Attribution audit 2 of 5 - post-translational modifications on KMT2A::AFF1: nothing survives at peptide level, and the one strong PTM result is explicitly about the wild-type protein
also called MLL-AF4 PTM audit · fusion phosphosite attribution · S516 attribution
Every chemical tag that gets added to a protein is identified by the short stretch of sequence it sits on, and every such stretch in the KMT2A::AFF1 fusion - except the one that spans the join - is letter-for-letter identical to a stretch in one of the two normal proteins. So a modification found by mass spectrometry cannot be said to be on the fusion rather than on the normal copy.
The same thing, for a scientist
This class fails the hardest of the five, and it fails for a reason that is computable rather than historical. **The structural argument.** A PTM is assigned to a protein by the peptide that carries it. In canonical KMT2A::AFF1 the KMT2A contribution runs to residue 1406 and the AFF1 contribution begins at residue 347 [E3]. Every tryptic peptide of the chimera therefore lies either entirely within KMT2A 1-1406 or entirely within AFF1 347-1210, **except the single peptide that spans the junction** - and for the canonical junction that peptide does not exist at zero missed cleavages at all (`ana-junction-peptide-detectability-kmt2a-fusions`). A modified peptide identified from a KMT2A-rearranged cell is thus, by sequence, equally consistent with the fusion and with the relevant wild-type parent. No search engine, no FDR threshold and no localisation score can break that tie, because the ambiguity is in the sequence and not in the spectrum. **The published PTM results, audited.** The strongest PTM finding on this protein is that **MLL is phosphorylated at serine 516 by ATR in response to genotoxic stress**, which blocks the SCF(Skp2) interaction and stabilises it [E1]. S516 lies inside the retained 1-1406 region, so the fusion carries the site. But the paper's claim is not that the fusion is phosphorylated: it is that **MLL fusions act as dominant negatives that abrogate ATR-mediated phosphorylation and stabilisation of wild-type MLL** [E1]. The measured object is the wild-type protein. Whether KMT2A::AFF1 itself is phosphorylated at S516 is not addressed, and a phospho-S516 antibody applied to a t(4;11) blast would report both species at once. Ubiquitination is the same shape. Polyubiquitination was demonstrated on **FLAG-tagged MLL co-expressed with HA-ubiquitin in 293T** [E2]. That is attributable - to a tagged wild-type construct in a kidney cell line. **Verdict. This class does NOT survive, and it is the one class where no partial survivor exists.** The size-resolution escape that rescues the composition and turnover classes is unavailable here, because a PTM is read off a peptide and peptides have no size dimension relative to the parent protein. The only two designs that would work are a phospho-antibody Western where the fusion band is resolved (which reports occupancy of a site, not its discovery) and PTM-aware targeted MS on the junction peptide itself.
What would fix it, and why it is cheap
The minimal experiment is a modified-junction-peptide PRM. The canonical KMT2A::AFF1 junction peptide VDFKQTYSNEVHCVEEILK carries one serine (S in QTYSNEV), one threonine and one tyrosine, all on the AFF1 side of the join. A PRM method monitoring the unmodified peptide at [M+H]+ 2281.1169 plus its +79.966 phospho forms would, in one injection, say whether the junction region of the fusion is phosphorylated - and the answer would be attributable by construction.
The second design is immunoprecipitate-then-blot with size resolution. Immunoprecipitate with any shared-epitope antibody, run a gel, and probe with a phospho-specific antibody. The fusion band at about 240 kDa and the wild-type MLL-N320 band are separated on the gel, so the phospho signal is assigned by position, not by the antibody. This is how the turnover class survives (ana-attribution-audit-protein-half-life-kmt2a-aff1) and it would work here too. It has not been done for any KMT2A fusion.
Neither needs a new reagent; both are a day of instrument time on material the field already makes.
Why this is an absence of testing and not a negative result
Nothing above says the fusion is unmodified. It says the assay that could have seen a fusion-specific modification has not been run. This graph already records that no phosphorylation site has ever been mapped on a KMT2A fusion protein (absence-no-phosphosite-mapped-on-any-kmt2a-fusion-protein); this entry supplies the mechanism for that absence. The sites are not missing because they were looked for and not found - they are missing because the standard shotgun workflow cannot assign them, and the targeted workflow that could has never been applied.
