Oncofusion

start / graph / ana-where-the-kmt2a-acetylation-site-sits-among-four-fusion-families-retained-ptm-precedents

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.q165-the-retained-acetylation-site

Ranked against four fusion families, the KMT2A acetylation site is the only candidate with no evidence on the fusion at all - and the strongest argument for it is one nobody has made

also called retained PTM precedent comparison · where K1130 sits

We compared the MLL1 tag site with the equivalent sites in four other cancer fusion proteins. Two of them have been taken all the way - one is the mechanism of a curative drug. The MLL1 site is the only one where nobody has even checked whether the fusion protein carries the tag.

The same thing, for a scientist

Five retained-portion PTM candidates across five fusion families, scored on the same six criteria. PML::RARA K160 SUMO and RUNX1::RUNX1T1 RUNX1 K43 acetylation are complete: site mapped, retention established, modification shown on the fusion, point mutant tested for a fusion phenotype, enzyme identified, and in PML's case a drug. EWSR1::FLI1 has a domain-level O-GlcNAc with no residue. NUP98 has nothing in its retained portion at all. KMT2A K1130/K1133 has the site and the retention and nothing else - and its single best argument, that a partner could change which enzymes reach the retained N-terminus, is the mechanism RUNX1::RUNX1T1 actually demonstrates and that nobody has proposed for KMT2A.

Go deeper
The scorecard
PML::RARA<br>PML K160 SUMORUNX1::RUNX1T1<br>RUNX1 K43 acKMT2A::X<br>K1130/K1133 acEWSR1::FLI1<br>EWS TAD O-GlcNAcNUP98::X
residue mapped✅✅✅❌ domain only—
in the retained portion✅ both breakpoints✅✅ 208 aa margin✅no site exists
in the 5' partner✅✅✅✅—
modification shown ON the fusion✅✅ patients + mice❌ never attempted✅ antibody, no residue—
point mutant → fusion phenotype✅ K160R✅ K43R, 145→328 d❌❌—
enzyme identified✅ PIAS1/RNF4✅ p300, docked by ETO NHR1❌ eraser SIRT1 only; writer unassigned❌ OGT presumed—
therapeutic link✅ arsenic⚠️ p300 inhibitors, +18 d❌❌—

Every one of the five sits in the 5' partner. With n = 4 informative families that is not a finding, but it is a consistent pattern and worth stating: in these fusions the regulated, retained PTM belongs to the gene that donates the N-terminus.

🔵 The argument for the KMT2A site that nobody has made

This beat was dispatched on the premise that retention is what makes K1130/K1133 interesting. It is not, and this comparison is what shows it.

  • Retention is at base rate. 47.6% of KMT2A's MS-evidenced acetyl-lysines are retained, against a lysine base rate of 45.8% (p = 1.00). The site is retained the way most of KMT2A's lysines are retained.
  • And the published consequence does not transfer. The fusion has no SET domain, and this beat further found that the catalytic consequence was never attached to the residue even in wild-type KMT2A.

The argument that does work is the RUNX1::RUNX1T1 argument, transplanted. In AML1-ETO, K43 is a RUNX1 residue and the acetyltransferase is recruited by NHR1 in the ETO partner. Delete NHR1 and K43 is not acetylated. The fusion manufactures the modification it depends on, by putting a writer next to a site that the wild-type protein does not normally present to it.

That mechanism is exactly available to KMT2A::AFF1. The fusion replaces everything C-terminal to ~1338 with AFF1, which brings AFF4, ELL, ELL2 and P-TEFb into a constitutive complex on the retained KMT2A N-terminus. Whether that changes which acetyltransferases and deacetylases reach K1130/K1133 - 17 residues from the CXXC - has never been asked. It is a question about the fusion, not about the site, and it does not require the SET domain to mean anything.

🟠 The honest counterweight: the same Science paper that established K43 tested p300 inhibitors in MLL-AF9 and reported no or minimal effect. A published negative for the acetyltransferase axis in a KMT2A fusion already exists and must be read before anyone proposes a p300-directed experiment here.

And the half of the question this comparison exposes

Four families were examined for what the fusion keeps. The sharper results came from what it loses:

familywhat the fusion subtracts
EWSR1::FLI120+ methylarginines. Wild-type EWSR1 is one of the most heavily arginine-methylated proteins in the nucleus; the fusion carries none of them
PML::RARAthe SIM and the exon-7a CK2 phosphorylation sites, in every form
KMT2A::AFF1all four SUMOylation sites, including K2528 - the single best-supported PTM site on the protein - plus ~1,100 residues of the wild-type N320 fragment
NUP98::Xthe entire mitotic CDK1/NEK6 phospho-regulation of NUP98

Every family loses a coherent regulatory layer, and in three of four it is a different layer from the one that was kept. A programme built only on retained sites is looking at half the object.

