Oncofusion

start / graph / risk-set-nup214-model-space-is-two-cell-lines

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

The entire experimental model space for SET::NUP214 is two cell lines, neither of them in DepMap, with no mouse model found - so every mechanism in this beat is n = 1 or n = 2

also called the two-cell-line problem · LOUCY and MEGAL · SET-NUP214 model scarcity

Almost everything known about this fusion comes from two laboratory cell lines, and one of them supplies nearly all of it. That is a very thin base for any conclusion, and it is also why several obvious experiments have never been done.

The same thing, for a scientist

**Of 141 human leukaemia and lymphoma cell lines screened, exactly two express SET-NUP214: LOUCY (T-ALL) and MEGAL (AML)** [E1]. Every mechanistic result this beat records comes from one of those two, or from exogenous expression of the cloned fusion in HeLa or HCT-116. **The scarcity compounds in four ways:** 1. **LOUCY carries the phenotype and MEGAL barely does.** MEGAL expresses less fusion protein, shows weaker CRM1/SET-Nup214 binding, and responds less to CRM1 inhibition [E2]. So the second line is not an independent replicate; it is a weaker version of the first. 2. **The founding knockdown worked in LOUCY and not in other T-ALL lines** [E3]. The authors report this as specificity, and it is - but it also means there is no line in which the result has been reproduced. 3. **Neither line is in DepMap.** A MEN1 query against the Open Targets DepMap service returns 1,258 screens, 8 of them T-lymphoblastic, and LOUCY (ACH-000104) and MEGAL are absent from both [E4]. So no CRISPR dependency datum exists for this fusion at all, and none can be looked up. 4. **No mouse model and no PDX was found** in the literature searched for this beat, and a single-centre series concludes the rearrangement in isolation is insufficient to induce leukaemia [E5] - which, if true, means a single-transgene model would not work even if someone built one. **The consequence for how this graph should be read:** the SET::NUP214 entries are a well-cited description of what two cell lines do. Any sentence in them that reads like a statement about the disease is over-reaching, and any negative - including this beat's menin negative - is a negative in two lines.

Go deeper
What this blocks, specifically
  • Any lineage-split dependency claim. The project's standing rule is that a DepMap number is not a result until it is lineage-split. Here there is no number to split.
  • Any statement that a drug does or does not work in this genotype. Two lines, one of them weak, is not a panel.
  • Any in vivo claim whatsoever, including about the differentiation block, the gamma-delta restriction, or CRM1 inhibition.
  • Reproducibility of the DOT1L co-IP, which has stood unrepeated since 2008 partly because there is nowhere else to repeat it.
What would fix it, and what it would cost

Cheapest: add LOUCY and MEGAL to the next dependency screen. Two lines is a marginal addition to any existing panel and would close a gap that currently has no datum in it at all.

Moderate: obtain primary material. The GRAALL cohort has eleven characterised patients and the Italian series has seven; a biobank request is the only route to a non-cell-line experiment that exists today.

Expensive and possibly futile: build a model. If the rearrangement in isolation is insufficient 5, a model needs a cooperating lesion, and the best-evidenced candidate is PHF6 loss, which is statistically associated with SET-NUP214 in patients [see cell-state-set-nup214-gamma-delta-arrest-and-steroid-resistance]. That is a testable design and nobody has published it.

Facts and figures
kindessential-function
risk subjectthe evidence base itself, not a drug
severityhigh for any claim generalised to the disease; low for claims stated as cell-line results
mitigationState the cell line in every claim (brief rule 73). Add LOUCY and MEGAL to the next dependency screen - two lines is a marginal cost against a gap with no datum in it. For anything beyond that, primary material from the GRAALL or Italian cohorts is the only existing route.
candidate cooperating lesion for a modelPHF6 loss, on the statistical association in patients
affects
[
  "fusion-set-nup214",
  "mechanism-set-nup214-crm1-is-the-address-not-the-fusion",
  "mechanism-set-nup214-dot1l-recruitment-is-convergent-not-af10-shared",
  "mechanism-set-nup214-recruits-wild-type-mll1",
  "ana-set-nup214-menin-bounded-absence"
]
model space
{
  "cell_lines": [
    "LOUCY (T-ALL)",
    "MEGAL (AML, weaker phenotype)"
  ],
  "lines_screened_to_find_them": 141,
  "in_depmap": false,
  "mouse_model_found": false,
  "pdx_found": false,
  "heterologous_systems_used": [
    "HeLa",
    "HCT-116"
  ]
}
How this could bite us back
  • Any negative result generated in these two lines - including a negative menin experiment - is a negative in two lines and must not be reported as a property of the disease.
  • Because LOUCY supplies nearly all the data, a single artefact in that line would propagate through the entire mechanistic literature for this fusion without any independent system to catch it.
  • The scarcity is self-reinforcing: with no model, nobody runs the experiment; with no experiment, nobody builds the model.
Still open
  1. Does any SET::NUP214 PDX exist in a repository that does not publish? PDX collections are catalogued by diagnosis, not by cryptic fusion, so one could exist and be invisible.
  2. Would a SET-NUP214 plus PHF6-loss mouse recapitulate the gamma-delta arrest? It is the best-supported model design available and nobody has published it.
  3. Has LOUCY drifted? It has been in culture since the 1980s and supplies essentially every functional result in this literature; nothing in this beat establishes that the fusion or the HOXA phenotype is stable across sources.
Every source behind this entry (5)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Of 141 human leukaemia and lymphoma cell lines tested, only the T-ALL line LOUCY and the AML line MEGAL expressed the SET(TAF-Ibeta)-NUP214 fusion gene transcript, and both express a 140 kDa fusion protein.

