Oncofusion

start / graph / hyp-menin-axis-separable-from-hoxa-programme

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q55-calm-af10

CALM-AF10 was meant to separate the menin axis from the HOXA programme - and the indirect evidence has now turned against separability, because KMT2A-free HOXA lesions keep proving menin-DEPENDENT

also called menin dependence versus HOXA dependence · the CALM-AF10 test

Two ideas have been quietly treated as one: that leukaemias need the menin protein, and that they run on the HOXA growth genes. CALM-AF10 turns on the same HOXA genes and needs the same DOT1L enzyme as the KMT2A leukaemias, but it contains no KMT2A at all - so if it does not need menin, the two ideas come apart.

The same thing, for a scientist

**Statement.** *Menin dependence and the HOXA/MEIS1 programme are separable.* PICALM::MLLT10 drives the HOXA5/7/9/10 + MEIS1 programme and is absolutely DOT1L-dependent [E1][E2] - with one qualification the brief's premise does not carry: the single published gene-level requirement test found **Hoxa5**, not Hoxa9, to be required, and found that requirement NOT shared with MLL-AF10 [E8], while containing **no KMT2A, no menin-binding motif and no CXXC/post-CXXC chromatin address**. If it is menin-independent, then 'HOXA-high' does not imply 'menin-sensitive', and the several entries in this graph that use the two interchangeably are overstated. Prior probability that CALM-AF10 is **menin-independent: 0.7**, resting on three indirect lines and no direct test. **The direct test has never been published** - `CALM-AF10 AND menin` returns 0 in PubMed with clean operand controls, and the 2026 comprehensive review of the disease never uses the word [E3][E4]. **This is a hypothesis about an untested question, not a summary of a settled one**, and the honest statement of the current position is: *probably menin-independent, nobody has looked, and one two-week experiment would settle it.*

Go deeper
The three indirect lines, and how much each is worth

1. Structural absence - worth a lot, but it is an argument from architecture. A menin-KMT2A inhibitor works by displacing the fusion's N-terminal MBM motifs from menin's central cavity. CALM-AF10 has no KMT2A sequence at all, so there is nothing for the drug to displace. The only route to menin dependence would be an indirect one - CALM-AF10 needing the wild-type menin-KMT2A1 complex to keep HOXA open, which is exactly how NPM1c and NUP98-rearranged leukaemias are menin-dependent. Nobody has tested that route here, so the architectural argument does not close the question; it only says the direct route is unavailable.

2. DepMap - worth little on its own, but it points the same way. In the CALM-AF10 line U-937: MEN1 -0.229, PSIP1/LEDGF +0.135, DOT1L -0.542; against KMT2A-rearranged means of -1.263, -0.553 and -0.604 and a non-KMT2A leukaemia MEN1 mean of -0.401 (SD 0.406, n=42) 5. The shape is DOT1L-dependent at the KMT2A-rearranged level, menin- and LEDGF-independent below the non-KMT2A average. n = 1 line, so this is a hypothesis generator and nothing more; it is reported because it is the only quantitative datum that exists.

3. The CRISPR-ChIP partition - worth a lot, with a system caveat that matters. H3K79 methylation partitions into a fusion-directed, menin-dependent oncogenic DOT1L complex and a separate endogenous DOT1L complex directed by MLLT10 6. CALM-AF10 reaches DOT1L through MLLT10's own OM-LZ interface - the half of the partition the screen assigns to the menin-independent unit. But that partition was established in murine MLL-AF9 AML (resolved from the study's GEO depositions, not stated in the abstract), so transferring it to a human T-lineage CALM-AF10 crosses species, lineage and fusion at once.

What would make the hypothesis wrong

The strongest counterweight is not a paper, it is a protocol. AUGMENT-101 cohort 2D enrols relapsed/refractory acute leukaemia with 'a genetic mutation expected to lead to HOX/MEIS upregulation' 7. The sponsor's operating hypothesis is therefore exactly the one this entry proposes to falsify: that HOX/MEIS dysregulation predicts menin sensitivity irrespective of mechanism. If a CALM-AF10 patient responds on that cohort, this hypothesis is in trouble.

The second counterweight is mechanistic and specific: wild-type MLL1 requirement. Several HOXA-driven, non-KMT2A leukaemias keep HOXA open through the endogenous menin-KMT2A1 complex. If CALM-AF10 does the same - and it must put HOXA into an active state somehow before DOT1L can maintain it - then it is menin-dependent by the indirect route and the separation fails.

