Oncofusion

start / graph / hyp-the-partner-not-the-site-is-what-would-make-a-retained-kmt2a-modification-fusion-specific

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.q165-the-retained-acetylation-site

The partner, not the site, is what could make a retained KMT2A modification fusion-specific - the AML1-ETO mechanism asked of KMT2A::AFF1

also called fusion changes the enzymatic environment of the retained N-terminus

A chemical tag on a part of MLL1 that the fusion keeps is, by itself, nothing special - the normal protein has it too. But in a different leukaemia fusion, the attached partner drags the tagging enzyme into place, so the fusion creates a tag the normal protein does not carry. The question is whether MLL1's partner does the same thing.

The same thing, for a scientist

Retention of K1130/K1133 in KMT2A fusions is at the lysine base rate (10/21 retained acetyl sites, p = 1.00 against a 45.8% lysine null) and the published consequence - tuning SET-domain methyltransferase activity - cannot transfer to a SET-less fusion. Both of this beat's framings for why the site matters therefore fail. The one framing that does not fail is imported from AML1-ETO, where p300 is docked by the NHR1 domain of the 3' ETO partner and acetylates K43 in the 5' RUNX1 portion, so that the fusion manufactures the modification it depends on. KMT2A::AFF1 replaces everything C-terminal to ~1338 with AFF1, bringing AFF4, ELL/ELL2 and P-TEFb into constitutive association with the retained N-terminus. The hypothesis is that this changes the acetyltransferase/deacetylase environment at K1130/K1133, making the occupancy of a shared-sequence site differ between fusion and wild-type KMT2A.

Go deeper
Why the two obvious framings both fail, and this one does not
framingverdict
"the site is retained, so it travels into the fusion"❌ true but uninformative. 47.6% of KMT2A's acetyl-lysines are retained against a 45.8% lysine base rate, exact p = 1.00. Retention is what most of them do.
"acetylation tunes MLL1 methyltransferase activity, so it tunes the fusion"❌ fails twice. The fusion has no SET domain; and this beat found the catalytic consequence was never attached to the residue even in wild-type KMT2A - no site mutant reached the HKMT assay.
"the partner changes which enzymes reach the retained N-terminus, so occupancy at a shared site differs between fusion and wild-type"✅ survives both objections. It does not require the SET domain, and it does not require the site to be special - it requires the environment to be different, which the fusion guarantees.

This is not a new idea, it is a borrowed one. In AML1-ETO, deleting NHR1 (ETO aa 245-436) abolishes p300 binding and K43 acetylation: the modification exists because of the fusion. Nobody has asked the equivalent question of KMT2A.

What follows if it is true, and what follows if it is false

🟢 If true, it supplies the thing this graph keeps looking for and rarely finds: a fusion-specific property of shared sequence. A site present on both proteins but differentially occupied is a discriminating feature that no sequence comparison would ever reveal, and it would generalise beyond this one lysine to every retained modification site.

🔴 If false - if occupancy at K1130/K1133 is the same in fusion and wild-type - then the retained acetylation site is simply a property of KMT2A that the fusion inherits, with no fusion-specific angle, and this beat's subject should be closed rather than extended. That is a real and likely outcome and the hypothesis is written so it can produce it.

🟠 And there is already a published result pointing the wrong way. The same Science 2011 paper that established RUNX1 K43 tested p300 inhibitors in MLL-AF9 and reported no or minimal effect, where the same compounds extended survival in the AE9a model. If the acetyltransferase axis mattered in a KMT2A fusion, that experiment had a reasonable chance of showing it and did not.

