Oncofusion

start / graph / exp-q173-ewsr1-znf384-versus-ewsr1-fli1-same-module-different-address

Experimentspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.q173-the-znf384-fusion-family

EWSR1::ZNF384 versus EWSR1::FLI1 - the same EWSR1 exon-7 module on two different DNA-binding proteins, and nobody has compared them

also called EWSR1 module portability experiment · same IDR different address

Two cancers - a bone tumour and a childhood leukaemia - appear to use the exact same piece of the EWSR1 protein, glued to two different DNA-binding proteins. Putting them side by side in one experiment would show whether that sticky piece does the same thing wherever it lands, or whether what it does depends entirely on where it is sent.

The same thing, for a scientist

**The natural experiment already exists in patients and has never been exploited.** Type 1 EWSR1::FLI1 joins EWSR1 exon 7 to FLI1 exon 6, and this graph's own `protein-ewsr1-domains` places the end of EWSR1 exon 7 at **codon 264** by Ensembl arithmetic [E1]. Martini et al. report EWSR1::ZNF384 as **EWSR1 exon 7 to CIZ exon 2 or exon 3** [E2]. **On that arithmetic the two fusions carry the identical EWSR1 fragment - residues 1-264, about 92% of the prion-like EWS activation domain - and differ only in the DNA-binding protein attached.** One causes a bone sarcoma and one a B-cell leukaemia. **This is the cleanest available test of whether a FET low-complexity domain is a portable activation module or an address-dependent one**, which is the question underlying this graph's `analogy-ewsr1-fli1-creates-its-addresses-at-polymorphic-repeats-while-kmt2a-fusions-occupy-invariant-cpg-islands`. **There is one partial precedent and it points away from portability:** Martini's 2002 experiment ran EWSR1-FLI1 as a positive control alongside EWSR1-CIZ and found the former induced its targets strongly while the latter induced nothing [E2].

Go deeper
Design

Step 0, and do not skip it. Align the junction sequence in Martini's Figure 1B to the current Ensembl EWSR1 transcript and confirm that 2002 'exon 7' is Ensembl exon 7 on ENST00000397938. The entire premise rests on this and it has not been checked. If the numbering differs, the two fragments differ by roughly the length of an exon and the experiment is not the one described here.

Constructs, all epitope-tagged, expressed at matched levels confirmed by western blot:

  • EWSR1(1-264)::FLI1 (type 1)
  • EWSR1(1-264)::ZNF384
  • Chimera swap 1: EWSR1(1-264)::ZNF384 with the ZNF384 zinc fingers replaced by the FLI1 ETS domain
  • Chimera swap 2: the reciprocal
  • FLI1 alone, ZNF384 alone, EWSR1(1-264) alone - the Martini both-halves controls

Cell systems, run in both because the point is address-dependence: a human CD34+ HSPC / NSG-SGM3 arm (the ZNF384 native context, validated by Dickerson et al.) and a mesenchymal or Ewing-permissive arm (the FLI1 native context).

Readouts:

  1. Occupancy by anti-tag CUT&RUN. Score GGAA-microsatellite occupancy and [GC]AAAAA-consensus occupancy separately for every construct. The prediction under portability is that the EWSR1 module contributes nothing to address choice; under address-dependence, EWSR1(1-264)::ZNF384 should show no GGAA-microsatellite preference at all.
  2. Condensate behaviour - the assays the Ewing comparator in this graph already defines: live-cell single-molecule imaging for hub formation, 1,6-hexanediol sensitivity, and tyrosine-mutant rescue of the EWSR1 module. None of these has ever been performed on any ZNF384 fusion.
  3. BAF/mSWI/SNF recruitment by co-IP and by ARID1A occupancy, since BAF retargeting is the established EWSR1::FLI1 mechanism and Dickerson's RIME list contains ARID1A for ZNF384 fusions.
  4. Transformation and lineage output in each host system.
What the outcomes would mean, and the trap in reading them
resultinterpretation
both fusions form hubs with the same hexanediol and tyrosine dependence, but occupy completely different genomic addresses set by their DNA-binding halvesThe FET module is portable; the address is the partner's. This is the tidy result and it makes the Ewing condensate literature importable to ZNF384.
EWSR1(1-264)::ZNF384 forms no hub and shows no BAF recruitmentThe module is address-dependent - it works at GGAA microsatellites because multivalent repeat arrays support it, and a single-site zinc-finger consensus does not. This would explain why FET partners are only ~12% of ZNF384 cases while dominating Ewing biology.
the chimera swaps behave like their DNA-binding half, not their EWSR1 halfstrong support for address-dependence and the more informative arm of the design
EWSR1(1-264)::ZNF384 forms hubs but they repress rather than activateconsistent with the Ewing optimum result already in this graph - Chong et al. found increasing low-complexity interaction toward phase separation REPRESSES EWS::FLI1 targets - and would be the most surprising and most useful outcome

