Three blots and one drug curve: does the ZMYND11::MBTD1 fusion protein still bind KMT2A, and does a menin inhibitor work in a leukaemia with no KMT2A rearrangement
also called the deleted-brake experiment · menin inhibitor in ZMYND11::MBTD1
Two experiments, run in order. First a single protein-pulldown asking whether the leukaemia fusion protein still grips the enzyme that its normal parent gene holds back - if it does, the idea is dead and nothing else needs doing. If it does not, then treat fusion-driven leukaemia cells with the approved drug revumenib and see whether they die, in a leukaemia the drug was never meant for.
The same thing, for a scientist
**Arm A, the gate: does the fusion retain the brake?** HEK293T transfected with FLAG-tagged (i) full-length ZMYND11, (ii) ZMYND11 aa 1-409, (iii) ZMYND11::MBTD1, (iv) FLAG empty vector. Anti-FLAG immunoprecipitation, immunoblot for endogenous KMT2A (N-terminal epitope) and for NSD1 as a second interactor the same mutation strips [E1]. **Readout:** KMT2A recovered per unit FLAG bait, normalised to full-length ZMYND11. **Decision rule:** if construct (ii) or (iii) recovers KMT2A at 50% or more of full-length, the premise of `hyp-zmynd11-mbtd1-aml-is-revumenib-sensitive-without-a-kmt2a-rearrangement` is refuted and arm B is not run. Cost tier low; one week; no animal. **Arm B, the drug, with the controls that separate three rival explanations.** Murine primary HSPCs transduced with ZMYND11::MBTD1 as in the founding study [E2], against a lineage-matched HOXA-high fusion-free comparator and a *ZMYND11*-wild-type myeloid culture. Treat with **VTP-50469** (the tool menin inhibitor this project's datasets use) across a dose range, 6 days, with **three orthogonal arms**: (1) *Men1* knockout, to establish the drug's effect is menin's; (2) I-BET151, the published lead, as a positive control for a working culture; (3) **doxycycline-inducible full-length ZMYND11 re-expression**, which is the arm unique to this hypothesis. **Readouts:** viability, CD11b/Gr-1 differentiation, colony replating, plus CUT&RUN for MLL1-N, menin and H3K4me3 at ZMYND11-bound transcription start sites. **The design's central weakness is stated rather than managed, because it cannot be managed.** There is no human ZMYND11::MBTD1 cell line [E3]. The only available model transduces a fusion cDNA into **wild-type** HSPCs, so both *Zmynd11* alleles are intact and the fusion is overexpressed - **the wrong genotype for a hypothesis about a disrupted allele.** In that model the ZMYND11 re-expression arm tests the converse question (does adding more ZMYND11 to an already-wild-type background blunt the drug?) which is weaker. **The genotype-correct version requires a knock-in allele or a patient-derived xenograft, and neither exists.** Arm B is therefore a screen for a phenomenon, and arm A plus the ZMYND11 re-expression arm are the only parts that test the stated mechanism. **Why it is still worth running.** Arm A is a week and settles a claim this graph currently rests on arithmetic. Arm B, whatever it shows, bears on the strongest threat in this corpus's queue - cycle 82's finding that a menin inhibitor moved H3K79me2 in a fusion-free line at 0.0057 from formally refuting fusion-dependence.
Arm A in full, with the reagent that decides it
Constructs. FLAG-ZMYND11 full length (aa 1-602); FLAG-ZMYND11 1-409, which is the fusion's retained 5' segment 2; FLAG-ZMYND11::MBTD1, the published construct (ZMYND11 1-409 + 16 UTR-derived residues + MBTD1 1-628) 2; FLAG-ZMYND11 R600W as an internal positive control for loss of binding 1; FLAG empty vector.
The reagent that decides the experiment is the KMT2A antibody, and this project has a rule about it. KMT2A is cleaved by taspase1 into N320 and C180 fragments, which UniProt models as separate entities with separate localisation comments, and which this project argues about constantly. An N-terminal-epitope antibody reads N320 plus full-length; a C-terminal one reads C180 plus full-length. The preprint's reporter used rTetR-MLL-C, the C-terminal region 1, so the functional restraint was demonstrated against the C-terminal fragment's activity while the interaction was demonstrated by mass spectrometry that does not distinguish fragments. Blot both ends. If ZMYND11 binds one fragment and not the other, that is a more informative result than the pulldown itself, and it is invisible to a single antibody.
