Oncofusion

start / graph / exp-aff1-chd-dimerisation-separation-of-function

Experimentmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q2b-atlas2

Separation-of-function test: does KMT2A::AFF1 need its C-terminal homology domain to dimerise?

also called CHD deletion and interface-mutant experiment · AFF1 CHD separation of function

Make three versions of the fusion protein — normal, one with the suspected sticky module cut off, and one with just a few surface letters changed so it keeps its shape but cannot stick — and see which ones can still turn blood cells cancerous. If only the normal one works, the sticky module is doing real work.

The same thing, for a scientist

A three-arm separation-of-function experiment testing whether the retained AFF1 C-terminal homology domain (949-1209) contributes to KMT2A::AFF1 transformation through dimerisation. Arm 1: full-length fusion. Arm 2: CHD truncation at residue 948. Arm 3: point mutants at the dimer interface. **Updated 2026-09-19: these no longer need designing.** An independent crystal structure of the same domain (called THD there, CHD here) reports that single alanine substitutions at AFF4 F1014 or Y1096 abolish dimerisation, names the equivalent AFF1 residues as **F1063 and Y1145**, shows they behave the same way, and - the part that matters most for a separation-of-function - reports that they do NOT disturb the interaction of AFF1/AFF4 with other super-elongation-complex subunits. We verified both residue identities directly against the UniProt sequences. Arm 3 is now AFF1 F1063A and Y1145A, singly and together. Readouts: co-immunoprecipitation of fusion with itself and with endogenous AFF4; serial replating of transduced mouse HSPCs; Hoxa9/Meis1 expression; and in vivo latency.

Go deeper
Design1 deeper

Constructs. Retrovirally expressed KMT2A::AFF1 (a standard junction, e.g. KMT2A exon 9 to AFF1 exon 4 as used in the Cleary-lab dimerisation work) in three forms:

  • FL — full length.

  • ΔCHD — stop at AFF1 residue 948, removing 949-1209 while retaining the pSER/TAD, the predicted NLS and the AF9/ENL-binding motif at 761-775. This is the crude arm and it has a known weakness: deleting 261 residues changes more than one thing.

  • CHD-mut — the informative arm, and it is already specified in the literature 4. Do not repeat the alignment. An independent 2.4 Å structure of the same domain gives the dimer interface as helices α4-α6 and shows that a single alanine substitution at AFF4 F1014 or Y1096 abolishes dimerisation; the equivalent AFF1 residues are named as F1063 and Y1145 and behave the same way. We checked both against the live UniProt sequences: AFF1 P51825 residue 1063 is indeed F and 1145 is indeed Y; AFF4 Q9UHB7 residue 1014 is F and 1096 is Y 7. Make F1063A, Y1145A and the double mutant. Circular dichroism and thermal shift are still required, but as confirmation rather than as a screen, and the AlphaFold-Multimer step drops from a design requirement to a cross-check.

    The same paper reports the property that makes this a real separation-of-function and not just a deletion: these mutations do not affect the interaction of AFF1/AFF4 with other super-elongation-complex subunits 4. So the point mutant removes dimerisation while leaving complex assembly intact — which is exactly what the ΔCHD arm cannot do.

    Two cautions. The published functional readout is HIV-1 proviral transactivation, not leukaemic transformation, so what transfers is the structural claim and the SEC-assembly control, not a prediction about self-renewal. And the two structures of this domain call it by different names — CHD in one paper, THD in the other — which is why this design was built without knowing the mutants existed. Search both.

Controls. A KMT2A::AFF1 carrying a mutation in the 761-775 AF9/ENL motif, as a positive control for a separation-of-function that should attenuate; and empty vector. Include a KMT2A::EPS15 arm if available, as a known oligomerisation-dependent fusion.

Readouts, in order of cost.

  1. Co-IP in 293T and in a fusion-expressing leukemia line: does the fusion pull down itself (differentially tagged) and endogenous AFF4, and is that CHD-dependent?
  2. Serial replating of transduced mouse bone-marrow HSPCs — the standard assay in which EPS15 and AFDN oligomerisation requirements were established, so results are directly comparable to that literature.
  3. Hoxa9 and Meis1 by qPCR at each replating.
  4. In vivo transplantation with latency and penetrance, only for arms that separate in vitro.

