Oncofusion

start / graph / absence-no-annotated-ptm-survives-the-nup98-breakpoint

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.q165-the-retained-acetylation-site

NUP98 fusions retain a 469-residue FG repeat region that carries ZERO annotated modification sites - the cleanest retained-portion PTM absence in the four fusion families checked

also called NUP98 FG repeat PTM absence · NUP98 O-GlcNAc not found

NUP98 fusion proteins keep the first 469 amino acids of NUP98 and throw away the rest. Every chemical modification anyone has ever catalogued on NUP98 is in the part that gets thrown away - the earliest one is at position 524. The retained part is completely blank.

The same thing, for a scientist

The retained NUP98 moiety in NUP98::NSD1, NUP98::KDM5A and NUP98::HOXA9 is approximately aa 1-469, spanning FG repeats 1 (1-156), the GLEBS motif (157-213) and part of FG repeats 2 (214-480). UniProt P52948 annotates 28 modified residues and the earliest is S524. There are no CARBOHYD features. The documented mitotic phosphorylation of NUP98 by CDK1 and NEK6 is entirely C-terminal to the breakpoint. A targeted search for O-GlcNAc site mapping on NUP98, and for any link between NUP98 O-GlcNAc and fusion condensates or leukaemogenesis, returned nothing across four query formulations.

Go deeper
The arithmetic
retained NUP98 moietyaa 1-469 (stated as "NUP98 ΔCT (amino acids 1-469)" in a primary methods section, 2)
UniProt P52948 length1,817 aa
annotated modified residues28
earliest annotated modified residueS524 (phosphoserine)
annotated modifications in aa 1-4690
CARBOHYD (glycosylation) features0

1

Mitotic phosphorylation, which is the one well-known regulated PTM layer on NUP98, is all lost. The CDK1 and NEK6 sites reported by Laurell et al. (Cell 2011) number in the isoform P52948-4 frame; resolved against canonical coordinates they fall at 546 and above, and three of them coincide with UniProt's curated S608, S612 and S623. The earliest is ~77 residues past the breakpoint, and that conclusion is robust to any plausible numbering offset 3.

🔴 The O-GlcNAc hypothesis is not supported by any retrievable paper

FG nucleoporins as a class are well known to be heavily O-GlcNAcylated, so the premise is biologically reasonable and the expectation that a NUP98-fusion O-GlcNAc paper exists is a natural one. It does not exist in Europe PMC. Four independent query formulations were run 4:

queryhitson-topic
NUP98 AND O-GlcNAc890 mapping a site on NUP98
NUP98 AND O-GlcNAcylation AND leukemia270
NUP98 AND OGT AND (phase separation OR condensate)130
nucleoporin AND O-GlcNAc AND GLFG210

The closest primary item is Zhu et al., PNAS 2025 (PMID 40773237), which maps an O-GlcNAc analogue across the nuclear pore by super-resolution and shows O-GlcNAc modulation changes mRNA export efficiency. It names no NUP98 residue, no fusion and no leukaemia.

🟠 Read the absence carefully. UniProt does not systematically curate O-GlcNAc, and O-GlcNAc site mapping is technically hard on FG repeats (low-complexity, few tryptic sites, labile modification). So "no annotated PTM in aa 1-469" is partly a statement about what is easy to measure. It remains true that nobody has published one.

What this contributes to the KMT2A question

Across the four fusion families examined in this beat, NUP98 is the one where the retained portion is genuinely empty. That matters two ways:

🟢 It makes the KMT2A case look better by comparison. KMT2A's retained portion carries 71 MS-evidenced sites and 14 UniProt-curated ones including two CK2 phosphosites with a demonstrated fusion phenotype. That is not a barren region.

🔴 And it is a warning about instruments. The NUP98 retained region is a long, low-complexity FG repeat - exactly the sequence class that bottom-up proteomics covers worst. An empty region may be an unmeasured region. The same caution applies, in weaker form, to any absence this beat has drawn from MS coverage of KMT2A.

