Oncofusion

start / graph / fusion-acin1-nutm1

Fusionmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q100-nutm1-comparator

ACIN1::NUTM1 - the commonest NUTM1 fusion in infants, cryptic to cytogenetics, and the only one with a human cell line

also called ACIN1-NUTM1 · ACIN1::NUT · ACINUS-NUTM1

This is the most common version of the NUTM1-rearranged leukaemia that affects babies without a KMT2A rearrangement. It joins the DNA-gripping front end of a splicing protein to almost all of NUTM1, and it is invisible to the standard chromosome test - it was found only when researchers sequenced the RNA.

The same thing, for a scientist

ACIN1::NUTM1 joins exon 3 or exon 4 of ACIN1 (14q11.2, Q9UKV3, 1341 aa) to exon 3 of NUTM1 (15q14, Q86Y26), retaining the functional **SAP DNA-binding domain of ACIN1 (aa 72-106, Pfam PF02037)** and most of NUTM1 [E1]. It is the **modal NUTM1 partner in infants - 17 of 39 cases with a known partner (44%)**, ahead of BRD9 (26%) and CUX1 (15%), and the third commonest in children (22%) [E2]. Both reported founding cases arose from **cryptic** rearrangements undetected by standard cytogenetics, and both were chemotherapy-sensitive [E1]. It is the only NUTM1 fusion with a human cell line: **KOPN32**, recognised in 2025 as the sole NUTM1-fusion line, with few copy-number alterations like KMT2A-rearranged lines, and - together with an inducible ACIN1::NUTM1 model - showing **direct** fusion-driven upregulation of **HOXA9, HOXA10, SKIDA1 and BMI1**, and high sensitivity to daunorubicin and vincristine [E3]. It is also the fusion behind the only locus-resolved epigenome in this subtype: in a BLUEPRINT case, HOXA9/HOXA10 expression comparable to two KMT2A-rearranged cases, with **decreased CpG methylation and high H3K4me3 and H3K27ac** over HOXA9-HOXA10, inside a region otherwise constitutively Polycomb-targeted in which two short HOXA9-overlapping regions **escape** repression [E2]. **The subnuclear trap is specific and measurable**: full-length ACIN1 is annotated in nuclear **speckles**, and both the speckle assignment and the SRPK2/Ser-1180 mechanism that moves it out of them are **uncoded** in UniProt - and Ser-1180 is far downstream of the exon-3/4 breakpoint, so **the fusion cannot be phospho-regulated that way and should not be assumed to be a speckle protein** [E4].

Go deeper
What is retained, and the one module that does the work

Two isoforms are reported, joining ACIN1 exon 3 or exon 4 to NUTM1 exon 3, and both retain the SAP domain of ACIN1 and most of NUTM1 1. Live UniProt places ACIN1's SAP domain at aa 72-106 (Pfam PF02037), and it is the protein's only annotated domain - the other 1,235 residues are six predicted disordered regions plus a C-terminal region (1210-1237) sufficient for RNPS1/SAP18 binding and ASAP-complex formation 4.

So the architecture of this fusion is unusually clean: one small DNA/chromatin-binding module from the 5' partner, and a disordered p300 activator from the 3' partner. That is the same logical shape as KMT2A::AFF1 - an address module plus an effector - and it is the reason the partner identity determines whether HOXA9 goes up while the NUTM1 moiety determines the global acetylation and methylation phenotype 52.

What is lost is the entire rest of ACIN1, which is most of what ACIN1 is known for: the ASAP complex, the exon junction complex association, the caspase-3-activated apoptotic chromatin condensation, the BCL2L1/Bcl-X splicing modulation, and the Ser-1180 SRPK2 site 4. This fusion is therefore not an ACIN1 gain-of-function in any of ACIN1's documented activities. It is a SAP domain, repurposed.

