Oncofusion

start / graph / interface-aff1-aff4-chd-heterodimer

Interfacemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q31-interfaces-structural

AFF1-AFF4 heterodimerisation through the C-terminal homology domain: the one SEC interface KMT2A::AFF1 certainly keeps, defined by a 2.2 A structure of the wrong subunit

also called AFF4 CHD · AFF1 CHD · C-terminal homology domain · 6R80

AFF1 and AFF4 clip together through a domain at their far end, and that domain is inside the leukaemic fusion protein. The structure of the clip was solved on AFF4; AFF1's copy has never been solved at all.

The same thing, for a scientist

Chen and Cramer solved AFF4 residues 899-1163 (PDB 6R80, 2.2 A) and found an eight-helix fold, distantly related to TPR motifs, that dimerises through helices alpha5/alpha6 and the connecting loop over a 1,655 square-angstrom interface. Analytical size-exclusion chromatography with static light scattering shows AFF4-CHD forms homodimers and that an MBP-tagged AFF1-CHD construct (AFF1 residues 938-1210) forms a heterodimer with AFF4-CHD; a five-point interface mutant (H1090/Y1096/V1097/F1103/L1104, to either alanine or aspartate) abolishes both the homodimer and the heterodimer. AFF1 is 1,210 residues long, so the AFF1 CHD occupies its extreme C-terminus - which is squarely inside KMT2A::AFF1 for every published breakpoint. This is therefore the SEC interface with the strongest positional claim to being retained by the index fusion, and there is no structure of it, no affinity for it, and no experiment testing it on a fusion protein.

Go deeper
Why this is the SEC interface that matters most for KMT2A::AFF1

Work through the SEC contacts by position on AFF1 (P51825, 1,210 aa):

contactAFF1 regionin the fusion?
P-TEFb / cyclin T1homologous to AFF4 2-73, i.e. AFF1 N-terminusno - 5' of the breakpoint
ELLbowhomologous to AFF4 301-351no - 5' of the breakpoint
AF4_int to AF9/ENL AHDAFF1 ~737-779yes
CHD heterodimerisationAFF1 938-1210yes

So of the four, two are lost with the AFF1 N-terminus and two are kept. The graph already has an entry for the AF4_int/AHD contact. The CHD contact has had none - and it is the one that would let a fusion recruit a full-length AFF4 molecule, which carries the P-TEFb and ELLbow contacts the fusion itself lost. Heterodimerisation is the mechanism by which KMT2A::AFF1 could reconstitute a complete SEC from a partial one.

What is measured and what is not

Measured: the AFF4 homodimer structure, its interface area, the five residues that carry it, and the fact that mutating them abolishes AFF1-AFF4 heterodimer formation in analytical SEC and static light scattering 1.

Not measured anywhere: a dissociation constant for AFF1-AFF4 heterodimerisation, a structure of AFF1-CHD, whether the KMT2A fusion moiety alters CHD dimerisation, and whether the heterodimer forms in leukaemia cells at endogenous stoichiometry.

Chen and Cramer also identify a surface loop of AFF4-CHD as a CDK9 substrate 1, so the dimerisation domain is itself phospho-regulated by the kinase the complex recruits - a feedback arrangement with no leukaemia-side investigation.

Facts and figures
partner agene-aff1
partner bgene-aff4
motif aAFF1 C-terminal homology domain, construct residues 938-1210 (P51825 is 1,210 aa, so the CHD is its extreme C-terminus)
motif bAFF4 C-terminal homology domain, crystallised 899-1163; dimerisation through helices alpha5/alpha6 and loop L(alpha5-alpha6)
interface area A21655
kd statusno dissociation constant published for AFF1-AFF4 heterodimerisation or for AFF4 homodimerisation; the evidence is analytical SEC with static light scattering
fate in fusionretained - the AFF1 CHD occupies residues ~938-1210 of a 1,210-residue protein and lies far C-terminal to every published AFF1 breakpoint
mechanistic significanceheterodimerisation with full-length AFF4 is the route by which a fusion that has lost AFF1's N-terminal P-TEFb and ELLbow contacts could still assemble a complete super elongation complex, in trans
druggabilitya large (1,655 square-angstrom) helix-helix interface carried by five hydrophobic and aromatic residues; large PPI interfaces of this class are hard but the hot-spot concentration is favourable, and no chemical matter has been reported
resolution notePDB metadata and the 6R80 entity sequence fetched from data.rcsb.org 2026-09-20 and aligned to UniProt Q9UHB7 fetched the same day; AFF1 construct boundaries taken from the fetched full text of PMID 31147444
structure details
6R802.2006R80: 2.26KN52.2006KN5: 2.26K7P2.4006K7P: 2.4
interface residues b
[
  "AFF4 H1090",
  "AFF4 Y1096",
  "AFF4 V1097",
  "AFF4 F1103",
  "AFF4 L1104"
]
kd measurements
[]
structures
[
  "6R80",
  "6K7P",
  "6KN5"
]
structure details — the numbers behind the chart
[
  {
    "pdb": "6R80",
    "title_fetched": "Structure of AFF4 C-terminal homology domain",
    "method": "X-ray",
    "resolution_A": 2.2,
    "contents": "single AFF4 entity, His-TEV tagged, within Q9UHB7 899-1163",
    "fetched": "2026-09-20"
  },
  {
    "pdb": "6K7P",
    "title_fetched": "Crystal structure of human AFF4-THD domain",
    "method": "X-ray",
    "resolution_A": 2.4,
    "fetched": "2026-09-20"
  },
  {
    "pdb": "6KN5",
    "title_fetched": "Crystal structure of AFF4 C-terminal domain",
    "method": "X-ray",
    "resolution_A": 2.2,
    "fetched": "2026-09-20"
  }
]
structures absent
[
  "no structure of AFF1-CHD",
  "no structure of an AFF1-AFF4 heterodimer",
  "no structure of any KMT2A fusion protein"
]
How this could bite us back
  • Compensation in trans: if the fusion recruits full-length AFF4 through this domain, then removing AFF1's own elongation contacts does nothing, because the AFF4 partner supplies them.
  • Sink effect: a fusion that binds AFF4 tightly could deplete AFF4 from normal SECs, so disrupting the heterodimer might release a pool of wild-type activity as well as removing a fusion function.
  • Undruggable geometry: a 1,655 square-angstrom helix-helix interface with no reported chemical matter is at the difficult end of the PPI spectrum.
  • Evidence class: heterodimerisation is shown by size-exclusion and light scattering on MBP-tagged constructs, not by structure or by any measurement in cells, so an intervention would be designed against an interface nobody has seen.
Still open
  1. Does KMT2A::AFF1 heterodimerise with endogenous AFF4 in leukaemia cells, and does that heterodimer carry P-TEFb? This is the single most consequential untested claim in this entry.
  2. What is the dissociation constant of the AFF1-AFF4 CHD heterodimer relative to either homodimer? If the heterodimer is preferred, the fusion is a sink for AFF4.
  3. Does the KMT2A moiety, or its chromatin tethering, change CHD dimerisation - for instance by raising local concentration at fusion target genes?
  4. Is the CDK9 phosphorylation site on the AFF4-CHD surface loop conserved in AFF1, and is it phosphorylated on the fusion?
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    AFF4-CHD (residues 899-1163) folds into eight helices and dimerises through helices alpha5/alpha6 over a 1,655 square-angstrom interface; a five-residue interface mutant abolishes both AFF4 homodimer and AFF1-AFF4 heterodimer formation, the AFF1-CHD construct used being AFF1 residues 938-1210; and a surface loop of AFF4-CHD is phosphorylated by CDK9 in vitro.

