Histone Modification

Why a tag on a protein controls a gene

DNA inside a human nucleus is two metres long, but the nucleus itself is six microns wide. The cell solves this packaging problem by wrapping its DNA around eight histone proteins to form a nucleosome — a bead, repeated millions of times along the genome. Each nucleosome has eight short, unstructured protein tails (mostly N-terminal) that poke out into the surrounding space, and it is on these tails that the chemistry happens.

The tail residues most relevant to gene regulation are lysines (denoted K) and arginines (R). Add an acetyl group (C₂H₃O) to a lysine and the residue's positive charge disappears; the histone's grip on the negatively charged DNA loosens, the nucleosome unsticks, and a stretch of DNA becomes available to RNA polymerase. Strip the acetyl back off and chromatin tightens. The reaction is freely reversible at physiological pH, which is the entire point: the cell needs to flip genes on and off thousands of times per cell cycle without ever cutting DNA.

Methylation works differently. A methyl group (CH₃) does not change charge; it adds a small hydrophobic bulge that creates a binding site. Whether the binding site activates or represses depends on which protein domain recognises it. H3K4me3 — three methyls on lysine 4 of histone H3 — recruits the basal transcription factor TAF3 and the CHD1 chromatin remodeller, signalling "promoter, transcribe me". H3K27me3 — three methyls on lysine 27 of the same histone — recruits CBX proteins of Polycomb Repressive Complex 1, which compact chromatin and lock the gene shut. Same chemistry, opposite outcome, because the readers are different.

The major marks at a glance

Mark	Effect	Where it sits	Writer	Eraser	Reader
H3K4me3	Activating	Active promoters	MLL / SET1 (KMT2 family)	KDM5 family (LSD2)	TAF3, CHD1 (PHD finger)
H3K9me3	Repressive (constitutive)	Pericentric heterochromatin, transposons	SUV39H1, SETDB1	KDM4 (JMJD2)	HP1 (chromodomain)
H3K27me3	Repressive (facultative)	Developmental gene bodies	EZH2 (PRC2)	UTX (KDM6A), JMJD3	CBX in PRC1 (chromodomain)
H3K27ac	Activating	Active enhancers and promoters	p300 / CBP (HAT)	HDAC1-3, SIRT1	BRD4 (bromodomain)
H3K36me3	Activating, transcription-coupled	Bodies of expressed genes	SETD2	KDM4 family	MRG15, PWWP-domain proteins
H4K16ac	Activating, anti-compaction	Open chromatin globally	MOF (KAT8)	SIRT1, HDAC1	BRD-containing factors

The first thing to take away from the table is that there is no single rule of the form "methylation = silencing" or "acetylation = activation". Acetylation is reliably activating — that part is consistent because the mechanism is electrostatic. Methylation is residue-specific. H3K4me3 always activates because it is read by the transcription machinery; H3K27me3 always silences because it is read by Polycomb. Cells exploit this dictionary, not a single rule.

Writer, reader, eraser — the three-part logic

Every mark in the table above belongs to a triplet of proteins. The writer catalyses deposition; the reader contains a binding domain that recognises the mark; the eraser reverses it. Understanding the triplet is the cleanest way to predict what a drug will do.

• Histone acetyltransferases (HATs). p300 and CBP are the most studied; they acetylate H3K27 and H3K18 at enhancers and promoters. The MYST family (MOF, Tip60) handles H4K16 and H4K5/8/12.
• Histone deacetylases (HDACs). Eighteen enzymes in four classes. HDAC1-3 are the workhorses of nuclear repression; sirtuins (SIRT1-7) are NAD-dependent and link acetylation status to cellular metabolism — when NAD runs low (ageing, fasting), sirtuin activity drops and acetylation accumulates.
• Histone methyltransferases (HMTs). SET-domain enzymes; ~50 in humans. Each is highly residue-specific: EZH2 only writes H3K27, SUV39H1 only H3K9, MLL only H3K4.
• Histone demethylases (KDMs). Two chemistries: LSD1 (KDM1A) uses FAD and only removes mono- or dimethyl marks; the JmjC family uses Fe(II) and α-ketoglutarate and can remove all three methylation states.
• Reader domains. Bromodomains bind acetyl-lysine; chromodomains, Tudor, PWWP, and PHD-finger domains bind methyl-lysines with residue specificity. A single chromatin remodeller often combines multiple readers in one polypeptide so it only acts where two marks coincide.

The same logic explains why every successful epigenetic drug we have to date hits one of these three nodes. Vorinostat and romidepsin inhibit HDACs, raising acetylation across the genome and selectively killing T-cell lymphoma cells. Tazemetostat inhibits EZH2 and is approved for tumours that depend on H3K27me3-driven silencing. JQ1 and clinical BET inhibitors block the bromodomain readers of acetyl-lysine and pull RNA polymerase off of MYC enhancers.

