Oncofusion

start / graph / exp-menin-inhibitor-h3k27me3-timecourse-index-fusion

Experimentspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q41-prc2-arm

Measure H3K27me3 after a menin inhibitor - the free pharmacodynamic nobody has read, and the one experiment that separates two mechanisms

also called menin inhibitor PRC2 pharmacodynamic · H3K27me3 timecourse under revumenib

Two different explanations exist for why menin inhibitors switch silenced genes back on, and they predict opposite things about the Polycomb 'off' tag. The early part of that measurement has now been found in public data for a different fusion, so what remains is to do it in the fusion this project is about, and for longer.

The same thing, for a scientist

**Design.** SEM and RS4;11 (KMT2A::AFF1) with MV4-11 (KMT2A::MLLT3) as a partner control and an infant-derived KMT2A::AFF1 PDX if available. Treat with a menin inhibitor (revumenib, and MI-538 to match PMID 40722046) at a fixed, target-engagement-verified exposure. Sample at 0, 6, 24, 48, 72 and 120 h. **Readouts, all calibrated:** spike-in-normalised CUT&RUN for **H3K27me3, H3K4me3, EZH2 and SUZ12**; western for EZH2 and EZH1 protein; RNA-seq. **Anchor loci: *HDAC7*** (the index-disease effector named by PMID 40722046) and a set of bivalent genes including **MHC class I** (the readout of PMID 36635503). **What it decides.** *Route 1* (`hyp-prc2-is-the-fusions-repressive-arm`): menin drives *Ezh2*, so EZH2 protein falls, EZH2/SUZ12 occupancy and H3K27me3 fall at *HDAC7*, and the fall lags the transcriptional change because H3K27me3 is lost by dilution. *Route 2* (PMID 36635503): PRC2 occupancy is roughly intact, H3K4me3 **rises at bivalent genes and falls at active genes**, and MHC class I comes up. The two are distinguishable in one experiment and are currently indistinguishable in every published one, because **no study reports H3K27me3 under a PHARMACOLOGICAL menin inhibitor** - Sparbier et al. 2023 (PMID 36635503) measured it in the MEN1-knockout arm and found total H3K27me3 unchanged, while its VTP50469 arm carries no H3K27me3, so the experiment below remains unperformed. **Cost tier: low.** The cell lines, the compounds and the antibodies are all standard, and the spike-in calibration is the only non-routine element - which this graph already insists on, having recorded that uncalibrated ChIP normalises away exactly the global change this experiment is looking for.

Go deeper
Controls, and the three ways this experiment could mislead

Controls:

  • Spike-in calibration is mandatory, not optional. A menin inhibitor is expected to change global chromatin state, and uncalibrated ChIP or CUT&Tag normalises global changes away by construction - this graph's own mark-h3k4me3 entry records a 2011 study whose four-mark negative result may be exactly that artefact.
  • A menin-binding-defective control: an inactive analogue, or a MEN1 M327I resistance-mutant line, so that an off-target effect of the compound cannot be read as menin biology.
  • A partner control: MV4-11 (KMT2A::MLLT3, myeloid) against SEM/RS4;11 (KMT2A::AFF1, B-lineage), because the two published effectors of the repressive arm differ by lineage (C/EBPalpha versus HDAC7) and this graph's standing finding is that KMT2A epigenetics results are usually myeloid.
  • A PRC2-inhibitor arm: valemetostat alone, to establish what the HDAC7 locus does when PRC2 catalysis is removed directly. Without it, a negative H3K27me3 result cannot distinguish 'menin does not license PRC2 here' from 'PRC2 was never on this locus'.

Three ways it misleads:

  1. Differentiation confounds everything after 72 h. Menin inhibition drives differentiation, and a differentiating cell changes H3K27me3 globally for reasons unrelated to the axis. The informative window is the first 24-48 h, and late timepoints must be interpreted against a differentiation-marker panel.
  2. H3K27me3 is diluted, not erased. The mark's loss after the writer is removed is replication-coupled, so a cell-cycle-arrested population will show no change even if the mechanism is real. Cell-cycle state must be measured, not assumed.
  3. EZH1 fills in. If menin-driven EZH2 falls and EZH1 covers the locus, H3K27me3 will not move even though Route 1 is correct. That is why EZH2 protein and EZH2 occupancy are primary readouts alongside the mark, and why the valemetostat arm matters.
Correction 2026-09-20: the H3K27me3 absence is narrower than this entry said

This entry asserted that no menin-inhibitor study in any system reports an H3K27me3 measurement. Read as written that is false, and the counter-example is the paper this graph already cites for the polycomb phenocopy.

