Oncofusion

start / graph / analogy-how-a-fusion-meets-baf-ss18-ssx-ews-fli1-and-the-kmt2a-silence

Mechanismstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q47-swi-snf

Two published ways a fusion oncoprotein engages BAF - be a subunit, or recruit it with a disordered domain - and KMT2A has never been tested for either

also called BAF retargeting by fusion oncoproteins · SS18-SSX and EWS-FLI1 comparison

In two other fusion-driven cancers, scientists worked out exactly how the fusion protein takes over the SWI/SNF machine: in one the fusion is part of the machine, in the other it grabs the machine with a sticky disordered tail. Nobody has asked whether MLL fusions do the second thing, even though their partners have exactly that kind of tail.

The same thing, for a scientist

**Route 1 - the fusion is a subunit.** SS18-SSX, 78 residues of SSX fused to the BAF subunit SS18, competes for assembly with wild-type SS18, produces a complex lacking BAF47, binds the SOX2 locus, reverses polycomb repression and activates SOX2; increasing wild-type SS18 reassembles normal complexes, retargets them off SOX2 and stops proliferation, and only two amino acids of SSX are required [E1]. The downstream consequence is a second dependency: cBAF perturbation makes synovial sarcoma and SMARCB1-null rhabdoid tumour synthetic-lethal for ncBAF, and BRD9 degradation reverses the oncogenic programme and inhibits tumour progression, with other sarcoma subtypes unaffected [E2][E3]. **Route 2 - recruitment by a prion-like domain.** EWS-FLI1 recruits BAF to tumour-specific enhancers; this is a neomorphic property compared with wild-type FLI1, depends on tyrosines required for EWSR1 prion-like-domain phase transitions, and **fusing short EWSR1 fragments to FLI1 is sufficient to recapitulate BAF retargeting and EWS-FLI1 activities** [E4]. **KMT2A: neither has been tested.** No KMT2A partner is a BAF subunit, so route 1 is structurally excluded. Route 2 is not excluded by anything - AFF1 and MLLT3 contribute large intrinsically disordered regions to their fusions - and the literature is silent: `BAF AND (MLL-AF4 OR MLL-AF9 OR KMT2A-AFF1) AND immunoprecipitation` returns 0 with operands and positive controls checked [E5]. **This entry records that silence as an untested question, not as a negative result.**

Go deeper
The experiment that transfers directly

Boulay et al.'s decisive construct is the smallest one: short EWSR1 fragments fused to FLI1 recapitulate BAF retargeting 4. The equivalent in this disease is a KMT2A-N-terminal fragment fused to a defined AFF1 or MLLT3 region, assayed for BAF co-immunoprecipitation and for BAF occupancy at fusion target loci, with the tyrosine-mutant control that paper established.

The standing caution this project applies to all condensate claims applies here in full: 1,6-hexanediol sensitivity and classical FRAP cannot distinguish phase separation from binding to spatially clustered sites, and hexanediol itself condenses chromatin in living cells. So the readout must be BAF occupancy and association, not droplet appearance - which is what the EWSR1 work measured and is why it is usable.

What the analogy does and does not license

Licensed: the architectural claim that perturbing one mSWI/SNF module creates dependence on another, which is what makes BRD9 a target in two cBAF-perturbed sarcomas 23 and is a plausible frame for the KMT2A-rearranged BRD9 dependency (rank 39 of 17,728) 6.

Not licensed: any statement that a KMT2A fusion binds, retargets or remodels BAF. Two named routes exist, one is structurally impossible here, the other is untested. Writing the analogy as a mechanism would be exactly the error this graph has made before - a mechanism written from a neighbouring disease and then cited as if it were measured in this one.

Facts and figures
mechanism classcross-family comparison of fusion-oncoprotein engagement with mSWI/SNF
kmt2a statusuntested question, not a negative result
routes
[
  {
    "route": "fusion is a BAF subunit",
    "example": "SS18::SSX in synovial sarcoma",
    "consequence": "altered complex lacking BAF47, SOX2 activation, ncBAF/BRD9 synthetic lethality",
    "applies_to_kmt2a": "structurally excluded - no KMT2A partner is a BAF subunit"
  },
  {
    "route": "fusion recruits BAF via a low-complexity/prion-like domain",
    "example": "EWSR1::FLI1 in Ewing sarcoma",
    "consequence": "BAF retargeted to tumour-specific enhancers; short EWSR1 fragments suffice",
    "applies_to_kmt2a": "NOT EXCLUDED and never tested; AFF1 and MLLT3 contribute large disordered regions"
  }
]
bounded absence
{
  "query": "BAF AND (MLL-AF4 OR MLL-AF9 OR KMT2A-AFF1) AND immunoprecipitation",
  "records": 0,
  "operand_controls": {
    "MLL-AF9 AND immunoprecipitation": 28,
    "SWI/SNF AND immunoprecipitation": 317,
    "leukemia": 405093
  },
  "date": "2026-09-20",
  "tool": "PubMed E-utilities esearch"
}
How this could bite us back
  • If a KMT2A fusion does recruit BAF, then a BAF-directed agent would be acting on the fusion's own machinery and the combination logic with menin inhibitors changes from complementary to overlapping - and could be antagonistic.
  • The synovial sarcoma precedent shows restoration of normal complex composition can be therapeutic; an inhibitor that destabilises complexes could in principle do the opposite of what is intended, and no leukaemia experiment has looked for that.
  • Borrowing a target from a cBAF-perturbed sarcoma into a disease with no known cBAF lesion is the kind of transfer that produces a clean preclinical story and a failed trial.
Still open
  1. Does a KMT2A::AFF1 or KMT2A::MLLT3 fusion co-immunoprecipitate any BAF subunit from a leukaemia line? This is a one-week experiment that the field appears never to have published.
  2. Is a KMT2A-N-terminal fragment fused to a short AFF1 disordered region sufficient to retarget BAF, as short EWSR1 fragments are for FLI1?
  3. Is the KMT2A-rearranged BRD9 dependency conditional on something resembling cBAF perturbation, as it is in rhabdoid tumour and synovial sarcoma, or does it have an independent cause?
Every source behind this entry (6)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    SS18-SSX competes for assembly with wild-type SS18, forming an altered BAF complex lacking BAF47 that binds and activates SOX2 by reversing polycomb repression; restoring wild-type SS18 reassembles normal complexes and stops proliferation, and the mechanism depends on only two amino acids of SSX.

