Oncofusion

start / graph / hyp-menin-dependency-tracks-available-mbm

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Qn-nup98-family

Menin dependency tracks the availability of an intact menin-binding motif, in cis or in trans — not the fusion family

also called MBM availability hypothesis · cis or trans menin tether · why NUP98-KMT2A is resistant

Menin drugs work when the leukaemia's key protein complex is held onto DNA by menin. KMT2A fusions carry that grip themselves; NUP98 fusions borrow it from the cell's normal copy — and the one NUP98 fusion that destroys the grip is the one the drugs fail against.

The same thing, for a scientist

**Statement.** A leukaemia is menin-inhibitor sensitive if and only if its oncogenic transcriptional programme is maintained by an MLL1 species that carries an intact menin-binding motif (KMT2A residues 6–25), whether that MBM is **in cis** (inside the oncoprotein, as in KMT2A fusions and arguably NPM1c-driven disease) or **in trans** (on endogenous wild-type MLL1 that the oncoprotein recruits, as in NUP98 fusions). It is **not** a property of the fusion family, the FG repeats, or HOXA expression as such. The hypothesis unifies four observations that otherwise look unrelated: KMT2A fusions retain the MBM in cis and are sensitive; NUP98 fusions recruit wild-type MLL1 in trans and are sensitive but less so, with a ~6-fold IC50 spread across partners [E2]; **NUP98::KMT2A deletes KMT2A exon 1, removing the MBM from the very MLL1 species the fusion supplies, and is resistant while also being HOXA-low** [E1]; and UBTF tandem-duplication AML, which has no KMT2A lesion at all, is menin-inhibitor sensitive [E4] — consistent with a trans dependency on endogenous menin-MLL1. The hypothesis predicts a stratification variable that is measurable today and is measured by nobody: not 'is there a NUP98 rearrangement' but 'is there an intact menin-MLL1 tether serving this programme'.

Go deeper
The cis/trans distinction, and what it buys
lesionMLL1 species serving the programmeMBM presentmenin-inhibitor response
KMT2A::AFF1, ::MLLT3 etc.the fusion itselfin cissensitive, approved indication
NUP98::NSD1, ::KDM5A, ::HOXA9endogenous wild-type MLL1, recruitedin transsensitive, graded, ~6-fold IC50 spread 2
NUP98::KMT2Athe fusion itself, exon 1 deletedabsentresistant 1
NPM1cendogenous wild-type MLL1in transsensitive, approved indication
UBTF tandem duplicationendogenous wild-type MLL1 (inferred)in transsensitive 4

The fourth row is the discriminating one. It is a NUP98 fusion, it retains the FG repeats that form condensates, and it should be sensitive on any model that says 'NUP98 fusions are menin-dependent'. It is resistant, and the one thing that distinguishes it is that it supplies its own MLL1 without the menin tether — so the endogenous tethered MLL1 is presumably outcompeted at those sites.

That is also why the reduced HOXA output of NUP98::KMT2A is not a separate puzzle but the same one: if the MBM-less fusion occupies KMT2A target sites and cannot be tethered, the programme it should be driving is not driven.

What this predicts, in order of cheapness
  1. Restoring the MBM should restore sensitivity. A NUP98::KMT2A construct with KMT2A exon 1 added back should regain both HOXA output and menin-inhibitor sensitivity. One construct, one viability curve.
  2. Removing it in trans should abolish sensitivity in NUP98 fusions. In a NUP98-NSD1 or NUP98-KDM5A model, replacing endogenous Kmt2a with an MBM-deleted allele should phenocopy a menin inhibitor and render the cells drug-insensitive. This is the single cleanest test and it separates 'the drug hits menin' from 'the drug hits something else'.
  3. MEIS1 level should modify the response independently of MBM status, because MEIS1 re-expression abrogates sensitivity downstream 2. So the hypothesis predicts a two-variable response map: MBM availability × MEIS1 output.
  4. A clinical corollary. NUP98-rearranged patients whose fusion partner is KMT2A should be excluded from menin-inhibitor trials, and NUP98-rearranged patients in general should be stratified by MEIS1 expression rather than pooled.
What would kill it
  • A NUP98::KMT2A model that responds to a menin inhibitor, particularly to genetic MEN1 deletion rather than a micromolar tool compound. The current resistance evidence is pharmacological only 1, which is the hypothesis's weakest support.
  • An MBM-addback NUP98::KMT2A that stays resistant. That would show the exon-1 deletion is not what causes the resistance and would point at the long latency, the dominant-negative behaviour, or something else entirely.
  • A menin-inhibitor-sensitive leukaemia with no MBM-bearing MLL1 species available — for example a KMT2A-null model that still responds. That would show the drug works through a menin function unrelated to the MLL1 tether, of which menin has several.
  • An MBM-deleted endogenous Kmt2a allele that does NOT phenocopy the drug in a NUP98 fusion model. That would break the trans limb directly.
Consequence added 2026-09-21: the CALM-AF10 panel this project keeps designing is now ASYMMETRIC