Sparbier et al. 2023 (PMID 36635503, Nature Cell Biology, full text PMC7614190) profiled H3K27me3 by CUT&Tag in MEN1-knockout versus control cells, reporting total H3K27me3, H3K4me3, H2AK119Ub, SUZ12 and EZH2 unchanged after MEN1 knockout, and at bivalent non-Menin-bound promoters increased H3K4me3 with reduced H3K27me3 and SUZ12.

The measurement is in the GENETIC arm, not the drug arm. The paper's VTP50469 experiments carry Menin ChIP-seq and RNA-seq and no H3K27me3.

So the defensible statement is the narrow one: no H3K27me3 measurement under a PHARMACOLOGICAL menin inhibitor has been reported (Sparbier et al. 2023, PMID 36635503, measured H3K27me3 by CUT&Tag in the MEN1-KNOCKOUT arm and found total H3K27me3 unchanged; its VTP50469 arm carries Menin ChIP-seq and RNA-seq and no H3K27me3). The experiment this graph keeps asking for - H3K27me3 under a clinical menin inhibitor, in the index fusion - is still unperformed, and knockout is not a substitute, because a drug that occupies a pocket and a deletion that removes the protein are different perturbations with different kinetics.

Correction 2026-09-20: the absence narrows a SECOND time

This entry's narrowed claim - no H3K27me3 measurement under a pharmacological menin inhibitor - is also false. PMID 38173411 treated human adipose-derived stem cell spheroids with 70 uM SNDX-5613 (revumenib) and blotted H3K27me3: "had no effect on H3K27me3".

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 Western blot exists, in a non-leukaemic system, and it was negative. Full reasoning and the bounded query in ana-the-h3k27me3-absence-narrows-a-second-time.

Facts and figures
designMenin-inhibitor timecourse in KMT2A::AFF1 lines (SEM, RS4;11) with a KMT2A::MLLT3 partner control (MV4-11) and an infant PDX if available; revumenib and MI-538 at target-engagement-verified exposure; sampling at 0, 6, 24, 48, 72, 120 h; arms include vehicle, an inactive analogue or MEN1 M327I resistance-mutant control, and a valemetostat-alone arm.
readoutSpike-in-calibrated CUT&RUN for H3K27me3, H3K4me3, EZH2 and SUZ12; western blot for EZH2 and EZH1 protein; RNA-seq; cell-cycle and differentiation-marker panels as interpretive controls. Anchor loci: HDAC7 and a bivalent set including MHC class I.
controlsspike-in calibration (mandatory); inactive analogue or MEN1 M327I resistance mutant; partner-fusion control for lineage; valemetostat-alone arm to establish the locus's behaviour under direct PRC2 inhibition; cell-cycle measurement because H3K27me3 loss is replication-coupled
cost tierlow
timeline3-4 months
why not doneNo menin-inhibitor study in any system reports an H3K27me3 measurement, despite the menin-EZH2 link being in print since 2013 and menin inhibitors having been in clinical development for a decade.
decision rule
{
  "Route_1_supported": "EZH2 protein falls; EZH2/SUZ12 occupancy and H3K27me3 fall at HDAC7, lagging the transcriptional change",
  "Route_2_supported": "PRC2 occupancy roughly intact; H3K4me3 rises at bivalent genes and falls at active genes; MHC class I de-repressed",
  "both": "the two routes operate at different loci, which is the most likely outcome and would still be the first measurement of either in this disease",
  "neither": "menin inhibition de-represses HDAC7 by a PRC2-independent route, and hyp-prc2-is-the-fusions-repressive-arm is falsified"
}
tests
[
  "hyp-prc2-is-the-fusions-repressive-arm"
]
the early half already exists found cycle 56
{
  "series": "GSE260742 (ChIP-seq subseries), PMID 41634341, Nature Cell Biology 2026",
  "timecourse_arm_verbatim": [
    "Timecourse Control - DMSO H3K27me3",
    "Timecourse Control - VTP50469 500nM 8h H3K27me3",
    "Timecourse Control - VTP50469 500nM 24h H3K27me3",
    "Timecourse Control - DMSO H2AK119ub",
    "Timecourse Control - VTP50469 500nM 8h H2AK119ub",
    "Timecourse Control - VTP50469 500nM 24h H2AK119ub",
    "Timecourse Control - VTP50469 500nM 4h H3K79me2",
    "Timecourse Control - VTP50469 500nM 8h H3K79me2",
    "Timecourse Control - VTP50469 500nM 16h H3K79me2",
    "Timecourse Control - VTP50469 500nM 24h H3K79me2"
  ],
  "n_h3k27me3_samples": 26,
  "arms": "MLLAF9 (murine FLAG), MOLM13 (NT and PB), NTvKO PCGF1 KO, K562, HUDEP-2, mESC, Timecourse",
  "no_index_fusion_line": "Checked for AFF1, AF4, RS4;11, SEM and MV4;11 across all 158 sample titles: zero. My first check omitted MV4;11 and had to be re-run -- MV4;11 IS a KMT2A::AFF1 line and the paper names it 24 times, but not in this ChIP subseries.",
  "what_this_changes": "The 0-24 h window of this design exists for MLL-AF9. What remains novel is the index fusion (SEM, RS4;11, an infant PDX) and the 48-120 h window, plus the EZH2/SUZ12 occupancy and protein readouts, which GSE260742 does not contain.",
  "a_geo_metadata_trap_found_alongside": "GSE260742's Series_pubmed_id field lists TWO ids: 41634341 (correct) and 12904788 - which is a real PubMed record for 'Low-loss hollow-core silica/air photonic bandgap fibre', Nature 2003. GEO appears to have put the PMC id (PMC12904788) into a pubmed_id field. This graph is NOT affected: its two occurrences of that digit string are the correct PMC12904788 accession, and a bare-digit grep matching them was my false alarm, not a bad citation."
}
How this could bite us back
  • A negative result would be ambiguous unless the cell-cycle and EZH1 controls are in place: H3K27me3 that does not move is equally consistent with 'menin does not license PRC2 here' and with 'the cells stopped dividing so the mark was not diluted'.
  • If Route 2 wins, the implication for infants is uncomfortable: menin inhibition activates bivalent genes broadly, including in normal and pluripotent cells, and that is a developmental exposure no response-rate endpoint would detect.
  • This entry's central absence claim has now been narrowed TWICE, on 2026-09-20 both times. Treat the remaining claim as provisional and re-run the bounded query before quoting it.
Still open
  1. Can this be run on banked patient samples from a menin-inhibitor trial rather than in cell lines? Paired pre- and on-treatment marrow from revumenib or ziftomenib trials would give the human answer directly, and CUT&RUN works on small cell numbers - the obstacle is sample access, not method.
  2. Does the full text of PMID 40722046 already contain the EZH2 ChIP or H3K27me3 measurement at HDAC7? The abstract does not report it and this beat could not read the full text; scoping this experiment should start there, because a substantial part of it may already exist.
Every source behind this entry (4)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Menin binds the Ezh2 promoter and promotes its expression, and menin depletion reduces EZH2 expression and EZH2 binding and repressive H3K27 methylation at C/EBPalpha target genes in MLL-AF9 leukaemia - the observation this experiment would test in a human index-fusion system.

