Oncofusion

start / graph / subtraction-eps15-eh-domains

Mechanismmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q11b-subtraction

EPS15 (AF1p): the EH domains may or may not be discarded, and either way nobody has shown it matters

also called MLL-EPS15 EH domains · MLL-AF1p · t(1;11)(p32;q23)

EPS15 normally helps cells pull receptors in from their surface, using three gripper modules at one end. Reports disagree about whether the leukemia fusion keeps those grippers, and no study has TESTED whether losing them matters - though one 2000 paper did propose that this endocytic axis is relevant to MLL-fusion leukaemia, and the transforming part identified so far is a different, retained segment that makes the fusion clump together.

The same thing, for a scientist

EPS15 (UniProt P42566, 896 aa) carries three EH domains at residues 15-104, 128-216 and 224-314 that bind Asn-Pro-Phe (NPF) motifs on endocytic adaptors. The transforming element of MLL-EPS15/MLL-AF1p is not their absence but the retained coiled-coil oligomerisation domain, which So et al. showed is necessary and sufficient for leukemogenic transformation. Crucially, EH-domain fate is not settled: the founding clinical report states the der(11) product fuses MLL to 'almost all' of AF1p, implying retention, while a 2009 case report describes an EH-domain-lacking variant as novel. Corrected 2026-09-19: this entry previously said no study had ever suggested that EH-domain loss might matter; So et al. (Leukemia 2000, PMID 10764144) showed the EH domains bind the NPF motifs of synaptojanin, which is strongly expressed in bone marrow and immature leukaemic lines, and proposed a link between MLL-fusion leukaemogenesis and inositol signalling. Nobody has tested EH-domain LOSS, which is the accurate statement. Verdict: addition for the established mechanism, and not determinable for any subtraction contribution — with the added caveat that the premise of loss is itself unverified.

Go deeper
What the EH domains do in the intact protein

EH (Eps15 homology) domains are signalling modules that recognise proteins containing Asn-Pro-Phe (NPF) sequences, binding the NPF peptide in a hydrophobic pocket between two alpha helices 1. In EPS15 the three EH domains connect the protein to the NPF-motif-bearing endocytic adaptors (epsin and relatives) that build clathrin coats, which is how EPS15 participates in clathrin-mediated endocytosis and receptor internalisation 1.

That is a well-characterised function, and it is entirely plausible as a thing worth losing. Plausibility is not evidence - but the haematopoietic NPF ligand has been identified (synaptojanin), so the loss is at least a testable proposition rather than an inert one.

Are they genuinely discarded? The sources disagree

This is the finding that matters most in this entry, because the rest of the argument depends on it and the graph currently asserts one side of a live disagreement.

  • Bernard et al. 1994, the founding molecular description of t(1;11)(p32;q23), states that the der(11) chromosome expresses the 1,368 N-terminal amino acids of HRX/MLL fused to 'almost all the AF-1p product' 2. Taken at face value, a fusion containing almost all of a 896-residue protein contains the EH domains at 15-314.
  • Shinohara et al. 2009 report a therapy-related ALL case explicitly titled as a novel MLL-AF1p/Eps15 fusion variant 'lacking the EH-domains' 3. It is described as novel precisely because it differs from the expected architecture.

Read together: EH-domain retention is breakpoint-dependent, and loss is a variant rather than the rule. This contradicts protein-eps15-domains, which assigns "fate": "lost" to all three EH domains without qualification. That node is another Queen's and is not edited here, but the discrepancy is recorded so it can be resolved rather than propagated.

What actually transforms: a retained oligomerisation domain

So et al. examined MLL fusions with proteins that normally reside in the cytoplasm (AF1p/EPS15 and GAS7) and found that 'the coiled-coil oligomerization domains of these proteins are necessary and sufficient for leukemogenic transformation induced by the respective MLL fusion proteins' 4. Dimerisation, not endocytosis and not the absence of endocytosis, is the mechanism.

Corrected 2026-09-19: this previously read 'No paper found in a dedicated search proposes or tests EH-domain loss as a contributing mechanism.' No paper TESTS EH-domain loss, but So et al. (Leukemia 2000, PMID 10764144) proposed that the EH-domain/NPF axis is relevant here, mapping EPS15 EH-domain binding to the NPF motifs of synaptojanin - a protein strongly expressed in bone marrow and immature leukaemic cell lines and also bound by two further MLL fusion partners, EEN and Abi-1. Given that a minimal coiled-coil fragment reproduces transformation on its own, the EH domains are demonstrably not required for the fusion to work — but 'not required for the gain' is a different statement from 'their loss contributes', and neither the add-back nor any separation-of-function experiment has been reported.

A coordinate caveat inherited from protein-eps15-domains and confirmed here: UniProt annotates no coiled coil anywhere in P42566, so the very element that does the transforming has no database coordinates. So et al. defined it by their own construct boundaries and prediction.

