question · intermediate · partly written
Can a gene be silenced permanently, and how does cellular memory work?
Can a gene be silenced permanently? In opened sources silencing lasts as long as machinery re-makes it, the longest measured engineered case is nearly a year in mouse liver, and every loop found has been undone by something.
Short answer from the sources
Silencing can last through many divisions and, engineered in mouse liver, for nearly a year. In each tested case it is kept by running machinery and something could undo it. No opened source shows a state that cannot be undone, or shows that none can exist. 123
Reported, not yet replicated
His hypothesis (amits-hypothesis): can a gene be silenced for ever
In his words: "there are a lot of phenomena that we still don't fully understand, like maybe cellular memory. How does that work? We know that there's gene silencing caused by permanent epigenetic regulations. In theory, could you permanently silence a gene? In theory, it makes sense, right?" 4
Amit's hypothesis
His hypothesis (amits-hypothesis): the knife picture
In his words: "If you think about a person with two arms and a knife, and they cut off both of their arms using the knife, now they can't cut off anything else. They've cut off the ability to cut off anything else, right? You can imagine this as a forever impact." 4
Amit's hypothesis
How a state is copied: DNA methylation more
The template is the old strand. Replication leaves a half-methylated site; UHRF1 ubiquitylates PAF15 (preferred in early S phase) or histone H3 (preferred in late S phase), and the doubly ubiquitylated protein recruits DNMT1 to methylate the new strand (Nishiyama et al. 2020, abstract). Replication-coupled maintenance depends on UHRF1-ligase 1 and PCNA-DNMT1 interactions, and a replication-uncoupled route, regulated by nucleosome occupancy and UHRF1 binding to methylated H3K9, can largely restore the methylome when the first is impaired. The authors state that mitotic inheritance is "quite robust" but imperfect, and that solo-WCGW sites and sites that lose methylation with age and in cancer show low maintenance efficiency (Ming et al. 2020, abstract).
Measured figures. At the hypermethylated FMR1 CpG island in human lymphocytes, hairpin-bisulfite PCR gave a maintenance efficiency of about 0.96 per site per cell division, and a de novo efficiency of 0.17 on that island (Laird et al. 2004, abstract). That is one locus. In HeLa cells more than half of methylated CpGs were maintained at the first time point after the fork, over 80% within 30 minutes, and the remainder over up to about 10 hours (Ming et al., full text per our source check). These time-course figures are a lag after the fork, not a per-division error rate, and the two kinds of figure must not be merged. A split by CpG context (isolated versus clustered CpGs, island versus not) is not published as numbers in the sources opened. 567
Reported, not yet replicated
How a state is copied: histone marks more
Parental histones are recycled and mixed with new unmarked ones. Positions of H3K4me3, H3K36me3, H3K79me3 and H3K27me3 are reproduced with high accuracy on new DNA and levels are restored over the cell cycle; about half of H3K4me3 regions recover within an hour and about 80% of H3K27me3 regions slowly, up to 24 hours (Reveron-Gomez et al. 2018; figures from the full text per our source check). PRC2 reads H3K27me3 through EED and is activated by it (Margueron et al. 2009, abstract). Whether this loop can hold a gene off without a DNA anchor is contested between two Drosophila papers (see the recycling record). 891011
Reported, not yet replicated
The longest-lived states more
X inactivation. Xist RNA was long thought dispensable once X inactivation is established but is now known to help maintain the inactive X. When Xist was deleted after inactivation in mouse embryonic cells, female pups showed partial X reactivation with no morbidity or mortality, and mice with Xist deleted in specific tissues had a normal life span; larger tumour burden appeared only under prolonged chronic stress of the gut (Yang et al. 2020, abstract). So the inactive X is held in part by continuing Xist and is not a one-way lock, though reactivation was partial.
Engineered silencing. In mouse liver one dose almost halved circulating PCSK9 for nearly a year and survived forced liver regeneration (Cappelluti et al. 2024, abstract): halving, not abolition.
Imprinting. In blood from many people, methylation at imprinted regions varies with factors including age, assisted reproduction, ancestry, pollutant exposure and sex, and CpGs closer to ZFP57 binding sites are less susceptible to change (Cecere et al. 2025, abstract). That shows imprints are not fixed in a population sample; it does not give how long an imprint lasts.
Durations of X inactivation and imprints, as measured lifespans, were not documented in any opened source. 3112
Reported, not yet replicated
The knife test: the nearest proposed loop, and what cut every loop found more
The nearest proposal to a state that blocks its own undoing is a self-reinforcing histone loop. Rea et al. (2000) showed SUV39H1 methylates H3 lysine 9 and that this is influenced by pre-existing modifications of the H3 tail; per the full text as read by our source check, Suv39h acts best on unmodified H3 tails, which new nucleosomes present, so K9 methylation "may even be self-propagating". That is a hypothesis of a loop, not shown to be unbreakable.
