Oncofusion

start / graph / localization-af9-cytoplasmic-pool-and-hsp90-dependent-nuclear-residence

Mechanismmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q75-dot1l-partners

AF9's nuclear residence is chaperone-dependent, the Hsp90 site is the SAME C-terminal region the fusion keeps, and UniProt records no cytoplasmic pool at all

also called AF9 Hsp90 client · MLLT3 cytoplasmic pool · AF9 novobiocin redistribution · Hsp90-Hsp70-p60/Hop and AF9

AF9 is usually described as a nuclear protein, but it needs a chaperone - a protein that escorts other proteins - to stay in the nucleus. Block that chaperone and AF9 moves out into the cytoplasm and comes off the gene it was controlling.

The same thing, for a scientist

**In mouse IMCD3 inner-medullary collecting-duct (kidney) cells, not in leukaemia**, AF9/MLLT3's nuclear residence depends on an Hsp90-Hsp70-p60/Hop chaperone complex: novobiocin (a C-terminal Hsp90 ATPase inhibitor) or Hsp90 siRNA - **but not 17-AAG**, which inhibits the N-terminal ATPase - redistributes AF9 from primarily nuclear to nuclear-plus-cytoplasmic, raises *Scnn1a*/ENaC-alpha mRNA and reduces AF9 occupancy at the ENaC-alpha promoter R2 region [E1][E2]. **UniProt P42568 gives AF9 Nucleus and Chromosome and no cytoplasmic value at all** [E3], which is the class of gap `localization-partner-baseline-compartment-is-often-miscalled` predicts. Three things from the full text are absent from the abstract and change how the result should be read in this project. **The Hsp90-binding region is AF9 residues 475-568 - the C-terminal region the paper itself states is necessary and sufficient for MLL-mediated transformation, i.e. the part RETAINED in KMT2A::MLLT3** [E2]. **The affinity column was run on THP-1 lysate, an MLL-AF9 leukaemia line** [E2]. And **AF9 protein abundance does not change under any of the three interventions**, so this is trafficking and not degradation [E2]. The affinity is weak - Kd approximately 3.5 uM by competitive ELISA [E2].

Go deeper
What the full text adds that the abstract does not have

This result already sits in gene-mllt3 as a field, written from the abstract. The full text was recovered here through route 4 of the unreadable test - Europe PMC fullTextXML returned HTTP 500 and efetch db=pmc returned well-formed XML with no <body>, while the PMC HTML page at PMC2852934 returned 42,046 characters of stripped text 2. Five things only in the body:

  1. The Hsp90-binding fragment is AF9-ct, residues 475-568 - isolated by FLAG affinity column. The paper's own opening states that the C-terminal ~90 residues of AF9 are necessary and sufficient for MLL-mediated leukaemic transformation. So the chaperone contact is on the piece of AF9 the fusion keeps, not the piece it loses.
  2. The starting material was THP-1, an MLL-AF9 leukaemia line. The interaction was discovered in a fusion-bearing leukaemia cell even though every functional experiment afterwards is in kidney epithelium.
  3. Kd ~3.5 uM by competitive ELISA with purified recombinant proteins, ATP-dependent (FLAG-AF9-ct co-precipitates with Hsp90 only in the presence of ATP-gamma-S). Micromolar, which the authors note is comparable to other Hsp90-client complexes.
  4. AF9 abundance is unchanged by novobiocin, 17-AAG or Hsp90 siRNA. Unlike a folding client, AF9 is not destabilised - the authors conclude Hsp90 traffics rather than folds it.
  5. A published non-replication in the paralogue: the authors note that affinity purification of ENL and its associated proteins by Slany and co-workers recovered Hsp70 but described neither Hsp90 nor Hop.
A subnuclear observation that a bounded query cannot reach - and a correction owed to gene-mllt3

gene-mllt3's subnuclear_localization field opens with 'NO MORPHOLOGICAL COMPARTMENT HAS BEEN ASSIGNED TO AF9', bounded by a PubMed query over MLLT3/AF9 crossed with subnuclear, nucleolus, speckle and nuclear-body terms. The body of this very paper says otherwise, in two places 2:

'in control-treated cells AF9 is localized primarily to the nucleus within discrete subnuclear speckles'

'Nuclear AF9 is shifted into DAPI-staining pockets or, more dramatically, is delocalized from the nucleus into the cytoplasm'

That is a morphological subnuclear description of endogenous AF9 by confocal immunofluorescence, plus a described relocation on chaperone inhibition. It is unquantified - the figure legend says the experiment was repeated in triplicate and one set is shown, with no scoring - and it is in mouse kidney epithelium, and the speckle observation is attributed to prior work (the AF4-bodies paper). But it exists, and no query over titles and abstracts could have found it, because this paper's abstract never uses the word. It is a textbook instance of brief rule 80: an index that covers one part of a document cannot bound a claim about the whole document.

The correction owed, stated to the standard the brief requires: gene-mllt3, field data.subnuclear_localization and evidence E_q72_mllt3_loc2's caveat, which says "'Proper subnuclear localisation' is the authors' phrase and the data behind it are a nuclear-versus-cytoplasmic partition plus a promoter ChIP, NOT a subnuclear compartment assignment." That caveat is wrong about the paper's contents. There is also a confocal panel showing discrete subnuclear speckles and their loss into DAPI pockets. It should say: no QUANTIFIED morphological compartment assignment for AF9 in any haematopoietic or KMT2A-rearranged cell; an unquantified confocal description of discrete subnuclear speckles and their chaperone-dependent loss exists in mouse kidney epithelium (PMID 20159978, full text read via PMC2852934). The [EXPLICIT NEGATIVE] sentence restricted to haematopoietic cells is unaffected and remains correct. Source: PMID 20159978, full text read this session through the PMC HTML route, not the abstract.

Why this matters for a KMT2A::MLLT3 fusion, and the reason to be careful

The reason to take it seriously. gene-mllt3's existing counter-note argues the result is hard to transfer to the fusion because AF9's only annotated NLS is at residues 295-300 and t(9;11) breaks fall downstream of exon 4, so the NLS is lost from the chimera. The Hsp90 contact does not have that problem: 475-568 is inside the retained AHD region. If Hsp90 traffics full-length AF9 by binding 475-568, it can in principle bind the fusion at the same site. Whether it then traffics anything is a different question - the fusion's nuclear address is supplied by KMT2A, and localization-kmt2a-mllt3-the-address-is-a-kmt2a-property carries that reasoning.

The reasons to be careful, and they are substantial.

