Oncofusion

start / graph / gene-nup98

Genestrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q126-the-partners-named-but-never-written

NUP98 - the mobile nucleoporin that leaves the pore to put H3K4me3 on haematopoietic promoters, and the shared 5' donor of seven fusions in this graph

also called nucleoporin 98 · Nup98 · NUP98-NUP96 precursor · NUP96 · Nup98-96 · NUP196 · ADAR2 (UniProt-listed gene-name synonym on P52948, NOT the RNA-editing enzyme ADARB1)

NUP98 is best known as part of the gateway in the wall of the cell nucleus, but a large share of it detaches from that gateway and works inside the nucleus, switching genes on. Seven different leukaemia fusions in this project all begin with the same piece of NUP98, and until now the graph described all seven and never described NUP98 itself.

The same thing, for a scientist

**NUP98 is UniProt P52948 (NUP98_HUMAN, 1,817 aa as the NUP98-NUP96 polyprotein precursor), HGNC:8068 `nucleoporin 98 and 96 precursor` at 11p15.4, primary Ensembl ENSG00000110713; `gene_exact:NUP98 AND organism_id:9606 AND reviewed:true` returned exactly one row, so the rule-62 several-rows trap does not fire, and HGNC independently carries P52948 and ENSG00000110713** [E1][E2]. One gene makes two nucleoporins: the precursor autocleaves at 880/881 into NUP98 and NUP96, which is why HGNC's gene name carries both numbers, and it is why a NUP98 residue number is meaningless without a stated length (`protein-nup98-domains` resolves the 1817-versus-920 offset). **The part this graph needed and did not have is that NUP98 is not only a pore protein.** Photobleaching of GFP-NUP98 in living cells shows it moving between the nuclear pore, the nucleoplasm and a discrete intranuclear structure the authors named the **GLFG body**, with mobility at both the pore and in the nucleoplasm **dependent on ongoing transcription by RNA polymerases I and II** [E3]. UniProt's own subcellular-location block records the same three addresses - nuclear membrane (nucleoplasmic side), nuclear pore complex, and nucleoplasm - and explicitly notes 'diffuse and discrete intranuclear foci (GLFG-body)' [E1]. The off-pore pool is the functional one for gene control: NUP98 binds developmentally regulated genes genome-wide in human cells, at the pore for genes in the initial stage of induction and **in the nuclear interior, away from the pores, for genes that are highly induced** [E4]. **And the epigenetics is haematopoietic and specific.** In haematopoietic cells NUP98 binds predominantly to transcription start sites and **recruits the Wdr82-Set1A/COMPASS complex; depleting NUP98 or WDR82 abolishes SET1A recruitment and ablates H3K4me3 at the adjacent promoters**, and a NUP98 fusion protein mislocalises H3K4me3 to abnormal regions and up-regulates the associated genes [E5]. That is the wild-type function the fusions corrupt, measured in the right lineage. **What makes it a poor drug target and a hard genetic experiment.** NUP98 is close to pan-essential: across **1,258 DepMap screens the mean Chronos gene effect is -0.809 and 91.3% of lines fall below -0.5** - against 28.9% for MEN1 on the same call [E6]. Separately, combined *Rae1* and *Nup98* haploinsufficiency in mice causes premature sister-chromatid separation, severe aneuploidy and untimely securin degradation, because the RAE1-NUP98 complex restrains APC/C-Cdh1 in early mitosis [E7]. **Losing NUP98 is not a clean loss of a transcriptional co-activator; it is loss of nucleocytoplasmic transport and of a mitotic brake at the same time.**

Go deeper
The subnuclear dimension, which is the reason this entry exists in the form it does

The PI's standing list separates subcellular from subnuclear localisation, and NUP98 is the case that shows why. Subcellularly it is a nuclear-envelope protein and nothing about that is interesting here. Subnuclearly it occupies three addresses that behave differently:

addressevidence classwhat it does there
nuclear pore complex, nucleoplasmic side / basketimmunofluorescence and immuno-EM, multiple studies 1transport, and pore-proximal binding of genes at the initial stage of induction 4
nucleoplasm, diffuseFRAP of GFP-NUP98; recovery indicates interaction with less-mobile nucleoplasmic components 3binds TSSs, recruits Wdr82-Set1A/COMPASS, sets H3K4me3 5; binds highly induced genes away from the pore 4
GLFG body, discrete intranuclear fociFRAP plus domain mapping; the GLFG domain is required for targeting there 3; colocalised with DHX9 18least understood of the three - see the caveat below

The dynamics are transcription-coupled, not constitutive. Inhibiting RNA polymerase I and II changes NUP98's mobility at the pore and in the nucleoplasm 3. A localisation statement for NUP98 that does not name a transcriptional state is therefore underspecified, which is a general point this graph's localisation entries should absorb.

