EP300::ZNF384 - the commonest ZNF384 fusion, and it throws away the acetyltransferase

FULL TEXT READ as the author-accepted PDF deposited in the University of Tsukuba institutional repository, which Unpaywall lists as a free location for this DOI; the Nature Leukemia page itself was not reachable. This is a Letter with no PubMed abstract. The 1,027 aa and 110 kDa figures are PREDICTED from the fusion cDNA, not measured on a gel in what was read. The frequency is a composite estimate from two differently ascertained screens (55 selected BCP-ALL samples without conventional abnormalities, then 346 further paediatric ALL cases by RT-PCR), which the authors combine into one percentage - it is not a rate from a consecutive series. 'Lacks the other domains such as the HAT domain' is the only domain-loss statement; KIX, TAZ2 and IBiD are never individually named as lost.

• The most important disagreement about this fusion is not about its structure but about what the structure implies, and the founding paper flags it itself: 'Both EWSR1-ZNF384 and TAF15-ZNF384 have transforming properties in NIH3T3 cells, but do not alter expressions of known ZNF384 target genes.' A fusion that retains the complete DNA-binding protein and a partner fragment, but does not change the DNA-binding protein's known targets, is not obviously acting as a transcriptional activator at those targets - and the HAT deletion removes the most obvious alternative mechanism. Gocho's own 'may be involved... but further functional evaluation is necessary' is the correct level of confidence, and later work has still not established what the retained EP300 fragment does. Martini A et al. 2002

FULL TEXT READ via Europe PMC fullTextXML for PMC5931087. Eight of nine identical breakpoints is the strongest statement of breakpoint invariance available for any ZNF384 fusion, but note the ninth: a ZNF384 exon 2 breakpoint, i.e. the junction is NOT literally invariant even within this one partner. The 5.7% and 1.3% figures come from transcriptome sequencing of two differently selected cohorts and are not population rates. The domain statement is a sentence of interpretation - the paper does not map the breakpoint to a residue.

• The bromodomain-loss statement in this paper and the HAT-loss statement in Gocho are both interpretations of an exon-6 breakpoint, and NO source this beat read gives an explicit retained-or-lost verdict for EP300's KIX, TAZ2 or IBiD/NCBD domains. Since an exon-6 breakpoint at roughly residue 449 lies 117 residues before KIX begins, those three are certainly lost on the arithmetic - but that is this beat's arithmetic, not a published claim, and it is recorded as such. The honest state of the literature is: TAZ1 retained, HAT lost, bromodomain lost, everything else unstated. Zhu L et al. 2023

FULL TEXT READ via Europe PMC fullTextXML for PMC8550960. **THE COMPARATOR IS 'ALL REMAINING ZNF384-REARRANGED PATIENTS', NOT TCF3::ZNF384.** No separate numeric EFS, OS or cumulative-incidence figure for TCF3::ZNF384 appears in the body text; per-partner curves are in Figure 1 and Supplementary Table 4, which were not retrieved. This is a retrospective consortium collection spanning diagnoses from 1992 to 2018, NOT a consecutive cohort, with no total B-ALL denominator stated anywhere in the paper; MPAL cases were excluded by design and 18 cases identified at relapse or in refractory disease were excluded from the outcome analysis; outcome data were missing for 15 patients and partner-gene information for 25.

• Overall survival did NOT differ significantly (93% vs 90%, P = 0.289), which the authors attribute to relapses in the other partner groups being salvageable. So the EP300 advantage is in relapse and event-free survival and does not translate into a survival advantage in this series - a distinction that gets lost when '4% versus 18% relapse' is quoted alone. Add the confounding: EP300::ZNF384 patients were six years older on median and half as likely to be allocated high-risk, and the comparison is observational across 26 years of changing protocols in 16 consortia. Hirabayashi S et al. 2021

Read from the PMC rendering of PMC9414741; per-experiment n values and statistics were not extracted. The knockin expresses an Ep300-exon6::Znf384-V5-HA cDNA minigene from Ep300 intron 5, so it models the fusion transcript but not the chromosomal rearrangement, the reciprocal product or the native ZNF384 regulatory context. NSG-SGM3 mice are cytokine-humanised and biased toward myeloid output.

• This item contains its own contradiction and it is the reason the item is worth citing. EP300::ZNF384 is the strongest oncoprotein in the mouse HSPC assay and is NEGATIVE in the human CD34+ assay in the same paper, where TCF3::ZNF384 is the only positive. Any claim that EP300::ZNF384 is intrinsically more or less transforming than the other partners therefore depends entirely on which of the two systems is chosen, and the paper does not reconcile them. This also cuts against reading the clinical outcome difference as a potency difference: the fusion with the BEST patient outcome is the one that transforms mouse HSPCs unaided. Dickerson KM et al. 2022

Coordinates fetched live in this session, not recalled. They are cited here only to locate the published domain-loss statements on the sequence; no source maps the EP300 exon-6 breakpoint to a residue, so the mapping of exon 6 onto these coordinates is done by this beat through the fusion protein's stated total length and is marked as derived.

• UniProt boundaries for EP300 are curated domain calls and other databases place several of them a few residues differently, so a breakpoint falling near an edge could be assigned either side. That does not affect the conclusions here, because the derived boundary at about residue 449 sits 117 residues clear of the nearest downstream domain (KIX at 566) and 32 residues clear of the nearest upstream one (TAZ1 ending at 417). No source disagrees with the placement of these domains in a way that would move the verdict. Gocho Y et al. 2015

**THE PRIMARY FULL TEXT IS UNREACHABLE** - Wiley is gated to every route tried, there is no PMC copy and no Wayback snapshot. The first clause of this claim is from the published ABSTRACT. **The mapping clause - that the EP300-binding region lies in the ZNF384 portion - is NOT from the paper**: it is from the first author's Japanese doctoral-thesis summary of the same work, deposited in the Nagoya University repository (k13347_abstract.pdf), which Unpaywall lists as the only free location for this DOI. That is a different document by the same author about the same experiments, and it is flagged as such rather than cited as the paper. The mapping resolution there is 'the ZNF384 region' versus 'the EP300 region' and nothing finer - no residues, no individual zinc fingers. The GST bait was an EP300 fragment, amino acids 1053-1860, which is a region the EP300::ZNF384 fusion protein does not itself contain.

• Wild-type ZNF384 also binds EP300 - the finding is quantitative, not qualitative. Both the published abstract ('associated more strongly with EP300 than Wild-Z') and the thesis summary describe the fusion as a stronger binder than a measurably-binding wild type, so 'the fusion gains an EP300 contact' would be an overstatement of it. The thesis also reports that the fusion bound TIP60 and GCN5 more strongly than wild-type ZNF384 as well, which argues the effect is not EP300-specific even though EP300 gave the largest transcriptional enhancement. And the third-party source for the mapping result is a thesis abstract, not a peer-reviewed figure. Yamada C et al. 2024