Oncofusion

start / graph / mechanism-dnapk-af9-s395-sec-release

Mechanismmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q25-partner-phospho

DNA-PKcs phosphorylates AF9 at S395 and releases it from the super elongation complex - the only AF9 regulatory element the fusion keeps

also called AF9 S395 · MLLT3 Ser395 · DNA-PK AF9 phosphorylation · AF9 SEC release

When a cell's DNA is damaged, a repair kinase puts a phosphate on one exact spot of AF9 and that makes AF9 let go of the machine that drives genes to be read out. That spot is inside the piece of AF9 the cancer fusion protein keeps, so the cancer protein probably has the same off-switch and nobody has checked.

The same thing, for a scientist

During genotoxic stress DNA-PKcs phosphorylates AF9 (MLLT3) at Serine 395, and phospho-S395 AF9 loses interaction with super elongation complex components including CDK9 and ELL, contributing to global transcriptional downregulation. The assignment rests on three convergent lines in the primary report: IR-induced phospho-S/T-Q signal on AF9 that is abolished by the DNA-PK inhibitor NU7441 at 2 uM; an S395A mutant with defective phosphorylation and unaltered SEC interaction after IR; and an S395D phosphomimic with reduced TFIID and SEC binding. **S395 is the single S/T-Q motif in all 568 residues of AF9**, and it lies C-terminal of residue 376, the breakpoint boundary for roughly 92% of MLL-AF9 patients. It is therefore on **retained** sequence. That matters because SEC engagement through the AHD is the accepted oncogenic function of KMT2A::MLLT3, and this is a kinase-controlled switch that turns that engagement off.

Go deeper
The switch as reported, in HEK293T cells
  1. Ionising radiation induces phospho-S/T-Q signal on AF9 1.
  2. AF9 interacts with endogenous DNA-PKcs; NU7441 (2 uM) abolishes the IR-induced phosphorylation 1.
  3. Phosphorylated AF9 shows reduced interaction with SEC components, explicitly CDK9 and ELL 1.
  4. An S395A mutant is phospho-defective and its SEC interaction is unaltered by IR; an S395D phosphomimic shows reduced SEC and TFIID binding 1.
  5. Cells expressing S395A show enhanced global nascent transcription after IR relative to wild-type - the phospho-site is doing the shutting down 1.

The paper's larger architecture is a two-modification switch: PCAF acetylates AF9 K339 to break Poly-Ser-mediated oligomerisation and TFIID binding, and DNA-PKcs phosphorylates S395 to break SEC binding. The two arms have opposite fates in the fusion - see ana-af9-regulatory-elements-lost-and-kept.

Retained or lost: retained, and this is the point

About 92% of MLL-AF9 patients break at or after AF9 residue 376 2. S395 is 19 residues C-terminal of that boundary, so in the great majority of KMT2A::MLLT3 patients the fusion protein contains S395.

That makes this the first kinase-gated contact this graph has recorded that sits on the partner's retained half. The previous phospho beat's central result, CK2 at KMT2A S136/S142, is on the retained half of KMT2A; this is the mirror image on the other side of the junction.

The contrast within AF9 itself is sharp. Of AF9's four characterised regulatory elements - the YEATS acyl-lysine reader (aa 1-138), the Poly-Ser oligomerisation domain (aa 149-196), the K339 acetylation switch, and S395 - only S395 is retained. The fusion discards the reader, the oligomerisation module and the acetylation switch, and keeps the phospho-switch that governs SEC binding, which is the one function the fusion actually needs.

The prediction nobody has tested, and why it is clinically loaded

KMT2A-rearranged patients receive DNA-damaging induction chemotherapy: anthracyclines, etoposide, cytarabine. If the fusion carries S395 and DNA-PKcs phosphorylates it, then induction chemotherapy should transiently phosphorylate KMT2A::MLLT3 and detach it from the SEC, suppressing the fusion's own output programme at the same time as it damages DNA.

And the reverse. DNA-PK inhibitors are in oncology trials as chemo- and radio-sensitisers. In a KMT2A::MLLT3 blast a DNA-PK inhibitor would block the release and keep the fusion engaged with the SEC during chemotherapy. The sensitising rationale and the fusion-specific effect point in opposite directions. This is a testable, cheap, and unasked question.

Neither direction has been measured. No paper in this beat's searches expresses a KMT2A fusion and asks what a PIKK does to it.