Orthogonal arm, cheap and worth doing first. AFF4 knockdown titration in a KMT2A::AFF1 leukemia line. If heterodimerisation is load-bearing, fusion activity should fall faster than a pure scaffolding model predicts.

What would make this uninterpretable

Three failure modes worth pre-empting, because this design has been run badly before in other systems.

The interface mutant does not fold. If CD or thermal shift shows the isolated mutant domain is destabilised, a negative transformation result means nothing. This is why the biophysics comes before the biology, not after. This risk is now much lower than when the node was written: the mutants are published single alanine substitutions that abolish dimerisation without disturbing SEC assembly 4, which is strong indirect evidence that the fold survives.

ΔCHD removes something other than the dimer surface. The AFF4 CHD also binds RNA and DNA and carries a CDK9-substrate loop. A ΔCHD phenotype could be any of those. Only the point mutant distinguishes them — which is why ΔCHD alone is not a sufficient experiment and is included mainly as a magnitude estimate.

Overexpression forces dimerisation that does not happen at endogenous levels. Retroviral constructs sit far above physiological concentration and weak dimers become obligate ones. Mitigate by including a dose series, or by knock-in at endogenous levels — and the endogenous-level option is no longer hypothetical: a CRISPR-edited human MLL::AF4 model in CD34+ cells has been published and recapitulates infant pro-B ALL features, with breakpoint site and cellular ontogeny both affecting leukaemogenic capacity 6.

Counter-evidence: what this design got wrong, and the arm it may be aiming at

Added 2026-09-19.

1. The hardest step is already done. This node's first open question asked whether a fold-preserving dimerisation-null mutant of the CHD can be designed at all, and treated the structural alignment as weeks of work with an uncertain outcome. It has been answered in the literature: single alanine substitutions at AFF4 F1014 or Y1096 each abolish dimerisation, the equivalent AFF1 residues are F1063 and Y1145, and crucially these mutations leave the interaction with other super-elongation-complex subunits intact 4. That is a published, validated separation of function. The question has been rewritten accordingly.

The reason this node did not know is worth recording on its own: the two structural papers on this domain call it by different names, CHD in one and THD in the other. A literature search anchored on 'C-terminal homology domain' returns one and not the other.

2. The arm most likely to separate may be the control. The closest published analogue of this experiment is a structure-function analysis of MLL-ELL, and it came out the other way round: the partner's canonical elongation domain was dispensable for immortalisation, while its EAF1-interaction domain was essential; MLL fused to only that minimal interaction domain still caused AML, and a heterologous MLL-EAF1 fusion immortalised as well as MLL-ELL and caused AML with shorter latency 5. EAF1 shares homology with the transactivation domains of AF4, LAF4 and AF5q31 — the AFF1 family.

If AFF1 behaves like ELL, then the load-bearing element is the protein-interaction and transactivation region in the AFF1 N-terminal half, which this design lists only as a positive control (the 761-775 AF9/ENL motif), and the CHD arm returns negative. That is still an informative experiment — but it is a different experiment, and the arms should be weighted for it. Recommendation: promote the 761-775 motif mutant from control to co-primary arm.

3. The coordinates are borrowed. UniProt annotates no domain over the AFF1 C-terminus at all; it marks 366-957 as disordered, which overlaps the stated CHD start at 949 7. The 949-1209 range is inherited from the AFF4 structural work, not curated for AFF1. It is very probably right — the domain has two crystal structures in the paralogue — but the experiment should not treat 948 as a sharp truncation point without checking where the fold actually starts in AFF1.

None of this kills the experiment. It makes it cheaper (the mutants exist), sharpens what the negative result would mean (the SEC-assembly control is already published), and shifts the prior towards a negative CHD arm. The node's kills_hypothesis_if is unchanged and is now testable in weeks rather than months.