Decision rule, pre-registered here so it cannot be moved afterwards. KMT2A signal per unit FLAG bait, mean of three independent transfections, normalised to full-length ZMYND11 = 1.0:
| result | verdict |
|---|---|
| fusion and 1-409 both below 0.2 | premise SUPPORTED; run arm B |
| either at or above 0.5 | premise REFUTED; do not run arm B |
| between 0.2 and 0.5 | UNRESOLVED; report as such and do not proceed on it |
The middle band exists because this project has a documented case of a verdict flipping across a fixed cut-point on an unrelated deviation, and a three-way rule with an explicit unresolved band is the fix.
Arm B, and the three rivals each control is for
| arm | what it is for | what a null result means |
|---|---|---|
| VTP-50469 dose-response in fusion-bearing culture | the phenomenon | no effect at all: the hypothesis's clinical prediction fails, cheaply |
| lineage-matched fusion-free HOXA-high comparator | separates rival 1, that the drug acts on wild-type KMT2A in myeloid cells generally | equal effect in both: the result is real but is not about this fusion, and must be reported as a general menin-inhibitor property |
| Men1 knockout | separates rival 3, off-target drug activity | drug works, knockout does not: the drug is not acting through menin and the experiment is uninterpretable, not negative |
| I-BET151 | positive control for a culture that can respond at all | nothing works: the culture is the problem |
| inducible full-length ZMYND11 re-expression | the ONLY arm unique to this hypothesis | drug effect unchanged by ZMYND11 restoration: the mechanism proposed is refuted even if the drug works |
| CUT&RUN for MLL1-N, menin, H3K4me3 at ZMYND11-bound TSSs | the molecular counterpart of the cortex H3K4me3 result | no H3K4me3 change: the de-restraint is not happening at those coordinates |
Two antibody hazards, both already logged in this project and both live here. The MLL1n reagent this graph's own datasets use is Bethyl A300-086A, an MLL1 N-terminus epitope, which reads wild-type MLL1 and any N-terminus-retaining fusion together - and in this experiment that is a feature rather than a bug, because wild-type MLL1 is the species the hypothesis is about. But it means the CUT&RUN cannot attribute signal to the fusion versus wild-type KMT2A on its own; a tagged fusion is needed for that, and the founding study's HA-tagged construct supplies one.
And the confound the design cannot remove. A 6-day exposure in a proliferating culture confounds chromatin effects with selection, which is the same criticism this corpus applied to its own cycle-82 K562 arm at 7 days. A 48-hour arm should be run alongside, and the two durations reported separately rather than pooled.
| design | ARM A (gate, cost tier low, one week, no animal): HEK293T transfection of FLAG-tagged full-length ZMYND11, ZMYND11 aa 1-409, ZMYND11::MBTD1, ZMYND11 R600W and empty vector; anti-FLAG immunoprecipitation; immunoblot for endogenous KMT2A with BOTH an N-terminal-epitope and a C-terminal-epitope antibody (to distinguish the taspase1 N320 and C180 fragments) and for NSD1. Three independent transfections. ARM B (cost tier medium-high, three to four months, animal only if extended in vivo): murine primary HSPCs transduced with ZMYND11::MBTD1 per PMID 33594072, versus a lineage-matched HOXA-high fusion-free comparator and a ZMYND11-wild-type myeloid culture; VTP-50469 dose-response at 48 h and 6 days reported separately; orthogonal arms Men1 knockout, I-BET151 positive control, and doxycycline-inducible full-length ZMYND11 re-expression at approximately endogenous level; CUT&RUN for MLL1-N, menin and H3K4me3 at ZMYND11-bound transcription start sites. |
|---|---|
| readout | ARM A: KMT2A immunoblot signal per unit FLAG bait, normalised to full-length ZMYND11 = 1.0, at both KMT2A epitopes. ARM B: viability, CD11b/Gr-1 differentiation, serial colony replating; and occupancy of MLL1-N, menin and H3K4me3 at the top quartile of ZMYND11-bound transcription start sites. |
| controls | ARM A: empty vector for bait specificity; ZMYND11 R600W as the published positive control for loss of KMT2A binding; NSD1 as a second interactor the same mutation strips, so a construct that loses KMT2A but keeps NSD1 is distinguishable from one that has simply misfolded. ARM B: lineage-matched fusion-free culture (separates the hypothesis from 'menin inhibitors act on wild-type KMT2A generally'); Men1 knockout (separates it from off-target drug activity); I-BET151 (confirms the culture can respond); inducible ZMYND11 re-expression (the only arm unique to this hypothesis); 48 h alongside 6 days (separates chromatin effect from selection). |
| cost tier | low for arm A, med-high for arm B |
| timeline | arm A one week; arm B three to four months in vitro, plus six to nine months if extended to transplant |