  • The functional system is mouse kidney collecting-duct epithelium and the target gene is an epithelial sodium channel. Nothing in this paper is haematopoietic (brief rule 73).
  • The 17-AAG negative is unexplained by the authors themselves - two inhibitors of the same chaperone give opposite answers, and the siRNA arm is what carries the result.
  • The imaging is unquantified and the ChIP is semi-quantitative PCR with ImageJ densitometry.
  • Kd ~3.5 uM is weak, and the AHD is a promiscuous surface: AF9 475-568 is also the binding site for DOT1L, AF4/AFF1 and BCOR, which compete for it. An Hsp90 contact at the same region is another competitor for a surface this graph already models as contested.
  • Nobody has tested any of it in a KMT2A-rearranged cell. (MLL OR KMT2A)[tiab] AND (HSP90 OR "heat shock protein 90")[tiab] returns 15 records, of which the haematology ones are FLT3-Hsp90 papers and none concerns AF9 4.
Facts and figures
proteinMLLT3 / AF9
uniprotP42568
systemmouse IMCD3 inner-medullary collecting-duct cells for all functional experiments; HeLa for the full-length co-IP; THP-1 (MLL-AF9 leukaemia) lysate for the original affinity column only
speciesmouse and human, tissue is kidney epithelium - NOT haematopoietic
localization shown for fusionNO for KMT2A::MLLT3 and for every other fusion. The Hsp90-binding region 475-568 IS retained in the fusion, and nobody has tested whether Hsp90 binds or traffics a KMT2A::MLLT3 chimera; the original affinity column was nevertheless run on THP-1, an MLL-AF9 line.
chaperone complex
[
  "HSP90AA1 / HSP90AB1 (both isoforms bind)",
  "HSP70",
  "p60/Hop (STIP1)"
]
hsp90 binding region AA
{
  "start": 475,
  "end": 568,
  "note": "AF9-ct; the paper states the C-terminal ~90 residues of AF9 are necessary and sufficient for MLL-mediated transformation, so this region is RETAINED in KMT2A::MLLT3"
}
affinity
{
  "kd_uM": 3.5,
  "method": "competitive ELISA, purified recombinant FLAG-AF9-ct and Hsp90beta, three independent experiments",
  "atp_dependent": true
}
interventions
[
  {
    "agent": "novobiocin",
    "target": "Hsp90 C-terminal ATPase",
    "af9_redistribution": "nuclear to nuclear+cytoplasmic",
    "enac_alpha_mrna": "increased, dose-dependent",
    "af9_chip_at_enac_r2": "decreased",
    "af9_abundance": "unchanged"
  },
  {
    "agent": "17-AAG",
    "target": "Hsp90 N-terminal ATPase",
    "af9_redistribution": "no observable change",
    "enac_alpha_mrna": "no effect",
    "af9_chip_at_enac_r2": "unchanged",
    "af9_abundance": "unchanged"
  },
  {
    "agent": "Hsp90 siRNA (both isoforms)",
    "target": "Hsp90 protein level",
    "af9_redistribution": "mimics novobiocin",
    "enac_alpha_mrna": "increased",
    "af9_chip_at_enac_r2": "decreased",
    "af9_abundance": "unchanged"
  }
]
localization
{
  "subcellular": "NUCLEUS AND CYTOPLASM. The cytoplasmic pool is chaperone-conditional: it appears when Hsp90 function is disrupted by novobiocin or siRNA. UniProt P42568 records Nucleus and Chromosome and no cytoplasmic value.",
  "subnuclear": "In mouse IMCD3 kidney cells, endogenous AF9 by confocal immunofluorescence is nuclear within DISCRETE SUBNUCLEAR SPECKLES, and on Hsp90 disruption shifts into DAPI-staining pockets or out to the cytoplasm. UNQUANTIFIED - triplicate with one set shown and no scoring - and in kidney epithelium, not blood. This observation is in the paper's BODY and not in its abstract, so no title/abstract query can reach it (brief rule 80)."
}
correction owed
{
  "entry": "gene-mllt3",
  "fields": [
    "data.subnuclear_localization (opening sentence)",
    "evidence E_q72_mllt3_loc2 caveats"
  ],
  "current_text": "'NO MORPHOLOGICAL COMPARTMENT HAS BEEN ASSIGNED TO AF9' and \"'Proper subnuclear localisation' is the authors' phrase and the data behind it are a nuclear-versus-cytoplasmic partition plus a promoter ChIP, NOT a subnuclear compartment assignment.\"",
  "should_say": "No QUANTIFIED morphological compartment assignment for AF9 in any haematopoietic or KMT2A-rearranged cell. An unquantified confocal description of discrete subnuclear speckles, and of their loss into DAPI-staining pockets on Hsp90 disruption, does exist - in mouse kidney epithelium.",
  "citation": "PMID 20159978, full text read 2026-09-21 via the PMC HTML route (PMC2852934) after fullTextXML 500 and a body-less efetch; the sentences are in the Results section under 'Hsp90 Affects the Subcellular Localization of AF9'",
  "read": "full text, not abstract",
  "note": "The [EXPLICIT NEGATIVE] sentence restricted to haematopoietic and KMT2A-rearranged cells is unaffected and remains correct. Not edited here - gene-mllt3 belongs to another beat."
}
How this could bite us back
  • This looks like a therapeutic opening - Hsp90 inhibitors are clinical agents and 'evict AF9 from the nucleus' is an attractive sentence. The only inhibitor that works here is NOVOBIOCIN, which is an antibacterial DNA gyrase inhibitor with poor Hsp90 selectivity, and the clinically developed Hsp90 inhibitors are the 17-AAG class, which does nothing in this assay.
  • Hsp90 is required by an enormous number of clients including FLT3, AKT and CDK4, so any Hsp90-directed intervention in leukaemia has effects that would swamp an AF9 trafficking phenotype and make attribution impossible.
  • If the AF9 AHD is a chaperone-recognised disordered segment rather than a specific Hsp90 partner, then every AHD-directed therapeutic proposal in this graph is proposing to drug a surface that a major chaperone system also engages, with unpredictable consequences for the chaperone's other clients.
  • The redistribution is unquantified imaging in one laboratory in one tissue with no replication in sixteen years. Treating it as established would put weight on a figure the authors themselves presented without scoring.
Still open
  1. Does Hsp90 bind a KMT2A::MLLT3 fusion protein? The binding region 475-568 is retained in the chimera and the experiment is a co-IP in a t(9;11) line - it has never been done, and it is the cheapest test of whether any of this transfers.
  2. If Hsp90 competes with DOT1L, AFF1 and BCOR for the AF9 AHD, does chaperone inhibition change AHD partner occupancy rather than localisation? That reframes the whole result as a competition experiment and is measurable by AHD-partner co-IP under novobiocin.
  3. Why does 17-AAG do nothing? The authors say they cannot explain it. A C-terminal-versus-N-terminal split this clean is either a real mechanistic constraint on the AF9 contact or an off-target action of novobiocin, and novobiocin is a DNA gyrase inhibitor with known Hsp90-independent effects.
  4. Is there an AF9 cytoplasmic pool in a haematopoietic cell at all? Every localisation experiment in this paper is kidney epithelium.
Every source behind this entry (4)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    PMID 20159978 is Lin JJ, Hemenway CS, 'Hsp90 directly modulates the spatial distribution of AF9/MLLT3 and affects target gene expression', J Biol Chem 2010, reporting that Hsp90 interacts directly and specifically with AF9 as part of an Hsp90-Hsp70-p60/Hop complex and that novobiocin, but not 17-AAG, or Hsp90 siRNA redistributes AF9 from a primarily nuclear to a cytoplasmic location, increasing ENaC-alpha expression and decreasing AF9 occupancy at the ENaC-alpha promoter.

