Oncofusion

start / graph / localization-flt3-itd-intracellular-retention-er-versus-golgi

Mechanismstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q57-localization2

FLT3-ITD signals from inside the cell, not from the surface - and whether it is the ER or the Golgi decides which pathway fires

also called FLT3-ITD ER retention · FLT3 mislocalization · FLT3 Golgi retention · 130 kDa FLT3 · compartment-dependent STAT5 activation

A healthy FLT3 receptor sits on the cell surface and waits for a signal. The mutant form found in leukemia never gets there: it is held inside the cell, and from there it switches on a different set of growth signals than it would from the surface. Drugs that block the kinase let it escape to the surface again, which means the drug changes where the protein is as well as what it does.

The same thing, for a scientist

FLT3 is the clearest case in this beat of localisation being the mechanism rather than a descriptor. Wild-type FLT3 matures to the plasma membrane; the internal-tandem-duplication mutant is largely retained in the early secretory pathway as an immature underglycosylated species [E1]. Compartment-restricted targeting shows the consequence is a **signalling switch**: ER-localised Flt3-ITD aberrantly activates STAT5 and its targets PIM1/2 but fails to activate PI3K and MAPK, while membrane-targeted Flt3-ITD strongly activates MAPK and PI3K with diminished STAT5 [E2]. The compartment identity is **contested**: endogenous FLT3-ITD in patient-derived AML lines accumulates in a **perinuclear region that co-stains with Golgi markers**, with ER-resident ITD activating STAT5 and Golgi-resident ITD activating AKT and ERK [E3]. ER retention also controls NOX4/p22phox-derived ROS, shown in **MV4-11 - a KMT2A::AFF1 line** [E4]. Mutant NPM1c is sufficient to move FLT3-D835Y from the surface to the ER and enable STAT5 activation [E5]. And clinically the key fact: **quizartinib and midostaurin reduce retention and increase plasma-membrane FLT3-ITD** [E3], so a kinase inhibitor is also a trafficking drug.

Go deeper
The switch, and why it is not a curiosity

Choudhary et al. targeted Flt3-ITD to defined compartments and quantified 12,186 phosphorylation sites 2. The result is categorical rather than quantitative:

where Flt3-ITD isSTAT5PI3K/AKTMAPK
endoplasmic reticulumstrongly activated, PIM1/2 upnot activatednot activated
plasma membranediminishedstrongly activatedstrongly activated

The same kinase, the same mutation, opposite downstream programmes. They also found spatially regulated tyrosine phosphorylation of the receptor itself - the compartment changes which of its own sites are phosphorylated.

This is the cleanest published demonstration in leukaemia of the principle this whole beat is built on: a protein's location is not an attribute of the protein, it is part of the mechanism, and an entry that records only 'FLT3: membrane' loses the biology.

ER or Golgi - the contradiction, recorded rather than resolved

The dominant account is ER retention: UniProt states it plainly ("constitutively activated mutant forms with internal tandem duplications are less efficiently transported to the cell surface and a significant proportion is retained in an immature form in the endoplasmic reticulum lumen") 1, and the compartment-targeting, NOX/ROS and NPM1c work all use ER as the intracellular site 245.

Yamawaki et al. disagree with the emphasis 3. Working with endogenous FLT3-ITD in leukaemia-patient-derived cell lines rather than with transfected constructs, they find the mutant accumulates in a perinuclear region that co-localises with Golgi markers, state that "the precise subcellular localization of mutant FLT3 remains unclear", and propose a sequential model: FLT3-ITD traffics biosynthetically, signals from the ER and from the Golgi, and the two compartments produce different outputs - ER to STAT5, Golgi to AKT and ERK.

That is not a flat contradiction; it is a partial reconciliation that reassigns the PI3K/MAPK arm from the plasma membrane to the Golgi. But it does conflict with 2's reading, where PI3K/MAPK activation is a plasma-membrane property. What would settle it: compartment-restricted rescue in cells with endogenous FLT3-ITD - an ER-retention mutant, a Golgi-retention mutant and a surface-forced mutant in the same background, scoring STAT5, AKT and ERK. All the constructs exist; the experiment has not been run head-to-head on endogenous protein.

Kinase-dependence is the other unreconciled detail. 3 reports that Golgi retention is dependent on FLT3-ITD tyrosine kinase activity and that TKIs release the receptor to the surface. A retention mechanism that the drug abolishes changes what 'target engagement' means.

The one measurement in a KMT2A-rearranged cell, and the lineage trap underneath it

4 is done in MV4-11, which carries t(4;11) KMT2A::AFF1 as well as FLT3-ITD. Tunicamycin and brefeldin A force ER retention of FLT3-ITD, and the consequence is loss of NOX4 and p22phox protein, NOX4 deglycosylation, and loss of the pro-survival ROS; PI3K/AKT signalling occurred only when FLT3-ITD was at the plasma membrane. So there is at least one KMT2A-rearranged system in which FLT3 compartment has been manipulated and read out.

