Oncofusion

start / graph / fusion-brd9-nutm1

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

BRD9::NUTM1, t(5;15)(p15.33;q14) - one fusion is enough to make infant B-ALL in a mouse, and it is the one that behaves best in a child

also called BRD9-NUTM1 · BRD9:NUTM1 · BRD9::NUT

This is a chromosome accident that joins part of BRD9, a chromatin-reading protein, to almost all of NUTM1, a protein that switches on a histone-modifying enzyme. It causes leukaemia in babies who do not have the more familiar KMT2A rearrangement - and unlike that one, children with this fusion nearly always survive standard chemotherapy.

The same thing, for a scientist

BRD9::NUTM1 joins the 5' portion of BRD9 (5p15.33, Q9H8M2, the defining bromodomain subunit of ncBAF) to almost the entire open reading frame of NUTM1 (15q14, Q86Y26). It is the second commonest NUTM1 fusion in infants - 10 of 39 infant cases with a known partner (26%) in the 85-case Ponte di Legno/Interfant series, behind ACIN1 at 44% [E1]. Breakpoints are variable on the BRD9 side, reported at **BRD9 exons 3, 8 and 14** in a 9-patient Russian infant series [E2], which matters because the surviving BRD9 elements differ between patients. Its standing among NUTM1 fusions is that it is the **only one functionally proven sufficient**: expression of BRD9-NUTM1 alone induces serially transplantable **prepro-B-like** leukaemia in vivo that recapitulates the molecular and immunophenotypic features of human NUTM1-r B-ALL, with the fusion establishing global H3K27 acetylation, new open chromatin regions co-opting B-lineage and stemness programmes including NF-kB and posterior HoxA, and global DNA hypomethylation [E3]. It is also the fusion that tracks with the most primitive human immunophenotype: **all three pro-B-ALL infants** in the international cohort carried BRD9::NUTM1, while the rest of the NUTM1-r infants were pre-B or common [E1]. Clinically it sits inside a subtype with 4-year EFS of 100% in 35 uniformly treated Interfant infants [E1], and with 2-year EFS of 80% (95% CI 52-100) in the 9-patient series where it is the commonest fusion [E2]. **The CXXC/menin logic of KMT2A does not apply here at all**: neither moiety carries an MBM-like motif or a CXXC domain, and the ncBAF biology `gene-brd9` is written for is a property of full-length BRD9, not obviously of the 5' fragment this fusion retains [E4].

Go deeper
What is actually retained, and why nobody should assume it is ncBAF

gene-brd9 already exists in this graph, written as the defining bromodomain subunit of non-canonical BAF and a KMT2A-rearranged dependency ranking 39 of 17,728 in the project's selectivity scan 4. That entry is about full-length BRD9 in a complex. This fusion is a different object and the entry must not be read across.

What is known: the fusion retains the 5' portion of BRD9 and almost the whole NUTM1 ORF 13. Reported BRD9 breakpoints are at exons 3, 8 and 14 2. UniProt places BRD9's bromodomain at aa 136-240 4, so a transcript breaking at exon 3 and one breaking at exon 14 very plausibly differ in whether the bromodomain survives - and this entry does not have the exon-to-residue mapping needed to say which. That mapping is the single most useful missing fact about this fusion and it is recorded as an open question rather than guessed.

What that uncertainty forbids: it forbids assuming the fusion is an ncBAF-directed oncoprotein, and it forbids assuming a BRD9 bromodomain ligand or degrader engages it. Both assumptions are natural - the 5' partner is a bromodomain protein with clinical-stage chemical matter - and neither is supported by anything read here.

The functional proof, and its limits

BRD9-NUTM1 is the representative fusion in the only functional study of this subtype 3. Expression of the fusion is described as sufficient to induce a serially transplantable prepro-B-like B-cell-progenitor leukaemia in vivo, faithfully recapitulating the key molecular and immunophenotypic features of the human disease. The mechanism reported is an aberrant chromatin state: global enhancement of H3K27 acetylation, distinctive open chromatin regions that co-opt B-lineage and stemness programmes including NF-kB and posterior HoxA genes, and a dual oncogenic role in which the fusion drives B-lymphoid commitment while conferring leukaemic stem cell properties.