What this changes inside this graph if it holds

Entries that treat 'menin dependence' and 'the HOXA programme' as one thing would need splitting into two claims with different evidence. The concrete consequence for the lab's own bet is favourable, not damaging: if menin dependence is a property of the KMT2A-menin contact specifically rather than of HOXA output generally, then the menin-interface hypothesis becomes more precise, not less supported - it stops being 'menin runs HOXA leukaemias' and becomes 'menin is the tether a KMT2A fusion needs'. The thing it damages is the breadth of the menin-inhibitor indication, not the mechanism.

Counter-evidence added 2026-09-21, and the title now overstates what survives

Five of this entry's eight evidence items carried no objection at all until this cycle, in the entry the whole menin-interface hypothesis is graded against. Writing them changes how the entry reads.

Two of the four load-bearing items do not discriminate between the arms. E1 shows a CALM-AF10 transgene drives Hoxa and leukaemia — the output KMT2A fusions also produce. E2 shows DOT1L dependence — which KMT2A-rearranged leukaemia also has. Neither tests menin. They establish that the HOXA programme runs in a KMT2A-free lesion, which nobody disputes, and they are silent on the question the entry is named for.

Two more are absences. E3 counts zero PubMed records for CALM-AF10 with menin; E4 notes a 2026 review does not mention menin. Both are facts about where the field has looked.

And the graph holds a positive result pointing the other way. Chromatin-bound CRM1 recruits SET-NUP214 and NPM1c onto HOXA, and NPM1c AML is clinically menin-inhibitor sensitive — so a KMT2A-free, CRM1-addressed, HOXA-driving lesion is already known to be menin-dependent. AUGMENT-101 cohort 2D encodes exactly that expectation by enrolling NPM1m alongside KMT2Ar.

What actually survives: that no published experiment has tested menin dependence in CALM-AF10, and that the one indirect number available (MEN1 gene effect in U-937, n=1) was bounded by this project's own power analysis as consistent with no difference. confidence_overall is already speculative and should stay there. The title — "the indirect evidence says they ARE separable" — is stronger than the entry now supports, and is flagged here rather than silently rewritten, because the title is what gets quoted and the change of reading is the finding.

CORRECTED 2026-09-21: the title is now fixed, not merely flagged, because the evidence moved

Last cycle this entry gained a note saying its title overstated what survived its own counter-evidence. The title is now corrected outright, because a decisive result was found.

PMID 34582559 (Blood 2022) is titled "The menin-MLL1 interaction is a molecular dependency in NUP98-rearranged AML." NUP98-HOXA9 and NUP98-JARID1A lines respond to VTP50469 at an IC50 the authors report as similar to KMT2A-rearranged and NPM1c cells, MLL1 and the NUP98 fusion are evicted from chromatin together, and the inhibitor prolongs survival in NUP98-NSD1 and NUP98-JARID1A PDX.

A KMT2A-free nucleoporin fusion, with no menin-binding motif of its own, driving HOXA, is menin-dependent. It reaches menin through wild-type MLL1, which every cell has. With NPM1c already established clinically (PMID 36922593), that is two independent KMT2A-free HOXA lesions on the menin-dependent side.

So "contains no KMT2A" is not the independent variable this control needs it to be. The comparison arm does not merely fail to support separability; it now leans against it.

And the experiment this entry proposes has become asymmetric. A menin-inhibitor dose-response in CALM-AF10 that comes back POSITIVE would no longer refute the handshake account — it would simply add CALM-AF10 to the NPM1c and NUP98 list. Only a negative result is now strongly informative. That is worth knowing before anyone runs it, and it lowers the expected value of an experiment three separate beats have independently designed.

confidence_overall stays speculative, which it should have been all along.

