Oncofusion

start / graph / hyp-wdr5-partition-nsl-versus-compass

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q5-interactome5

WDR5 is partitioned, not shared: NSL and SET1/MLL complexes compete for one pool, so perturbing NSL changes how much WDR5 reaches the wild-type KMT2A complex the fusion depends on

also called WDR5 partition hypothesis · NSL versus COMPASS competition for WDR5

WDR5 is a small hub protein that several big chromatin machines need. A structural study showed two of those machines cannot hold it at the same time. If that is true inside a leukaemia cell, then interfering with one machine should change how much WDR5 is available to the other - including the normal KMT2A complex that the leukaemia fusion relies on.

The same thing, for a scientist

The KANSL1/WDR5/KANSL2 structure establishes that **WDR5's interactions with the MOF-containing NSL complex and with MLL/COMPASS methyltransferase complexes are mutually exclusive**, and the authors propose that WDR5 is not a shared subunit but a component that assembles distinct complexes [E1]. Two earlier studies describe WDR5 and HCF1 as **shared** subunits of NSL and MLL/SET and reason from that sharing to functional crosstalk, reporting NSL as acting upstream of MLL/SET to promote H3K4me2 [E2][E3]. Both accounts can be true at once only if the coupling is **competitive**. **The hypothesis:** in a KMT2A-rearranged cell, WDR5 is limiting and partitioned between NSL and SET1/MLL complexes, so that acute depletion of the NSL scaffold KANSL1 **increases** WDR5 occupancy in SET1/MLL complexes and at their target loci, and acute depletion of a SET1/MLL scaffold correspondingly increases WDR5 in NSL. **Nobody has measured this in any mammalian cell, let alone a leukaemic one.** The prediction has direct therapeutic sign consequences: this graph carries a WIN-site target dossier, and if the partition is real, an NSL-directed agent would push WDR5 towards the complex the disease depends on. Prior probability is set at **0.25** - the structural claim is solid, but 'mutually exclusive in a crystal and in binding assays' becoming 'limiting and competitive in a cell' is a substantial extra step that this graph has seen fail before.

Go deeper
Why this is worth a hypothesis node rather than a sentence

Because the two possible worlds make opposite predictions about an intervention the project is already considering.

  • If WDR5 is shared (the 2010/2013 framing 23), NSL and MLL/SET are collaborators, NSL acts upstream, and reducing NSL activity reduces H3K4 methylation at co-regulated genes. An NSL-directed agent would push in the therapeutic direction.
  • If WDR5 is partitioned (the 2014 structure 1), NSL and MLL/SET are competitors for a limiting hub, and reducing NSL releases WDR5 to SET1/MLL complexes. An NSL-directed agent would push in the anti-therapeutic direction.

The graph currently has no position and cannot get one from the literature, because the two published accounts were never reconciled by their authors. A hypothesis node with a falsifier is the honest way to hold that, and the experiment that decides it is small.

There is a second reason. This graph's WDR5 entry establishes that WDR5 binds the part of KMT2A every fusion discards, and that WDR5 is the single worst-scoring gene in the lineage-controlled selectivity ranking (+0.1984, rank 17,587 of 17,728 4). Those facts have been used to argue that WDR5 is not a target. The partition hypothesis does not challenge that conclusion but changes what WDR5 measures: if WDR5 is a limiting shared resource, its own dependency profile would be expected to look flat regardless of how important the partition is, because every cell needs the hub and only the allocation differs.

What makes it falsifiable, and the result that would kill it

The hypothesis is killed by any of the following:

  1. Acute KANSL1 degradation does not increase WDR5 co-immunoprecipitation with RBBP5/ASH2L (the WRAD core shared by SET1/MLL complexes) within a timeframe too short for transcriptional adaptation. If the pools do not exchange, there is no partition to exploit or fear.
  2. WDR5 is not limiting. If free WDR5 is in excess of the combined binding capacity of NSL and SET1/MLL complexes, exclusivity at the level of a single WDR5 molecule has no consequence at the level of a cell. A simple quantification of WDR5 against the summed abundance of its partners would settle it.
  3. WDR5 occupancy at HOXA9 and MEIS1 is unchanged by KANSL1 loss. Exchange without redistribution to the relevant loci would make the partition real but irrelevant here.

Any one of those outcomes closes the question. That is a cheap falsification profile, which is the main argument for running it.