The histone code is combinatorial

Marks rarely act alone. A typical active promoter carries H3K4me3, H3K27ac, and H3K9ac; a gene body adds H3K36me3; an active enhancer is H3K4me1 plus H3K27ac. Switching one mark for another changes the meaning of the locus. Embryonic stem cells exploit this with bivalent chromatin: developmental promoters carry both activating H3K4me3 and repressive H3K27me3 simultaneously. The gene is poised — silent now, but ready to switch on the moment the cell commits to a lineage and the H3K27me3 mark is removed.

The combinatorial nature is also why chromatin immunoprecipitation experiments now use sequencing readouts (ChIP-seq, CUT&RUN) rather than antibody alone — biologists need to know which marks coincide, not just which marks exist. The ENCODE and Roadmap Epigenomics projects mapped the genome-wide distribution of dozens of marks across hundreds of cell types and produced the chromatin state segmentations that anchor most modern enhancer annotation.

Histone modification vs DNA methylation

	Histone modification	DNA methylation
Substrate	Histone tail residues (K, R)	Cytosine base in DNA
Number of marks	Dozens, combinatorial	Mostly one (5-methylcytosine)
Reversibility	Fast, minutes	Slower, requires replication or TET
Inheritance through replication	Reader-writer feedback	DNMT1 copies the parent strand
Direction of effect	Activating or silencing	Mostly silencing in promoters
Drugs	HDACi, EZH2i, BET inhibitors	5-azacytidine, decitabine

The two systems are intertwined. H3K9me3 recruits the DNA methyltransferase DNMT3A to its target loci; conversely, the DNA-bound MeCP2 protein recruits HDACs to deacetylate flanking histones. Cells use the redundancy as a safety net — silencing a transposon should not depend on a single mark.

Where it shows up in the clinic

• Cancer. EZH2 gain-of-function mutations drive lymphomas; SETD2 loss accelerates clear-cell renal carcinoma; KDM6A loss appears in 25% of bladder cancers. Tazemetostat (FDA 2020) directly targets the EZH2 dependency.
• Cutaneous T-cell lymphoma. Vorinostat (2006) and romidepsin (2009), both HDAC inhibitors, were the first epigenetic therapies approved.
• Rett syndrome. Mutations in MECP2 — the protein that reads methylated DNA and recruits HDACs — cause severe neurological regression in girls around 18 months of age.
• Cardiac hypertrophy. p300 acetylates GATA4 in stressed cardiomyocytes; small-molecule p300 inhibitors are in trials to slow heart failure.
• HIV latency. The provirus integrates into host chromatin and is silenced by H3K9me3 and HDAC activity. "Shock-and-kill" strategies use HDAC inhibitors to wake latent virus so the immune system can clear it.

Variants and exotic marks beyond acetyl/methyl

• Phosphorylation. H3S10ph is the mitotic mark; condensin and other mitotic regulators read it. H2AX phosphorylation (γH2AX) flags double-strand breaks for repair.
• Ubiquitination. H2AK119ub is deposited by PRC1 (the BMI1/RING1 complex) and reinforces Polycomb silencing. H2BK120ub on gene bodies activates transcription.
• SUMOylation, ADP-ribosylation, crotonylation, lactylation. Newer marks; lactylation in particular links the lactate output of glycolytic cells (including macrophages and tumour cells) to transcriptional changes.
• Histone variants. H3.3 replaces canonical H3 at active genes and telomeres; CENP-A replaces H3 at centromeres; macroH2A replaces H2A at silent X-linked regions. The chromatin is the variant plus its marks, not the marks alone.
• Histone clipping. Some cells (granulocytes, ageing yeast) cleave histone tails altogether, removing whole stretches of marks at once.

Pitfalls and easy misreadings

• "Methylation silences." Only some methyl marks silence. H3K4me3, H3K36me3, and H3K79me2 all activate. Always specify the residue.
• "Histone code = deterministic." The same combination of marks does not always predict the same expression outcome — chromatin context, sequence-specific transcription factors, and 3D contacts all matter. The code is more like a probability landscape than a lookup table.
• "Marks cause expression." Many marks are downstream of, not upstream of, transcription. H3K36me3 is deposited by SETD2 riding along with elongating RNA polymerase. Removing the mark sometimes has surprisingly mild effects because it is a consequence rather than a cause.
• "Reversible means safe." HDAC inhibitors hit dozens of substrates, including non-histone proteins (p53, tubulin, HSP90); cardiotoxicity and thrombocytopenia are well documented.
• Lamarckian temptation. Histone marks reset largely (though not entirely) in primordial germ cells. Headlines about "epigenetic inheritance of stress" overstate a small, debated effect; the genome remains the dominant inherited substrate.