    PMID 23540691DOI
    What this evidence cannot show

    Synovial sarcoma systems; the wild-type SS18 rescue is over-expression.

    Evidence that argues against this (1)
    • This route requires the fusion partner to be a complex subunit and therefore cannot occur for any KMT2A fusion; it is a boundary case rather than a template. Boulay G et al. 2017
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    ncBAF complexes are synthetic-lethal targets specific to synovial sarcoma and malignant rhabdoid tumour, both of which have cBAF perturbation, and ncBAF maintains gene expression at retained CTCF-promoter sites in a manner distinct from fusion-oncoprotein-bound complexes.

    A non-canonical SWI/SNF complex is a synthetic lethal target in cancers driven by BAF complex perturbation. · Michel BC, D'Avino AR, Cassel SH, Mashtalir N, et al. · Nat Cell Biol 2018
    PMID 30397315DOI
    What this evidence cannot show

    Sarcoma and rhabdoid lines only; the synthetic lethality is conditional on cBAF perturbation.

    Evidence that argues against this (1)
    • KMT2A-rearranged leukaemia has no known cBAF perturbation, so if the conditionality is strict the analogy does not reach this disease. Chatterjee SS et al. 2018
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    A domain-focused CRISPR screen identified the BRD9 bromodomain as a critical dependency in synovial sarcoma; BRD9 is a component of SS18-SSX-containing BAF complexes, co-localises extensively with SS18-SSX on the genome, and a small-molecule BRD9 degrader downregulates the oncogenic programme and inhibits tumour progression in vivo while other sarcoma subtypes are unaffected.

    Targeted degradation of BRD9 reverses oncogenic gene expression in synovial sarcoma. · Brien GL, Remillard D, Shi J, Hemming ML, et al. · Elife 2018
    PMID 30431433DOI
    What this evidence cannot show

    Synovial sarcoma; the degrader's selectivity claim rests on the sarcoma-subtype panel tested.

    Evidence that argues against this (1)
    • The BRD9 dependency there is tied to BRD9 being in the fusion-containing complex; no KMT2A fusion has been shown to be in any BAF complex, so the mechanistic basis does not transfer even though the target does. Hohmann AF et al. 2016
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    EWS-FLI1 recruits BAF to tumour-specific enhancers as a neomorphic property compared with wild-type FLI1, dependent on tyrosines necessary for EWSR1 prion-like-domain phase transitions, and fusing short EWSR1 fragments to FLI1 is sufficient to recapitulate BAF retargeting and EWS-FLI1 activities.

    Cancer-Specific Retargeting of BAF Complexes by a Prion-like Domain. · Boulay G, Sandoval GJ, Riggi N, Iyer S, et al. · Cell 2017
    PMID 28844694DOI
    What this evidence cannot show

    Ewing sarcoma; the phase-transition attribution rests on tyrosine mutants. This project's standing condensate caution applies: the common assays cannot separate phase separation from binding to clustered sites, so the recruitment is better supported than the physical model for it.

    Evidence that argues against this (1)
    • If the mechanism is really clustered-site binding rather than phase separation, then the transferability to KMT2A fusions depends on sequence-specific targeting rather than on having a disordered domain - and KMT2A fusions target through CXXC and elongation machinery, not through a microsatellite. Shi J et al. 2013
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    No published experiment tests whether a KMT2A fusion engages BAF: `BAF AND (MLL-AF4 OR MLL-AF9 OR KMT2A-AFF1) AND immunoprecipitation` returns 0 records.

    What this evidence cannot show

    Bounded absence. Bounded to PubMed on 2026-09-20 over that term set. It cannot see an unlabelled co-IP inside a paper framed at complex level, a supplementary mass-spectrometry table, or an unpublished negative result - and an unpublished negative is a likely outcome for an experiment this obvious.

    Evidence that argues against this (1)
    • The graph's own KMT2A IntAct pull (218 rows, 101 human partners) contains no SWI/SNF subunit, which is consistent with the search but inherits the same limitation: IntAct curates what was published. Mo X et al. 2022
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    BRD9 ranks 39 of 17,728 for KMT2A-rearranged selectivity with pan-essentiality above zero, and is one of four SWI/SNF panel genes depleted in both lineage halves.

    SWI/SNF panel lineage-split DepMap dependency scan, 28 genes · Oncofusion knowledge graph, analyses/swi-snf-lineage-split/run.py · DepMap Chronos gene effect via Open Targets GraphQL 2026
    What this evidence cannot show

    Eight myeloid and five B-lineage index lines; z is not an effect size and BRD9's raw difference is -0.3896.

    Evidence that argues against this (1)
    • A BRD9 dependency in KMT2A-rearranged lines may have nothing to do with the sarcoma architecture; the published AML mechanism for BRD9 is MYC transcription in AML generally, not a fusion-conditional synthetic lethality. Hohmann AF et al. 2016
Connected to