The GLFG body carries a real caveat and it is not a minor one. GLFG bodies were characterised in cells overexpressing exogenous Nup98 or in a HeLa subline with unusually high endogenous Nup98; a later study found that a second HeLa-derived line, HEp-2, forms GLFG bodies and expresses truncated NUP98 forms fused to a non-coding region of 11q22.1, with massive amplification of the juxtaposed NUP98 and 11q22.1 loci 9. So in at least two of the lines the structure is named after, the structure travels with a NUP98 genomic rearrangement. It should not be quoted as an ordinary feature of a normal cell without saying so.

What the seven fusion entries assumed about NUP98 and never cited

This corpus holds seven fusion-nup98-* entries and names NUP98 in 221 node files, and had no gene-nup98 (python scripts/is-symbol-covered.py NUP98 -> NUP98 no match under 9 candidate names, run 2026-09-23). The consequence is measurable rather than rhetorical: of the seven primary sources this entry is built on, scripts/check-citation-reuse.py reports six as NOT CITED HERE - genuinely new to this graph - Griffis 2002 (11950939), Franks 2017 (29269482), Liang 2013 (23468646), Capitanio 2017 (28221134), Jeganathan 2005 (16355229) and Romana 2016 (26685999). Only the NUP98-HOXA9 mouse model (11157742) was already cited, and it is a fusion paper.

So the graph had imported NUP98's leukaemia biology in full and its normal biology not at all. That is the partner-import pattern this project already names for KMT2A::AFF1 evidence, appearing here on the 5' side of a different family.

One concrete thing the missing entry cost. localization-nuclear-pore-nup98-fg-comparator frames the NUP98 fusion story as 'a fusion repurposing a compartment-defining module' and describes the FG IDR as creating the permeability barrier of the pore. That is true and incomplete: a substantial NUP98 pool is already off the pore and already on chromatin, already recruiting an H3K4 methyltransferase complex, in normal haematopoietic cells 5. On that reading a NUP98 fusion is less a nucleoporin being repurposed for chromatin, and more a chromatin protein that lost its autoproteolytic C-terminus and kept its chromatin-facing N-terminus. This entry does not claim the second reading is correct - both are consistent with the cited data - but the corpus currently states only the first, and no experiment in this beat's reading discriminates them.

Cell type and cell state

Cell type. NUP98 is ubiquitously expressed and screened in all 1,258 DepMap lines 6, so a 'cell type' claim for the gene itself is weak by default. The claims worth recording are the ones where the function was measured in a named cell type: haematopoietic progenitor cells for the TSS/Wdr82-Set1A/H3K4me3 mechanism 5, and human embryonic stem cells undergoing differentiation for the genome-wide developmental-gene association 4. Both are the lineage this project cares about or its upstream.

Cell state. Two states change what NUP98 is doing, and both are documented:

  • Transcriptional state. Mobility at the pore and in the nucleoplasm is dependent on ongoing Pol I and Pol II transcription 3; and pore-proximal versus interior binding partitions by how strongly a gene is induced 4. NUP98's address is a function of transcription, not a fixed property.
  • Cell-cycle state. NUP98 dissociates from the disassembling NPC early in prophase 1, and in early mitosis it is in a complex with RAE1 and APC/C-Cdh1 inhibiting securin ubiquitination 7. Its mitotic job is not its interphase job.

A NUP98 fusion protein retains the N-terminal FG/GLFG region and therefore the chromatin-facing part, and loses the peptidase domain - see protein-nup98-domains for residue coordinates and the isoform-numbering conversion, which this entry deliberately does not restate.