Facts and figures
statementMenin dependence and the HOXA/MEIS1 programme are separable: PICALM::MLLT10 drives HOXA/MEIS1 and requires DOT1L while containing no KMT2A, no menin-binding motif and no CXXC address, and is predicted to be menin-inhibitor insensitive.
originliterature-gap
prior probability0.7
what it would change in this graphClaims that use 'menin-dependent' and 'HOXA-driven' interchangeably would have to be split. The menin-interface hypothesis itself would become sharper, not weaker: menin dependence would be a property of the KMT2A-menin contact rather than of HOXA output.
relation to ledger item GLedger item G asks whether cohesin-mutant, KMT2A-wild-type AML is menin-inhibitor sensitive. That is the same separation asked of a different lesion class, and CALM-AF10 is the cleaner instrument: cohesin mutation is a background alteration in a heterogeneous genome, whereas CALM-AF10 is a single defined driver with a known HOXA output and a known DOT1L dependence. If only one of the two can be run, run this one - and if both are run, a shared result across two mechanistically unrelated KMT2A-wild-type HOXA-high lesions is far stronger than either alone.
premise qualificationThe brief's premise - that CALM-AF10 drives 'the same HOXA/MEIS1 programme' - holds at cluster and output level and fails at the level of which gene is required: Hoxa5 knockout attenuates CALM-AF10 and not MLL-AF10 (PMID 16921363, read in full text). The menin experiment must therefore read out the whole HOXA cluster, so that a negative cannot be attributed to a programme difference instead of to the absent menin contact.
predictions
[
  "Revumenib or VTP-50469 at full target engagement produces no growth inhibition and no HOXA/MEIS1 collapse in U-937 or in a murine CALM-AF10 model, while the same exposure collapses both in a KMT2A-rearranged control run side by side.",
  "MEN1 CRISPR knockout in a CALM-AF10 line gives a gene effect near zero while MEN1 knockout in a matched KMT2A-rearranged line is strongly depleting.",
  "H3K79me2 at HOXA in a CALM-AF10 line is unchanged by menin inhibition, because it belongs to the MLLT10-directed unit.",
  "If a CALM-AF10 patient is ever treated on a HOX/MEIS eligibility criterion, they do not respond."
]
experiments
[
  "exp-menin-inhibitor-dose-response-calm-af10-panel"
]
analyses
[
  "ana-depmap-men1-dependency-calm-af10-u937",
  "ana-calm-af10-menin-bounded-absence"
]
falsifiers
[
  "A menin inhibitor collapses HOXA/MEIS1 or kills cells in a CALM-AF10 model at exposures matched to a KMT2A-rearranged control - the separation fails and menin dependence tracks the HOXA programme rather than the KMT2A contact.",
  "MEN1 knockout depletes a CALM-AF10 line as strongly as it depletes KMT2A-rearranged lines.",
  "CALM-AF10 is shown to require wild-type KMT2A1 to establish or maintain HOXA, which would make it menin-dependent by the indirect NPM1c/NUP98 route even without a menin-binding motif.",
  "A CALM-AF10 patient responds to a menin inhibitor on AUGMENT-101 cohort 2D or an equivalent HOX/MEIS-defined cohort.",
  "The HOXA programmes turn out to be different enough - different required cluster members, different MEIS1 dependence - that a menin-negative result in CALM-AF10 cannot be attributed to the missing menin contact rather than to the programme difference."
]
How this could bite us back
  • If the hypothesis is right, HOX/MEIS-based eligibility criteria will enrol patients who cannot benefit, exposing them to differentiation syndrome and QTc risk for no mechanistic reason.
  • If it is wrong, the menin indication is broader than currently written - which is good for patients and bad for this graph's mechanistic story, because it would mean menin dependence does not require a menin-binding motif.
Still open
  1. Does CALM-AF10 require wild-type KMT2A1 for HOXA maintenance? This is the single question that decides the hypothesis and it has no published answer.
  2. Has any menin-inhibitor programme screened a CALM-AF10 line and not published it? U-937 is a standard control line in cell-line panels, so the data may exist as an unremarked negative in supplementary tables.
  3. If CALM-AF10 is menin-independent, what keeps HOXA in an active, DOT1L-accessible state in the first place? CTCF is required for CALM-AF10 leukaemia, which suggests an architectural answer rather than a menin one.
Every source behind this entry (9)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In transgenic mice, a CALM-AF10 transgene drives Hoxa cluster overexpression and acute leukaemia.

    PMID 17804713DOI
    What this evidence cannot show

    Murine transgenic, long latency; cooperating events are required.

    Evidence that argues against this (1)
    • DRIVING HOXA IS THE SHARED OUTPUT, NOT THE DISCRIMINATING ONE. That a CALM-AF10 transgene produces Hoxa overexpression and leukaemia establishes the programme runs; it does not test menin at all, and a KMT2A fusion does the same thing. An item that measures only the output common to both arms cannot support a claim about what separates them. It is also a murine transgenic, with expression from a non-endogenous promoter and a cell of origin set by the construct. Oncofusion knowledge graph (internal entry) 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In mouse models, CALM-AF10 transformation requires Dot1l for both initiation and maintenance, and EPZ004777 collapses Hoxa and Meis1 expression in CALM-AF10-transformed cells.

