ZMYND11::MBTD1 AML is menin-inhibitor sensitive despite having no KMT2A rearrangement, because the translocation deletes ZMYND11's MYND domain and with it a physical brake on wild-type KMT2A
also called the deleted brake hypothesis · revumenib in a fusion with no KMT2A
A rare leukaemia is caused by joining two genes, neither of which is the gene the approved drug revumenib targets - so nobody expects the drug to work on it. This entry argues the opposite: the gene at the front of that fusion normally holds back revumenib's target, and the join throws away exactly the piece that does the holding, which should leave the target running free and the drug useful.
The same thing, for a scientist
**The claim.** ZMYND11::MBTD1 AML responds to a menin inhibitor, and does so for a reason that has nothing to do with a KMT2A fusion. **The argument in four steps, each with a source.** (1) Wild-type ZMYND11 physically restrains KMT2A: it co-immunoprecipitates KMT2A and NSD1, deleting its MYND domain collapses most of its interactome, the commonest disease-associated missense R600W selectively strips KMT2A and NSD1, *Zmynd11*-knockout cortex gains H3K4me3 specifically at ZMYND11-bound transcription start sites, and full-length ZMYND11 inhibits an rTetR-MLL-C-driven luciferase reporter [E1]. (2) Losing ZMYND11 is therefore partly a KMT2A gain-of-function, and the transcriptional consequences of degrading ZMYND11 in primary cortical neurons are **significantly attenuated by revumenib** [E1]. (3) UniProt places the MYND-type zinc finger at **563-598** of a 602-residue protein, and the leukaemic fusion retains only ZMYND11 **aa 1-409** [E2][E3]. (4) The translocation is balanced, so one *ZMYND11* allele is disrupted and the chimera it produces lacks the brake while keeping the H3.3K36me3-reading head that puts it at ZMYND11's addresses [E3]. **So the prediction is not 'the drug hits the fusion'. It is 'the drug hits the wild-type KMT2A that the fusion has stopped restraining, at the coordinates the fusion occupies.'** That is a different mechanism from everything in this graph's menin arm, and it is testable with commercial reagents. **It directly contradicts the expectation written in this graph one cycle ago.** `fusion-zmynd11-mbtd1` open question 5 says: *'Is ZMYND11::MBTD1 AML menin-independent? Nobody has asked. It has no CXXC domain and no menin-binding motif, so the expectation is yes.'* That reasoning is sound and this hypothesis says it is incomplete, because it reasons only about the fusion protein and not about what the translocation removed. **And it converges with the corpus's current strongest threat.** Cycle 82's salvaged K562 arm found a menin inhibitor moving H3K79me2 about as much in a line with **no KMT2A fusion** as in fusion-bearing murine cultures - 0.0057 from formally refuting fusion-dependence. If menin inhibitors act materially on wild-type KMT2A, this hypothesis is the mechanistic story for why, and the two results are independent. **If it holds, the drug's indication widens and the argument that its selectivity comes from the fusion weakens - and those are opposite-signed consequences for this laboratory.** **Prior probability 0.2.** The mechanistic chain is clean but every link is in the wrong system: a preprint, mouse neurons, HEK293T, and a domain-boundary inference nobody has tested on the fusion protein.
What has to be true, and which links are weakest
| link | status | weakest point |
|---|---|---|
| ZMYND11 binds KMT2A | shown, in mouse brain and HEK293T, in a preprint 1 | not peer reviewed; the mutant comparison is on overexpressed protein |
| the binding is MYND-dependent | shown, by dMYND and R600W 1 | dMYND collapses most interactions, so it may destabilise rather than delete a specific contact |
| the fusion lacks the MYND domain | inference: 409 < 563 23 | a feature-table interval is not a folding unit; no co-immunoprecipitation on the fusion protein or on aa 1-409 exists |
| losing ZMYND11 de-restrains KMT2A | shown as an H3K4me3 gain at ZMYND11-bound TSSs in knockout cortex 1 | cortex, not blood; a mark gain is not a transcription-factor-activity measurement |
| blocking KMT2A helps | shown: revumenib attenuates the ZMYND11-degradation transcriptome 1 | partial, 24 h, one timepoint, neurons |
| the same holds in a fusion-bearing myeloid cell | never tested | no human ZMYND11::MBTD1 cell line exists |
The link to attack first is row three, because it is the only one that is pure inference and the only one that can be settled by a single blot. If the fusion protein still binds KMT2A, the hypothesis inverts: the fusion would be retaining a brake as well as adding an accelerator, and a menin inhibitor's effect would be unpredictable in sign.
