Revumenib does not move H3K27me3 in MLL-fusion leukaemia, in vitro or in vivo: the absence narrows a THIRD time and the experiment's prior collapses
also called H3K27me3 absence · experiment C · SNDX-5613 H3K27me3 · stale correction queue
For many cycles this project has said that its cheapest and most valuable experiment is to measure a particular repressive chromatin mark after giving a menin inhibitor, **because nobody had ever done it.** That claim had already been narrowed once. Re-checking it today it is wrong again: somebody has done it - in stem cells taken from human fat, not in leukaemia - and the drug **did not move the mark at all.** The experiment is still worth doing, but the sentence justifying it has to change, and the one existing result points away from the expected answer.
The same thing, for a scientist
Ledger experiment **C** rests on the assertion that **no menin-inhibitor study in any system has ever reported an H3K27me3 measurement.** That was narrowed on 2026-09-20 to *no measurement under a PHARMACOLOGICAL menin inhibitor*, because Sparbier et al measured it in a MEN1-knockout arm. **The narrowed version is also false.** **PMID 38173411** (Chang et al, *Stem Cells Transl Med* 2024) treated human adipose-derived stem cell spheroids with **70 uM SNDX-5613 - revumenib, named as such in the paper** - and blotted for H3K27me3. Verbatim: *"The SNDX-5613 effectively inhibited the trimethylation of H3K4 and H3K9 but **had no effect on H3K27me3**, HDAC5, and pHDAC4/5/7."* **So a pharmacological menin inhibitor has been assayed against H3K27me3 and the result was negative.** The claim must now be stated by system and by assay, and the one datapoint available cuts against the hypothesis the experiment was designed to test. **NARROWED A THIRD TIME, 2026-09-20, by rule 80 applied to this entry's own claim.** The bounded queries behind the first two narrowings searched **titles and abstracts**. A Europe PMC **full-text** search (`BODY:`) for menin-inhibitor names with H3K27me3 returns **105 records** where the abstract-level query returned 3. In one of them, **PMID 42170936**, MLL-fusion leukaemia cells treated with **Revumenib** show that the drug *"don't affect the total levels of H3K4me3, H3K79me2, and H3K27me3"*, and MV4;11 tumours from treated mice show reduced H3K36me3 **but not H3K4me3 or H3K27me3**. **So a pharmacological menin inhibitor has been assayed against H3K27me3 in the index-adjacent disease, in vitro AND in vivo, and the result is negative.**
What the defensible statement is now
No locus-resolved H3K27me3 measurement under a pharmacological menin inhibitor has been reported in a KMT2A-rearranged leukaemia. One bulk measurement exists, by Western blot, in human adipose-derived stem cells, and it was negative.
Three things that statement preserves and the old one did not:
- The assay matters. A bulk Western blot of total H3K27me3 is a very different instrument from locus-resolved ChIP or CUT&Tag with spike-in calibration. Global levels can be flat while redistribution is large - that is the normal case for this mark.
- The system matters (rule 73). Adipose-derived stem cells have no KMT2A fusion, and the menin-KMT2A axis there is wild-type biology. A negative there is weak evidence about a fusion line and strong evidence about nothing else.
- The direction matters. The experiment was motivated by the idea that menin inhibition phenocopies polycomb disruption through wild-type KMT2A/B. This is the first measurement bearing on it and it is negative, which the ledger should say rather than continuing to advertise the experiment as untouched.
And the leukaemia measurement is one antibody away in a built system
Checking the same bounded query surfaced PMID 34663924 (Janssens et al, Nat Genet 2021), automated CUT&Tag in KMT2A-rearranged leukaemia. It profiles H3K27me3 extensively - bivalent H3K4me3/H3K27me3 signatures at oncofusion-binding sites, and single-cell H3K27me3 scores at HOXA9 - and it runs a menin-inhibitor arm: H3K4me3 peaks in SEM cells altered by 30 uM VTP50469 for 72 h.
But the two are not crossed. The drug arm profiled H3K4me3 only; EZH2 is not mentioned anywhere in the paper. So the absence survives for the index disease - and it survives in a platform that already had the antibody, the cell line and the drug. That makes experiment C cheaper than the ledger claims, not less necessary: it is one additional CUT&Tag target in an assay these authors automated.
How this was found, which is the transferable part
From the stale-correction queue, which is ledger priority 2 and had never been worked. scripts/check-stale-corrections.py reports 95 un-struck assertions across 525 entries carrying a correction marker, 52 of them sitting BEFORE the correction. This entry is one item off that list.
The queue is a list of sentences that were true when written. That is the failure mode it exists to catch, and it caught one on the first read - on the project's own top-ranked experiment, a claim that had already been corrected once and was still wrong. A claim that has been narrowed once is not thereby safe; the narrowing fixed the counter-example that had been found, not the class of counter-example.
