The ELLbow: AFF4 301-351 wrapped over the ELL2 occludin domain at 2.0 A, Kd 86 nM - and no structure of ELL itself, the partner in t(11;19)
also called ELLbow · AFF4 ELLBow · ELL2 occludin domain · 5JW9
The scaffold of the elongation complex folds a fifty-residue stretch into an elbow shape that lies across a curved surface on ELL2. The protein fused to KMT2A in t(11;19) leukaemia is ELL, not ELL2 - and ELL has never been solved with any scaffold at all.
The same thing, for a scientist
PDB 5JW9 (2.00 A) is the ELL2 C-terminal occludin-fold domain bound to the AFF4 ELLbow. The deposited AFF4 entity aligns to Q9UHB7 residues ~301-352 and the ELL2 entity to O00472 ~519-640 against the fetched UniProt sequences. Fluorescence polarisation gives the wild-type competitor peptide Kd = 86 nM; AFF4 hydrophobic residues V316, I319, L320, M323, W327, L331, I334 and P342 line the interface, with I319D/L320D costing over 25-fold and the I319D/L320D/M323D triple mutant losing roughly 50-fold. On the ELL2 side, the single mutants H559E, H608E, N619A and K625T are each tolerated while the quadruple mutant abolishes binding in both pull-down and FP. The gap this entry exists to name: RCSB search on 2026-09-20 returns zero entries containing ELL (P55199) together with AFF4 or with ELL2, and only one entry containing P55199 at all. KMT2A::ELL is a recurrent t(11;19) fusion and its partner's scaffold interface has no structure.
Residues, both sides
AFF4 (Q9UHB7): ELLbow = 301-351; buried hydrophobics V316, I319, L320, M323, W327, L331, I334, P342 1.
ELL2 (O00472): occludin-fold domain, deposited span ~519-640; the polar quartet H559, H608, N619, K625 is individually dispensable and collectively required 1.
ELL (P55199): nothing. The homologous region of ELL aligns readily by sequence - ELL 500-580 and ELL2 519-599 are recognisably related - but sequence homology is a modelling premise, not a structure.
A tracer worth flagging
The FP assay is described as monitoring displacement of a fluorescently labelled wild-type AFF1 ELLbow peptide by unlabelled AFF4 301-351 mutants, with the unlabelled wild-type competitor at Kd 86 nM 1. Read literally, the tracer is AFF1 and the competitors are AFF4, which would make this the only affinity measurement in the SEC literature that touches AFF1 at all. Read as a typographical slip for AFF4, it is an all-AFF4 assay. The published text does not disambiguate, and this entry does not resolve it: the 86 nM is recorded as the affinity of the unlabelled wild-type ELLbow competitor, and the tracer's identity is flagged as uncertain rather than assumed.
Why the ELL/ELL2 distinction is not pedantry here
This graph already holds epi-adult-kmt2a-aml-ell-survival-advantage, concept-anti-ell-antibodies-and-the-paralogue-problem and a KMT2A::ELL fusion entry. ELL and ELL2 are separate genes with separate expression patterns and separate antibodies, and the field has repeatedly conflated them. The structural literature conflates them too, but in one direction only: everything solved is ELL2. Any statement about how KMT2A::ELL engages an AFF scaffold is currently an inference across a paralogue boundary.
| partner a | gene-aff4 |
|---|---|
| partner b | gene-ell2 |
| motif a | AFF4 ELLbow, residues 301-351: an N-terminal helix followed by an extended hairpin |
| motif b | ELL2 C-terminal occludin-fold domain, deposited span ~519-640 |
| kd NM | 86 |
| fate in fusion | retained in KMT2A::AFF1 by position on the AFF1 side if the AFF1 ELLbow is homologous and C-terminal to the breakpoint, but this has not been demonstrated for AFF1 and no AFF1 ELLbow structure or affinity exists |
| paralogue gap | KMT2A::ELL is the recurrent t(11;19) fusion and ELL, not ELL2, is its partner; the entire structural description of this interface is ELL2 |
| druggability | an imperfectly packed interface with a cavity, explicitly noted by the authors - which is an unusually favourable starting point for a peptidomimetic or small molecule, and nobody has pursued it in oncology |
| resolution note | PDB metadata and entity sequences fetched from data.rcsb.org 2026-09-20; spans aligned to UniProt Q9UHB7 and O00472 fetched the same day; absences established by RCSB search API with the queries recorded in caveats |
interface residues a
[ "AFF4 V316", "AFF4 I319", "AFF4 L320", "AFF4 M323", "AFF4 W327", "AFF4 L331", "AFF4 I334", "AFF4 P342" ]
interface residues b
[ "ELL2 H559", "ELL2 H608", "ELL2 N619", "ELL2 K625" ]
kd measurements
[
{
"ligand": "wild-type ELLbow competitor peptide",
"kd_nM": 86,
"method": "fluorescence polarisation displacement",
"source_pmid": "28134250",
"note": "the labelled tracer is described in the text as an AFF1 ELLbow peptide while the competitors are AFF4 301-351; the discrepancy is unresolved in the published text"
},
{
"ligand": "AFF4 I319D/L320D",
"fold_loss": ">25",
"method": "fluorescence polarisation",
"source_pmid": "28134250"
},
{
"ligand": "AFF4 I319D/L320D/M323D",
"kd_nM": ">3000",
"fold_loss": "~50",
"method": "fluorescence polarisation",
"source_pmid": "28134250"
}
]structures
[ "5JW9" ]
structure details
[
{
"pdb": "5JW9",
"title_fetched": "The Crystal Structure of ELL2 Oclludin Domain and AFF4 peptide",
"method": "X-ray",
"resolution_A": 2.003,
"contents": "AFF4 64-mer (Q9UHB7 ~301-352 plus tags) + ELL2 122-mer (O00472 ~519-640)",
"fetched": "2026-09-20"
}
]structures absent
[ "no structure of ELL (P55199) with AFF4 or AFF1", "no structure of ELL with ELL2", "P55199 appears in exactly one PDB entry, 2DOA" ]
- Paralogue substitution: ELL, ELL2 and ELL3 can all occupy the SEC, so blocking the occludin contact on one paralogue invites replacement by another.