Connections this would change if it held
  • The WIN-site dossier. SETD1B carries its own WIN motif at 1745-1750 5, and KANSL1 and KANSL2 bind WDR5 through conserved linear motifs 1. A WIN-site inhibitor is therefore a pan-partition agent that would release WDR5 from everything at once, which is a different pharmacology from what a KMT2A-directed WIN inhibitor is usually imagined to be.
  • The NUP98 cross-family result. Five NUP98 fusions bind both MLL1 and NSL 6. If WDR5 is partitioned, those fusions are engaging two assemblies that compete for the same hub, which is a strange arrangement and might be the point - or might indicate the partition does not operate the way the structure suggests.
  • The MSL/NSL discrimination. MOF sits in both MSL and NSL and writes different marks in each 3. MSL does not contain WDR5. So a WDR5 perturbation should shift MOF activity towards MSL-type output (H4K16ac only) and away from NSL-type output (H4K5/K8ac as well), which is a measurable, mark-level prediction that nothing in this literature has tested.
Facts and figures
statementIn KMT2A-rearranged leukaemia cells WDR5 is limiting and partitioned between the NSL complex and SET1/MLL complexes, so acute loss of the NSL scaffold KANSL1 increases WDR5 occupancy in SET1/MLL complexes and at their target loci, and shifts MOF output from NSL-type (H4K5/K8/K16ac) towards MSL-type (H4K16ac only).
originliterature-gap
prior probability0.25
therapeutic sign consequenceIf true, an NSL-directed agent RELEASES WDR5 towards the SET1/MLL complexes the disease depends on - the anti-therapeutic direction. If false, NSL and MLL/SET are collaborators and an NSL-directed agent pushes the therapeutic way. The two worlds differ in sign, not magnitude.
strongest existing objectionFive NUP98 fusions bind MLL1 and NSL simultaneously (E6), which is awkward for a strict partition model, though not incompatible with it since the two assemblies could each carry their own WDR5.
predictions
[
  "Acute KANSL1 degradation increases WDR5 co-immunoprecipitation with RBBP5 and ASH2L within 2-6 hours, before transcriptional adaptation.",
  "WDR5 CUT&RUN signal at HOXA9 and MEIS1 increases after acute KANSL1 loss.",
  "H4K5ac and H4K8ac fall while H4K16ac is relatively preserved after KANSL1 loss, because MOF activity redistributes towards the MSL configuration.",
  "The reciprocal holds: acute degradation of a SET1/MLL scaffold increases WDR5 in NSL immunoprecipitates.",
  "Quantitative proteomics shows total WDR5 is below the summed abundance of its NSL and SET1/MLL binding partners - that is, the hub is limiting."
]
experiments
[
  "exp-wdr5-pool-exchange-kansl1-degron"
]
analyses
[
  "ana-interactome5-enumeration"
]
falsifiers
[
  "Acute KANSL1 degradation does not increase WDR5 association with RBBP5/ASH2L on a timescale shorter than transcriptional adaptation.",
  "WDR5 is present in molar excess of its binding partners, so exclusivity at the molecular level has no cellular consequence.",
  "WDR5 occupancy at HOXA9 and MEIS1 is unchanged by KANSL1 loss, making any exchange irrelevant to this disease.",
  "The MSL/NSL mark discrimination (H4K16ac only versus H4K5/K8/K16ac) fails to hold in cells, removing the mark-level readout."
]
How this could bite us back
  • If the hypothesis is true, the most obvious intervention it suggests - inhibit NSL - is contraindicated, because it feeds WDR5 to the complex the disease relies on. A hypothesis whose confirmation closes a door is still worth testing, but it should be described that way from the start.
  • A WIN-site inhibitor would act on every WDR5-binding motif at once, including SETD1B's and KANSL1's, so its net effect on the partition is not predictable from a KMT2A-centric model.
  • The experiment proposed uses an acute degron. Degron systems can perturb the tagged protein's abundance at baseline, and a partial baseline depletion of KANSL1 would blunt the very exchange being measured.
  • Even a confirmed partition would be a mechanism, not a window. Every protein involved is broadly required, and nothing in this hypothesis identifies a KMT2A-selective vulnerability.
Still open
  1. Is WDR5 limiting in a leukaemic blast? Everything here turns on that and it has never been measured in this cell type.
  2. Does WDR5 exchange between complexes on a timescale that a degron experiment can resolve, or is the partition set at assembly and effectively static?
  3. If the partition is real, does it explain why WDR5's own dependency profile is so flat despite its central position? A limiting hub would look flat under any allocation.
  4. Would a WIN-site inhibitor release WDR5 from NSL as well as from SET1/MLL complexes, and what net direction would that produce? Nothing in the WIN-inhibitor literature read here addresses the NSL half.
Every source behind this entry (7)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The KANSL1/WDR5/KANSL2 structure shows WDR5 is recruited into the NSL complex via conserved linear motifs of KANSL1 and KANSL2; structure-based KANSL1 mutants in transgenic flies show the KANSL1-WDR5 interaction is required for NSL assembly, efficient recruitment to target promoters and fly viability; and the interactions of WDR5 with the MOF-containing NSL complex and with MLL/COMPASS histone methyltransferase complexes are mutually exclusive, leading the authors to propose that WDR5 assembles distinct histone-modifying complexes rather than being a shared subunit.