And now the trap, which is a rule-73 problem in the opposite direction. This entire literature is about FLT3-ITD, an AML lesion. Infant and paediatric KMT2A-rearranged B-ALL characteristically overexpresses wild-type FLT3 and carries activating point mutations far more often than ITD; MV4-11 is a KMT2A::AFF1 line but it is a monocytic AML-type line with an ITD, which is not the common infant-ALL configuration. A compartment result obtained on ITD does not transfer to overexpressed wild-type FLT3, because the whole phenomenon is caused by the ITD's failure to mature - and overexpressed wild-type receptor, by 1's own logic, goes to the surface. Any FLT3-inhibitor rationale in infant KMT2A::AFF1 ALL therefore inherits none of this localisation mechanism, and should not cite it.

A second lesion can relocalise the receptor

Rudorf et al. show that NPM1c - the cytoplasm-mislocalised nucleophosmin mutant that defines a large AML subtype - moves FLT3-D835Y from the cell surface to the ER, and that FLT3-TKD activates STAT5 exclusively in the presence of NPM1c, in primary murine cells and in patients with both mutations 5. This is a co-mutation acting through compartment rather than through pathway, and it is the mechanistic explanation offered for why NPM1 status has prognostic relevance specifically for FLT3-TKD patients.

It is worth stating what this implies generally: co-occurring lesions can change where a target is. This graph holds RAS/FLT3 co-mutation entries; none of them records a localisation consequence, and on this evidence at least one class of co-mutation has one.

Facts and figures
genegene-flt3
uniprotP36888
length AA993
compartment subcellularwild-type: plasma membrane, single-pass type I, fully glycosylated. FLT3-ITD: retained intracellularly in the early secretory pathway as an immature underglycosylated species - ER per UniProt and the construct literature, Golgi-marker-positive perinuclear region per the endogenous-protein study.
compartment subnuclearnot applicable - FLT3 is a receptor tyrosine kinase and has no nuclear pool reported in any source read here
contradiction recordedER (UniProt, PMID 19854140, 31186273) versus Golgi (PMID 34811450, endogenous receptor, Golgi-marker co-stain). Yamawaki et al. propose sequential signalling - ER to STAT5, Golgi to AKT/ERK - which partially reconciles them but conflicts with the membrane attribution of PI3K/MAPK in PMID 19854140. Resolving experiment: ER-, Golgi- and surface-restricted FLT3-ITD in one endogenous-ITD background, scoring STAT5, AKT and ERK together.
drug changes localizationquizartinib (AC220) and midostaurin (PKC412) markedly decrease FLT3-ITD retention and increase plasma-membrane levels of the mutant (PMID 34811450)
shown for fusionPARTLY - the NOX/ROS compartment experiment (PMID 27870947) was done in MV4-11, a KMT2A::AFF1-positive line. But the lesion imaged is FLT3-ITD, not the fusion, and no study asks whether a KMT2A fusion itself changes FLT3 localisation.
lineage trapThe whole localisation literature concerns FLT3-ITD, an AML lesion. Infant/paediatric KMT2A-rearranged B-ALL characteristically OVEREXPRESSES wild-type FLT3 or carries activating point mutations rather than ITD, and overexpressed wild-type receptor traffics to the surface. None of this compartment mechanism transfers to that setting.
patient material datatrue
uniprot evidence codes
{
  "Membrane": "no ECO code on the location values; the ITD note is narrative",
  "Endoplasmic reticulum lumen": "no ECO code on the location value"
}
signalling by compartment
{
  "ER": "STAT5, PIM1/2",
  "Golgi": "AKT, ERK (per PMID 34811450)",
  "plasma_membrane": "MAPK and PI3K (per PMID 19854140); NOX4/p22phox-derived ROS (per PMID 27870947)"
}
methods
[
  "compartment-targeted constructs plus global phosphoproteomics (12,186 sites)",
  "co-immunofluorescence with Golgi markers on endogenous receptor",
  "trafficking inhibitors tunicamycin and brefeldin A",
  "surface versus total receptor glycoform analysis",
  "murine co-expression models and patient samples"
]
species and systems
[
  "human AML cell lines including MV4-11 (t(4;11) KMT2A::AFF1 plus FLT3-ITD)",
  "murine bone marrow models",
  "human AML patients with FLT3-TKD plus NPM1c"
]
How this could bite us back
  • A FLT3 kinase inhibitor is also a trafficking drug: it releases FLT3-ITD to the plasma membrane. On withdrawal or at trough concentrations the receptor is then positioned to fire MAPK and PI3K, the pathways the intracellular pool could not reach - a structural reason to expect rebound rather than simple regrowth.
  • Compartment-dependent signalling means a pharmacodynamic assay reading phospho-STAT5 could show full inhibition while a relocated receptor drives AKT and ERK, and the assay would report success.
  • Surface FLT3 is the epitope for FLT3-directed antibodies, CAR-T and bispecifics. A kinase inhibitor that increases surface FLT3 could raise on-target antigen - a rational combination - but also raises it on normal FLT3-expressing progenitors, which is where the toxicity is.
  • Glycosylation inhibitors have been proposed to exploit FLT3-ITD's maturation defect. N-glycosylation is required by a large fraction of the surface proteome, so the window is narrow and the toxicity is not FLT3-specific.
  • Importing this mechanism into infant KMT2A::AFF1 ALL would be a category error: the lesion there is usually not an ITD, and the entire phenomenon depends on the ITD failing to mature.
Still open
  1. ER or Golgi? The two best studies disagree, one using targeted constructs and one using endogenous receptor. A three-way compartment-restriction experiment in an endogenous FLT3-ITD background would settle it and has not been done.
  2. Does a KMT2A fusion change FLT3 localisation? KMT2A-rearranged leukaemias have the highest FLT3 expression of any subtype, and nobody has asked whether the receptor traffics normally in them.
  3. In infant KMT2A::AFF1 B-ALL, where FLT3 is overexpressed wild-type, is any of it intracellular? If it all reaches the surface, the compartment mechanism is irrelevant there and FLT3-inhibitor rationale must rest on kinase activity alone.
  4. If quizartinib and midostaurin push FLT3-ITD to the surface, does drug withdrawal produce a surface-rich, MAPK/PI3K-primed receptor pool - a rebound with a different signalling character than the pre-treatment disease?
  5. Do other co-mutations in this graph's RAS/FLT3 set change FLT3 localisation the way NPM1c does? None has been tested.
Every source behind this entry (5)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt P36888 (human FLT3, 993 aa) annotates Membrane, single-pass type I, and Endoplasmic reticulum lumen, with the note that constitutively activated mutant forms with internal tandem duplications are less efficiently transported to the cell surface and a significant proportion is retained in an immature form in the ER lumen, and that the activated kinase is rapidly targeted for degradation.