Read from the abstract only. The transduction system, the recipient genotype, the number of animals, the latency and the penetrance are all unread here. 'Sufficient' in a retroviral or transgenic overexpression system is a statement about that system, and the human disease arises prenatally from a single fusion event in a fetal progenitor 5 - a context no overexpression model reproduces. The convergence with the human immunophenotype (prepro-B in the mouse; the three pro-B human infants all BRD9::NUTM1 1) is the strongest argument that the model is faithful, and it is a correspondence of two small observations rather than a formal validation.

Cell type, cell state, localization and the epigenetic readout

Cell type. A B-cell progenitor at the prepro-B/pro-B stage. The mouse disease is prepro-B-like 3; in humans, of 32 NUTM1-r infants with immunophenotype data 16 were pre-B, 12 common, 3 pro-B and one biphenotypic, and all three pro-B cases were BRD9::NUTM1 1. The cell of transformation is a fetal progenitor: a patient-specific NUTM1 fusion and two IG/TR leukaemic markers were detectable in that patient's umbilical cord blood at birth 5.

Cell state. Simultaneously B-lymphoid-committed and stem-like - the 'dual oncogenic role' 3 - which is a different shape from the KMT2A-rearranged differentiation block this graph holds as cell-state-differentiation-block, where the lesion arrests a progenitor rather than committing it while conferring self-renewal.

Localization. NOT MEASURED FOR THIS FUSION. No imaging of BRD9::NUTM1 in any cell is reported in anything read here. What is known by family resemblance: BRD3-NUT and BRD4-NUT are strictly nuclear, because the BRD moiety retains NUT on chromatin against NUT's own leptomycin-sensitive NES 6. Whether the BRD9 fragment does the same is untested, and it is not a formality - the NES is in the retained NUTM1 moiety of every one of these fusions.

Epigenetic readout. Global H3K27ac enhancement plus global DNA hypomethylation independent of the 5' partner, and new open chromatin regions 3; plus posterior HOXA activation, confirmed for BRD9-NUTM1 specifically in the human cohort 1. The marks are written by p300, activated allosterically by the two NUTM1 TADs 7; neither BRD9 nor NUTM1 writes or erases anything.

The contrast with KMT2A::AFF1, stated as a contrast

Same disease, same age band, drawn from the same cohorts - the NUTM1-r cases in the Interfant-99/-06 and Ponte di Legno series are by construction the KMT2A-wild-type arm of the trials this graph's index fusion defines 1.

KMT2A::AFF1BRD9::NUTM1
what the 5' side contributesAT-hooks, a CXXC CpG-binding module and the menin-binding motifa 5' fragment of a bromodomain protein, of variable extent (exons 3, 8, 14)
what the 3' side contributesAFF1 pSER/ALF, recruiting the elongation machinery and DOT1Lalmost the whole NUTM1 ORF - a disordered p300 activator
the markH3K79me2/3 (DOT1L)H3K27ac (p300), plus global DNA hypomethylation
meninthe drug target of the entire programmenot involved in any way reported here
HOXA9upup 1
chemosensitivitypoor; MRD below 0.05% in 48%profound; MRD below 0.05% in 91.1% 13
4-year EFS~36-37% 2-year EFS in an independent infant series100% for the NUTM1-r arm as a whole (n=35) 12

Where this is a genuine mechanistic contrast: the mark, the writer, the reader, the druggable node, and the presence or absence of a menin interface all differ. Where overclaiming would be easy and wrong: both run a posterior-HOXA programme, and the graph should not conclude from the outcome difference that HOXA activation is unimportant - what differs is the route to it and, per the Blood 2026 mechanism, a dependence on active transcription that makes the cells chemo-fragile 3. Nothing read here establishes why one HOXA-driven infant leukaemia is curable and the other is not, and that is the most interesting unanswered question in this comparison.