The trap: this experiment will produce a condensate readout for a ZNF384 fusion where none exists today, and a positive will be tempting to generalise to the whole family. It cannot be. FET partners are about 12% of ZNF384 cases - 17 TAF15 and 6 EWSR1 of 193 partner-typed - and EP300 and TCF3, at 74%, have no prion-like domain of this kind. A condensate mechanism established for EWSR1::ZNF384 would be a mechanism for a minority of a minority.

Facts and figures
designSTEP 0 (prerequisite, cheap, not yet done): align the EWSR1-CIZ junction sequence in Martini 2002 Figure 1B to Ensembl ENST00000397938 to confirm that 2002 exon 7 is the current exon 7. THEN: epitope-tagged, expression-matched parallel expression of EWSR1(1-264)::FLI1, EWSR1(1-264)::ZNF384, two domain-swap chimeras exchanging the ETS and zinc-finger DNA-binding modules, and the three single-half controls (FLI1 alone, ZNF384 alone, EWSR1(1-264) alone), in both a human CD34+ HSPC/NSG-SGM3 system and a Ewing-permissive mesenchymal system.
readoutAnti-tag CUT&RUN with GGAA-microsatellite and [GC]AAAAA-consensus occupancy scored separately; live-cell single-molecule imaging for hub formation; 1,6-hexanediol sensitivity; tyrosine-mutant rescue of the EWSR1 module; BAF recruitment by co-IP and ARID1A occupancy; transformation and lineage output in each host.
controlsThe three Martini single-half constructs, which are the historical both-halves control and were reported only as 'not shown'; expression matching by western blot before any readout; the reciprocal domain swaps, which are the arm that actually discriminates portability from address-dependence; a Ewing cell line as a positive control for every condensate assay, since all of them were developed there.
cost tierhigh
timeline12-18 months, dominated by the two host systems and the imaging arm. Step 0 is one afternoon and gates everything.
measurements
count0.000count: 0proportion0.119proportion: 0.119residue264.0residue: 264
measurements — the numbers behind the chart
[
  {
    "quantity": "residue",
    "subject": "EWSR1 fragment shared by EWSR1::FLI1 type 1 and EWSR1::ZNF384",
    "value": 264,
    "units": "aa (last residue, derived)",
    "n": null,
    "universe": "Ensembl exon arithmetic on ENST00000397938 computed inside this graph for EWSR1::FLI1 and applied to EWSR1::ZNF384 on the basis that both use EWSR1 exon 7",
    "null": null,
    "note": "DERIVED AND UNVERIFIED ACROSS TRANSCRIPT MODELS. This single number is the premise of the whole experiment and step 0 exists to check it."
  },
  {
    "quantity": "proportion",
    "subject": "FET-family partners among partner-typed ZNF384 fusions",
    "value": 0.119,
    "units": "fraction",
    "n": 193,
    "universe": "TAF15 17 plus EWSR1 6 of 193 partner-typed Ponte di Legno cases",
    "null": 0.741,
    "note": "the null is the EP300-plus-TCF3 share, 74.1%. This bounds how far any condensate result from this experiment can generalise: to about an eighth of the family, and probably less, since cytogenetically visible FET fusions are over-ascertained."
  },
  {
    "quantity": "count",
    "subject": "published condensate, FRAP, hexanediol or hub-imaging experiments on any ZNF384 fusion protein",
    "value": 0,
    "units": "studies",
    "n": null,
    "universe": "this beat's reading: 112 PubMed records enumerated for ZNF384 with fusion or rearrangement terms in title and abstract, plus targeted searches on CIZ, NMP4, EWSR1 and TAF15 with leukemia",
    "null": null,
    "note": "A BOUNDED ABSENCE AT TITLE AND ABSTRACT LEVEL ONLY. A condensate experiment reported as a supplementary panel of a paper about something else would not be found this way, and this project's rule 80 records exactly that failure mode for measurement-type queries."
  }
]
How this could bite us back
  • The likeliest misuse of a positive result is importing Ewing condensate-directed therapeutics into ZNF384-rearranged leukaemia. This graph's own comparator says that would be hazardous in a specific way: increasing low-complexity-domain interaction toward phase separation REPRESSES EWS::FLI1 targets, so the dose-response is not monotone and a condensate-disrupting agent could increase target-gene output. Any such agent must be dose-scanned in both directions.
  • The FET low-complexity domain is a shared, wild-type module present in FUS and EWSR1 as well as TAF15, so an agent aimed at it is aimed at housekeeping protein chemistry in every cell, not at a neoantigen. FUS mutations cause ALS; the essential-function toxicity is not hypothetical.
  • And a scope warning that belongs on the proposal, not the paper: whatever this experiment finds applies to at most an eighth of ZNF384-rearranged patients, and nothing at all to the EP300 and TCF3 fusions that are three quarters of them.
Still open
  1. Step 0 first: is Martini's EWSR1 exon 7 the same exon as Ensembl's? The whole design collapses to an ordinary comparison of two unrelated fusions if it is not, and the check costs an afternoon.
  2. Is there a mesenchymal system in which EWSR1::ZNF384 can even be expressed meaningfully, and a haematopoietic system in which EWSR1::FLI1 can? The design asks each fusion to be tested outside its native lineage, which is the point, but a fusion that is simply toxic or silent in a foreign cell yields an uninterpretable null.
  3. Does ZNF384's [GC]AAAAA consensus occur in multivalent arrays anywhere in the genome, analogous to GGAA microsatellites? If it does, those loci are where a FET::ZNF384 fusion would be predicted to build a hub, and the analysis can be done computationally before any construct is made.
  4. Would the answer even matter for patients? FET partners are about 12% of this family at most. A positive result would be mechanistically illuminating and therapeutically narrow, and that should be said out loud when the experiment is proposed rather than after it is funded.
Every source behind this entry (4)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    In the type 1 EWSR1::FLI1 fusion, EWSR1 exon 7 joins FLI1 exon 6, and recomputing the Ensembl EWSR1 coding exons on transcript ENST00000397938 places the end of exon 7 at codon 264, so EWSR1 contributes residues 1-264: most of the prion-like EWS activation domain (annotated 1-285, with 31 degenerate tandem repeats over 8-285), with the IQ motif, the RRM at 361-447, the RanBP2-type zinc finger at 518-549 and the PY-NLS at 639-656 all lost. UniProt independently annotates an EWSR1 translocation site at residue 265 for EWSR1/ATF1.

