Oncofusion

start / graph / concept-three-addresses-one-output-what-set-nup214-does-to-the-menin-interface-hypothesis

Explainermoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q58-set-nup214

Three addresses, one output: what SET::NUP214 actually does to the menin-interface hypothesis, and why it is a weaker test than CALM-AF10

also called three routes to HOXA · addressing versus output · the SET-NUP214 verdict

Three different leukaemia fusions all switch on the same HOXA genes, but they find those genes in three different ways. Comparing how they get there - rather than what they turn on - is what tells you whether the menin drug target is essential or incidental.

The same thing, for a scientist

**The menin-interface hypothesis** holds that a KMT2A fusion is leukaemogenic because it combines the **menin contact** (MBM1/MBM2) with the **CXXC/post-CXXC address**. Testing it means finding lesions that reach the same output without those parts. This beat was the second such test. Here is what it returned. | | address | menin contact in the lesion | DOT1L route | tested against menin | |---|---|---|---|---| | **KMT2A::AFF1** | CXXC/post-CXXC, direct DNA binding | **yes, in the fusion** | via AFF1/SEC | extensively | | **CALM-AF10** | AF10 OM-LZ plus a CRM1 export requirement | **no** | AF10 OM-LZ, defined interface | **0 records** | | **SET::NUP214** | **borrowed** - chromatin-bound CRM1 is already at HOX and the FG repeats dock onto it | **no in the fusion, YES in recruited wild-type MLL1** | unmapped surface, AF10 absent from the complex | **0 records** | **Three findings, in descending order of how much they change things.** **1. SET::NUP214 is a WEAKER test than the brief assumed, and this is a correction the beat owes.** The fusion has no menin-binding motif, but it **recruits wild-type MLL1**, which has two. The inference 'no MBM in the fusion, therefore menin-independent' does not hold for a lesion that recruits the protein carrying the motif. **CALM-AF10 remains the cleaner instance; SET::NUP214 is an open question, not a second confirmation.** **2. DOT1L recruitment across these lesions is CONVERGENT, not shared.** CALM-AF10 and KMT2A::MLLT10 use one defined interface (AF10 OM-LZ). SET::NUP214 has no OM-LZ and the founding study looked for AF10 in its complex and did not find it. **That is a more interesting claim than shared recruitment**: it says the *enzyme* is the constraint, not the *contact*, and it predicts that a PPI inhibitor against the OM-LZ/DOT1L contact would do nothing here while a catalytic DOT1L inhibitor should work. **3. The addressing comparison is the real content, and it cuts in the hypothesis's favour.** KMT2A binds DNA. CALM-AF10 uses a protein interface plus an export requirement. SET::NUP214 uses **no address of its own at all** - CRM1 gets there first and the fusion docks on it. **Three lesions reach one output by three categorically different addressing mechanisms**, which is evidence that HOXA-high is an *output* and not a mechanism, and that convergence on HOXA says almost nothing about whether the menin axis is required. **All comparisons here cross lineage and age: KMT2A::AFF1 is infant B-lineage; CALM-AF10 and SET::NUP214 are mostly adult T-lineage.**

Go deeper
What would have to be true for this to threaten the hypothesis

The menin-interface hypothesis is threatened if HOXA activation can be shown to be leukaemogenic with the menin axis demonstrably intact and unused. Neither comparator currently shows that, for different reasons:

  • CALM-AF10: nobody has run the menin experiment, so the separation is inferred from composition, not measured.
  • SET::NUP214: nobody has run it either, and the composition argument is now unavailable because of the recruited MLL1.

So the honest position is that the hypothesis has not been tested against either comparator, and that the graph's confidence in the separation should not have risen on the strength of a second KMT2A-free fusion existing. Existence is not a test.

The hypothesis is supported, weakly, by the addressing diversity: if HOXA-high were sufficient for leukaemia by itself, the address would not matter, and yet the three lesions differ sharply in outcome, lineage, age and drug sensitivity.