Facts and figures
gene5BRD9
gene3NUTM1
cytobandt(5;15)(p15.33;q14)
cytoband noteDerived from the two genes' HGNC locations fetched live 2026-09-22 (BRD9 5p15.33 HGNC:25818; NUTM1 15q14 HGNC:29919). The sources read for this entry identify the fusion by RNA-seq and do not print a karyotypic designation, and NUTM1 fusions are frequently CRYPTIC and invisible to standard cytogenetics, so this band pair is the genes' addresses rather than a reported translocation.
uniprot5Q9H8M2
uniprot3Q86Y26
prognosisFavourable, at the SUBTYPE level and not demonstrated per partner. NUTM1-r: 4-year EFS 100% in 35 uniformly treated Interfant infants versus 74.0% in 126 NUTM1-wildtype/KMT2A-wildtype infants (p=0.001); 4-year EFS 92.1% and OS 100% in 32 children aged 1 or over; all cases reached CR; MRD below 0.05% in 41/45 (91.1%) against 48% in KMT2A-rearranged infant ALL; no transplants in first remission (PMID 34211097). In the independent 9-patient series where BRD9::NUTM1 is the commonest fusion, NUTM1-r 2-year EFS was 80% (95% CI 52-100) against 36-37% for KMT2A partners (PMID 40000924). NEITHER study reports survival by NUTM1 partner, so a BRD9::NUTM1-specific prognosis does not exist.
cell typeB-cell progenitor at the prepro-B/pro-B stage. The mouse disease driven by this fusion alone is prepro-B-like (PMID 41460961), and all three pro-B-immunophenotype infants among 32 NUTM1-r infants with immunophenotype data carried BRD9-NUTM1, the remainder being pre-B (16) or common (12) (PMID 34211097). The cell of transformation is a FETAL progenitor: a NUTM1 fusion has been detected in cord blood at birth (PMID 39099338, CUX1::NUTM1).
cell stateB-lymphoid-committed AND leukaemic-stem-cell-like at the same time - the reported 'dual oncogenic role' (PMID 41460961). This is a different shape from the KMT2A-rearranged differentiation block this graph holds as cell-state-differentiation-block: commitment plus self-renewal rather than arrest.
epigenetic readoutGlobal enhancement of H3K27 ACETYLATION plus GLOBAL DNA HYPOMETHYLATION that is independent of the 5' fusion partner, with distinctive open chromatin regions co-opting B-lineage and stemness programmes including NF-kB and posterior HoxA genes (PMID 41460961). Posterior HOXA activation is confirmed for BRD9-NUTM1 specifically in human cases (PMID 34211097). The acetyltransferase is p300, allosterically activated by the retained NUTM1 TADs at 403-470 (PMID 36690674) - NEITHER moiety of this fusion writes or erases a mark itself: BRD9 contributes a reader module (if retained at all) and NUTM1 contributes an activator of a writer. The DNA hypomethylation has NO established mechanism in anything read here and no DNMT or TET contact is reported for either moiety.
menin relationshipNONE. Neither moiety carries an MBM-like RWRFP motif or a CXXC domain, and no menin contact is reported for BRD9, NUTM1 or this fusion in anything read here. This fusion is the KMT2A-WILD-TYPE comparator, and its existence is why the Interfant cohorts have a KMT2A-wild-type arm at all.
shown for fusionthis entry IS a fusion entry; the analogous field on the component genes reads: p300 recruitment shown for BRD4::NUTM1 and not for BRD9::NUTM1; in-vivo leukaemogenesis shown for BRD9-NUTM1 and for no other NUTM1 fusion
diagnosticsNUTM1 fusions are frequently cryptic and missed by standard cytogenetics. Partner-independent screens are break-apart FISH, RNA sequencing, NUTM1 immunohistochemistry and NUTM1 RNA expression (3' exon expression is highly specific to fusion cases) (PMID 31515871, 34211097). Since new partners are still being found, an unbiased approach is preferred over partner-specific RT-PCR.
diseases
[
  "infant B-cell precursor ALL (KMT2A-wild-type)",
