Oncofusion

start / graph / mechanism-set-nup214-dot1l-recruitment-is-convergent-not-af10-shared

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

SET::NUP214 recruits DOT1L by a different surface from CALM-AF10 and KMT2A::MLLT10: the founding paper tested for AF10 in the complex and did not find it

also called SET-NUP214 DOT1L · convergent DOT1L recruitment · no OM-LZ in SET-NUP214

Three different leukaemia fusions all end up using the same enzyme, DOT1L, to switch on the HOXA genes. Two of them grab it with the same molecular handle; SET::NUP214 does not have that handle, so it must be grabbing the enzyme some other way.

The same thing, for a scientist

**CALM-AF10 and KMT2A::MLLT10 both recruit DOT1L through the AF10 octapeptide-motif/leucine-zipper (OM-LZ) region** - a defined protein-protein interface. **SET::NUP214 contains no AF10 and no OM-LZ.** The founding paper anticipated exactly this objection and tested it: in LOUCY, an anti-NUP214 immunoprecipitate blotted for AF10 was **negative**, while an anti-DOT1L immunoprecipitate blotted for SET recovered the fusion [E1]. The authors' own conclusion is unambiguous: > *'this interaction seemed independent of the presence of AF10, since an interaction between SET-NUP214 and AF10 could not be demonstrated in LOUCY ... An OM-LZ region as present in AF10 was not identified in SET-NUP214, and therefore another OM-LZ-like region in SET-NUP214 or an indirect interaction between SET-NUP214 and DOT1L may be facilitated by other SET-NUP214-interacting proteins.'* [E1] **So DOT1L recruitment across the HOXA-high fusions is CONVERGENT, not shared.** Two lesions (CALM-AF10, KMT2A::MLLT10) use one interface; a third reaches the same enzyme by an unmapped and possibly indirect route. That is a different and stronger claim than 'they all use DOT1L', because convergence on an enzyme by independent surfaces says the *enzyme* is the constraint, whereas a shared interface would say the *contact* is. **The mark is measured, and only in this one paper.** H3K79 dimethylation by ChIP is present at HOXA1, HOXA3, HOXA9, HOXA10 and HOXA11 promoters in LOUCY, absent in the control line SKW3, and **partially lost** on SET-NUP214 knockdown; H3 acetylation at the same locus is unchanged in the same time frame [E1]. **This is human paediatric T-ALL cells and the LOUCY line. Not infant B-lineage KMT2A::AFF1.**

Go deeper
How strong is the SET-NUP214/DOT1L interaction result, honestly

Weaker than the headline reads, in four specific ways.

  1. It is one co-IP, in one cell line, in one paper, in 2008. SET-NUP214 AND DOT1L returns exactly 1 PubMed record (2026-09-20) and it is that paper. Eighteen years later nobody has repeated it.
  2. No interface is mapped, no domain is deleted, no point mutant is made. The authors say so themselves: whether the interaction is direct or bridged 'remains to be established'.
  3. The paper's own text is internally inconsistent about the DOT1L IP. The Figure 6C legend describes 'CRM1 and DOT1L immunoprecipitates in LOUCY' and the Methods list a gift rabbit anti-hDOT1L used for IP - yet the Results state that 'reciprocal DOT1L IP experiments to study AF10 involvement could not be performed, as suitable DOT1L antibodies for IP experiments are not available.' Both sentences are in the published full text. The most charitable reading is that the DOT1L IP was performed but could not be blotted for AF10; that reading is an inference, not something the paper states, and it is recorded here as an ambiguity rather than resolved.
  4. The AF10 negative is an absence of a band. A failed co-IP in one line, with one antibody, is the weakest form of a negative - although it is the correct experiment and the authors ran it unprompted, which is more than most papers do.

Net: the convergence claim is the right shape and its evidence is thin. It should be written as convergent on current evidence, with the falsifier stated.

