Oncofusion

start / graph / ptm-kmt2a-k2528-sumoylation

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

KMT2A K2528 SUMO2 - the single best-supported modification site on the whole protein, in the N320 fragment, and LOST in every fusion

The most reliably detected chemical tag on the entire MLL1 protein is not one of the acetylation spots at all - it is a SUMO tag two-thirds of the way along. Every leukaemia fusion throws it away, even though the fusion is often described as behaving like the front half of MLL1.

The same thing, for a scientist

Q03164 K2528 carries a SUMO2 isopeptide cross-link (UniProt CROSSLNK, ECO:0007744 PubMed 28112733) and is reported in EIGHT independent PRIDE deposits - more than any other modification site on KMT2A, and eight times the support of K1130. It lies at 2528, which is inside the taspase1 N320 fragment (chain 1..2718) but far distal to the breakpoint cluster, so it is present on wild-type KMT2A's N-terminal fragment and ABSENT from every KMT2A fusion protein. This is a sharp and, as far as this graph can tell, unremarked asymmetry: the fusion is routinely reasoned about as an N320-like species, and it is missing the best-evidenced PTM that N320 carries.

Go deeper
The asymmetry, stated plainly

This graph reasons constantly about the taspase1 products N320 (aa 1-2718) and C180 (aa 2719-3969), and about KMT2A fusions as species that resemble N320 with a partner attached. The UniProt chain boundaries confirm those coordinates 2.

K2528 is in N320 and is not in any fusion. Between the breakpoint cluster (ends ~1607) and the taspase1 CS2 site (2718) there are ~1,100 residues that wild-type N320 has and the fusion does not - and that stretch contains the best-supported modification site on the protein.

wild-type KMT2A N320KMT2A::AFF1
aa 1-1337✅✅
aa 1338-2718, incl. K2528 SUMO2✅❌
taspase1 cleavage at 2666/2718✅❌ (no CS1/CS2 in the fusion)

So "the fusion is N320 plus a partner" is wrong by about 1,100 residues, and one of the things inside those 1,100 residues is the protein's most reproducibly observed PTM.

How strong is the site, and how strong is the inference

🟢 The site is strong. Eight PRIDE deposits (PXD004927, PXD005296, PXD006545, PXD008003, PXD011932, PXD023841, PXD028050, PXD034631) plus a UniProt CROSSLNK annotation with an MS evidence code. Nothing else on KMT2A comes close - the next best is acetyl-K2958 at four 1.

🔴 The inference is weak. Best-supported in large-scale MS means most detectable, which is a property of the peptide and the enrichment chemistry as much as of the biology. SUMO-remnant enrichment methods have strong sequence preferences, and a site that many SUMO studies find may simply sit in a favourable tryptic context. No functional consequence of KMT2A K2528 SUMOylation is documented anywhere, and this entry makes no claim that one exists. What is solid is the coordinate and the retention call.

Facts and figures
uniprotQ03164
position2528
residueK
modificationSUMO2 isopeptide cross-link
fusion statusLOST
in taspase1 fragmentN320 (chain 1..2718)
uniprot curatedtrue
functional consequencenone documented anywhere
shown for fusionN/A - the residue is not in the fusion protein
taspase1 chains
{
  "N320": "1..2718",
  "C180": "2719..3969"
}
pride deposits
[
  "PXD004927",
  "PXD005296",
  "PXD006545",
  "PXD008003",
  "PXD011932",
  "PXD023841",
  "PXD028050",
  "PXD034631"
]
measurements
[
  {
    "quantity": "independent PRIDE deposits reporting this site",
    "subject": "Q03164 K2528 SUMOylation",
    "value": 8,
    "units": "deposits",
    "n": 174,
    "universe": "all 174 MS-evidenced site/modification calls on Q03164",
    "null": null,
    "note": "the maximum over the whole protein; the next best is acetyl-K2958 at 4 and acetyl-K1130 at 1"
  },
  {
    "quantity": "residues present in wild-type N320 but absent from a classical KMT2A fusion",
    "subject": "Q03164 aa 1608-2718 (post-cluster to taspase1 CS2)",
    "value": 1111,
    "units": "residues",
    "n": 1,
    "universe": "UniProt chain 1..2718 minus the retained+variable region",
    "null": null,
    "note": "the stretch that makes 'the fusion is N320 plus a partner' a ~1,100-residue approximation"
  }
]
How this could bite us back
  • Calling K2528 'the best-supported PTM site on KMT2A' invites the reading that it is the most important one. Deposit count measures detectability. This entry exists to record a coordinate and a retention call, not to promote a site.
Still open
  1. Does losing K2528 SUMOylation matter? Nothing is known about the site's function, so this is a clean, well-specified gap rather than a competing hypothesis: a SUMO-IP western on wild-type KMT2A N320 versus KMT2A::AFF1 would establish whether the fusion is SUMO-hypomodified at all, and that has never been reported.
  2. Are the eight deposits methodologically independent, or is this one SUMO-enrichment chemistry repeated eight times? The answer changes how much weight 'best-supported site on the protein' can carry.
Every source behind this entry (2)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    Q03164 K2528 SUMOylation is reported in eight independent PRIDE deposits, more than any other modification site on KMT2A.

    EBI Proteins API - proteomics-ptm, Q03164 (KMT2A) - K2528 SUMOylation, 8 deposits · EMBL-EBI Proteins API / PRIDE-reprocessed MS evidence 2026
    What this evidence cannot show

    Deposit count measures reproducible detectability, not biological importance. SUMO-site enrichment methods (SUMO-remnant, His-SUMO) have strong and shared sequence biases, so eight deposits finding the same site is less independent than eight deposits sounds. The eight were read from the endpoint's dbReferences and their individual project metadata was NOT reviewed.

    Evidence that argues against this (1)
    • If the eight deposits share a method, they are one observation repeated, not eight. That was not checked.
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt annotates Q03164 K2528 as a glycyl-lysine isopeptide cross-link with SUMO2, and annotates the taspase1-derived chains as 1..2718 and 2719..3969.

    What this evidence cannot show

    Residue identity verified against the live UniProt Q03164 FASTA (3,969 aa) on 2026-09-24. Canonical isoform only. The CROSSLNK carries ECO:0007744 (combinatorial/MS evidence), not ECO:0000269, so no curator has read a low-throughput experiment establishing it.

    Evidence that argues against this (1)
    • The UniProt CROSSLNK carries ECO:0007744, which is MS-derived - so it is NOT independent corroboration of the eight PRIDE deposits in [E1], it is very likely a curated restatement of the same evidence class. This entry therefore has one line of evidence presented as two. The taspase1 chain boundaries are the firmer half: K2528 sits 190 residues inside the 1..2718 chain, so the 'in N320' call survives any plausible error in the cleavage-site annotation.
Connected to