  "paediatric B-ALL"
]
breakpoints
[
  {
    "id": "bp-brd9-exon-variable-nutm1",
    "gene5_exon_last": 3,
    "domains_retained_5": [],
    "domains_lost_5": [],
    "domains_retained_3": [
      "TAD1 (transactivation domain 1)",
      "TAD2 (transactivation domain 2)"
    ],
    "frequency_pct": null,
    "population": "infant BCP-ALL, KMT2A-wild-type",
    "phenotype_notes": "BRD9 breakpoints are reported at exons 3, 8 AND 14 in a 9-patient series (PMID 40000924); this record carries the earliest of the three as gene5_exon_last and the other two are listed in data.brd9_breakpoint_exons. NO source read here maps those exons onto BRD9's bromodomain at aa 136-240, so whether the bromodomain is retained is UNKNOWN and is deliberately not asserted. frequency_pct is null because the 9-patient series does not report the per-exon split. The NUTM1 side retains almost the entire open reading frame in every reported NUTM1 fusion, which is why the TADs at 403-470 are listed as retained.",
    "domains_5_note": "LEFT EMPTY DELIBERATELY. BRD9 breakpoints are reported at exons 3, 8 and 14 (PMID 40000924) and no source read here maps those exons onto BRD9's bromodomain at aa 136-240, so no retained or lost 5' domain can be named. A placeholder string here was flagged by scripts/check-joins.mjs as an unresolvable domain reference, which is the correct behaviour: an undetermined boundary belongs in prose, not in a domain lane."
  }
]
brd9 breakpoint exons
[
  3,
  8,
  14
]
frequency
{
  "infant_nutm1_r_partner_share_pct": 26,
  "infant_nutm1_r_partner_share_n": "10 of 39 infant cases with a known partner (PMID 34211097)",
  "paediatric_partner_share": "not among the three commonest paediatric partners, which are CUX1 (28%), ZNF618 (28%) and ACIN1 (22%) of 32 children with a known partner (PMID 34211097)",
  "nutm1_r_share_of_kmt2a_wildtype_infants_pct": 21.7,
  "nutm1_r_share_of_paediatric_b_all_pct": "0.28-0.86% across three population-representative cohorts; the paper's discussion gives roughly 3-5% in infants and 0.4-0.9% in children"
}
localization
{
  "subcellular": "NOT MEASURED for BRD9::NUTM1. Inferred-by-family-resemblance only: BRD3-NUT and BRD4-NUT are strictly nuclear because the BRD moiety retains NUT on chromatin against NUT's own leptomycin-sensitive NES (PMID 17934517). The BRD9 fragment's chromatin-binding capacity is undetermined [E4], so this inference is weaker than for the BET fusions and is recorded as an inference.",
  "subnuclear": "NOT MEASURED. No imaging of BRD9::NUTM1 in any cell exists in anything read here. BRD4-NUT forms discrete nuclear foci corresponding to ~100 hyperacetylated megadomains up to 2 Mb (PMID 26220994); whether BRD9::NUTM1 forms foci at all in a B progenitor is unknown, and the Blood 2026 chromatin data report a GLOBAL H3K27ac rise and new open regions rather than megadomains (PMID 41460961), which is compatible with either.",
  "shown_for_fusion": "BRD4::NUTM1 and BRD3::NUTM1 (strictly nuclear, foci); NOT shown for BRD9::NUTM1"
}
epigenetic marks
[
  "H3K27ac",
  "DNA methylation (global, direction: DOWN)",
  "H4K5ac (BRD9 bromodomain's normal ligand, retention in the fusion undetermined)"
]
How this could bite us back
  • A BRD9 degrader is the obvious idea and may be exactly wrong here. BRD9::NUTM1 retains only a 5' fragment of BRD9, breaking at exon 3, 8 or 14, and a degrader engages a ligandable bromodomain surface the fusion may not have. The likely outcome of a BRD9 degrader in a BRD9::NUTM1 patient is full degradation of wild-type BRD9 - with whatever ncBAF-dependent toxicity that carries - and no effect on the oncoprotein.
