Oncofusion

start / graph / fusion-nup98-kmt2a

Fusionmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Qn-nup98-family

NUP98::KMT2A, inv(11)(p15q23) — the NUP98 fusion that contains KMT2A and is menin-inhibitor RESISTANT

also called NUP98-MLL · NUP98::MLL · NUP98-KMT2A · inv(11)(p15q23) · iNUP98-KMT2A

This rare rearrangement joins the NUP98 protein to almost the whole of KMT2A — the very gene whose fusions menin drugs were built for. It is the one NUP98 fusion those drugs do not work on, because the join deletes the exact 20 amino acids that menin grips.

The same thing, for a scientist

**The most informative negative result in this beat.** NUP98::KMT2A arises from an **inversion, inv(11)(p15q23), not a translocation** — both genes sit on chromosome 11 (NUP98 at 11p15, KMT2A at 11q23) — and was described in two adult AML patients in whom it was the isolated chromosomal abnormality [E1]. It fuses the NUP98 FG repeats to **almost the entire KMT2A open reading frame** while **deleting KMT2A exon 1**, which encodes the menin/LEDGF interaction region: UniProt Q03164 places the **menin-binding motif at residues 6–25** and the LEDGF/PSIP1-binding IBM1/IBM2 at 123–134 and 147–152, all lost [E2][E5]. Everything downstream is retained — AT-hooks, CXXC, PHD1–4, bromodomain, FYRN/FYRC, the **WIN motif (3762–3767)** and the **SET domain (3829–3945)** [E5]. The functional consequence is the opposite of every other NUP98 fusion: in an inducible transgenic mouse, leukaemic blasts express **very low levels of Hoxa5/a7/a9/a10, Meis1, Eya1 and Pbx3** compared with normal marrow or with KMT2A-ENL/KMT2A-AF9 blasts, matching the low HOXA seen in the two patients, and the cells are **resistant to the menin inhibitor MI-2-2 at 3–12 µM and to JQ1 at 0.05–0.5 µM**, concentrations at which KMT2A-AF9 cells arrest [E2][E1]. The authors' own conclusion: *"targeting KMT2A might not be suitable for efficient therapeutic interference with NUP98-KMT2A AML"* [E2].

Go deeper
Why this entry matters more than its patient count

Two patients is not an epidemiology. It is, however, a natural experiment that no designed study has matched.

The claim this beat defends is that NUP98-rearranged leukaemia is the second menin-dependent fusion family, because NUP98 fusions borrow wild-type MLL1 and menin tethers it (mechanism-nup98-fusion-menin-mll1-dependency). If that is right, then a NUP98 fusion that carries KMT2A but lacks the menin-binding motif should be the exception — and it is. It has low rather than high HOXA output and it is drug-resistant 2.

That is exactly the pattern a menin-interface model predicts and it is not what a generic "NUP98 FG repeats transform cells" model predicts. A naive condensate-only model would expect NUP98::KMT2A to behave like the others, since it retains the FG repeats that form condensates. It does not.

So this fusion functions as a positive control for the menin hypothesis and a negative control for the FG-repeat hypothesis at the same time, which is why it earns an entry despite n=2.

The mouse model, with the numbers

Inducible transgenic iNUP98-KMT2A: the human fusion ORF in the Col1a1 locus under a tetracycline-responsive element, rtTA in Rosa26, doxycycline-driven 2.

  • Median latency 80 weeks (n=22), range 13–106; over 90% developed disease.
  • 6 of 22 developed transplantable AML, median latency 62 weeks; transplant into wild-type recipients induced disease in 100% of recipients.
  • Bone-marrow transplant of naïve transgenic marrow accelerated disease to median 32 weeks (n=5, p=0.0007); sublethal 600 cGy irradiation accelerated to median 26 weeks (p=0.0003); doxycycline withdrawal extended to median 49 weeks (n=6, p=0.002).
  • Phenotype is MDS-like: reduced marrow cellularity, anaemia, raised WBC, extramedullary haematopoiesis, multilineage dysplasia, LSK expansion with a competitive repopulation advantage (+20% CD45.2 chimerism by 25 weeks, p=0.019).
  • FLT3-ITD did not cooperate — retroviral FLT3-ITD did not accelerate disease, in direct contrast to NUP98-HOXD13 where FLT3-ITD knock-in gives fully penetrant AML.
  • No spontaneous Nras/Kras/Npm1 mutations in three AML mice, unlike the NHD13 model.

The proposed mechanism is cell-cycle interference and escape from replicative senescence — MEFs propagated beyond 40 passages, G1 accumulation, downregulation of Sirt1, Igf1r, Rbl2, Tert, Foxo3, Prkcd — with the authors stating the fusion "interferes with cell cycle progression rather than primarily blocking differentiation" 2. That is a different disease mechanism from every other NUP98 fusion in this graph.