The subnuclear question, and why the partner's baseline compartment is the wrong answer

This graph already holds localization-partner-baseline-compartment-is-often-miscalled. ACIN1::NUTM1 is a textbook case of it, and the measurement is cheap enough to have been made here.

Live UniProt for Q9UKV3 gives three subcellular locations - Nucleus (SL-0191), Nucleus speckle (SL-0186) and Nucleus, nucleoplasm (SL-0190) - plus a note that phosphorylation on Ser-1180 by SRPK2 redistributes it from the nuclear speckles to the nucleoplasm. Every one of those four statements carries NO evidence code at all - not ECO:0000269, not ECO:0000305, nothing - while the same entry's FUNCTION field carries five ECO:0000269 citations 4. Rule 91's second clause says read the uncoded comments and record them as uncoded; that is what this entry does.

Two consequences for the fusion:

  1. Ser-1180 is not in the fusion. The breakpoint is at ACIN1 exon 3 or 4, and the retained module ends around residue 106. A regulatory phosphosite at 1180 is ~1,074 residues downstream of anything the fusion contains. Whatever mechanism moves full-length ACIN1 between speckles and nucleoplasm cannot operate on this oncoprotein.
  2. Nothing has been imaged. No localization experiment on ACIN1::NUTM1 in any cell appears in anything read for this entry. Assuming the fusion sits in speckles because ACIN1 does would be importing an uncoded annotation about a protein that is 92% absent from the fusion.

The NUTM1 side pulls the other way and is also unmeasured here: native NUT shuttles nucleus-to-cytoplasm through a leptomycin-sensitive NES, and BRD-NUT fusions are strictly nuclear only because the BET partner's bromodomains hold them on chromatin 6. A SAP domain binding DNA could do the same job, or not. The honest record is: subnuclear address unmeasured, and the two obvious inferences point in different directions.

The only locus-resolved epigenome in this subtype

One ACIN1-NUTM1 patient in the international series had BLUEPRINT data, and it is the only place in the NUTM1 literature read here where the chromatin is resolved at a locus rather than globally 2:

  • RNA-seq: high HOXA-locus expression, particularly HOXA9 and HOXA10, "comparable with levels observed in two cases with KMT2A rearrangement".
  • Bisulfite sequencing: decreased methylation around HOXA9 and HOXA10.
  • ChIP-seq: high H3K4me3 and H3K27ac there (chromatin state E11, indicative of an active transcription start site).
  • And the structure of the locus: an alternating pattern of activating (H3K4me1, H3K4me3) and repressing (H3K27me3) modifications, with the region "constitutively targeted by the Polycomb complex" (state E7) and two short activated regions overlapping HOXA9 that escape the repression.

That last observation is the interesting one for this project, because it is a different topology from what KMT2A-rearranged disease is usually described as doing. The fusion does not appear to clear Polycomb from the HOXA cluster; it appears to punch two small holes in it. The inducible model then shows the causal direction: ACIN1::NUTM1 induction alone raises HOXA9, HOXA10, SKIDA1 and BMI1 3.

n=1 for the epigenome, one cell line plus one inducible model for the causality. Neither is a substitute for the other and together they are still one fusion in a handful of backgrounds.

Cryptic by construction, and what that does to every frequency estimate

Both founding cases were cryptic: "derived from cryptic chromosomal rearrangements undetected by standard cytogenetic approaches", found by transcriptome analysis 1. ACIN1 sits at 14q11.2 and NUTM1 at 15q14, so the expected derivative chromosomes are small and the exchange is easily missed.

The founding paper's response was to propose partner-independent assays - NUTM1 immunohistochemistry and NUTM1 RNA expression in bone marrow - precisely because the partner list was expected to grow 1. It did: from two partners then to twelve now.

The practical consequence, and it is a bias with a direction: an infant whose ACIN1::NUTM1 is not looked for is counted as B-other, joins the KMT2A-wild-type comparator arm of an infant trial as an unclassified good-outcome case, and thereby raises the comparator's survival and shrinks the apparent NUTM1 advantage. Every frequency and every comparator figure in this subtype inherits the sequencing coverage of the contributing centres.

