Oncofusion

start / graph / mechanism-set-nup214-crm1-is-the-address-not-the-fusion

Mechanismstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q58-set-nup214

SET::NUP214 has no address of its own: chromatin-bound CRM1 is already at HOX, and the fusion's FG repeats dock onto it

also called CRM1 recruits SET-Nup214 · chromatin-bound CRM1 · XPO1 as a chromatin platform

You would expect a leukaemia fusion to find the HOXA genes itself and drag its helpers there. For SET::NUP214 it is the other way round: the export carrier CRM1 is already sitting on those genes, and the fusion sticks to it.

The same thing, for a scientist

In the human T-ALL line **LOUCY**, ChIP-seq shows **CRM1 selectively bound at the HOXA and, more strongly, the HOXB cluster**, with no such accumulation in K562 or HL60 [E1]. The decisive experiment is the order test: **knocking down SET-Nup214 down-regulates HOX genes but does not significantly reduce chromatin-bound CRM1**, so CRM1 is upstream and functions as the recruiter of the fusion, not its cargo [E1]. The pharmacology agrees. **KPT-330 (selinexor)** at 1,000 nM for 24 h removes both the CRM1 and the Nup214 ChIP peaks almost completely and reduces HOX expression, with LOUCY more sensitive than K562 [E1]. Leptomycin B, which binds CRM1 irreversibly and does not degrade it, does the same and additionally degrades SET-Nup214 [E1]. In MEGAL, the AML line carrying the same fusion, CRM1 and SET-Nup214 preferentially occupy the **HOXA9** region and the KPT-330 effect is present but weaker, tracking the line's lower fusion expression [E1]. **Why this matters to the menin-interface hypothesis.** The hypothesis holds that a KMT2A fusion is leukaemogenic because it combines the menin contact with the **CXXC/post-CXXC address**. SET::NUP214 has neither. What it has instead is a **borrowed address**: an exportin that was already on the locus. That is a genuinely different class of targeting from sequence-specific DNA binding, and it means the comparison between SET::NUP214 and KMT2A::AFF1 is not 'same job, different protein' but 'same output, categorically different addressing'. **This is human adult/paediatric T-ALL and AML cell lines. It is not infant B-lineage KMT2A::AFF1 and it is not a patient.**

Go deeper
The order-of-recruitment experiment, stated exactly

'Reduction of SET-Nup214 caused down-regulation of HOX genes ... but did not significantly affect the amount of CRM1 that bound to HOX regions, suggesting that CRM1 functions as a recruiter of SET-Nup214.' 1

This is a one-directional result and should be read as such. It shows CRM1 does not need the fusion to be at HOX. It does not show that the fusion cannot also stabilise CRM1 there, because a knockdown leaves residual protein, and it does not address what puts CRM1 on HOX in the first place.

The companion result from the same group is that CRM1 does the same job for NUP98-HOXA9 in mouse ES cells, across all four Hox clusters, and for NPM1c via its NES 1. So the platform is general and the fusion-specific part is only the docking chemistry: nucleoporin-CRM1 for SET-Nup214, NES-CRM1 for NPM1c.

HOXB, which this graph has not been looking at

The ChIP-seq finds CRM1 binding especially strong at HOXB, visible at genome-wide scale on chromosome 17, alongside HOXA on chromosome 7 1. This graph's entire HOX layer is HOXA. A second independent report finds SET-NUP214 driving HOXC overexpression in acute megakaryoblastic leukaemia through a hypermethylation landscape 2.

So across three papers the fusion is associated with HOXA, HOXB and HOXC. Any claim that SET::NUP214 'activates HOXA' is a claim about the cluster that was measured, and the founding ChIP-qPCR measured only HOXA promoters because that is what it was looking for.