  • The standard of care is already close to a cure in this subtype. Any targeted agent must beat 4-year EFS of 100% in 35 uniformly treated infants with no transplants; the honest development path is de-escalation, and a new agent that adds toxicity is a net harm even if it improves a biomarker.
  • If chemosensitivity really rests on a dependence on active transcription, the mechanistically obvious drugs here - BET inhibitors, CDK9 inhibitors, p300 HAT inhibitors - all reduce transcription and could ANTAGONISE the chemotherapy that currently cures these children. That antagonism should be tested in vitro before any of them is combined with induction.
  • Attacking p300 attacks a protein whose loss is developmentally lethal, in an infant. The interface the NUTM1 moiety uses, TAZ2, is the same surface p53 TAD2 and p63 TAD use, so an interface inhibitor would be expected to perturb p53 signalling - in a leukaemia whose current cure depends on chemotherapy-induced apoptosis.
  • The patients who might actually need a targeted agent are the outliers - the reported high-risk-feature and poor-outcome NUTM1-r cases - and they cluster in the HOXA9-NEGATIVE partner group, not with BRD9. Developing a BRD9-directed agent would target the patients least likely to need it.
Still open
  1. Which BRD9 residues does each breakpoint retain? Breakpoints at BRD9 exons 3, 8 and 14 against a bromodomain at aa 136-240 almost certainly means some patients' fusions keep the bromodomain and some do not. This is a one-afternoon question - map the three exon boundaries to residues in the canonical transcript - and it decides whether a BRD9 bromodomain ligand or degrader can engage the fusion at all, patient by patient.
  2. Does BRD9::NUTM1 recruit p300, and does it form nuclear foci? Both are assumed from the BET fusions and neither has been shown for this fusion. A co-IP and an immunofluorescence in the mouse model from the Blood 2026 paper would settle it.
  3. Is the fusion nuclear if the bromodomain is gone? NUTM1's leptomycin-sensitive NES is retained in every one of these fusions, and nuclear retention in BRD-NUT fusions is attributed to the BRD moiety's chromatin binding. An exon-3 BRD9 fusion could be substantially cytoplasmic, which would be the first reported instance of a NUTM1 fusion that is not nuclear.
  4. Why does this fusion, alone among NUTM1 fusions, track with the pro-B immunophenotype in humans and produce a prepro-B disease in the mouse? n=3 in humans, but the mouse agrees, and the correspondence suggests the 5' partner sets the differentiation stage rather than just the chromatin address.
  5. Is a BRD9::NUTM1 blast sensitive to a menin inhibitor? It expresses HOXA9 with no KMT2A fusion and no menin interface. A negative answer would localise menin-inhibitor activity to the KMT2A complex; a positive one would relocate it to HOXA dependence, which would widen the drug class's indication and weaken the interface rationale.
Every source behind this entry (7)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Among 85 NUTM1-rearranged B-ALL cases from the Ponte di Legno Childhood ALL Working Group and Interfant-99/-06, BRD9 was the partner in 10 of 39 infant cases with a known partner (26%), second to ACIN1 (17, 44%) and ahead of CUX1 (6, 15%); all three pro-B-immunophenotype infants of 32 with immunophenotype data carried BRD9-NUTM1; BRD9-NUTM1 was associated with HOXA9 expression; and across the subtype 4-year event-free survival was 100% in 35 uniformly treated Interfant infants against 74.0% (95% CI 65.1-81.0) in 126 NUTM1-wildtype/KMT2A-wildtype infants (p=0.001), with all cases in complete remission, 41 of 45 (91.1%) with MRD below 0.05% against 48% in KMT2A-rearranged infant ALL, and no transplants in first remission.