    PMID 23349306DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. Mouse MLL-AF9, genetic menin depletion rather than a menin inhibitor, myeloid effector. The result is thirteen years old and has not been reproduced in a human system or with a clinical-grade menin inhibitor, which is the gap this experiment addresses.

    Evidence that argues against this (2)
    • A competing mechanism predicts PRC2 occupancy is roughly intact after menin inhibition and that the moving mark is H3K4me3 at bivalent genes. Both mechanisms predict de-repression, so an experiment measuring only expression would confirm neither - which is the design constraint that makes the chromatin readouts primary here. Sparbier CE et al. 2023
    • SUPERSEDED 2026-09-20 (second narrowing). A pharmacological menin inhibitor HAS now been assayed against H3K27me3: PMID 38173411 treated human adipose-derived stem cells with 70 uM SNDX-5613 (revumenib) and reported by Western blot that it 'had no effect on H3K27me3'. The defensible statement is now narrower again - no LOCUS-RESOLVED H3K27me3 measurement under a pharmacological menin inhibitor in a KMT2A-REARRANGED LEUKAEMIA - and the single existing datapoint is negative, in a non-leukaemic system, by bulk blot. See [[ana-the-h3k27me3-absence-narrows-a-second-time]]. Chang MM et al. 2024
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In t(4;11) infant B-ALL, HDAC7 is silenced by EZH2 together with KMT2A::AFF1 and the menin inhibitor MI-538 restores HDAC7 expression - providing the specific anchor locus and the specific compound for this experiment, while leaving the chromatin step unmeasured.

    HDAC7 induction combined with standard-of-care chemotherapy provides a therapeutic advantage in t(4;11) infant B-cell acute lymphoblastic leukemia. · de Barrios O, Ocón-Gabarró I, Gusi-Vives M, Collazo O, Meler A, Romecín PA, et al. · Biomark Res 2025
    PMID 40722046DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. The abstract does not report EZH2 ChIP or H3K27me3 at HDAC7 and runs no PRC2 inhibitor arm, which is precisely why this experiment is proposed. It is possible the full text contains some of these measurements; this beat could not read it and the experiment should be scoped against the full text before being run.

