WRAD reaches KMT2A fusions only as a passenger on a recruited wild-type complex
also called second-shell WRAD · WRAD arrives attached to wild-type KMT2A
The WRAD proteins may well be found sitting next to the leukemia fusion - not because they grip it, but because they are already attached to the normal KMT2A the fusion is known to need. If that is right, a test that only asks 'is WRAD nearby?' will say yes and mean nothing.
The same thing, for a scientist
The direct fusion has no WRAD-binding surface, but it has two documented routes to one. It requires prebinding of a wild-type KMT2A complex for stable association with HOXA9 [E1], and the wild-type Kmt2a allele is required for MLL-AF9 leukaemogenesis and maintenance [E2]. A wild-type complex is WRAD-loaded by definition. The hypothesis is therefore that **WRAD subunits are second-shell neighbours of every KMT2A fusion, recovered by any method with a spatial rather than a binary readout, and absent from any method that requires a direct contact**. The strongest support is from a sister fusion family: BioID of NUP98-PHF23, a fusion containing no KMT2A sequence, recovers KMT2A/MLL1, KMT2B, WDR5 and RBBP5 together, and affinity purification of five NUP98 fusions recovers WDR5 alongside MLL1 [E3][E4]. In both, WRAD arrives with a recruited methyltransferase rather than on its own account. The hypothesis matters because it predicts that the experiment everyone would run to settle the WRAD question - a proximity map of KMT2A::AFF1 - would return a POSITIVE that means nothing, and it names in advance the control that would be needed.
Predictions
- A TurboID map of KMT2A::AFF1 recovers WDR5 and RBBP5 at moderate abundance, and the recovery collapses when the wild-type KMT2A allele is degraded while the fusion is untouched. Under the rival model (no route at all) the recovery is absent in both conditions; under the rival model (a direct non-SET contact) the recovery survives wild-type degradation.
- A stringent co-immunoprecipitation of tagged KMT2A::AFF1 does not recover WRAD, because the second shell does not survive stringency that a direct contact would.
- WRAD recovery scales with wild-type KMT2A abundance, so a taspase1 or UBE2O perturbation that stabilises wild-type MLL1 5 increases it.
- The reciprocal-positive line behaves differently: in a t(4;11) line expressing AF4-MLL, WRAD recovery on the reciprocal product survives wild-type degradation, because that product has its own interface.
This graph already holds the concept that proximity counts are detectability rather than membership, and the observed WRAD split across one 50-bait proximity deposition - WDR5 11, ASH2L 9, RBBP5 0, DPY30 0 - is exactly what a second-shell, abundance-weighted recovery of an obligate complex looks like 6.
Why this is worth writing down rather than assuming
The field has treated the WRAD-versus-fusion question as binary for twenty-two years: either they associate or they do not. A second-shell model is neither, and it changes what counts as an answer. It also explains the otherwise awkward pattern that WRAD subunits are functionally required in KMT2A-rearranged cells (Wdr5 deletion attenuates MLL-AF9 leukaemogenesis 7) while having no plausible contact with the fusion: the requirement is the wild-type allele's requirement, transmitted.
If true, the therapeutic reading is the least attractive of the three: a WRAD agent works in these cells by degrading normal KMT2A function, in infants, with the fusion untouched.
| statement | WRAD subunits associate with KMT2A fusion proteins only indirectly, as passengers on a recruited wild-type KMT2A/COMPASS complex, and therefore appear in spatial assays and disappear in stringent binary ones. |
|---|---|
| origin | literature-gap |
| prior probability | 0.45 |
predictions
[ "TurboID of KMT2A::AFF1 recovers WDR5 and RBBP5; recovery collapses on acute degradation of the wild-type KMT2A allele", "Stringent co-IP of tagged KMT2A::AFF1 does not recover any WRAD subunit", "WRAD recovery increases when wild-type MLL1 is stabilised (UBE2O or taspase1 perturbation)", "In a reciprocal-positive t(4;11) line, WRAD recovery on the der(4) product survives wild-type KMT2A degradation" ]
experiments
[ "exp-wrad-coip-and-degron-timed-proximity-across-kmt2a-fusion-classes" ]
analyses
[ "ana-wrad-fusion-evidence-enumeration" ]
falsifiers
[ "A stringent co-immunoprecipitation of tagged KMT2A::AFF1 recovers WDR5 or RBBP5 stoichiometrically - the association is direct and the hypothesis is wrong", "TurboID of KMT2A::AFF1 recovers no WRAD subunit above control even with the wild-type allele intact - there is no route at all and the hypothesis is unnecessary", "WRAD recovery in a proximity map is unchanged by acute wild-type KMT2A degradation - the route is not the wild-type complex", "The reported NUP98-fusion-to-MLL1 recruitment is confirmed to be an overexpression artefact, removing the only positive precedent for the mechanism" ]
- If WRAD reaches the fusion only through wild-type KMT2A, then a WRAD-directed agent acts by impairing normal KMT2A function in an infant with no marrow reserve, and the fusion is untouched.
