TAF15 (RBP56, TAF2N) - the FET-family, prion-like partner that donates a low-complexity N-terminus to ZNF384

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

• TAF15 is one of three FET paralogues with overlapping roles in transcription and RNA processing, so any therapeutic strategy aimed at the FET low-complexity domain is aimed at a shared, wild-type, housekeeping module and not at a neoantigen. The essential-function toxicity here is not speculative: FUS mutations cause ALS and EWSR1 has documented developmental requirements.
• The Ewing comparator in this graph shows that INCREASING low-complexity-domain interaction toward phase separation REPRESSES EWS::FLI1 target genes. A drug designed to disrupt condensates could therefore land on the wrong side of an optimum and increase target-gene output. Any condensate-directed intervention against a FET::ZNF384 fusion must be dose-scanned in both directions, not assayed at one concentration.
• This entry supports no intervention of its own. The interventional content of the ZNF384 family sits in target-flt3-in-znf384-rearranged-leukaemia, where the bite-back is recorded against the actual drug.

1. Where exactly is TAF15 truncated in TAF15::ZNF384, and does the fusion carry the whole low-complexity region 1-237, the RRM at 234-320, or the RanBP2 zinc finger at 354-385? Nyquist 2011 says two different TAF15 breakpoints exist but the abstract does not name them and this beat could not reach the full text. Until that is settled no condensate argument about this fusion can be made, because the size of the disordered module is unknown.
2. Does TAF15::ZNF384 form condensates or dynamic hubs, and does it retarget BAF the way EWSR1::FLI1 does? The Ewing comparator in this graph establishes a narrow optimum of low-complexity-domain interaction beyond which target genes are REPRESSED, so the prediction is not monotone and the experiment is worth doing precisely because the naive direction may be wrong.
3. EWSR1 and TAF15 both partner ZNF384; FUS, the third FET protein, has not been reported to in anything this beat read. Is that a real biological asymmetry or an ascertainment artefact of how few TAF15::ZNF384 and EWSR1::ZNF384 cases exist? A systematic search of fusion catalogues for FUS::ZNF384 was not run here and should be.
4. Caspase-3/-7 cleave TAF15 inside its low-complexity region at 106-DQPD/Y-110 and cleave the TAF15-CIZ/NMP4 fusion at a single site. If chemotherapy-induced apoptosis severs the disordered module from the ZNF384 DNA-binding half, does the cleaved fragment still bind chromatin, and is a DNA-binding domain with no activation module a dominant negative or an inert occupant?

1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   UniProtKB Q92804 (RBP56_HUMAN, gene TAF15, synonyms RBP56 and TAF2N) is a reviewed 592 aa human entry whose feature table records: Disordered 1-237, 324-356 and 373-592; Domain RRM 234-320; Zinc finger RanBP2-type 354-385; Region 407-575 '21 X approximate tandem repeats of D-R-[S,G](0,3)-G-G-Y-G-G'; compositional bias Low complexity at 1-21, 28-91, 104-129 and 140-150, Gly residues at 176-187, 341-350 and 415-578; and PRMT1-dependent asymmetric dimethylarginine at residues 206, 528, 535 and 570.

   UniProtKB entry Q92804 (RBP56_HUMAN) - sequence, feature table and compositional bias; retrieved live 2026-09-24 by q173 · UniProt Consortium 2026
   ▸What this evidence cannot show

   Every coordinate here was read out of the live JSON in this session, not recalled. The 'Disordered' and 'Low complexity' features are sequence-composition predictions, not measurements, so the boundary at residue 237 is a convention with no experimental edge - exactly as this graph already records for the EWSR1 EAD at 1-285 in protein-ewsr1-domains. No PDB structure covers the low-complexity region; pdb_ids is empty in this entry for that reason.