    What this evidence cannot show

    A database record. gene_exact:FLT3 AND organism_id:9606 AND reviewed:true returned one reviewed record, gene_primary FLT3, accession P36888, length 993. The ITD note is unusual for UniProt in that it describes a MUTANT's localisation - useful here, but it means the record mixes wild-type and mutant behaviour in one comment, which an automated import would flatten.

    Evidence that argues against this (1)
    • The record names the ER and not the Golgi. Yamawaki et al., working on endogenous rather than transfected receptor, place the bulk of FLT3-ITD in a Golgi-marker-positive perinuclear region and state that the precise localisation remains unclear. The database's single-compartment statement is therefore ahead of the primary literature.
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Flt3-ITD localised at the endoplasmic reticulum aberrantly activates STAT5 and upregulates its targets PIM1/2 but fails to activate PI3K and MAPK signalling, whereas membrane targeting of Flt3-ITD strongly activates MAPK and PI3K with diminished STAT5 phosphorylation; global phosphoproteomics quantifying 12,186 phosphorylation sites confirmed compartment-dependent pathway activation and revealed spatial regulation of the receptor's own tyrosine phosphorylation pattern.

    Mislocalized activation of oncogenic RTKs switches downstream signaling outcomes. · Choudhary C, Olsen JV, Brandts C, Cox J, Reddy PN, Bohmer FD, Gerke V, Schmidt-Arras DE, Berdel WE, Muller-Tidow C, Mann M, Serve H · Mol Cell 2009
    PMID 19854140DOI
    What this evidence cannot show

    RULE 73: engineered compartment-targeted Flt3-ITD constructs in cell lines, and Flt3-ITD is an AML lesion - not KMT2A-rearranged infant B-ALL, where FLT3 is typically overexpressed wild-type or point-mutated rather than ITD. Forced targeting establishes what the kinase does in a compartment, not what fraction of endogenous receptor is in each compartment. Read: abstract only (PubMed efetch, 2026-09-20).

    Evidence that argues against this (1)
    • Yamawaki et al. reassign PI3K/MAPK activation to the GOLGI pool of endogenous FLT3-ITD rather than to the plasma membrane. If that is right, the membrane-targeting arm of this experiment may have been reporting what an artificially surface-tethered kinase does rather than what the endogenous surface pool does, and the clean two-compartment picture becomes a three-compartment one.
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In leukaemia-patient-derived cell lines, endogenous FLT3-ITD but not wild-type FLT3 accumulates in a perinuclear region that co-localises with Golgi markers; retention there is dependent on FLT3-ITD tyrosine kinase activity, quizartinib (AC220) and midostaurin (PKC412) markedly decrease retention and increase plasma-membrane levels of the mutant, and FLT3-ITD in the ER activates STAT5 while FLT3-ITD in the Golgi can activate AKT and ERK.

