ELL2 — domain architecture and residue coordinates

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.

• Blocking super-elongation-complex assembly hits a machine that normal cells need for essentially all Pol II pause release. The therapeutic index for an ELL2-AFF interface inhibitor is not obviously better than for a CDK9 inhibitor, and CDK9 inhibitors have been toxicity-limited.
• ELL and ELL2 are paralogues with the same fold, so an inhibitor of one is likely to hit the other, and redundancy between them is a ready-made resistance route: losing the targeted paralogue's contribution may simply shift the complex to the other.
• The concave-surface interface is large. Small molecules rarely displace a 50-residue element that buries most of a concave face, and the reported cavity may be too shallow to anchor real affinity.

1. The cavity Qi and colleagues observed at the imperfectly packed AFF4-ELL2 interface has been framed only as an HIV latency-reversal opportunity. Is it a tractable site in KMT2A-rearranged leukemia, where blocking super-elongation-complex assembly would be partner-agnostic in the same way menin inhibition is? A Boltz or docking pass against PDB 5JW9 is a cheap first look.
2. Is the AFF1-ELL2 interface the same as the solved AFF4-ELL2 interface? The ELLBow is conserved between AFF1 and AFF4 but the AFF1 complex has not been solved, and KMT2A::AFF1 biology runs through AFF1.
3. Do ELL and ELL2 contribute interchangeably to the KMT2A::AFF1 super elongation complex, or is there a paralogue preference? If one dominates, the 19-residue coordinate difference stops being a bookkeeping nuisance and starts being a targeting decision.
4. Does the seven-residue N-terminal extension in the 5JW9 construct (519-525) contribute to AFF4 binding? It was included in the crystallised construct and sits outside every database OCEL boundary. If it does, the docking surface for the super-elongation complex is slightly larger than any annotation shows.

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

   UniProt O00472 (ELL2_HUMAN) is 640 aa with an OCEL domain at 526-636 and disordered regions at 172-202, 290-320 and 343-490; the paralogue ELL (P55199) is 621 aa with its OCEL domain at 507-617.

   UniProtKB entries O00472 (ELL2_HUMAN) and P55199 (ELL_HUMAN) · UniProt Consortium · UniProtKB 2026
   ▸What this evidence cannot show

   Retrieved 2026-09-18. Both entries were fetched and their gene names and lengths checked rather than recalled. The disordered regions are MobiDB-lite predictions, not measurements.

   ▸Evidence that argues against this (2)

   • The node calls ELL2's coordinates 'among the most reliable in this beat because the functionally critical domain has a 2.0 A crystal structure with its partner bound'. The structure does not settle the boundary. The crystallised construct is 519-640, wider than UniProt's 526-636 at both ends, and the profile databases spread the N-terminal boundary from 526 to 536. The reliability claim holds for the fold's existence and for the AFF4 interface; it does not hold for the bar's endpoints. InterPro Consortium; EMBL-EBI 2026
   • Independent check of the identity trap this node flags: UniProt's live record for O00472 gives the primary gene name ELL2 and a length of 640 aa, and annotates the OCEL domain at 526-636 and nothing else - no ELL-N region and no winged-helix feature. The ELL-N (11-291) and WH (201-291) bars therefore come entirely from InterPro/Pfam models (PF10390 and IPR019464 at 11-291; SSF46785 at 201-291), not from UniProt, which the node does not currently say. InterPro Consortium; EMBL-EBI 2026
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   The 2.0 Å crystal structure of the human ELL2 C-terminal domain bound to a 50-residue AFF4 element (the ELLBow) shows an occludin-like arch-shaped fold whose concave surface is largely occupied by the AFF4 element; the interface is imperfectly packed, leaving a cavity the authors suggest as a potential small-molecule site.

   Structural basis for ELL2 and AFF4 activation of HIV-1 proviral transcription · Qi S, Li Z, Schulze-Gahmen U, Stjepanovic G, Zhou Q, Hurley JH · Nat Commun 8:14076 2017
   PMID 28134250DOI
   ▸What this evidence cannot show

   The AFF-family partner in this structure is AFF4, not AFF1, and the functional assays are HIV-1 Tat-mediated proviral transcription, not KMT2A-rearranged leukemia. AFF1 and AFF4 are paralogues and the ELLBow element is conserved, but the AFF1-ELL2 interface has not itself been solved. The suggested cavity is an observation about crystal packing quality, not a validated druggable pocket.

