Oncofusion

start / graph / hyp-set-nup214-hoxa-runs-through-recruited-wild-type-mll1

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q58-set-nup214

Hypothesis: SET::NUP214 activates HOXA partly through recruited wild-type MLL1, which would make a KMT2A-unrearranged leukaemia menin-inhibitor-sensitive

also called the wild-type MLL1 route · menin dependence without a KMT2A rearrangement

If SET::NUP214 switches on HOXA by borrowing the normal, unbroken KMT2A protein, then the menin drugs that pull KMT2A off chromatin should work in this leukaemia too - even though it has no KMT2A rearrangement at all.

The same thing, for a scientist

**Statement.** In SET::NUP214-positive leukaemia cells, a measurable fraction of HOXA activation is contributed by **wild-type MLL1 recruited into the fusion's nuclear bodies through the SET acidic region**, and that fraction is menin-dependent, so a clinical menin inhibitor should lower HOXA expression and induce differentiation in LOUCY and MEGAL. **Origin: literature-gap, sharpened by a contradiction inside this beat's own brief.** SET::NUP214 was dispatched as a KMT2A-free comparator on the reasoning that a fusion with no menin-binding motif cannot need menin. Two papers place wild-type MLL1 in the mechanism [see `mechanism-set-nup214-recruits-wild-type-mll1`], and wild-type MLL1 carries MBM1/MBM2. Nobody has tested it in any register [see `ana-set-nup214-menin-bounded-absence`]. **Prior probability: 0.25.** Low, and deliberately so. The competing route - chromatin-bound CRM1 as the recruiter - has direct genome-wide evidence and a working pharmacology, while the MLL1 route rests on one abstract-level reporter assay and one abstract-level condensate claim. But the graph already holds a precedent in the adjacent family: **NUP98 fusions, which are also nucleoporin fusions forming nuclear bodies, are menin/MLL1-dependent and a NUP98-NSD1 patient achieved complete remission on a menin inhibitor.** If the MLL1 recruitment is a property of the nucleoporin nuclear body rather than of the NUP98 partner, SET::NUP214 should behave the same way. **This would be a human adult/paediatric T-ALL and AML result, two steps from infant B-lineage KMT2A::AFF1.**

Go deeper
What distinguishes this from the rival model

The rival is mechanism-set-nup214-crm1-is-the-address-not-the-fusion: CRM1 sits on HOX first, the fusion docks on it, and everything downstream follows from that. The two are not mutually exclusive - CRM1 could deliver the fusion and the fusion could then bring MLL1 - and the experiment below is designed to read them separately rather than to pick one.

The discriminating prediction is ordering under inhibitor: if MLL1 is downstream of CRM1 delivery, selinexor should remove MLL1 from HOXA and menin inhibition should not remove the fusion. If the two routes are parallel, each inhibitor removes its own arm and only the combination collapses HOXA.

Why the answer matters beyond this fusion

Menin-inhibitor eligibility is currently written around KMT2A rearrangement and NPM1 mutation. This graph already records that partner-blind eligibility is a risk and that no trial has stratified by partner. A positive result here would add a third genotype that has no KMT2A rearrangement at all, and would say that the biomarker for menin dependence is 'wild-type MLL1 doing load-bearing work at HOXA', not 'KMT2A is broken'.

A negative result is worth almost as much: it would make SET::NUP214 the clean second instance alongside CALM-AF10 that the coordinator wanted, and sharpen the menin-interface hypothesis by showing that HOXA-high plus DOT1L plus CRM1 is achievable with the menin axis untouched in two mechanistically unrelated lesions.

Facts and figures
statementIn SET::NUP214-positive leukaemia cells, part of HOXA activation is contributed by wild-type MLL1 recruited through the SET acidic region, and that part is menin-dependent, so a clinical menin inhibitor lowers HOXA and induces differentiation in LOUCY and MEGAL.
originliterature-gap
prior probability0.25
rival modelmechanism-set-nup214-crm1-is-the-address-not-the-fusion
rivals are exclusivefalse
predictions
[
  "Menin co-immunoprecipitates with SET-NUP214 from LOUCY nuclear extract.",
  "Menin and wild-type MLL1 are detectable by CUT&RUN at the HOXA9 and HOXA10 promoters in LOUCY and not in the fusion-negative comparator SKW3.",
  "A clinical menin inhibitor (revumenib or ziftomenib) lowers HOXA9/HOXA10 transcript in LOUCY by a measurable but INCOMPLETE amount - partial, because the CRM1 arm remains.",
  "Menin inhibition does NOT reduce chromatin-bound CRM1 or the fusion's own HOXA occupancy, because it acts on the recruited protein and not on the platform.",
  "Menin inhibitor plus selinexor is more than additive on HOXA suppression if the two arms are parallel."
]
falsifiers
[
  "Menin is not detectable in a SET-NUP214 immunoprecipitate from LOUCY at a sensitivity that recovers CRM1 in the same lysate - this alone reduces the hypothesis to a reporter-assay artefact.",
  "A clinical menin inhibitor at exposures that displace menin from chromatin in a KMT2A::AFF1 control line produces no change in HOXA9/HOXA10 transcript in LOUCY or MEGAL.",
  "Wild-type MLL1 is not detectable at HOXA promoters in LOUCY by CUT&RUN.",
  "The Cigdem 2021 MLL construct turns out to be KMT2B or KMT2D rather than KMT2A, which would move the claim to a different protein that this graph already records as a separate menin binder with a different affinity."
]
experiments
[
  "exp-menin-in-the-set-nup214-complex-and-menin-inhibitor-in-loucy"
]
analyses
[
  "ana-set-nup214-menin-bounded-absence"
]
How this could bite us back
  • Even if true, the predicted HOXA reduction is partial by construction, and a partial reduction in a disease with 100% early chemotherapy resistance is unlikely to be a single-agent story.
  • Proposing a menin inhibitor for a genotype with no trial data invites off-label use in an adult T-ALL population where the toxicity profile - differentiation syndrome, QTc, the endocrine exposures this graph documents - has never been characterised.
  • A positive co-IP would be seized on as evidence for menin-inhibitor eligibility long before a functional experiment existed; the entry should be written so that the co-IP is explicitly insufficient.
Still open
  1. If wild-type MLL1 can be recruited into a nucleoporin nuclear body, is that a general property of FG-repeat condensates? DEK-NUP214 and SQSTM1-NUP214 would test it and neither has been looked at this way.
  2. Is the recruited MLL1 the full-length protein or the processed N-terminal fragment? Only the N-terminal fragment carries MBM1/MBM2, and the two are separated by taspase cleavage.
  3. Does menin inhibition in a HOXA-high, KMT2A-unrearranged leukaemia have any clinical precedent at all outside NPM1-mutant AML?
Every source behind this entry (3)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    SET-NUP214 interacts with MLL via the SET acidic region and the two cooperatively enhance the HOXA10 promoter, with neither half of the fusion sufficient alone.