Facts and figures
uniprotP52948
length AA1817
gene primaryNUP98
hgncHGNC:8068
hgnc namenucleoporin 98 and 96 precursor
ensemblENSG00000110713
cytoband11p15.4
accession resolutiongene_exact:NUP98 AND organism_id:9606 AND reviewed:true returned exactly one row, P52948, primary gene name NUP98; HGNC:8068 independently carries P52948 and ENSG00000110713. Both fetched live 2026-09-23. Rule 62 does not fire. NOTE: UniProt's gene-name field carries the legacy synonym 'ADAR2' on this record - it is not the RNA-editing enzyme.
polyprotein noteOne gene, two nucleoporins. The 1,817-aa precursor autocleaves at 880/881 into NUP98 (N-terminal) and NUP96 (C-terminal), which is why HGNC names the gene 'nucleoporin 98 and 96 precursor' and why NUP96 is an alias of NUP98. Residue coordinates, the isoform-numbering offset and domain-database disagreements are in protein-nup98-domains and are deliberately not restated here.
subcellular localizationnuclear envelope / nuclear pore complex, nucleoplasmic side, at or near the nuclear basket
subnuclear localizationTHREE addresses, not one: (1) the NPC nucleoplasmic basket; (2) diffuse nucleoplasm, where it binds transcription start sites; (3) the GLFG body, discrete intranuclear foci requiring the GLFG domain for targeting and colocalised with DHX9. Movement between them is transcription-dependent (Pol I and Pol II).
localization shown for fusionFor the WILD-TYPE protein the three-address pattern is shown in non-haematopoietic cultured cells (GFP-Nup98 FRAP, PMID 11950939). For a FUSION, the off-pore nucleoplasmic condensate is established for NUP98-HOXA9, NUP98-NSD1 and NUP98-PHF23 in this corpus's existing entries; NO published experiment found in this beat places a NUP98 FUSION protein in a GLFG body, or tests whether the fusion joins the wild-type protein's nucleoplasmic pool or forms a distinct one.
epigenetic readoutH3K4me3 at transcription start sites. Wild-type NUP98 recruits the Wdr82-Set1A/COMPASS complex to promoters in haematopoietic progenitors; NUP98 or WDR82 depletion abolishes SET1A chromatin recruitment and ablates promoter H3K4me3 (PMID 29269482). This is a WRITER-RECRUITMENT role, not catalysis: NUP98 has no methyltransferase activity of its own.
writer of that markSETD1A (Set1A/COMPASS), recruited via WDR82. NUP98 is the recruiter, not the writer.
coverage note deliberately not in the titleWritten cycle 88 after scripts/is-symbol-covered.py NUP98 returned 'no match under 9 candidate names' while NUP98 was named in 221 node files across seven fusion entries. Six of the seven primary sources used here were NOT previously cited anywhere in this corpus (scripts/check-citation-reuse.py, 2026-09-23). Recorded in data, per the standing rule that a coverage claim about this graph does not belong in a title.
relationship to existing entriesprotein-nup98-domains holds residue coordinates, isoform numbering and domain-database disagreements; localization-nuclear-pore-nup98-fg-comparator holds the fusion-condensate comparator argument. This entry holds the GENE and the WILD-TYPE protein's own biology - polyprotein identity, the three subnuclear addresses, the transcription-dependent dynamics, the Wdr82-Set1A/H3K4me3 recruitment function, the mitotic RAE1 role and the essentiality profile - none of which those two entries carry. It deliberately does not restate their content.
alias symbol
[
  "NUP96",
  "Nup98-Nup96",
  "Nup98-96",
  "NUP196"
]
prev symbol
[]
cell types where function was measured
[
  "haematopoietic progenitor cells (PMID 29269482)",
  "human embryonic stem cells during differentiation (PMID 23468646)",
  "mouse haemopoietic stem cells, for the fusion (PMID 11157742)"
]
cell states
[
  "transcriptionally active versus Pol I/Pol II-inhibited - changes NUP98 mobility at the pore and in the nucleoplasm",
  "initial induction versus high induction - changes whether a target gene meets NUP98 at the periphery or in the nuclear interior",
  "interphase versus early mitosis - NUP98 leaves the disassembling NPC in prophase and joins RAE1/APC-Cdh1"
]
essentiality
{
  "depmap_source": "Open Targets GraphQL target.depMapEssentiality, fetched 2026-09-23",
  "depmap_release": null,
  "open_targets_dataVersion": "year 26, month 06, iteration null",
  "n_screens": 1258,
  "chronos_mean": -0.809,
  "chronos_median": -0.7888,
  "frac_below_minus_0_5": 0.9126,
  "frac_below_minus_1_0": 0.1932,
  "leukaemia_strata": {
    "Acute Myeloid Leukemia": {
      "n": 35,
      "mean": -0.8168
    },
    "B-Cell Acute Lymphoblastic Leukemia": {
      "n": 13,
      "mean": -0.7335
    },
    "T-Lymphoblastic Leukemia/Lymphoma": {
      "n": 8,
      "mean": -0.8624
    },
    "Acute Leukemias of Ambiguous Lineage": {
      "n": 1,
      "mean": -0.8225
    }
  },
  "comparator_men1_same_call": {
    "chronos_mean": -0.3616,
    "frac_below_minus_0_5": 0.2893
  },
