Oncofusion

start / graph / hyp-the-hoxa-meis-independent-cbx5-programme-is-an-unasked-menin-inhibitor-resistance-route

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q110-h3k9me3-writer-layer

A published MLL leukaemia stem-cell programme immortalises progenitors WITHOUT Hoxa or Meis - Myb, Hmgb3 and Cbx5 - and nobody has ever asked what a menin inhibitor does to it, although this graph reads the drug's effect almost entirely through HOXA9 and MEIS1

also called Myb/Hmgb3/Cbx5 programme · the Hoxa/Meis-independent route · ESC-like LSC maintenance programme

The only targeted drugs for this leukaemia work by switching off two genes, HOXA9 and MEIS1. A 2009 paper found three other genes that together keep MLL leukaemia stem cells self-renewing without needing HOXA9 or MEIS1 at all. If that route is active in patients, it is a way for the leukaemia to survive the drug - and it has never been tested.

The same thing, for a scientist

**The published result.** In a mouse model of human MLL AML, leukaemia stem cells are maintained by a transcriptional subprogramme more like embryonic than adult stem cells; **Myb, Hmgb3 and Cbx5 are critical components of it, and co-expressing the three suffices for Hoxa/Meis-INDEPENDENT immortalisation of myeloid progenitors** [E1]. The authors call it *ancillary to the leukaemia-initiating MLL/Hox/Meis programme*. **Why that is a resistance hypothesis nobody has stated.** This graph's therapeutic spine is the menin-KMT2A interface: revumenib, ziftomenib, bleximenib and enzomenib all displace menin from the fusion. The founding paper describes the mechanism as displacement of menin from protein complexes and reduced MLL chromatin occupancy at **select genes**, and names neither HOXA9 nor MEIS1 in its abstract [E2] - reading the drug as a HOXA9/MEIS1 agent is this graph's shorthand, and it is the shorthand this hypothesis exploits, so it is worth being exact about. A self-renewal programme that is **by construction** independent of Hoxa and Meis is a route the drug's mechanism does not reach. The paper is from 2009, predates menin inhibitors entirely, and makes no claim about resistance - so the connection has never been drawn, and the query that would draw it has never been run: `CBX5[tiab] AND revumenib[tiab]`, `HMGB3[tiab] AND menin[tiab]` and the equivalents return nothing this beat could find. **What makes it testable cheaply, and therefore worth writing.** One of the three genes, **MYB**, already has an entry here recording *"MLL-AF9 in mouse; nothing in KMT2A::AFF1 or B-lineage"* - the same system limitation. Published transcriptional datasets under menin inhibitors exist and this graph has already analysed several of them for *HOXA* and *MEIS1* response. Asking whether *CBX5*, *MYB* and *HMGB3* move in those same datasets is a re-analysis, not an experiment: if the three are **unchanged or up** while *HOXA9* and *MEIS1* collapse, the programme survives the drug in the cells where the drug works, which is the observation the hypothesis needs. **Prior probability: 0.15.** Low, and the reasons are specific: the result is murine, myeloid, and rests on retroviral co-expression of three genes at non-physiological levels; the initiating MLL fusion is not stated in the abstract; there is no loss-of-function arm for Cbx5 alone; CBX5's DepMap dependency is flat (mean -0.083, 0.4% of 1,258 lines below -0.5) and has two paralogues within eight residues; and nothing places the programme in a human B-lineage KMT2A::AFF1 cell. It is written because **the cost of asking is one re-analysis and the cost of not asking has already been seventeen years.**

