ZNF384::TEX41 - the case that breaks the family's architecture: ZNF384 is the 5' partner, the 3' partner is a lncRNA, and the fusion is out of frame
also called ZNF384-TEX41 · ZNF384-LINC00953
Every textbook description of this leukaemia subtype says ZNF384 is the second half of the fusion and gets a powerful new front end. In one child it was the other way round: ZNF384 was the front half, the back half came from a gene that does not make a protein at all, and the join was out of frame - yet the leukaemia looked exactly like the others.
The same thing, for a scientist
In a consecutive Czech cohort of 643 children with BCP-ALL or B/myeloid MPAL, Zaliova et al. found three cases with the ZNF384-rearranged gene-expression signature but no canonical fusion. One of them (patient 1) carried a **noncanonical fusion with ZNF384 as the 5' partner**: the large 5' portion of ZNF384, including the coding sequence for all of its zinc fingers, fused to the 3' portion of **TEX41**. The fusion is **out of frame** and terminates early, contributing only **nine amino acids encoded by TEX41 intron 1** - annotated as **ZNF384 p.481_516 > SFQDYCACQ**. TEX41 (NCBI Gene **401014**, 2q22.3, alias LINC00953) is a **long intergenic non-protein-coding RNA** and is described in the paper as a gene not shown to be recurrently involved in leukaemia [E1][E2]. **This single case contradicts three separate parts of the standard description of the ZNF384 family at once**: ZNF384 is not always 3', the partner does not always contribute a coactivator or an intrinsically disordered activation module (here it contributes nine residues of nonsense from a non-coding gene), and the product is not an in-frame chimera. What it preserves is the zinc-finger array and the transcriptional output - the case co-clustered with canonical ZNF384-rearranged leukaemias by unsupervised hierarchical clustering on RNA-seq [E1].
What the fusion actually is
Verbatim from the paper 1:
"Routine analysis of RNA-seq data using in-house bioinformatic pipelines for single-nucleotide variants, indels, and fusion gene detection revealed noncanonical ZNF384 fusion (ZNF384 as the 5' partner) in 1 patient"
"Patient 1 harbored a novel ZNF384-TEX41 fusion consisting of the large 5' portion of ZNF384 (including the coding sequence for all 6 zinc fingers) fused to the 3' portion of TEX41, a gene not shown to be recurrently involved in leukemia. The out-of-frame fusion led to an early stop codon with only 9 amino acids (AAs) encoded by TEX41 intron 1 (ZNF384 p. 481_516 > SFQDYCACQ)."
So the protein product is, functionally, a C-terminally truncated ZNF384 that stops at residue 480 and carries a nine-residue nonsense tail. TEX41 is not a partner in any meaningful protein sense: it is the genomic sequence that happens to supply a stop codon.
Read against gene-znf384's UniProt record (Q8TF68, 577 aa, eight annotated C2H2 fingers ending at 455), the truncation at 480 removes only the last 97 residues, all of them C-terminal to the whole zinc-finger array. That is the same part of the protein removed by the last-exon indels in the two sister cases (variant-znf384-last-exon-indel), and it is why all three phenocopy the canonical fusions.
Why this is the most informative case in the family, and the trap in reading it
The intuitive model of a ZNF384 fusion is DNA-binding protein + borrowed coactivator = deregulated transcription. This case has a DNA-binding protein and no borrowed anything, and it still produces the subtype's gene-expression signature and its B/myeloid immunophenotype 13.
The conclusion this licenses is narrow and should be stated narrowly: the common element across canonical fusions and these noncanonical lesions is loss of the ZNF384 C-terminus, with the zinc fingers retained - not acquisition of a specific activation domain. The paper says so more cautiously than that:
"DNA binding is likely preserved in ZNF384 with novel aberrations, as the 6 zinc fingers are mostly intact. However, as the more distal C-terminal portion of the protein affected by the novel aberrations lacks well-defined functional domains, functional studies would be necessary to directly demonstrate the biological consequences of these aberrations and to compare them with canonical ZNF384 fusions."
The trap: n = 1. One out-of-frame fusion in one child does not establish that the C-terminus is the thing that matters in the canonical fusions, where a real coactivator IS fused on and where interface-znf384-fusion-med23 records a gained Mediator contact. Both mechanisms can be true, in different cases, producing the same expression signature by different routes. What the case DOES establish, with no room for argument, is that the architecture is not invariant.