    PMID 34211097DOI
    What this evidence cannot show

    Read in full from the Europe PMC open-access full text (PMC8478641); PubMed carries no abstract for this letter. The outcome figures are for the NUTM1-rearranged subtype as a whole, NOT for BRD9::NUTM1 specifically - this study does not report survival by partner, so no partner-specific prognosis can be drawn from it, and the 'all three pro-B cases were BRD9' observation is n=3. Retrospective; only the 35 Interfant infants are uniformly treated, and a 100% Kaplan-Meier estimate at n=35 with no events does not have a lower bound of 100%.

    Evidence that argues against this (1)
    • Partner identity does appear to matter within the subtype in at least one direction: HOXA9 expression was low or absent in 4 of 5 paediatric ZNF618-NUTM1 cases in this same study, and a ZNF618::NUTM1 case has been reported with a poor outcome. So 'NUTM1-r is favourable' should not be read as 'every NUTM1 fusion is favourable', and BRD9's membership of the HOXA9-positive group is a biological rather than a prognostic statement. Chen X et al. 2025
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In 175 infants with BCP-ALL diagnosed 2010-2023 at a single centre, the KMT2A-wild-type group was 33 cases (17.7%), of which NUTM1-r was 9 (27%) with 2-year event-free survival of 80% (95% CI 52-100) against 36% (21-59) for KMT2A::AFF1 and 37% (23-60) for other KMT2A partners; the most common NUTM1 fusion in that group was BRD9::NUTM1, with breakpoint variability at BRD9 exons 3, 8 and 14.

    Genetic Diversity in KMT2A-r and KMT2A-Wt Groups: Assessing the Prognostic Value of Markers in BCP-ALL Among Infants · Ilyasova K, Zerkalenkova E, Soldatkina O, et al. · International Journal of Laboratory Hematology 2025
    PMID 40000924DOI
    What this evidence cannot show

    The NUTM1-r arm is n=9 and its confidence interval reaches 100%; the authors say so themselves ('the small cohort size limit the statistical power'). The 80% versus 36% contrast must not be made to carry more than n=9 supports, and the log-rank p between the KMT2A-wild-type subgroups was 0.9. Single centre, retrospective, one country. THE BRD9-COMMONEST CLAIM IS COHORT-SPECIFIC AND DOES NOT GENERALISE: in the 85-case international series ACIN1 is the commonest infant partner at 44% with BRD9 second at 26% [E1], and an independent 2025 paper calls ACIN1::NUTM1 'the most frequent fusion' in NUTM1-R infant ALL. The exon-3/8/14 breakpoint observation is from 9 patients and no exon-to-residue mapping is given.

    Evidence that argues against this (1)
    • ACIN1, not BRD9, is the most frequent NUTM1 partner in infants in the largest series and in an independent 2025 report, so this cohort's 'most common fusion' is a property of nine patients. Tamai M et al. 2025
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Expression of BRD9-NUTM1 is sufficient to induce serially transplantable prepro-B-like B-cell-progenitor leukaemia in vivo, recapitulating the key molecular and immunophenotypic features of human NUTM1-rearranged B-ALL; NUTM1 fusions establish an aberrant chromatin state with global enhancement of H3K27 acetylation and distinctive open chromatin regions co-opting B-lineage and stemness programmes including NF-kB and posterior HoxA genes, feature global DNA hypomethylation irrespective of the 5' fusion partner, and confer profound chemosensitivity 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, not the full text. The transduction system, recipient genotype, animal numbers, latency and penetrance behind 'sufficient' are all unread, as is the assay behind 'global DNA hypomethylation' and the definition of 'profound sensitivity to chemotherapy'. 'Sufficient' in an overexpression model is a statement about that model; the human disease arises prenatally in a fetal progenitor [E5], which no overexpression model reproduces. BRD9-NUTM1 is described as 'a representative fusion', so the in-vivo result is one fusion standing for ten.

    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; its existence is recorded rather than the primary paper being presented as unchallenged. Shiloh R et al. 2026
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    BRD9 (Q9H8M2, 597 aa) carries a bromodomain at aa 136-240 with an H4K5ac/H4K8ac-binding region at aa 214-216 and is the defining subunit of the non-canonical BAF (ncBAF) complex; it is 5p15.33 by HGNC (HGNC:25818), and this graph's existing gene-brd9 entry classifies it as a KMT2A-rearranged dependency and drug target rather than as a demonstrated interactor.

