EPS15 / AF1p — domain architecture, and the one coordinate nobody has published

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

• Wild-type EPS15 is required for clathrin-mediated endocytosis. Anything targeting its coiled coil risks broad endocytic toxicity in every tissue, which is a poor profile for a rare fusion.
• A coiled-coil homodimerisation interface is a notoriously hard small-molecule target, and the So et al. result that synthetic dimerisers substitute implies the requirement is for dimerisation per se rather than for a specific surface — meaning a blocker would have to out-compete a high-avidity interaction with no unique pharmacophore.
• Recording a 227-residue bracket as if it were a domain in a figure would be worse than leaving the lane empty. This entry's own data could be misused that way if the `evidence_basis` and `boundary_precision` fields are dropped downstream.

1. Where exactly is the EPS15 coiled coil? Reading the So et al. 2003 full text for their construct boundaries, or running a coiled-coil predictor over residues 315-541 and labelling the output as prediction, would replace a 227-residue bracket with something usable — this is the cheapest unfinished task in this beat.
2. Is InterPro's unendorsed UIM at 442-462 the same amphipathic helix that a coiled-coil predictor would call? If so, the oligomerisation domain has a candidate location and the two annotations stop conflicting.
3. Synthetic dimerisation modules substitute for the EPS15 coiled coil in transformation assays. Does the converse hold — does a dimerisation-blocking agent revert established KMT2A::EPS15 disease, or is oligomerisation only required for initiation?
4. The EH domains are retained in the fusion, and nobody has asked what three intact NPF-binding endocytic modules do inside the nucleus while tethered to CpG islands by KMT2A. So et al. showed they are not required for transformation; that is not the same as showing they are inert. Do they sequester NPF-motif partners (synaptojanin, epsin, stonin) away from the membrane?
5. What is the actual EPS15 residue at the KMT2A::EPS15 junction? 'Almost all' (Bernard 1994) is the best the literature offers. A single RNA-seq junction read from any t(1;11) case would fix it and would also settle whether EH1 survives intact.

1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   UniProt P42566 (EPS15_HUMAN) is 896 aa with EH domains at 15-104, 128-216 and 224-314, EF-hands at 48-83, 160-195, 223-258 and 262-292, a DAB2-interaction region at 2-330, a disordered region at 542-606, fifteen approximate D-P-F repeats at 599-827, an SH3-binding motif at 768-774, and UIMs at 851-870 and 877-896 — and it annotates no coiled-coil feature anywhere in the sequence.

   UniProtKB entry P42566 (EPS15_HUMAN), epidermal growth factor receptor substrate 15 · UniProt Consortium · UniProtKB 2026
   ▸What this evidence cannot show

   Retrieved 2026-09-18 with the coiled-coil feature type explicitly requested, so the absence is a genuine absence in the entry rather than an artefact of the query. Absence of annotation is not absence of structure: UniProt's coiled-coil calls are conservative and a real coiled coil can go uncalled.

   ▸Evidence that argues against this (1)

   • The fourth EF-hand has no support outside UniProt, and the feature list omits a second disordered region. Of UniProt's four EF-hand calls (48-83, 160-195, 223-258, 262-292) the fetched InterPro record reproduces only three - IPR002048 gives 48-83, 160-195 and 223-258, PROSITE PS50222 the same three, SMART SM00054 gives 52-80, 164-192 and 227-255 - and no member database annotates anything at 262-292, so one of the four bars in this lane is single-sourced. The same UniProt record also carries TWO Disordered regions, 542-606 and 760-848; only the first is listed here, and the omitted one contains the SH3-binding motif at 768-774 and runs to within three residues of UIM 1. UniProt Consortium 2026
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   InterPro maps EH domains (IPR000261) at 8-104, 121-217 and 217-314, EF-hand domains, and a ubiquitin-interacting-motif conserved site (IPR003903) at 442-462, 851-870 and 877-896 to P42566; it maps no coiled-coil entry.

   InterPro entry-to-protein matches for UniProt P42566 · InterPro / EMBL-EBI · InterPro 2026
   ▸What this evidence cannot show

   InterPro's member databases do not systematically include coiled-coil predictors, so the absence of a coiled-coil match here is weaker evidence than the UniProt absence. The third UIM call at 442-462 is a conserved-site match and UniProt does not endorse it. Retrieved 2026-09-18. CORRECTION 2026-09-19: 'InterPro maps no coiled-coil entry' is wrong as a statement about the fold. InterPro assigns SUPERFAMILY SSF90257 ('Myosin rod fragments', the canonical two-stranded coiled coil) at 335-447 and CATH-Gene3D G3DSA:1.10.287.1490 at 329-523 - the region the leukaemia literature calls the EPS15 coiled coil. What is true is that no entry is LABELLED 'coiled coil'.