- Wrong paralogue: the recurrent leukaemia fusion partner is ELL and every structure is ELL2, so structure-guided design here starts from a homology model of the wrong protein.
- Viral framing: the entire interface literature is HIV-1 proviral transcription, so the functional assays validate a viral readout rather than a leukaemic one.
- Scaffold redundancy: AFF1 and AFF4 both carry ELLbow-like regions, so removing one scaffold's contact may be compensated by the other.
- Does AFF1 have an ELLbow, and does it bind ELL2 or ELL with comparable affinity? Neither has been measured.
- Does KMT2A::ELL use the occludin domain to recruit an AFF scaffold, and if so is the arrangement the same as ELL2's? The fusion retains ELL's C-terminus, so this is a live and unanswered question about a real recurrent fusion.
- Is the cavity at the AFF4-ELL2 interface, which the authors note is imperfectly packed, large enough to be a small-molecule site?
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The AFF4 ELLbow (residues 301-351) binds the ELL2 occludin domain with Kd 86 nM by fluorescence polarisation; AFF4 I319D/L320D loses over 25-fold and the I319D/L320D/M323D triple mutant about 50-fold, while ELL2 H559E/H608E/N619A/K625T abolishes binding only as a quadruple mutant.
Structural basis for ELL2 and AFF4 activation of HIV-1 proviral transcription. · Qi S, Li Z, Schulze-Gahmen U, et al. · Nat Commun 2017What this evidence cannot show
The crystal used a fusion construct, which the authors themselves flag as the reason they validated binding separately by FP. The tracer peptide is described as AFF1 while the competitors are AFF4 and the text does not resolve the discrepancy. All work is in an HIV-1 proviral transcription frame.
Evidence that argues against this (1)
- The authors note the AFF4-ELL2 interface is imperfectly packed and leaves a cavity, which is an argument that the crystallised arrangement is one of several and that a drug-like ligand might occupy the interface without the peptide geometry being the whole story. Qi S et al. 2017
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
PDB 5JW9 is a 2.00 A crystal of an ELL2 domain with an AFF4 peptide; the AFF4 entity maps to Q9UHB7 ~301-352 and the ELL2 entity to O00472 ~519-640.
PDB 5JW9: The Crystal Structure of ELL2 Oclludin Domain and AFF4 peptide · Qi S, Li Z, Schulze-Gahmen U, et al. · RCSB Protein Data Bank 2017What this evidence cannot show
Title, method, resolution and both entity sequences fetched from data.rcsb.org 2026-09-20; residue spans are this cycle's alignment to UniProt sequences fetched the same day, and the AFF4 entity carries an N-terminal GAMGS tag and a C-terminal GSGSGS linker that must be stripped first. The RCSB entry title contains the typo 'Oclludin'. INTERACTION PROVENANCE (Q76, 2026-09-21, live IntAct REST): every one of the six IntAct records for the AFF4-ELL2 pair is MI:1060 spoke-expanded and typed MI:0914 association, from PMIDs 21729782, 26496610, 28514442 and 33961781 - so in the curated interaction record this pair has NEVER been observed as a binary interaction, and 5JW9 plus the FP displacement in E1 are the entire basis for calling it a contact. That is not a weakness of the pair, it is the reason this entry exists: a spoke-expanded row plus a 2.0 A structure is strong, and a spoke-expanded row alone is not. A spoke row is still a real observation that the prey was in the pulldown; what it cannot support on its own is a direct-contact claim. See ana-two-thirds-of-the-curated-interactome-is-a-row-nobody-observed.
Evidence that argues against this (1)
- A single-chain crystallisation construct was used; the bimolecular arrangement is supported by the solution FP data rather than by the crystal itself. Qi S et al. 2017
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
No PDB entry contains ELL (P55199) together with AFF4 (Q9UHB7) or with ELL2 (O00472); P55199 appears in exactly one entry in the whole PDB.
PDB 5JW9: The Crystal Structure of ELL2 Oclludin Domain and AFF4 peptide · Qi S, Li Z, Schulze-Gahmen U, et al. · RCSB Protein Data Bank 2017What this evidence cannot show
RCSB search API, exact_match on reference_sequence_identifiers.database_accession, run 2026-09-20: P55199 alone returns 1 entry (2DOA); P55199 + Q9UHB7 and P55199 + O00472 each return an empty result set. A deposit lacking a UniProt cross-reference would be invisible to this query, so the absence is bounded by the annotation, not by the structure.
Evidence that argues against this (1)
- ELL and ELL2 are homologous over the relevant region, so a homology model is possible and the absence of a deposit is not the absence of a usable structural hypothesis - only of an experimental one. Qi S et al. 2017
AFF4 ELLbow 301-351, Kd 86 nM
ELL2 occludin domain ~519-640
no structure of ELL with any AFF scaffold exists
- related to →KMT2A::ELL (MLL-ELL), t(11;19)(q23;p13.1)
the t(11;19) partner's scaffold interface is unsolved; ELL2 is the stand-in
the same ELL/ELL2 conflation, in the structural literature
the other characterised ELL-family interface in this graph
the queries establishing that ELL appears in one PDB entry
AFF4-ELL2 is 6 of 6 spoke-expanded in IntAct and is nonetheless a solved crystal structure, PDB 5JW9 - the clearest case that the flag does not grade the evidence