Every loop found in the opened sources was cut by something: targeted DNA demethylation relieved engineered silencing (Amabile et al. 2016); Xist deletion partly reactivated the inactive X (Yang et al. 2020); removal of the Polycomb response element let H3K27me3 dilute and repression be lost (Laprell et al. 2017); and the germline erases most methylation (Hackett et al. 2013). In each, "permanent" meant faithfully re-made over many divisions, by machinery that has to keep running.
The nearest thing to a lock is the irreversibly silent state in a three-state model fitted to single-cell time-lapse data (Bintu et al. 2016, abstract). It is a state of a fitted model; the abstract does not identify a molecular basis for it. 1323111415
Our inference
Where the evidence differs from his picture more
Two points, set beside his words and not in place of them. First, "permanent" in the papers opened here means a state faithfully re-made through divisions, and in each tested case something could undo it (see the knife test). Whether any state is permanent in the sense of unbreakable is not shown, and it is not shown to be impossible either: no experiment has run long enough or targeted a candidate lock. Second, the knife picture needs an act that removes the ability to undo itself; the closest proposal (Rea 2000) is a loop that reinforces itself, which is a weaker thing than removing the means of reversal. The question stays open. 131523
Our inference
Durations of the classic long-lived states, measured more
The follow-up search found no direct measurement, in years or per division, of how long an inactive X or an imprint lasts. In animals X inactivation leaks (at least 23% of X genes incompletely inactivated in human tissues) but largely holds; in cultured female stem cells XIST was lost over roughly 8 to 27 passages in one medium and no opened paper shows the lost state returning without gene editing. So permanent is supported as high fidelity with a small tail of failures, not as a measured lifetime. 161718
Reported, not yet replicated
Observation
Genes can be silenced by marks that persist through many cell divisions: methylation re-made at about 0.96 per site per division at one measured island, silencing that outlasted editor expression through division and differentiation in culture, and an engineered silencing in mouse liver that almost halved a circulating protein for nearly a year. In every tested case something could reverse it. 71912
Reported, not yet replicated
Hypotheses
Memory is a continuously re-made copy
A silenced state persists only while copying machinery keeps running; it can be undone by removing the writers, the template or the marks. 23115
Predicts. Silencing decays when the maintenance machinery or the template is removed, and can be reversed by targeted demethylation.
Would falsify. A silenced state that persisted unchanged through loss of its maintenance machinery and could not be reversed by any known intervention.
Reported, not yet replicated
A self-reinforcing histone loop
A methyltransferase that acts best on unmodified histone tails could make K9 methylation self-propagating (Rea et al. 2000, as a possibility). A hypothesis, not shown unbreakable. 131011
Predicts. Marks re-appear on new nucleosomes without a DNA anchor.
Would falsify. Mark dilution and loss of repression when the DNA anchor is removed, as one Drosophila paper reports (Laprell et al. 2017); the other (Coleman and Struhl 2017) reports memory without copying to new nucleosomes, so the question is contested.
Contested
An irreversibly silent state
A single-cell model fitted to time-lapse data has three states: active, reversibly silent, irreversibly silent (Bintu et al. 2016). The third state is a fitted model state, not an identified molecular lock. 15
Predicts. Cells that enter the third state do not reactivate within the observation time.
Would falsify. Reactivation of cells in that state by a known intervention, or a molecular identification showing it is a slow reversible state.
Reported, not yet replicated
What is missing
Measured durations
X inactivation and imprint durations as lifespans, with failure rates: not found in the opened sources. 312
Reported, not yet replicated
Per-context fidelity
Methylation copying by CpG context is not published as numbers in the opened sources. 6
Reported, not yet replicated
What is known
Methylation and histone marks are copied through division with high but imperfect fidelity; engineered silencing has lasted nearly a year in mouse liver and about 200 days in culture; the inactive X partly reactivates when Xist is removed after establishment; the germline erases most methylation. 612314
Reported, not yet replicated
What is not
What would settle it
Still open here
- Can a state be unbreakable, and what experiment would show it? No opened source shows one.
- How long do engineered silenced states last in humans? A trade-press report of a company phase I result exists but is unverified and is not used here.
- Measured durations of X inactivation and imprints: a follow-up search found none; see questions/how-long-do-x-inactivation-and-imprints-last.
Not yet written: Who works on it.
Linked from
- How DNA methylation is copied through replication
- DNA methylation
- Epigenome editing
- Histone recycling and the PRC2 read-write loop
- Revusiran
- siRNA
- Cellular memory
- Can a gene delivered by AAV be turned down or off after dosing in people?