    PMID 23138183DOI
    What this evidence cannot show

    Murine bone marrow. Establishes the DOT1L half of the dissociation; the menin half is untested.

    Evidence that argues against this (1)
    • DOT1L DEPENDENCE IS EVIDENCE OF SIMILARITY, NOT OF SEPARABILITY, and the two are not alternatives. KMT2A-rearranged leukaemia is also DOT1L-dependent and its HOXA/MEIS programme also collapses under EPZ004777, so this result places CALM-AF10 alongside the KMT2A fusions rather than opposite them. Nothing in this experiment tested menin in the same system, which is the measurement the hypothesis needs and still lacks. Oncofusion knowledge graph (internal entry) 2026
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    As of 2026-09-20, `CALM-AF10 AND menin` returns 0 PubMed records while `CALM-AF10` alone returns 81 and `menin` alone returns 2,237; `PICALM AND menin` = 0 and `AF10 AND menin` = 0; `MLLT10 AND menin` = 3, all about KMT2A::MLLT10.

    What this evidence cannot show

    Bounded absence with operand controls, per brief rule 67, and with the drug-name list per rule 72. It bounds the published literature, not the experiment: a negative screening result inside a pharmaceutical menin programme would never appear in PubMed.

    Evidence that argues against this (1)
    • A COUNTED ZERO IS BOUNDED BY ABSTRACTS (rule 80), AND MORE IMPORTANTLY IT IS THE WRONG KIND OF EVIDENCE FOR THIS CLAIM. That nobody has published on CALM-AF10 and menin is a fact about what the field has looked at, not about whether the dependence exists - and this graph holds a positive result pointing the other way: chromatin-bound CRM1 recruits SET-NUP214 AND NPM1c onto HOXA, and NPM1c AML is clinically menin-inhibitor sensitive. So a KMT2A-free, CRM1-addressed, HOXA-driving lesion is ALREADY KNOWN to be menin-dependent. An absence of study cannot outweigh a positive instance of the thing it is being used to deny. Oncofusion knowledge graph (internal entry) 2026
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.review

    A 2026 comprehensive review of PICALM::MLLT10 leukaemia covering mechanism, prognosis and treatment does not mention menin at all.

    PICALM::MLLT10 translocated leukemia. · Cullen JM, et al. · FEBS Lett 2026
    PMID 41532548DOI
    What this evidence cannot show

    REVIEW, cited as a map of field attention rather than as biological evidence. Full text fetched and searched 2026-09-20 (69,891 characters).

    Evidence that argues against this (1)
    • A REVIEW'S SILENCE MEASURES THE FIELD'S ATTENTION, NOT THE BIOLOGY, and reviews lag their literature by a year or more. A 2026 PICALM::MLLT10 review omitting menin is consistent with menin dependence being real and simply untested in this lesion - which is precisely what this graph says elsewhere. Using the omission as evidence for separability converts a gap in the literature into a biological claim, the same move rule 80 exists to prevent. Oncofusion knowledge graph (internal entry) 2026
  5. E5weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.dataset

    In DepMap CRISPR leukaemia screens, the CALM-AF10 line U-937 shows MEN1 -0.229 and PSIP1 +0.135 against KMT2A-rearranged means of -1.263 and -0.553, with DOT1L at -0.542 versus a KMT2A-rearranged mean of -0.604.

    What this evidence cannot show

    One cell line. See ana-depmap-men1-dependency-calm-af10-u937 for the full method, the second line with no data, and why no selectivity statistic is computed.

    Evidence that argues against this (1)
    • The AUGMENT-101 protocol encodes the opposite hypothesis: cohort 2D enrols on 'a genetic mutation expected to lead to HOX/MEIS upregulation', i.e. it treats HOX/MEIS output as the predictor of menin sensitivity irrespective of whether a menin-binding motif is present. A response in such a patient would falsify this entry. ClinicalTrials.gov 2026
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    CRISPR-ChIP partitions H3K79 methylation into a menin-dependent, MLL-fusion-directed oncogenic DOT1L complex and a separate endogenous DOT1L complex whose activity is directed by MLLT10.