The link that decides whether any of this reaches a patient is row six, and it has a reagent problem this entry cannot solve. There is no human ZMYND11::MBTD1 line. The only available system is the murine transduced-HSPC model from the founding paper, in which wild-type Zmynd11 is intact on both alleles and the fusion is overexpressed - which is the wrong genotype for this hypothesis, because the hypothesis is partly about the disrupted allele. exp-revumenib-and-menin-eviction-in-zmynd11-mbtd1-aml states the design and says so.
Three rival explanations that would produce the same observation
If a menin inhibitor turned out to work in a ZMYND11::MBTD1 model, at least three other stories explain it, and the experiment has to separate them.
Rival 1: the drug is simply active against wild-type KMT2A in myeloid cells generally, fusion or no fusion. This is the cycle-82 K562 reading, and it predicts the same response in a ZMYND11-wild-type myeloid line of similar lineage. Discriminator: the effect should be no larger in the fusion-bearing culture than in a matched fusion-free one. If it is larger, rival 1 is insufficient.
Rival 2: MEN1 or KMT2A is a dependency of the cell state rather than of the lesion. MEN1 outscores KMT2A itself in DepMap in KMT2A's own lines, and this graph already records that. A HOXA-high, differentiation-blocked myeloid progenitor may need the menin-KMT2A axis whatever put it in that state. Discriminator: the response should track HOXA-programme dependence, not fusion genotype, and should appear in NUP98-fusion and CALM-AF10-like HOXA-high models too.
Rival 3: off-target or differentiation-agnostic toxicity. Revumenib at high concentration is not a clean probe. Discriminator: a MEN1 M327I-type binding-site mutant, or MEN1 knockout, should phenocopy the drug. If genetic MEN1 loss does nothing while the drug does something, the drug is not working through menin.
And the discriminator specific to THIS hypothesis, which none of the rivals predicts: the response should depend on ZMYND11 status, not on the fusion. Restoring full-length ZMYND11 - or the MYND domain alone, which represses without activating 1 - should reduce the drug's effect, because it restores the brake the drug is substituting for. That is the one prediction that is this hypothesis and nothing else.
| statement | ZMYND11::MBTD1 AML is sensitive to menin-KMT2A interface inhibition, not because the fusion engages menin - it has no CXXC domain and no menin-binding motif - but because the translocation deletes ZMYND11's MYND-type zinc finger (563-598; the fusion retains only aa 1-409) and with it a physical restraint on wild-type KMT2A, leaving KMT2A de-restrained at the H3.3K36me3-marked addresses the fusion's retained PWWP occupies. |
|---|---|
| origin | literature-gap |
| origin note | Two published literatures that do not cite each other. The ZMYND11-KMT2A restraint is a 2025 preprint on neurodevelopment that never mentions MBTD1 or any fusion; the fusion boundary is a 2021 leukaemia paper that does not know ZMYND11 binds KMT2A. MBTD1[tiab] AND KMT2A[tiab] returns 0 records. |
| prior probability | 0.2 |
| prior note | The mechanistic chain is unusually clean for a novel hypothesis and every link sits in the wrong system: a preprint, mouse cortical neurons, HEK293T, and one step that is pure domain-boundary arithmetic with no experiment behind it. 0.2 rather than lower because the revumenib rescue is a direct pharmacological result in the right direction, and rather than higher because no haematopoietic cell has ever been involved. |