  "mouse_ko": "Combined Rae1 and Nup98 HAPLOINSUFFICIENCY (double heterozygote) gives premature sister-chromatid separation and severe aneuploidy (PMID 16355229). A Nup98-only null phenotype was not established in this beat's reading and is not asserted here."
}
fusions donating this gene as the 5 prime partner in this corpus
[
  "fusion-nup98-ddx10",
  "fusion-nup98-hoxa9",
  "fusion-nup98-hoxd13",
  "fusion-nup98-kdm5a",
  "fusion-nup98-kmt2a",
  "fusion-nup98-nsd1",
  "fusion-nup98-top1"
]
How this could bite us back
  • NUP98 itself is not a target and the numbers are not marginal: 91.3% of 1,258 DepMap lines fall below -0.5 and the AML stratum (-0.8168, n=35) is if anything slightly worse than the whole-panel mean. An agent that degraded NUP98 would take nucleocytoplasmic transport with it in every cell in the patient. Any therapeutic reading of this entry must be about the FUSION protein or about a partner, never about the wild-type gene.
  • The FG/GLFG region every NUP98 fusion retains is the same region wild-type NUP98 uses to bind DHX9 and to reach the GLFG body. An FG-repeat-directed agent - the thing the condensate literature keeps implying - would therefore hit the wild-type protein's own chromatin-facing functions, including the Wdr82-Set1A/H3K4me3 route in normal haematopoietic progenitors, which is exactly the compartment a paediatric leukaemia therapy has to preserve.
  • The mitotic role is a second, independent toxicity axis that the transcription-focused literature routinely omits. RAE1-NUP98 restrains APC/C-Cdh1 in early mitosis, and reducing the pair causes severe aneuploidy in mice. A partial-knockdown strategy therefore risks inducing aneuploidy in normal tissue - which, in a child, is a second-malignancy liability rather than a tolerable on-target effect.
  • This entry's own strongest mechanistic result is a knockdown result. 'Depletion of Nup98 abolishes Set1A recruitment' is compatible with NUP98 being a required scaffold and with NUP98 depletion being globally disruptive enough to move anything measured; no degron or separation-of-function allele appears in this beat's reading. Anyone building on the Wdr82-Set1A axis should establish the acute, reversible version of that experiment first.
Still open
  1. Does a NUP98 FUSION protein still recruit Wdr82-Set1A/COMPASS, or has it swapped that route for the KMT2/MLL partitioning its FG condensate is known to perform? The wild-type function (PMID 29269482) and the fusion-condensate composition (PMID 39922194) were established by different groups in different systems and nobody appears to have asked whether the fusion retains the wild-type job. A SET1A ChIP in a NUP98-fusion line against a wild-type haematopoietic progenitor is the experiment, and it is cheap.
  2. Do NUP98 fusions retain the RAE1-binding GLEBS motif in a functional state, and does the fusion therefore carry a mitotic liability? GLEBS is at 157-213, below every annotated NUP98 fusion breakpoint, so the motif is present on sequence. Whether the fusion binds RAE1, and whether it perturbs APC/C-Cdh1 restraint of securin, was not found tested in this beat's reading - and aneuploidy in a NUP98-rearranged leukaemia would be a phenotype nobody currently attributes to the fusion.
  3. Is the GLFG body a real compartment in a primary haematopoietic cell, or an overexpression and rearrangement artefact? Every reported instance involves exogenous overexpression, an unusually high-expressing HeLa subline, or a line carrying a NUP98 genomic rearrangement (PMID 26685999). This matters directly, because 'the fusion makes nuclear condensates' and 'NUP98 already makes an intranuclear body' are close enough that the graph could conflate them.
  4. Does the off-pore NUP98 pool depend on the FG repeats for chromatin binding in the way the fusion condensate does? If the wild-type off-pore pool needs the same module the fusions retain, the FG region is a chromatin-targeting module the fusion inherits rather than a phase-separation device it repurposes, and the two framings make different predictions about an FG-repeat-directed agent.
  5. What fraction of cellular NUP98 is off the pore at steady state in a haematopoietic progenitor, and does that fraction change with differentiation state? Nothing in this beat's reading quantifies the partition, so every mechanistic claim about the nucleoplasmic pool is currently a claim about an unmeasured fraction.
Every source behind this entry (10)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    NUP98 is UniProt P52948, a 1,817-residue NUP98-NUP96 polyprotein precursor; its subcellular-location block records nucleus membrane (peripheral, nucleoplasmic side), nucleus/nuclear pore complex and nucleus/nucleoplasm, with notes that it localises at or near the nucleoplasmic basket, dissociates from the disassembled NPC early in prophase of mitosis, and is colocalised with DHX9 in diffuse and discrete intranuclear foci (GLFG-body).