The CRM1 convergence with CALM-AF10, held at the strength it deserves

Q55 separated three convergence claims by strength and called CRM1 the strongest. This beat's reading supports that ordering and sharpens it:

  1. CRM1 (strongest). For CALM-AF10, a CRM1-AF10 chimera causes leukaemia and mutating the CRM1 residues that contact NUP214 abolishes it. For SET::NUP214, the fusion literally contains the NUP214 surface CRM1 docks against, and CRM1 is shown by ChIP-seq to be the upstream recruiter. The two lesions touch the same physical contact from opposite sides, and both have loss-of-function evidence on that contact.
  2. DOT1L (middling, and now demoted to convergent). See mechanism-set-nup214-dot1l-recruitment-is-convergent-not-af10-shared.
  3. HOXA (weakest). And now weaker still, because the cluster identity is not stable across papers.
Facts and figures
mechanism classchromatin targeting by borrowed platform
fusionfusion-set-nup214
address moduleNUP214 FG repeats docking onto chromatin-bound CRM1/XPO1 - NOT a sequence-specific DNA-binding domain
contrast with kmt2aKMT2A fusions retain the CXXC and post-CXXC region, which binds unmethylated CpG DNA directly. SET::NUP214 has no DNA-binding domain at all.
order of recruitmentCRM1 first: SET-Nup214 knockdown does not reduce chromatin-bound CRM1
clusters reported
[
  "HOXA (LOUCY, MEGAL)",
  "HOXB (LOUCY, strongest CRM1 signal)",
  "HOXC (acute megakaryoblastic leukaemia, title-only source)"
]
pharmacology
{
  "KPT-330 (selinexor)": "1,000 nM, 24 h - CRM1 and Nup214 ChIP peaks almost completely lost; HOX expression falls; LOUCY more sensitive than K562",
  "leptomycin B": "depletes HOX-bound CRM1 and SET-Nup214 and additionally degrades the fusion protein"
}
systems
[
  "LOUCY (human T-ALL)",
  "MEGAL (human AML)",
  "K562, HL60, HEL (comparators)"
]
How this could bite us back
  • Selinexor works here by inhibiting a common-essential export receptor, so the demonstrated anti-leukaemic effect in fusion-positive lines is confounded by general CRM1 dependence unless a lineage-matched fusion-negative control is run.
  • Nuclear bodies reassemble shortly after the CRM1 antagonist clears, so intermittent dosing may give intermittent mechanism.
  • If CRM1 is the address and the fusion is the cargo, then resistance can come from anything that restores HOX expression downstream of the platform - including the wild-type MLL1 route this beat also documents.
  • CRM1 inhibition removes every NES cargo's export simultaneously; a HOX transcriptional readout after 24 h at 1 microM cannot be attributed to the chromatin-local effect without a separation-of-function reagent.
Still open
  1. What puts CRM1 on the HOX clusters in the first place? The recruiter has a recruiter, and nothing in this literature identifies it. If it is a lineage-restricted factor, that would explain why SET-NUP214 knockdown does nothing in other T-ALL lines.
  2. Does chromatin-bound CRM1 sit at HOXA in KMT2A-rearranged cells? Nobody has looked. If it does, the two mechanisms share a platform and the menin axis and the CRM1 axis are two handles on one locus.
  3. Is the HOXB signal, which is the strongest CRM1 peak in LOUCY, functionally required? Every functional readout in this literature is HOXA.
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, ChIP-seq with two independent anti-CRM1 antibodies shows CRM1 selectively bound at HOXA and especially HOXB cluster regions and absent from K562 and HL60; SET-Nup214 knockdown down-regulates HOX genes without significantly reducing chromatin-bound CRM1, indicating CRM1 is the recruiter; KPT-330 at 1,000 nM for 24 h removes both CRM1 and Nup214 peaks almost completely and lowers HOX expression, with LOUCY more sensitive than K562; leptomycin B likewise depletes HOX-bound CRM1 and SET-Nup214 and degrades the fusion; and in the SET-Nup214-positive AML line MEGAL both proteins preferentially occupy HOXA9 with a milder KPT-330 effect tracking lower fusion expression.