   ▸Evidence that argues against this (1)

   • A later cryo-EM structure shows the ELL2 surface that actually does the work on polymerase is a different one. ELL2-EAF1 bound to an RNA polymerase II elongation complex at 2.8 A shows the ELL2-EAF1 dimerisation module contacting the Pol II lobe - the same lobe surface used by the initiation factor TFIIF - with stimulation of elongation requiring that module plus the ELL2 linker tethering it to the Pol II protrusion, i.e. the super elongation complex stimulates elongation allosterically through contacts entirely separate from the concave occludin surface AFF4 occupies. The cited paper's own cavity measurement also bounds the druggability suggestion hard: the interfacial cavity is 36 cubic angstroms with a narrow mouth, smaller than a single benzene ring at about 80. BOUNDED ABSENCE: PubMed `AFF4 ELL2 interface small molecule inhibitor super elongation complex latency reversing agent screen` returns 0 records; no published ligand or fragment screen against that cavity exists. Chen Y et al. 2021
3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   InterPro maps the Pfam ELL N-terminal model (PF10390 / IPR019464) to O00472 residues 11-291, a winged-helix DNA-binding-superfamily module to 201-291, and the occludin homology domain (IPR010844 / PF07303) to 526-636 and 532-633 respectively.

   InterPro entry-to-protein matches for UniProt O00472 · InterPro / EMBL-EBI · InterPro 2026
   ▸What this evidence cannot show

   HMM hit boundaries. PF10390 at 11-291 is a family-level model that spans a region containing at least one predicted disordered segment (172-202), so it should not be drawn as a single folded domain. Retrieved 2026-09-18.

   ▸Evidence that argues against this (1)

   • The winged helix at 201-291 is one of two mutually inconsistent calls in the same fetched record, and the other is filed under a misleading name. InterPro returns the winged-helix module at 201-291 under SUPERFAMILY SSF46785 and IPR036390, but at 194-292 under CATH-Gene3D G3DSA:1.10.10.2670 and InterPro IPR042065 - seven residues wider at the N-terminus and one at the C-terminus - and the CATH-Gene3D entry carries the name 'E3 ubiquitin-protein ligase' (its InterPro parent is 'E3 ubiquitin-protein ligase ELL-like'), a transferred label with nothing to do with ELL2's function. Reporting only 201-291 picks the narrower of two and drops the fact that the wider one is annotated as an unrelated activity. InterPro Consortium; EMBL-EBI 2026
4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   PDB 2E5N is the solution NMR structure of the ELL2 ELL_N2 domain (chain A, residues 200-292); PDB 5JW9 is the crystal structure of the ELL2 occludin domain with an AFF4 peptide (ELL2 chain B, 519-640); PDB 7OKX and 7OKY are composite cryo-EM structures of the active Pol II-DSIF-ELL2-EAF1 elongation complex containing an ELL2 CONSTRUCT annotated 1-640 in SIFTS (chain M), of which only 139 residues are actually modelled.

   PDB entries 2E5N, 5JW9, 7OKX and 7OKY, with UniProt O00472 chain mappings · RCSB Protein Data Bank; UniProt Consortium · RCSB Protein Data Bank 2021
   ▸What this evidence cannot show

   Titles and methods were confirmed against the RCSB data API on 2026-09-18. 7OKX is explicitly a composite structure, so its 1-640 chain range reflects the modelling assembly rather than continuous experimental density; large parts of ELL2 are predicted disordered and are unlikely to be resolved. Construct boundaries in 2E5N and 5JW9 are expression choices, not fold boundaries. CORRECTION 2026-09-19: this item previously said 7OKX contains full-length ELL2. PDBe polymer coverage for chain M gives observed author residues 7-19, 25-73, 75-104 and 110-156 only - 139 of 640, none in the OCEL domain. 1-640 is the deposited construct.