    What this evidence cannot show

    Fetched live 2026-09-23 via gene_exact:NUP98 AND organism_id:9606 AND reviewed:true, which returned exactly ONE row. RULE 91 APPLIES AND BITES HARD HERE, so the coded citations must not be quoted as NUP98 localisation papers without checking: of the ECO:0000269 PMIDs attached to the nuclear-membrane and basket statements, PMID 11839768 is Frosst et al., 'Tpr is localized within the nuclear basket of the pore complex and has a role in nuclear protein export' (J Cell Biol 2002) and PMID 12802065 is Hase & Cordes, 'Direct interaction with nup153 mediates binding of Tpr to the periphery of the nuclear pore complex' (Mol Biol Cell 2003) - both TPR papers in which a curator read a Nup98 observation - and PMID 33097660 is a SARS-CoV-2 Orf6 paper. All three titles were fetched by efetch this session. The founding observation for the mobile off-pore pool is PMID 11950939 (E3), which is NOT among the location-comment citations. UniProt's gene-name synonym list includes the string 'ADAR2' on this record; that is a legacy synonym and must not be read as the RNA-editing enzyme.

    Evidence that argues against this (1)
    • The same UniProt record's location comments are the clearest available instance of rule 91 in this corpus: the experimental evidence codes point at two TPR papers and a coronavirus paper rather than at the NUP98 dynamics literature. An agent filtering P52948 on ECO:0000269 and stopping there would record NUP98 as a nuclear-basket protein and would never see the transcription-dependent nucleoplasmic pool that is the entire reason NUP98 matters in this graph. Frosst P et al. 2002
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    HGNC:8068 assigns the symbol NUP98 the name 'nucleoporin 98 and 96 precursor' at cytogenetic location 11p15.4, with primary Ensembl gene id ENSG00000110713, UniProt P52948, alias symbols NUP96, Nup98-Nup96, Nup98-96 and NUP196, and no previous symbol.

    HGNC record for NUP98 (HGNC:8068) · HUGO Gene Nomenclature Committee · HGNC REST 2026
    What this evidence cannot show

    Fetched live 2026-09-23; exactly one document returned. NUP96 is listed as an ALIAS of NUP98 because both proteins come from this one precursor, which means a literature or database search on 'NUP96' reaches this gene and a reader may mistake a NUP96 result for a NUP98 result. This is a collision worth adding to data/symbol-collisions.json, where a grep for NUP96 currently returns no row. Rule 78: the HGNC primary ensembl_gene_id is the id used for the DepMap join in E6.

    Evidence that argues against this (1)
    • A single primary Ensembl id makes NUP98 look better resolved than it is. protein-nup98-domains already records that the live Ensembl gene record lists 54 protein-coding transcripts with 44 distinct protein lengths from 93 to 1848 aa, and that the Ensembl canonical transcript encodes 1800 aa - 17 residues shorter than the UniProt canonical. So ENSG00000110713 fixes the GENE and fixes nothing about which protein a given paper means, and HGNC's tidy one-row answer actively conceals that. Ensembl; EMBL-EBI 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Nup98 is found both at the nuclear pore complex and within the nucleus, where it associates with a novel intranuclear structure the authors named the GLFG body; the GLFG domain is required for targeting to that structure; photobleaching of GFP-Nup98 in living cells shows Nup98 is mobile and moves between these localisations, with a recovery rate indicating interaction with less-mobile nucleoplasmic components; and the mobility of Nup98 both within the nucleus and at the pore is dependent on ongoing transcription by RNA polymerases I and II.