Facts and figures
statementKMT2A fusion partners change which acetyltransferases and deacetylases reach the retained KMT2A N-terminus, so acetylation occupancy at K1130/K1133 differs between the fusion protein and wild-type KMT2A in the same cell, despite the two sharing that sequence exactly.
originanalogy
analogy sourceRUNX1::RUNX1T1 - p300 docked by ETO NHR1 acetylates RUNX1 K43 (PMID 21764752)
prior probability0.2
kills the beat if falsetrue
predictions
[
  "Acetyl-K1130/K1133 stoichiometry, measured on immunoprecipitated fusion versus immunoprecipitated wild-type KMT2A from the same KMT2A-rearranged cell, differs by at least 2-fold in either direction.",
  "The acetyltransferase/deacetylase composition of the interactome at the retained KMT2A N-terminus differs between a fusion-positive and a fusion-negative cell.",
  "If occupancy is higher on the fusion, a CBP/p300 inhibitor reduces it preferentially on the fusion; if lower, a SIRT1 inhibitor raises it preferentially on the fusion."
]
experiments
[
  "exp-resolve-acetyl-k1130-occupancy-on-the-fusion-versus-wild-type-kmt2a"
]
analyses
[
  "ana-where-the-kmt2a-acetylation-site-sits-among-four-fusion-families-retained-ptm-precedents",
  "ana-every-ms-evidenced-kmt2a-modification-site-flagged-retained-or-lost-against-the-breakpoint"
]
falsifiers
[
  "Acetyl-K1130/K1133 stoichiometry measured by targeted PRM on fusion-IP and wild-type-IP material from the SAME KMT2A::AFF1 cell line (SEM or RS4;11 for the lymphoid index disease, MV4-11 for myeloid) differs by LESS than 1.5-fold, with 3 biological replicates and a synthetic heavy acetyl-peptide standard. A ratio inside 1.5-fold kills the hypothesis outright, because the hypothesis is entirely about a difference in occupancy at a shared site.",
  "Comparative IP-MS of the retained KMT2A N-terminus in a fusion-positive versus an isogenic fusion-negative cell recovers the SAME set of acetyltransferases and deacetylases (CREBBP, EP300, KAT2A/B, KAT5, SIRT1, HDAC1/2/3) with no member changing by more than 2-fold in spectral counts. If the enzyme environment is unchanged, the proposed mechanism has no substrate.",
  "ALREADY PARTLY RESOLVED IN THE WRONG DIRECTION: p300 inhibitors (Lys-CoA-Tat, C646) had no or minimal effect in MLL-AF9 in PMID 21764752, where the same compounds extended median survival in the AE9a AML1-ETO model. This is not decisive - MLL-AF9 is myeloid and not the index fusion, the compounds are weak and the readout was survival rather than site occupancy - but it is a published result pointing against the hypothesis and it should be quoted whenever the hypothesis is."
]
measurements
[
  {
    "quantity": "prior probability",
    "subject": "this hypothesis",
    "value": 0.2,
    "units": "probability",
    "n": null,
    "universe": "author's prior",
    "null": null,
    "note": "set low deliberately: the mechanism is borrowed from one fusion, acetyltransferases are not documented SEC components, and a published p300-inhibitor negative already exists in a KMT2A fusion"
  }
]
How this could bite us back
  • If occupancy does differ, the intervention still acts on an enzyme (CBP/p300 or SIRT1) with hundreds of substrates, not on the site. A demonstrated difference in occupancy does not produce a tractable target; it produces a biomarker at best.
  • The hypothesis is constructed to survive the two objections that killed the simpler framings. A hypothesis designed around known objections is at risk of being unfalsifiable in practice, which is why the first falsifier is written with a specific ratio, a specific assay and a specific cell line.
Still open
  1. Can fusion and wild-type KMT2A be separated cleanly enough to measure occupancy at a shared site? Fusion-specific immunoprecipitation requires either a partner-directed antibody (AFF1 C-terminal) that will also pull wild-type AFF1, or an epitope-tagged fusion knock-in. The separation problem is the hardest part of testing this and it is not solved.
  2. Do acetyltransferases appear in any published KMT2A-fusion interactome? If the answer is no across every deposited dataset, the hypothesis is weakened before any new experiment - and this graph holds enough interactome entries to check.
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In AML1-ETO, the NHR1 domain of the 3' ETO partner docks p300 and is required for acetylation of K43 in the 5' RUNX1 portion; deleting NHR1 abolishes both p300 binding and K43 acetylation.

    The leukemogenicity of AML1-ETO is dependent on site-specific lysine acetylation · Wang L, Gural A, Sun XJ, et al., Roeder RG, Nimer SD · Science 333(6043):765-9 2011
    PMID 21764752DOI
    What this evidence cannot show

    Read by a worker agent; title and journal verified against Europe PMC by this beat. This establishes the mechanism in ONE fusion. Transplanting it to KMT2A::AFF1 is an analogy with no supporting data in that system, which is why this is a hypothesis node.

    Evidence that argues against this (1)
    • The same paper reports p300 inhibitors had no or minimal effect in MLL-AF9. That is a published negative for the acetyltransferase axis in a KMT2A fusion and it is the single strongest argument against this hypothesis.
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    Retention of acetylation sites in KMT2A is indistinguishable from the lysine base rate: 10 of 21 MS-evidenced acetylation sites are retained (47.6%) against a retained-lysine share of 45.75%, exact two-sided binomial p = 1.00.

    What this evidence cannot show

    n = 21 acetylation sites gives this test very little power; it cannot exclude a real enrichment of moderate size. What it does exclude is the claim that the retention is striking, which is the claim the beat was dispatched on.

    Evidence that argues against this (1)
    • A deeper acetylome could change 10/21 substantially in either direction. The current evidence supports 'not obviously enriched', not 'known to be unenriched'.
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    KMT2A::AFF1 replaces KMT2A sequence C-terminal to approximately residue 1338 with AFF1, which brings AFF4, ELL/ELL2 and P-TEFb into constitutive association with the retained KMT2A N-terminus.

    What this evidence cannot show

    The breakpoint coordinate is verified here from the exon map and UniProt's own SITE breakpoint annotations. The SEC composition statement is standard in this corpus and is not re-derived in this entry.

    Evidence that argues against this (1)
    • Acetyltransferases are not a documented part of the SEC. The hypothesis requires that SEC recruitment changes the local enzyme environment indirectly - through residence time, chromatin context or condensate partitioning - which is a longer causal chain than the direct NHR1-p300 docking that inspired it.
Connected to