    What this evidence cannot show

    support_location is 'this-project' because 264 is arithmetic computed inside this graph, not a figure printed in any paper. The source entry itself records that the junction codon is chimeric so '264 versus 265 is a convention', and that the EAD boundary at 1-285 is a compositional convention with no structural support - InterPro covers no model over residues 1-225. The figure was computed for EWSR1::FLI1 and its application to EWSR1::ZNF384 is an assumption this experiment's step 0 exists to check.

    Evidence that argues against this (1)
    • The weak joint is the exon numbering across two eras of transcript models, and it is load-bearing for the whole design. Martini 2002 gives exon numbers against a 2002 model; this graph's 264 comes from a current Ensembl model. If the numbering differs by one exon, the two EWSR1 fragments are not identical and the 'same module' premise fails. This beat did NOT run that check, which is why it is step 0 of the design rather than a background assumption. Martini A et al. 2002
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    EWSR1::ZNF384 joins EWSR1 exon 7 to CIZ (ZNF384) exon 2, with a variant joining EWSR1 exon 7 to CIZ exon 3; in the same study EWSR1-FLI1 served as a positive control and induced mmp3, eat2 and manic fringe strongly in NIH3T3 cells while neither TET-CIZ fusion induced any of them, and neither TET-CIZ fusion transactivated a CIZ-binding-site reporter despite binding the CIZ consensus in EMSA.