A counted absence that is accurate as a measurement and false as a conclusion - recorded because it nearly went into this entry

Asked whether anyone has measured a histone mark on SET::NUP214, PubMed says no:

query (esearch, 2026-09-20)records
SET-NUP214 AND (H3K79 OR "H3-K79" OR "K79" OR "lysine 79")0
SET-NUP214 AND (H3K79 OR H3K4)0
SET-NUP214 AND DOT1L (positive control, identical shape)1
SET-NUP214 AND histone21

The control returns a one, so rule 67 passes and the zero is a real measurement. And the conclusion it invites is false. PMID 18299449 contains an H3K79 dimethylation ChIP across five HOXA promoters, before and after knockdown. It is invisible to every K79 query because the abstract does not mention the mark - the data are in Figure 6A and the Results text, and PubMed indexes titles, abstracts and MeSH, not figures.

The transferable form: a PubMed counted absence is bounded by abstracts, not by papers. Rules 70-72 bound the name list; this bounds the corpus. Where a measurement is a supporting panel rather than a headline - which is where most chromatin measurements live - the index cannot see it, and the query that returns zero is the same query that returns the paper containing the answer under a different operand.

What a proximity interactome did and did not find

A BioID proximity-labelling study of SET-NUP214 reports that mitochondrial proteins and metabolic regulators are the significantly overrepresented categories in its proximal interactome, with RNA processing and translation-associated proteins shared with NUP98-HOXA9, and p53-programme perturbation in both 2. Neither DOT1L nor MLL1 nor menin appears in its results - both DOT1L and MLL1 are named only in its introduction, citing the prior literature.

That is a weak negative and must be read as one. The experiment used HCT-116 colorectal carcinoma cells transiently transfected with an exogenous BirA-tagged fusion 2. A chromatin methyltransferase that acts in a T-lymphoid context need not be in proximity in a colon line, and BioID has a labelling radius and an expression-level dependence that both cut against recovering a transient chromatin contact.

Facts and figures
mechanism classenzyme recruitment
fusionfusion-set-nup214
enzymegene-dot1l
verdictCONVERGENT, not shared
af10 om lz presentfalse
af10 in complex testedtrue
af10 in complex resultnot detected (anti-NUP214 IP blotted for AF10, LOUCY)
interface mappedfalse
direct or bridgedunresolved; the founding authors explicitly leave both open
h3k79me2 at hoxapresent in LOUCY, absent in SKW3, partially lost on fusion knockdown
h3 acetylation at hoxaunchanged by knockdown within the experiment's time frame
replication count1
dot1l inhibitor testedfalse
falsifierFinding an OM-LZ-like segment in SET or in the retained NUP214 FG region that binds the DOT1L coiled coil would convert this from convergent back to shared.
comparison
{
  "CALM-AF10": "DOT1L via the AF10 OM-LZ, a defined interface",
  "KMT2A::MLLT10": "DOT1L via the same AF10 OM-LZ",
  "SET::NUP214": "DOT1L by an unmapped surface, AF10 absent from the complex"
}
How this could bite us back
  • If DOT1L recruitment is convergent rather than interface-shared, a protein-protein-interaction inhibitor designed against the AF10 OM-LZ/DOT1L contact would be expected to do nothing in SET::NUP214 - the enzyme is shared but the contact is not, and a PPI drug targets the contact.
  • Conversely a catalytic DOT1L inhibitor should work across all three fusions if the enzyme is truly the constraint, which makes this the cheap discriminating experiment and also the one nobody has run.
  • H3K79 loss on knockdown was only PARTIAL and only measured once; designing around a mark whose dependence on the fusion is partial risks a drug that removes the mark without removing the programme.
  • The graph already records that DOT1L inhibition deepens an existing H3K79 deficit in some contexts; a T-ALL with only partial fusion-dependent H3K79 may have the same exposure.
Still open
  1. Does SET-NUP214 bind DOT1L directly? A recombinant pull-down with purified components would settle in one experiment a question that has stood since 2008.
  2. If the interaction is bridged, what is the bridge? The obvious candidate given this beat's other findings is wild-type MLL1, which SET-NUP214 does bind and which sits in complexes with DOT1L-associated machinery - and if MLL1 is the bridge, the DOT1L route and the menin route are the same route.
  3. Is H3K79me2 at HOXA in a SET-NUP214 patient sample, as opposed to LOUCY? The only measurement is in one cell line and has never been repeated.
  4. Would pinometostat do anything to LOUCY? Zero published records, and LOUCY is a standard, distributed line.
Every source behind this entry (4)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In the human T-ALL line LOUCY, an anti-DOT1L immunoprecipitate blotted with anti-SET recovered SET-NUP214, establishing an in vivo interaction; an anti-NUP214 immunoprecipitate blotted with anti-AF10 did NOT, and the authors state the DOT1L interaction appeared independent of AF10 and that no OM-LZ region as present in AF10 was identified in SET-NUP214, proposing either another OM-LZ-like region or an indirect interaction bridged by other proteins; ChIP with an anti-dimethyl-H3K79 antibody showed the mark at HOXA1, HOXA3, HOXA9, HOXA10 and HOXA11 promoters in LOUCY, absent in SKW3, and partially lost after SET-NUP214 knockdown, while H3 acetylation at the same locus was unaltered in the same time frame.

