Oncofusion

start / graph / ana-the-clock-sirt1-acetylation-site-is-at-kmt2a-1130-1133-and-lies-208-residues-inside-the-retained-portion

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.

The clock/SIRT1-regulated MLL1 acetylation site maps to KMT2A K1130 and K1133 in human numbering, just N-terminal to the CXXC and 208 residues before the breakpoint cluster - so it is retained in every KMT2A fusion

also called K1130 K1133 verified · the retained acetylation site · MLL1 acetylation is inside the fusion · SIRT1 site at 1130/1133

A 2015 paper found that the MLL1 protein is switched on and off by adding and removing acetyl groups at two specific spots, under the control of the body clock and cellular energy levels. Those two spots turn out to sit inside the exact piece of the protein that every leukaemia fusion keeps.

The same thing, for a scientist

🔴 **AMENDED, cycle 157, on four counts - two of them fatal to how this entry framed its own significance (q165, verified by the king).** **1. The species question is CLOSED, in my favour.** PMC4732879 Results says verbatim *"These are K636, K1130, K1133 and K1235 in the human isoform of MLL1"*, and the Methods give the immunising peptide as `NH2-APPIK(ac)PIK(ac)PVTR`, which matches Q03164 uniquely at 1126-1137. The caveat below about "conserved rather than human-mapped" is SUPERSEDED. Mouse P55200 is 3,714 aa, so mouse positions are certainly not 1130/1133. **2. "It affects the methyltransferase activity of MLL1" is NOT supported at the residue level, and this entry quotes it in bold.** A full-text audit finds the HKMT assay (Fig 4c) uses full-length MLL1 with or without SIRT1 - **no site mutant**. The only K-to-R constructs are antibody validation on a fragment and a peptide competition. **No K-to-Q anywhere.** The catalytic consequence was never attached to these residues **even in wild-type KMT2A**. **3. RETENTION IS AT BASE RATE - the central premise fails its own null.** The retained third is 33.7% of the protein by length but carries **45.75% of its lysines**. Acetylation retention is **10 of 21 = 47.6%, exact two-sided p = 1.000**. **"The site is retained" distinguishes nothing.** A length-based null would have made every PTM class look enriched. **4. A regulated retained PTM with a named enzyme and a fusion phenotype ALREADY EXISTS, and this corpus already holds it.** UniProt curates **S136 and S142 as "Phosphoserine; by CK2", ECO:0000269 PubMed 29997176** (Sharma et al., PNAS 2018;115(30):E7053-62), whose abstract states *"elimination of IBM phosphorylation sites on MLL1 disrupts the oncogenic potential of primary MLL1-rearranged leukemic cells"*. **The acetylation site is the UNSTUDIED SIBLING of a PROVEN one**, not the first candidate of its kind - and this entry framed it as the latter. **Also:** UniProt lists K636, K1130 and K1235 in MOD_RES but **NOT K1133**, and K1130 evidence is **19608861 (Choudhary 2009)**, not the 2015 paper - the site predates it. **What survives:** the coordinates, the retention fact itself, and the observation that nobody has asked about the fusion. **What does not:** that retention is remarkable, and that the published consequence attaches to the residue. q160 surfaced PMID **25751424** as a never-cited paper and described it as *"MLL1 acetylated at K1130/K1133, in the portion every fusion retains"*. I repeated that in a king log **without checking the coordinates**, which was careless: **1130 is close enough to the KMT2A breakpoint cluster that the claim could easily have been false.** It is not. Verified end to end [E1]: | | | |---|---| | the paper | Aguilar-Arnal, Katada, Orozco-Solis & Sassone-Corsi, *Nat Struct Mol Biol* 2015;22(4):312-8, PMID 25751424 | | the claim | *"mammalian MLL1 is acetylated at two conserved residues, **K1130 and K1133**"*; cyclic, **controlled by the clock and by SIRT1**, and it **affects the methyltransferase activity of MLL1** | | human sequence | **UniProt Q03164, 3,969 aa** - position 1130 is **K**, position 1133 is **K**; residues 1121-1145 read `DAEPLAPPI`**K**`PI`**K**`PVTRNKAPQEPP` | | CXXC zinc finger | **1147-1195** (exons 5-6) | | leukaemia breakpoint cluster | **~1338-1607** (exons 8-14) | 🟢 **So the site sits 17 residues N-terminal to the CXXC domain and ~208 residues N-terminal to the start of the breakpoint cluster.** It is retained in **every** KMT2A fusion, including KMT2A::AFF1 [E1][E2]. 🔴 **And the obvious objection has to be stated first, because it is strong.** The paper's functional readout is **methyltransferase activity**, and **KMT2A::AFF1 loses the SET domain entirely** (SET is 3829-3945, far distal to the breakpoint). **"Acetylation tunes MLL1 methyltransferase activity" therefore cannot transfer to the fusion.** What transfers is only the *site*, not the published consequence of modifying it [E2]. 🔵 **Why it is still worth a beat.** The site is **CXXC-proximal** - 17 residues from a domain this graph already documents as a DNA-binding module retained in full - so DNA binding, complex assembly and chromatin residence are live alternative consequences that the paper did not test and nobody has tested in a fusion. **A regulated, NAD+-responsive modification inside retained sequence is a candidate determinant the menin interface does not explain**, and unlike this project's occupancy line it does not depend on GSE252378. 🟠 **What is NOT established here, and I am not going to imply it.** Nobody has shown the fusion is acetylated at these residues; nobody has shown the site does anything other than tune SET activity; and the paper's own experiments are substantially **mouse liver**, with the residues described as *conserved* rather than as human-mapped. **The human mapping in the table above is mine, from the UniProt sequence, not the paper's.**