    Nup98 is a mobile nucleoporin with transcription-dependent dynamics · Griffis ER, Altan N, Lippincott-Schwartz J, Powers MA · Mol Biol Cell 2002
    PMID 11950939DOI
    What this evidence cannot show

    Abstract read this session; full text not read. This is a GFP-tagged overexpression study in cultured cells, which is the standard caveat for FRAP-based nucleoporin work and is the same overexpression condition E9 shows can itself produce GLFG bodies. The transcription dependence is established by polymerase inhibition, so it is a pharmacological perturbation with the usual off-target exposure. No haematopoietic cell is involved.

    Evidence that argues against this (1)
    • The structure this paper names is not established as a feature of ordinary cells. A later study reports that GLFG bodies had previously been seen only in cells overexpressing exogenous Nup98 or in a HeLa subline with unusually elevated endogenous Nup98, and that the HEp-2 line's GLFG bodies come with truncated NUP98 forms fused to non-coding 11q22.1 and amplification of the juxtaposed loci. See E9. Romana S et al. 2016
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    At a genome-wide level the nuclear pore protein NUP98 associates with developmentally regulated genes active during human embryonic stem cell differentiation; overexpression of a dominant-negative NUP98 fragment decreases expression of NUP98-bound genes; and there are two modes of regulation distinguished by spatial localisation - genes at the initial stage of developmental induction can associate with NUP98 embedded in nuclear pores at the nuclear periphery, whereas highly induced genes interact with NUP98 in the nuclear interior away from the pores.

    Dynamic association of NUP98 with the human genome · Liang Y, Franks TM, Marchetto MC, Gage FH, Hetzer MW · PLoS Genet 2013
    PMID 23468646DOI
    What this evidence cannot show

    Abstract read this session; full text not read. The functional arm is a dominant-negative NUP98 fragment, not a knockout or a degron, so the direction of effect is established more firmly than its magnitude. The system is human embryonic stem cells and their differentiation, not haematopoietic cells, and the periphery-versus-interior assignment rests on the localisation assays used in that paper.

    Evidence that argues against this (1)
    • The periphery-versus-interior distinction is assay-limited in a way the two-mode model does not acknowledge. A chromatin-association readout cannot distinguish a gene contacting a pore-embedded NUP98 from a gene contacting a nucleoplasmic NUP98 molecule that happens to lie near the envelope, and the functional arm is a dominant-negative fragment that could act by titrating NUP98's partners rather than by removing NUP98 from its sites. A degron or an anchoring mutant would separate these; neither was used.
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In haematopoietic cells Nup98 binds predominantly to transcription start sites and recruits the Wdr82-Set1A/COMPASS complex, which is required for deposition of H3K4me3; depletion of Nup98 or Wdr82 abolishes Set1A recruitment to chromatin and ablates H3K4me3 at adjacent promoters; and expression of a Nup98 fusion protein implicated in aggressive AML causes mislocalisation of H3K4me3 to abnormal regions and up-regulation of the associated genes.

    Nup98 recruits the Wdr82-Set1A/COMPASS complex to promoters to regulate H3K4 trimethylation in hematopoietic progenitor cells · Franks TM, McCloskey A, Shokirev MN, Benner C, Rathore A, Hetzer MW · Genes Dev 2017
    PMID 29269482DOI
    What this evidence cannot show

    Abstract read this session; full text NOT read, so the identity of the specific AML fusion used in the mislocalisation arm is not stated here and must be checked before any partner-specific claim is built on it. The depletion arms are knockdowns rather than degrons, and 'abolishes' and 'ablates' are the authors' words for their own measurements. Haematopoietic progenitor system, not primary patient blasts.