| what would make this project care | If falsifiers 1 and 3 both point the hypothesis's way, then a menin inhibitor works in a leukaemia with no KMT2A rearrangement, through de-restrained wild-type KMT2A. That widens the drug's indication and simultaneously weakens the claim that its selectivity derives from the fusion - opposite-signed consequences, and the second is the one this laboratory's central hypothesis is exposed to. It would also supply the mechanism for cycle 82's K562 result, which found a menin inhibitor moving H3K79me2 in a fusion-free line at 0.0057 from formally refuting fusion-dependence. |
| why the only available model is the wrong one | There is no human ZMYND11::MBTD1 cell line. The founding murine model transduces a fusion cDNA into wild-type HSPCs, so both Zmynd11 alleles are intact and the fusion is overexpressed. That genotype cannot test a hypothesis about a disrupted allele. An honest version of falsifier 3 in that model tests the CONVERSE - whether adding ZMYND11 back on top of an already-wild-type background changes the drug response - which is a weaker experiment, and the experiment entry says so. |
| absence class | never-measured |
| absence class note | Nobody has given a menin inhibitor to any ZMYND11::MBTD1 system, and nobody has asked whether the fusion protein binds KMT2A. Both need new material - constructs, cells, drug - so never-measured. ZMYND11[tiab] AND MBTD1[tiab] AND menin[tiab] returns 0, and the drug-name arm of that query was NOT run, which is a rule-72 limitation stated in E4's caveats. One sub-arm is cheaper than never-measured and is named rather than claimed: the deposited ZMYND11::MBTD1 genomic datasets could be asked whether fusion peaks coincide with known menin or MLL1-N occupancy in any published dataset, which would be measured-not-analysed - but no accession was looked up in this beat and none is named. |
predictions
[ "A menin inhibitor reduces proliferation and induces myeloid differentiation in a ZMYND11::MBTD1-transformed myeloid culture, at a potency comparable to or greater than in a matched fusion-free culture of the same lineage.", "MEN1 genetic ablation phenocopies the drug in that culture. If the drug works and MEN1 knockout does not, the effect is not through menin and the hypothesis is not what is being tested.", "Re-expressing full-length ZMYND11, or the MYND domain alone, REDUCES the menin inhibitor's effect, because it restores the brake the drug is substituting for. This is the prediction unique to this hypothesis and shared with none of its rivals.", "The ZMYND11::MBTD1 fusion protein does NOT co-immunoprecipitate KMT2A, and neither does ZMYND11 aa 1-409, while full-length ZMYND11 does.", "H3K4me3 is elevated at ZMYND11-bound transcription start sites in fusion-bearing cells relative to matched controls, and falls under menin inhibition - the haematopoietic counterpart of the Zmynd11-cKO cortex result." ]
experiments
[ "exp-revumenib-and-menin-eviction-in-zmynd11-mbtd1-aml" ]
analyses
[]
falsifiers
[ "SCOREABLE FALSIFIER 1, the cheapest and the one to run first. In HEK293T co-transfected with FLAG-tagged full-length ZMYND11, FLAG-ZMYND11 aa 1-409 and FLAG-ZMYND11::MBTD1, anti-FLAG immunoprecipitation followed by KMT2A immunoblot: if the 1-409 fragment or the fusion recovers KMT2A at 50% or more of the full-length signal, the central premise is REFUTED and the hypothesis fails. Fires on a ratio, in one cell line, in one blot.", "SCOREABLE FALSIFIER 2. In a ZMYND11::MBTD1-transformed murine myeloid culture treated with VTP-50469 at 300 nM for 6 days, against a lineage-matched fusion-free culture treated identically: if the fusion-bearing culture's viability reduction is NOT greater than the comparator's by at least 25 percentage points, the hypothesis is not distinguishable from rival 1 (the drug acts on wild-type KMT2A in myeloid cells generally) and must be reported as such rather than as support.", "SCOREABLE FALSIFIER 3, the one unique to this hypothesis. In the same culture, doxycycline-inducible re-expression of full-length