Oncofusion

start / graph / interface-ywhaq-aff1-s588-14-3-3

Interfacemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q22-phospho-contacts

14-3-3 theta binds AF4 at Ser588 and rides into the MLL-AF4 complex - the phospho-gated contact the fusion really does carry

also called 14-3-3 theta AF4 · YWHAQ AFF1 S588 · 14-3-3 MLL-AF4

14-3-3 proteins only stick to partners that have been phosphorylated. One of them, 14-3-3 theta, sticks to AF4 at a single spot and is found in a complex with the MLL-AF4 fusion; removing it makes t(4;11) leukaemia cells stop growing and die.

The same thing, for a scientist

Fioretti et al. report that **AF4 and 14-3-3 theta are nuclear interactors, that 14-3-3 theta binds Ser588 of AF4, and that 14-3-3 theta forms a complex with MLL-AF4**. In RS4;11 and MV4-11, 14-3-3 theta knockdown **decreased MLL-AF4 target gene expression, induced apoptosis and hampered proliferation**, while impairing recruitment of AF4 - but **not** of MLL-AF4 - to target chromatin. AFF1 S588 lies far C-terminal to any credible AFF1-side junction (this graph's own record places AFF1 'exon 4' at codons 54-353 or 354-368 depending on numbering), so **S588 is retained in the KMT2A::AFF1 fusion protein**. It is also independently observed as a phosphosite in the PRIDE re-analysis of AFF1. **This, not S821, is where the 14-3-3 story in KMT2A leukaemia actually is** - and it sits on the partner side, which means it is specific to t(4;11) rather than shared by every KMT2A fusion.

Go deeper
The sequence context, and why it is not a canonical motif

AFF1 580-600 reads PGKRSCQKSPAQQEPPQRQTV, putting S588 in a K-S-P context. Scanning the whole of P51825 for the canonical 14-3-3 modes

mode 1  R S X pS X P
mode 2  R X X X pS X P

returns zero matches anywhere in AFF1, so S588 is a non-canonical 14-3-3 site. That is consistent with the fraction of validated 14-3-3 clients that bind through divergent motifs, and it is a direct warning about the method used on KMT2A in cycle 52: a canonical-motif scan would have missed the one 14-3-3 contact in this system that has actually been demonstrated by pulldown.

S588-P589 is a proline-directed site, which is a CDK/MAPK consensus rather than a CK2 or basophilic one. No kinase is assigned. Given that AFF1 sits in a P-TEFb-containing complex, CDK9 is the obvious candidate and is untested.

What the knockdown does and does not show

Knockdown of 14-3-3 theta in t(4;11) cells decreased MLL-AF4 target gene expression, induced apoptosis and reduced proliferation 1. That is a functional dependency in the right cells.

But the same experiment found that 14-3-3 theta knockdown impaired chromatin recruitment of AF4 and not of MLL-AF4. The authors' own interpretation is that 14-3-3 theta triggers the transactivating function and subsequent degradation of AF4, and that MLL-AF4 activity depends on cellular availability of 14-3-3 theta. So the mechanism is argued to run through wild-type AF4, with the fusion's dependence being indirect - the same shape as the CK2-taspase1 route, where the intervention acts on the wild-type partner and the fusion feels it second-hand.

A reader who wants this as a fusion handle needs the experiment the paper does not report: does 14-3-3 theta bind MLL-AF4 directly through the retained S588, and is that binding lost on S588A?

How this explains the KMT2A 14-3-3 proximity result

The graph records KMT2A as prey of six 14-3-3 family members including YWHAQ, reproduced in a second study through YWHAG. Cycle 52 looked for a motif on KMT2A to explain it and found only chance-level matches, none of them phosphorylated.

There is a simpler explanation that the motif scan could not reach: the KMT2A-14-3-3 association is indirect, bridged by an AFF-family scaffold whose 14-3-3 site is known and non-canonical. YWHAQ is the same paralogue in both observations. AFF1 and AFF4 co-occupy chromatin with KMT2A-family complexes in ordinary cells, so a bridged recovery in an interactome or proximity experiment needs no KMT2A motif at all.

That is a hypothesis, stated as one in hyp-kmt2a-14-3-3-association-is-bridged-by-aff-family, and it is falsifiable in a single experiment.