    Evidence that argues against this (1)
    • Rule 65 applies to the anchor locus. HDAC9 carries HDAC7 as a synonym and its protein name reads 'Histone deacetylase 7B'; in this disease HDAC7 is low and to be induced while HDAC9 is high and to be repressed, so primers, antibodies and CUT&RUN peak calls must be validated against HDAC7 = Q8WUI4 (952 aa) and not merely against the symbol. UniProt 2026
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    H3K27me3 change from baseline has already been used as a pharmacodynamic endpoint in a 139-patient first-in-human phase 1/2 of an EED-directed PRC2 inhibitor, so the assay exists at clinical scale and the readout proposed here is not novel technology.

    A first-in-human phase 1/2 dose-escalation study of MAK683 (EED inhibitor) in patients with advanced malignancies. · Ribrag V, Iglesias L, De Braud F, Ma B, Yokota T, Zander T, et al. · Eur J Cancer 2025
    PMID 39793445DOINCT02900651
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. The abstract states pharmacodynamic profiles were measured by H3K27me3 change from baseline but does not state the tissue, the assay or whether the endpoint tracked response - so 'the assay exists' is a weaker claim than 'a validated assay exists'. No leukaemia cohort; adults with solid tumours and lymphoma.

    Evidence that argues against this (1)
    • A bulk H3K27me3 endpoint sums the EZH1- and EZH2-deposited pools and cannot resolve locus-level redistribution, which is exactly what both mechanisms here predict. The clinical precedent supports feasibility, not sufficiency - this experiment's readouts must be locus-resolved and spike-in calibrated, not a bulk blot. Neff T et al. 2012
  4. E_k56_partly_donestrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    THE EARLY TIMEPOINTS OF THIS EXPERIMENT ALREADY EXIST, in another fusion. GSE260742, the ChIP-seq subseries of PMID 41634341 (Neville D, Ferguson DT, Heikamp EB, et al., 'DOT1L provides transcriptional memory through PRC1.1 antagonism', Nature Cell Biology 2026), contains 158 ChIP samples including an explicit timecourse arm under VTP-50469 at 500 nM: H3K27me3 at DMSO, 8 h and 24 h; H2AK119ub at the same three points; and H3K79me2 at DMSO, 4, 8, 16 and 24 h. Twenty-six H3K27me3 samples in total, spanning DMSO, VTP-50469 500 nM and the DOT1L inhibitor SGC0946 5 uM, in murine FLAG-MLLAF9, MOLM13 (with a PCGF1-knockout arm), K562, HUDEP-2 and mESC. CRITICALLY FOR THIS ENTRY: the ChIP subseries contains NO KMT2A::AFF1 line - checked for AFF1, AF4, RS4;11, SEM and MV4;11, all zero - so the index-fusion scope of this experiment survives intact. What does not survive is the claim that the measurement has never been done. This graph cites PMID 41634341 in 36 entries.

    DOT1L provides transcriptional memory through PRC1.1 antagonism. · Neville D, Ferguson DT, Heikamp EB, et al. · Nature Cell Biology 2026
    PMID 41634341DOI
    What this evidence cannot show

    Established from the GEO sample titles of GSE260742, read 2026-09-22, not from the paper's figures: what those H3K27me3 tracks SHOW at 8 and 24 h was not analysed here, only that they exist. The timecourse arm's cell system is labelled only 'Timecourse Control' and is inferred to be the murine MLLAF9 system from the series design statement rather than from the sample title, so the line identity of the timecourse specifically is not certain. 8 h and 24 h do not cover this design's 48, 72 and 120 h points, and the existing data is spike-in status unknown.

    Evidence that argues against this (2)
    • This makes the proposed experiment BETTER justified and cheaper, not redundant: there is now a direct comparator in a non-AFF1 fusion, so the index-fusion arm becomes a difference measurement rather than a first look. But it also means the novelty claim in this entry's original framing was wrong, and the honest reading is that the project proposed an experiment whose early half was sitting in a dataset it cites 36 times.
    • A reader could conclude the reverse - that because the early timepoints exist in MLL-AF9, the index-fusion version is unnecessary. That would be wrong for the reason this whole project exists: the 5' partner determines the complex and the address, MLL-AF9 is myeloid and AFF1 is lymphoid, and the two decision routes this experiment discriminates were both established in AFF1-context systems.
Connected to