Third narrowing, and this time the experiment's prior goes with it
What survives of the claim is now very thin. Only the words locus-resolved are holding it up: PMID 42170936 is immunoblotting of global levels, not ChIP or CUT&Tag, and a redistribution of H3K27me3 at constant total level is exactly what a blot cannot see.
But the motivation for ledger experiment C is now in poor shape, and that matters more than the claim. The experiment was proposed because menin inhibition might phenocopy polycomb disruption through wild-type KMT2A/B. There are now two independent negative bulk measurements:
| system | drug | result |
|---|---|---|
| human adipose-derived stem cells | SNDX-5613 (revumenib), 70 uM | no effect on H3K27me3 |
| MLL-fusion leukaemia cells and MV4;11 xenograft tumours | Revumenib | no change in H3K27me3, in vitro and in vivo |
The second is in the right disease. If menin inhibition were producing a substantial polycomb shift, a global blot in an MLL-fusion line would be a reasonable place to see it, and it does not.
The honest status of experiment C: the question is still legitimate, because redistribution at constant global level is real and common for this mark. But it is no longer unmeasured, its two existing measurements both point the same unhelpful way, and it should not sit near the top of a priority list as though untouched. Downgrade it and say why.
And note what found this. Rule 80 — a PubMed zero is bounded by abstracts, not by papers — was written this morning from another beat's finding. It has now changed an answer three times on the day it was written: the SET::NUP214 H3K79 zero, the localization beat's DOT1L imaging absence, and this. The measurements a mechanism graph wants really are, overwhelmingly, supporting panels.
bounded query
{
"target": "(menin inhibitor OR revumenib OR SNDX-5613 OR VTP50469 OR ziftomenib OR KO-539 OR MI-503 OR MI-463 OR bleximenib OR JNJ-75276617) AND (H3K27me3 OR \"H3K27 trimethylation\")",
"target_hits": 3,
"on_topic_after_title_and_full_text_check": 2,
"controls": {
"same shape with H3K79": 6,
"menin inhibitor alone": 666,
"H3K27me3 alone": 7000
},
"run": "PubMed esearch, 2026-09-20"
}stale correction queue — the numbers behind the chart
{
"entries_with_marker": 525,
"un_struck_assertions": 95,
"sitting_before_the_correction": 52
}- This does not retire experiment C and it does not license anyone to say menin inhibition leaves H3K27me3 alone. A bulk Western blot in fat-derived stem cells at 70 uM is not a test of locus-level redistribution in a KMT2A-rearranged blast. What it retires is the SENTENCE the experiment was advertised with.
- And it is a warning about this project's own corrections. This claim had already been narrowed once, five cycles ago, and the narrowed version was still false. Narrowing a claim to exclude the counter-example you found does not protect you from the class of counter-example you did not look for.
- This claim has now been narrowed three times in one day - from 'no study in any system', to 'none under a pharmacological inhibitor', to 'none locus-resolved in a KMT2A-rearranged leukaemia'. Each narrowing was forced by a paper found with a slightly better search. Treat the surviving version as provisional and run the full-text query before quoting it.
- Does H3K27me3 REDISTRIBUTE at menin-inhibitor-silenced loci while global levels stay flat? That is now the entire surviving question, and both existing measurements are blots that cannot answer it.
- Does the GSK126 arm of PMID 38173411 show a bulk H3K27me3 change in the same blot? That is the positive control that decides whether the negative under SNDX-5613 means anything.
- Would one extra CUT&Tag target - H3K27me3 - added to the already-automated VTP50469 arm of PMID 34663924 settle experiment C outright? It appears to be the cheapest remaining route.
- How many of the other 94 un-struck assertions in the stale-correction queue have gone the same way? One read found one on the top-ranked experiment.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.journal-article
In human adipose-derived stem cell spheroids, treatment with 70 uM SNDX-5613 (revumenib, a pharmacological menin-KMT2A interaction inhibitor) reduced H3K4 and H3K9 trimethylation by Western blot but had no effect on H3K27me3, establishing that a pharmacological menin inhibitor has been assayed against H3K27me3 and that the reported result was negative.
Histone Trimethylations and HDAC5 Regulate Spheroid Subpopulation and Differentiation Signaling of Human Adipose-Derived Stem Cells. · Chang MM, Hong YK, Hsu CK, et al. · Stem Cells Transl Med 2024What this evidence cannot show
Read from the full text (PMC10940829), Fig. 5B with quantification in Supplementary Fig. S14A. **The system is human adipose-derived stem cells, not a KMT2A-rearranged leukaemia** - no fusion, wild-type menin-KMT2A biology, and a differentiation assay rather than a leukaemia one (rule 73). **The assay is a bulk Western blot of total H3K27me3**, which cannot detect redistribution at constant global level - the expected behaviour of this mark. 70 uM is far above the concentration used in leukaemia work. **Strong, not established**: the negative is reported as a specificity control for other inhibitors rather than as a result in its own right, so it received no dose-response and no replication as a primary endpoint.