Facts and figures
kinasegene-prkdc
kinase nameDNA-PKcs (PRKDC)
substrategene-mllt3
uniprotP42568
siteS395
motifS/T-Q (PIKK consensus); the only such motif in AF9
gated contactsuper elongation complex binding (CDK9, ELL) and TFIID binding
directionphosphorylation REDUCES SEC interaction
mapped inwild-type AF9, HEK293T
mapped in fusionfalse
retained in fusiontrue
retained basisS395 is C-terminal of AF9 residue 376; about 92% of MLL-AF9 breakpoints fall at or after 376
retained call confidencestrong for the sequence fact, untested for the fusion protein
observed in peptideatlasfalse
observed in uniprot curatedfalse
inhibitor usedNU7441 2 uM
systemHEK293T, ionising radiation 10 Gy
gene symbolMLLT3
How this could bite us back
  • A DNA-PK inhibitor given as a chemosensitiser would, on this mechanism, block the phosphorylation that detaches the fusion from the SEC - preserving the oncogenic elongation complex at the exact moment chemotherapy is trying to kill the cell. If real, this is an antagonism between two drugs that would plausibly be combined.
  • If the reverse is true and chemotherapy already phospho-inactivates the fusion, then part of the benefit attributed to cytotoxicity is really transient target inactivation, and a menin inhibitor given concurrently would be acting on a fusion that is already partly disengaged. Combination scheduling would matter in a way nobody is modelling.
  • Pushing S395 phosphorylation deliberately - the apparent therapeutic direction - means activating a DNA-damage kinase in a paediatric patient, which is a genotoxic strategy in a population whose main long-term problem is second malignancy.
  • The whole mechanism is one laboratory, one cell line, no fusion protein, and several single-replicate panels. Building anything on it without an independent replication in a leukaemic cell would repeat a pattern this project criticises elsewhere.
Still open
  1. Is KMT2A::MLLT3 phosphorylated at S395 in a t(9;11) leukaemia cell after an anthracycline or ionising radiation? One phospho-S/T-Q blot on an immunoprecipitated fusion protein answers this.
  2. Does phospho-S395 reduce the fusion's binding to AFF1/AFF4 and CDK9 as it does for wild-type AF9, or does the KMT2A half hold the complex together through menin and LEDGF regardless?
  3. Would an S395A knock-in in a KMT2A::MLLT3 model make the leukaemia resistant to genotoxic chemotherapy, by removing the fusion's chemo-induced off-switch?
  4. Do DNA-PK inhibitors, now in clinical trials as chemosensitisers, antagonise chemotherapy specifically in KMT2A::MLLT3 disease?
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    DNA-PKcs-mediated phosphorylation of AF9 at Serine 395 during genotoxic stress reduces AF9 interaction with super elongation complex components including CDK9 and ELL; NU7441 abolishes the IR-induced phosphorylation, an S395A mutant shows defective phosphorylation with unaltered SEC interaction and enhanced global nascent transcription after IR, and an S395D phosphomimic shows reduced SEC and TFIID interaction.

    PMID 38750015DOI
    What this evidence cannot show

    Full text read via Europe PMC fullTextXML (PMC11096357) on 2026-09-19. Entirely HEK293T; several panels underlying the S395 arm are stated by the authors as n=1 or n=2 replicates. Ectopic AF9 is used for most interaction work with endogenous confirmation for a subset. No haematopoietic cell, no KMT2A fusion protein, and no patient material anywhere in the study. The S395D phosphomimic reduces both TFIID and SEC binding, which complicates the clean separation the model proposes between the acetylation arm (TFIID) and the phosphorylation arm (SEC).

    Evidence that argues against this (1)
    • Every experiment in the primary report is on wild-type AF9 in HEK293T cells. The KMT2A::MLLT3 fusion protein is never expressed, never immunoprecipitated and never blotted, so the claim that the fusion carries this switch is a sequence inference from the breakpoint distribution, not a measurement. This graph's own subtraction entry for MLLT3 makes the parallel point about the YEATS domain: the sequence fact is firm and the causal claim is not. Hu H et al. 2023
  2. 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 sequence C-terminal of 376 - including S395 - is retained in the fusion protein.

    PMID 37295550DOI
    What this evidence cannot show

    A review cited for a pooled breakpoint distribution, read at abstract level plus this graph's subtraction-mllt3-yeats entry which uses the same figure. A distribution is not a guarantee for an individual patient, and a minority of breaks fall elsewhere.

    Evidence that argues against this (1)
    • Breakpoint position within the partner is one of the parameters clinical studies of KMT2A-rearranged disease have generally not modelled, so the functional consequence of falling outside the 92% is unmeasured rather than shown to be absent. An untested variable is not a null one. Kabra A et al. 2022
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    S395 is the only S/T-Q motif in the entire 568-residue AF9 sequence, so the PIKK-family kinase that acts on AF9 has no alternative site to choose within the protein.

    What this evidence cannot show

    A sequence computation done in this session. Motif uniqueness constrains which residue a PIKK can hit but says nothing about whether a PIKK hits the protein at all, and non-canonical PIKK sites exist.

    Evidence that argues against this (1)
    • S395 appears in neither the PeptideAtlas/PTMeXchange re-analysis for P42568 (7 phosphosites, none at 395) nor UniProt's curated table (5 phosphosites, none at 395). A functionally decisive site can be entirely absent from steady-state phosphoproteomics, which is a caution against this beat's own atlas-based reasoning as much as against this claim. EMBL-EBI 2026
Connected to