    PMID 19166587DOI
    What this evidence cannot show

    A 2009 panel of one repository's holdings. A line added since, or held elsewhere, would not appear, and the survey used RT-PCR for a specific junction, so an alternative junction could have been missed - the same paper notes the transcript requires alternative splicing, which is exactly the situation in which a junction-specific assay can fail.

    Evidence that argues against this (1)
    • Alternative SET-NUP214 fusion transcripts have been molecularly characterised in a patient, so a junction-specific RT-PCR screen is not guaranteed to detect every fusion-positive line - the 2 of 141 figure is a lower bound on the available model space rather than a complete census. Kim J et al. 2010
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    MEGAL shows lower expression of SET-Nup214 than LOUCY, weaker CRM1 and SET-Nup214 binding at HOX clusters confined mainly to HOXA9, and a milder HOX-expression response to KPT-330 than LOUCY.

    PMID 31755865DOI
    What this evidence cannot show

    The authors state the reason for MEGAL's lower KPT-330 sensitivity remains unknown, so attributing it to expression level is their inference from a ratio comparison across two supplementary figures, not a measured dose-response. Full text read via Europe PMC (PMC6874418).

    Evidence that argues against this (1)
    • PUBLISHED CONFLICTING RESULT on the scope of this risk. The model space for NUP214-FUSION biology is not two cell lines, even though the space for SET::NUP214 specifically is: FKH-1 carries DEK::NUP214, supports CRISPR deletion of XPO1 with a strong dependency readout, and has a patient-derived xenograft that responds to eltanexor with molecular clearance in marrow. That is a third line, a genetic perturbation and an in vivo arm. It does not contain SET, so it cannot test anything SET-specific - but every claim in this beat that is really about NUP214, CRM1 or the FG repeats has a third model available and this entry says it does not. Charles Cano F et al. 2025
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    siRNA knockdown of SET-NUP214 abolished HOXA expression, inhibited proliferation and induced differentiation in LOUCY but not in other T-ALL lines.

    PMID 18299449DOI
    What this evidence cannot show

    The 'other T-ALL lines' are fusion-negative, so this is a specificity control and is the correct design - it is listed here as a limit on replication, not as a criticism. Full text read via the PMC HTML route (PMC2343598).

    Evidence that argues against this (1)
    • PUBLISHED CONFLICTING RESULT about what the one positive line establishes. The specificity control is correctly designed, and the consequence is that the fusion-dependence claim has exactly one positive observation. The only available replicate is not silent, it is discordant: MEGAL expresses less SET-Nup214, shows weaker CRM1 and fusion binding at the HOX clusters confined mainly to HOXA9, and responds more mildly to KPT-330 - and the authors state the reason is unknown. So the second fusion-positive line differs on the pathway that is supposed to be the mechanism, and nobody has repeated the knockdown in it. Oka M et al. 2019
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    Neither LOUCY (DepMap ACH-000104) nor MEGAL appears in the DepMap CRISPR gene-effect dataset served by the Open Targets Platform: a MEN1 query returns 1,258 screens with a gene effect, of which 8 are annotated T-Lymphoblastic Leukemia/Lymphoma, and neither line is present by DepMap id or by cell-line name.

    What this evidence cannot show

    Fetched 2026-09-20. The DepMap release version is NOT stated in the API response, so the release is unverified, and a line present in a different or newer release would not be seen here. The DepMap id for LOUCY was resolved from Cellosaurus CVCL_1380 rather than assumed, and the absence was checked both by id and by name across all 1,258 screens.

    Evidence that argues against this (1)
    • PUBLISHED CONFLICTING RESULT on the register, not on the fetch. 'Not in DepMap' understates the genetic-dependency data that exist for this fusion family, because targeted CRISPR work outside DepMap has been published: deletion of XPO1 in DEK::NUP214-positive FKH-1 revealed a strong dependency, with chromatin co-localisation and target-gene readouts. A dependency question about a NUP214 fusion can therefore be asked genetically today; it just cannot be asked from a pooled screen that never included the relevant lines. The gap is one of screening coverage, and the correct response is a targeted experiment rather than a note that the resource is silent. Charles Cano F et al. 2025
  5. E5weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    A single-centre series reports that SET-NUP214 rearrangement in isolation is insufficient to induce leukemia.

    PMID 25956045DOI
    What this evidence cannot show

    TITLE ONLY, and flagged as such throughout this beat. PubMed carries no abstract, Europe PMC returns no PMCID so routes 2-4 are structurally unavailable, and the DOI resolves to a paywall. Recorded as a known-but-unreadable item: the identifier and what the title asserts, nothing more. It is used here only to note that the one published statement on this question points away from a single-lesion model.

    Evidence that argues against this (1)
    • PUBLISHED CONFLICTING OBSERVATION, in the only register a title-only claim can be opposed in. The case literature contains the pattern this title argues against: an identical SET::NUP214 with PHF6 and EZH2 mutations tracking one clone across two simultaneous presentations, acute undifferentiated leukaemia in the marrow and T-lymphoblastic lymphoma in the mediastinum. Neither report settles sufficiency - both are single centres, and the co-mutations in the case report are themselves candidate cooperating lesions, which is closer to supporting insufficiency than refuting it. The objection is narrower and firmer: a title nobody in this project has been able to read should not be quoted as the field's position on whether one lesion is enough. Wei Q et al. 2025
Connected to