ZMYND11 at approximately endogenous level: if the menin inhibitor's effect on viability is UNCHANGED (within 10 percentage points) by ZMYND11 restoration, then the drug is not substituting for the deleted brake and the mechanism proposed here is REFUTED even if the drug works.", "SCOREABLE FALSIFIER 4. If MEN1 knockout in the fusion-bearing culture produces no viability or differentiation phenotype while VTP-50469 does, the drug effect is off-target and the hypothesis is untested rather than refuted - but it must then be reported as an instrument failure and not as a negative result.", "SCOREABLE FALSIFIER 5, on the human side and requiring no wet work. If a patient with karyotype-confirmed t(10;17)(p15;q21)/ZMYND11::MBTD1 AML has received a menin inhibitor with no response, that is a single-patient refutation of the clinical claim. No such case appears in the twelve-to-thirteen case literature, and finding one is a question for the authors of the 2026 series rather than an experiment." ]
- This hypothesis is a threat to its own laboratory's central claim, and that is the honest reason to run it. If a menin inhibitor works in a leukaemia with no KMT2A rearrangement, then 'the drug works because the fusion cannot let go of menin' stops being the explanation, and the selectivity story that justifies the interface programme has to be rebuilt around something else.
- A drug that acts by substituting for a deleted tumour-suppressor brake is acting on wild-type machinery in every cell that has it. There is no fusion-only interface in this scenario, so the therapeutic index argument that makes menin inhibitors attractive does not transfer - and the on-target toxicity would land wherever KMT2A restrains development, which the same preprint says is the brain.
- If the prediction is right for the wrong reason - rival 2, a HOXA-high differentiation-blocked state needing the menin-KMT2A axis whatever caused it - then menin inhibitors are a state-directed rather than a lesion-directed therapy. That is a larger indication and a much weaker biomarker, and it would make fusion genotype the wrong patient-selection criterion for a drug approved on fusion genotype.
- The experiment cannot be done properly. There is no human cell line, the murine model has the wrong genotype, and the disease has twelve or thirteen reported cases. Any result will be a result about a transduced mouse progenitor, and this project's commonest documented synthesis failure is generalising exactly that.
- Does the ZMYND11::MBTD1 fusion protein bind KMT2A? One anti-FLAG immunoprecipitation with a KMT2A blot, three constructs, one cell line. It is falsifier 1, it is the cheapest experiment in this entry, and it decides whether the rest is worth doing.
- Is wild-type ZMYND11 expressed from the untranslocated allele in fusion-positive patient blasts? The translocation is balanced and nobody has looked. If the remaining allele is silenced or the fusion is dominant-negative over it, the de-restraint is larger than gene dosage alone would predict; if the remaining allele compensates, the hypothesis weakens. This is answerable from any RNA-sequencing already performed on a fusion-positive case.
- Has any ZMYND11::MBTD1 patient received a menin inhibitor? Several of the twelve-to-thirteen reported patients relapsed after transplant in the era when revumenib was available. A question to the authors of the 2026 Haematologica series, not an experiment, and a single non-responder refutes the clinical claim.
- Does the ZMYND11-KMT2A contact exist in a haematopoietic cell at all? Every demonstration is neuronal or HEK293T. If ZMYND11 does not bind KMT2A in a myeloid progenitor - because a competing partner occupies the MYND surface there, for instance - the whole chain breaks in a cell type nobody has examined.