The one experiment that resolves both comparators at once

A single plate. LOUCY and MEGAL (SET::NUP214), a CALM-AF10 line, a KMT2A::AFF1 line as the menin-inhibitor positive control, and a HOXA-low T-ALL line. Revumenib and ziftomenib dose-response, HOXA transcript and differentiation readout, plus menin CUT&RUN at HOXA.

Every cell of that table is currently empty in the published literature, and filling it would convert two inferred separations into two measured ones - or overturn both. See exp-menin-in-the-set-nup214-complex-and-menin-inhibitor-in-loucy and the sister CALM-AF10 experiment entry.

Facts and figures
concept classsynthesis / hypothesis assessment
verdict on the beats questionSET::NUP214 does NOT provide a second independent instance of menin-free HOXA activation, because it recruits wild-type MLL1. The separation claim continues to rest on CALM-AF10 alone, and CALM-AF10's version is inferred from composition rather than measured.
hypotheses
[
  "hyp-menin-axis-separable-from-hoxa-programme",
  "hyp-set-nup214-hoxa-runs-through-recruited-wild-type-mll1"
]
experiments
[
  "exp-menin-in-the-set-nup214-complex-and-menin-inhibitor-in-loucy"
]
comparison
{
  "KMT2A::AFF1": {
    "address": "CXXC / post-CXXC, direct unmethylated-CpG DNA binding",
    "menin_contact_in_lesion": true,
    "dot1l_route": "via AFF1 and the super elongation complex",
    "disease": "infant B-lineage ALL",
    "menin_tested": "extensively"
  },
  "CALM-AF10": {
    "address": "AF10 OM-LZ plus a CRM1-dependent export requirement",
    "menin_contact_in_lesion": false,
    "dot1l_route": "AF10 OM-LZ, a defined interface",
    "disease": "T-ALL and some AML, mostly adult",
    "menin_tested": "no published record"
  },
  "SET::NUP214": {
    "address": "borrowed - chromatin-prebound CRM1, docked by NUP214 FG repeats; no DNA-binding domain at all",
    "menin_contact_in_lesion": false,
    "menin_contact_in_recruited_partner": "yes - wild-type MLL1 MBM1/MBM2",
    "dot1l_route": "unmapped surface; AF10 not detected in the complex",
    "disease": "T-ALL and some AML, mostly adult",
    "menin_tested": "no published record"
  }
}
what the beat does add
[
  "DOT1L recruitment is convergent, not interface-shared - which makes the enzyme rather than the contact the constraint",
  "Three lesions reach one output by three categorically different addressing mechanisms, so HOXA-high is an output and not a mechanism",
  "A concrete, cheap experiment that would resolve both comparators on one plate"
]
How this could bite us back
  • This synthesis is built on two abstract-level papers for its most consequential claim (the MLL1 recruitment) and on one 2008 co-IP for its second (convergent DOT1L). It should be revisited as soon as either is read in full or repeated.
  • A reader taking the comparison table as settled would draw exactly the wrong clinical conclusion in either direction - offering a menin inhibitor because MLL1 is recruited, or withholding one because there is no rearrangement. Neither is supported.
  • The convergent-DOT1L claim, if correct, weakens the case for DOT1L PPI inhibitors across the HOXA-high family while strengthening the case for catalytic ones - and the graph currently carries DOT1L PPI inhibition as a modality entry without this distinction.
Still open
  1. Should the graph's confidence in the menin/HOXA separation be revised downward, given that the second supporting instance turned out to have a menin-carrying partner in its complex?
  2. Are there other lesions in this graph where 'no menin-binding motif in the fusion' has been read as menin-independence without checking what the fusion recruits? That audit has not been run.
  3. Is the addressing diversity itself the finding? Three unrelated address mechanisms converging on one cluster in one lineage suggests the HOXA locus is unusually permissive in arrested thymocytes, which is a cell-state claim rather than a lesion claim.
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In the human T-ALL line LOUCY, chromatin-bound CRM1 occupies HOXA and HOXB independently of the fusion - SET-Nup214 knockdown lowers HOX expression without reducing bound CRM1 - so the fusion is recruited to chromatin by CRM1 rather than carrying its own DNA-binding address, and CRM1 inhibition removes both proteins from HOX and lowers HOX expression.

