Oncofusion

start / graph / ana-the-kmt2a-k1130-k1133-numbering-is-human-and-the-immunising-peptide-settles-it-independently

Analysisestablished confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.q165-the-retained-acetylation-site

The K1130/K1133 numbering is HUMAN, stated in the paper's own words, and the immunising peptide proves it a second time without relying on that sentence

also called K1130 species numbering · MLL1 acetylation site human or mouse

A 2015 paper said a protein called MLL1 is switched on and off by a chemical tag at two spots, numbered 1130 and 1133 - but proteins are numbered differently in mice and humans, and most of that paper's experiments were in mice. Reading the full text settles it: the numbers are the human ones, and the exact stretch of protein the authors used to make their detection antibody matches the human sequence letter for letter.

The same thing, for a scientist

The single open question this graph could not close from the abstract was whether K1130/K1133 in PMID 25751424 is human or mouse numbering - the experiments are substantially mouse liver and mouse embryonic fibroblasts, and the paper's own abstract says only 'two conserved residues'. The full text (PMC4732879) states it outright, and a second, independent line of evidence inside the Methods confirms it. Everything this graph has built on the human mapping stands.

Go deeper
The paper says it in one sentence, and nobody had read that sentence

Results, section SIRT1-mediated circadian deacetylation of MLL1:

"While MLL1 acetylation has not been studied previously, mass spectrometry (MS) profiling for global protein acetylome identified four lysine residues as acetylated in MLL1 [ref 35]. These are K636, K1130, K1133 and K1235 in the human isoform of MLL1 [ref 35]."

"in the human isoform of MLL1" is the answer, in the authors' own words 1. It is not an inference from conservation and it is not this graph's mapping - it is the paper's.

The immunising peptide confirms it without using that sentence at all

This matters because a single sentence can be a slip, and because the graph's rule is to verify a residue claim against the UniProt sequence rather than a paper's text. The Methods give the antigen used to raise the site-specific antibody:

"The rabbit polyclonal anti acetyl-lysines K1130 and K1133 from human MLL1 was generated by immunizing rabbits with KHL-conjugates of the peptide NH2-APPIK(ac)PIK(ac)PVTR (Millipore)." 1

Aligning APPIKPIKPVTR to the live UniProt Q03164 sequence (3,969 aa, fetched 2026-09-24) gives an exact, unique match at residues 1126-1137 2:

position  1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137
residue     A    P    P    I    K    P    I    K    P    V    T    R

The two acetyl-lysines in the antigen fall at exactly 1130 and 1133 of the human protein. The peptide is an independent physical object - it was synthesised and injected into rabbits - so this is a check that does not share any failure mode with reading the numbering sentence.

What this closes, and the one thing it does NOT close

🟢 Closed. ana-the-clock-sirt1-acetylation-site-is-at-kmt2a-1130-1133-and-lies-208-residues-inside-the-retained-portion flagged species numbering as its residual risk and said "the human mapping in the table above is mine, from the UniProt sequence, not the paper's". It is now also the paper's. The 208-residue margin to the breakpoint cluster, the 17-residue distance to the CXXC, and the retained-in-every-fusion conclusion all stand unchanged.

🟠 Not closed, and it is a different question. The paper's experiments are still substantially mouse (mouse liver at ZT timepoints, Sirt1-null MEFs). The human numbering tells you which residues the antibody recognises; it does not tell you that the mouse residues the antibody detected in mouse liver are at 1130/1133 of mouse Kmt2a. Mouse KMT2A (UniProt P55200) is 3,714 aa - 255 residues shorter than human - so the mouse positions are almost certainly NOT 1130/1133. The paper does not give them, and this analysis did not attempt to derive them, because the alignment was not run. Any future claim about the mouse site needs its own mapping.

Why the abstract was genuinely ambiguous - this was not a careless read

The abstract says "mammalian MLL1 is acetylated at two conserved residues, K1130 and K1133". "Mammalian" plus "conserved" plus mouse-liver figures is exactly the signature of a paper that numbers in the model organism and asserts conservation. The prior was reasonable; the answer happens to run the other way. Recording this because the graph's standing instinct - if the figures are mouse, suspect mouse numbering - is a good instinct that was wrong here, and a rule that is wrong sometimes is more useful when its failure is written down.