Three separate beats have independently designed the same experiment — a menin-inhibitor dose-response across an index KMT2A-r line, U-937 (CALM-AF10), LOUCY (SET::NUP214) and an NPM1c control. This hypothesis makes that panel asymmetric, and nobody had said so.

Under the cis/trans account, a positive CALM-AF10 result is close to uninformative: it would mean CALM-AF10 recruits an MBM-bearing MLL1 species in trans, exactly as NUP98 fusions do, and would place it alongside NPM1c and NUP98 rather than distinguishing anything. A negative result is the informative one — a HOXA-driving leukaemia that is genuinely menin-independent would be the first, and would bound the hypothesis sharply.

That changes what the experiment is worth, not whether to run it. It should be run as a test of this hypothesis rather than as the separability control it was originally conceived as — and whoever allocates the material should know that the likely outcome is the uninformative one, because every KMT2A-free HOXA lesion tested so far has come back menin-dependent.

The corresponding update to hyp-menin-axis-separable-from-hoxa-programme is that its premise — 'contains no KMT2A, therefore should be menin-free' — is not the independent variable it needs to be. Menin dependence is a property of MBM availability, not of fusion family, and that entry's title has been corrected accordingly.

Facts and figures
statementA leukaemia is menin-inhibitor sensitive if and only if its oncogenic transcriptional programme is maintained by an MLL1 species carrying an intact menin-binding motif, in cis (inside the oncoprotein) or in trans (endogenous wild-type MLL1 recruited by it).
origindata-pattern
prior probability0.45
distinguishes from rivalThe rival hypothesis is that menin inhibition works on any HOXA/MEIS1-driven myeloid leukaemia. That rival predicts NUP98::KMT2A should be insensitive because it is HOXA-LOW, not because it lacks the MBM - so the addback experiment separates them: under this hypothesis addback restores both HOXA and sensitivity; under the rival, HOXA level is the only variable and the MBM is incidental.
predictions
[
  "Adding KMT2A exon 1 back to a NUP98::KMT2A construct restores HOXA output and menin-inhibitor sensitivity.",
  "Replacing endogenous Kmt2a with an MBM-deleted allele in a NUP98-NSD1 or NUP98-KDM5A model phenocopies a menin inhibitor and renders the cells drug-insensitive.",
  "Response is modified independently by MEIS1 output, giving a two-variable map of MBM availability by MEIS1 level.",
  "NUP98-rearranged patients whose partner is KMT2A should not respond to menin inhibitors and should be excluded from trials."
]
experiments
[
  "exp-mbm-addback-and-trans-deletion-panel"
]
analyses
[
  "ana-depmap-nup98-men1-power"
]
falsifiers
[
  "A NUP98::KMT2A model that responds to genetic MEN1 deletion.",
  "An MBM-addback NUP98::KMT2A that remains resistant.",
  "A menin-inhibitor-sensitive leukaemia with no MBM-bearing MLL1 species available, for example a KMT2A-null model that still responds.",
  "An MBM-deleted endogenous Kmt2a allele that fails to phenocopy the drug in a NUP98-fusion model."
]
How this could bite us back
  • If the hypothesis is right, the escape route nobody is watching is loss or mutation of the endogenous KMT2A menin-binding motif in trans, which pocket-directed resistance sequencing would not detect.
  • It also predicts that a patient with a NUP98 rearrangement and a coincidental KMT2A N-terminal lesion is not a candidate, which no eligibility criterion currently captures.
  • The clinical corollary - exclude NUP98::KMT2A patients - rests on two reported patients and one mouse model, and acting on it would deny a menin inhibitor to a patient on the strength of a single tool-compound experiment.
Still open
  1. Has anyone ever built an MBM-deleted endogenous Kmt2a allele? If such a mouse exists, the trans limb of this hypothesis could be tested in a month rather than a year.
  2. Does NPM1c-driven leukaemia fit the trans limb, and if so is its greater sensitivity relative to NUP98 fusions explained by how much MLL1 it recruits?
  3. Is MEIS1 output measurable in a diagnostic sample in a way that could be used to stratify a trial?
  4. Does the hypothesis predict anything about resistance? A subclone that loses the MBM in trans should escape without a MEN1 pocket mutation, which nobody is screening for.
Every source behind this entry (4)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    NUP98::KMT2A deletes KMT2A exon 1, which encodes the menin-binding motif, and is resistant to the menin inhibitor MI-2-2 at 3-12 uM while expressing reduced rather than elevated HOXA cluster genes.

