Four regulatory elements on AF9, and the fusion keeps exactly one - the phospho-switch that controls SEC binding
also called AF9 regulatory element subtraction · MLLT3 retained regulation · AF9 YEATS Poly-Ser K339 S395
AF9 has four known control knobs, and the KMT2A::MLLT3 cancer protein throws away three of them and keeps one. The one it keeps is the switch that decides whether AF9 grabs the gene-reading machine, which is the opposite of what you would expect if the discarded parts were the important ones.
The same thing, for a scientist
AF9 (P42568) carries four experimentally characterised regulatory elements: the YEATS acyl-lysine reader (aa 1-138), the Poly-Ser oligomerisation domain (aa 149-196), the PCAF-acetylated K339 that controls oligomerisation and TFIID binding, and the DNA-PKcs-phosphorylated S395 that controls SEC binding. About 92% of MLL-AF9 breakpoints fall at or after residue 376. Three of the four elements therefore lie in discarded sequence and **only S395 is retained**. This is a sharper subtraction than the YEATS story this graph already records, and it points the other way: what the fusion keeps is not simply the AHD docking surface but a kinase-controlled input to that surface. It also explains an otherwise odd fact - an AF9 fragment spanning 270-568, approximately the fusion-retained portion, **loses the ability to self-associate** in the primary report, because the Poly-Ser oligomerisation domain is not in it.
The four elements and their fate
| element | residues | function | fate in KMT2A::MLLT3 |
|---|---|---|---|
| YEATS domain | 1-138 | reads H3K9ac; required for TFIID interaction together with Poly-Ser; recruits DOT1L to chromatin in the wild-type protein | lost |
| Poly-Ser oligomerisation domain | 149-196 | drives AF9 self-association; a 270-568 fragment lacking it cannot self-associate; required for TFIID binding | lost |
| K339 (PCAF acetylation) | 339 | acetylation monomerises AF9 and reduces TFIID binding; K339Q mimic fails to restore target-gene expression | lost |
| S395 (DNA-PKcs phosphorylation) | 395 | phosphorylation reduces SEC binding (CDK9, ELL); S395A preserves SEC binding after damage | retained |
Element boundaries and functions from the primary report 1; breakpoint boundary from the review this graph already uses 2.
The Poly-Ser boundary is worth stating carefully. The authors describe the deletion fragment that retains it as 149-196 aa; a direct scan of P42568 finds an uninterrupted run of 42 serines at 149-190, and a second shorter run at 383-391 3. The functional call and the sequence agree.
Why this is not the YEATS story again
This graph's subtraction-mllt3-yeats entry records that AF9 discards its reader and keeps the AHD, and judges the causal question open because no in-cis add-back has been done. That framing treats the retained half as a passive docking module.
The present analysis says something different about the same breakpoint: the retained half contains a regulated element, and the discarded half contains the constitutive machinery. Oligomerisation, reader function and the acetylation brake all go; the kinase input to SEC binding stays.
A consequence worth stating plainly: the fusion cannot oligomerise through AF9. Whatever multivalency KMT2A::MLLT3 has must come from the KMT2A half or from its partners, not from the Poly-Ser domain that makes wild-type AF9 self-associate. Condensate and hub models of this fusion that invoke AF9-driven multivalency are invoking a domain the fusion does not contain.
The one number that could overturn this
Everything here hangs on the 376 boundary. If a substantial minority of KMT2A::MLLT3 patients break N-terminal of 339, K339 would be retained and the acetylation arm would be in play too; if a minority break C-terminal of 395, those patients' fusions would have no regulatory element at all. The 92% figure comes from a pooled review distribution, not from a residue-resolved primary series, and this graph has no entry that maps MLLT3 breakpoints to residue in a patient cohort.
That is the measurement that would make this analysis either solid or wrong, and it is a retrospective bioinformatic analysis of existing fusion-transcript data, not a new experiment.
| status | run |
|---|---|
| breakpoint boundary AA | 376 |
| breakpoint patient fraction PCT | 92 |
| elements total | 4 |
| elements retained | 1 |
| elements lost | 3 |
| uniprot | P42568 |
| result summary | Three of four characterised AF9 regulatory elements are discarded by the fusion and only the S395 phospho-switch is retained. The fusion therefore cannot oligomerise through AF9's Poly-Ser domain, which an AF9 270-568 fragment is shown to be unable to do. |
| gene symbol | MLLT3 |
elements — the numbers behind the chart
[
{
"name": "YEATS domain",
"start": 1,
"end": 138,
"fate": "lost",
"function": "acyl-lysine reader; with Poly-Ser required for TFIID interaction"
},
{
"name": "Poly-Ser oligomerisation domain",
"start": 149,
"end": 196,
"fate": "lost",
"function": "self-association; required for TFIID interaction",
"sequence_note": "uninterrupted 42-serine run at 149-190 in P42568"
},
{
"name": "K339 acetylation switch",
"start": 339,
"end": 339,
"fate": "lost",
"function": "PCAF acetylation monomerises AF9 and reduces TFIID binding"
},
{
"name": "S395 phospho-switch",
"start": 395,
"end": 395,
"fate": "retained",
"function": "DNA-PKcs phosphorylation reduces SEC (CDK9, ELL) binding"
}
]- A YEATS-domain inhibitor is the obvious pharmacological handle on this family, and this analysis says the AF9 YEATS domain is not in the fusion at all. A YEATS inhibitor in KMT2A::MLLT3 disease would be acting on wild-type AF9 and on ENL, not on the oncoprotein.