Facts and figures
designThree-arm separation-of-function in retrovirally transduced mouse bone-marrow HSPCs: full-length KMT2A::AFF1 vs CHD truncation at AFF1 residue 948 vs the PUBLISHED fold-preserving dimer-interface point mutants AFF1 F1063A and Y1145A (singly and combined), taken from Tang 2020 rather than designed de novo. Controls: 761-775 AF9/ENL-motif mutant, empty vector, and a KMT2A::EPS15 oligomerisation-dependent comparator. Isolated-domain CD and thermal shift required before any mutant enters a cell.
readoutCHD-dependent fusion self-association and endogenous AFF4 co-immunoprecipitation; serial replating colony counts across 3-4 platings; Hoxa9 and Meis1 qPCR at each plating; disease latency and penetrance on transplantation for arms separating in vitro.
controlsFold-integrity controls (CD, thermal shift) on every mutant domain; expression-level matching by Western across arms; an expression dose series to guard against overexpression-forced dimerisation; the 761-775 motif mutant as a positive separation-of-function control.
cost tiermed
timelineStructural alignment and mutant design 2-4 weeks; construct generation and biophysics 6-8 weeks; co-IP and replating 3 months; in vivo arm a further 4-6 months.
cheap first stepAFF4 knockdown titration in an existing KMT2A::AFF1 leukemia line, PLUS - now the cheapest step of all - express KMT2A::AFF1 carrying the published F1063A and Y1145A substitutions and co-immunoprecipitate. The mutants are specified in the literature and need only be synthesised, so the first informative experiment is a site-directed mutagenesis and a co-IP, not a structural campaign.
kills hypothesis ifFold-verified interface point mutants transform indistinguishably from full-length fusion.
coordinate provenanceThe AFF1 CHD range 949-1209 is inherited from the AFF4 structural literature. UniProt annotates NO domain over the AFF1 C-terminus (P51825 carries only disorder at 1-45, 73-314, 366-957 and 1098-1119, plus a PARP1-interaction region at 601-900), and its disorder annotation to 957 overlaps the stated domain start at 949. Verify where the AFF1 fold actually starts before treating 948 as a truncation point.
prior on outcomeShifted towards a NEGATIVE CHD arm. The closest published analogue (MLL-ELL, PMID 11463848) found the partner's canonical domain dispensable and its protein-interaction domain essential, with a heterologous MLL-EAF1 fusion transforming at least as well. Recommendation: promote the 761-775 AF9/ENL-motif mutant from control to co-primary arm.
endogenous model existsYes, as of 2023: a CRISPR-edited human MLL::AF4 model in CD34+ cells recapitulating infant pro-B ALL features (PMID 37756522). This node's second open question assumed the knock-in route was prohibitive; it is no longer hypothetical, though its tractability for a three-arm design is unknown from the abstract.
precedent assays
[
  "serial replating (So 2003)",
  "minimal-region mapping (Liedtke 2010)"
]
published interface mutants
{
  "aff4": {
    "F1014": "alanine substitution abolishes AFF4 dimerisation and reduces transactivation",
    "Y1096": "alanine substitution abolishes AFF4 dimerisation and reduces transactivation"
  },
  "aff1": {
    "F1063": "equivalent residue; affects transactivation of HIV-1 provirus",
    "Y1145": "equivalent residue; affects transactivation of HIV-1 provirus"
  },
  "sec_assembly_control": "These mutations have NO effect on the interaction of AFF1/AFF4 with other SEC subunits (Tang 2020, PMID 32128251). That is the published separation-of-function control this design was going to have to generate itself.",
  "residue_identity_verified": "Checked against live UniProt sequences 2026-09-19: AFF1 P51825 (1210 aa) residue 1063 = F, residue 1145 = Y; AFF4 Q9UHB7 (1163 aa) residue 1014 = F, residue 1096 = Y. Primary gene names confirmed as AFF1 and AFF4.",
  "readout_caveat": "The published functional readout is HIV-1 proviral transactivation, not leukaemic transformation. The structural claim and the SEC-assembly control transfer; a prediction about self-renewal does not.",
  "naming_trap": "The same domain is called CHD (Chen & Cramer 2019, PDB 6R80) and THD, 'TPRL with Handle Region Dimerization Domain' (Tang 2020). A search on one name misses the other. This node was written without knowing the mutants existed for exactly that reason."
}
counter evidence pass
{
  "queen": "Q2b-atlas2",
  "cycle": 31,
  "date": "2026-09-19",
  "checked": [
    "PMID 31147444 (Chen & Cramer) metadata re-fetched and abstract read back - the node's E1 claim is supported, and the abstract confirms the domain ALSO binds RNA and DNA and carries a CDK9 substrate loop, which is why the deletion arm is uninformative",
    "PMID 32128251 (Tang 2020) metadata fetched, abstract read",
    "PMID 11463848 (Luo 2001) metadata fetched, abstract read",
    "PMID 37756522 (Bueno 2023) metadata fetched, abstract read",
    "UniProt P51825 and Q9UHB7 live sequences fetched and individual residues 1063/1145 and 1014/1096 read back"
  ]
}
How this could bite us back
  • A negative result here is weakly informative: the mouse progenitor assay could miss a contribution that matters in human fetal pro-B cells, which is the actual cell of origin for infant t(4;11).
  • A positive result creates a target whose wild-type counterpart is used by every cell for normal pause release, so success in this experiment does not imply a therapeutic window.
  • The published dimerisation-null mutants were validated on HIV-1 proviral transactivation. Importing them into a leukaemia experiment imports an assumption that the same interface does the same job in both settings, and if it does not, a clean negative result here would be a false negative with a citation attached.
Still open
  1. ANSWERED, 2026-09-19. The original question asked whether a fold-preserving dimerisation-null CHD mutant can be designed at all. It can, and it has been: single alanine substitutions at AFF4 F1014 or Y1096 abolish dimerisation, the AFF1 equivalents are F1063 and Y1145, and they leave SEC-subunit interaction intact [E4]. The live question is now narrower: does a dimerisation-null AFF1 CHD behave the same way inside a KMT2A fusion as it does in free AFF1, where the fusion's N-terminal KMT2A moiety supplies its own targeting and avidity?
  2. Should the endogenous-level version be done instead? A CRISPR-edited human MLL::AF4 model in CD34+ cells now exists and recapitulates infant pro-B ALL features [E6], so the knock-in route this node dismissed on cost is at least established. Whether it tolerates a three-arm allelic series is not answerable from the published Letter.
  3. Does the in vitro replating assay have the sensitivity to detect a partial contribution? If CHD dimerisation contributes 30% of the fusion's activity rather than being required, serial replating may not resolve it and a competitive in vivo assay would be needed.
  4. If the CHD arm comes back negative and the 761-775 AF9/ENL-motif arm separates - which the MLL-ELL precedent predicts [E5] - does that mean AFF1 contributes to KMT2A::AFF1 purely as an adaptor? A heterologous MLL-EAF1 fusion transformed at least as well as MLL-ELL, so the analogous test for AFF1 is whether KMT2A fused to a minimal AF9/ENL-binding module alone reproduces the disease.
Every source behind this entry (7)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The AFF4 CHD structure (PDB 6R80, 2.2 Å) defines the dimer interface whose residues would be transplanted onto AFF1 by structural alignment, and establishes that the domain mediates AFF4 homodimerisation and AFF1-AFF4 heterodimerisation.