    PMID 37679565DOI
    What this evidence cannot show

    Abstract only (paywalled; DOI route returned a 303-character stub). System resolved from GEO: the H3K79me2 CRISPR-ChIP and menin-inhibitor ChIP-seq arms are MURINE MLL-AF9 AML; the human arms are K562 and cover H3K4me3/Pol II. Transfer to CALM-AF10 crosses species, lineage and fusion.

    Evidence that argues against this (1)
    • The partition was established in murine MLL-AF9 AML, resolved from the study's GEO depositions rather than stated in the abstract; its human arms cover H3K4me3 and Pol II in K562, not H3K79me2. Applying it to a human T-lineage CALM-AF10 crosses species, lineage and fusion partner simultaneously. Gilan O et al. 2023
  7. E7established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.trial_registry

    AUGMENT-101 cohort 2D enrols relapsed/refractory acute leukaemia with a genetic mutation expected to lead to HOX/MEIS upregulation, exemplified by KMT2Ar, NPM1m and NUP98r.

    NCT04065399
    What this evidence cannot show

    Read from the raw ClinicalTrials.gov v2 API 2026-09-20. Eligibility wording encodes a sponsor hypothesis; it is not evidence that the hypothesis is right.

    Evidence that argues against this (1)
    • THIS ITEM CUTS AGAINST THE HYPOTHESIS AS EASILY AS FOR IT, AND THE NPM1m ARM IS WHY. AUGMENT-101 cohort 2D enrols leukaemias expected to show HOX/MEIS upregulation and exemplifies them with KMT2Ar, NPM1m and NUP98r. **NPM1-mutant AML is CRM1-addressed and is menin-inhibitor responsive**, so the trial's own eligibility logic is that a KMT2A-free HOX/MEIS lesion can be menin-dependent. That PICALM::MLLT10 is not named is an eligibility and attention fact - this graph separately records that menin and DOT1L trial eligibility excludes it - and eligibility criteria are written from what sponsors expect to enrol, not from what is mechanistically excluded. Oncofusion knowledge graph (internal entry) 2026
  8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Hoxa5 knockout attenuates the transformation capability of CALM-AF10 in murine bone marrow but has no effect on that of MLL-AF10, and Hoxa9 was undetectable in the CALM-AF10-transformed cells.

    PMID 16921363DOI
    What this evidence cannot show

    Murine; read in full text 2026-09-20 (PMC4425349). It complicates rather than defeats the hypothesis: the two fusions differ in WHICH HOXA gene they require, which is an argument that they reach the cluster by different routes - the same direction as menin independence - but it also weakens the clean 'identical programme' framing the test relies on.

    Evidence that argues against this (1)
    • If CALM-AF10 and a KMT2A fusion require different HOXA genes, then CALM-AF10 is a less clean control than the brief assumes, and a negative menin result in it could be attributed to the programme difference rather than to the missing menin contact. The experiment must therefore report HOXA cluster-wide expression, not a single gene. Caudell D et al. 2007
  9. E9strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.journal-article

    NUP98-rearranged AML, a KMT2A-free fusion driving HOXA with no menin-binding motif of its own, is dependent on the menin-MLL1 interaction: NUP98-HOXA9 and NUP98-JARID1A lines respond to VTP50469 at an IC50 reported as similar to KMT2A-rearranged and NPM1c cells, MLL1 and the NUP98 fusion are evicted from chromatin together, and the inhibitor prolongs survival in NUP98-NSD1 and NUP98-JARID1A patient-derived xenografts.

    The menin-MLL1 interaction is a molecular dependency in NUP98-rearranged AML. · Heikamp EB, Henrich JA, Perner F, et al. · Blood 2022
    PMID 34582559DOI
    What this evidence cannot show

    Preclinical throughout - lines and PDX, no clinical arm for NUP98, unlike NPM1c where the evidence is a registrational trial. 'Similar IC50' is the authors' characterisation and this entry has not re-derived the comparison from their figures. VTP50469 is a tool compound closely related to revumenib rather than the clinical agent. Read at abstract level for this item.

    Evidence that argues against this (1)
    • THIS DOES NOT SHOW CALM-AF10 IS MENIN-DEPENDENT, and the inference to it is by analogy rather than measurement. NUP98 fusions and CALM-AF10 reach HOXA by different routes - the NUP98 FG repeats and the CALM-AF10 OM-LZ/DOT1L contact are not the same mechanism - so a menin requirement in one does not transfer. What it removes is the argument that a KMT2A-free lesion must be menin-free, which is a weaker and different thing. Issa GC et al. 2023
Connected to