| gene5 | ZNF384 |
|---|---|
| gene3 | TEX41 |
| cytoband | ZNF384 12p13 to TEX41 2q22.3; the paper reports the fusion from RNA-seq and this entry records no karyotype |
| uniprot5 | Q8TF68 |
| join gap note | scripts/check-joins.mjs reports 'no protein entry for 3' gene TEX41' for this fusion and WILL CONTINUE TO. That is correct and must not be fixed. TEX41 (NCBI Gene 401014) is a long intergenic non-protein-coding RNA, it has no UniProt accession, and this fusion is out of frame - the TEX41 side contributes nine residues translated from intron 1 before a stop codon, which is not a protein and has no domains. Creating a protein entry for TEX41 to make the 3D viewer lane draw would be fabricating an object that does not exist. The lane should render as a truncated ZNF384 with a nine-residue tail, or not at all. |
| frame | OUT OF FRAME - an early stop codon nine residues into TEX41 intron 1 sequence |
diseases
[ "childhood B-other BCP-ALL with the ZNF384-rearranged expression signature" ]
breakpoints
[
{
"id": "bp-znf384-480-tex41-intron1",
"gene5_exon_last": null,
"gene3_exon_first": null,
"protein_annotation": "ZNF384 p.481_516 > SFQDYCACQ",
"frequency_pct": null,
"population": "1 of 643 consecutive Czech childhood BCP-ALL/B-myeloid MPAL, Dec 2010 to Dec 2020",
"domains_retained_5": [
"all annotated ZNF384 C2H2 zinc fingers (the paper counts 6; UniProt Q8TF68 annotates 8, ending at residue 455)"
],
"domains_lost_5": [
"the ZNF384 C-terminus distal to residue 480, which the paper notes lacks well-defined functional domains"
],
"domains_retained_3": [],
"phenotype_notes": "ZNF384-rearranged expression signature by unsupervised hierarchical clustering; myeloid-marker positive; CD66c positive, which is unusual for canonical ZNF384r"
}
]measurements — the numbers behind the chart
[
{
"quantity": "count",
"subject": "noncanonical ZNF384 fusions with ZNF384 as the 5' partner",
"value": 1,
"units": "patients",
"n": 1,
"universe": "643 consecutive children diagnosed with BCP-ALL or B/myeloid MPAL in the Czech Republic between December 2010 and December 2020, of whom 158 were classified B-other",
"null": null,
"note": "the point of this number is not its size but its existence: the canonical description of the family states ZNF384 is always the 3' partner"
},
{
"quantity": "residue",
"subject": "last ZNF384 residue retained in ZNF384::TEX41",
"value": 480,
"units": "aa position",
"n": 1,
"universe": "protein annotation ZNF384 p.481_516 > SFQDYCACQ given in the paper's Results; ZNF384 canonical is 577 aa (UniProt Q8TF68)",
"null": null,
"note": "the truncation removes 97 C-terminal residues, all of them distal to the annotated zinc-finger array which ends at 455"
},
{
"quantity": "count",
"subject": "amino acids contributed by the TEX41 side of the fusion",
"value": 9,
"units": "aa",
"n": 1,
"universe": "the out-of-frame junction reads into TEX41 intron 1 and terminates; sequence SFQDYCACQ",
"null": null,
"note": "for comparison, EP300 contributes on the order of hundreds of residues in the canonical fusion. Nine residues of out-of-frame intronic sequence is not a functional module."
}
]- This entry touches no intervention directly, but it carries a diagnostic bite-back that does have clinical consequences: a patient with this lesion is negative on every ZNF384 fusion assay, including RNA-seq fusion calling, and would be classified B-other. If ZNF384-rearranged status is ever used to assign therapy - and the Ponte di Legno data make that a live question - then the assay defines the population, and this class of patient falls outside it while having the same disease.
- A fusion panel that reports 'ZNF384 rearrangement: not detected' on this case is not wrong about the fusion; it is wrong about the patient. Any protocol built on fusion detection needs the expression signature as a backstop, which is exactly what Zaliova recommend.
- Is a C-terminally truncated ZNF384 sufficient to produce the subtype, or does it require the same cooperating lesions the canonical fusions require? Dickerson's mouse work showed EP300::ZNF384 alone made serially transplantable leukaemia while other ZNF384 fusions needed NRAS G12D; nobody has put a p.481-truncated ZNF384 into an HSPC at all.
- Do the canonical fusions ALSO work by removing the ZNF384 C-terminus, with the fused coactivator a second and separable event? The experiment is a clean one: express full-length ZNF384 truncated at 480 alongside EP300::ZNF384 in the same HSPC system and compare occupancy, expression signature and transformation. If the truncation alone reproduces the signature, the 'partner supplies the activation module' story is wrong for the whole family.
- How many ZNF384-rearranged-like cases in published cohorts are noncanonical lesions that fusion callers missed? Zaliova found two of their three by MANUAL read-level reanalysis after automated indel and fusion callers missed them, and reported that no similar ZNF384 mutation appears in any published genomic study or public database. That is an instrument-failure claim about the field, and it is testable by reanalysing deposited ZNF384r-like RNA-seq.
- Does the truncated protein still bind the canonical [GC]AAAAA consensus, and does it still contact Mediator? The zinc fingers are intact so DNA binding is expected, but the MED23 contact recorded in interface-znf384-fusion-med23 was found for a fusion oncoprotein and not for wild-type ZNF384, so a truncation with no partner is the discriminating reagent.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Zaliova et al. identified a noncanonical ZNF384 fusion with ZNF384 as the 5' partner: ZNF384-TEX41, comprising the large 5' portion of ZNF384 including the coding sequence for all 6 zinc fingers fused to the 3' portion of TEX41; the fusion is out of frame, generates an early stop codon and contributes only 9 amino acids encoded by TEX41 intron 1, annotated ZNF384 p.481_516 > SFQDYCACQ. The patient co-clustered with canonical ZNF384-rearranged leukaemias in unsupervised hierarchical clustering of RNA-seq expression data.