   ▸Evidence that argues against this (1)

   • InterPro maps two coiled-coil-class folds, under names that do not contain the words coiled coil. The fetched record for P42566 includes SUPERFAMILY SSF90257 'Myosin rod fragments' at residues 335-447 and CATH-Gene3D G3DSA:1.10.287.1490 at 329-523 - the myosin rod is the canonical long two-stranded coiled coil, and these are exactly the residues the KMT2A-fusion literature calls the EPS15 coiled coil. The correct statement is not that InterPro finds no coiled coil but that it assigns a coiled-coil-class fold across 113-219 residues that UniProt leaves entirely unannotated, which makes 'no coiled-coil feature' the outlier rather than the consensus. InterPro Consortium; EMBL-EBI 2026
3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   The coiled-coil oligomerisation domains of AF1p/Eps15 and GAS7 are necessary and sufficient for leukemogenic transformation induced by the respective MLL fusion proteins, and homodimerisation of MLL by synthetic dimerisation modules mimics bona fide MLL fusions, activating Hox genes and enhancing haematopoietic progenitor self-renewal.

   Dimerization contributes to oncogenic activation of MLL chimeras in acute leukemias · So CW, Lin M, Ayton PM, Chen EH, Cleary ML · Cancer Cell 4:99-110 2003
   PMID 12957285DOI
   ▸What this evidence cannot show

   The published abstract does not give residue boundaries for the Eps15 coiled coil, and we did not obtain the full text, so we cannot cite a residue range from this paper. Retroviral overexpression in mouse progenitors; the synthetic-dimeriser result shows dimerisation is sufficient in that assay, not that it is the only thing the real fusion does. KMT2A::EPS15 is a rare partner, so the clinical weight is light.

   ▸Evidence that argues against this (1)

   • The dimerisation model does not extend to the dominant KMT2A partners. AFF1, MLLT3 and MLLT1 act by recruiting the super elongation complex and DOT1L, not by oligomerising KMT2A, so oligomerisation explains a minority class rather than the disease. Meyer C et al. 2023
4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   The second EH domain of EPS15 was solved by NMR (PDB 1EH2, chain A residues 121-219) and the structure defines an Asn-Pro-Phe binding pocket.

   PDB entry 1EH2 — Structure of the second Eps15 homology domain of human Eps15; primary citation: Structure and Asn-Pro-Phe binding pocket of the Eps15 homology domain · RCSB Protein Data Bank · RCSB Protein Data Bank / Science 1998
   ▸What this evidence cannot show

   Title, method and chain range confirmed against the RCSB data API on 2026-09-18. We did not independently verify a PMID for the 1998 Science primary citation, so we cite the structure rather than the paper. This domain is lost in the fusion and is included only for completeness of the lane.

   ▸Evidence that argues against this (1)

   • The chain stops four residues short of 219 and the pocket was defined with no NPF peptide in the entry. SIFTS maps 1EH2 chain A to P42566 121-219 at PDB residue numbers 6-104, but PDBe polymer coverage gives coordinates only for residue numbers 6-100 across the 20 deposited models, i.e. UniProt 121-215 - so 216-219 have no coordinates and the ordered extent falls one residue short of UniProt's own EH-2 end at 216. RCSB reports polymer_entity_count = 1 with a single bound calcium ion and no peptide, so the Asn-Pro-Phe binding pocket is a surface inferred from the apo domain, not a complex that was observed. PDBe / SIFTS et al. 2026
5. E5established 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 - including the AT-hook, snRNP and methyltransferase similarities - fused to almost all of the AF-1p product, a predicted 98 kDa acidic protein highly similar to 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 9(4):1039-45 1994
   PMID 8134107
   ▸What this evidence cannot show

   ABSTRACT ONLY - the 1994 Oncogene full text was not obtainable and this journal/year is not in PMC. 'Almost all' is the authors' own wording and is not a residue-level junction coordinate, so EH1 (residues 15-104) could in principle be clipped at its N-terminal edge. Two patients only, both myeloid. The indexed abstract also contains an internal inconsistency, writing 't(1;11)(p32;q11)' in the body where the title says (p32;q23). This citation settles retention versus loss of the EH region; it does not give a breakpoint residue. **Regraded strong -> established on 2026-09-20** under `concept-the-confidence-grading-rule`: the residue-level architecture of the der(11) product is what the paper reports from its own cloning and sequencing.