    PMID 31147444DOI
    What this evidence cannot show

    AFF4 coordinates, not AFF1. The alignment step is itself an assumption and should be checked with an AlphaFold-Multimer or Boltz model of the AFF1 CHD dimer before committing to mutants.

    Evidence that argues against this (2)
    • The alignment step this design treats as the hard part is already published. An independent crystal structure of the same AFF4 domain gives the dimer interface as helices alpha4-alpha6 and shows that single alanine substitutions at AFF4 F1014 or Y1096 abolish dimerisation, with the corresponding AFF1 residues named as F1063 and Y1145 and shown to behave the same way. Arm 3 should therefore START from AFF1 F1063A and Y1145A rather than repeating the alignment, and the AlphaFold-Multimer step drops from a design requirement to a cross-check. Tang D et al. 2020
    • The two structures of this domain use different names for it - CHD in one, THD ('TPRL with Handle Region Dimerization Domain') in the other. Anyone searching on one name misses the other, and this node was built entirely on the CHD literature. Tang D et al. 2020
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Serial replating of retrovirally transduced mouse bone-marrow progenitors is the established assay in which the oligomerisation requirement for MLL-AF1p/EPS15 and MLL-GAS7 was demonstrated, with synthetic dimerisation modules shown to substitute.

    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

    Mouse progenitor assays reward self-renewal specifically and have historically over-called transformation for constructs that do not produce leukemia in vivo, which is why the in vivo arm is retained despite its cost.

    Evidence that argues against this (1)
    • The serial-replating precedent may be pointing at the wrong arm. In the closest published structure-function analysis of an SEC-component MLL fusion, the partner's canonical elongation domain was DISPENSABLE for immortalisation while its protein-interaction domain was essential - and MLL fused to a heterologous EAF1 transactivation domain caused AML with shorter latency than the natural fusion. If the AF4 family behaves like ELL, the arm that separates is the 761-775 AF9/ENL-binding motif that this design lists as a control, and the CHD arm is the one that comes back negative. Luo RT et al. 2001
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The AF6 RA1 precedent shows that a minimal partner module can be both necessary and sufficient for MLL-fusion transformation, and that self-association rather than the module's canonical ligand-binding function can be the operative property.

