Oncofusion

start / graph / hyp-menin-inhibition-creates-a-bmi1-dependency

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q56-homeodomain

If HOXA9 substitutes for BMI1 at CDKN2A, then a menin inhibitor should CREATE a BMI1 dependency that does not exist at baseline

also called BMI1 synthetic lethality under menin inhibition · HOXA9-BMI1 buffering

BMI1 looks useless as a target in this leukemia: knocking it out does nothing, because the fusion's own product HOXA9 does BMI1's job of silencing the brake gene. But that is exactly the setup for a two-drug trick - take HOXA9 away with a menin inhibitor and BMI1 should suddenly become essential.

The same thing, for a scientist

Two published facts, held together, make a specific prediction nobody has tested. **(1)** In murine MLL-AF9, Hoxa9 regulates the *p16Ink4a/p19Arf* locus as Bmi1 does and substitutes for it: MLL-AF9 transforms *Bmi1*-null cells readily, where non-Hox-activating fusions senesce, and either *Bmi1* or *Hoxa9* alone restores leukaemogenesis in the double-null background [E1]. **(2)** Menin inhibition collapses the *HOXA*/*MEIS1* programme in KMT2A-rearranged cells, which is the agreed mechanism of the approved agent in this disease. If HOXA9 is the buffer that makes BMI1 dispensable, removing HOXA9 should unmask a BMI1 requirement. The prediction is quantitative and falsifiable: **BMI1 knockout should be near-neutral in untreated KMT2A-rearranged cells (live DepMap: +0.0585 across 1,258 lines, -0.0103 across 57 leukaemia lines [E2]) and should become costly under a menin inhibitor.** The same logic predicts a CDKN2A-dependence: the effect should be abolished in *CDKN2A*-deleted cells, which is 8.5% of infant patients [E3] and a testable isogenic arm.

Go deeper
The argument in four steps
  1. BMI1's effector in leukaemic stem cells is repression of the Ink4a/Arf locus. Standing position of the polycomb field and restated as such in the MLL-AF9 literature 4.
  2. HOXA9 regulates the same locus. Measured, in murine MLL-AF9: Hoxa9 regulates p16Ink4a/p19Arf and can overcome AML1-ETO-induced senescence 1.
  3. The two are functionally substitutable there. MLL-AF9 could not initiate leukaemia in Bmi1(-/-)Hoxa9(-/-) mice, but either gene alone restored it 1. That is the definition of a buffer pair.
  4. A menin inhibitor removes the HOXA programme. Therefore, in a menin-inhibitor-treated KMT2A-rearranged cell, the buffer is gone and BMI1 should become the only remaining repressor of the locus.

The prediction: BMI1 loss, neutral at baseline, becomes costly under menin inhibition; and the effect is CDKN2A-dependent, so it disappears in a CDKN2A-null background.

Why it is worth writing rather than assuming someone has done it: step 4 has never been joined to steps 1-3 in print in anything read this cycle. The BMI1 literature in this disease is from 2003-2011; the menin inhibitors arrived after it and the two literatures have not met.

What would kill this hypothesis, cheaply and early

This is written to be killed cheaply, per the standing instruction to compute the null before publishing an entry on a screen difference.

  • If BMI1 loss does not derepress CDKN2A in a KMT2A-rearranged cell even at baseline, the axis is not operating here at all and steps 1-2 fail in this system. One qPCR/western experiment, no menin inhibitor needed. Do this first.
  • If the HOXA collapse under a menin inhibitor does not derepress CDKN2A either, then HOXA9 is not holding the locus down in human KMT2A-rearranged cells and step 3 does not transfer from mouse.
  • If BMI1 knockout remains neutral under menin inhibition, the hypothesis is dead as stated. It could still survive as a PCGF-family hypothesis if PCGF1/PRC1.1 is the arm doing the work - this graph already treats PRC1.1 as central - which is why the experiment should knock out PCGF1 alongside BMI1 rather than BMI1 alone.
  • If the effect appears but is CDKN2A-independent, then it is real and the mechanism is wrong, which is also informative: the Yuan data already say Ink4a/Arf deletion only partially rescues Bmi1-null LSCs 4.
Why this is the most useful thing BMI1 could be in this disease

As a single target, BMI1 in KMT2A-rearranged leukaemia is weak and this beat says so plainly in gene-bmi1: fitness-neutral in 98.6% of human cancer lines, bypassed by the fusion's own output, and with a normal-HSC requirement of similar magnitude to its leukaemic one.

As a combination target it inverts. A dependency that is created by the first drug is, by construction, absent in untreated normal tissue - so the therapeutic index comes from the combination rather than from the gene, which is the only place it could have come from here.

And it addresses a real clinical problem: menin-inhibitor resistance. This graph already carries resistance-prc1-1-loss-menin-inhibitor-nup98 and mechanism-menin-inhibition-shifts-the-kmt2a-polycomb-balance, both of which say the polycomb arm is where the response to menin inhibition is decided. This hypothesis says which polycomb subunit to test and why, and predicts the direction.