    PMID 31558671DOI
    What this evidence cannot show

    A single laboratory, a transgenic overexpression model, and a tool compound at micromolar concentrations with no genetic MEN1 deletion arm. Only two patients with this fusion have ever been molecularly confirmed. The mapping from 'exon 1 lost' to 'residues 6-25 lost' combines the paper with UniProt annotation rather than coming from one source.

    Evidence that argues against this (1)
    • The model's 80-week median latency and MDS-like phenotype mean its cells may differ from other NUP98-fusion leukaemias for reasons unrelated to the menin tether, so attributing resistance specifically to MBM loss is an inference the paper supports but does not prove by rescue. Kaltenbach S et al. 2010
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    NUP98-fusion leukaemia depends genetically on Men1 and Kmt2a (CRISPR dropout comparable to the essential-gene control Rpa3), with a graded pharmacological sensitivity (VTP50469 IC50 84 nM for NUP98-KDM5A versus 517 nM for NUP98-HOXA9), a differentiation rather than apoptotic phenotype, MEIS1 re-expression abrogating sensitivity, and 2 of 3 PDX models responding.

    The menin-MLL1 interaction is a molecular dependency in NUP98-rearranged AML · Heikamp EB, et al. · Blood 139(6):894-906 2022
    PMID 34582559DOI
    What this evidence cannot show

    Full-text details were read in a delegated research pass in this session. CRISPR dropout of Men1 and Kmt2a establishes a requirement but not that the requirement runs through the MBM interface specifically; no MBM-deleted Kmt2a allele was used, which is exactly the experiment this hypothesis needs.

    Evidence that argues against this (1)
    • Genetic dropout of Kmt2a and Men1 is equally consistent with a requirement for the MLL1 complex as a whole, independent of the menin tether. Without an MBM-deletion allele, 'menin dependency' and 'MLL1 dependency' are not separated by any published experiment in this disease. Xu H et al. 2016
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt annotates the KMT2A menin-binding motif at residues 6-25 and LEDGF-interaction motifs at 123-134 and 147-152, all within the region encoded by exon 1.

    What this evidence cannot show

    Primary gene name and sequence length were read back from the fetched record because UniProt's gene_exact: search matches synonyms. The assignment of residues 6-25 to exon 1 is standard but was not verified against a transcript structure in this session.

    Evidence that argues against this (1)
    • UniProt feature boundaries are curated from primary literature of varying age and the exact MBM limits differ between structural papers; a few residues either way would not change this hypothesis, but a substantially different boundary would change which breakpoints are predicted to be resistant. UniProt Consortium 2026
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    AML with UBTF tandem duplications, which carries no KMT2A lesion, is sensitive to menin inhibitors - consistent with a dependency on endogenous, MBM-bearing menin-MLL1 acting in trans.

    PMID 37890156DOI
    What this evidence cannot show

    Cited from the title and abstract; the primary data were not read in this session. The inference that the dependency is on endogenous MBM-bearing MLL1 is this hypothesis's, not the paper's, and has not been tested in that system.

    Evidence that argues against this (1)
    • That a third, unrelated lesion class is also menin-sensitive could equally mean that menin inhibition works on any HOXA/MEIS1-driven myeloid leukaemia regardless of tether topology - which would make the cis/trans distinction decorative rather than explanatory. Wenge DV et al. 2025
Connected to