    PMID 31755865DOI
    What this evidence cannot show

    Human leukaemia CELL LINES only - LOUCY (T-ALL), MEGAL (AML), with K562, HL60 and HEL as comparators. No patient sample, no xenograft, no normal thymocyte. The recruiter conclusion rests on a knockdown, which leaves residual fusion protein, so it excludes an obligate requirement and not a contribution. KPT-330 at 1,000 nM for 24 h is a high, non-clinical exposure and CRM1 inhibition has pleiotropic effects on every NES cargo in the cell, so the HOX change is not necessarily a chromatin-local effect. The article is a Research Advance building on the same group's 2016 NUP98-HOXA9 paper, so the platform model and this test share authors and reagents. Full text read via Europe PMC (PMC6874418).

    Evidence that argues against this (1)
    • The same paper reports KPT-330-sensitive CRM1 binding signals at HOX regions in HEL cells, which do NOT carry SET-Nup214, and states that the biological significance of this remains to be determined. Chromatin-bound CRM1 at HOX is therefore not by itself a fusion-specific event. Oka M et al. 2019
  2. E2weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    A SET-NUP214-induced hypermethylation landscape promotes abnormal overexpression of HOXC cluster genes in acute megakaryoblastic leukaemia.

    PMID 39634129DOI
    What this evidence cannot show

    TITLE ONLY. PubMed carries no abstract text for this record. It is open access in PMC (PMC11615591) and was not read in full within this beat's scope - it is flagged in the memo as the highest-value unread item, because a DNA-methylation mechanism and a HOXC output are both outside every model this graph currently holds for SET::NUP214. Carried at `weak` and nothing has been written about what the paper probably says beyond its title.

    Evidence that argues against this (1)
    • PUBLISHED CONFLICTING RESULT, and it is in the data this entry's own model is built on rather than in the title-only item. The CRM1-address model was measured against HOX clusters in LOUCY, and that measurement reports CRM1 selectively bound at the HOXA and ESPECIALLY THE HOXB cluster regions. A title asserting a HOXC output in acute megakaryoblastic leukaemia is therefore not the first cluster this model fails to name - the published occupancy already puts the strongest CRM1 signal at a cluster this entry does not discuss. 'Chromatin-bound CRM1 is already at HOXA' is an under-description of the occupancy data, and any statement that the address explains the HOXA output has to say why HOXB carries more CRM1 and less attention. Oka M et al. 2019
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Targeted CRM1 inhibition disperses SET-NUP214 and DEK-NUP214 nuclear bodies and re-localises both fusions throughout the nucleoplasm; cell viability, metabolism and proliferation of leukaemia cell lines harbouring these fusions are compromised by CRM1 inhibition, an effect sustained after clearance of the antagonist - but the nuclear bodies themselves reestablish shortly after the inhibitor is removed.

    PMID 32934780DOI
    What this evidence cannot show

    Abstract-level for this beat. Cell lines only, no in vivo work, and viability/metabolism/proliferation readouts in lines that carry a common-essential target's inhibitor cannot distinguish fusion-directed killing from general CRM1 dependence without a fusion-negative control matched for lineage. The paper reports the body-reassembly and the sustained-effect results together, which are in tension and are recorded here as stated.

    Evidence that argues against this (1)
    • METHODOLOGICAL DEFEATER plus the experiment that repairs it, published in the sibling fusion. Dispersal-plus-viability in NUP214-fusion lines cannot separate fusion-directed killing from general CRM1 dependence, because XPO1 is common-essential and this study has no lineage-matched fusion-negative comparator. The design that does separate them exists: CRISPR DELETION of XPO1 in DEK::NUP214-positive FKH-1, with co-localisation of XPO1 and the fusion at chromatin measured directly and target-gene downregulation as the readout, carried through to a patient-derived xenograft. That is a genetic dependency with a molecular endpoint rather than a pharmacological dispersal with a viability endpoint, and this entry's CRM1-as-address claim rests on the weaker of the two. Charles Cano F et al. 2025
Connected to