    PMID 31147444DOI
    What this evidence cannot show

    Heterodimerisation is shown by analytical size-exclusion chromatography and static light scattering on an MBP-tagged AFF1-CHD, with no dissociation constant; MBP tags are themselves capable of influencing oligomeric behaviour. The AFF1-CHD construct boundaries were designed from alignment to the AFF4 structure rather than determined experimentally. Full text reached by the PMC HTML route (64,077 characters) after fullTextXML returned HTTP 500 and efetch db=pmc returned no body element.

    Evidence that argues against this (1)
    • An independent crystal form of the AFF4 C-terminal domain, solved at 2.4 A in the context of HIV-1 proviral transcription, frames AFF4 dimerisation primarily as a determinant of transactivation strength rather than of AFF-family heterodimer composition; the two papers agree the domain dimerises and differ on what dimerisation is for. Tang D et al. 2020
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    PDB 6R80 is a 2.2 A crystal of the AFF4 C-terminal homology domain, a single polymer entity mapping to Q9UHB7 within 899-1163.

    PDB 6R80: Structure of AFF4 C-terminal homology domain · Chen Y, Cramer P · RCSB Protein Data Bank 2019
    What this evidence cannot show

    Title, method, resolution and the entity sequence fetched from data.rcsb.org 2026-09-20; the entity carries an N-terminal His6-TEV tag which must be stripped before alignment to UniProt Q9UHB7, fetched the same day. A homodimer in a crystal can be a lattice contact; the solution light-scattering data are what make it a dimer.

    Evidence that argues against this (1)
    • A second AFF4 C-terminal domain crystal form exists (6K7P, 2.4 A) from a different group, so the fold is reproduced independently - but neither deposit contains AFF1. Tang D et al. 2020
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    No PDB entry contains AFF1 (P51825) together with AFF4 (Q9UHB7); AFF1 appears in exactly one PDB entry, the AFF1/AF9 chimera 2LM0.

    PDB 2LM0: Solution structure of the AF4-AF9 complex · Leach BI, Kuntimaddi A, Schmidt CR, et al. · RCSB Protein Data Bank 2013
    What this evidence cannot show

    RCSB search API, exact_match on reference_sequence_identifiers.database_accession, run 2026-09-20. P51825 alone returns one entry; P51825 + Q9UHB7 returns an empty result set. Bounded by UniProt cross-annotation. Independently reproduced 2026-09-21 by intersecting the UniProt PDB cross-reference lists for P51825 and Q9UHB7: empty. INTERACTION PROVENANCE (Q76, 2026-09-21, live IntAct REST): the curated interaction record is no better than the structural one. All eight IntAct records for the AFF1-AFF4 pair are MI:1060 spoke-expanded and typed MI:0914 association, from PMIDs 20153263 and 21729782 - both focused n-ary complex purifications, neither a binary binding experiment - so IntAct holds NO observed binary AFF1-AFF4 interaction either. Taken with E1's analytical SEC and light scattering on an MBP-tagged construct, the heterodimer rests on one biophysical experiment in one paper and on co-purification; it has no Kd, no structure and no curated binary record. That does not make it false - a spoke row is a real observation that both proteins were in the same pulldown - but this entry's own title calls it 'the one SEC interface KMT2A::AFF1 certainly keeps', and the certainty is positional (the CHD is C-terminal to every breakpoint), not evidential. See ana-aff1-has-one-observed-binary-interaction-and-forty-two-residues-of-structure.

    Evidence that argues against this (1)
    • The AFF1 CHD is highly conserved with AFF4's - the interface residues are described as highly conserved among AFF-family CHDs - so the absence of an AFF1 structure is a weaker objection here than it would be for a divergent region. Chen Y et al. 2019
Connected to