    PMID 31755865DOI
    What this evidence cannot show

    Human leukaemia cell lines only. The recruiter conclusion rests on a knockdown that leaves residual fusion. The same paper finds KPT-330-sensitive CRM1 at HOX in HEL cells, which carry no SET-Nup214, so chromatin-bound CRM1 at HOX is not by itself fusion-specific. Full text read via Europe PMC (PMC6874418).

    Evidence that argues against this (1)
    • The addressing story is not exclusive: a 2021 report has SET-NUP214 binding MLL through the SET acidic region and cooperating with it on the HOXA10 promoter, which would be an address contributed by a recruited partner rather than by CRM1. The two models are compatible and nobody has run the experiment that orders them. Cigdem S et al. 2021
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In LOUCY, SET-NUP214 co-immunoprecipitates with DOT1L but not with AF10; the authors state that no OM-LZ region as present in AF10 was identified in SET-NUP214 and propose either another OM-LZ-like region or an indirect interaction bridged by other proteins - whereas for CALM-AF10 and MLL-AF10 the DOT1L interaction depends on the AF10 octapeptide-motif/leucine-zipper region.

    PMID 18299449DOI
    What this evidence cannot show

    One cell line, one paper, 2008, unrepeated. The AF10 result is a negative co-IP, the weakest form of negative. No interface was mapped in either direction, so 'different surface' is an inference from the absence of an OM-LZ sequence rather than a measured alternative contact. Full text read via the PMC HTML route (PMC2343598).

    Evidence that argues against this (1)
    • PUBLISHED CONFLICTING RESULT, and the founding authors left room for it themselves. They state that no OM-LZ region as present in AF10 was identified in SET-NUP214 and propose either another OM-LZ-like region OR an indirect interaction bridged by other proteins. A bridge has since been named: SET-NUP214 interacts with MLL through the SET acidic region of the fusion and the two cooperatively enhance HOXA10 promoter activity, with neither the SET region nor the NUP214 region alone sufficient. An MLL1-bridged DOT1L contact is not 'a different surface for DOT1L' - it is a different ARCHITECTURE, in which the fusion reaches the H3K79 machinery through the same protein KMT2A fusions are built from. That converges the two arms rather than separating them, which is the opposite of what this entry's three-addresses framing implies. Cigdem S et al. 2021
  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 as of 2026-09-20, against operand controls of 2,237 for menin and 54 for SET-NUP214 and a positive control of 1 through the identical shape; neither LOUCY nor MEGAL is present in the DepMap CRISPR set served by Open Targets; and the founding paper contains no menin blot.

    What this evidence cannot show

    An absence of testing, not a negative result, and bounded by abstracts rather than by papers - a menin blot in a supplementary figure would be invisible to every query above. Conference abstracts, preprints and supplementary cell-line panels were not searched.

    Evidence that argues against this (1)
    • PUBLISHED CONFLICTING RESULT against reading this absence as neutral. The nearest lesion class in which the question HAS been asked answered yes: menin-MLL1 inhibition with VTP50469 is active in NUP98-HOXA9 and NUP98-JARID1A leukaemia at an IC50 the authors report as similar to KMT2A-rearranged and NPM1c cells, evicts both MLL1 and the fusion from chromatin, and prolongs survival in NUP98-NSD1 and NUP98-JARID1A xenografts. A nucleoporin fusion carrying no menin-binding motif of its own was menin-dependent. An untested lesion of the same architectural class therefore does not sit at even odds, and this item must not be quoted as though absence of testing were evidence of independence. Heikamp EB et al. 2022
Connected to