Facts and figures
statementThe Myb/Hmgb3/Cbx5 embryonic-like self-renewal programme, shown in mouse MLL AML to sustain leukaemia stem cells and to immortalise myeloid progenitors independently of Hoxa and Meis, is active in human KMT2A-rearranged leukaemia and constitutes an intrinsic resistance route to menin inhibition, whose mechanism acts through HOXA9 and MEIS1 and therefore cannot reach it.
originliterature-gap
prior probability0.15
why a menin inhibitor cannot reach itA menin inhibitor's mechanism is displacement of menin from protein complexes with reduced MLL chromatin occupancy at select genes [E2], and the downstream consequence this graph and most of the field read out is loss of the HOXA9/MEIS1 transcriptional programme. A self-renewal programme demonstrated to immortalise progenitors WITHOUT Hoxa or Meis is, by the construction of the experiment that defined it, not downstream of THAT axis. The step this argument has not established - and which would break it - is whether MYB, HMGB3 and CBX5 are themselves MLL-fusion-bound genes, in which case the drug would reach them directly and the independence is irrelevant. Nobody has checked.
the analysis that would test it and why it is cheapThis graph has already re-analysed published menin-inhibitor transcriptional datasets for HOXA and MEIS1 response - two such analyses are named above. Adding CBX5, MYB and HMGB3 to the same gene list in the same datasets is a re-analysis with no new data and no wet work. It is the single cheapest test of a resistance hypothesis this beat found.
what this hypothesis is notIt is NOT a claim that a menin inhibitor fails, and it is not an argument against the menin-interface hypothesis. Menin inhibitors are approved and effective in relapsed/refractory KMT2A-rearranged acute leukaemia and this graph holds the measurement that their transcriptional response is nine-fold fusion-selective. It is a claim that ONE published self-renewal route was defined as independent of the axis the drug acts on, and that nobody has looked at it under the drug. Intrinsic resistance mechanisms in this disease are a live clinical problem; this is a named, testable candidate that costs a re-analysis.
relation to known resistanceThis graph already records MEN1 M327I and related pocket mutations as acquired resistance to menin inhibitors. A Hoxa/Meis-independent self-renewal programme would be a DIFFERENT class: intrinsic rather than acquired, transcriptional rather than structural, and invisible to the MEN1 sequencing that is used to detect the known mechanism.
localizationnucleus - all three genes encode nuclear chromatin or transcription regulators (MYB a transcription factor, HMGB3 a high-mobility-group chromatin protein, CBX5 the H3K9me3 reader). No subcellular measurement is part of this hypothesis.
subnuclear localizationnot measured and not part of the hypothesis. CBX5's own compartment is centromeric and pericentromeric heterochromatin (gene-cbx5); whether the self-renewal programme involves that compartment is unknown, and the chromodomain-dependence prediction above is the test that would begin to say.
cell typethe source result is MURINE MYELOID: retrovirally immortalised mouse myeloid progenitors and mouse MLL AML leukaemia stem cells. The hypothesis is ABOUT human infant KMT2A::AFF1 B-lineage blasts, in which the programme has never been looked for. That distance is the principal reason the prior is 0.15.
cell stateSELF-RENEWAL is the state, measured in the source as serial immortalisation and LSC frequency - endpoints a proliferation screen does not capture, which is why CBX5's flat DepMap score is not counter-evidence to the source result.
epigenetic readoutH3K9me3 via CBX5's chromodomain is the candidate epigenetic mechanism and it is UNTESTED - the source paper treats Cbx5 as a chromatin regulatory factor in a gene-expression programme and does not ask whether its chromodomain or its H3K9me3 binding is required. HMGB3 is an HMG-box chromatin protein, not a mark reader or writer.
shown for fusionMLL AML in mouse, as a PRO-LEUKAEMIC self-renewal programme (PMID 19200802); the abstract states LSC frequency correlates with the initiating MLL oncogene but does not name which fusion the immortalisation experiments used. Nothing in KMT2A::AFF1, nothing in B-lineage, nothing in human material, and nothing under any menin inhibitor in any system.
the three genes
{
  "MYB": "has an entry in this graph (gene-myb), whose shown_for_fusion field reads 'MLL-AF9 in mouse; nothing in KMT2A::AFF1 or B-lineage' - the identical system limitation",
  "HMGB3": "no entry in this graph as of 2026-09-22, checked against a full id listing of 2,103 ids",
  "CBX5": "written by this beat as gene-cbx5; the H3K9me3 reader, DepMap-flat, with two paralogues within eight residues"
}
predictions
[
  "In published transcriptional data from KMT2A-rearranged cells treated with a menin inhibitor, CBX5, MYB and HMGB3 are unchanged or increased while HOXA9 and MEIS1 fall.",
  "In cells that survive or regrow under sustained menin inhibition, expression of the three-gene programme is higher than in the drug-naive population.",
  "Forced co-expression of MYB, HMGB3 and CBX5 in a KMT2A-rearranged human line reduces the effect of a menin inhibitor on colony formation or self-renewal without restoring HOXA9 or MEIS1.",
