Oncofusion

start / graph / localization-kmt2a-mllt3-the-address-is-a-kmt2a-property

Mechanismmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q29-mllt3-yeats-loss

Where KMT2A::MLLT3 sits in the nucleus: two papers, both overexpression, both from before 2005 - and the fusion carries no AF9 import signal

also called MLL-AF9 subnuclear localisation · MLL-AF9 nucleolus

Only two published studies have ever looked at where this fusion protein sits inside the nucleus, and both pushed extra copies of it into cells that do not have the disease. Both found it goes where the KMT2A half tells it to go.

The same thing, for a scientist

The whole published subnuclear-localisation literature for KMT2A::MLLT3 is **two papers, from 2000 and 2004, both overexpression** - a bounded absence measured this session with the exact queries recorded in `data`. Conditionally expressed MLL-AF9 localises to nucleoplasm and nucleolus in a punctate pattern indistinguishable from an amino-terminal MLL fragment and from endogenous MLL, in U937 and HeLa, which the authors read as the MLL amino terminus conferring a common subnuclear target on MLL chimeras [E1]. The same paper places MLL and the N-terminal fragment in the interphase nuclear matrix and on mitotic chromosomal scaffolds colocalised with topoisomerase II [E1]. The one partner-side observation is from the other direction: AF4 and AF9 interact at discrete non-nucleolar nuclear foci the authors call **'AF4 bodies'**, and expression of the MLL-AF4 fusion protein **alters the subnuclear localisation of AF9** [E2]. So a KMT2A fusion can redistribute AF9 in trans. A sequence fact makes the KMT2A-dominant reading more interesting: **the only nuclear localization signal annotated on AF9 is at residues 295-300 and is therefore discarded by 98.4% of t(9;11) breakpoints** [E3][E4]. That annotation is ECO:0000255, a sequence-analysis prediction with no experimental support, so the honest statement is that **the fusion carries no predicted AF9 import signal and its nuclear address has only ever been attributed to the KMT2A half** - consistent, unmeasured, and cheap to test.

Go deeper
What has actually been imaged, and in what
observationsystemyear
MLL-AF9 in nucleoplasm and nucleolus, punctate, like the MLL N-terminal fragment and endogenous MLLconditional expression in U937 and HeLa, immunofluorescence + confocal + immunoelectron microscopy2000 1
MLL and MLL3AT in the interphase nuclear matrix; colocalised with topoisomerase II on mitotic chromosomal scaffoldssame2000 1
AF4 and AF9 interact at discrete non-nucleolar 'AF4 bodies'; MLL-AF4 alters AF9's subnuclear localisationtransfection2004 2

That is the entire list. No published image shows endogenous KMT2A::MLLT3 protein in a t(9;11) cell line or a patient sample, for the same reason as every other KMT2A fusion: there is no junction-specific antibody, and an anti-AF9-C reagent sees wild-type AF9 and an internal-promoter AF9 isoform as well as the fusion (concept-anti-mllt3-antibodies-and-the-third-species-nobody-counts).

U937 is a particularly awkward host for this experiment in hindsight: it carries PICALM-MLLT10, a DOT1L-dependent HOXA-axis fusion, so the 2000 experiment expressed one KMT2A fusion in a cell already driven by a related one.

The sequence argument, and its evidence code

UniProt P42568 annotates a nuclear localization signal at residues 295-300 - the only import signal on AF9 3. It lies 76 residues N-terminal of the residue-376 breakpoint boundary that about 92% of MLL-AF9 patients break at or after, and well downstream of MLLT3 exon 4, past which 98.4% of breaks fall 4. So on the annotation as it stands, the fusion protein contains no AF9 nuclear import signal.

The evidence code has to travel with that statement. It is ECO:0000255 - inferred from sequence analysis, with no experimental evidence and no cited publication. A PubMed search for an experimental AF9 NLS returns nothing on point: AF9 nuclear localization signal MLLT3 returns 4 records, of which one is about Dot1a's three NLSs in a different organism's kidney epithelium, one is about Diversin, one is an oestrogen/AID gene-proximity paper and one is the AF4-bodies paper 2.

So this is a prediction that has never been tested, on a protein whose nuclear behaviour has been imaged twice, in a fusion whose localisation has never been measured endogenously. Three layers of absence stacked on one another, which is worth stating precisely rather than resolving by assumption.

What this does and does not license

Licensed: the claim that the fusion's nuclear and subnuclear address has only ever been attributed to the KMT2A moiety, in overexpression, and that the partner contributes no annotated import signal.

Not licensed: that the AF9 moiety contributes nothing to localisation. The AF4-bodies result shows a KMT2A fusion can move AF9 in trans 2, and a fusion's residence in chromatin-associated foci is set by what it binds as much as by an import signal. A 2026 report finds that the RNA-binding protein AKAP95 forms condensates that partially co-condense with the translocated MLL1 fragment and that AKAP95 loss impairs MLL-AF9-driven leukemogenesis 5; that is a statement about co-condensation with the KMT2A portion, and it is the only modern condensate observation touching this fusion.