Facts and figures
scriptcomputed inline (Europe PMC fullTextXML + UniProt REST), 2026-09-24
statusrun
methodFetched PMC4732879 full-text XML, stripped tags, sentence-split, and extracted every sentence containing K11xx. Separately fetched Q03164.fasta and located the substring APPIKPIKPVTR.
result summaryNumbering is HUMAN, stated explicitly in the paper and independently confirmed by an exact unique match of the immunising peptide APPIKPIKPVTR to Q03164 residues 1126-1137.
questionIs K1130/K1133 in PMID 25751424 human or mouse numbering?
answerhuman
quoted sentenceThese are K636, K1130, K1133 and K1235 in the human isoform of MLL1 [35].
immunising peptideNH2-APPIK(ac)PIK(ac)PVTR
peptide match q031641126-1137, exact and unique
human uniprotQ03164
human length AA3969
mouse uniprotP55200
mouse length AA3714
mouse positions derivedfalse
resolves open question inana-the-clock-sirt1-acetylation-site-is-at-kmt2a-1130-1133-and-lies-208-residues-inside-the-retained-portion
relationship to existing entriesAnswers the first open question of the king's analysis entry, which explicitly deferred species numbering to this beat. Does not restate that entry's coordinate work; it supplies the missing premise that makes it safe.
inputs
[
  "Europe PMC fullTextXML PMC4732879 (113,427 bytes, HTTP 200, fetched 2026-09-24)",
  "UniProt REST Q03164.fasta (3,969 aa, fetched 2026-09-24)"
]
measurements
[
  {
    "quantity": "peptide match start residue",
    "subject": "APPIKPIKPVTR in Q03164",
    "value": 1126,
    "units": "residue",
    "n": 1,
    "universe": "UniProt Q03164 canonical, 3969 aa",
    "null": null,
    "note": "unique occurrence; the two acetyl-lysines in the antigen fall at 1130 and 1133"
  },
  {
    "quantity": "human minus mouse KMT2A length",
    "subject": "Q03164 vs P55200",
    "value": 255,
    "units": "residues",
    "n": 2,
    "universe": "UniProt canonical sequences",
    "null": null,
    "note": "why mouse numbering would not have been 1130/1133 - the mouse positions were NOT derived here"
  }
]
How this could bite us back
  • Settling the numbering makes the site LOOK safer to build on than it is. The numbering was never the weak link - the weak link is that nothing connects this site to a fusion, and a resolved species question can create false momentum toward an intervention that still has no target validation.
Still open
  1. What are the mouse Kmt2a positions equivalent to human K1130/K1133? The paper's in vivo circadian data are mouse liver, so any attempt to reproduce the in vivo result needs them, and this analysis deliberately did not guess - it requires a Q03164/P55200 alignment that has not been run.
  2. Does the site-specific antibody (Millipore, raised for this paper) still exist as a catalogue item? Every future experiment on this site in a fusion cell either uses it or needs a replacement reagent, and a custom antibody from a 2015 paper is the single most likely thing to have disappeared.
Every source behind this entry (2)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    PMID 25751424 states that K636, K1130, K1133 and K1235 are positions in the HUMAN isoform of MLL1, and its site-specific antibody was raised against the peptide NH2-APPIK(ac)PIK(ac)PVTR described as 'from human MLL1'.

    NAD(+)-SIRT1 control of H3K4 trimethylation through circadian deacetylation of MLL1 · Aguilar-Arnal L, Katada S, Orozco-Solis R, Sassone-Corsi P · Nature Structural & Molecular Biology 22(4):312-8 2015
    PMID 25751424DOI
    What this evidence cannot show

    Read from the Europe PMC full-text XML of the NIHMS author manuscript (PMC4732879), not the typeset Nature version. Author manuscripts occasionally differ from the version of record in wording; the numbering sentence and the Methods antigen sequence are both in the manuscript text and are unlikely to have been altered in proof, but they were not checked against the publisher PDF.

    Evidence that argues against this (1)
    • The paper's in vivo experiments are mouse liver and Sirt1-null MEFs, and the abstract describes the residues only as 'conserved' in 'mammalian MLL1'. A reader of the abstract alone would reasonably infer mouse numbering, and that inference would be wrong. The mouse orthologue (UniProt P55200) is 3,714 aa and its equivalent positions are NOT 1130/1133; this analysis did not derive them.
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    In human UniProt Q03164 (KMT2A, 3,969 aa) the sequence APPIKPIKPVTR occurs at residues 1126-1137, placing lysines at exactly 1130 and 1133; residues 636, 1130, 1133 and 1235 are all lysine.

    What this evidence cannot show

    Verified against the canonical isoform only. KMT2A has annotated isoforms; a residue number in a paper that used a non-canonical construct would not necessarily correspond. The paper's constructs are described as MLL1 truncations from a full-length Flag-MLL1-myc template, consistent with the canonical sequence.

    Evidence that argues against this (1)
    • Two assumptions sit under this that a FASTA cannot test. First, the canonical isoform: if the paper's Flag-MLL1 construct was built on a cDNA differing from Q03164, the numbering could shift. This graph's own exon map records that Ensembl ENST00000534358 translates to 3,972 aa against UniProt's 3,969, and that the two are identical only over residues 1-1603 - so the protection here is a separate fact, not part of this item, and it would not hold for a site past 1603. Second, uniqueness of APPIKPIKPVTR was checked within Q03164 ONLY, not proteome-wide; an occurrence in another protein would mean the site-specific antibody has a cross-reactivity problem nobody has tested.
Connected to