    PMID 18299449DOI
    What this evidence cannot show

    ONE cell line, ONE paper, 2008, never repeated - `SET-NUP214 AND DOT1L` returns exactly 1 PubMed record. No interface mapped, no deletion mutant, no point mutant; the authors state that whether the DOT1L interaction is direct or bridged remains to be established. The AF10 result is a negative co-IP, the weakest form of negative. The published text is internally inconsistent about whether a DOT1L IP was feasible: the Figure 6C legend and the Methods describe one, while the Results state that reciprocal DOT1L IP experiments could not be performed for want of a suitable antibody; this entry records that ambiguity rather than resolving it. The H3K79 loss is 'partial' and explicitly bounded by the limited time frame of the siRNA experiment. Full text read via the PMC HTML route (PMC2343598) after Europe PMC fullTextXML returned HTTP 500.

    Evidence that argues against this (1)
    • The same authors' proposed alternative - 'another OM-LZ-like region in SET-NUP214' - would make the recruitment SHARED rather than convergent after all. Nothing in the eighteen years since has tested for such a region, so the convergent reading is the better-supported one but is not established. Van Vlierberghe P et al. 2008
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    BioID proximity labelling of SET-NUP214 and NUP98-HOXA9 expressed in HCT-116 cells found that both fusions associate with regulators of RNA processing and translation and perturb the p53 transcriptional programme, that NUP98-HOXA9 is additionally linked to Wnt, MAPK, ER signalling and the cytoskeleton, and that mitochondrial proteins and metabolic regulators are significantly overrepresented in the SET-NUP214 proximal interactome; DOT1L and MLL1 appear in the paper only in its introduction citing prior work, not among its own results.

    PMID 32664447DOI
    What this evidence cannot show

    The system is HCT-116 COLORECTAL CARCINOMA cells transiently transfected with an exogenous BirA(R118G)-tagged fusion cloned from LOUCY RNA - not a leukaemia cell, not a lymphoid cell, not the endogenous fusion. The absence of DOT1L from the results is therefore a weak negative: a T-lymphoid chromatin contact need not occur in a colon line, and BioID's labelling radius and expression dependence both work against recovering a transient chromatin interaction. The tagged fusion did localise correctly (intranuclear foci plus nuclear rim), which is the relevant positive control for the construct but not for the chromatin context. Full text read via Europe PMC (PMC7407662).