Facts and figures
partnerEPS15 (AF1p)
partner uniprotP42566
fusionKMT2A::EPS15, t(1;11)(p32;q23) — no fusion node exists in this graph
candidate discarded regionEH1 15-104, EH2 128-216, EH3 224-314
intact function of candidate discarded regionBind Asn-Pro-Phe motifs on endocytic adaptors; clathrin-mediated endocytosis and receptor internalisation
genuinely discarded at observed breakpointsno — canonically RETAINED (settled 2026-09-18)
retention disputeSETTLED 2026-09-18 by the verification beat, in favour of retention. The founding report (Bernard et al., Oncogene 1994;9(4):1039-45, PMID 8134107, abstract read) states the der(11) chromosome expresses the 1,368 N-terminal amino acids of HRX fused to 'almost all the AF-1p product', AF-1p being a 98 kDa protein highly similar to murine eps15 — so the EH region is in the canonical fusion. Shinohara et al. (PMID 19665789) report an EH-lacking variant as NOVEL, which corroborates retention as the norm rather than undermining it. Residual uncertainty: 'almost all' is not a residue-level junction, so EH1 (starting at residue 15) could in principle be clipped; and the 1994 full text was not obtainable.
retained oncogenic modulecoiled-coil oligomerisation domain, necessary and sufficient for transformation, with no UniProt or InterPro coordinates
loss shown to contribute to transformationfalse
verdictaddition
verdict rationaleThe only demonstrated mechanism is gained oligomerisation. A subtraction contribution is not merely unproven, its premise (that the EH domains are lost) is itself disputed in the primary literature.
experiment to settleTwo-step. First, settle the sequence question: type the EPS15 breakpoint in the reported t(1;11) cases and determine what fraction retain EH1-EH3. Only if loss is common is the second step worth doing — restore the EH domains in cis to MLL-EPS15 and score myeloid progenitor immortalisation, with an NPF-binding-dead EH mutant as the control that separates 'restoring a fold' from 'restoring endocytic engagement'.
graph corrections raised
[
  "protein-eps15-domains assigns fate 'lost' to all three EH domains without qualification. The founding clinical report (Bernard 1994) says the fusion contains almost all of AF1p, and the EH-lacking architecture is reported as a novel variant. That node's fate assignments should be downgraded to breakpoint-dependent."
]
graph corrections resolved
[
  "RESOLVED 2026-09-18. protein-eps15-domains has been corrected: EH1, EH2 and EH3 now carry fate 'retained' rather than 'lost', with Bernard 1994 (PMID 8134107) and Shinohara 2010 (PMID 19665789) added as evidence E5/E6, and its status moved from 'disputed' back to 'verified'. The correction raised by this node was upheld."
]
How this could bite us back
  • Dimerisation interfaces are notoriously hard to drug, and the EPS15 one has no structure, no coordinates and no chemical matter. Classifying this partner as 'addition' identifies the mechanism without identifying anything actionable.
  • If EH-domain retention varies between patients, any partner-matched therapeutic reasoning about EPS15 fusions is being applied to a heterogeneous group that nobody has stratified.
Still open
  1. What fraction of published MLL-EPS15 cases retain the EH domains? Until that is known, every statement in this graph about EPS15 domain fate is a coin flip dressed as a fact.
  2. Does the coiled-coil oligomerisation domain of EPS15 have residue coordinates anywhere in the primary literature? So et al. must have used defined constructs; reading their methods would replace this graph's 227-residue exclusion bracket with a real number.
  3. Do EH-retaining and EH-lacking MLL-EPS15 fusions differ in lineage? The one EH-lacking case reported was lymphoblastic and therapy-related, while the founding EH-retaining cases were myeloid — an n of 3, but a suggestive one.
  4. EPS15 loss-of-function affects EGFR internalisation. Is there any receptor-trafficking phenotype in MLL-EPS15 blasts that would give the subtraction hypothesis a cheap first test?
Every source behind this entry (4)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    EH domains are eukaryotic signalling modules that recognise proteins containing Asn-Pro-Phe (NPF) sequences, with the NPF peptide bound in a hydrophobic pocket between two alpha helices.

    Structure and Asn-Pro-Phe binding pocket of the Eps15 homology domain · de Beer T, Carter RE, Lobel-Rice KE, Sorkin A, Overduin M · Science 1998
    PMID 9721102DOI
    What this evidence cannot show

    NMR structure of a single EH domain; establishes the binding mechanism, not the physiological consequence of losing all three in a leukemia context. Identifier and abstract verified via PubMed on 2026-09-18.