   ▸Evidence that argues against this (1)

   • Chain M has coordinates for 139 residues, not 640, and none of them are in the OCEL domain. PDBe polymer coverage for 7OKX chain M gives observed author residues 7-19, 25-73, 75-104 and 110-156 - four fragments, 21.7% of the sequence, all N-terminal, with nothing from 157 to 640 and therefore nothing from the OCEL domain at 526-636. SIFTS's 1-640 is the deposited construct, not the model. The entry is also not a human complex: of its 18 polymer entities only three (O00472 ELL2, Q96JC9 EAF1, O00267 SPT5H) are human, with the Pol II subunits mapping to Sus scrofa proteins, at 3.3 A. PDBe / SIFTS et al. 2026
5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

   ELL, not ELL2, is the recurrent KMT2A translocation partner catalogued in the Frankfurt recombinome series.

   The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, Burmeister T, Gröger D, Sutton R, et al. · Leukemia 37:988-1005 2023
   PMID 37019990DOI
   ▸What this evidence cannot show

   We use this citation to establish that ELL is the catalogued recurrent partner. We did not exhaustively verify that ELL2 never appears anywhere in the 107 in-frame fusions listed, so the correct statement is 'ELL is the recurrent partner', not 'ELL2 is never a partner'.

   ▸Evidence that argues against this (1)

   • ELL2, not ELL, is the ELL-family subunit of the super elongation complex that KMT2A fusion proteins hijack, so its absence from the translocation catalogue makes it the wild-type effector the fusions recruit rather than a bystander. The complex is defined as containing P-TEFb, ELL2, AFF4/AFF1, ENL and AF9, recruited by HIV-1 Tat and by MLL proteins respectively, with ENL and AF9 sitting in separate ELL2-containing complexes whose YEATS domains target the complex to Pol II via the PAF complex. BOUNDED ABSENCE, searched 2026-09-19: `"MLL-ELL2" OR "KMT2A-ELL2" OR "KMT2A::ELL2" OR "MLL::ELL2"` returns 0 records; `(KMT2A[tiab] OR MLL[tiab]) AND ELL2[tiab]` returns 6 (42748909, 38493200, 21873227, 21832080, 11418481, 11238904), none reporting a KMT2A::ELL2 fusion; `ELL2[tiab] AND (fusion OR translocation OR leukemia OR leukaemia)[tiab]` returns 18, none reporting ELL2 as a fusion partner. WHAT WOULD FALSIFY IT: a molecularly confirmed KMT2A::ELL2 transcript or genomic breakpoint. TRAP: PMID 38493200 prints 'KMT2A::ELL2' verbatim in its abstract, but its Results describe the patients as carrying KMT2A::ELL (NM_006532.3, ELL on 19p13.11, not ELL2 on 5q15) - the abstract is mislabelled and a naive search returns it as a false positive. He N et al. 2011
6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   The OCEL domain's boundaries are disputed by up to ten residues at the N-terminus and four at the C-terminus: UniProt and PROSITE PS51980 give 526-636, SUPERFAMILY SSF144292 528-635, Pfam PF07303 532-633 and CATH-Gene3D G3DSA:6.10.140.340 536-636; the winged-helix module is 201-291 in SUPERFAMILY and InterPro IPR036390 but 194-292 in CATH-Gene3D and InterPro IPR042065.

   InterPro member-database domain matches for UniProt O00472 (Pfam, PROSITE, CATH-Gene3D, SUPERFAMILY, PANTHER) · InterPro Consortium; EMBL-EBI · InterPro 2026
   ▸What this evidence cannot show

   Profile-model extents, not measurements. Retrieved 2026-09-19 (database record; no full text read). None of these disagreements changes any fate assignment in this node, because ELL2 is present in the super-elongation complex as a whole protein rather than being cut by a breakpoint.

   ▸Evidence that argues against this (1)

   • The experimental structure puts the ordered fold outside all four of those windows at one end, and the spread is wider than ten residues once it is included. PDBe polymer coverage for 5JW9 chain B - the only crystal structure of this domain - gives observed author residues 525-639, so order begins one residue before the earliest profile call (526) and runs three residues past the latest C-terminal call (636), ending only at 639 of a 640-residue protein. Against CATH-Gene3D's 536-636 that is an 11-residue N-terminal discrepancy and a 3-residue C-terminal one; against Pfam's 532-633 it is 7 and 6. The profile spread is a disagreement among models none of which matches the density. PDBe / SIFTS et al. 2026
7. E7established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   The construct in the OCEL-AFF4 crystal structure 5JW9 is wider than the stored OCEL bar at both ends: SIFTS maps its ELL2 chain B to O00472 residues 519-640, i.e. seven residues N-terminal of UniProt's 526 and running to the protein's C-terminus rather than stopping at 636; the bound AFF4 chain A maps to Q9UHB7 301-351.