    Evidence that argues against this (1)
    • This result puts an H3K4 methyltransferase complex downstream of WILD-TYPE NUP98 and a DIFFERENT one - the KMT2/MLL family - inside the NUP98-FUSION condensate, and the graph should not merge them. PMID 39922194, already cited in this corpus, shows the FG repeats of NUP98 oncofusions selectively partition KMT2/MLL H3K4 methyltransferases, histone acetyltransferases and BRD4, while the partner's PHD finger supplies targeting. Wild-type NUP98 recruiting Wdr82-SET1A and a NUP98 fusion condensate concentrating KMT2/MLL are two distinct H3K4 routes, and no published experiment in this beat's reading tests whether the fusion still does the SET1A job. Ahn JH et al. 2025
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    NUP98 is near-pan-essential in DepMap: over 1,258 screens the mean Chronos gene effect is -0.809, the median -0.7888, with 91.3% of lines below -0.5 and 19.3% below -1.0. On the same call and the same denominator, MEN1 is -0.3616 with 28.9% below -0.5, NSD1 -0.1231 with 2.9%, JAZF1 -0.3954 with 27.7%, MEAF6 -0.3565 with 24.3% and PHF1 -0.0700 with 0.2%.

    What this evidence cannot show

    Fetched live 2026-09-23; computed by this beat and persisted with the per-gene numbers and the leukaemia strata in analyses/q126-the-partners-named-but-never-written/depmap-essentiality.json (rule 90: the identities and strata are on disk, not just the summary). RULE 83: the API returns meta.dataVersion {year 26, month 06, iteration null}, which is the OPEN TARGETS platform release and NOT a DepMap release name such as 24Q2, so no DepMap release is recorded and none is invented. The figures are means over all lineages; per rule 84 a whole-panel mean is not a lineage result. NUP98 leukaemia strata, reported as counts and not as contrasts: AML n=35 mean -0.8168, B-ALL n=13 mean -0.7335, T-ALL n=8 mean -0.8624 - no leukaemia-specific sparing, which is what near-pan-essentiality predicts.

    Evidence that argues against this (1)
    • A CRISPR knockout score is the wrong instrument for the question anyone would actually ask about NUP98 in a NUP98-rearranged leukaemia, and quoting -0.809 as though it settled druggability would be an error. The knockout removes the pore protein, nucleocytoplasmic transport and the mitotic RAE1-NUP98 brake together; no published perturbation in this beat's reading separates the chromatin-facing off-pore pool from the pore pool, and the FG-repeat region the fusions retain is shared by both. Jeganathan KB et al. 2005
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Combined Rae1 and Nup98 haploinsufficiency in mice results in premature separation of sister chromatids, severe aneuploidy and untimely degradation of securin; Rae1 and Nup98 form a complex with Cdh1-activated APC in early mitosis and specifically inhibit APC(Cdh1)-mediated ubiquitination of securin.

    The Rae1-Nup98 complex prevents aneuploidy by inhibiting securin degradation · Jeganathan KB, Malureanu L, van Deursen JM · Nature 2005
    PMID 16355229DOI
    What this evidence cannot show

    Abstract read this session; full text not read. The in vivo phenotype requires COMBINED Rae1 and Nup98 haploinsufficiency - it is a double-heterozygote result, not a Nup98 single-allele result, and this entry must not be read as saying that losing one NUP98 allele causes aneuploidy on its own. Mouse, not human. Whether the RAE1-binding GLEBS motif is retained in NUP98 fusions is a residue question answered in protein-nup98-domains (GLEBS at 157-213, below every annotated fusion breakpoint), so the fusion protein plausibly retains a RAE1-binding surface - a prediction this beat did NOT find tested anywhere.

    Evidence that argues against this (1)
    • The genotype is a Rae1/Nup98 DOUBLE heterozygote, so the securin phenotype cannot be attributed to NUP98 rather than to RAE1, or to the pair rather than to either. A reader could reasonably hold that RAE1 is the APC/C-facing partner and NUP98 is a stabilising passenger - the paper's own model has them acting as a complex - in which case the 'NUP98 is a mitotic brake' framing this entry uses in bite_back is an over-attribution. No Nup98-only allele is cited here and none was found in this beat.
  8. E8moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Beyond their role at nuclear pore complexes some nucleoporins function in the nucleoplasm; intranuclear Nup98 binds the DExH/D-box helicase DHX9 through its FG/GLFG region in an RNA-facilitated manner, stimulates DHX9 ATPase activity, and Nup98 and DHX9 bind interdependently to similar gene loci and their transcripts.