    PMID 34320268DOI
    What this evidence cannot show

    ABSTRACT-LEVEL ONLY; not in PMC. A reporter assay in an unstated system with MLL unresolved to an accession. This is the weakest kind of support for a hypothesis of this consequence, which is why the prior is set at 0.25.

    Evidence that argues against this (1)
    • The rival model has much stronger evidence. Genome-wide ChIP-seq in LOUCY shows chromatin-bound CRM1 at HOXA and HOXB independently of the fusion - knocking down SET-Nup214 lowers HOX genes without reducing bound CRM1 - and CRM1 inhibition removes the fusion from chromatin and lowers HOX expression. A complete account of HOXA activation is available with no MLL1 in it, so the MLL1 arm is an addition the data do not currently require. Oka M et al. 2019
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The phase-separated nuclear bodies of NUP98::HOXA9 and SET::NUP214 promote the condensation of MLL1 together with CRM1 and co-localise with them on chromatin.

    PMID 37516964DOI
    What this evidence cannot show

    ABSTRACT-LEVEL ONLY; Europe PMC returns pmcid null. The MLL1 claim is made jointly for two fusions and the extent to which it was demonstrated for SET::NUP214 in its own right is unresolved. Standing condensate caveat applies: hexanediol and classical FRAP cannot distinguish LLPS from binding to clustered sites.

    Evidence that argues against this (1)
    • SPECIFIC MECHANISTIC DEFEATER, and it predicts the OPPOSITE sign from this hypothesis. The hypothesis needs SET::NUP214 to deliver an MLL1 that still has menin on it. But the one paper that maps where SET touches MLL puts the contact at the AMINO-TERMINAL REGION of MLL, via TAF-Ibeta's acidic region - the same end of MLL1 that carries the menin-binding motifs MBM1 and MBM2. Nothing has tested whether menin and SET/TAF-Ibeta can occupy MLL1's N-terminus at the same time. If they cannot, SET::NUP214 would DISPLACE menin from MLL1 rather than recruit a menin-loaded complex, and a menin inhibitor would be redundant in this lesion instead of effective. The hypothesis and its inverse are both consistent with every published observation here, and the experiment that separates them - menin blot on an MLL1 immunoprecipitate from LOUCY, with and without fusion knockdown - has not been run. Shimoyama T et al. 2005
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    PubMed returns 0 records for SET-NUP214 or LOUCY against a menin and menin-inhibitor name list on 2026-09-20, against operand controls of 2,237 for menin and 54 for SET-NUP214 and a positive control of 1 for SET-NUP214 AND DOT1L through the identical shape; and neither LOUCY nor MEGAL is present in the DepMap CRISPR set served by Open Targets.

    What this evidence cannot show

    An absence of testing, not a negative result. Bound: titles, abstracts and MeSH only, and one DepMap release whose version the API does not state. Conference abstracts, preprints and supplementary cell-line panels were not searched.

    Evidence that argues against this (1)
    • PUBLISHED RESULT THAT WITHDRAWS THE ASYMMETRY THIS ITEM IS USED FOR. An absence of testing supports no hypothesis, and the only comparable lesion class that has been tested went the other way: VTP50469 is active in NUP98-HOXA9 and NUP98-JARID1A leukaemia at an IC50 similar to KMT2A-rearranged and NPM1c cells, and evicts MLL1 and the fusion from chromatin together. That is a positive result for menin dependence in a nucleoporin fusion, which supports the recruited-MLL1 half of this hypothesis and simultaneously removes the untested-therefore-independent reading that the rest of this beat leans on. The item should be read as arguing for this hypothesis and against the entry it was originally written to support. Heikamp EB et al. 2022
Connected to