Evidence that argues against this (1)
- A global Western blot is close to the weakest possible test of this hypothesis. The menin-polycomb argument predicts REDISTRIBUTION of H3K27me3 at specific loci, which leaves bulk levels unchanged; the same paper's own EZH2-inhibitor arm (GSK126) is the positive control that would show whether the blot could detect a change at all, and that comparison is not drawn. So this bounds the claim 'nobody has measured it' without bearing much on the biology. Sparbier CE et al. 2023
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.journal-article
In KMT2A-rearranged leukaemia cell lines and patient samples, automated CUT&Tag profiling reported H3K27me3 at oncofusion-binding sites including bivalent H3K4me3/H3K27me3 signatures and single-cell H3K27me3 scores at HOXA9, and separately reported an arm treating SEM cells with 30 uM VTP50469 for 72 hours, but the drug arm profiled H3K4me3 and the two were not crossed.
Automated CUT&Tag profiling of chromatin heterogeneity in mixed-lineage leukemia. · Janssens DH, Meers MP, Wu SJ, et al. · Nat Genet 2021What this evidence cannot show
Read from the full text (PMC8571097); H3K27me3 occurs 57 times and EZH2 not at all. **This is a negative claim about what a paper does NOT contain**, made from the full text including extended-data legends but without inspecting every supplementary table, so an unreported H3K27me3-plus-drug panel cannot be excluded absolutely. The VTP50469 arm is a single cell line (SEM) at one concentration and timepoint.
Evidence that argues against this (1)
- The bivalency this paper maps is itself the argument that the crossed experiment would matter: if oncofusion targets are already H3K4me3/H3K27me3 bivalent and menin inhibition moves H3K4me3 at those sites, then H3K27me3 is the obvious second axis and its absence from the drug arm may reflect a deliberate scoping decision rather than an oversight - which would mean the authors had a reason this entry does not know. Sparbier CE et al. 2023
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.journal-article
In MLL-fusion leukaemia cells, treatment with the pharmacological menin inhibitor Revumenib alone or combined with the SETD2 inhibitor EZM0414 did not affect total levels of H3K4me3, H3K79me2 or H3K27me3 by immunoblot, and MV4;11 xenograft tumours from treated mice showed reduced H3K36me3 but unchanged H3K4me3 and H3K27me3.
Blocking SETD2 Enhances the Therapeutic Efficiency of Menin Inhibitor in MLL-Fusion Leukemia. · Li A, Zhang Y, Li Z, et al. · Cancer Sci 2026What this evidence cannot show
Read from the full text (PMC13394958), Figures S4A/B and 8E. **These are immunoblots of GLOBAL histone levels, not locus-resolved ChIP or CUT&Tag**, so a redistribution of H3K27me3 at constant total level would be invisible - which is the normal behaviour of this mark and the reason the remaining question is legitimate. The H3K27me3 blots are **controls for the paper's SETD2 argument**, not a primary endpoint, so they received no dose-response and no replication as a result in their own right. MLL-fusion AML (MV4;11 and NPM1-mut lines), not infant B-lineage KMT2A::AFF1 (rule 73).
Evidence that argues against this (1)
- A negative global blot is weak evidence against locus-level change, and this entry should not be read as closing the question. The published argument for expecting a polycomb shift is about redistribution at menin-inhibitor-silenced targets, which a whole-cell histone blot averages away entirely. What the two negatives establish is that no LARGE global shift occurs - not that nothing happens at the loci that matter. Chang MM et al. 2024
NARROWS ITS PREMISE A SECOND TIME: a pharmacological menin inhibitor has been assayed against H3K27me3 and the result was negative - in adipose stem cells, by bulk blot.
THE FIRST NARROWING WAS NOT ENOUGH. That entry's surviving claim is also false.
CARRIES THE SAME SUPERSEDED CLAIM, now corrected in place.
THE MARK ITSELF: the only pharmacological-menin-inhibitor measurement of it is negative and non-leukaemic.
- related to →Trithorax against Polycomb at HOXA: the textbook antagonism, and the two ways this disease breaks it
THE HYPOTHESIS THIS BEARS ON: the one existing datapoint cuts against it, weakly.
A BOUNDED ABSENCE CAN STILL BE WRONG: this one was bounded, corrected once, and still false.
SAME QUEUE, SAME SESSION: a second advertised experiment checked against the literature. This one's absence SURVIVED its bounded query, which is what a working queue looks like.