Explicitly out of scope: general subnuclear compartment biology, which another beat owns this cycle. Everything above is specific to MLLT3 or to the MLL-AF9 chimera.

And the condensate caution applies in full: this graph records that 1,6-hexanediol sensitivity and classical FRAP cannot distinguish liquid-liquid phase separation from binding to spatially clustered sites, and that hexanediol itself immobilises and condenses chromatin. A punctate MLL-AF9 pattern in an overexpressing HeLa cell is weak evidence about the endogenous protein in a blast, in either direction.

Facts and figures
fusionfusion-kmt2a-mllt3
endogenous fusion subnuclear localisation measuredfalse
junction specific antibody existsfalse
scope noteDeliberately restricted to MLLT3-specific and MLL-AF9-specific observations. General subnuclear compartment biology is owned by a parallel beat this cycle and is not duplicated here.
bounded negatives
MLLT3 nucleolus0.000MLLT3 nucleolus: 0MLL-AF9 subnuclear1.000MLL-AF9 subnuclear: 1MLL-AF9 immunofluorescenc…1.000MLL-AF9 immunofluorescence nuclear: 1(MLL-AF9 OR KMT2A-MLLT3) …3.000(MLL-AF9 OR KMT2A-MLLT3) AND (nuclear speckle OR nuclear body OR nucleolus): 3AF9 nuclear localization …4.000AF9 nuclear localization signal MLLT3: 4
published observations
[
  {
    "what": "MLL-AF9 in nucleoplasm and nucleolus, punctate, matching the MLL N-terminal fragment and endogenous MLL",
    "system": "conditional expression in U937 and HeLa",
    "pmid": "11069025",
    "year": 2000,
    "title": "The amino terminus targets the mixed lineage leukemia (MLL) protein to the nucleolus, nuclear matrix and mitotic chromosomal scaffolds.",
    "authors": "Caslini C, Alarcòn AS, Hess JL, et al.",
    "venue": "Leukemia",
    "title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
  },
  {
    "what": "AF4 bodies: AF4 and AF9 interact at discrete non-nucleolar foci; MLL-AF4 alters AF9 subnuclear localisation",
    "system": "transfection, tagged proteins",
    "pmid": "14603337",
    "year": 2004,
    "title": "MLL fusion partners AF4 and AF9 interact at subnuclear foci.",
    "authors": "Erfurth F, Hemenway CS, de Erkenez AC, et al.",
    "venue": "Leukemia",
    "title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
  }
]
af9 nls
{
  "start": 295,
  "end": 300,
  "uniprot": "P42568",
  "evidence_code": "ECO:0000255 (sequence analysis prediction)",
  "experimentally_tested": false,
  "retained_in_fusion": false
}
bounded negatives — the numbers behind the chart
[
  {
    "query": "MLL-AF9 subnuclear",
    "count": 1,
    "hit": "14603337",
    "date": "2026-09-20"
  },
  {
    "query": "MLL-AF9 immunofluorescence nuclear",
    "count": 1,
    "hit": "11069025",
    "date": "2026-09-20"
  },
  {
    "query": "(MLL-AF9 OR KMT2A-MLLT3) AND (nuclear speckle OR nuclear body OR nucleolus)",
    "count": 3,
    "hits": [
      "14603337",
      "11681416",
      "11069025"
    ],
    "note": "11681416 is an antisense knockdown paper, not a localisation study",
    "date": "2026-09-20"
  },
  {
    "query": "MLLT3 nucleolus",
    "count": 0,
    "warninglist": "No items found",
    "date": "2026-09-20"
  },
  {
    "query": "AF9 nuclear localization signal MLLT3",
    "count": 4,
    "note": "none on point; hits are Dot1a NLSs in kidney epithelium, Diversin, an oestrogen/AID gene-proximity paper, and the AF4-bodies paper",
    "date": "2026-09-20"
  }
]
How this could bite us back
  • If the fusion's nuclear address is entirely a KMT2A property, then localisation is not a partner-selective handle and any strategy built on mislocalising the partner moiety fails for this fusion before it starts.
  • Two overexpression papers from 2000 and 2004 are a thin foundation for any therapeutic argument about where this protein acts, and a reader should treat the nucleolar localisation in particular as unreplicated in a leukaemic cell.
Still open
  1. Is the predicted AF9 NLS at 295-300 functional? A single GFP-fusion import assay would answer a question that has been annotated but never asked.
  2. Does an endogenously tagged KMT2A::MLLT3 allele in THP-1 or MOLM-13 localise like overexpressed MLL-AF9 in HeLa? The knock-in reagents exist in this graph's `concept-epitope-tagged-knock-in-at-the-endogenous-fusion-allele` and the experiment has not been done for this fusion.
  3. Do the two t(9;11) B-ALL lines localise the fusion differently from the nine monocytic ones? Nobody has imaged either.
Every source behind this entry (5)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Conditionally expressed MLL-AF9 and MLL-AF4 chimeric proteins localise to nucleoplasm and nucleolus in the same punctate pattern as a 50 kDa amino-terminal MLL fragment and as endogenous MLL, and that fragment and endogenous MLL are present in interphase nuclear matrices and colocalise with topoisomerase II on mitotic chromosomal scaffolds - indicating a common subnuclear target conferred by the amino terminus of MLL.