    Evidence that argues against this (1)
    • The entry previously claimed total silence on whether EH-domain loss could matter; that claim was too strong and has been corrected in the body (2026-09-19). The exception is haematopoietic. So et al. showed that the EH domains of Eps15 bind the NPF motifs of synaptojanin, that synaptojanin is strongly expressed in bone marrow and in immature leukaemic cell lines but only weakly in peripheral blood leukocytes and absent in a mature B line, and that two other MLL fusion partners with SH3 domains (EEN and Abi-1) bind the same protein through its proline-rich regions - proposing on that basis a link between MLL-fusion leukaemogenesis and inositol signalling. That is a named, bone-marrow-expressed EH ligand and an explicit proposal that the endocytic axis matters in this disease, which makes EH-domain fate a testable question rather than an inert structural fact. A PubMed search this session over the title/abstract phrases EH domain and EH domains combined with leukemia or leukaemia returns 2 records, both title-checked: this one and the 2009 EH-lacking case report at E3. A companion search over EPS15 or AF1p together with MLL or KMT2A returns 15 records. So CW et al. 2000
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    In t(1;11)(p32;q23), the der(11) chromosome expresses the 1,368 N-terminal amino acids of HRX/MLL fused to almost all of the AF-1p product, AF-1p being the human homologue of murine Eps15.

    A novel gene, AF-1p, fused to HRX in t(1;11)(p32;q23), is not related to AF-4, AF-9 nor ENL · Bernard OA, Mauchauffe M, Mecucci C, Van den Berghe H, Berger R · Oncogene 1994
    PMID 8134107
    What this evidence cannot show

    Two myeloid leukemia cases from 1994; 'almost all' is the authors' wording and no residue boundary is given, so this supports retention only qualitatively. Identifier and abstract verified via PubMed on 2026-09-18. **Regraded moderate -> established on 2026-09-20** under `concept-the-confidence-grading-rule`: the same reported architecture as the sibling entries, which grade it established.

    Evidence that argues against this (1)
    • Directly contradicted for at least one patient by Shinohara et al. 2009 (E3), who report an MLL-AF1p/Eps15 variant lacking the EH domains and describe it as novel. The two reports are best reconciled as breakpoint heterogeneity rather than as one being wrong. Shinohara A et al. 2009
  3. E3weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    A therapy-related acute lymphoblastic leukemia case carried an MLL-AF1p/Eps15 fusion variant lacking the EH domains, reported as novel.

    A novel MLL-AF1p/Eps15 fusion variant in therapy-related acute lymphoblastic leukemia, lacking the EH-domains · Shinohara A, Ichikawa M, Ueda K, Takahashi T, Hangaishi A, Kurokawa M · Leukemia Research 2009
    PMID 19665789
    What this evidence cannot show

    A single-case letter. The abstract is not indexed in PubMed, so only the title was retrievable; the claim rests on the title alone, and confidence is set to weak accordingly. A full-text read is required before this is used for anything load-bearing. Identifier verified via PubMed on 2026-09-18.

    Evidence that argues against this (1)
    • Two conflicting case descriptions are the wrong instrument for a question that a breakpoint resource can answer. Meyer et al. sequenced MLL genomic breakpoints in 2,345 patients, characterised 94 partner genes at the molecular level and catalogued 135 distinct rearrangements, with the MLL-side breakpoint shifting by age and lineage. Whether the EH domains are inside or outside MLL-EPS15 fusions is a distribution, not a yes/no, and it is recoverable from series of this kind rather than from a 1994 'almost all' and a 2009 single-case letter whose abstract is not even indexed. Until that is done, this entry's 'not determinable' is correct but understates how cheaply it could be made determinable. Meyer C et al. 2018
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    For MLL fusions with normally cytoplasmic proteins including AF1p/Eps15 and GAS7, the coiled-coil oligomerisation domains of those partners are necessary and sufficient for leukemogenic transformation induced by the respective MLL fusion proteins.

    Dimerization contributes to oncogenic activation of MLL chimeras in acute leukemias · So CW, Lin M, Ayton PM, Chen EH, Cleary ML · Cancer Cell 2003
    PMID 12957285DOI
    What this evidence cannot show

    Murine progenitor transformation assays. UniProt annotates no coiled coil in P42566, so the transforming element has no database coordinates and its boundaries come from the authors' own constructs and prediction. Identifier and abstract verified via PubMed on 2026-09-18.

    Evidence that argues against this (1)
    • Dimerisation is not the only published mechanism available to a cytoplasmic-partner fusion, and the competing one comes from the other canonical member of that class. Yokoyama et al. showed that MLL-AF6 does NOT form the MLL/AEP hybrid complexes that AF4- and ENL-family fusions form, yet still constitutively recruits AEP (AF4-family, ENL-family and P-TEFb) to target chromatin by an alternative mechanism, and framed bypass of the normal AEP-recruitment route as the strategy most frequently used by MLL oncoproteins. If a cytoplasmic partner can deliver the elongation machinery without oligomerising into it, then 'the coiled coil is necessary and sufficient' is a statement about what the immortalisation assay measured, not about what the fusion does in a patient - and the EH domains remain untested under either model. Yokoyama A et al. 2010
Connected to