- Does NSD1 belong in this story? R600W strips NSD1 as well as KMT2A, and NUP98::NSD1 is a fusion this graph already holds. A ZMYND11 that restrains both KMT2A and NSD1, losing its brake in a fusion, sits upstream of two of this graph's fusion families at once - and gene-nsd1 does not exist as an entry.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.preprint
In a 2025 preprint, ZMYND11 co-immunoprecipitates KMT2A and NSD1 from mouse brain and HEK293T; deletion of the MYND domain drastically reduces most ZMYND11 interactions and the ZRSID-associated R600W substitution selectively removes the top interactors including KMT2A and NSD1; Zmynd11 conditional-knockout cortex gains H3K4me3 at highly-ZMYND11-bound transcription start sites but not at lowly-bound ones with no change in H3K4me1, and strongly-bound genes are up-regulated; full-length ZMYND11 inhibits activation of a 9xTetO-minCMV-luciferase reporter driven by rTetR-MLL-C while the MYND domain alone represses without activating; and 24-hour dTAG degradation of endogenous ZMYND11 in primary cortical neurons produces transcriptional defects that revumenib significantly attenuates, with revumenib's own small effects anti-correlated with those of ZMYND11 loss and no rescue of dTAG off-target genes in wild-type neurons.
ZMYND11 Restrains KMT2A to Enable a Neuronal Developmental Program. · Greben AW, Wu XS, Robb JE, Vogel IR, et al. · bioRxiv 2025What this evidence cannot show
A PREPRINT, not peer reviewed, and filed at moderate for that reason alone. Full text read this session via PMC12636335. THE SYSTEM IS ENTIRELY NEURONAL AND HETEROLOGOUS: mouse cortical neurons, mouse cortex, HEK293T. No haematopoietic cell appears in it. The revumenib rescue is explicitly PARTIAL, at one timepoint. The mutant interactome comparison is FLAG-IP of overexpressed ZMYND11, and the authors themselves note overexpression recovers lower-affinity interactions. The paper does not mention MBTD1 or any fusion - zero occurrences of MBTD1 in its full text - so its application to ZMYND11::MBTD1 is entirely this entry's inference. gene-zmynd11 E5 already cites this record at abstract level.
Evidence that argues against this (1)
- The authors' own framing makes a competing prediction awkward for this hypothesis. They position ZMYND11 as a tumour suppressor whose loss phenocopies part of a KMT2A gain of function - which means a ZMYND11-disrupting translocation could be oncogenic through de-restrained KMT2A quite independently of the chimeric protein, and in that case the fusion protein and the drug target sit in different causal chains. This hypothesis needs both to be true at the same coordinates, and nothing shows that. Wen H et al. 2014
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
UniProt Q15326 (ZMY11_HUMAN) is 602 amino acids with a MYND-type zinc finger at residues 563-598, a PWWP domain at 280-331, a bromodomain at 149-255, a PHD-type zinc finger at 100-148 and a SAMD1-like winged-helix domain at 6-82.
UniProtKB Q15326 (ZMY11_HUMAN), Zinc finger MYND domain-containing protein 11; feature table fields ft_domain and ft_zn_fing · UniProt 2026What this evidence cannot show
Fetched live 2026-09-23 with an explicit field list. Domain boundaries are curation and are revised between releases. Note that residue 600 - the ZRSID hotspot whose substitution abolishes KMT2A binding - lies OUTSIDE the 563-598 feature interval, and the preprint describes it as at the C-terminal end of the MYND domain; the structural element and the feature interval need not coincide, and this entry's argument uses the retained/lost boundary at 409 rather than the exact MYND limits, so the discrepancy does not affect it.
Evidence that argues against this (1)
- The whole hypothesis rests on treating a feature-table interval as a folding and binding unit, which is the same class of error this project has logged for PDB residue numbering and for ASH2L's isoform offset. If the KMT2A-binding surface extends N-terminal of residue 409, or if the MYND domain stabilises a contact made elsewhere - which the preprint's own dMYND result permits, since it collapses most interactions rather than one - then the fusion could retain KMT2A binding and the hypothesis fails at its cheapest link. Greben AW et al. 2025
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The functional ZMYND11-MBTD1 construct is ZMYND11 amino acids 1-409 joined through a 16-residue UTR-derived segment to the entire 628-residue MBTD1 open reading frame; the fusion requires an intact H3.3K36me3-binding PWWP domain and the Tip60-interaction surface for transformation of murine haematopoietic stem/progenitor cells, and the resulting AML is sensitive to the BET inhibitor I-BET151 in vitro and in vivo.