    PMID 20159978DOI
    What this evidence cannot show

    Title, both authors, journal, year and DOI copied character-by-character from a live NCBI efetch XML record fetched 2026-09-21, reading only PubmedData/ArticleIdList/ArticleId (brief rule 81). A two-author single-laboratory paper from 2010 with no replication in the sixteen years since. RENAL EPITHELIUM, MOUSE for every functional experiment - not leukaemia, not haematopoietic, no KMT2A fusion in any functional arm (brief rule 73).

    Evidence that argues against this (1)
    • The same paper records that affinity purification of the AF9 paralogue ENL and its associated proteins by Slany and co-workers recovered Hsp70 but described neither Hsp90 nor Hop, so the closest paralogue's own interactome does not carry the chaperone this result depends on. Lin JJ et al. 2010
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Read in full text: the Hsp90-binding fragment is AF9 residues 475-568, isolated from THP-1 (an MLL-AF9 leukaemia line) lysate on a FLAG-AF9-ct column; binding is direct and ATP-dependent with a competitive-ELISA Kd of approximately 3.5 micromolar; AF9 protein abundance is unchanged by novobiocin, 17-AAG or Hsp90 siRNA, so the effect is trafficking rather than destabilisation; in control IMCD3 cells AF9 is nuclear within discrete subnuclear speckles and under novobiocin or Hsp90 siRNA it shifts into DAPI-staining pockets or out to the cytoplasm; and Hsp90 as well as AF9 loses occupancy at the R2 region of the ENaC-alpha promoter.

