Attribution audit 1 of 5 - who co-purifies with KMT2A::AFF1: the inventory does not survive, the individual partner blots mostly do
also called MLL-AF4 complex composition audit · AEP attribution · fusion interactome attribution
Lists of the proteins that stick to the KMT2A::AFF1 fusion were made with antibodies that also stick to the ordinary versions of those same proteins, so a mass-spectrometry list cannot say which proteins were on the fusion and which were on the normal copy. The one thing that does work is a Western blot, because there the fusion is a different size and can be seen on its own.
The same thing, for a scientist
Four reagent classes carry every published composition claim for the KMT2A::AFF1 protein complex, and they audit differently. **(1) Partner-antibody immuno-affinity purification plus mass spectrometry, in a cell that carries no KMT2A rearrangement.** The AEP complex - AFF1/AF4, AFF4/AF5q31, MLLT1/ENL, CDK9 and cyclin T1 - was purified from **K562** nuclear extract with an anti-AF4 monoclonal raised against AF4 aa 782-979, and reproduced with an anti-AF5q31 monoclonal [E1]. K562 carries BCR::ABL1 and no KMT2A fusion. That purification is fully attributable - to **wild-type AF4**. It is not a fusion inventory, though it is routinely cited as the composition of the fusion complex. **(2) Reciprocal co-immunoprecipitation read out by Western blot, in patient-derived t(4;11) lines.** The claim that the fusion sits in a hybrid MLL/AEP complex rests on reciprocal IP in SEM-K2 and other leukaemia lines with anti-AF4, anti-AF5q31, anti-ENL and anti-MLL-N [E1]. Every one of those binds retained sequence and cannot separate fusion from wild-type parent **at the antibody step**. It can at the **detection** step: the fusion runs at about 240 kDa and wild-type AF4 at 125 and 145 kDa, and the 240 kDa species appears only in t(4;11) lines [E2]. A co-IP western showing partner X in a lane with a resolved 240 kDa fusion band is a genuine attributable statement that X is in a complex containing the fusion. **(3) Epitope-tagged overexpression.** FLAG-tagged MLL fusions in pCMV5 were used for IP [E1]; HA-tagged MLL::AF9 supports anti-HA pulldown and anti-HA ChIP [E4]. The tag is on the fusion and nowhere else, so this is attributable - at a stoichiometry the endogenous protein never reaches, and with a tag that is itself perturbing if placed N-terminally [E4]. **(4) The one SET-retaining affinity purification is unreadable.** Benedikt et al. report an 11-component AF4 complex and a complex of at least 16 constituents for AF4-MLL [E3]. That molecule is the **reciprocal** AF4::KMT2A, not KMT2A::AFF1, and the paper failed all four readability routes on 2026-09-20, so the purification reagent, the expression system and the gene-symbol-level membership cannot be checked. **Verdict. The complex inventory does NOT survive; single-partner claims read on a size-resolved blot DO.** AP-MS has no size dimension: every tryptic peptide of the fusion except the junction peptide is shared with a wild-type parent, so a protein identified in an anti-AF4 pulldown from SEM cannot be assigned to the fusion rather than to the more abundant wild-type AF4 in the same lysate [E2].
Why the abundance ratio makes this worse than a symmetric ambiguity
Li, Frestedt and Kersey reported that in t(4;11) lines the fusion proteins were present in lesser quantity than the wild-type AF4 2. An anti-AF4 pulldown from such a lysate is therefore dominated by the wild-type species, and any co-purifying protein is, on prior odds, a wild-type AF4 interactor. The direction of the bias is knowable and it points away from the fusion.
That is also why the AEP inventory and the fusion inventory are so easily conflated. They were generated with the same antibody, and in the rearranged cell that antibody mostly reports the same complex it reported in K562.
The minimal experiment that would fix it
Two options, in increasing order of cost.
- Junction-peptide PRM on the IP eluate. Run the existing anti-AF4 or anti-MLL-N pulldown from SEM unchanged, then ask by targeted mass spectrometry whether the junction peptide is in the eluate and at what abundance relative to a wild-type-only peptide. For canonical KMT2A::AFF1 that peptide is
VDFKQTYSNEVHCVEEILK, 19 residues, [M+H]+ 2281.1169, unique against both wild-type parents and all 20,416 reviewed human proteins (ana-junction-peptide-detectability-kmt2a-fusions). This turns an unattributable pulldown into a quantified mixture without changing the reagent and without a new antibody. - Anti-FLAG or anti-HA AP-MS from an epitope tag knocked in at the endogenous fusion allele. The only design that gives a fusion-specific inventory at endogenous stoichiometry. Nobody has published one for KMT2A::AFF1 (
concept-epitope-tagged-knock-in-at-the-endogenous-fusion-allele).
