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

FULL TEXT READ from an Internet Archive snapshot dated 2009-09-29 of the original HighWire full-text page, because aacrjournals.org is Cloudflare-blocked to every route tried; NO DOI exists for this pre-DOI AACR article in PubMed, Europe PMC or Crossref, so the archived URL and the PMID are the only locators. Seven cases total, of which five had t(12;17) and two t(12;22). Exon-level breakpoints only - the paper gives NO amino-acid coordinates for any junction. The negative transactivation result was obtained in NIH3T3 and Ba/F3 cells, neither of which is a human B-cell precursor. expected_absent_terms names ZNF384 because the gene is called CIZ/NMP4 throughout.

• THE TARGET LIST IS THE WEAKNESS AND IT IS DECISIVE. The transactivation experiment tested the CIZ targets known in 2002 - a reporter carrying CIZ binding sites, plus mmp3, mmp7, eat2 and manic fringe - and this graph's own gene-znf384 entry records that UniProt still flags the MMP/COL1A1 promoter targets as inferred 'By similarity' rather than experimentally established in human. Later work identified real ZNF384-fusion targets that could not have been tested then: SALL4 and ID2 (Yamamoto 2019), RGS1 and CREB5 (Yamada 2024) and FLT3 through a distal enhancer (Zhao 2022). Martini et al. offer this reading themselves - 'or that MMPs are not the true targets of CIZ'. The negative therefore bounds the mechanism much less tightly than it first appears. Yamada C et al. 2024

The body text is paywalled (Elsevier) and the old journal domain is now a squatted site; the breakpoint statement was read from the FIGURE 3 LEGEND recovered from an Internet Archive snapshot of the journal's article page, hence support_location 'figure'. Nucleotide positions are in the paywalled Results and were NOT obtained. The FISH probes were RP11-151M4 over ZNF384 and RP11-362K1 over TAF15 at 17q12. n = 2 cases.

• The abstract's own closing sentence is the appropriate caution and it has not been answered in the fifteen years since: 'Whether the two variants of the TAF15-ZNF384 fusion that these correspond to are in any way hematologically or prognostically different, is unknown.' Two different TAF15 breakpoints, exon 6 and exon 9, retain different amounts of the TAF15 low-complexity region, and if that region is the functional contribution then the two variants should differ. Nobody has separated them in any cohort, including the 17 TAF15 cases in the Ponte di Legno series, which reports no breakpoint-level analysis. Nyquist KB et al. 2011

FULL TEXT READ via Europe PMC fullTextXML for PMC8550960. The MRD zero is from Table 1 and the denominator for the TAF15 arm is not stated in the row as extracted, so '0 of the TAF15 group' cannot be turned into a rate and no confidence interval is offered. MRD data existed for only 77 of 218 patients overall, and the difference across partner groups was not significant (p = 0.491). No separate survival figure is published for TAF15::ZNF384; the paper's outcome contrast is EP300 versus all remaining patients.

• A zero MRD count in a subgroup whose denominator is not stated is exactly the kind of number that should not travel, and it is recorded here only because it is the one TAF15-specific clinical datum in the largest series. Applying the rule of three to a plausible denominator of around eight TAF15 patients with MRD data bounds the true positivity rate below roughly 37%, which is not informative. This item should be read as 'no TAF15-specific outcome signal has been published', not as 'TAF15::ZNF384 has good MRD clearance'. Hirabayashi S et al. 2021

ABSTRACT ONLY. The inference that 'ZNF384-TAF15' in the title is chromosome order rather than a 5'-3' fusion name is THIS BEAT'S READING, not a statement in the paper: it rests on the fact that the order matches the karyotype, that the evidence is FISH, and that no junction is reported. A reader who wanted to argue that this case genuinely had a 5'-ZNF384 fusion could not be refuted from the abstract - they could only be told that nothing in it supports the claim either. The patient died of gram-negative sepsis during induction.

• The one case in the literature where a 5'-ZNF384 fusion IS sequenced and confirmed - ZNF384::TEX41, out of frame, in Zaliova et al. - shows that such fusions are not impossible, only rare and detectable when a junction is read. So the trap runs both ways: a chromosome-ordered name must not be read as an orientation claim, and 'the name is chromosome order' must not become a reflex that dismisses a real inversion. The discriminator is whether a sequenced junction with a stated reading frame exists. Zaliova M et al. 2021

A REVIEW, read in full text via Europe PMC, and cited here as a map of the breakpoint range rather than as a primary source for any individual case. 'Between exons 4 and 10' is a range compiled across reports the review does not individually enumerate at that point, so it cannot be converted into a distribution. The SYQG terminology is Martini's, carried forward. Note that this review's statement that the cytoband ranges 'from q11 to q21' confirms the banding inconsistency recorded in gene-taf15 E3 rather than resolving it.

• The review gives no residue coordinates, and neither does any primary source this beat reached, so how much of the TAF15 low-complexity region (UniProt Disordered 1-237) the fusion actually carries is unknown for every reported case. That matters because the whole FET analogy to EWSR1::FLI1 rests on the size and composition of the retained disordered module. A review sentence naming an exon range is not a substitute, and a beat that treats 'SYQG domain retained' as settling the architecture has assumed the thing the Ewing comparator says is quantitative. UniProt Consortium 2026