Epigenetics and chromatin dynamics: DNA methylation, histones, remodellers, and 3D genome organisation

Short version. Epigenetics is the study of heritable or persistent changes in gene activity that do not involve changes in DNA sequence. The principal layers are DNA methylation, covalent histone modifications, ATP-dependent chromatin remodelling, and three-dimensional chromatin organisation. Together they specify which subset of the genome is accessible and active in a given cell. The Roadmap Epigenomics reference epigenomes (Kundaje et al. 2015) provide the genome-scale catalogue across human tissues; the loop-extrusion model (Sanborn 2015; Fudenberg 2016) provides the mechanistic account of TADs.

DNA methylation

The most extensively studied DNA modification in mammals is 5-methylcytosine (5mC) at CpG dinucleotides. Genome-wide, around 70 to 80% of CpGs in human somatic tissues are methylated; the unmethylated minority is concentrated at CpG islands — CG-dense regions of about 1 kb that overlap roughly 60% of mammalian gene promoters and that are typically kept unmethylated by transcriptional machinery. CpG-island methylation is associated with stable transcriptional silencing, classically illustrated by tumour-suppressor gene silencing in cancer and by the establishment of imprinting. The biology is reviewed comprehensively in Bird 2002, Genes Dev 16:6.

Three DNA methyltransferases write the modification. DNMT1 is the maintenance enzyme: it preferentially methylates hemi-methylated CpG sites, copying the methylation pattern of the parental strand onto the newly synthesised daughter strand after replication. This is the molecular mechanism by which methylation patterns are heritable through cell division. DNMT3A and DNMT3B are the de novo methyltransferases that establish new methylation patterns, especially during gametogenesis and early post-implantation development. DNMT3L is a non-catalytic regulatory cofactor.

Active demethylation is catalysed by the TET family of dioxygenases (TET1, TET2, TET3), which oxidise 5mC to 5-hydroxymethylcytosine (5hmC), then to 5-formylcytosine and 5-carboxylcytosine, which are removed by base-excision repair (Tahiliani et al. 2009, Science 324:930). 5hmC has additional roles as a stable mark, particularly in neurons and embryonic stem cells. The standard genome-wide assay for cytosine methylation is bisulfite sequencing, which deaminates unmethylated cytosines to uracil while leaving 5mC intact; high-resolution variants include whole-genome bisulfite sequencing (WGBS), reduced-representation bisulfite sequencing (RRBS), and oxidative bisulfite sequencing for distinguishing 5mC from 5hmC.

Histone modifications

Histone proteins package DNA into nucleosomes (147 bp wrapped around a histone octamer of two each of H2A, H2B, H3, and H4). The histone tails project outward from the nucleosome and are subject to a large vocabulary of covalent modifications — acetylation, methylation, phosphorylation, ubiquitination, sumoylation — written, read, and erased by dedicated enzymes. A handful of marks dominate functional-genomics interpretation:

• H3K4me3 — trimethylation of lysine 4 on histone H3 — marks active or poised RNA polymerase II promoters; ChIP-seq peaks are sharp and centred on transcription start sites.
• H3K4me1 — monomethylation of the same residue — marks enhancers (poised or active).
• H3K27ac — acetylation of lysine 27 on H3 — distinguishes active enhancers from poised ones (which carry H3K4me1 but lack H3K27ac).
• H3K27me3 — trimethylation of the same residue — is the canonical Polycomb-repressive mark deposited by PRC2.
• H3K9me3 — trimethylation of lysine 9 on H3 — marks constitutive heterochromatin (pericentromeric, telomeric, transposon-rich).
• H3K36me3 — trimethylation of lysine 36 on H3 — marks transcribed gene bodies.

ChromHMM is the standard hidden-Markov model approach for integrating these marks into a small number of discrete chromatin states ("active TSS", "strong enhancer", "Polycomb repressed", "weak transcription", "constitutive heterochromatin", and so on) at high genome-wide resolution. The Roadmap Epigenomics Consortium applied ChromHMM to histone-modification, DNA methylation, and accessibility data across 111 reference epigenomes (Kundaje et al. 2015, Nature 518:317) to produce the per-tissue chromatin-state annotations now in routine use.

ATP-dependent chromatin remodellers

Nucleosome positioning is dynamic, and four families of ATP-dependent chromatin remodellers actively reposition, evict, or restructure nucleosomes: SWI/SNF (BAF and PBAF in mammals) uses ATP hydrolysis to slide or evict nucleosomes, opening regulatory regions; ISWI spaces nucleosomes regularly along chromatin and assembles chromatin during replication; CHD remodellers include the NuRD complex (which couples remodelling to histone deacetylation); INO80 remodellers exchange canonical H2A for the variant H2A.Z at promoters and play roles in DNA-damage response. The biology is reviewed in Ho et al. 2014, Nature 512:449, and a striking proportion of human malignancies harbour mutations in SWI/SNF subunits (notably ARID1A, SMARCA4, SMARCB1), establishing chromatin remodelling as a major axis of disease biology in research contexts.