    FLT3-ITD transduces autonomous growth signals during its biosynthetic trafficking in acute myelogenous leukemia cells. · Yamawaki K, Shiina I, Murata T, Tateyama S, Maekawa Y, Niwa M, Shimonaka M, Okamoto K, Suzuki T, Nishida T, Abe R, Obata Y · Sci Rep 2021
    PMID 34811450DOI
    What this evidence cannot show

    RULE 73: AML cell lines carrying FLT3-ITD; no KMT2A-rearranged B-lineage system, no primary patient material. Compartment assignment rests on co-immunofluorescence with Golgi markers at confocal resolution, and ER and Golgi are adjacent perinuclear structures. The pathway attributions come from trafficking-inhibitor treatments (brefeldin A-class reagents), which perturb the whole secretory system and not only the receptor. Read: abstract only (PubMed efetch, 2026-09-20).

    Evidence that argues against this (1)
    • This directly disputes the field's standard ER-only account, including UniProt's. It is graded strong because it uses ENDOGENOUS receptor where the earlier work used targeted constructs, but the trafficking-inhibitor logic is indirect and the ER/Golgi distinction is at the edge of confocal resolution. The two accounts are recorded here side by side; this entry does not pick a winner, and says what would.
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In MV4-11, a human leukaemia line carrying t(4;11) KMT2A::AFF1 together with FLT3-ITD, the trafficking inhibitors tunicamycin and brefeldin A induce ER retention of FLT3-ITD, which decreases NOX4 and p22phox protein, deglycosylates NOX4 and degrades p22phox, abolishing NOX-generated pro-survival reactive oxygen species; PI3K/AKT signalling occurred only when FLT3-ITD was expressed at the plasma membrane.

    PMID 27870947DOI
    What this evidence cannot show

    RULE 73 in a subtle form: MV4-11 IS KMT2A::AFF1-positive, which makes this the only measurement in this entry performed in a KMT2A-rearranged cell - but the lesion being studied is FLT3-ITD, and MV4-11 is a monocytic AML-type line, not the infant B-lineage configuration where KMT2A::AFF1 usually sits and where FLT3 is typically overexpressed wild-type rather than ITD. Retention is forced pharmacologically with tunicamycin (an N-glycosylation inhibitor) and brefeldin A (which collapses the Golgi into the ER); both have large pleiotropic effects and brefeldin A in particular cannot be used to argue cleanly about ER versus Golgi. Read: abstract only (PubMed efetch, 2026-09-20).

    Evidence that argues against this (1)
    • Brefeldin A collapses Golgi into ER, so 'ER retention' produced by brefeldin A cannot distinguish an ER effect from loss of the Golgi pool - which is precisely the compartment Yamawaki et al. say matters. The NOX4/ROS result is robust as a consequence of blocking maturation; its compartment attribution is not.
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Mutant NPM1c alters the cellular localisation of FLT3-D835Y from the cell surface to the endoplasmic reticulum, and FLT3-TKD activates STAT5 exclusively in the presence of NPM1c; co-expression produces an aggressive murine myeloproliferative disease with 31.5-day latency, and aberrant STAT5 activation is also seen in patients with combined FLT3-TKD and NPM1c mutations.

    NPM1c alters FLT3-D835Y localization and signaling in acute myeloid leukemia. · Rudorf A, Muller TA, Klingeberg C, Kreutmair S, Poggio T, Gorantla SP, Ruckert T, Schmitt-Graeff A, Gengenbacher A, Paschka P, Baldus C, Zeiser R · Blood 2019
    PMID 31186273DOI
    What this evidence cannot show

    RULE 73: murine models plus human NPM1c/FLT3-TKD AML patients. NPM1c-mutant AML and KMT2A-rearranged leukaemia are largely mutually exclusive entities, so this is an adjacent-disease mechanism, imported here as a demonstration that a co-mutation can act through compartment. The lesion is FLT3-TKD (D835Y), not ITD, so it is not interchangeable with the rest of this entry. Read: abstract only (PubMed efetch, 2026-09-20).

    Evidence that argues against this (1)
    • The direction of causality is asserted from co-expression rather than from an NPM1c-degron timecourse, so an alternative reading - that NPM1c changes the cell state and the receptor follows - is not excluded. More importantly for this graph: no equivalent experiment asks whether a KMT2A fusion changes FLT3 localisation, although KMT2A-rearranged leukaemias are the archetype of high FLT3 expression.
Connected to