    Self-association mediated by the Ras association 1 domain of AF6 activates the oncogenic potential of MLL-AF6 · Liedtke M, Ayton PM, Somervaille TCP, Smith KS, Cleary ML · Blood 116:63-70 2010
    PMID 20395419DOI
    What this evidence cannot show

    AFDN is a cytoplasmic partner with no independent transcription-elongation function, so the inference that a similar design will be as clean for AFF1 — which has a well-supported alternative mechanism — is optimistic.

    Evidence that argues against this (1)
    • The AFDN RA1 precedent is the optimistic reading of the design; the MLL-ELL precedent is the pessimistic one and is a closer analogue, because ELL like AFF1 is a genuine elongation factor with an alternative mechanism available, whereas AFDN is a cytoplasmic protein with none. The node's own caveat on this item anticipated the objection; the MLL-ELL result makes it concrete. Luo RT et al. 2001
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The structural-alignment step this experiment proposes has already been done and functionally validated. An independent 2.4 A crystal structure of the AFF4 C-terminal domain (named there the THD, 'TPRL with Handle Region Dimerization Domain') identifies helices alpha4, alpha5 and alpha6 as the dimer interface, shows that SINGLE alanine substitutions at AFF4 Phe1014 or Tyr1096 each abolish dimerisation and each reduce transactivation, identifies the equivalent AFF1 residues as Phe1063 and Tyr1145 and shows they behave the same way - and, decisively for this design, reports that these mutations have NO effect on the interaction of AFF1/AFF4 with other subunits of the super-elongation complex.

    Structural and functional insight into the effect of AFF4 dimerization on activation of HIV-1 proviral transcription · Tang D, Chen C, Liao G, Liu J, Liao B, Huang Q, Chen Q, Zhao J, Jiang H, Duan J, Huang J, Wang K, Wang J, Zhou C, Chu W, Li W, Sun B, Li Z, Dai L, Fu X, Cheng W, Xue Y, Qi S · Cell Discov 6:7 2020
    PMID 32128251DOI
    What this evidence cannot show

    ABSTRACT ONLY for the structural detail; the functional readout is HIV-1 proviral transactivation in a reporter assay, NOT leukaemic transformation, so the separation of function is established for transcription of one viral promoter and not for fusion-driven self-renewal. We independently verified the residue identities against the live UniProt sequences: AFF1 P51825 (1,210 aa) residue 1063 is F and 1145 is Y; AFF4 Q9UHB7 (1,163 aa) residue 1014 is F and 1096 is Y. Note also that this paper calls the domain THD while the structure this node's design is built on calls it CHD; they are the same module and a literature search on one name will miss the other.

    Evidence that argues against this (1)
    • AFF1 and AFF4 are not interchangeable in the very assay that established the separation of function. In HIV transactivation experiments AFF4 efficiently activates viral transcription while its homolog AFF1 is substantially lower, and the difference traces to differential recruitment to the promoter and differential association with the PAF complex; P-TEFb phosphorylation of AFF4 itself modulates AFF1/AFF4 dimerisation. So reading AFF4 Phe1014/Tyr1096 phenotypes onto AFF1 Phe1063/Tyr1145 assumes an equivalence that the same experimental system contradicts, and a null effect on other super-elongation-complex subunits measured on AFF4 does not license the same claim for AFF1. Kuzmina A et al. 2017
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In the closest published analogue of this experiment, the canonical function of the partner domain turned out to be the dispensable one: a structure-function analysis of MLL-ELL found that the elongation domain of ELL was dispensable for immortalisation of myeloid progenitors while the EAF1-interaction domain was essential, that MLL fused to only the minimal EAF1-interaction domain still caused AML in mice, and that a heterologous MLL-EAF1 fusion immortalised progenitors as well as MLL-ELL and caused AML with SHORTER latency. EAF1 shares a limited region of homology with the transactivation domains of AF4, LAF4 and AF5q31.