Dominant-negative on wild-type KMT2A

The authors interpret the reduced HOXA output as the fusion acting dominant-negatively on normal KMT2A 2. The structural reading is available: a protein carrying the CXXC domain, the PHD fingers, the WIN motif and an intact SET domain but no menin tether would be expected to occupy KMT2A target sites and assemble a COMPASS-like complex while being unable to be held there by menin, and to compete with endogenous KMT2A for WDR5 and for the same sites.

That reading is consistent with the data but is not demonstrated: no competition assay, no WIN-mutant version of the fusion, and no endogenous-KMT2A occupancy measurement is reported. It is recorded here as the mechanism the authors propose, not as an established one.

Nomenclature, and a trap

The brief this beat was dispatched under described this lesion as t(11;11). It is not: NUP98 is at 11p15 and KMT2A at 11q23, both on chromosome 11, so the rearrangement is an inversion, inv(11)(p15q23) 1. Any pipeline that searches for a t(11;11) NUP98::KMT2A will find nothing.

A second trap: PubMed's query parser breaks on translocation parentheses. NUP98 KMT2A MLL fusion t(11;11) leukemia returns 0 records; NUP98-MLL fusion gene acute myeloid leukemia returns 1 (the founding paper) and NUP98::KMT2A returns 1 (the functional paper). Searching this lesion under the current symbol pair alone would have concluded it does not exist.

inv(11)(p15q23) has been reported cytogenetically in peripheral T-cell lymphoma, childhood ALL, myeloma and hairy cell leukaemia, but fusion expression has been molecularly confirmed in only two AML patients 2. Cytogenetic reports of the inversion should not be counted as NUP98::KMT2A cases.

Facts and figures
gene5NUP98
gene3KMT2A
cytobandinv(11)(p15q23)
lesion typeinversion, NOT a translocation (both genes on chromosome 11)
common misnomert(11;11) - incorrect; do not use
uniprot5P52948
uniprot3Q03164
length AA 51817
length AA 33969
hoxa outputREDUCED (opposite of every other NUP98 fusion)
menin inhibitor sensitivityRESISTANT (MI-2-2, 3-12 uM, no growth impairment)
bet inhibitor sensitivityRESISTANT (JQ1, 0.05-0.5 uM)
flt3 itd cooperationnone observed
mouse median latency weeks80
prognosisunknown; two reported patients, no cohort outcome data
diseases
[
  "AML",
  "MDS"
]
breakpoints
[
  {
    "id": "bp-nup98-kmt2a-inv11",
    "domains_retained_5": [
      "FG repeats 1 (1-156)",
      "GLEBS (157-213)",
      "FG repeats 2 (214-480)"
    ],
    "domains_lost_5": [
      "peptidase S59 autoproteolytic domain (738-880)",
      "Nup96 chain (881-1817)"
    ],
    "domains_retained_3": [
      "AT-hooks (169-180, 217-227, 301-309)",
      "CXXC (1147-1195)",
      "PHD1-4",
      "bromodomain (1635-1765)",
      "FYRN (2018-2074)",
      "FYRC (3666-3747)",
      "WIN motif (3762-3767)",
      "SET (3829-3945)"
    ],
    "domains_lost_3": [
      "menin-binding motif MBM (6-25)",
      "LEDGF-interaction IBM1 (123-134)",
      "LEDGF-interaction IBM2 (147-152)"
    ],
    "phenotype_notes": "Loss of exon 1 removes the menin and LEDGF tethers; HOXA output is REDUCED, not elevated, and the cells are menin-inhibitor resistant."
  }
]
frequency
{
  "molecularly_confirmed_patients": 2,
  "note": "inv(11)(p15q23) reported cytogenetically in other malignancies, but fusion expression confirmed only in 2 AML patients; absent from the 160-patient paediatric NUP98 landscape cohort"
}
How this could bite us back
  • If NUP98-rearranged leukaemia is enrolled on menin-inhibitor trials by cytogenetic NUP98 rearrangement alone, a NUP98::KMT2A patient would be enrolled and would be expected to fail, diluting the very subgroup this beat is arguing for.
  • This fusion is a warning that 'NUP98-rearranged' is not one disease for drug-response purposes; partner-blind eligibility criteria are a design flaw, not a convenience.
  • The dominant-negative model, if right, means this leukaemia is driven partly by LOSS of normal KMT2A function - and a menin inhibitor, which also removes KMT2A from chromatin, would be expected to make it worse rather than better.
Still open
  1. Does genetic MEN1 deletion, as opposed to MI-2-2, spare NUP98-KMT2A cells? The resistance claim currently rests on a micromolar tool compound and a genetic arm would make it decisive.
  2. Does NUP98-KMT2A form FG condensates? If it does and still behaves differently, the condensate model loses its explanatory monopoly; if it does not, the retained FG repeats are doing something else. Nobody has run the assay on this fusion.
  3. Is the dominant-negative interpretation right? A WIN-motif-mutant version of the fusion, or a measurement of endogenous KMT2A occupancy in the presence of the fusion, would test it.
  4. Would a WDR5 or SET-domain-directed agent work here where menin inhibition does not? This fusion retains both, which makes it the one NUP98 fusion for which the WIN-site strategy is on-mechanism.
  5. How many of the cytogenetically reported inv(11)(p15q23) cases in lymphoma, ALL, myeloma and hairy cell leukaemia actually express the fusion? Nobody has gone back with a molecular assay.
Every source behind this entry (4)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    NUP98-MLL (NUP98::KMT2A) was identified in two adult AML patients with inv(11)(p15q23) as the isolated chromosomal abnormality, and the patients' samples showed low levels of HOXA gene expression.

