymethylcytosine is present in Purkinje neurons and the brain. Science 324 (2009) 929-930.

15. Heijmans, B. T. et al. — Persistent epigenetic differences associated with prenatal exposure to famine in humans. Proc. Natl. Acad. Sci. USA 105 (2008) 17046-17049.