   ▸Evidence that argues against this (1)

   • A variant fusion lacking the EH domains does exist, so EH retention is canonical rather than universal. Shinohara A et al. 2010
6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   An MLL-AF1p/Eps15 fusion variant lacking the EH domains was reported in therapy-related acute lymphoblastic leukemia and described as novel, implying that the canonical MLL-AF1p fusion retains them.

   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 34(2):e62-3 2010
   PMID 19665789DOI
   ▸What this evidence cannot show

   TITLE AND PUBMED METADATA ONLY - this is a one-page Letter/Case Report with no abstract indexed in PubMed and no full text obtained. n=1. The inference 'novel implies the canonical form differs' rests on the authors' framing in the title, not on a sentence they are quoted as writing. Used only as corroboration of E5, never on its own.

   ▸Evidence that argues against this (1)

   • The EPS15 breakpoint splits by lineage, so an EH-domain-lacking fusion is the B-ALL norm rather than a one-off. A 75-year-old woman with t(1;11)(p32;q23) B-ALL had KMT2A exon 8 fused in frame to EPS15 exon 12, and the report makes the pattern explicit: considering three reported B-ALL cases, EPS15 breakpoints were markedly different between AML (exon 2) and B-ALL (exons 10-12). The t(1;11) fusion was originally defined in AML as KMT2A exon 8 to EPS15 exon 2, which retains all three EH domains; the known lymphoid cases cluster instead at EPS15 exons 10-12, downstream of the EH-domain region. 'Novel, EH-domain-lacking' therefore describes a recurrent lymphoid-lineage breakpoint class, and the canonical-retains-them inference holds only for the AML breakpoint. Yamamoto K et al. 2021

• related to →KMT2A::AFDN (MLL-AF6), t(6;11)(q27;q23)

  Counter-evidence added cycle 41: the EPS15 breakpoint splits by lineage - AML breaks at EPS15 exon 2 (retaining all three EH domains) and the known B-ALL cases at exons 10-12 (losing them), so an 'EH-domain-lacking' fusion is a recurrent lymphoid class rather than a one-off (Yamamoto 2021, PMID 33647817). InterPro also DOES assign a coiled-coil-class fold to EPS15 335-447/329-523, contrary to this entry's earlier 'no coiled coil anywhere'.

• fuses with →KMT2A / MLL1 — the gene that breaks

  EPS15/AF1p is the 3' partner in KMT2A::EPS15, t(1;11)(p32;q23)

• supports →Hypothesis — forced-oligomerisation interfaces are the mutation-resistant drug targets the fusion field has been ignoring

  the EPS15 coiled coil is necessary and sufficient for MLL-AF1p transformation, and synthetic dimerisers substitute

• analogous to →AFDN / AF6 — domain architecture and residue coordinates

  two cytoplasmic partners that both transform by self-association rather than by elongation hijack

• ← related toWhich partner, in which patient: KMT2A partner frequencies across six age-lineage strata

  EPS15 is 2.2% of infant ALL, the commonest infant fusion whose mechanism is oligomerisation rather than elongation hijack

• ← analogous toKMT2A::AFF1 oligomerises through the retained AFF1 CHD, putting the commonest fusion in the oligomerisation class

  same proposed mechanism, better-shaped interface: a folded eight-helix domain rather than a coiled coil

• ← related toNon-elongation KMT2A partners considered and not written: what the searches actually returned, partner by partner

  EPS15 flagged as a gap: 55 patients, no fusion entry

• ← part ofKMT2A::EPS15 (MLL-AF1p), t(1;11)(p32;q23) — the infant-ALL cytoplasmic partner

  almost all of the EPS15 product in myeloid cases; EH domains lost at the exon 10-12 B-ALL junctions

• ← fuses withKMT2A / MLL1 — the gene that breaks

  t(1;11)(p32;q23); 55 patients, rank 8, 22 of them infant ALL

• ← refutesEPS15 (AF1p): the EH domains may or may not be discarded, and either way nobody has shown it matters

  that node assigns fate 'lost' to all three EH domains; the founding clinical report describes the fusion as containing almost all of AF1p and EH loss is reported elsewhere as a novel variant

• ← refutesEPS15 (AF1p): the EH domains may or may not be discarded, and either way nobody has shown it matters

  Founding report (Bernard 1994, PMID 8134107) says the der(11) product joins MLL to 'almost all' of AF-1p — EH1-3 fate corrected lost → retained, dispute closed

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

  Same EH-domain-fate question, settled the same way from the same primary source