A second consequence follows and it constrains this graph's own phospho work. Every claim of the form 'site X is on retained partner sequence and therefore on the fusion' is an inference from breakpoint position, not an observation of the fusion. That inference is sound where the breakpoint distribution is tight, and it is exactly as strong as the breakpoint data - no stronger.
| claim class | post-translational modifications of the KMT2A::AFF1 protein |
|---|---|
| verdict | does not survive; the only class of the five with no partial survivor |
reagent classes found
[
{
"reagent": "shotgun phosphoproteomics, peptide-level identification",
"system": "patient blasts and cell lines",
"fusion_attributable": "no",
"why": "every tryptic peptide except the junction peptide is shared with a wild-type parent; sequence ambiguity, not spectral"
},
{
"reagent": "phospho-specific antibody (anti-pS516 class)",
"system": "any rearranged cell",
"fusion_attributable": "no",
"why": "epitope is in retained sequence; reads fusion and wild-type together unless resolved by size on a gel"
},
{
"reagent": "anti-FLAG IP of FLAG-MLL with HA-ubiquitin",
"system": "293T transient transfection",
"fusion_attributable": "yes, but for wild-type MLL",
"why": "tagged construct is wild-type MLL, not the fusion"
}
]minimal fix
[ "PTM-aware PRM on the junction peptide VDFKQTYSNEVHCVEEILK, monitoring +79.966 forms of its S/T/Y residues", "IP with any shared-epitope antibody, then phospho-blot with the ~240 kDa fusion band resolved from wild-type MLL-N320" ]
retained region AA — the numbers behind the chart
{
"kmt2a_last_residue": 1406,
"aff1_first_residue": 347,
"kmt2a_full_length": 3969
}sites inside retained region
[
{
"site": "S516",
"kinase": "ATR",
"evidence": "E1",
"measured_on_fusion": false
}
]- A kinase inhibitor chosen because it removes a phosphosite 'on the fusion' may in fact be removing that site on wild-type KMT2A or wild-type AFF1 in every cell of the body, because no published measurement separates the two - so the therapeutic index inferred from such a biomarker is not measured, it is assumed.
- Pharmacodynamic assays built on a phospho-specific antibody in this disease will report a mixture, and because the fusion is the minor species the assay is biased towards reporting the wild-type response to the drug.
- If a fusion-specific PTM does exist and is the point of therapeutic leverage, the standard workflow is structurally incapable of finding it, so the absence of such a finding is not reassurance that no such leverage exists.
- Is the KMT2A::AFF1 junction peptide itself modified? It carries a serine, a threonine and a tyrosine on the AFF1 side, and a PRM method monitoring the +79.966 forms would answer it in one injection with attribution guaranteed by construction.
- Is S516 phosphorylated on the fusion protein as well as on wild-type MLL? The dominant-negative model predicts engagement of the fusion by ATR machinery, but the experiment distinguishing the two species - a phospho-S516 blot with the fusion band resolved - has never been run.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
MLL is phosphorylated at serine 516 by ATR in response to genotoxic stress in S phase, which disrupts its interaction with SCF(Skp2) and stabilises it; MLL fusions are reported to act as dominant negatives that abrogate the ATR-mediated phosphorylation and stabilisation of wild-type MLL, so the measured phospho-species in that work is wild-type MLL and not the fusion.
Phosphorylation of MLL by ATR is required for execution of mammalian S-phase checkpoint. · Liu H, Takeda S, Kumar R, et al. · Nature 2010What this evidence cannot show
Full abstract read verbatim from the Europe PMC core record; the PMC full text (PMC2940944) returned HTTP 500 on fullTextXML and was not pursued further because the attribution point turns on the abstract's own framing, which is explicit. The fusion model used for the checkpoint phenotype is a murine Mll-CBP knock-in mimicking t(11;16), not KMT2A::AFF1. S516 is within KMT2A 1-1406 and so is retained in KMT2A::AFF1 [E3], but no measurement of S516 occupancy on a fusion protein is reported.