Polycomb and Trithorax

Polycomb-group (PcG) and Trithorax-group (TrxG) complexes are the canonical machinery of long-range, heritable transcriptional memory in metazoans. PRC2 contains EZH1/EZH2 (the catalytic subunit), SUZ12, and EED, and trimethylates H3K27 to deposit the H3K27me3 repressive mark. PRC1 contains BMI1 (and paralogues), CBX, RING1A/B, and ubiquitinates H2AK119; classical PRC1 reads H3K27me3 via its CBX subunit, anchoring repression at PRC2-marked sites. The two complexes together produce stable, heritable repression of developmental regulators — classically the Hox gene clusters — that is propagated through cell divisions. The biology is reviewed in Margueron & Reinberg 2011, Nature 469:343. Trithorax complexes (the MLL/COMPASS family) deposit H3K4me3 and counteract Polycomb repression, maintaining the active state of genes in which their resolution has gone the active-direction.

X-inactivation

X-chromosome inactivation is the canonical case of heritable, monoallelic, chromosome-wide silencing, and was first described in genetic terms by Mary Lyon (Lyon 1961, Nature 190:372). In female placental mammals, one of the two X chromosomes is transcriptionally silenced early in development. Initiation is driven by the long non-coding RNA XIST, which is transcribed from the future inactive X and coats the chromosome in cis. XIST recruits PRC2 and other repressive machinery (HNRNPK, SPEN, LBR, etc.), excludes RNA polymerase II, and establishes a heterochromatic state with H3K27me3, H3K9me3, and DNA methylation that is then propagated through cell divisions. Because the choice of which X is silenced is stochastic and clonally heritable, female mammals are mosaic for X-linked variation — the mechanism behind tortoiseshell coat patterns in cats and the variable expression of X-linked recessive variants in heterozygous females. This is relevant to several pedigree analyses; Mendelian inheritance covers the inheritance-pattern implications.

Genomic imprinting

Imprinting is the phenomenon by which a small set of mammalian genes (around 100 in humans) is expressed exclusively from the maternally or paternally inherited allele. The parental-origin information is set as differential DNA methylation at imprinting control regions in the germline (DNMT3A and DNMT3L play central roles in establishing this), and propagated through somatic divisions. The classical disease consequences are Prader-Willi and Angelman syndromes (loss of paternal or maternal expression at chromosome 15q11-q13, respectively) and Beckwith-Wiedemann syndrome (chromosome 11p15.5). Imprinting is the mechanism by which the same DNA sequence can produce different phenotypes depending on parental origin — a phenomenon visible in pedigrees and accommodated explicitly in some inheritance analyses.

Topologically associating domains and 3D genome organisation

The genome is not a linear string; in three-dimensional space it folds into a hierarchical structure visible in Hi-C contact maps. At the megabase scale, the genome partitions into A and B compartments — A predominantly active, gene-rich, and accessible; B predominantly inactive and heterochromatic. At the sub-megabase scale, the genome partitions into topologically associating domains (TADs): regions within which DNA-DNA contacts are enriched and across whose boundaries contacts are depleted. TADs were described independently in mouse and human (Dixon et al. 2012, Nature 485:376; Nora et al. 2012, Nature 485:381) and are remarkably conserved across cell types and species — suggesting that they reflect a fundamental aspect of chromosome folding rather than a cell-specific regulatory phenomenon.

The mechanistic explanation for TADs is the loop-extrusion model. Cohesin loads onto chromatin and extrudes a loop, reeling DNA through itself bidirectionally until it stalls at a CTCF-bound site oriented in the appropriate direction (CTCF motifs at TAD boundaries are convergent). The model was articulated computationally by Sanborn and colleagues (Sanborn et al. 2015, PNAS 112:E6456) and by Fudenberg and colleagues (Fudenberg et al. 2016, Cell Rep 15:2038), and reproduces Hi-C contact maps with high fidelity. Live-cell imaging and single-molecule biophysical experiments subsequently confirmed that purified cohesin extrudes DNA loops in vitro. TAD boundaries constrain which enhancers contact which promoters, and disrupting boundaries can reroute regulatory contacts — the molecular basis of several developmental disorders studied at research scale.

Why this matters for pedigree-level analysis

Several pedigree-visible phenomena are epigenetic in origin: imprinting disorders show non-Mendelian inheritance patterns that depend on parental origin; X-linked recessive variants show variable expressivity in heterozygous females because of stochastic X-inactivation; some forms of variable expressivity reflect epigenetic modifier loci. Family-history-based pedigree modelling does not directly model methylation states, but understanding these mechanisms is essential context. The educational pillar regulation of gene activity and the companion pages on cis-regulatory elements and gene regulatory networks place the chromatin layer in its broader regulatory context.

Frequently asked questions

Related reading

• Regulation of gene activity (pillar)
• Cis- and trans-regulatory elements
• Gene regulatory networks
• Mendelian inheritance calculator
• Complex disease pedigree software