- A positive proximity result would be read by most readers as settling the WRAD question in favour of association, when under this hypothesis it settles nothing - which makes running the obvious experiment without the degron arm actively harmful to the field's understanding.
- Stabilising wild-type MLL1 is itself a proposed therapeutic strategy; if it increases WRAD delivery to fusion target loci, the two strategies interfere.
- Is there any fusion-bearing system in which wild-type KMT2A can be acutely degraded without killing the cell? The wild-type allele is required for maintenance, so the degron window may be too short to read a proximity map in.
- Would a second-shell WRAD recovery have any functional consequence at all, or is it an epiphenomenon of the wild-type complex being where it already needs to be?
- Does the same second-shell logic apply to menin? Menin is a direct contact of the fusion AND of the wild-type protein, so a proximity map cannot separate the two pools either.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
KMT2A fusion proteins lack the PHD fingers and require prebinding of a wild-type KMT2A complex, together with CXXC recognition of DNA, for stable association with HOXA9.
Multiple interactions recruit MLL1 and MLL1 fusion proteins to the HOXA9 locus in leukemogenesis. · Milne TA, Kim J, Wang GG, et al. · Mol Cell 2010What this evidence cannot show
A locus-specific study at HOXA9 in limited models; generality across partners and across the genome is assumed.
Evidence that argues against this (1)
- A genome-wide study reports that wild-type MLL1 and MLL-AF9 occupy distinct chromatin regions with divergent transcriptional outputs, which is difficult to reconcile with a model in which the wild-type complex licenses the fusion at the same sites. Xu J et al. 2016
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The wild-type Mll allele is required for MLL-AF9-induced leukemogenesis and for maintenance of MLL-AF9-transformed cells, and menin recruits both wild-type MLL and the fusion to HOX loci.
MLL-AF9-induced leukemogenesis requires coexpression of the wild-type Mll allele. · Thiel AT, Blessington P, Zou T, et al. · Cancer Cell 2010What this evidence cannot show
Mouse genetics with MLL-AF9 only; no lymphoid or t(4;11) model, and no human primary system.
Evidence that argues against this (1)
- Stabilising wild-type MLL1 displaces MLL chimeras from chromatin and delays leukaemic progression, so the wild-type allele can also act against the fusion - the relationship is not simply permissive. Liang K et al. 2017
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
BioID proximity labelling of the NUP98-PHF23 oncofusion, which contains no KMT2A sequence, recovers KMT2A/MLL1, KMT2B, WDR5 and RBBP5, with the paper's model being that the FG repeats partition the KMT2/MLL H3K4 methyltransferases into a chromatin-bound condensate.
The phenylalanine-and-glycine repeats of NUP98 oncofusions form condensates that selectively partition transcriptional coactivators. · Ahn JH, Guo Y, Lyons H, et al. · Mol Cell 2025What this evidence cannot show
A different fusion family. A proximity hit is a neighbour, not a binding partner, and recovery is abundance-weighted and compartment-structured. The WRAD subunit identities are in the data rather than the abstract, which is why this paper is invisible to any WRAD-keyed literature search.
Evidence that argues against this (1)
- Unbiased AP-MS of five Strep-HA-tagged NUP98 fusions at near-physiological levels in HL-60 explicitly failed to detect the reported MLL1 and NSL complex interactions, so the recruitment premise of this hypothesis is contested in the family it is borrowed from. Terlecki-Zaniewicz S et al. 2021
- E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Five NUP98 fusions containing no KMT2A sequence bind MOF, PHF20, OGT1, HCF1, WDR5 and MLL1 by biotin affinity purification, and conditional Mll1 deletion delays NUP98-HOXA9 leukaemogenesis.
NUP98 Fusion Proteins Interact with the NSL and MLL1 Complexes to Drive Leukemogenesis. · Xu H, Valerio DG, Eisold ME, et al. · Cancer Cell 2016What this evidence cannot show
Targeted western blots after affinity purification in overexpression systems, rather than unbiased mass spectrometry. WDR5's recovery is most simply explained as a passenger on recruited MLL1, which is the very point this hypothesis makes, so the citation is being used for the pattern rather than for a direct claim.