Evidence that argues against this (1)
- If the fusion is a dominant negative because it sequesters the ATR-dependent machinery, the fusion would have to be engaged by that machinery - which would imply the fusion IS phosphorylated or at least bound. The dominant-negative model therefore carries an untested prediction about the fusion's own modification state, and the experiment that would test it is a phospho-S516 blot with the fusion band resolved by size. Liu H et al. 2007
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
MLL polyubiquitination was demonstrated by co-expressing FLAG-MLL and HA-ubiquitin in 293T cells followed by anti-FLAG immunoprecipitation, and MLL-AF4 resistance to degradation was assessed on FLAG-MLL-AF4 transiently expressed in 293T and purified with anti-FLAG.
Bimodal degradation of MLL by SCFSkp2 and APCCdc20 assures cell cycle execution: a critical regulatory circuit lost in leukemogenic MLL fusions. · Liu H, Cheng EH, Hsieh JJ. · Genes & development 2007What this evidence cannot show
Read verbatim from the PMC HTML page for PMC1993870 after fullTextXML returned HTTP 500 and efetch db=pmc returned a document with no <body> element; the HTML route gave 91,654 characters. The ubiquitination experiment is on tagged wild-type MLL in a non-haematopoietic line, and the MLL-AF4 experiment is an in vitro degradation assay on immunoprecipitated tagged protein, not a measurement of ubiquitin occupancy on endogenous fusion.
Evidence that argues against this (1)
- The same paper reached the endogenous fusion by a different route and succeeded there, so the tagged-overexpression limitation is specific to the ubiquitination experiment rather than to the paper: in SEM cells separated by centrifugal elutriation, MLL-AF4 protein did not fluctuate through the cell cycle while wild-type MLL did. That endogenous result is attributable and is audited in ana-attribution-audit-protein-half-life-kmt2a-aff1. Liu H et al. 2007
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In the canonical KMT2A::AFF1 chimera the KMT2A contribution ends at residue 1406 and the AFF1 contribution begins at residue 347, so every tryptic peptide of the chimera except the junction-spanning peptide is sequence-identical to a peptide of one wild-type parent.
Acute mixed-lineage leukemia t(4;11)(q21;q23) generates an MLL-AF4 fusion product. · Domer PH, Fakharzadeh SS, Chen CS, et al. · Proc Natl Acad Sci U S A 1993What this evidence cannot show
The junction anchor is GenBank AAC37520.1 from this paper, and the residue arithmetic is this project's own computation in analyses/junction-peptide-detectability, independently reproduced by the coordinator to four decimal places. Breakpoints vary between patients, so the specific numbers 1406 and 347 apply to the canonical chimera; the structural argument - that only the junction peptide is unshared - holds for every in-frame breakpoint.
Evidence that argues against this (1)
- The argument would fail if a fusion-specific PTM changed the mass of an otherwise shared peptide in a way that distinguished it - but that is circular, because one cannot know a modification is fusion-specific without first attributing the peptide. It would also fail for a protein-level rather than peptide-level readout, which is exactly the size-resolved blot proposed as the fix. UniProt Consortium 2026
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
UniProt's reviewed entry for human KMT2A (Q03164) is 3969 residues, so the retained 1-1406 region is roughly the first 35% of the protein and the taspase1 cleavage sites near residues 2666 and 2718 are outside it.
UniProtKB Q03164 (KMT2A_HUMAN), Histone-lysine N-methyltransferase 2A, 3969 aa; modified-residue and sequence features · UniProt Consortium 2026What this evidence cannot show
Database fact, fetched from the REST endpoint. Cited only for sequence length and feature positions. UniProt feature coordinates are curated from primary literature and occasionally revised; the load-bearing use here is only that the taspase1 sites lie well beyond residue 1406, which is not a marginal call.
Evidence that argues against this (1)
- A database fact is opposable on scope: UniProt's canonical isoform is one of several KMT2A isoforms, and a breakpoint mapped onto a different isoform numbering would shift every residue number. Any entry quoting a KMT2A residue number should say which isoform it is numbered against. Domer PH et al. 1993
audits every published PTM claim about this protein by reagent, and finds no attributable one
supplies the mechanism for that absence: peptide-level identification cannot break a sequence tie
the proposed fix is a PTM-aware extension of that PRM assay
same reagent problem, different escape: composition is rescued by size resolution and PTM assignment is not
that atlas infers fusion content from breakpoint position; this entry bounds how strong such an inference can be
both classes lose the size dimension that rescues the composition and turnover classes
explains why the PTM class is the one with no partial survivor
verdict 2 of 5