Contrast with KMT2A::AFF1, and one place the analogy actually holds

Contrast. Different 3' effector class (a p300 activator, not an elongation/DOT1L recruiter), different mark (H3K27ac and DNA hypomethylation, not H3K79me2/3), no menin interface, no CXXC, and the opposite prognosis in the same trials 2.

And one structural analogy that does hold, which is worth saying because this beat is a contrast beat and the temptation is to make everything differ. Both fusions are address module + effector: KMT2A contributes AT-hooks and a CXXC CpG-binding module and the partner contributes the effector; ACIN1 contributes a SAP DNA-binding domain and NUTM1 contributes the effector. In both, the 5' partner decides where and the 3' partner decides what. That is why NUTM1's global phenotypes (H3K27ac, DNA hypomethylation) are reported as partner-independent while the HOXA phenotype is partner-dependent 52 - the same division of labour, running in the opposite direction from the KMT2A case, where the 5' partner is the constant.

What this does not license. It does not license calling ACIN1 a functional analogue of KMT2A - SAP is not CXXC, it is not CpG-selective, and no occupancy data for this fusion exist at all. And it does not license explaining the outcome difference, which remains unexplained by anything read here.

Facts and figures
gene5ACIN1
gene3NUTM1
cytobandt(14;15)(q11.2;q14)
cytoband noteDerived from the two genes' HGNC locations fetched live 2026-09-22 (ACIN1 14q11.2, HGNC:17066; NUTM1 15q14, HGNC:29919). The founding report calls the rearrangement CRYPTIC and undetected by standard cytogenetics, so this band pair is the genes' addresses and NOT a reported karyotype.
uniprot5Q9UKV3
uniprot3Q86Y26
prognosisFavourable at the SUBTYPE level and not demonstrated per partner - neither the 85-case international series nor the 175-infant series reports survival by NUTM1 partner. The two founding ACIN1-NUTM1 cases were described as chemotherapy-sensitive (PMID 31515871), and ACIN1 sits in the HOXA9-positive group where no poor-outcome case has been reported. The reported poor-outcome NUTM1-r cases are ZNF618::NUTM1 and CUX1::NUTM1, both outside this partner.
cell lineKOPN32 - the ONLY human cell line carrying a NUTM1 fusion, recognised as ACIN1::NUTM1-positive in 2025, established from a RELAPSED infant ALL case; few copy-number alterations; highly sensitive to daunorubicin and vincristine among eight standard agents (PMID 40884014). An ACIN1::NUTM1-inducible ALL model was built from cDNA subcloned out of KOPN32.
cell typeB-cell precursor. The two founding cases are infant ALL; KOPN32 is a B-precursor line from a relapsed infant case. Immunophenotype is not reported per partner except that the three pro-B cases in the international series were BRD9-NUTM1 rather than ACIN1-NUTM1, so ACIN1 cases are pre-B or common by elimination within that cohort of 32.
cell stateB-lymphoid-committed with leukaemic stem cell properties, per the subtype-level functional work (PMID 41460961), which used BRD9-NUTM1 rather than this fusion.