  "The programme is detectable in human infant KMT2A::AFF1 B-lineage blasts - which is the first thing to check, since it has only ever been shown in murine myeloid cells.",
  "If CBX5's contribution is chromodomain-dependent, an H3K9me3-binding point mutant fails to support the programme; if it is not, then the programme is not an H3K9me3 story at all and this hypothesis belongs entirely to the resistance literature rather than to this beat's."
]
experiments
[]
analyses
[
  "ana-the-menin-inhibitor-response-is-nine-fold-fusion-selective-and-does-not-touch-hoxa-or-meis1-in-either-kmt2a-aff1-line",
  "ana-the-potent-menin-inhibitor-collapses-what-the-weak-one-left-untouched-and-hoxa9-was-never-measured-by-either"
]
falsifiers
[
  "THE PROGRAMME COLLAPSES WITH THE DRUG. If CBX5, MYB and HMGB3 fall alongside HOXA9 and MEIS1 in menin-inhibitor-treated KMT2A-rearranged cells, the programme is downstream of the axis the drug acts on and this hypothesis is dead.",
  "IT IS ABSENT FROM THE INDEX DISEASE. If the three-gene programme is not expressed in human KMT2A::AFF1 B-lineage blasts, it is a property of murine myeloid MLL-AF9 biology and carries no clinical implication here.",
  "THE ORIGINAL RESULT DOES NOT SURVIVE PHYSIOLOGICAL EXPRESSION. If CBX5 knockout or degradation in an MLL-rearranged line has no effect on self-renewal, then the 2009 sufficiency result is an over-expression artefact - and the flat DepMap dependency is already weakly consistent with that.",
  "PARALOGUE SUBSTITUTION. If CBX1 or CBX3 can replace CBX5 in the immortalisation assay, then 'Cbx5' in the original result means 'an HP1 protein', the specificity this hypothesis assumes is absent, and CBX1's four-times-deeper dependency becomes the thing to explain instead."
]
How this could bite us back
  • THIS IS A RESISTANCE HYPOTHESIS, NOT A TARGET NOMINATION, AND CBX5 IS A PARTICULARLY BAD TARGET. It is DepMap-flat (mean -0.083, 0.4% of 1,258 lines below -0.5), it has two paralogues within eight residues that dimerise with it, and forced local concentration of HP1-alpha is sufficient to nucleate PML nuclear bodies and induce ALT-like recombination biomarkers - a genome-instability liability in a population whose therapeutic goal is fewer rearrangements. Nothing here supports dosing any of the three genes.
  • MYB is the worst of the three to reason about therapeutically, and this graph already knows it: MYB is required for normal haematopoiesis, and a 'targetable' three-gene self-renewal programme whose first member is MYB is an essential-function problem before it is an oncology opportunity.
  • Do not let this hypothesis be quoted as evidence that menin inhibitors fail. They are approved and effective in relapsed/refractory KMT2A-rearranged acute leukaemia, and this graph holds the measurement that their transcriptional response is nine-fold fusion-selective. The claim here is narrow: one published self-renewal route was defined as independent of the axis the drug acts on, and nobody has looked at it under the drug.
  • The evidence base is one 2009 mouse abstract plus an internal citation whose underlying datasets this beat did not re-read. Prior 0.15. If a later cycle finds the re-analysis is not actually cheap, the hypothesis should be demoted rather than pursued, and this entry's own E2 caveat says so.
Still open
  1. Do CBX5, MYB and HMGB3 change in published transcriptional data from KMT2A-rearranged cells under a menin inhibitor? This is the cheapest resistance question in this beat and it needs no new data - but it needs the existing analysis entries opened first, which this beat did not do.
  2. Is the Myb/Hmgb3/Cbx5 programme expressed in human infant KMT2A::AFF1 B-lineage blasts at all? Public infant ALL expression data would answer it, and a negative would close the hypothesis cleanly.
  3. Why has HMGB3 no entry in this graph? It is one of three genes in a published MLL leukaemia stem-cell programme and is absent from 2,103 ids. If this hypothesis survives its first test, HMGB3 is the next entry to write.
  4. Is CBX5's role in the programme chromodomain-dependent? If not, this hypothesis has nothing to do with H3K9me3 and should be re-filed under resistance rather than under the H3K9 beat that produced it.
  5. Can CBX1 or CBX3 substitute for CBX5 in the immortalisation assay? The three paralogues are within eight residues and the DepMap panel says they are not equivalent; a substitution experiment would say whether the 2009 result is about HP1-alpha or about HP1 activity in general.
Every source behind this entry (2)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In a MOUSE model of human MLL AML, the transcription/chromatin regulatory factors Myb, Hmgb3 and Cbx5 are critical components of a leukaemia stem-cell maintenance programme more akin to embryonic than adult stem cells, and co-expressing the three suffices for Hoxa/Meis-INDEPENDENT immortalisation of myeloid progenitors, establishing the cooperative and essential role of an ESC-like LSC maintenance programme ancillary to the leukaemia-initiating MLL/Hox/Meis programme; leukaemia stem cells exhibit variable frequencies that correlate with the initiating MLL oncogene; and enriched expression of LSC-maintenance and ESC-like programme genes in normal myeloid progenitors and poor-prognosis human malignancies links aberrantly self-renewing progenitor-like cancer stem-cell frequency to prognosis in human cancer.