- If the retained element is a kinase-controlled off-switch, then the pharmacological direction that would disable the fusion is to ACTIVATE a DNA-damage kinase. That is genotoxic therapy in children, with second malignancy as the price.
- Any programme that reasons from AF9's biology to the fusion's biology inherits three domains the fusion does not have. That includes the oligomerisation-based condensate arguments this graph holds elsewhere.
- Map MLLT3 breakpoints to residue in a real patient cohort. Everything here depends on a pooled 92% figure taken from a review.
- If KMT2A::MLLT3 cannot self-associate through Poly-Ser, what supplies multivalency for the hubs and condensates the field attributes to this fusion? The KMT2A half, AFF1/AFF4, or nothing?
- Is K339 acetylation retained in the minority of patients whose breakpoint falls N-terminal of 339, and do those patients behave differently?
- Does re-supplying the Poly-Ser domain in cis to KMT2A::MLLT3 change transformation? That is the same in-cis add-back design this graph already specifies for the YEATS domain.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
AF9 contains an N-terminal YEATS domain, a Poly-Ser domain at 149-196 whose deletion abolishes self-association (an AF9 270-568 fragment lacking it cannot self-associate), a PCAF-acetylated K339 whose acetylation reduces oligomerisation and TFIID binding, and a DNA-PKcs-phosphorylated S395 whose phosphorylation reduces SEC binding; both Poly-Ser and YEATS are required for TFIID interaction but not for SEC interaction.
Post-translational modification-dependent oligomerization switch in regulation of global transcription and DNA damage repair during genotoxic stress. · Talukdar P, et al. · Nat Commun 2024What this evidence cannot show
Full text read via Europe PMC fullTextXML (PMC11096357) on 2026-09-19. All mapping is by deletion and point mutants of ectopically expressed AF9 in HEK293T cells, with endogenous confirmation for a subset; many individual panels are n=1 or n=2 as stated in the legends. The Poly-Ser 149-196 boundary is the boundary of a deletion construct, not of a folded domain, and the authors note Poly-Ser is not predicted to be structured. No KMT2A fusion is assayed.
Evidence that argues against this (1)
- A structural review of MLLT3 and MLLT1 from an independent laboratory attributes AF9's functional repertoire to the intrinsically disordered AHD undergoing coupled binding and folding with AF4, DOT1L, BCOR and CBX8, and does not treat Poly-Ser oligomerisation or K339 acetylation as part of the regulatory account. The four-element framing used here is largely one laboratory's. Kabra A et al. 2022
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Approximately 92% of MLL-AF9 patients have breakpoints at or after AF9 residue 376, so the YEATS domain, the Poly-Ser domain and K339 lie in discarded sequence while S395 lies in retained sequence.
The YEATS domain epigenetic reader proteins ENL and AF9 and their therapeutic value in leukemia. · Hu H, et al. · Exp Hematol 2023What this evidence cannot show
A review cited for a pooled distribution. There is no residue-resolved primary breakpoint series behind this figure in this graph, and the whole analysis is sensitive to it: a minority breaking N-terminal of 339 would retain the acetylation arm.
Evidence that argues against this (1)
- The counterpart partner falsifies any general rule drawn from this: about 84.1% of MLL-ENL breakpoints fall at or before ENL residue 5, so the paralogue retains everything AF9 discards. Whether an element is retained is a fact about the breakpoint distribution of that partner, not about the element. Hu H et al. 2023
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
The AF9 canonical sequence contains an uninterrupted run of 42 serines at residues 149-190 and a second run of 9 serines at 383-391, consistent with the 149-196 Poly-Ser boundary the primary report uses.
UniProtKB P42568 (AF9_HUMAN) canonical FASTA, scanned for S-runs of length 4 or more, 2026-09-19 · UniProt Consortium 2026What this evidence cannot show
A sequence computation done in this session. It corroborates the construct boundary but does not independently establish the domain's function.
Evidence that argues against this (1)
- Six of AF9's nine union phosphosites (288, 294, 301, 309, 321, 364) sit between the Poly-Ser run and the breakpoint boundary and are all lost, so the region the fusion discards is also where nearly all of AF9's observed phosphorylation occurs. That is an equally good reading of the same coordinates and it says the fusion is phospho-impoverished rather than phospho-selective. EMBL-EBI 2026
a fifth discarded element: the only annotated AF9 NLS
- related to →KMT2A::MLLT3 (MLL-AF9), t(9;11)(p21.3;q23.3)
three of four AF9 regulatory elements discarded
extends the YEATS subtraction to three further elements
S395 is the one retained element
element-level map of AF9
the lost arm in the four-element map
per-element retention analysis