epigenetic readoutTHE ONLY LOCUS-RESOLVED EPIGENOME IN THIS SUBTYPE, and it is n=1. In a BLUEPRINT-profiled ACIN1-NUTM1 case: DECREASED CpG METHYLATION around HOXA9 and HOXA10 with HIGH H3K4me3 and H3K27ac (BLUEPRINT chromatin state E11, an active TSS), HOXA9/HOXA10 expression comparable to two KMT2A-rearranged cases, and an ALTERNATING pattern of activating (H3K4me1, H3K4me3) and repressing (H3K27me3) marks across a HOXA region otherwise constitutively Polycomb-targeted (state E7), with TWO SHORT HOXA9-OVERLAPPING REGIONS ESCAPING the repression (PMID 34211097). Causality from the inducible model: ACIN1::NUTM1 induction alone raises HOXA9, HOXA10, SKIDA1 and BMI1 (PMID 40884014). Subtype-level and partner-independent: global H3K27ac enhancement and global DNA hypomethylation (PMID 41460961). NEITHER MOIETY WRITES OR ERASES A MARK: the SAP domain binds DNA and the NUTM1 moiety allosterically activates p300. THE TOPOLOGY IS WORTH NOTING AS A CONTRAST - the fusion appears to punch two small holes in a Polycomb-repressed HOXA cluster rather than clear it, which is a different description from the KMT2A-rearranged HOXA state and is based on one patient's segmentation calls.
acin1 function lostACIN1's documented activities are all C-terminal of the retained SAP domain and are therefore LOST in the fusion: auxiliary component of the exon junction complex, RNA-binding subunit of the ASAP complex, caspase-3-activated apoptotic chromatin condensation, BCL2L1/Bcl-X splicing modulation suppressing pro-apoptotic Bcl-X(S), and regulation of cyclin A1 in leukaemia cells (UniProt Q9UKV3 FUNCTION, five ECO:0000269 citations). This fusion is NOT an ACIN1 gain-of-function in any of them.
menin relationshipNONE. No MBM-like motif and no CXXC domain in either moiety; no menin contact reported for ACIN1, NUTM1 or this fusion. HOXA9 and HOXA10 are nonetheless activated, which is the fact that constrains a necessity claim about the KMT2A-menin interface.
diagnosticsCRYPTIC - both founding cases were invisible to standard cytogenetics and were found by transcriptome analysis. RT-PCR for the fusion supports MRD monitoring once the breakpoint is known, but partner-independent detection at diagnosis (break-apart FISH, RNA-seq, NUTM1 IHC, NUTM1 RNA expression) is what the founding paper recommends (PMID 31515871).
diseases
[
  "infant B-cell precursor ALL (KMT2A-wild-type)",
  "paediatric B-ALL"
]
breakpoints
[
  {
    "id": "bp-acin1-exon-3-nutm1-exon-3",
    "gene5_exon_last": 3,
    "gene3_exon_first": 3,
    "domains_retained_5": [
      "SAP (SAF-A/B, Acinus and PIAS DNA-binding domain)"
    ],
    "domains_lost_5": [
      "ASAP-complex region (1210-1237, RNPS1/SAP18 binding)",
      "Ser-1180 SRPK2 phosphosite",
      "the caspase-3 cleavage and apoptotic-chromatin-condensation function"
    ],
    "domains_retained_3": [
      "TAD1 (transactivation domain 1)",
      "TAD2 (transactivation domain 2)"
    ],
    "frequency_pct": null,
    "population": "infant BCP-ALL, KMT2A-wild-type",
    "phenotype_notes": "Isoform 1 of two. The founding report identifies ACIN1 exon 3 OR exon 4 joined to NUTM1 exon 3, both retaining the functional DNA-binding SAP domain and most of NUTM1 (PMID 31515871). frequency_pct is null because neither isoform's share is reported. The domains_lost_5 list follows from the SAP domain sitting at aa 72-106 while everything else annotated in ACIN1 is C-terminal of residue 1000 (UniProt Q9UKV3, fetched live); it is an inference from domain position, not a statement the source makes."
  },
  {
    "id": "bp-acin1-exon-4-nutm1-exon-3",
    "gene5_exon_last": 4,
    "gene3_exon_first": 3,
    "domains_retained_5": [
      "SAP (SAF-A/B, Acinus and PIAS DNA-binding domain)"
    ],
    "domains_lost_5": [