    PMID 19200802DOI
    What this evidence cannot show

    Abstract level. Murine, MYELOID, retroviral co-expression of three genes together at non-physiological levels; the abstract does not name which MLL fusion the immortalisation experiments used. No loss-of-function arm for Cbx5 alone is reported in the abstract, so this is a three-gene sufficiency result and not a Cbx5 requirement result. Published 2009, before menin inhibitors existed, and it makes no claim about drug resistance - the resistance reading is this entry's inference and is the reason this is a hypothesis node.

    Evidence that argues against this (2)
    • CBX5's DepMap dependency is flat - mean -0.083 across 1,258 screens with 0.4% of lines below -0.5, and a blood-arm mean of -0.080 - while its paralogue CBX1 is four times deeper. If HP1-alpha were broadly required for leukaemic self-renewal, a panel containing many leukaemia lines should show something. The honest reading is that a 21-day proliferation screen does not measure serial self-renewal, but the flat score is the strongest available argument that the 2009 sufficiency result is an over-expression phenotype.
    • The word 'independent' is doing heavy work and is an experimental claim about one assay, not a claim about a patient. 'Immortalises myeloid progenitors without Hoxa or Meis' is compatible with the programme being unable to sustain an established leukaemia that already depends on HOXA9 and MEIS1 - in which case removing the HOXA axis with a drug would still kill the cells and the programme would be irrelevant to resistance.
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    A menin-MLL inhibitor works by displacing menin from protein complexes and reducing MLL chromatin occupancy at select genes, and the phenotype follows from the resulting change in gene expression: VTP50469 is potent, highly selective and orally bioavailable, MLL-rearranged cell lines were selectively responsive, loss of MLL binding led to changes in gene expression, differentiation and apoptosis, and patient-derived xenografts from MLL-r AML and MLL-r ALL showed dramatic reductions in leukaemia burden with multiple mice engrafted with MLL-r ALL remaining disease free for over a year. So the drug class acts by removing a transcription-factor complex from a defined set of genes, which is why a self-renewal programme defined as independent of that gene set's canonical output is not predicted to be affected by it.

    PMID 31821784DOI
    What this evidence cannot show

    Abstract level, and fetched live on 2026-09-22 rather than copied from a sibling entry. This paper establishes the drug class's MECHANISM, not the resistance claim: it does not measure CBX5, MYB or HMGB3, does not name HOXA9 or MEIS1 in its abstract, and reports profound efficacy including durable disease-free survival in MLL-r ALL xenografts. It is cited here only to establish that the mechanism is displacement of a complex from select genes. This graph's own two menin-inhibitor re-analyses (named in data.analyses) were identified from an id listing and their datasets were NOT re-read by this beat, so the claim that adding three genes to their gene lists is cheap is UNVERIFIED and is flagged as such in data.the_analysis_that_would_test_it_and_why_it_is_cheap.

    Evidence that argues against this (2)
    • This paper is the strongest argument AGAINST the hypothesis mattering clinically. Multiple mice engrafted with MLL-rearranged ALL remained disease free for more than a year on VTP50469. If a Hoxa/Meis-independent self-renewal programme were sufficient to sustain the leukaemia, durable single-agent remissions of that length would be harder to obtain - so either the programme is not present in those models, or it is not sufficient.
    • The mechanism is displacement of MLL from SELECT genes, not from HOXA9 and MEIS1 alone, and the abstract names neither. Framing the drug as a HOXA9/MEIS1 agent is this graph's shorthand rather than the founding paper's claim, and if menin inhibition also removes the fusion from genes in the Myb/Hmgb3/Cbx5 programme then the independence argument fails at its first step. Nobody has checked whether those three genes are MLL-fusion-bound.
Connected to