   SIFTS UniProt residue-level mappings for PDB entries 5JW9, 2E5N (construct ranges read back from the PDBe mappings API) · PDBe / SIFTS; EMBL-EBI · PDBe SIFTS 2026
   ▸What this evidence cannot show

   SIFTS reports the deposited sequence range rather than the observed model, so 519-640 is the construct the authors crystallised and not a claim that all of it is ordered. Read back 2026-09-19 from the PDBe mappings API (database record; no full text read).

   ▸Evidence that argues against this (1)

   • Six of those seven residues were never seen. PDBe polymer coverage for the same chain gives coordinates only for author residues 525-639, so the crystallised construct does extend to 519 but residues 519-524 have no density - the ordered fold begins at 525, one residue before UniProt's 526, not seven. The C-terminal half of the claim survives and is the stronger part: order runs to 639, three residues past UniProt's 636 and six past Pfam's 633. Reporting the construct range as if it were evidence about the domain edge inflates the N-terminal discrepancy sevenfold. PDBe / SIFTS et al. 2026

• related to →Step 3 — The hijack: the fusion drags the super elongation complex to paused RNA polymerase II

  Counter-evidence on the ELL2 druggable-cavity suggestion: the surface that does the work on polymerase is a different one. ELL2-EAF1 bound to a Pol II elongation complex at 2.8 A contacts the Pol II lobe via the dimerisation module plus the ELL2 linker, separate from the concave occludin surface AFF4 occupies (Chen 2021, PMID 34265249); and the interfacial cavity is 36 cubic angstroms, smaller than a benzene ring.

• related to →ELL — rank 7 on selectivity and the deepest common-essential in the top ten, with one NMR fragment to its name

  Bounded absence recorded cycle 41: no KMT2A::ELL2 fusion exists in PubMed (three queries returning 0, 6 and 18 records, none reporting one). TRAP worth knowing: PMID 38493200 prints 'KMT2A::ELL2' in its abstract but its Results describe KMT2A::ELL (NM_006532.3, ELL on 19p13.11) - the abstract is mislabelled and returns as a false positive.

• related to →The little elongation complex (ELL + ICE1 + ICE2 + ZC3H8) — the other place ELL works, and why ELL's pan-essentiality is probably not about the fusion

  Qualification: the 2.0 A OCEL-AFF4 structure settles the fold and the interface but not the bar's endpoints. The crystallised construct (5JW9) is O00472 519-640, wider than UniProt's 526-636 at both ends, and the databases spread the N-terminal boundary from 526 to 536. Only OCEL is a UniProt feature at all; ELL-N and the winged-helix module come from Pfam/InterPro models.

• related to →ELL2 — a SEC subunit that almost no cell needs, next to a paralogue almost every cell needs

  residue-level domain coordinates for ELL2 (UniProt O00472, 640 aa)

• binds →AFF4 — the textbook SEC scaffold that KMT2A::AFF1 cells need LESS than other leukemias

  ELL2 OCEL domain (526-636) binds the AFF4 ELLBow; 2.0 A crystal structure PDB 5JW9

• part of →Step 3 — The hijack: the fusion drags the super elongation complex to paused RNA polymerase II

  ELL2 is a core super-elongation-complex subunit in the assembly the fusion recruits

• related to →The ELL/EAF activation module (ELL2, EAF1, EAF2) — a general elongation stimulator, not a KMT2A-fusion dependency

  paralogue caution: ELL's OCEL domain is 507-617, ELL2's is 526-636; coordinates are not interchangeable

• related to →KMT2A::ELL (MLL-ELL), t(11;19)(q23;p13.1)

  the recurrent KMT2A partner is ELL, not ELL2 — flagged because the two are routinely conflated

• ← bindsAFF1 / AF4 — domain architecture and residue coordinates

  AFF-family ELLBow element docks on the ELL2 OCEL concave face; solved for AFF4, inferred for AFF1