- How long do X inactivation and imprints last, and how often do they fail?
- Is the irreversibly silent state of an engineered gene a molecular lock, and how long do silenced states last?
Sources
- Cappelluti, M.A., Mollica Poeta, V., Valsoni, S., et al.. Durable and efficient gene silencing in vivo by hit-and-run epigenome editing. Nature (2024). paper; opened 2026-11-17 all records citing this
- Amabile, A., Migliara, A., Capasso, P., et al.. Inheritable silencing of endogenous genes by hit-and-run targeted epigenetic editing. Cell (2016). paper; opened 2026-11-17 all records citing this
- Yang, L., Yildirim, E., Kirby, J.E., et al.. Widespread organ tolerance to Xist loss and X reactivation except under chronic stress in the gut. Proceedings of the National Academy of Sciences (2020). paper; opened 2026-11-17 all records citing this
- Amit's dictation of 2026-11-17: what the atlas must contain. Kept in the atlas repository; not published.
- Nishiyama, A., Mulholland, C.B., Bultmann, S., et al.. Two distinct modes of DNMT1 recruitment ensure stable maintenance DNA methylation. Nature Communications (2020). paper; opened 2026-11-17 all records citing this
- Ming, X., Zhang, Z., Zou, Z., et al.. Kinetics and mechanisms of mitotic inheritance of DNA methylation and their roles in aging-associated methylome deterioration. Cell Research (2020). paper; opened 2026-11-17 all records citing this
- Laird, C.D., Pleasant, N.D., Clark, A.D., et al.. Hairpin-bisulfite PCR: assessing epigenetic methylation patterns on complementary strands of individual DNA molecules. Proc Natl Acad Sci U S A (2004). paper; opened 2026-11-17 all records citing this
- Reveron-Gomez, N., Gonzalez-Aguilera, C., Stewart-Morgan, K.R., et al.. Accurate recycling of parental histones reproduces the histone modification landscape during DNA replication. Molecular Cell (2018). paper; opened 2026-11-17 all records citing this
- Margueron, R., Justin, N., Ohno, K., et al.. Role of the polycomb protein EED in the propagation of repressive histone marks. Nature (2009). paper; opened 2026-11-17 all records citing this
- Coleman, R.T., Struhl, G.. Causal role for inheritance of H3K27me3 in maintaining the OFF state of a Hox gene. Science (2017). paper; opened 2026-11-17 all records citing this
- Laprell, F., Finkl, K., Muller, J.. Propagation of Polycomb-repressed chromatin requires sequence-specific recruitment to DNA. Science (2017). paper; opened 2026-11-17 all records citing this
- Cecere, F., Relator, R., Levy, M., et al.. Identification of genetic and non-genetic modifiers of genomic imprinting through screening of imprinted DMR methylation in humans. Epigenetics & Chromatin (2025). paper; opened 2026-11-17 all records citing this
- Rea, S., Eisenhaber, F., O'Carroll, D., et al.. Regulation of chromatin structure by site-specific histone H3 methyltransferases. Nature (2000). paper; opened 2026-11-17 all records citing this
- Hackett, J.A., Sengupta, R., Zylicz, J.J., et al.. Germline DNA demethylation dynamics and imprint erasure through 5-hydroxymethylcytosine. Science (2013). paper; opened 2026-11-17 all records citing this
- Bintu, L., Yong, J., Antebi, Y.E., et al.. Dynamics of epigenetic regulation at the single-cell level. Science (2016). paper; opened 2026-11-17 all records citing this
- Tukiainen T, Villani AC, Yen A, et al.. Landscape of X chromosome inactivation across human tissues. (2017). paper; opened 2026-11-17 all records citing this
- Cloutier M, Kumar S, Buttigieg E, et al.. Preventing erosion of X-chromosome inactivation in human embryonic stem cells. (2022). paper; opened 2026-11-17 all records citing this
- Mekhoubad S, Bock C, de Boer AS, et al.. Erosion of dosage compensation impacts human iPSC disease modeling. (2012). paper; opened 2026-11-17 all records citing this
- Nunez, J.K., Chen, J., Pommier, G.C., et al.. Genome-wide programmable transcriptional memory by CRISPR-based epigenome editing. Cell (2021). paper; opened 2026-11-17 all records citing this
Where it sits on the map
History of this record (9)
This record has been changed 9 times, on the dates below. The note kept against each change is written for the people who maintain the atlas, in their own shorthand, so it is not shown here; it is in the record itself.
- Kind
- amit
- Register
- observed-unexplained
- Status
- partly written
- Last updated
- 2026-11-19
- Record
- content/questions/can-a-gene-be-silenced-forever.json