    Structural analysis of the KANSL1/WDR5/KANSL2 complex reveals that WDR5 is required for efficient assembly and chromatin targeting of the NSL complex · Dias J, Van Nguyen N, Georgiev P, Gaub A, Brettschneider J, Cusack S, Kadlec J, Akhtar A · Genes & Development 28(9):929-42 2014
    PMID 24788516DOI
    What this evidence cannot show

    Abstract fetched from PubMed and read on 2026-09-19; full text not read, every identifier copied from that fetched record. Crystallography plus **Drosophila** genetics. 'Mutually exclusive' is a statement about binding-site overlap and binding data, not a measurement of pool allocation in a mammalian cell, and certainly not in a leukaemic one. **The entire hypothesis rests on extending that statement to cellular stoichiometry, which is the step most likely to fail.**

    Evidence that argues against this (1)
    • Two earlier studies treat WDR5 and HCF1 as shared subunits of NSL and MLL/SET and build a functional-crosstalk argument on that sharing, which is a different model with different predictions. Neither side reconciled the accounts, and this hypothesis exists because the field left the question open rather than because it answered it. Cai Y et al. 2010
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Depletion of the NSL subunits hMOF or NSL1/KANSL1 reduced both H4K16ac and H3K4me2 near the ANKRD2 transcription start site, whereas depletion of RbBP5 reduced H3K4me2 only - consistent with NSL acting upstream of MLL/SET to regulate H3K4me2 at certain promoters.

    Crosstalk between NSL histone acetyltransferase and MLL/SET complexes: NSL complex functions in promoting histone H3K4 di-methylation activity by MLL/SET complexes · Zhao X, Su J, Wang F, Liu D, Ding J, Yang Y, Conaway JW, Conaway RC, Cao L, Wu D, Wu M, Cai Y, Jin J · PLoS Genetics 9(11):e1003940 2013
    PMID 24244196DOI
    What this evidence cannot show

    Abstract fetched from PubMed and read on 2026-09-19; full text not read, identifiers copied from that fetched record. Single exemplar locus (ANKRD2, a skeletal-muscle gene) in human cell lines, by knockdown. **This experiment cannot distinguish 'NSL acts upstream' from 'NSL and MLL/SET compete for a limiting shared component'** - which is precisely the ambiguity this hypothesis proposes to resolve.

    Evidence that argues against this (1)
    • If the true relationship is competitive, this result's direction is puzzling: removing a competitor should INCREASE H3K4me2, not decrease it. Either the competition model is wrong, or the ANKRD2 promoter needs NSL for something else - most plausibly the acetylation that precedes methylation - and that is a specific, testable alternative. Dias J et al. 2014
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The NSL complex comprises nine subunits, of which WDR5 and HCF1 are shared with MLL/SET-family complexes and MCRS1 with INO80; assembly of MOF into MSL versus NSL controls substrate specificity, with MSL-associated MOF acetylating nucleosomal H4 almost exclusively on K16 and NSL-associated MOF also acetylating K5 and K8.

    Subunit composition and substrate specificity of a MOF-containing histone acetyltransferase distinct from the male-specific lethal (MSL) complex · Cai Y, Jin J, Swanson SK, Cole MD, Choi SH, Florens L, Washburn MP, Conaway JW, Conaway RC · Journal of Biological Chemistry 285(7):4268-72 2010
    PMID 20018852DOI
    What this evidence cannot show

    Abstract fetched from PubMed and read on 2026-09-19; full text not read, identifiers copied from that fetched record. Proteomics plus in vitro nucleosomal HAT assays. Used here for two things only: the roster (which contains WDR5) and the mark-level discrimination between MSL and NSL output, which supplies one of the hypothesis's testable predictions.

    Evidence that argues against this (1)
    • The mark discrimination was measured in vitro on nucleosomal substrates and has not, in anything read here, been demonstrated in cells. If it does not hold in vivo, the H4K5/K8-versus-K16 readout proposed as a test of this hypothesis would not work. Dias J et al. 2014
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    In the project's genome-wide KMT2A-rearranged selectivity scan, WDR5 has a difference of +0.1984 at rank 17,587 of 17,728 on a pan-essentiality of -1.159; KANSL1 is +0.0685 at rank 15,913 and KANSL2 +0.2265 at rank 17,631. None of the WDR5-binding proteins in this beat shows a KMT2A-selective requirement.