Facts and figures
partner aYWHAQ (14-3-3 theta)
partner bgene-aff1
motif bAFF1 Ser588, non-canonical; sequence context PGKRSCQK-S588-PAQQEPPQRQTV
fate in fusionretained
tested on fusionPartially - 14-3-3 theta forms a complex with MLL-AF4 and knockdown reduces MLL-AF4 target gene expression, but knockdown did not impair MLL-AF4 chromatin recruitment and S588A was not tested on the fusion
canonical motif scanzero mode-1 or mode-2 matches across the whole of P51825
phospho sites
[
  {
    "protein": "AFF1",
    "uniprot": "P51825",
    "residue": "S588",
    "kinase": "unassigned; S-P proline-directed consensus, CDK9 untested",
    "fate": "retained"
  }
]
cell lines
[
  "RS4;11",
  "MV4-11",
  "HEK293"
]
structures
[]
How this could bite us back
  • Knockdown of a 14-3-3 paralogue in a leukaemia line kills the cells, but 14-3-3 proteins are among the most abundant and most pleiotropic scaffolds in the cell; a dependency shown by knockdown is not evidence of a targetable interface.
  • If 14-3-3 binding is chaperone-like, blocking it would promote aggregation of the fusion rather than inactivating it, and an aggregated oncoprotein is not obviously a treated one.
  • The paper's own data place the mechanism on wild-type AF4. A therapy built on it would depend on wild-type AFF1 the way the CK2-taspase1 route depends on wild-type KMT2A, and would be lost by the same class of event.
  • This is a single-laboratory finding in two cell lines. It is the strongest 14-3-3 result in the system and it is also the least replicated thing this beat is resting weight on.
Still open
  1. Does 14-3-3 theta bind MLL-AF4 directly, and is that binding lost on S588A? The paper shows a complex, not a direct retained-side contact, and this is the experiment that would turn it into a fusion handle.
  2. Which kinase writes AFF1 S588? It is a proline-directed site in a P-TEFb-containing complex, which makes CDK9 the obvious and untested candidate.
  3. Is the equivalent residue in AFF4 phosphorylated, and does 14-3-3 bind it? AFF4 carries 53 observed phosphosites and substitutes for AFF1 in the SEC.
  4. Has any group replicated the 14-3-3 theta dependency in t(4;11)? This beat found no independent report.
Every source behind this entry (3)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    AF4 and 14-3-3 theta are nuclear interactors; 14-3-3 theta binds Ser588 of AF4 and forms a complex with MLL-AF4; in t(4;11)-positive RS4;11 and MV4-11 cells, 14-3-3 theta knockdown decreased MLL-AF4 target gene expression, induced apoptosis and hampered proliferation, and impaired chromatin recruitment of AF4 but not of MLL-AF4.

    PMID 31493143DOI
    What this evidence cannot show

    Abstract read 2026-09-19; the Europe PMC record gives PMC12994319 but marks it not open access, and full text was not retrieved, so the binding assay format, whether S588A abolishes binding, and whether the MLL-AF4 complex was shown to be direct are not established here. Two cell lines, both t(4;11). A single-laboratory result with no independent replication found in this beat's searches.

    Evidence that argues against this (2)
    • The paper's own result that 14-3-3 theta knockdown impairs recruitment of AF4 but NOT of MLL-AF4 argues that the fusion is not the direct client. On that reading, 14-3-3 theta is a handle on wild-type AF4 and the fusion's dependence is a consequence of wild-type AF4 biology, which is a much weaker basis for a fusion-directed therapy than the title suggests. Fioretti T et al. 2019
    • The interactome study that makes KMT2A a 14-3-3 client concludes the family's principal role is chaperone-like, preventing client aggregation. If 14-3-3 theta binding to the AF4 moiety is about solubility rather than targeting, occupying that site pharmacologically would promote aggregation of the fusion rather than inhibit it - a different outcome from the one a knockdown phenotype implies. Segal D et al. 2023
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    AFF1 Ser588 is independently observed as a phosphorylation site in the PRIDE/PTMeXchange re-analysis served by the EMBL-EBI Proteins API, which reports 22 phosphosites on P51825; a scan of the full AFF1 sequence for canonical 14-3-3 binding modes 1 and 2 returns zero matches, so S588 is a non-canonical site.

    What this evidence cannot show

    Fetched and scanned 2026-09-19 against the canonical P51825 sequence (1,210 aa). The motif scan uses the two classical modes only and does not cover mode III or the divergent motifs that a substantial share of validated clients use, so 'non-canonical' here means 'not mode 1 or mode 2', not 'not a motif'. Ensembl canonical ENSP00000378578 is 1,218 aa, an eight-residue offset from UniProt numbering, which this graph's AFF1 domain entry already records.

    Evidence that argues against this (1)
    • That a canonical-motif scan returns zero for a protein with a demonstrated 14-3-3 contact is itself the strongest available argument that the same scan's negative result on KMT2A carries little information. The method used to close the KMT2A question fails on the positive control provided by its own partner. Segal D et al. 2023
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    AFF1 Ser588 lies C-terminal to every credible AFF1-side fusion junction and is therefore retained in the KMT2A::AFF1 fusion protein: this graph's AFF1 domain record places the junction at AFF1 'exon 4', spanning codons 54-353 on Ensembl numbering or codons 354-368 on first-coding-exon numbering.

    What this evidence cannot show

    This is a within-graph derivation, not a primary source: the retained/lost call follows from the graph's own recorded junction uncertainty rather than from a sequenced fusion transcript. The graph's own entry states the AFF1-side junction residue is not known to better than about +/-300 aa. S588 sits 220 residues beyond the most C-terminal of the two candidate junctions, which is why the call is safe despite that uncertainty, but a rare breakpoint further into AFF1 would change it.

    Evidence that argues against this (1)
    • The same graph entry records a live dispute about the AFF1 side: its fusion node lists the ALF/NHD domain as RETAINED at both breakpoints while its domain node calls it LOST, because the two use different exon-numbering conventions. A retained/lost call derived from a junction the graph itself disputes should be re-derived from a sequenced patient transcript before anything depends on it. Oncofusion knowledge graph (internal) 2026
Connected to