Facts and figures
scriptworker-agent literature search plus direct UniProt flat-file parse in this beat, 2026-09-24
statusrun
fusion familyNUP98::NSD1, NUP98::KDM5A, NUP98::HOXA9
uniprotP52948
length AA1817
annotated mod res total28
annotated mod res in retained portion0
earliest annotated mod res524
carbohyd features0
mitotic sites retained0
o glcnac site mapped paper foundfalse
result summaryZero annotated modification sites in the retained NUP98 portion. The cleanest retained-portion PTM absence of the four fusion families examined - and partly an artefact of how badly proteomics covers FG repeats.
absence classliterature-absence-bounded
absence class basisBounded against UniProt P52948, which annotates 28 modified residues of which the earliest is S524, downstream of the approximately aa 1-469 retained moiety; no CARBOHYD features.
retained portion AA
[
  1,
  469
]
retained domains
{
  "FG_repeats_1": [
    1,
    156
  ],
  "GLEBS": [
    157,
    213
  ],
  "FG_repeats_2_partial": [
    214,
    469
  ]
}
mitotic kinases documented
[
  "CDK1",
  "NEK6"
]
measurements
[
  {
    "quantity": "annotated modified residues in the retained NUP98 portion",
    "subject": "P52948 aa 1-469",
    "value": 0,
    "units": "sites",
    "n": 28,
    "universe": "all 28 MOD_RES/CROSSLNK/CARBOHYD features on P52948",
    "null": null,
    "note": "the retained portion is 25.8% of the chain; under a length null 7.2 sites would be expected, so this is a real depletion and not a small-n zero"
  },
  {
    "quantity": "margin between the fusion breakpoint and the earliest annotated PTM",
    "subject": "P52948 aa 469 to S524",
    "value": 55,
    "units": "residues",
    "n": 1,
    "universe": "UniProt P52948 feature table",
    "null": null,
    "note": "so the conclusion survives any breakpoint below about 520"
  }
]
How this could bite us back
  • This is a clean negative and clean negatives invite the inference that NUP98 fusions have no PTM-based vulnerability. The honest reading is that NUP98's retained region is the hardest sequence class in the proteome to measure, and the absence should be treated as provisional.
Still open
  1. Is the retained NUP98 FG region actually unmodified, or merely unmeasurable by bottom-up proteomics? A top-down or middle-down MS approach on the NUP98 moiety, or a chemoenzymatic O-GlcNAc labelling experiment, would distinguish these - and the distinction decides whether this is a fact about biology or about instruments.
  2. If the retained NUP98 portion carries no regulated PTM, what does regulate NUP98-fusion condensate formation? The condensate literature treats the FG repeats as a constitutive multivalent element; a fusion whose defining module has no known regulatory input is an unusual claim and may simply reflect that nobody has looked.
Every source behind this entry (4)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt P52948 (NUP98, 1,817 aa) annotates 28 modified residues, the earliest at S524, and no carbohydrate features; nothing is annotated in aa 1-469.

    What this evidence cannot show

    Parsed directly from the UniProt flat file in this beat, independently of the worker agent that first reported it. Canonical isoform only. UniProt curates O-GlcNAc inconsistently and does not systematically annotate it, so an absence of CARBOHYD features is weak evidence about glycosylation specifically.

    Evidence that argues against this (1)
    • The retained NUP98 region is a long low-complexity FG/GLFG repeat, the sequence class bottom-up proteomics covers worst - few tryptic cleavage sites, highly repetitive peptides that cannot be uniquely assigned. Absence of annotation here is substantially an absence of measurement.
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The NUP98 moiety of NUP98-HOXA9 is aa 1-469.

    PMID 34383740DOI
    What this evidence cannot show

    A single paper's construct boundary, read by a worker agent and not verified in the primary text by this beat. NUP98 fusion breakpoints vary by partner - NUP98::NSD1, NUP98::KDM5A and NUP98::HOXA9 do not all break at the same residue - so 469 is representative rather than universal. The conclusion is insensitive to this: the earliest annotated PTM is at 524, so any breakpoint below ~520 gives the same answer.

    Evidence that argues against this (1)
    • NUP98 fusion breakpoints are NOT the same across partners - NUP98::NSD1, NUP98::KDM5A and NUP98::HOXA9 do not all break at the same NUP98 residue - so 'aa 1-469' is one construct's boundary generalised to a family. The counter-evidence that would matter is a NUP98 fusion whose breakpoint falls past residue 524, which would give that fusion a retained phosphosite and break the entry's headline. No such fusion was searched for here. The claim survives only because the margin is 55 residues; it would not survive a 60-residue error.
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The mitotic CDK1 and NEK6 phosphorylation sites on NUP98 reported by Laurell et al. (Cell 2011) map to canonical positions 546 and above, all C-terminal to the fusion breakpoint.

    Phosphorylation of Nup98 by multiple kinases is crucial for NPC disassembly during mitotic entry · Laurell E, Beck K, Krupina K, Theerthagiri G, Bodenmiller B, Horvath P, Aebersold R, Antonin W, Kutay U · Cell 144:539-50 2011
    PMID 21335236DOI
    What this evidence cannot show

    THE INDIVIDUAL SITE NUMBERS ARE SECOND-HAND. The full text returned HTTP 403 at both Cell and ScienceDirect, so the reported sites (CDK1: T529, T536, S595, S606, T653; NEK6: S591, S822) come from secondary sources. They were reconciled by noting they score 1/7 against canonical P52948 and 7/7 against isoform P52948-4, with a +17 offset above canonical residue 393 - and that three of them then coincide with UniProt's curated S608, S612 and S623. The reconciliation is internally consistent but the raw numbers were never read from the paper.

    Evidence that argues against this (1)
    • If the isoform reconciliation is wrong, the individual site positions are wrong. The conclusion - all mitotic sites are lost - survives any offset up to about 50 residues, so it is robust even if the arithmetic is not.
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    Four Europe PMC query formulations for NUP98 O-GlcNAcylation, including searches pairing it with leukaemia, OGT, phase separation and GLFG repeats, return no paper mapping an O-GlcNAc site on NUP98 or linking NUP98 O-GlcNAc to fusion condensates or leukaemogenesis.

    What this evidence cannot show

    Run by a worker agent; the hit counts and the on-topic screen were not independently re-run by this beat. The four queries were unquoted free-text rather than field-scoped phrases, so they run BROAD - which makes the zero conservative rather than fragile. No PubMed-side control operand was run, so the possibility of an instrument failure is not formally excluded.

    Evidence that argues against this (1)
    • O-GlcNAc on FG nucleoporins is textbook biochemistry established largely before site-mapping was possible, much of it in the wheat-germ-agglutinin era. A modern site-mapped paper may be absent because the question is considered settled at the class level, not because nobody believes it.
Connected to