      "ASAP-complex region (1210-1237, RNPS1/SAP18 binding)",
      "Ser-1180 SRPK2 phosphosite"
    ],
    "domains_retained_3": [
      "TAD1 (transactivation domain 1)",
      "TAD2 (transactivation domain 2)"
    ],
    "frequency_pct": null,
    "population": "infant BCP-ALL, KMT2A-wild-type",
    "phenotype_notes": "Isoform 2 of two, from the same founding report. No functional difference between the two isoforms is reported."
  }
]
acin1 sap domain
{
  "name": "SAP (SAF-A/B, Acinus and PIAS)",
  "start": 72,
  "end": 106,
  "pfam": "PF02037",
  "source": "UniProt Q9UKV3, fetched live 2026-09-22; ACIN1's ONLY annotated domain"
}
frequency
{
  "infant_nutm1_r_partner_share_pct": 44,
  "infant_nutm1_r_partner_share_n": "17 of 39 infant cases with a known partner (PMID 34211097) - the MODAL infant partner",
  "paediatric_partner_share_pct": 22,
  "paediatric_partner_share_n": "7 of 32 children aged 1 or over with a known partner (PMID 34211097)"
}
localization
{
  "subcellular": "NOT MEASURED for this fusion. The two inferences point in DIFFERENT directions: the NUTM1 moiety brings a leptomycin-sensitive NES that in BRD-NUT fusions is overridden by the partner's chromatin binding (PMID 17934517), while the ACIN1 moiety contributes only a 35-residue SAP DNA-binding domain rather than a double bromodomain. Whether that is enough to retain the oncoprotein in the nucleus is untested.",
  "subnuclear": "NOT MEASURED, and the partner's baseline compartment must NOT be substituted for it. Full-length ACIN1 is annotated in Nucleus speckle (SL-0186) as well as nucleoplasm, and UniProt notes that SRPK2 phosphorylation of Ser-1180 moves it out of speckles - but ALL FOUR of those localization statements are UNCODED in UniProt (fetched 2026-09-22), and Ser-1180 is roughly 1,074 residues downstream of the fusion breakpoint, so the fusion cannot be regulated that way. No imaging of ACIN1::NUTM1 exists in anything read here.",
  "shown_for_fusion": "NOT shown for ACIN1::NUTM1; shown for BRD4::NUTM1 and BRD3::NUTM1 (strictly nuclear, discrete foci)",
  "partner_baseline_caution": "This is an instance of localization-partner-baseline-compartment-is-often-miscalled, with the extra wrinkle that the partner's own annotation is uncoded AND the regulatory site that produces its compartment switch is not retained."
}
epigenetic marks
[
  "H3K27ac",
  "H3K4me3",
  "H3K4me1",
  "H3K27me3",
  "DNA methylation (decreased at HOXA9/HOXA10; global hypomethylation subtype-wide)"
]
How this could bite us back
  • All the pharmacology of this fusion is eight drugs in one cell line derived from a RELAPSED patient. A relapse-derived line is a poor model of a chemosensitive disease, and any claim that KOPN32's sensitivity profile reflects the subtype's curability has the causality backwards - this line came from the child in whom the cure failed.
  • The obvious targeted node is p300, activated by the retained NUTM1 TADs. p300 loss is developmentally lethal and its TAZ2 surface is shared with p53 TAD2 and p63 TAD, so a p300-directed agent in an infant with a subtype already cured by chemotherapy is a poor risk trade. BMI1 is the better-behaved downstream node and is untested here.
  • If the fusion punches small activating holes in a Polycomb-repressed HOXA cluster, then a Polycomb-directed intervention - an EZH2 inhibitor, for instance - would be expected to make the situation WORSE by removing the repression that still covers most of the locus. That is the opposite of how EZH2 inhibition is usually reasoned about in leukaemia and it follows directly from the one epigenome available.