    What this evidence cannot show

    HGNC location 5p15.33 fetched live 2026-09-22 from rest.genenames.org; the BRD9 domain boundaries and ncBAF membership are quoted from this graph's own gene-brd9 entry, which sourced them from live UniProt in cycle 41. THE CRITICAL LIMIT: all of this is about FULL-LENGTH BRD9. BRD9::NUTM1 retains only a 5' portion, breaking at exon 3, 8 or 14 [E2], and no source read here maps those exons onto the 136-240 bromodomain. Whether the fusion retains a functional bromodomain is therefore UNKNOWN, and nothing in this entry should be read as saying it does.

    Evidence that argues against this (1)
    • This graph's collision registry already records that querying BRD9 against leukaemia returns BRD9::NUTM1 fusion papers rather than ncBAF papers, i.e. the fusion literature and the complex literature are distinct bodies of work about the same symbol. That is exactly the reason not to transfer ncBAF conclusions to this fusion. oncofusion-knowledge-graph, data/symbol-collisions.json 2026
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    NUTM1-rearranged infant BCP-ALL arises prenatally: in an umbilical cord blood sample from one infant patient, the patient-specific CUX1::NUTM1 fusion gene and two IG/TR leukaemic markers were already present and detectable at birth.

    Prenatal origin of NUTM1 gene rearrangement in infant B-cell precursor acute lymphoblastic leukaemia · Bardini M, Fazio G, Abascal LC, et al. · British Journal of Haematology 2024
    PMID 39099338DOI
    What this evidence cannot show

    ONE patient, and the fusion is CUX1::NUTM1, not BRD9::NUTM1. It is cited here for the class-level claim that a NUTM1 fusion can be a prenatal event in this disease, which bears on how any model should be read, and NOT as evidence about BRD9::NUTM1 specifically. Backtracking studies are also subject to sensitivity limits in the opposite direction - failure to detect in cord blood would not have proven a postnatal origin.

    Evidence that argues against this (1)
    • A commentary on backtracking childhood leukaemia to birth appeared in the same journal and frames these single-patient backtracking results as a 'battle of addition', i.e. as accumulating case-level evidence rather than as a settled class-level fact. British Journal of Haematology 2024
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Native NUT shuttles between nucleus and cytoplasm through an NLS and a leptomycin-sensitive NES, whereas BRD3-NUT and BRD4-NUT are strictly nuclear because the BRD moiety retains NUT in the nucleus via chromatin interactions, with FRAP showing markedly slower nuclear diffusion for the fusions.

    PMID 17934517DOI
    What this evidence cannot show

    The 5' partners here are BRD3 and BRD4, double-bromodomain BET proteins in carcinoma cells - NOT BRD9, and not a lymphoid cell. Applying this to BRD9::NUTM1 is an inference from family resemblance and is labelled as one; the BRD9 fragment retained by the fusion may or may not bind chromatin at all [E4].

    Evidence that argues against this (1)
    • The same paper is the reason the inference is uncertain: nuclear retention is attributed specifically to the BRD moiety's chromatin interaction, so a 5' fragment that has lost its bromodomain would be predicted NOT to be retained - and the BRD9 breakpoints vary between patients. Ilyasova K et al. 2025
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The NUTM1 moiety retained in these fusions contains two transactivation domains, TAD1 (403-418) and TAD2 (419-470), which bind the p300 TAZ2 domain and allosterically activate the acetyltransferase, producing acetylation-driven condensation on chromatin comprising H3K27 and H3K18 acetylation together with BRD4L/S, CDK9, MED1 and RNA polymerase II.

    PMID 36690674DOI
    What this evidence cannot show

    NMR structures 7XEZ and 7XFG, verified live against RCSB 2026-09-22; the work is on BRD4-NUT in a carcinoma setting. It applies to BRD9::NUTM1 only because nearly the whole NUTM1 ORF is retained in both, so the TADs at 403-470 are present; the p300 recruitment has not been demonstrated for BRD9::NUTM1 itself, and the Blood 2026 leukaemia work reports the H3K27ac consequence without naming the writer [E3].

    Evidence that argues against this (1)
    • Condensate formation and transcriptional output by BRD4-NUT are reported to be driven by overlapping but distinct molecular features, so a fusion with a different 5' partner could reproduce one and not the other. Scientific Reports 2023
Connected to