    PMID 11069025DOI
    What this evidence cannot show

    Conditional overexpression in U937 and HeLa, read out by immunofluorescence, confocal and immunoelectron microscopy in 2000. Neither cell is a t(9;11) leukaemia, and U937 carries PICALM-MLLT10, a related HOXA-axis fusion. No AF9-only or NLS-mutant comparator is included, so the paper cannot say what the AF9 moiety contributes. Overexpression is the condition under which punctate nuclear patterns are least informative.

    Evidence that argues against this (1)
    • MLL-AF4 expression alters the subnuclear localisation of AF9 in trans, which shows that partner localisation in a KMT2A-rearranged cell is not simply dictated by the MLL amino terminus of the fusion that is present. Erfurth F et al. 2004
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    AF4 and AF9, the two most common MLL fusion partners, interact at discrete non-nucleolar nuclear foci termed AF4 bodies that do not coincide with known nuclear structures; the interaction is maintained by the MLL-AF4 fusion protein, and expression of MLL-AF4 can alter the subnuclear localisation of AF9.

    MLL fusion partners AF4 and AF9 interact at subnuclear foci. · Erfurth F, Hemenway CS, de Erkenez AC, et al. · Leukemia 2004
    PMID 14603337DOI
    What this evidence cannot show

    Transfected, tagged proteins in a heterologous system in 2004; 'AF4 bodies' are defined negatively, by not coinciding with the nuclear structures the authors examined. The experiment concerns MLL-AF4 and wild-type AF9, not the MLL-AF9 fusion, so its relevance here is that a KMT2A fusion can redistribute AF9 rather than that it redistributes itself.

    Evidence that argues against this (1)
    • A punctate pattern in a transfected cell is exactly the observation this graph's condensate rules warn about: overexpressed low-complexity-domain proteins form hubs without detectable phase separation at physiological expression, and the assays used here cannot distinguish a body from clustered binding sites. Yan X et al. 2026
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    UniProt P42568 (AF9_HUMAN, primary gene name MLLT3, 568 aa) annotates exactly one nuclear localization signal, at residues 295-300, with evidence code ECO:0000255 (inferred from sequence analysis) and no cited publication.

    What this evidence cannot show

    A predicted feature, not a measured one. ECO:0000255 means a sequence-analysis inference; no experiment anywhere in this beat's searches tests AF9 nuclear import against any residue. A predicted NLS may be non-functional, and a protein of 63 kDa can also enter the nucleus passively or by piggybacking on a partner.

    Evidence that argues against this (1)
    • AF9 is a 568-residue, 63 kDa protein that binds several obligate nuclear partners, so its nuclear accumulation could be entirely carrier-mediated and the predicted NLS irrelevant; conversely, that same argument means the fusion's nuclear entry would not require an AF9 NLS at all. UniProt 2026
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    AF9 genomic breaks in t(9;11) occur almost exclusively downstream of MLLT3 exon 4 (98.4%, n=442), and in the recombinome archive KMT2A::MLLT3 accounts for 612 of 3,401 genotyped KMT2A-rearranged patients, so the AF9 sequence N-terminal of the break - including residues 295-300 - is absent from the fusion in essentially all of them.

    The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al. · Leukemia 2023
    PMID 37019990DOI
    What this evidence cannot show

    The 98.4% figure is from PMID 36435883 and is recorded with its own caveats in `breakpoint-mllt3-t-9-11-distribution-and-the-two-patients-who-kept-yeats`; the recombinome citation here supports the cohort size and the partner's frequency, read from PMC10169636 on 2026-09-20. Genomic breakpoints, not protein measurements.

    Evidence that argues against this (1)
    • The residue-level boundary (approximately 376) comes from a review's summary rather than from a residue-resolved primary series, so 'residues 295-300 are absent' is an inference from an exon-level distribution plus a reviewed residue estimate, not a direct measurement in any patient. Meyer C et al. 2023
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    AKAP95 forms transcriptional condensates, interacts with the MLL1 translocated fragment with partial co-condensation that strengthens AKAP95 binding at MLL-AF9 target genes, and its loss downregulates MLL-AF9 targets and impairs MLL-AF9-driven leukemogenesis while being dispensable for normal and regenerative haematopoiesis.

    PMID 41501053DOI
    What this evidence cannot show

    2026 report, abstract read; the co-condensation is with the translocated MLL1 fragment - the KMT2A half - not with the AF9 moiety, so it does not bear on what the partner contributes to localisation. Condensate claims in this graph carry a standing caution that the commonest assays cannot distinguish phase separation from binding to clustered sites.

    Evidence that argues against this (1)
    • Because the reported co-condensation is with the KMT2A-derived fragment, this result is partner-agnostic by construction and would be expected to hold for KMT2A::AFF1 and KMT2A::MLLT1 equally; it is therefore not evidence about MLLT3-specific localisation. Yan X et al. 2026
Connected to