What this does NOT license
It does not license saying the AEP components are absent from the fusion complex. They very probably are present - the fusion retains the AFF1 C-terminal region that carries the AFF4/ENL/P-TEFb contacts, and the tagged-overexpression IPs support it 1. The audit finding is narrower and it is about evidentiary weight: the number 11, the number 16, and any gene-symbol-level membership list are properties of a measurement that could not tell the two species apart.
| claim class | composition of the KMT2A::AFF1 protein complex |
|---|---|
| verdict | the inventory does not survive; size-resolved single-partner blots do |
reagent classes found
[
{
"reagent": "anti-AF4 / anti-AF5q31 monoclonal immuno-affinity purification plus MS",
"system": "K562, no KMT2A rearrangement",
"fusion_attributable": "no",
"why": "no fusion is present in the cell; the inventory is a wild-type AF4 inventory"
},
{
"reagent": "anti-AF4 / anti-ENL / anti-MLL-N reciprocal co-IP plus Western",
"system": "SEM-K2 and other t(4;11) lines",
"fusion_attributable": "partially",
"why": "the antibody cannot discriminate, but the ~240 kDa fusion band is resolved from 125/145 kDa wild-type AF4 in the readout"
},
{
"reagent": "anti-FLAG IP of FLAG-tagged MLL fusions",
"system": "transient transfection, pCMV5",
"fusion_attributable": "yes",
"why": "tag is unique to the fusion; not endogenous stoichiometry"
},
{
"reagent": "unstated affinity purification of AF4-MLL",
"system": "unknown - paper unreadable by all four routes",
"fusion_attributable": "unknown",
"why": "reagent not readable, and the molecule is the reciprocal product"
}
]minimal fix
[ "junction-peptide PRM on the existing IP eluate (VDFKQTYSNEVHCVEEILK, MH+ 2281.1169)", "anti-FLAG or anti-HA AP-MS from an epitope tag knocked in at the endogenous fusion allele" ]
unreadable
[
{
"pmid": "21030982",
"routes_tried": [
"EPMC search EXT_ID core: pmcid null",
"fullTextXML: no PMCID; PMC21030982 returns HTTP 500",
"efetch db=pmc: PMCID not available",
"publisher HTML nature.com/articles/leu2010249: HTTP 200, abstract plus paywall only"
],
"date": "2026-09-20",
"title": "The leukemogenic AF4-MLL fusion protein causes P-TEFb kinase activation and altered epigenetic signatures.",
"authors": "Benedikt A, Baltruschat S, Scholz B, et al.",
"venue": "Leukemia",
"year": 2011,
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
}
]- If a drug is chosen because its target is 'in the MLL-AF4 complex', and that membership comes from an AP-MS list that could not separate fusion from wild-type AF4, the drug may be aimed at a wild-type AF4 interaction the fusion does not make - a target-selection error no amount of potency repairs.
- The converse risk is toxicity: a partner that sits in the wild-type AF4 complex in every cell of the body is a worse therapeutic target than the same partner would be if it were fusion-specific, and the current evidence base cannot tell those two situations apart.
- Menin-inhibitor pharmacodynamic assays that read 'loss of the fusion complex' by co-IP from patient blasts inherit this ambiguity, so a partial signal could be movement of the wild-type AF4 complex under drug rather than eviction of the fusion.
- What fraction of an anti-AF4 pulldown from SEM is fusion and what fraction is wild-type AF4? Junction-peptide PRM against a wild-type-only AFF1 peptide would give the ratio in one run and would recalibrate every composition claim in this graph at once.
- Does any component of the AEP inventory bind the fusion but not wild-type AF4, or the reverse? Nothing in the current evidence base could have detected such a difference, so the assumption that the two complexes are compositionally identical has never been tested.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The AEP complex (AF4, AF5q31, ENL, CDK9, cyclin T1) was immuno-affinity purified from K562 cells - a line with no KMT2A rearrangement - using an anti-AF4 monoclonal, and the claim that MLL chimeras form hybrid MLL/AEP complexes rests on reciprocal co-immunoprecipitation with the same shared-epitope antibodies in human leukaemia lines plus IP of FLAG-tagged MLL fusions expressed from pCMV5.
A higher-order complex containing AF4 and ENL family proteins with P-TEFb facilitates oncogenic and physiologic MLL-dependent transcription. · Yokoyama A, Lin M, Naresh A, et al. · Cancer cell 2010What this evidence cannot show
Read verbatim from the PMC body (PMC2824033 via efetch db=pmc, <body> present, 58,695 characters of stripped text). The paper is not making an attribution error - it states its systems plainly and its co-IP westerns do resolve species by size. The audit point concerns how the resulting inventory is cited downstream. Note also that neither the AF4 nor the AF5q31 purification recovered DOT1L or AF10, which other groups had reported as ENL partners, so complex inventories of this era disagree with one another.