    Human Nup98 regulates the localization and activity of DExH/D-box helicase DHX9 · Capitanio JS, Montpetit B, Wozniak RW · eLife 2017
    PMID 28221134DOI
    What this evidence cannot show

    Abstract read this session; full text not read. Confidence is moderate rather than strong because the transcriptional consequence rests on a reporter assay - 'suggests Nup98 supports DHX9-stimulated transcription' is the authors' own hedge - and the binding is characterised in vitro plus in cultured cells. The interaction surface is the FG/GLFG region, the exact region every NUP98 fusion retains, so this is an interactor a NUP98 fusion would be predicted to bring with it, and this beat found no test of that prediction.

    Evidence that argues against this (1)
    • DHX9 is an abundant, promiscuous DExH helicase recovered in a great many RNA-dependent interactomes, and this interaction is explicitly RNA-FACILITATED - which is precisely the condition under which a shared-RNA bridge produces a reproducible co-purification that is not a direct functional partnership. The ATPase stimulation is the strongest part of the paper and the transcriptional consequence is a reporter assay the authors hedge. An RNase-resistant direct-binding measurement would settle it and is not cited.
  9. E9moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    GLFG bodies had previously been reported only in cells overexpressing exogenous Nup98 or in a HeLa subline expressing unusually elevated endogenous Nup98; the HEp-2 HeLa subline contains distinct GLFG bodies and expresses additional truncated forms of Nup98 fused to a non-coding region of chromosome 11q22.1, with massive amplification of juxtaposed NUP98 and 11q22.1 loci, and minor co-amplifications were also observed in other HeLa sublines.

    PMID 26685999DOI
    What this evidence cannot show

    Abstract read this session; full text not read. The author list is given as the four names present in the fetched efetch record followed by 'et al.' rather than completed from recall. This is a cell-line cytogenetics study in HeLa derivatives; it does NOT show that GLFG bodies are always artefactual, and it does not address whether a physiological GLFG body exists in a haematopoietic cell. Recorded because the graph should not assert the GLFG body as a normal subnuclear compartment without this qualification.

    Evidence that argues against this (1)
    • This does NOT establish that GLFG bodies are artefactual, and the entry must not be read that way. Griffis et al. produced GLFG bodies by GFP-NUP98 overexpression in cells with no reported NUP98 rearrangement, so overexpression alone is sufficient - which means a rearrangement is not necessary, and the Romana finding is better read as 'high NUP98 levels, however achieved, produce the structure' than as 'the structure requires a rearrangement'. Whether a physiological GLFG body exists at endogenous levels in any cell remains open in both directions. Griffis ER et al. 2002
  10. E10strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    NUP98-HOXA9 expressed in mouse haemopoietic stem cells causes a myeloproliferative disease that progresses to AML, and the NUP98 portion of the fusion protein is responsible for converting the clinically silent pre-leukaemic phase seen with Hoxa9 alone into a chronic, stem-cell-derived myeloproliferative disease.

    NUP98-HOXA9 expression in hemopoietic stem cells induces chronic and acute myeloid leukemias in mice · Kroon E, Thorsteinsdottir U, Mayotte N, Nakamura T, Sauvageau G · EMBO J 2001
    PMID 11157742DOI
    What this evidence cannot show

    Abstract read this session. Retroviral overexpression in mouse bone marrow, so expression level is not physiological and the human disease context is inferred. Included for one reason only: it is the cleanest published statement that the NUP98 MOIETY, not the homeobox partner, supplies a distinct transforming contribution - the premise the cross-partner comparison in ana-q126-nup98-shared-address-cross-partner-scan rests on. Already cited in fusion-nup98-hoxa9; per rule 82 the conclusion drawn there was read before this reuse.

    Evidence that argues against this (1)
    • 'The NUP98 portion is responsible' identifies a requirement, not a specificity, and later work shows the requirement is generic. Replacing the NUP98 FG repeats with an unrelated phase-separating IDR from FUS reproduces the enhanced genome-wide binding and target activation, which means the 2001 result is compatible with the NUP98 moiety supplying any sufficiently condensate-forming disordered region rather than something NUP98-specific. This corpus already carries that result and it materially weakens the 'the shared 5' partner is the shared mechanism' reading built on E10. As cited in protein-nup98-domains and localization-nuclear-pore-nup98-fg-comparator 2021
Connected to