Facts and figures
statementIn KMT2A-rearranged leukaemia, BMI1 is dispensable at baseline because HOXA9 substitutes for it in repressing the CDKN2A/CDKN2B locus. Pharmacological collapse of the HOXA programme with a menin inhibitor should therefore convert BMI1 from a neutral gene into a required one, and the effect should be abolished in a CDKN2A-null background.
originliterature-gap
prior probability0.3
why not already doneThe BMI1-in-MLL literature is 2003-2011; the menin inhibitors are later. No paper found this cycle joins them. The bounded check: PubMed for menin-inhibitor names combined with BMI1 was not run as a counted query this cycle and is listed as an open action rather than claimed as an absence.
predictions
[
  "BMI1 knockout or knockdown is near-neutral in untreated KMT2A-rearranged lines (already measured: +0.0585 all lines, -0.0103 leukaemia lines) and becomes costly under revumenib or a tool menin inhibitor",
  "CDKN2A/p16INK4a transcript and protein rise more on BMI1 loss PLUS menin inhibition than on either alone (a synergy at the locus, not only in the phenotype)",
  "the growth effect is abolished or strongly attenuated in an isogenic CDKN2A-null derivative",
  "PCGF1 knockout shows the same or a stronger effect than BMI1, if PRC1.1 rather than canonical PRC1 is the arm holding the locus - this is the most likely way the hypothesis is right in spirit and wrong in the named gene",
  "the effect is larger in a serial-replating or transplantation readout than in a growth curve, because the published BMI1 phenotype is self-renewal"
]
experiments
[
  "exp-bmi1-knockout-under-menin-inhibition-kmt2a-aff1"
]
analyses
[]
falsifiers
[
  "BMI1 loss does not derepress CDKN2A in a KMT2A-rearranged cell at baseline - kills steps 1-2 in this system with one western blot and no menin inhibitor",
  "menin inhibition alone does not derepress CDKN2A in a KMT2A-rearranged cell - kills step 3, meaning HOXA9 is not holding the locus down in human cells",
  "BMI1 knockout remains fitness-neutral under menin inhibition in both a growth and a serial-replating readout - kills the hypothesis as stated",
  "the combination effect is identical in CDKN2A-null cells - the phenotype would be real and this mechanism wrong"
]
How this could bite us back
  • The intended effect is derepressing CDKN2A, which induces senescence rather than death. A senescent blast is alive, secretory, and potentially supportive of its neighbours - so a successful experiment could produce a reservoir rather than a clearance.
  • BMI1 is required by normal HSCs [E2 counter-evidence, and gene-bmi1 E1]. The combination's therapeutic index depends entirely on normal HSCs NOT being under menin inhibition's HOXA collapse - and menin inhibitors do affect normal haematopoiesis, so the assumption that the created dependency is leukaemia-specific needs testing, not assuming.
  • If the answer is PCGF1 rather than BMI1, the combination collides with resistance-prc1-1-loss-menin-inhibitor-nup98, which records PRC1.1 LOSS as a menin-inhibitor resistance mechanism. Inhibiting the same complex therapeutically would then be pushing in the direction resistance already goes.
  • There is no BMI1 inhibitor with credible clinical standing read for this entry. The hypothesis is genetic; converting it into a drug combination requires chemical matter this beat did not assess.
  • A created dependency is created only while the first drug is on. Intermittent dosing, a drug holiday or a pharmacokinetic trough restores the HOXA buffer and with it the BMI1 dispensability - so the combination has a schedule-dependence that a constant-exposure experiment will not reveal.
Still open
  1. Has anyone already measured BMI1 or PCGF1 dependency under a menin inhibitor? A counted PubMed query over the full menin-inhibitor name list (menin inhibitor/inhibitors/inhibition, revumenib, ziftomenib, bleximenib, enzomenib, VTP-50469, VTP50469, MI-503, MI-538, MI-2-2, SNDX-5613) crossed with BMI1 and PCGF1 was NOT run this cycle and should be run before the experiment is funded - rule 72 says the name list is the instrument.
  2. Is the relevant PCGF here BMI1 (PCGF4, canonical PRC1) or PCGF1 (PRC1.1)? This graph treats PRC1.1 as central and the two arms have never been compared head-to-head at the CDKN2A locus in this disease.
  3. Does the HOXA collapse under menin inhibition actually reach the CDKN2A locus, or only the canonical HOXA/MEIS1 targets? A p16INK4a western in a revumenib time-course is a one-week experiment and would settle whether this hypothesis has a premise.
Every source behind this entry (4)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In murine cells, MLL-AF9, which drives Hoxa7 and Hoxa9, readily transforms Bmi1(-/-) cells while AML1-ETO and PLZF-RARalpha, which do not activate Hox, trigger senescence; Hoxa9 regulates expression of the p16(Ink4a)/p19(ARF) locus and could overcome senescence induced by AML1-ETO; and MLL-AF9 could not initiate leukaemia in Bmi1(-/-)Hoxa9(-/-) mice, but either gene alone restored its ability to establish leukaemic stem cells in the double-null background.