    Recurrent rearrangement of the Ewing's sarcoma gene, EWSR1, or its homologue, TAF15, with the transcription factor CIZ/NMP4 in acute leukemia. · Martini A, La Starza R, Janssen H, Bilhou-Nabera C, Corveleyn A, Somers R, Aventin A, Foa R, Hagemeijer A, Mecucci C, Marynen P · Cancer Res 2002
    PMID 12359745
    What this evidence cannot show

    FULL TEXT READ from an Internet Archive snapshot of the AACR HighWire page dated 2009-09-29; aacrjournals.org is Cloudflare-blocked and no DOI exists for this pre-DOI article. The comparison was made in NIH3T3 fibroblasts, a system native to neither disease, and the targets assayed were EWSR1-FLI1 targets plus the CIZ targets known in 2002. So the existing partial comparison is confounded: the positive control was scored on ITS OWN targets, which is the correct design for a control and the wrong design for a comparison of two fusions. expected_absent_terms names ZNF384 because the gene is CIZ/NMP4 throughout.

    Evidence that argues against this (1)
    • This item is the only existing comparison and it is a weak one, in a direction that matters. Martini scored both fusions on a target list assembled from EWSR1-FLI1 biology and 2002-era CIZ biology, so 'EWSR1-CIZ induced nothing' may mean the right targets were not on the list - and later work found real ZNF384-fusion targets (SALL4, ID2, RGS1, CREB5, FLT3) none of which could have been tested. The proposed experiment must therefore score both fusions on BOTH target sets and on unbiased occupancy, or it will repeat the 2002 asymmetry with better instruments. Zhao X et al. 2022
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    The condensate assays this experiment would apply are already defined and controlled in the EWSR1::FLI1 system: the EWSR1 prion-like domain retargets the BAF/mSWI/SNF complex to tumour-specific GGAA-microsatellite enhancers in a manner dependent on tyrosines required for phase transitions; live-cell single-molecule imaging shows low-complexity-domain hubs that stabilise DNA binding and recruit Pol II with no detectable phase separation at physiological concentrations; and EWS::FLI1 requires a narrow optimum of low-complexity-domain interaction, with increases toward phase separation REPRESSING GGAA-microsatellite target genes and nucleolar sequestration inhibiting transcription and transformation.

    What this evidence cannot show

    support_location is 'this-project' because this item cites an existing entry in this graph as the assembled source rather than re-deriving the three primary papers, which this beat did not re-read. Anyone running the experiment should go to those primaries directly. The assays were developed in Ewing models; their transfer to a haematopoietic system is not routine, and hexanediol in particular is a blunt reagent with documented off-target effects on chromatin and on kinase signalling.

    Evidence that argues against this (1)
    • The Ewing comparator's own headline is the reason to expect a non-obvious answer, and it should be read as a warning rather than a foundation: Chong et al. found NO detectable phase separation under physiological conditions, and found that pushing the system toward phase separation REPRESSES targets. So an experiment designed to ask 'does EWSR1::ZNF384 form condensates?' is asking a question the Ewing field has already partly reframed. The better-posed version asks about hub formation and DNA-binding stabilisation at a defined occupancy, not about demixing. Zhao X et al. 2022
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    FET-family partners account for only about 12% of partner-typed ZNF384 fusions - TAF15 17 and EWSR1 6 of 193 cases in the Ponte di Legno series - while EP300 (83) and TCF3 (60) together account for 74%.

    Clinical characteristics and outcomes of B-ALL with ZNF384 rearrangements: a retrospective analysis by the Ponte di Legno Childhood ALL Working Group. · Hirabayashi S, Butler ER, Ohki K, Kiyokawa N, Bergmann AK, Moricke A, Boer JM, Cave H, Cazzaniga G, Yeoh AEJ, et al., Moorman AV, Manabe A · Leukemia 2021
    PMID 33692463DOI
    What this evidence cannot show

    FULL TEXT READ via Europe PMC fullTextXML for PMC8550960. This is cited to bound the GENERALISABILITY of the proposed experiment, not to support its design. A retrospective consortium collection with heterogeneous detection methods across 26 years, partner information missing for 25 of 218 patients, MPAL excluded by design.

    Evidence that argues against this (1)
    • The 12% figure could be an underestimate if FET partners are harder to detect - but the likelier bias runs the other way. TAF15::ZNF384 is associated with a recognisable t(12;17) karyotype and has been looked for since 2002, whereas EP300::ZNF384 is cryptic to G-banding (five of six founding cases had a normal karyotype) and was only found in 2015. A partner spectrum assembled partly by FISH and karyotype therefore OVER-represents the cytogenetically visible FET fusions, which makes 12% a ceiling rather than a floor and strengthens the generalisability warning. Gocho Y et al. 2015
Connected to