Evidence that argues against this (1)
- The same paper's headline cuts the other way for the underlying biology: AEP is normally recruited to MLL-target chromatin. Wild-type AF4 and wild-type MLL therefore occupy overlapping complexes at overlapping loci, which is what makes the shared-epitope ambiguity biologically real rather than merely formal. Yokoyama A et al. 2010
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In t(4;11) leukaemia lines the MLL-AF4 fusion protein is detected as a roughly 240 kDa band distinct from wild-type AF4 at 125 and 145 kDa, and the fusion protein is present in lesser quantity than wild-type AF4.
AF4 encodes a ubiquitous protein that in both native and MLL-AF4 fusion types localizes to subnuclear compartments. · Li Q, Frestedt JL, Kersey JH. · Blood 1998What this evidence cannot show
Abstract read in full via the Europe PMC core record; the paper has no PMCID and the full text was not reached, so the basis of 'lesser quantity' (densitometry or visual inspection) could not be checked and the statement should be treated as qualitative. Rabbit polyclonals raised against GST-AF4 immunogens, 1998 reagents.
Evidence that argues against this (1)
- Relative abundance measured on a blot depends on epitope accessibility in each species, so a fusion whose AF4 portion folds differently could blot less efficiently without being less abundant. No orthogonal quantification of fusion versus wild-type AF4 protein abundance - for example junction-peptide PRM against a wild-type-only peptide - has been published for any KMT2A fusion. Yokoyama A et al. 2010
- E3weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper
An affinity purification reports that the wild-type AF4 complex consists of 11 binding partners and the AF4-MLL complex of at least 16 constituents including P-TEFb kinase, H3K4me3 and H3K79me3 methyltransferases, a protein arginine N-methyltransferase and a histone acetyltransferase; the paper is paywalled and has no PMCID, so the purification reagent and the expression system cannot be checked.
The leukemogenic AF4-MLL fusion protein causes P-TEFb kinase activation and altered epigenetic signatures. · Benedikt A, Baltruschat S, Scholz B, et al. · Leukemia 2011What this evidence cannot show
UNREADABLE BY ALL FOUR ROUTES, checked 2026-09-20. Europe PMC core record returns pmcid null and isOpenAccess N. NCBI ID converter returns 'Identifier not found in PMC'. elink dbfrom=pubmed db=pmc returns only pubmed_pmc_refs (48 citing articles) and no pubmed_pmc link. efetch db=pmc returns 'The following PMCID is not available'. The publisher page nature.com/articles/leu2010249 returns HTTP 200 and 18,543 characters of stripped text that is abstract plus paywall notice only. Only the abstract was read, and nothing beyond it is asserted here. NOTE ALSO: the molecule is AF4::KMT2A, the RECIPROCAL product - it is SET-retaining precisely because it carries the KMT2A C-terminus - so it is not KMT2A::AFF1 and the brief's framing of it as a KMT2A fusion complex needs that qualification.
Evidence that argues against this (1)
- This graph's entry complex-af4-mll-assembles-a-sixteen-component-elongation-machine describes this purification as being done 'on over-expressed protein'. That methods detail is not in the abstract and the full text is unreadable by all four routes, so it is an inference from the model system rather than a read fact and should be marked as such in that entry. Benedikt A et al. 2011
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
An epitope tag placed on the fusion gives fully attributable pulldowns and chromatin profiling: HA-FKBP12-tagged MLL::AF9 supported anti-HA ChIP-seq and anti-HA Western blot in which no wild-type species carries the epitope.
MLL::AF9 degradation induces rapid changes in transcriptional elongation and subsequent loss of an active chromatin landscape. · Olsen SN, Godfrey L, Healy JP, et al. · Molecular cell 2022What this evidence cannot show
Read verbatim from the PMC body (PMC9044330 via efetch db=pmc, <body> present). The construct is retrovirally overexpressed in murine LSK cells and in human CD34+ cells, not knocked in, and the paper's own Limitations section states that 'future studies should include endogenously degron-tagged MLL::AF9'. It is MLL::AF9, not KMT2A::AFF1, and is cited here as an existence proof for the reagent class.
Evidence that argues against this (1)
- The same paper found that an N-terminally tagged MLL::AF9 did not reproduce the leukaemogenesis kinetics of the untagged construct, which the authors attribute to interference with chromatin localisation because the DNA-binding motifs and the menin-binding domain are N-terminal. A tag is therefore attributable and potentially perturbing at the same time, and tag position has to be justified experimentally rather than assumed neutral. Olsen SN et al. 2022
audits every published claim about which proteins co-purify with this fusion, by reagent class
the reagent property that drives the verdict is established there
the proposed fix is the computed junction peptide run as PRM on the existing IP eluate
the 11-and-16-component inventory rests on a paper unreadable by all four routes, and describes the reciprocal product rather than KMT2A::AFF1
the wild-type AF4 pool is the background term that defeats the pulldown
same reagent problem, different escape: composition is rescued by size resolution and PTM assignment is not
anti-tag AP-MS is the proposed fix for the composition class
the composition audit's fourth reagent class is unauditable because this paper cannot be read
anti-HA AP-MS is the direct test of the composition audit's verdict
verdict 1 of 5