    Functional crosstalk between Bmi1 and MLL/Hoxa9 axis in establishment of normal hematopoietic and leukemic stem cells · Smith LL, Yeung J, Zeisig BB, Popov N, Huijbers I, Barnes J, Wilson AJ, Taskesen E, Delwel R, Gil J, Van Lohuizen M, So CW · Cell Stem Cell 8:649-662 2011
    PMID 21624810DOI
    What this evidence cannot show

    Abstract read this cycle. MURINE MLL-AF9 MYELOID disease, germline knockouts, and a transformation/initiation readout rather than a maintenance readout. The hypothesis extrapolates from germline absence of Hoxa9 to pharmacological collapse of the HOXA programme in an established human B-lineage leukaemia - two steps, either of which could fail. The paper's own closing sentence already suggests synergistic targeting of both molecules, so the direction of this hypothesis is theirs; what is added here is the specific pharmacological partner and the CDKN2A-dependence prediction.

    Evidence that argues against this (1)
    • A second 2011 paper found Bmi1 indispensable for MLL-AF9 leukaemia in vivo even with Hoxa9 present, which if correct means the buffer is incomplete at baseline and the menin inhibitor would have less to unmask than this hypothesis assumes. Yuan J et al. 2011
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Fetched live on 2026-09-20, DepMap Chronos gene effect for BMI1 has mean +0.0585 across 1,258 cell lines and -0.0103 across 57 leukaemia-annotated lines, with 1.43% of lines at or below -0.5 - establishing the baseline neutrality this hypothesis predicts a menin inhibitor will overturn.

    DepMap CRISPR (Chronos) gene effect for BMI1 (ENSG00000168283), retrieved via the Open Targets Platform GraphQL API · Broad Institute DepMap; Open Targets Platform; query run by this project 2026-09-20 · Database query 2026
    What this evidence cannot show

    Untreated lines only - DepMap has no menin-inhibitor arm, which is precisely why this hypothesis requires a new experiment rather than a re-analysis. 21-day fitness in immortalised lines, no self-renewal readout, no infant-derived line, and the leukaemia subset is a coarse text match on diseaseFromSource. The baseline is therefore a baseline for PROLIFERATION, and if the unmasked dependency is a self-renewal dependency this assay would miss it in both arms.

    Evidence that argues against this (1)
    • If BMI1's requirement in this disease is entirely a self-renewal requirement, a proliferation baseline of zero is uninformative and the predicted 'creation' of a dependency could be invisible in the same assay - so the proposed experiment must include a serial-replating or transplantation arm, not only a growth curve. Lessard J et al. 2003
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    CDKN2A is deleted in about 8.5% (4/47) of infant KMT2A-rearranged ALL against 28% (5/18) of infant ALL without a KMT2A rearrangement, so the locus this hypothesis turns on is genomically intact in the large majority of index-disease patients and a CDKN2A-null isogenic arm is a realistic negative control rather than the common case.

    The landscape of somatic mutations in infant MLL-rearranged acute lymphoblastic leukemias · Andersson AK, Ma J, Wang J, Chen X, Gedman AL, Dang J, Nakitandwe J, Holmfeldt L, Parker M, et al. · Nature Genetics 47:330-337 2015
    PMID 25730765DOI
    What this evidence cannot show

    The frequency is quoted from this graph's gene-cdkn2a entry, not read from the paper by this beat; the PMID was fetched fresh from PubMed efetch on 2026-09-20 for the title, DOI, venue and authors. n=47 against n=18, P=0.058, underpowered, and no clonality data.

    Evidence that argues against this (1)
    • If the relapsing clone acquires CDKN2A deletion under treatment - which no study read here has measured, because clonality was not reported - then the very patients who relapse would be the ones in whom this combination stops working. Andersson AK et al. 2015
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    One of the major targets of polycomb group proteins including Bmi1 in leukaemic stem cells is the Ink4a/Arf tumour suppressor gene locus; and in murine MLL-AF9, deletion of both Ink4a and Arf only partially restored the leukaemogenic capacity of Bmi1-null leukaemic stem cells, so the locus is a major but not exclusive effector.

    Bmi1 is essential for leukemic reprogramming of myeloid progenitor cells · Yuan J, Takeuchi M, Negishi M, Oguro H, Ichikawa H, Iwama A · Leukemia 25:1335-1343 2011
    PMID 21527932DOI
    What this evidence cannot show

    Abstract read this cycle. MURINE MLL-AF9 myeloid, GMP cell of origin, germline knockout. 'Only partially restored' is unquantified in the abstract, so the size of the Ink4a/Arf-independent component is unknown from what was read - which matters, because if it is large then the CDKN2A-dependence arm of this hypothesis will read as a partial rather than a clean negative.

    Evidence that argues against this (1)
    • If a large part of BMI1's effect is Ink4a/Arf-independent, then the mechanism proposed here is the wrong explanation even if the predicted synthetic lethality is observed - and the experiment must therefore measure CDKN2A derepression directly, not infer it from the phenotype. Smith LL et al. 2011
Connected to