  • The fusion is cryptic, so the patients who would be enrolled in any trial of a targeted agent are precisely those whose centres sequence RNA. That is not a random sample of the disease, and the trial population would be systematically better-resourced than the patients the drug is meant to help.
  • ACIN1's normal job includes suppressing formation of the pro-apoptotic Bcl-X(S) isoform and inducing apoptotic chromatin condensation after caspase-3 activation. The fusion loses both. Whether the REMAINING wild-type ACIN1 allele is haploinsufficient for them is unexamined, and if it is, the leukaemic cell may be closer to the apoptotic threshold than a normal cell - which is a candidate explanation for the chemosensitivity, and also a warning that any agent stabilising ACIN1 function would be expected to reduce it.
Still open
  1. Where is ACIN1::NUTM1 in the cell? A single immunofluorescence in KOPN32 - which exists and is available - would answer subcellular and subnuclear at once, and would test whether a 35-residue SAP domain can hold the NUTM1 NES in check the way a double bromodomain does. Nobody has imaged this fusion.
  2. Does ACIN1::NUTM1 occupy the HOXA cluster directly? The BLUEPRINT case shows an active, hypomethylated HOXA9-HOXA10 island inside a Polycomb-repressed region, and the inducible model shows the fusion causes the expression change, but no ChIP or CUT&RUN for the fusion protein exists. Without it, 'the SAP domain addresses HOXA' is a hypothesis with the right shape and no data.
  3. Is KOPN32 menin-inhibitor sensitive? It is HOXA9-high, has no KMT2A fusion and no menin interface, and it is the only NUTM1-fusion line in existence. This is a one-plate experiment that would settle a question Boer 2021 raised in 2021 and that bears directly on whether menin inhibitors work through the interface or through HOXA dependence.
  4. Does the fusion's SAP domain bind DNA sequence-specifically, and if so where? SAP domains are AT-hook-adjacent scaffold-attachment modules rather than sequence readers, which would predict a broad, AT-rich or matrix-associated distribution rather than a promoter-selective one - a testable difference from the CpG-island selectivity of KMT2A's CXXC.
  5. Why is BMI1 upregulated by both HOXA9-positive and HOXA9-negative NUTM1 fusions? BMI1 is the shared downstream node across the whole subtype and is proposed to involve EP300 binding the NUTM1 moiety, which would make it the partner-independent output. It is also the one downstream node with existing pharmacology, and it has not been tested in this subtype.
Every source behind this entry (6)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Two cases of chemotherapy-sensitive non-KMT2A-rearranged infant ALL carried identical ACIN1-NUTM1 fusions derived from cryptic chromosomal rearrangements undetected by standard cytogenetics; two isoforms were identified, joining exon 3 or exon 4 of ACIN1 to exon 3 of NUTM1, and both fusion transcripts contained the functional DNA-binding SAP domain of ACIN1 and most of NUTM1; RT-PCR detection allowed MRD monitoring, and the authors predicted NUTM1 fusions to be recurrent in infant ALL and proposed partner-independent screening by NUTM1 immunohistochemistry and NUTM1 RNA expression.