    PMID 20159978DOI
    What this evidence cannot show

    Full text recovered 2026-09-21 through route 4 of the unreadable test - Europe PMC fullTextXML returned HTTP 500 and efetch db=pmc returned XML with no <body>; the PMC HTML page for PMC2852934 returned 42,046 characters of stripped text (tags stripped before entities unescaped, brief rule 61). Specific weaknesses stated by the paper itself: the Coomassie gel in Fig. 1 is 'the original and only replicate'; the immunofluorescence in Fig. 5B is triplicate with one set shown and NO QUANTIFICATION; the ChIP in Fig. 7 is semi-quantitative PCR quantified by ImageJ densitometry. The 17-AAG negative is explicitly unexplained by the authors. Kd ~3.5 uM is a weak, micromolar affinity measured on a 94-residue fragment, not on full-length AF9.

    Evidence that argues against this (1)
    • AF9 475-568 is the AHD, the same surface DOT1L, AFF1/AF4 and BCOR bind competitively. A micromolar Hsp90 contact at a promiscuous, intrinsically disordered surface - the paper's own in silico analysis calls the region highly unstructured - is exactly the profile of a chaperone recognising an unfolded segment, which is what chaperones do to any exposed disordered region and need not be a specific functional partnership. Lin JJ et al. 2010
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.analysis

    UniProtKB P42568 annotates AF9/MLLT3 with Nucleus and Chromosome and carries no cytoplasmic value, so the cytoplasmic pool reported in PMID 20159978 is absent from the reference database entirely - the class of gap this graph records in localization-partner-baseline-compartment-is-often-miscalled.

    What this evidence cannot show

    The UniProt content is quoted from this graph's own gene-mllt3 entry, which fetched P42568 on 2026-09-21 and recorded the location comments verbatim; I did NOT re-fetch P42568 in this session, so this is a second-hand database reading and is marked moderate-to-strong rather than established for that reason. Brief rule 91 applies to the codes gene-mllt3 recorded (ECO:0000255 PROSITE rule plus ECO:0000269 on PMID 27105114 for Nucleus): a code means a curator read an experiment, not that the paper is about AF9's compartment.

    Evidence that argues against this (1)
    • A database omission is not evidence that the omitted pool exists - it is equally consistent with UniProt's curators having read PMID 20159978 and judged an unquantified immunofluorescence in mouse kidney insufficient to annotate a human protein's compartment, which would be a defensible call rather than a gap. Lin JJ et al. 2010
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    Counted absence, PubMed E-utilities esearch run 2026-09-21: (MLL OR KMT2A)[tiab] AND (HSP90 OR "heat shock protein 90")[tiab] returns 15 records, of which the haematological ones concern FLT3-Hsp90 synergy, 17-AAG with etoposide, and an MLL1-HSF1 coactivator result; none tests the AF9-Hsp90 axis, and none is in a KMT2A-rearranged model of AF9 localisation.

    What this evidence cannot show

    Fifteen titles were read; no abstract was read, so a paper doing this in a supporting figure would be missed (brief rule 80). The query names HSP90 and its spelled-out form only - HSP90AA1, HSP90AB1, geldanamycin, 17-AAG, novobiocin and 'heat shock protein 90 inhibitor' were not run, and brief rule 72 says a drug-class absence is bounded by its name list, so this is a weaker bound than the record count suggests.

    Evidence that argues against this (1)
    • Hsp90 inhibitors have been tested extensively in AML including KMT2A-rearranged lines, and those papers are framed by drug name (17-AAG, ganetespib, geldanamycin) rather than by HSP90 as a gene symbol - so the true literature reachable by a fuller name list is certainly larger than 15 and this absence should not be quoted as 'Hsp90 is unstudied in this disease'. NCBI PubMed (queries run 2026-09-21) 2026
Connected to