ZMYND11-MBTD1 induces leukemogenesis through hijacking NuA4/TIP60 acetyltransferase complex and a PWWP-mediated chromatin association mechanism. · Li J, Galbo PM, Gong W, Storey AJ, et al. · Nat Commun 2021What this evidence cannot show
Murine primary HSPCs with a retrovirally overexpressed human cDNA and syngeneic transplant. CRITICALLY FOR THIS HYPOTHESIS, that model has TWO INTACT WILD-TYPE Zmynd11 ALLELES and an overexpressed fusion, so it is the wrong genotype for a hypothesis that is partly about the disrupted allele - the only available system cannot cleanly test the claim. The construct boundary 1-409 is the construct's, and whether the patient transcript's boundary is identical is not established: the construct sums to 1,053 residues against a transcript-derived 1,054, a one-residue discrepancy neither source explains. Full text read via PMC7886901.
Evidence that argues against this (1)
- The published therapeutic lead for this fusion is BET inhibition, which works in the murine model and has a complete downstream mechanism. A menin inhibitor prediction has to compete with an agent that already has an efficacy result in the same system, and if both work the experiment cannot easily attribute the effect - both converge on the same transcriptional output. The design has to include a BET arm for that reason, and a genetic MEN1 arm to show the menin effect is menin's. Devoucoux M et al. 2022
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
Bounded literature absence, 2026-09-23: no indexed title or abstract has asked this question. ZMYND11[tiab] AND MBTD1[tiab] AND menin[tiab] returns 0 records; MBTD1[tiab] AND KMT2A[tiab] returns 0; and ZMYND11[tiab] AND KMT2A[tiab] returns 4, of which three are Cornelia-de-Lange or neurodevelopmental-disorder gene lists in which both symbols appear incidentally and the fourth is the preprint cited as E1.
What this evidence cannot show
One [tiab] tag per term, never the parenthesised group form; every querytranslation read back; pubmed_safe.py raises rather than returning a zero on an empty translation, so these are genuine zeros for the strings searched. THE DRUG-NAME ARM WAS NOT RUN, and under rule 72 that is a real limitation: 'menin' alone was the operand, not revumenib, ziftomenib, VTP-50469, SNDX-5613, bleximenib or the nominalisations 'menin inhibitor', 'menin inhibitors', 'menin inhibition'. A paper reporting a revumenib result in this fusion without using the word menin in its abstract would be missed. Titles and abstracts only; rule 80 applies.
Evidence that argues against this (1)
- The drug-name gap is not hypothetical for this question. Rule 72 exists in this project because a menin-name list returned 0 where the bare operand returned 8, and the failure was in the name list rather than the query mechanics. A menin-inhibitor result in a rare fusion is exactly the kind of thing that would be published as a case report naming only the trade or code name, so this absence is weaker than the clean zeros make it look. Greben AW et al. 2025
generated by the architectural comparison, from the MYND-domain inference
arm A gates the premise on one blot; arm B tests the phenomenon with three rival-separating controls
contradicts that entry's stated expectation that the fusion should be menin-INDEPENDENT because it has no CXXC and no MBM - the expectation reasons only about the fusion protein, not about what the translocation removed
predicts activity in a leukaemia with no KMT2A rearrangement, through de-restrained wild-type KMT2A rather than through a fusion
the prediction is only about menin if MEN1 ablation phenocopies the drug; falsifier 4 makes that an explicit arm
the whole chain runs through ZMYND11's MYND-domain restraint of KMT2A and the fusion boundary at aa 409
the preprint proposes a SETD2 to H3K36me3 to ZMYND11 to KMT2A-inhibition axis, noting KMT2A-rearranged leukaemias carry SETD2 mutations more often than non-rearranged ones