    Cryptic recurrent ACIN1-NUTM1 fusions in non-KMT2A-rearranged infant acute lymphoblastic leukemia · Pincez T, Landry JR, Roussy M, et al. · Genes, Chromosomes and Cancer 2020
    PMID 31515871DOI
    What this evidence cannot show

    Two patients. 'Chemotherapy-sensitive' in two cases is an observation, not a survival estimate. The recurrence prediction rested at the time on public genomic datasets and a literature review rather than on a cohort; it has since been confirmed at scale [E2], so it is retrospectively vindicated rather than independently established here. The paper gives exon-level breakpoints and does NOT give residue boundaries for the retained ACIN1 fragment; the SAP domain coordinates in this entry come from live UniProt [E4], and the inference that exon 3/4 retains aa 72-106 while losing everything C-terminal is this entry's, based on the domain's position near the N-terminus.

    Evidence that argues against this (1)
    • The proposed partner-independent screens rest on NUTM1 being transcriptionally silent outside testis. That specificity has not been tested against the NUTM2-family paralogues, which UniProt and IntAct list as separate NUT-family genes; antibody or expression cross-reactivity with a NUTM2 product would be invisible to this assay design. UniProtKB 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Among 71 NUTM1-rearranged B-ALL cases with a known partner in the Ponte di Legno/Interfant series, ACIN1 was the commonest partner in infants (17 of 39, 44%) ahead of BRD9 (10, 26%) and CUX1 (6, 15%), and third commonest in children aged 1 or over (7 of 32, 22%); ACIN1 belongs to the HOXA9-positive partner group, and in a BLUEPRINT-profiled ACIN1-NUTM1 case HOXA9 and HOXA10 expression was comparable to two KMT2A-rearranged cases, with decreased CpG methylation and high H3K4me3 and H3K27ac (chromatin state E11) over HOXA9-HOXA10 inside a region otherwise constitutively Polycomb-targeted, in which two short HOXA9-overlapping regions escaped repression.

    PMID 34211097DOI
    What this evidence cannot show

    Read in full from the Europe PMC open-access full text (PMC8478641); PubMed indexes no abstract for this letter. THE EPIGENOME IS n=1 - one ACIN1-NUTM1 patient with BLUEPRINT data, compared against two KMT2A-rearranged cases and a handful of other subtypes - and the chromatin states are BLUEPRINT segmentation calls (E11, E7) rather than hand-curated peaks. Outcome is not reported by partner in this study, so nothing here supports an ACIN1-specific prognosis. The partner frequencies are from cases in which a partner was identified, so a partner that is systematically harder to detect would be underrepresented.

    Evidence that argues against this (1)
    • The same study found HOXA9 low or absent in 4 of 5 paediatric ZNF618-NUTM1 cases, so HOXA9 activation is partner-dependent and the ACIN1 result must not be generalised to the subtype. It also reports that both HOXA9-positive and HOXA9-negative fusions upregulate BMI1, so the shared downstream node is BMI1 rather than HOXA9. Boer JM et al. 2021
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    ACIN1::NUTM1 is the most frequent fusion in NUTM1-rearranged infant ALL; KOPN32, established from a relapsed infant ALL case, is the only cell line carrying a NUTM1 fusion, and has few copy-number alterations like KMT2A-rearranged lines; both KOPN32 and an ACIN1::NUTM1-inducible ALL model built from the KOPN32-derived fusion cDNA show upregulation of HOXA9, HOXA10, SKIDA1 and BMI1, indicating direct involvement of the fusion; and KOPN32 is highly sensitive to daunorubicin and vincristine among eight standard agents tested.

    PMID 40884014DOI
    What this evidence cannot show

    One cell line and one inducible model, so the HOXA9/HOXA10/SKIDA1/BMI1 result is a direct causal demonstration in a single genetic background. Eight agents in one line is the entire published pharmacology of this subtype. KOPN32 derives from a RELAPSED infant ALL case - a patient in whom chemotherapy had already failed - which is an awkward foundation for a model of a chemosensitive subtype, and its drug-response profile should not be read across to the clinic. The paper's 'most frequent fusion' claim agrees with the 85-case series [E2] and contradicts a 9-patient series in which BRD9 was commonest.