Go deeper
The check I should have run before writing it into a log

🔴 I wrote "residues in the portion every fusion retains" into the cycle-150 king log on the strength of an agent's one-line summary. KMT2A breakpoints cluster at exons 8-14, residues ~1338-1607. A site at 1130 is 208 residues from that boundary - close enough that being wrong was a real possibility, and the statement would then have inverted the whole point: a modification site lost from the fusion is evidence about wild-type MLL1 and nothing else.

🔵 Three things had to hold and all three were checked separately: that the residue numbers are human rather than mouse (fetched Q03164 and read positions 1130 and 1133 - both K); that the breakpoint cluster is where this graph says it is (ana-kmt2a-exon-to-domain-map, ~1338-1607); and that 1130 < 1338. Only the third was obvious.

🟠 The residual risk is the one I could not close from a sequence. The paper's numbering could still be mouse and coincidentally land on two lysines in human - mouse Kmt2a and human KMT2A differ in length, so a shared 1130/1133 is not guaranteed by conservation alone. Finding two lysines at exactly the two stated positions in the human protein makes coincidence unlikely but does not exclude it. The full text settles it and is q165's first task.

Facts and figures
retained in fusiontrue
SET retained in fusionfalse
transfers to fusionthe SITE transfers; the published CONSEQUENCE (methyltransferase activity) does not
correctscycle-150 king log, which asserted the retention without checking the coordinates
paper
{
  "pmid": "25751424",
  "doi": "10.1038/nsmb.2990",
  "journal": "Nat Struct Mol Biol 2015;22(4):312-8",
  "authors": "Aguilar-Arnal L, Katada S, Orozco-Solis R, Sassone-Corsi P"
}
uniprot
{
  "accession": "Q03164",
  "length_aa": 3969,
  "residue_1130": "K",
  "residue_1133": "K",
  "context_1121_1145": "DAEPLAPPIKPIKPVTRNKAPQEPP",
  "fetched": "2026-09-24",
  "http": 200,
  "bytes": 4140
}
coordinates
{
  "CXXC": "1147-1195",
  "breakpoint_cluster": "~1338-1607 (exons 8-14)",
  "SET": "3829-3945",
  "margin_to_breakpoint_start": 208,
  "distance_to_CXXC": 17
}
measurements
[
  {
    "quantity": "margin between the acetylation site and the start of the KMT2A breakpoint cluster",
    "value": 208,
    "unit": "residues",
    "subject": "K1130/K1133 against a breakpoint cluster at ~1338-1607, in UniProt Q03164 numbering",
    "universe": "the site is also 17 residues N-terminal to the CXXC domain (1147-1195); the SET domain (3829-3945) is lost from the fusion"
  }
]
amendment cycle 157
{
  "species": "HUMAN, closed twice: explicit sentence in Results + immunising peptide APPIK(ac)PIK(ac)PVTR matching Q03164 1126-1137",
  "methyltransferase_claim": "NOT residue-level - HKMT assay uses full-length MLL1 with no site mutant; no K-to-Q construct exists",
  "retention_base_rate": {
    "retained_fraction_by_length": 0.337,
    "retained_fraction_of_lysines": 0.4575,
    "acetylation_retained": "10/21 = 47.6%",
    "exact_two_sided_p": 1,
    "verdict": "retention distinguishes nothing"
  },
  "proven_sibling": {
    "sites": "S136, S142",
    "enzyme": "CK2",
    "uniprot_evidence": "ECO:0000269 PubMed 29997176",
    "paper": "Sharma et al., PNAS 2018;115(30):E7053-62, doi 10.1073/pnas.1803909115",
    "phenotype": "elimination of IBM phosphorylation sites on MLL1 disrupts oncogenic potential of primary MLL1-rearranged leukemic cells",
    "already_in_corpus": "the phospho-contacts beat, cited by 15 entries"
  },
  "uniprot_modres": "K636, K1130, K1235 present; K1133 ABSENT; K1130 evidence is 19608861 (Choudhary 2009)",
  "by": "q165, every point re-verified by the king against live UniProt and PubMed",
  "immunising_peptide": "NH2-APPIK(ac)PIK(ac)PVTR, matching Q03164 uniquely at 1126-1137"
}
Still open
  1. Settle the species numbering from the full text (PMC4732879). If K1130/K1133 is mouse numbering, the human site must be re-mapped and every statement here re-checked - this is q165's first task and the one thing a sequence fetch could not close.
  2. Ask whether the site does anything other than tune SET activity. It is 17 residues from the CXXC, so DNA binding, complex assembly and chromatin residence are the obvious candidates, and the fusion retains all of them while losing SET.
  3. Test whether the fusion is acetylated there at all. Nobody has asked; a mass-spectrometry re-analysis of an existing KMT2A-rearranged proteomics deposit might answer it without new wet work.
  4. Enumerate every known modification site in the retained portion with Q03164 coordinates and a retained/lost flag. This one was nearly mis-assigned; the others have never been checked against the breakpoint at all.
Every source behind this entry (2)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.journal-article