    Evidence that argues against this (1)
    • PUBLISHED CONFLICTING RESULT that establishes this assay's false-negative rate for exactly this class of partner. The BioID did not recover DOT1L; it also did not recover MLL1. MLL1 is a demonstrated SET-NUP214 partner in two other studies - a direct interaction mapped to the SET acidic region of the fusion with cooperative activation of the HOXA10 promoter, and co-condensation of MLL1 with the fusion and CRM1 on chromatin. So the study's negatives include a partner that is positive elsewhere, in the same protein class, by two independent assays. A proximity-labelling negative from a colorectal line therefore cannot be used to argue that the DOT1L contact is made by a different surface, or that it is absent. Cigdem S et al. 2021
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    PubMed esearch on 2026-09-20 returns 0 records for `SET-NUP214 AND (H3K79 OR "H3-K79" OR "K79" OR "lysine 79")` and 0 for `SET-NUP214 AND (H3K79 OR H3K4)`, against 1 for `SET-NUP214 AND DOT1L` and 21 for `SET-NUP214 AND histone` through the identical query shape - while PMID 18299449, which the DOT1L query returns, contains an H3K79 dimethylation ChIP in its Figure 6A.

    What this evidence cannot show

    Bounded absence with a positive control through the identical shape (rule 67 satisfied) and a spelling variant list (rule 70/72 satisfied: H3K79, H3-K79, K79, 'lysine 79'). The bound that is NOT satisfied by any of those rules, and is the point of this item, is the corpus bound: PubMed indexes titles, abstracts and MeSH, so a measurement reported only in a figure is invisible. Not searched: Google Scholar full text, Europe PMC full-text search, or the supplementary material of the 54 SET-NUP214 records.

    Evidence that argues against this (1)
    • PUBLISHED RESULT THAT REMOVES THE FRAMEWORK THIS ABSENCE IS INTERPRETED IN. The graph reads an H3K79 silence for SET::NUP214 against a partition of H3K79 methylation into a menin-dependent, MLL-fusion-directed oncogenic DOT1L complex and a separate endogenous DOT1L complex directed by MLLT10. SET::NUP214 contains NEITHER KMT2A NOR MLLT10, so neither arm of that partition describes it, and the published framework makes no prediction for this lesion at all. That matters more than the corpus bound: even if the measurement were recovered from a figure, there is currently no model in which to place it, and calling this fusion's DOT1L recruitment 'convergent' asserts a third route that the partition neither contains nor excludes. Gilan O et al. 2023
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    No published study has tested a DOT1L inhibitor against SET-NUP214: `(SET-NUP214 OR TAF-I-NUP214) AND (pinometostat OR EPZ-5676 OR EPZ004777 OR "DOT1L inhibitor" OR "DOT1L inhibition")` returns 0 records in PubMed as of 2026-09-20.

    What this evidence cannot show

    Bounded absence. Fusion name list: SET-NUP214, SET::NUP214 (PubMed normalises these to the same 54 records), TAF_I-NUP214, TAF-I-NUP214. Drug name list: pinometostat, EPZ-5676, EPZ004777, 'DOT1L inhibitor', 'DOT1L inhibition' - singular, plural and nominalisation per rule 72. Positive control through the identical shape: `SET-NUP214 AND DOT1L` = 1. NOT searched: SGC0946, EPZ-004777 with the alternative hyphenation, unnamed tool compounds, or the LOUCY line as the query subject rather than the fusion - though `LOUCY AND DOT1L` also returns only PMID 18299449.

    Evidence that argues against this (1)
    • PUBLISHED CONFLICTING RESULT about what the absence means. An untested drug class is only a gap if the drug would have been expected to work, and the clinical record argues against that expectation even where the H3K79 rationale is strongest: pinometostat in adult advanced acute leukaemia, enriched for MLL rearrangement, produced complete remission in 2 of 51 patients, both with t(11;19), and the authors characterise the activity as modest. Nobody testing a DOT1L inhibitor in a fusion with weaker and less direct DOT1L evidence, in a disease with two cell lines and no mouse model, is failing to do an obvious experiment. The honest reading of this zero is triage, not oversight. Stein EM et al. 2018
Connected to