| the limitation that cannot be designed around | No human ZMYND11::MBTD1 cell line is known to exist, and Cellosaurus was NOT queried in this beat to confirm that. The only published model transduces a fusion cDNA into WILD-TYPE murine HSPCs, so both Zmynd11 alleles are intact and the fusion is overexpressed. A hypothesis partly about a disrupted allele cannot be tested in that genotype. In it, the ZMYND11 re-expression arm tests the converse question - whether adding ZMYND11 on top of a wild-type background blunts the drug - which is a weaker experiment. The genotype-correct version needs a Zmynd11-fusion knock-in allele or a patient-derived xenograft, and neither exists. |
| what the experiment cannot show | It cannot show that a patient would respond. It cannot attribute a CUT&RUN signal to the fusion versus wild-type KMT2A without the HA-tagged construct, because the MLL1-N reagent this project uses (Bethyl A300-086A) reads both. And a positive result in a transduced murine progenitor is exactly the inference this project's rule 73 exists to constrain: it would be a result about a mouse cell, and the first clause of every claim drawn from it has to say so. |
| status | planned |
| absence class | never-measured |
| absence class note | Both arms need new material - constructs, cells, drug, and for arm B an animal - so never-measured. Two cheaper checks are named rather than done and are NOT never-measured: a Cellosaurus query for any ZMYND11::MBTD1 line (one request, and it could remove this design's central limitation), and a question to the authors of the 2026 Haematologica series about whether any of their relapsed patients received a menin inhibitor. Both are afternoon-scale and both should precede arm B. |
decision rule prearm
{
"fusion_and_1_409_both_below_0_2": "premise SUPPORTED - run arm B",
"either_at_or_above_0_5": "premise REFUTED - do not run arm B; amend hyp-zmynd11-mbtd1-aml-is-revumenib-sensitive-without-a-kmt2a-rearrangement with the outcome written INSIDE the falsifier",
"between_0_2_and_0_5": "UNRESOLVED - report as unresolved and do not proceed on it. The explicit middle band exists because this corpus has a documented case of a verdict flipping across a fixed cut-point on an unrelated deviation."
}- The experiment is designed to test a claim that, if true, undermines this laboratory's own selectivity argument. That is the reason to run it and the reason it will be tempting to report a null cheaply. The pre-registered decision rule and the explicit unresolved band are in the data block precisely so that the verdict cannot be moved after the fact.
- A positive arm B result with no ZMYND11-dependence in the re-expression arm would be the worst outcome to interpret: the drug works, the mechanism proposed is refuted, and the finding becomes 'menin inhibitors work in a fusion-free HOXA-high myeloid state' - which is a much larger and much less well-controlled claim than the one the experiment set out to test, and which the design's comparators are only just adequate to support.
- Arm B's drug exposure in a proliferating primary culture confounds chromatin effect with selection at 6 days, which is the same criticism this corpus levelled at its own K562 arm at 7 days. Running 48 h alongside is in the design, but a 48-hour viability readout in a slowly-transforming culture may have no dynamic range, so the two durations may not both be informative.
- Everything here is a mouse progenitor with an overexpressed cDNA and two intact wild-type Zmynd11 alleles. The honest ceiling on arm B is 'a phenomenon exists in this model', and the model is the wrong genotype for the hypothesis.
- Does a ZMYND11::MBTD1 cell line exist? Cellosaurus was not queried in this beat. One request, and a positive answer removes this design's central limitation and makes the whole experiment better.
- Which KMT2A fragment does ZMYND11 bind - N320, C180 or only the uncleaved protein? The preprint's mass spectrometry cannot say and its reporter used only the C-terminal region. Blotting both epitopes in arm A answers it for free, and the answer would be more informative than the pulldown it comes from, because this project's arguments about KMT2A constantly turn on which fragment is meant.
- Is there a lineage-matched HOXA-high fusion-free comparator culture that is genuinely matched? The comparator arm carries the weight of separating this hypothesis from the general-menin-activity rival, and 'lineage-matched' is doing unexamined work - a NUP98-fusion or CALM-AF10-like model shares the HOXA programme but not the genotype, and a Zmynd11-wild-type myeloid culture shares the genotype but not the programme. The design needs both and says so, but which is the primary comparator is unresolved.
- Should the first move be the wet experiment at all? Asking the authors of the 2026 series whether any of their relapsed patients received a menin inhibitor costs one email and could refute the clinical prediction outright, or supply a single-patient signal no mouse can. It is named here and in the hypothesis, and it should precede arm B.