   ▸Evidence that argues against this (1)

   • The N-terminal region is annotated as disordered by a predictor, and this graph's own EWSR1 entry established that InterPro returns no model of any kind over the equivalent EWSR1 region, i.e. the domain has no independent structural support and no sharp edge. The same caution must apply to TAF15 1-237: read it as 'roughly here'. This is a limitation of the annotation rather than a disagreement between sources - no source was found that assigns TAF15 a folded N-terminal domain, and none is expected to. UniProt Consortium 2026
2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

   The founding report of the ZNF384 fusion family describes TAF15 and EWSR1 - two of the three TET/FET proteins - as recurrent partners of the transcription factor CIZ/NMP4 (now ZNF384) in acute leukaemia, arising through t(12;17)(p13;q11) or the variant t(12;22)(p13;q12); the fusions transform NIH3T3 cells.

   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

   ABSTRACT ONLY - the full text of this 2002 Cancer Research paper was not retrievable through PMC or Europe PMC in this session, so no breakpoint, exon or residue number is taken from it here. expected_absent_terms names ZNF384 deliberately: the gene was called CIZ/NMP4 in 2002 and this paper does not contain the string ZNF384, which is precisely why a ZNF384-only literature search misses the founding report - the same failure mode this project already recorded for KMT2A versus MLL.

   ▸Evidence that argues against this (1)

   • This paper's own conclusion cuts against the reading a reader will bring to it. Verbatim from the abstract: 'The fusions possess transforming properties in NIH3T3 cells but do not affect the expression of CIZ target genes, suggesting a contribution to oncogenesis that is independent of the transactivating properties of the fusion protein.' The intuitive model - a DNA-binding protein supercharged by a fused activation domain - is contradicted by the founding experiment, which found no change in CIZ target-gene expression. Any claim that the FET N-terminus works here as a transactivation module has to answer this. Martini A et al. 2002
3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   TAF15::ZNF384 is a recurrent, independently confirmed fusion in acute lymphoblastic leukaemia and in childhood mixed-phenotype acute leukaemia, reported by at least four independent groups; two different TAF15 breakpoints have been described.

   Identification of the TAF15-ZNF384 fusion gene in two new cases of acute lymphoblastic leukemia with a t(12;17)(p13;q12). · Nyquist KB, Thorsen J, Zeller B, Haaland A, Troen G, Heim S, Micci F · Cancer Genet 2011
   PMID 21504714DOI
   ▸What this evidence cannot show

   Abstract only. The abstract states the translocation 'has been reported in 25 cases and its putative molecular consequence, the formation of a TAF15-ZNF384 fusion gene, in only six cases' and that the results 'suggest that two different breakpoints in TAF15 may be involved' - but the abstract does not give the exon numbers, so NO breakpoint coordinate is recorded here. Independent confirmations read at abstract level only: Kim J et al. Ann Lab Med 2016 (PMID 27139620), Yao L et al. Cancer Genet 2017 in childhood MPAL (PMID 28279306), Georgakopoulos N et al. In Vivo 2018 (PMID 30150451).

   ▸Evidence that argues against this (1)

   • THE CYTOBAND IS NOT AGREED, and a reader matching on karyotype string will miss cases. The same TAF15::ZNF384 event is written t(12;17)(p13;q11) by Martini 2002 (PMID 12359745) and Grammatico 2013 (PMID 23206227), t(12;17)(p13;q12) by Nyquist 2011 (PMID 21504714) and Kim 2016 (PMID 27139620), and t(12;17)(p13;q21) by Georgakopoulos 2018 (PMID 30150451) and Thorsen 2011 (PMID 21962896). Three different 17q bands for one fusion. TAF15 is placed at 17q12 by the gene record, so q11 and q21 look like banding-resolution differences rather than different genes, but this beat did not resolve which is correct and no claim in this graph should be keyed to the band. Georgakopoulos N et al. 2018
4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   A reviewed, organism-restricted gene_exact query for TAF15 returns TWO human records and the first is not TAF15: Q16514 (TAF12_HUMAN, 161 aa) carries TAF15 as a legacy gene-name alias and is returned ahead of Q92804 (RBP56_HUMAN, 592 aa), which is the FET-family TAF15.