    The elongation domain of ELL is dispensable but its ELL-associated factor 1 interaction domain is essential for MLL-ELL-induced leukemogenesis · Luo RT, Lavau C, Du C, Simone F, Polak PE, Kawamata S, Thirman MJ · Mol Cell Biol 21:5678-5687 2001
    PMID 11463848DOI
    What this evidence cannot show

    ABSTRACT ONLY. MLL-ELL and myeloid progenitors, not MLL-AF4 and pro-B cells, and 2001 retroviral constructs at non-physiological expression. The inference is by analogy within the ELL/AF4 partner family, which is exactly the kind of inference this graph is meant to flag rather than rely on. It is cited as a prior on which arm will separate, not as a result about AFF1.

    Evidence that argues against this (1)
    • The ELL precedent predicts the wrong prior for this family. In the AFF proteins small sequence differences do change which genes are controlled: AFF1 and AFF4 reside in separate super elongation complexes with largely non-overlapping gene target sets, and the functional difference between them in Tat transactivation was traced to a single amino-acid variation that alters the affinity each confers on Tat for P-TEFb. A design that expects the partner's canonical elongation function to be the dispensable arm is borrowing an intuition from ELL that does not hold where it has been tested within the AFF family. Lu H et al. 2015
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    A human genome-editing MLL::AF4 model now exists: CRISPR-edited CD34+ haematopoietic cells generate a pro-B acute lymphoblastic leukaemia that recapitulates key cellular and molecular features of infant patients, and both cellular ontogeny and the MLL breakpoint site influence its leukaemogenic capacity.

    A human genome editing-based MLL::AF4 ALL model recapitulates key cellular and molecular leukemogenic features · Bueno C, Torres-Ruiz R, Velasco-Hernandez T, Molina O, Petazzi P, Martinez A, Rodriguez V, Vinyoles M, Cantilena S, Williams O, Vega-Garcia N, Rodriguez-Perales S, Segovia JC, Quintana-Bustamante O, Roy A, Meyer C, Marschalek R, Smith AL, Milne TA, Fraga MF, Tejedor JR, Menéndez P · Blood 142:1752-1756 2023
    PMID 37756522DOI
    What this evidence cannot show

    ABSTRACT ONLY, and the item is a Blood Letter rather than a full paper, so the abstract is two sentences and carries no numbers. Cited here only to establish that the endogenous-level human model this node treats as prohibitively expensive has been built and published; whether it is tractable for a three-arm separation-of-function is a separate question this node cannot answer from the abstract.

    Evidence that argues against this (1)
    • A model that starts from postnatal CD34+ cells cannot test the developmental window in which these fusions demonstrably act. Co-expression of MLL-AF4 and AF4-MLL in human embryonic stem cell differentiation acts at the hemato-endothelial specification step, strongly promoting the emergence of endothelial- and hemogenic-primed precursors with H3K79 methylation specifically enriched over HOX-A cluster genes - a stage that no cord-blood or adult CD34+ editing model contains. The letter cited here says itself that cellular ontogeny influences leukaemogenic capacity; that is a limitation on the model, not only a finding from it. Bueno C et al. 2019
  7. E7established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt annotates no domain feature at all over the AFF1 C-terminal homology domain: P51825 carries only disorder regions (1-45, 73-314, 366-957, 1098-1119) and a functional region at 601-900 required for PARP1 interaction and recruitment to DNA damage sites. The 949-1209 coordinates this experiment uses are therefore inherited from the AFF4 structural literature, not from a curated AFF1 annotation.

    What this evidence cannot show

    Live records fetched and read back on 2026-09-19 (database record; no full text). Primary gene names confirmed as AFF1 and AFF4 respectively, against the synonym-match trap. The absence of a UniProt domain feature does not mean the domain is not real - it plainly is, since it has two crystal structures - but it does mean no independent curation anchors the AFF1 boundaries, and that the disorder annotation reaching to residue 957 overlaps the stated domain start at 949.

    Evidence that argues against this (1)
    • Absence of a UniProt domain feature is not absence of an experimental map, and the map that exists is of a different kind. The AFF4 axis has been mapped by binding rather than by fold: it is a structurally disordered central scaffold that recruits CycT1, ELL2 and ENL/AF9 through short hydrophobic regions distributed along its length, allowing dynamic and flexible assembly. For a protein of that architecture a single 949-1209 HMM window is the wrong unit of description regardless of which database supplies it, and the disorder annotation running to residue 957 is consistent with the experimental picture rather than in tension with it. Chou S et al. 2013
Connected to