    Evidence that argues against this (1)
    • A 175-infant single-centre series reports BRD9::NUTM1, not ACIN1::NUTM1, as the most common NUTM1 fusion in its KMT2A-wild-type group. That group is n=9, so the disagreement is very probably sampling, but it is a real disagreement in the published record and this graph should carry both. Ilyasova K et al. 2025
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProtKB Q9UKV3 (ACINU_HUMAN, apoptotic chromatin condensation inducer in the nucleus) is 1341 aa with exactly one annotated domain - SAP at aa 72-106, Pfam PF02037 - six predicted disordered regions, and a region at 1210-1237 sufficient for RNPS1/SAP18 binding and ASAP-complex formation; it is localized to Nucleus, Nucleus speckle and Nucleus nucleoplasm with a note that phosphorylation on Ser-1180 by SRPK2 redistributes it from nuclear speckles to the nucleoplasm, and ALL FOUR of those localization statements carry NO evidence code, while the FUNCTION field carries five ECO:0000269 citations; the uncoded TISSUE SPECIFICITY field states that the Ser-1180-phosphorylated form is highly expressed and phosphorylated in patients with myeloid haematologic malignancies.

    What this evidence cannot show

    Fetched live 2026-09-22 by gene_exact:ACIN1 AND organism_id:9606 AND reviewed:true, one reviewed hit, accession Q9UKV3 and length 1341 read back from the record, then re-fetched by accession for the reference list. The absence of evidence codes on the localization comments is a property of the database record as served on that date and is reported as such; it does not mean no experiment exists - Acinus's nuclear speckle residence is described in the primary literature this entry has not read. IT DOES MEAN a fusion beat cannot treat 'ACIN1 is a speckle protein' as an experimentally coded fact, and the SRPK2/Ser-1180 mechanism is irrelevant to the fusion regardless of its coding, because Ser-1180 is not retained.

    Evidence that argues against this (1)
    • The uncoded comments are also where the interesting statements live, which is rule 91's point in the other direction: the Ser-1180/myeloid-malignancy observation in TISSUE SPECIFICITY is uncoded and is the only line in the record connecting this protein to haematologic malignancy at all. Discarding uncoded lines would have discarded it. Nature 1999
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    NUTM1-rearranged leukaemia features global DNA hypomethylation irrespective of the 5' fusion partner and an aberrant chromatin state of globally enhanced H3K27 acetylation with distinctive open chromatin regions co-opting B-lineage and stemness programmes including NF-kB and posterior HoxA genes; the fusions drive B-lymphoid commitment while conferring leukaemic stem cell properties, and the cells are profoundly chemosensitive in a manner linked to a dependence on active transcription.

    Comprehensive molecular and functional analysis of NUTM1-rearranged leukemia · Nishimura K, Isobe T, Shigehiro T, et al. · Blood 2026
    PMID 41460961DOI
    What this evidence cannot show

    Read from the published abstract only. The in-vivo work uses BRD9-NUTM1, not ACIN1::NUTM1, so the leukaemogenic-sufficiency result does not transfer to this fusion; what transfers is the partner-independent claim about DNA hypomethylation, and that claim's strength depends on how many partners were profiled, which is unread here.

    Evidence that argues against this (1)
    • An editorial comment on this paper appeared in the same journal and has not been read for this entry; recorded so the primary paper is not presented as unchallenged. Shiloh R et al. 2026
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Native NUT carries both a nuclear localization sequence and a leptomycin-sensitive nuclear export sequence and shuttles between nucleus and cytoplasm, whereas BRD3-NUT and BRD4-NUT are strictly nuclear because the BRD moiety retains NUT in the nucleus through chromatin interactions.

    PMID 17934517DOI
    What this evidence cannot show

    The 5' partners are BET proteins in carcinoma cells; neither ACIN1 nor a lymphoid cell appears in this paper. It is cited here for one specific purpose - to establish that the NUTM1 moiety brings an active NES that the 5' partner has to overcome - and NOT to assert that ACIN1::NUTM1 is nuclear, which nobody has measured.

    Evidence that argues against this (1)
    • The paper attributes nuclear retention specifically to the BRD moiety's chromatin binding. ACIN1's contribution to the fusion is a 35-residue SAP domain, not a double bromodomain, so the prediction for this fusion is weaker retention - and possibly a partly cytoplasmic oncoprotein, which would be unprecedented in this family and has not been looked for. UniProtKB 2026
Connected to