    Human KMT2A, UniProt Q03164, is 3,969 aa and carries lysine at both position 1130 and position 1133, with residues 1121-1145 reading DAEPLAPPIKPIKPVTRNKAPQEPP - matching the two conserved acetylated residues K1130 and K1133 reported for mammalian MLL1 by Aguilar-Arnal et al.

    PMID 25751424DOI
    What this evidence cannot show

    The sequence check confirms that human KMT2A has lysines at the two stated positions; it does NOT confirm that the paper used human numbering. The experiments are substantially mouse liver and the abstract says 'mammalian MLL1' and 'conserved residues'. Mouse Kmt2a and human KMT2A differ in length, so identical numbering across species is not guaranteed. Only the abstract was read for this entry; the full text (PMC4732879) was not.

    Evidence that argues against this (1)
    • Finding lysines at two specified positions in a 3,969-residue protein is less improbable than it sounds - lysine is roughly 6% of residues, so two specified positions both being K has a chance probability around 0.4%, small but not negligible, and the positions were not chosen blind. The sequence check raises confidence; it does not by itself establish the numbering is human.
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    In this graph's own exon-to-domain map for KMT2A, the CXXC zinc finger occupies residues 1147-1195 (exons 5-6) and the leukaemia breakpoint cluster region occupies approximately residues 1338-1607 (exons 8-14), placing K1130/K1133 seventeen residues N-terminal to the CXXC and about 208 residues N-terminal to the start of the breakpoint cluster, and therefore within the portion retained in every KMT2A fusion; the SET domain at 3829-3945 is lost from the fusion, so the paper's methyltransferase-activity readout cannot transfer to it.

    What this evidence cannot show

    The breakpoint cluster is given as an approximate range (~1338-1607) because it is an interval over exons 8-14 rather than a fixed boundary, and individual patient breakpoints fall at different points within it - so '208 residues inside the retained portion' is a margin to the START of the cluster, and the true margin for any given fusion is larger. Both coordinate sources are this graph's own entries; neither was re-derived from Ensembl or UniProt independently this session.

    Evidence that argues against this (2)
    • Being retained is necessary but nowhere near sufficient. The retained portion is ~1,300 residues containing many modification sites, and this graph has repeatedly found that a real biochemical property of wild-type KMT2A fails to transfer to the fusion - the KAT8/MOF case, and q160's finding that all 11 KMT2A-H3C1 structural records map to a cassette the fusion deletes. Retention earns a hypothesis, not a finding.
    • The paper's entire functional claim is about methyltransferase activity, which the fusion cannot have. Reading the site as relevant to the fusion requires positing an untested second function for it, and a corpus that has been burned by plausible-mechanism reasoning should treat that as the weakest kind of lead until someone measures acetylation on the fusion itself.
Connected to