    Genome-wide KMT2A-rearranged selectivity scan, 17,728 genes (DepMap Chronos via Open Targets) · Broad Institute DepMap / Open Targets Platform; analyses/depmap-kmt2a-selective-scan-full 2026
    What this evidence cannot show

    Rows read out of analyses/depmap-kmt2a-selective-scan-full/results.json on 2026-09-19 in a scripted pass. Eleven KMT2A-rearranged against 43 other leukaemia lines, AML-heavy, none infant-derived. WDR5's pan-essentiality of -1.159 places it near the floor where the difference statistic compresses. **This row is cited as a reason the hypothesis is hard to test genetically, not as support for it.**

    Evidence that argues against this (1)
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The SETD1B catalytic domain is required in MLL-rearranged leukaemia: its disruption narrows H3K4me3 breadth and reduces MYC expression.

    PMID 40341256DOI
    What this evidence cannot show

    The SETD1B leukaemia result is from an abstract fetched from PubMed and read on 2026-09-19 (full text not read); the WIN-motif coordinates are from the UniProt record Q9UPS6 fetched the same day and recorded in `gene-setd1b`. **The two facts come from different sources and are combined here by this entry**, which is an inference of the graph and not a claim of either source. **Split 2026-09-19.** This item previously bundled three UniProt sequence annotations with the paper's biological finding under the paper's citation alone. The full text (PMC12208907, 147,057 characters) contains **zero occurrences of 1745, 1798, 1827, 'WIN motif', 'RxxxRR' or 'UniProt'** - those coordinates are UniProt's and are now cited as such in E9. Surfaced by `scripts/reading-queue.py`; fifth instance of a citation supporting only part of its claim.

    Evidence that argues against this (2)
    • That SETD1B has its own WIN motif does not establish that SETD1B competes for the same WDR5 pool as NSL under physiological stoichiometry - only that it uses the same pocket. The hypothesis needs the pool to be limiting, and no measurement of that exists. Dias J et al. 2014
    • Bundling a database coordinate with a wet-lab finding under one citation makes the pair look like a single sourced statement, and it is the failure mode `verify-citations.py` is structurally blind to: the PMID resolves, the title matches, and half the sentence still has no source. The claim-support score caught it only because the residue numbers could not appear in the abstract. UniProt Consortium 2026
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Five NUP98 fusion proteins containing no KMT2A - NUP98-HOXA9, NUP98-HOXD13, NUP98-NSD1, NUP98-PHF23 and NUP98-TOP1 - physically interact with MLL1 and with the NSL complex, and MLL1 is required for the proliferation of NUP98-HOXA9 cells in vitro and in vivo.

    NUP98 Fusion Proteins Interact with the NSL and MLL1 Complexes to Drive Leukemogenesis · Xu H, Valerio DG, Eisold ME, Sinha A, Koche RP, Hu W, Chen CW, Chu SH, Brien GL, Park CY, Hsieh JJ, Ernst P, Armstrong SA · Cancer Cell 30(6):863-878 2016
    PMID 27889185DOI
    What this evidence cannot show

    Abstract fetched from PubMed and read on 2026-09-19; full text not read, identifiers copied from that fetched record. Murine models and human expression signatures; no KMT2A fusion tested. Cited here because a fusion engaging both assemblies simultaneously is an awkward fact for a strict partition model and is recorded as such rather than omitted.

    Evidence that argues against this (1)
    • If NUP98 fusions engage MLL1 and NSL at once, then either the two assemblies coexist on the same fusion without sharing a WDR5 molecule - which is possible and is exactly what the structure permits - or the partition is weaker in cells than in the crystal. This observation is the strongest existing argument against the hypothesis and it is recorded in the hypothesis itself. Dias J et al. 2014
  7. E9established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt annotates SETD1B (Q9UPS6, 1,966 aa) with a WDR5-interaction (WIN) motif at residues 1745-1750, an RxxxRR motif at 1798-1803, a SET domain at 1827-1944 and a Post-SET domain at 1950-1966, with a disordered region at 1767-1800 spanning the gap between the two motifs.

    What this evidence cannot show

    Fetched by the coordinator from the UniProt REST API on 2026-09-19 and read off the feature table directly, not recalled. UniProt sequence features are annotations of varying provenance - some experimental, some by similarity - and this fetch did not record the evidence code behind each one, so the coordinates are reliable as UniProt's position and not automatically as an experimental result.

    Evidence that argues against this (2)
    • A WIN motif annotated in the sequence is not a demonstrated WDR5 interaction in cells, and this graph has a standing rule that a database fact is opposable on scope. The motif's presence makes a SETD1B-WDR5 contact possible and says nothing about whether it happens in a KMT2A-rearranged cell, at what occupancy, or in competition with the other WIN-motif proteins that share the same site. UniProt Consortium 2026
    • The RxxxRR motif sits inside a region UniProt itself calls disordered (1767-1800), so the two annotations overlap in a way that complicates reading either as a fixed structural element. A motif in a disordered segment is a sequence pattern, not a binding site with a conformation. Leukemia 2025
Connected to