Facts and figures
scriptsynthesis over this beat's five family entries
statusrun
families compared5
families informative4
kmt2a criteria met3
kmt2a criteria total7
all five sites in 5prime partnertrue
best untested argument for kmt2athe AML1-ETO mechanism transplanted: the 3' partner recruits the writer to a site in the retained 5' portion, so a retained site can be fusion-specific through its enzymatic environment rather than through its sequence. KMT2A::AFF1 brings AFF1/AFF4/ELL/P-TEFb to the retained N-terminus and nobody has asked whether that changes which acetyltransferases reach K1130/K1133.
published negative againstp300 inhibitors had no or minimal effect in MLL-AF9 in the same Science 2011 paper that established K43
criteria
[
  "residue mapped",
  "in the retained portion",
  "in the 5' partner",
  "modification shown ON the fusion",
  "point mutant tested for a fusion phenotype",
  "enzyme identified",
  "therapeutic link"
]
complete precedents
[
  "PML::RARA PML K160 SUMO",
  "RUNX1::RUNX1T1 RUNX1 K43 acetylation"
]
what each family loses
{
  "EWSR1::FLI1": "20+ methylarginines - the entire PRMT-written layer",
  "PML::RARA": "the SIM and the exon-7a CK2 phosphorylation sites, in every form",
  "KMT2A::AFF1": "all four SUMOylation sites including K2528, plus ~1,100 residues of the wild-type N320 fragment",
  "NUP98::X": "all mitotic CDK1 and NEK6 phosphorylation"
}
measurements
[
  {
    "quantity": "of seven criteria met by the KMT2A acetylation site",
    "subject": "KMT2A K1130/K1133 against the beat's scorecard",
    "value": 3,
    "units": "criteria",
    "n": 7,
    "universe": "the seven criteria satisfied by PML K160",
    "null": null,
    "note": "mapped, retained, 5' partner. Not met: shown on the fusion, mutant phenotype, enzyme, therapeutic link."
  },
  {
    "quantity": "of five families whose retained PTM site is in the 5' partner",
    "subject": "PML::RARA, RUNX1::RUNX1T1, KMT2A::X, EWSR1::FLI1, NUP98::X",
    "value": 4,
    "units": "families",
    "n": 4,
    "universe": "the 4 families with any candidate site; NUP98 has none",
    "null": null,
    "note": "4 of 4, but n=4 and the families were chosen because they are well studied - this is a pattern to notice, not a result"
  }
]
How this could bite us back
  • A scorecard that puts the site of interest last is easy to read as 'the work has not been done yet, so do it'. The other reading is that four other groups looked at analogous sites and the two that paid off had properties - a single-copy dependency, a clean in vivo model, an identified writer - that KMT2A K1130 does not have.
  • The transplanted AML1-ETO mechanism points at p300/CBP inhibitors, and the paper that supplies the mechanism also supplies the negative result for that drug class in a KMT2A fusion. Anyone who cites the first half without the second is building on a paper that already contradicts them.
Still open
  1. Does KMT2A::AFF1 change which acetyltransferases and deacetylases reach the retained KMT2A N-terminus, relative to wild-type KMT2A? This is the AML1-ETO mechanism asked of this graph's index fusion, it does not depend on the SET domain, and it is the strongest untested argument this beat found. A comparative IP-MS of the interactomes at the retained N-terminus in a fusion-positive versus fusion-negative cell would address it.
  2. Is there a fusion family where a retained-PTM programme was attempted and failed? The two successes here are famous and the three blanks are unstudied; the missing denominator is what would tell anyone whether this is a productive line of work or two lucky cases.
  3. Should this graph systematically enumerate what each fusion LOSES rather than what it keeps? Three of four families here have a sharper subtraction than retention story, and for KMT2A the subtraction (all four SUMO sites, and the ~1,100 residues between the cluster and taspase1 CS2) is not written down anywhere else in the corpus.
Every source behind this entry (2)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    PML::RARA K160 SUMOylation and RUNX1::RUNX1T1 RUNX1 K43 acetylation each satisfy all six criteria (mapped, retained, shown on the fusion, point mutant tested for a fusion phenotype, enzyme identified, and for PML a therapeutic link), while KMT2A K1130/K1133 satisfies only mapping and retention.

    What this evidence cannot show

    A scorecard across five families assembled in one beat, with three of the five families researched by a worker agent whose residue claims were independently re-verified against UniProt but whose full-text reads were not. The six criteria were chosen by this beat and are not a standard; a different set would rank the families differently. Four informative families is far too few to support any statement about fusions in general, including the 5'-partner observation.

    Evidence that argues against this (1)
    • Selection bias runs strongly through this comparison. PML::RARA and RUNX1::RUNX1T1 were chosen precisely because they are the known success cases; the denominator - fusion families where a retained-PTM programme was attempted and failed, or was never attempted - is not observable. A scorecard over two famous successes and three blanks cannot say how often this line of work pays off.
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In AML1-ETO the acetyltransferase p300 is docked by the NHR1 domain of the 3' ETO partner and acetylates K43 in the 5' RUNX1 portion; deleting NHR1 abolishes both p300 binding and K43 acetylation, so the fusion creates the modification it depends on.

    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; the title and journal were verified against Europe PMC by this beat. The NHR1-deletion result establishes that NHR1 is required, not that it is sufficient or that the equivalent applies in any other fusion. Transplanting this mechanism to KMT2A::AFF1 is an analogy and is labelled as one.

    Evidence that argues against this (1)
    • The same paper reports that p300 inhibitors had no or minimal effect in MLL-AF9. That is a direct published negative for an acetyltransferase-directed intervention in a KMT2A fusion, and it argues against the transplanted mechanism before it is tested.
Connected to