Loading
Old version — revision 1
This is a fixed snapshot of Epigenome editing, saved by Import as part of the initial corpus import. It is not edited and it is not updated; the article may have changed since.
Edit summary: Initial import of content/epigenome-editing.md — the filesystem corpus, unchanged. Not an edit.
Targeted modification of the chemical marks that control gene expression, switching a gene off or on without altering the underlying DNA sequence.
Epigenome editing changes how much of a gene a cell makes without changing the gene itself. A programmable DNA-binding protein is aimed at a promoter or enhancer and carries with it an enzyme that writes or erases a regulatory mark — DNA methylation, histone modification — or that simply recruits the cell's own repression machinery. The sequence is left intact, which means the change is in principle reversible, and which also means it can be lost. That distinguishes it sharply from Base editing and Prime editing, which alter the genome permanently and cannot be undone once installed.
Every epigenome editor is two interchangeable halves bolted together: something that finds a chosen stretch of DNA, and something that acts on the chromatin once it arrives. The halves are developed independently, which is why advances in targeting from nuclease work transfer directly to this field.
The most common targeting scaffold is catalytically dead Cas9, produced by disabling both nuclease domains of the enzyme described in CRISPR–Cas9. It retains full guide-RNA-programmed binding and none of the cutting. Zinc-finger proteins and TALE arrays, the platforms that preceded CRISPR, remain in use for the same purpose and have two advantages for therapy: they are considerably smaller, which eases packaging, and being closer to human protein scaffolds they may provoke less of an immune response than a bacterial enzyme. The tradeoff is that retargeting them requires protein engineering rather than swapping twenty nucleotides of RNA.
What is fused to the binding module determines the effect.
The distinction that matters clinically is between effects that persist only while the editor is present and effects that persist after it is gone. Simple repressor fusions do the former. Combining a KRAB domain with methyltransferases produces silencing that survives cell division and differentiation after a single transient exposure — the property that turned epigenome editing from a laboratory technique into a therapeutic proposition.1
The tools arrived in a predictable order: repression first, because recruiting existing silencing machinery is easy;2 activation next, because it needed multiplexed activation domains to reach useful expression levels; and durable, heritable marks last, because those required combining several enzymatic activities at once.3 The pivotal demonstration for medicine was that a transient dose could leave a permanent regulatory state — a "hit-and-run" edit — since a therapy that requires continuous expression of a bacterial protein is difficult to make safe.4
Work in mice then showed that a single systemic dose could silence a liver gene for the better part of a year without touching the DNA sequence.5 That result, together with the observation that reversal is possible with a demethylating editor, defined the commercial pitch: the durability of gene therapy with an off switch.
As of 2026 the field has moved into first-in-human testing, with results still limited. Several venture-funded companies pursue durable epigenetic silencing — one focused on chronic hepatitis B, where the aim is to silence both the viral episome and integrated viral DNA that antivirals cannot clear; another on facioscapulohumeral muscular dystrophy, where the disease is caused by inappropriate expression of a normally silent gene and where restoring silence is a more natural fix than cutting. Zinc-finger repressors have been developed for chronic pain and for prion disease, where lowering expression of a single protein is the whole therapeutic goal, though how far each has advanced into human testing is not firmly established. None of these has produced anything comparable to the approval record of nuclease-based Somatic gene therapy, and no epigenetic editor has reached the regulatory stage of Casgevy.
Consolidation has been rapid. Companies in the sector have merged or refocused since 2024, and the small number of programmes in the clinic means that a single safety event would change the field's trajectory. Readers should treat any specific programme status here as possibly outdated.
Why silencing is the natural first targetFor a large class of diseases the problem is too much of something — a viral genome, a toxic protein, a gene that should be off. Turning expression down requires no template, no repair pathway and no correction of a specific mutation, so one editor can serve every patient with the condition regardless of their individual genotype.
Durability is locus-dependent and imperfectly predictable. Promoters rich in CpG dinucleotides accept and propagate methylation well; CpG-poor promoters often do not, and some loci resist stable silencing entirely. Activation is harder than repression: raising a gene's output to a therapeutically meaningful, sustained level is much less reliable than shutting it down, which is why activation-based programmes lag repression-based ones.
Delivery imposes the same constraints as elsewhere in the field. The fusion proteins are large — larger than Cas9 alone — so single-vector packaging in AAV vectors is often impossible, and most in vivo work uses Lipid nanoparticles carrying messenger RNA, which mainly reaches the liver. Repeat dosing, if durability proves shorter than hoped, would run into anti-Cas9 immunity.
Because dCas9 binds tolerantly at sequences that would not be cut, epigenome editors can perturb transcription at sites where a nuclease would produce no off-target lesion at all. Off-target binding is transient and reversible in principle, but if the effector deposits durable methylation the perturbation may not be. Silencing can also spread beyond the intended element to neighbouring genes, and long-range enhancer contacts mean the affected gene is not always the nearest one.
A structural concern is verification. Sequencing shows that the genome is unchanged, which is reassuring but also means the standard safety assay is blind to the edit. Confirming that an epigenetic therapy did what was intended, and nothing else, requires genome-wide methylation and expression profiling of the treated tissue — feasible in a mouse, difficult in a patient.
Reversibility is a design goal, not a guaranteeErasing an installed mark has been demonstrated in cultured cells and in animals. Whether a clinician could reliably reverse a silencing therapy in a patient, in the tissue where it was delivered, has not been shown.
Epigenome editing is often confused with Epigenetic reprogramming, which uses pluripotency transcription factors to reset a cell's epigenome globally rather than at a chosen locus. The two share a premise — that cell state is written in modifiable marks — and little else. Targeted editing changes one element and leaves identity alone; Partial reprogramming changes identity and leaves the specific targets unchosen. Companies such as NewLimit work between the two, screening transcription factor combinations for effects on cell state rather than editing single regulatory elements, and readings from an Epigenetic clocks are sometimes used as an outcome measure in both approaches despite the unresolved question of what such clocks actually track.
The near-term test is durability in humans. Mouse liver is a permissive setting: hepatocytes divide slowly and the delivery problem is solved. Whether a single dose can hold a gene off for years in a dividing tissue, and whether the immune system tolerates the machinery, will determine whether epigenome editing becomes a category of medicine or a laboratory technique with a brief clinical detour. A heritable-germline application is not on any credible roadmap, and the frameworks in Governance of human genome editing would treat a durable epigenetic change to an embryo no differently from Human germline editing proper, since the marks would be transmitted to every cell of the resulting person.
paperNuñez, J. K. et al. "Genome-wide programmable transcriptional memory by CRISPR-based epigenome editing." Cell, 2021.↩Demonstrated in cultured human cells; the memory reported is measured across cell divisions in a dish, not in an organism.
paperGilbert, L. A. et al. "CRISPR-mediated modular RNA-guided regulation of transcription in eukaryotes." Cell, 2013. ↩
paperLiu, X. S. et al. "Editing DNA methylation in the mammalian genome." Cell, 2016. ↩
paperAmabile, A. et al. "Inheritable silencing of endogenous genes by hit-and-run targeted epigenetic editing." Cell, 2016. ↩
paperCappelluti, M. A. et al. "Durable and efficient gene silencing in vivo by hit-and-run epigenome editing." Nature, 2024.↩The long-lasting silencing is in mouse liver, the tissue lipid nanoparticles reach best and one that divides slowly.