- 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 by FLAG-IP with quantitative tandem-mass-tag mass spectrometry; deletion of the MYND domain drastically reduces most ZMYND11 interactions while the R600W substitution selectively removes the top interactors including KMT2A and NSD1; full-length ZMYND11 inhibits activation of a 9xTetO-minCMV-luciferase reporter driven by rTetR-MLL-C, and the MYND domain alone represses without activating; and revumenib significantly attenuates the transcriptional defects caused by 24-hour dTAG degradation of endogenous ZMYND11 in primary cortical 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. Full text read this session via PMC12636335. Neuronal and heterologous systems only; the paper does not mention MBTD1 or any fusion. THIS EXPERIMENT'S ARM A IS A DIRECT ADAPTATION OF THE PREPRINT'S OWN ASSAY, which is its main methodological strength: the constructs, the FLAG-IP and the R600W control are all the authors' and are reported to work in HEK293T, so arm A is a replication with two added constructs rather than a new method. Note that the interaction was demonstrated by mass spectrometry, which does not distinguish the taspase1 N320 and C180 KMT2A fragments, while the functional reporter used only the C-terminal region.
Evidence that argues against this (1)
- Adapting an assay from a preprint means inheriting its unpublished failure modes. The mutant interactome comparison was performed on OVEREXPRESSED ZMYND11 in 293T, and the authors themselves note that overexpression recovers lower-affinity interactions - so a 293T FLAG-IP may report a KMT2A contact for the 1-409 fragment that does not occur at endogenous levels, which would produce a false REFUTATION of the hypothesis at arm A. The decision rule's middle band exists partly for this reason and an endogenous-level replication would be needed to act on a refutation. Wen H et al. 2014
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The murine model for this fusion transduces primary haematopoietic stem/progenitor cells with a construct comprising ZMYND11 amino acids 1-409, a 16-residue UTR-derived segment and the entire 628-residue MBTD1 open reading frame, conferring indefinite ex vivo self-renewal and causing AML on transplant with a mean latency of 154 days in 8 of 9 mice; the resulting AML is sensitive to the BET inhibitor I-BET151 in vitro and in vivo, with blocked proliferation, elevated cell death and terminal myeloid differentiation.
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
This item supplies the model and the positive-control arm. THE MODEL'S GENOTYPE IS THE DESIGN'S CENTRAL LIMITATION AND IS NOT FIXABLE WITHIN IT: wild-type Zmynd11 is intact on both alleles and the fusion is overexpressed from a retrovirus, so the model cannot represent the allele disruption the hypothesis is partly about. Transformation had long latency and incomplete penetrance, so culture-to-culture variability should be expected. Full text read via PMC7886901; identifiers fetched from PubMed 2026-09-23.
Evidence that argues against this (1)
- If I-BET151 and VTP-50469 both work in this culture, the experiment struggles to attribute the effect, because both converge on the same transcriptional output - the fusion's acetylation is read by BRD4, and de-restrained KMT2A would feed the same programme. The Men1-knockout arm is the only thing separating them, and a genetic arm in a primary transduced culture is technically the hardest part of the design. Devoucoux M et al. 2022
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
No human ZMYND11::MBTD1 leukaemia cell line exists and none has been used in either mechanism study; the entity comprises twelve assembled cases plus one further 2026 paediatric case report, with detection depending on conventional karyotype followed by deliberate molecular confirmation because neither gene is typically included in targeted RNA-sequencing fusion panels and clinical FISH probes are not routinely available.
Acute myeloid leukemia with t(10;17)(p15;q21)/ ZMYND11::MBTD1: a subtype with minimal differentiation, CD7/CD56 expression, and generally poor outcomes in adults. · Wei Q, Pemmaraju N, Wang SA, Hu S, et al. · Haematologica 2026What this evidence cannot show
Retrospective case series; the 'no cell line' statement is an absence across the two mechanism papers and this series rather than a positive assertion by any of them, and a line could exist unpublished or in a biobank this beat did not query - Cellosaurus was NOT searched, which is a cheap check this project has previously found decisive and which is named here rather than done. Cited to justify why the design is stuck with a transduced murine model. Identifiers fetched from PubMed 2026-09-23.
Evidence that argues against this (1)
- The absence of a cell line is claimed from three papers' silence, and a Cellosaurus query would settle it in one request. This project has a documented case where an entry said a negative Cellosaurus query 'was not run' while resting an absence on a weaker literature search, and running it took one request. Until that is done, 'no human cell line exists' is the weakest sentence in this entry and the design choice it justifies may be unnecessary. Liang H et al. 2026
arm A gates the premise on one blot; arm B tests the phenomenon with three rival-separating controls
arm A blots endogenous KMT2A at both epitopes, so it also asks which taspase1 fragment ZMYND11 binds
the comparator arm is a lineage-matched HOXA-high myeloid culture, which is the state this graph's KMT2A-r AML entries describe