Evidence that argues against this (1)
- The same associations were not detected by unbiased AP-MS of the same five fusions at near-physiological levels, which is the stronger design for a composition question. Terlecki-Zaniewicz S et al. 2021
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
UBE2O regulates the stability of wild-type MLL1 in response to interleukin-1 signalling, and pharmacologically stabilising wild-type MLL1 displaces MLL chimeras from some target genes, delaying leukaemic progression - so wild-type KMT2A abundance in a fusion-bearing cell is a manipulable variable.
Therapeutic Targeting of MLL Degradation Pathways in MLL-Rearranged Leukemia. · Liang K, Volk AG, Haug JS, et al. · Cell 2017What this evidence cannot show
Mouse leukaemia models and cell lines; the pharmacology is early and the displacement is partial and gene-selective.
Evidence that argues against this (1)
- An independent route to the same endpoint - inhibiting taspase1-dependent cleavage - also stabilises MLL1 and displaces chimeras, and the two mechanisms have not been reconciled, so 'wild-type MLL1 abundance' is controlled by at least two pathways that may not act on the same pool. Zhao Z et al. 2019
- E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
Across one 50-bait proximity deposition, obligate complex members split widely: WDR5 11, ASH2L 9, RBBP5 0, DPY30 0, and for PAF1c CDC73 25, RTF1 15, PAF1 6, CTR9 3, LEO1 0 - a pattern showing that proximity recovery tracks detectability rather than complex membership.
Proximity-labelling bait-prey counts across a 50-bait deposition; ana-kmt2a-proximity-promiscuity-25-of-50 and concept-proximity-labelling-is-not-co-purification in this graph · Analysis run by this project over a published proximity deposition · Database 2026What this evidence cannot show
Restated from this project's own earlier analysis rather than recomputed; a single deposition, and curated proximity rows are usually spoke-expanded bait-to-prey assignments rather than binary interactions.
Evidence that argues against this (1)
- An alternative reading of the same split is that WDR5 and ASH2L genuinely have larger free pools or additional complexes - WDR5 in NSL, ASH2L across six SET1-family enzymes - so the split could be real biology rather than pure detectability. Yang Z et al. 2018
- E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Genetic deletion of Wdr5 impairs MLL-AF9 leukaemia cell growth and attenuates leukemogenesis in vivo, establishing a functional requirement for a WRAD subunit in a KMT2A-fusion-driven leukaemia without any demonstrated contact with the fusion.
Loss of Wdr5 attenuates MLL-rearranged leukemogenesis by suppressing Myc targets. · Liu L, Guo X, Wang Y, et al. · Biochim Biophys Acta Mol Basis Dis 2023What this evidence cannot show
One murine model; the mechanism given is ribosomal rather than fusion-related, which is itself compatible with this hypothesis being irrelevant to why WDR5 is required.
Evidence that argues against this (1)
- If the WDR5 requirement is ribosomal and p53-mediated, it is not transmitted through the wild-type KMT2A allele at all and this hypothesis explains nothing about why WDR5 matters here. Howard GC et al. 2024
the module's internal geometry, none of which contacts the fusion directly
a DPY30 dependency in this disease is not attributable to COMPASS while a second cellular pool exists and no separation-of-function allele does
partially - AKAP95 reaches the fusion directly through the retained fragment, not through WRAD, so a WRAD-adjacent protein can have a first-shell route
if the DPY30 route contributes nothing, AKAP95 is another case of WRAD not reaching the fusion - but by a route that does
- depends on →A BioID hit is a neighbour, not a complex member - and the two methods are known to disagree
The hypothesis is only meaningful under the four proximity rules: a second-shell passenger is exactly what a spatial assay reports as a hit and a binary assay misses.
- depends on →KMT2A / MLL1 — the gene that breaks
The proposed route is the wild-type KMT2A allele, which KMT2A fusions require for stable chromatin association and for leukaemia maintenance.
The only precedent: BioID of NUP98-PHF23 recovers MLL1, KMT2B, WDR5 and RBBP5 together, with WRAD arriving on a recruited methyltransferase.
The observed WRAD split across one 50-bait deposition - WDR5 11, ASH2L 9, RBBP5 0, DPY30 0 - is what abundance-weighted second-shell recovery of an obligate complex looks like.
- tests →Settle the WRAD question: tagged co-IP plus degron-timed proximity across three fusion classes
The degron arm is the experiment designed to decide this hypothesis in one pass.
- related to →Proximity-labelling map of KMT2A::AFF1 — the experiment nobody has run in twenty years of interactome work
Warns the existing TurboID proposal that a WRAD positive without a wild-type-allele degron arm would be uninterpretable.