   UniProtKB REST search executed live 2026-09-24 by q173: query=gene_exact:TAF15 AND organism_id:9606 AND reviewed:true, fields=accession,id,length,protein_name,gene_names · UniProt Consortium 2026
   ▸What this evidence cannot show

   This is a claim about this project's own query, not about a paper's contents, hence support_location 'this-project'. The result ordering is UniProt's default relevance ordering and could change; what will not change is that two reviewed human records answer to the symbol. The discriminator is length: 161 aa versus 592 aa.

   ▸Evidence that argues against this (1)

   • The collisions-registry check returned nothing for TAF15, and 'nothing' could mean either 'no row' or 'the check is broken' - this project's rule-88 history contains exactly one instance of that read failing silently in the false-clear direction. The check used here was the corrected, case-sensitive, word-boundary form over d['collisions'] and d['non_gene_antigens'], and it was validated in the same run by returning a known-present row for TCF3, so the instrument was demonstrably working when it returned no row for TAF15. The honest residual risk is that a TAF15 row exists in some other file this beat did not read. Oncofusion Knowledge Graph 2026
5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   TAF15 is cleaved by caspases-3 and -7 at a single consensus site, 106-DQPD/Y-110, and the oncogenic TAF15-CIZ/NMP4 fusion proteins are also cleaved by caspases-3 and -7 at one single site.

   TAF15 and the leukemia-associated fusion protein TAF15-CIZ/NMP4 are cleaved by caspases-3 and -7. · Alves J, Wurdak H, Garay-Malpartida HM, Harris JL, Occhiucci JM, Belizario JE, Li J · Biochem Biophys Res Commun 2009
   PMID 19426707DOI
   ▸What this evidence cannot show

   Abstract only, and the work is an in-vitro caspase-substrate screen with site-directed mutagenesis plus staurosporine-treated Jurkat cells - not a leukaemia model and not a ZNF384-rearranged cell. The site 106-DQPD/Y-110 falls inside the N-terminal low-complexity region (UniProt Disordered 1-237), so cleavage would sever the prion-like module from the rest of the protein; the abstract does not say which TAF15-CIZ/NMP4 variant was tested, where its single site lies, or which side the ZNF384 moiety ends up on.

   ▸Evidence that argues against this (1)

   • The abstract's own internal disagreement is the counter-evidence, verbatim: 'Site-directed mutagenesis revealed the consensus sequence 106DQPD/Y110 as the only site recognized by these caspases. Surprisingly, TAF15 was cleaved at more than one site in staurosporine-treated Jurkat cells.' A single-site claim derived from recombinant protein is contradicted by the cellular experiment reported in the same abstract. Nothing has been found that tests caspase cleavage of TAF15::ZNF384 in a leukaemia cell, and this beat did not run a dedicated caspase-plus-ZNF384 search, so this item is a lead and not a result. Alves J et al. 2009

• fuses with →ZNF384 - a C2H2 transcription factor that is the invariant 3' half of a whole B-ALL fusion subtype, including AKAP8::ZNF384

  TAF15::ZNF384, t(12;17); 17 of 193 partner-typed Ponte di Legno cases (9%)

• analogous to →EWSR1 — domain architecture and residue coordinates

  FET-family paralogue: both donate an N-terminal prion-like low-complexity region to a DNA-binding partner

• related to →EWSR1::FLI1 — the comparator that shows condensate formation has an optimum, and that forcing a fusion into the nucleolus switches it off

  the comparator that would have to be repeated for TAF15::ZNF384; no condensate experiment exists for any ZNF384 fusion

• part of →TAF15::ZNF384 - the FET-partner version, where the naming order in the literature is a trap and the founding experiment says the transactivation domain is not doing the work

  TAF15 is the 5' partner, contributing its SYQG-rich N-terminal transactivating domain

• ← part ofTAF15::ZNF384 - the FET-partner version, where the naming order in the literature is a trap and the founding experiment says the transactivation domain is not doing the work

  TAF15 exon 6 or exon 9 contributes the SYQG-rich N-terminal transactivating domain

• ← related toEWSR1::ZNF384 versus EWSR1::FLI1 - the same EWSR1 exon-7 module on two different DNA-binding proteins, and nobody has compared them

  the other FET partner; the same design would apply to TAF15::ZNF384 once its breakpoint residue is known