    NUP98-MLL fusion in human acute myeloblastic leukemia · Kaltenbach S, et al. · Blood 116(13):2332-2335 2010
    PMID 20558618DOI
    What this evidence cannot show

    n = 2, both adults; the low-HOXA observation in patient samples is from a two-patient series with no matched controls described in the abstract. The fusion has not been found again in any of the large paediatric NUP98 cohorts, so its true frequency is unknown and may be below the detection floor of every published series.

    Evidence that argues against this (1)
    • The largest paediatric NUP98 cohort (160 NUP98-rearranged of 2,235) enumerates NSD1, KDM5A and 13 other partners and does not report KMT2A among them, so in children this fusion is either absent or vanishingly rare, and conclusions drawn from two adults may not transfer to the paediatric population this project is about. Bertrums EJM et al. 2023
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In an inducible iNUP98-KMT2A transgenic mouse, leukaemic blasts express very low levels of Hoxa5/a7/a9/a10, Meis1, Eya1 and Pbx3 compared with normal marrow or KMT2A-ENL/KMT2A-AF9 blasts; the cells are resistant to the menin inhibitor MI-2-2 (3-12 uM) and to JQ1 (0.05-0.5 uM) at doses that arrest KMT2A-AF9 cells; median disease latency is 80 weeks (n=22) with 6 of 22 developing transplantable AML; and FLT3-ITD does not cooperate.

    PMID 31558671DOI
    What this evidence cannot show

    Single laboratory, single model, and the model is a transgenic overexpression system rather than a knock-in at the endogenous locus, so expression level is not physiological. MI-2-2 is a tool compound, not revumenib or any clinical agent, and micromolar concentrations are high enough that failure to respond could reflect compound properties rather than target biology - no genetic MEN1 deletion arm is reported. The very long latency means secondary events are certain to have accumulated.

    Evidence that argues against this (2)
    • Resistance to a tool compound at micromolar concentrations is weaker evidence than a genetic test would be. The founding menin-dependency study in other NUP98 fusions used CRISPR knockout of Men1 and Kmt2a alongside a compound; no equivalent genetic arm exists for NUP98-KMT2A, so 'menin-independent' is supported pharmacologically and not genetically. Heikamp EB et al. 2022
    • The FG repeats retained by this fusion are the module reported to drive condensate formation and leukaemic transformation in other NUP98 fusions, yet this fusion behaves differently. Either the condensate model does not generalise, or this fusion forms condensates that do not produce the same output - neither has been tested, because no condensate assay has been run on NUP98-KMT2A. Chandra B et al. 2022
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    NUP98 is P52948, 1817 aa, with FG repeats 1 at residues 1-156, the GLEBS/RAE1-interaction region at 157-213, FG repeats 2 at 214-480 and the autoproteolytic peptidase S59 domain at 738-880; its only UniProt gene-name synonym is ADAR2, a historical naming collision.

    What this evidence cannot show

    The primary gene name was read back from genes[0].geneName.value and the length from sequence.length, because UniProt's gene_exact: search matches synonyms. UniProt does NOT annotate an RRM or RNA-binding domain in P52948, so the C-terminal RNA-binding region described in the leukaemia literature is deliberately not given residue boundaries here.

    Evidence that argues against this (1)
    • The ADAR2 synonym on P52948 is a live trap for automated alias matching: an alias sweep keyed on NUP98's UniProt synonym list would pull in RNA-editing literature that has nothing to do with this gene. Alias searches in this project have failed in both directions before, and this is the inflation direction. UniProt Consortium 2026
  4. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    KMT2A is Q03164, 3969 aa, with the menin-binding motif annotated at residues 6-25 and LEDGF-interaction motifs at 123-134 and 147-152 - all encoded by exon 1 and therefore deleted by this fusion - while the AT-hooks, CXXC (1147-1195), bromodomain (1635-1765), WIN motif (3762-3767) and SET domain (3829-3945) are retained.

    What this evidence cannot show

    The mapping from 'KMT2A exon 1 is lost' to 'residues 6-25 are lost' combines the fusion paper's exon-level statement with UniProt's residue-level annotation. The two were not checked against a single source giving both, so the residue claim is an inference from two verified facts rather than a directly reported one.

    Evidence that argues against this (1)
    • If the breakpoint were slightly different between the two reported patients, the retained KMT2A N-terminus could differ, and neither patient's exact junction sequence is available to check. A fusion retaining even a partial menin-binding motif would change the interpretation entirely. Kaltenbach S et al. 2010
Connected to