A novel class of ZNF384 aberrations in acute leukemia. · Zaliova M, Winkowska L, Stuchly J, Fiser K, Triska P, Zwyrtkova M, Hrusak O, Starkova J, Sramkova L, Stary J, Trka J, Zuna J · Blood Adv 2021What this evidence cannot show
FULL TEXT READ, via Europe PMC fullTextXML for PMC8579251, retrieved 2026-09-24; the sentences quoted in this entry are from the Results and discussion section of that XML. n = 1 for this fusion and it has been reported once. The fusion junction was called from RNA-seq and the supplemental material (which the paper says holds the detailed molecular analysis) was NOT retrieved, so the exon-level breakpoint is not recorded here - only the protein-level annotation the main text gives. No functional experiment was performed on this protein.
Evidence that argues against this (1)
- The paper itself declines to conclude what this entry is tempted to conclude, and it is right to: 'as the more distal C-terminal portion of the protein affected by the novel aberrations lacks well-defined functional domains, functional studies would be necessary to directly demonstrate the biological consequences of these aberrations and to compare them with canonical ZNF384 fusions.' No functional study has been done. It remains possible that the expression-signature match is driven by something this beat has not identified - a second lesion, or a lineage/cell-of-origin effect - rather than by the ZNF384 truncation, and with n = 1 nothing excludes that. Zaliova M et al. 2021
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
TEX41 (testis expressed 41) is NCBI Gene 401014, located at 2q22.3 on chromosome 2, with the alias LINC00953, and is annotated as 'long intergenic non-protein coding RNA 953 | testis expressed 41 (non-protein coding)'.
What this evidence cannot show
The Gene ID was resolved by esearch on the symbol in this session and NOT taken from memory. That mattered: the first Gene ID this beat reached for was wrong and resolved to a WITHDRAWN chromosome-15 record ('DISCONTINUED: This record has been withdrawn by NCBI'), which would have put the partner on the wrong chromosome. The record carries no RefSeq summary text, so the only functional statement available is the 'non-protein coding' designation itself.
Evidence that argues against this (1)
- A 'non-protein coding' annotation is a statement about the current RefSeq model, not proof that no peptide is made from the locus; lncRNA loci with small open reading frames are a recognised class. But the direction of that uncertainty does not rescue the coactivator framing here, because Zaliova's fusion is OUT OF FRAME and terminates nine residues in, so whatever TEX41 might encode elsewhere, this chimera does not contain it. No source was found claiming a protein product for TEX41; a targeted search for TEX41 peptides was not run. NCBI Gene 2026
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
All three patients with noncanonical ZNF384 aberrations in this cohort expressed myeloid markers, CD10 expression was weak in two of them, and two of the three were classified as MPAL - matching the immunophenotype associated with canonical ZNF384 fusions. Unusually, all three also expressed CD66c, which is rarely expressed in ZNF384-rearranged ALL.
A novel class of ZNF384 aberrations in acute leukemia. · Zaliova M, Winkowska L, Stuchly J, Fiser K, Triska P, Zwyrtkova M, Hrusak O, Starkova J, Sramkova L, Stary J, Trka J, Zuna J · Blood Adv 2021What this evidence cannot show
n = 3 across all three noncanonical lesions, of which only ONE is the ZNF384-TEX41 fusion; the marker statements are pooled across the three and the paper's main text does not break CD66c or CD10 down per patient. The per-antigen percentages live in the paper's Table 1, which the retrieved XML renders only as a caption, so no numeric marker percentage is recorded in this entry. 'Rarely expressed in ZNF384r ALL' is the authors' characterisation, referenced to their reference 8, which this beat did not open.
Evidence that argues against this (1)
- CD66c is the one marker that does NOT match: the authors flag it as rarely expressed in canonical ZNF384-rearranged ALL and present in all three noncanonical cases. Three of three on a marker said to be rare is the kind of observation that either means the noncanonical lesions are a distinct entity after all, or means nothing at n = 3 - and the paper does not have the numbers to tell. It is recorded here because the phenocopy claim is the entry's headline and this is the observation that partly undercuts it. Zaliova M et al. 2021
ZNF384 is the 5' partner here - the one reported case that inverts the family's architecture
an out-of-frame truncation at ZNF384 p.481 produces the subtype's expression signature with no partner contribution
the sequenced counterexample that refutes 'ZNF384 is always the 3' partner'
contrast: AKAP8::ZNF384 is a canonical 5'-partner fusion, this one inverts the order and is out of frame
the truncation construct reproduces this patient's protein exactly: ZNF384 p.481_516 > SFQDYCACQ
- analogous to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
the boundary case that tests the rule: ZNF384 sits 5', the 3' partner is a lncRNA and the product is out of frame - the architecture KMT2A::AFF1 never violates