Oncofusion

start / graph / localization-igf2bp3-cytoplasmic-rnp-granules-p-bodies-stress-granules

Mechanismmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q57-localization2

IGF2BP3 works in cytoplasmic RNA granules - the only compartment in this beat where a drug is already pointed, and nobody has looked at it in a leukaemia cell

also called IGF2BP3 localization · IMP3 granules · KOC · P-body stress granule IGF2BP3 · oncofetal RNA-binding protein compartment

IGF2BP3 is an RNA-binding protein switched on by KMT2A fusions, and it binds the fusion's own messenger RNA. Its job happens in small droplets of RNA and protein in the cytoplasm, including the ones cells build under stress. There is now a drug against it that works in a mouse MLL-Af4 leukemia - and no published picture of where the protein sits in any leukemic cell.

The same thing, for a scientist

UniProt O00425 lists Nucleus and Cytoplasm **with no evidence codes at all**, and adds two experimentally coded cytoplasmic compartments - **P-body** and **stress granule** - plus a long uncoded note placing IGF2BP3 in lamellipodia of the leading edge, the perinuclear region, beneath the plasma membrane, at the connecting piece and tail of spermatozoa, and colocalised with CD44 mRNA in RNP granules [E1]. The coded compartments come from the paper establishing the IGF2BP family as m6A readers that stabilise and store target mRNAs and are recruited to stress granules under oxidative stress [E2]. This matters here for a concrete reason: **IGF2BP3 binds both reciprocal t(4;11) fusion transcripts**, reducing KMT2A::AFF1 protein abundance while increasing AFF1::KMT2A [E3], and a small-molecule IGF2BP3 inhibitor (I3IN-002) that disrupts protein-RNA binding shows anti-leukaemic activity in a syngeneic MLL-Af4 model [E4]. **The compartment in which all of that happens has never been imaged in a leukaemic cell** [E5].

Go deeper
What is coded and what is not

The UniProt record, fetched 2026-09-20, is an unusually clean example of the evidence-code point this beat keeps making:

locationevidence
Nucleusnone
Cytoplasmnone
Cytoplasm, P-bodyECO:0000269, PubMed 29476152
Cytoplasm, Stress granuleECO:0000269, PubMed 29476152
note: lamellipodia, perinuclear, beneath plasma membrane, spermatozoa tail, CD44 mRNA RNP granulesnone

So the two specific compartments are experimentally supported and the two general ones are not. That inverts the usual expectation and means a graph copying 'nucleus and cytoplasm' would import the uncoded half and drop the coded half.

The note is also worth reading for what it says about cell-type dependence: "the subcytoplasmic localization is cell specific and regulated by cell contact and growth." A compartment that moves with cell state is a compartment that has to be measured in the cell type of interest, which for this graph is a blast.

Why the compartment is intervention-relevant here

Three results make IGF2BP3 the one gene in this beat where localisation touches an intervention directly.

  1. IGF2BP3 binds the fusion transcripts themselves. Both reciprocal t(4;11) transcripts - KMT2A::AFF1 (MA4) and AFF1::KMT2A (A4M) - are bound by IGF2BP3, and the consequence is asymmetric: reduced MA4 protein and increased A4M 3. An RNA-binding protein that sets the ratio of the two reciprocal fusion proteins is acting at the level of mRNA fate, which happens in cytoplasmic RNP compartments.
  2. MA4 induces RNA-binding protein genes as a counter-regulatory response 3. So the fusion and its regulator form a loop, and the loop runs through the cytoplasm.
  3. There is a drug. I3IN-002 inhibits the IGF2BP3-RNA interaction, is tolerated in mice, and has anti-leukaemic activity in a syngeneic MLL-Af4 transplantation model, with thermal-shift and DARTS evidence of on-target engagement 4.

A compound whose mechanism is disruption of protein-RNA binding has a compartment-dependent pharmacodynamic question: does it act on the diffuse cytoplasmic pool, on P-bodies, on stress granules, or on all three - and does it change granule number or composition? None of this is known, and in-situ binding of IGF2BP3 to target mRNA was measured in 4 without a compartment readout.

The stress-granule angle is not decorative

IGF2BP3 is recruited to stress granules under oxidative stress 12, and its family function is to stabilise and store target mRNAs including MYC. Storage is a compartment-level idea: an mRNA in a stress granule is protected and translationally silent.

That sets up a specific resistance hypothesis this graph does not contain. Chemotherapy and oxidative stress induce stress granules. If IGF2BP3 sequesters fusion-target or fusion-encoding transcripts into granules under treatment, those transcripts survive the insult and are released afterwards - a compartment-based mechanism of transcript persistence that would look like transcriptional rebound and be invisible to every assay currently used. It is testable in a week with granule imaging plus single-molecule FISH in a treated t(4;11) line.

Facts and figures
genegene-igf2bp3
uniprotO00425
length AA579
compartment subcellularcytoplasm, including P-bodies and stress granules (both experimentally coded); nucleus and general cytoplasm are annotated with NO evidence code; uncoded note adds lamellipodia of the leading edge, perinuclear region, beneath the plasma membrane, spermatozoa connecting piece and tail, and CD44 mRNA RNP granules
compartment subnuclearnot applicable in practice - no nuclear subcompartment is reported, and the nuclear annotation itself carries no evidence code
state dependenceUniProt states the subcytoplasmic localisation is cell specific and regulated by cell contact and growth; recruitment to stress granules is induced by oxidative stress
fusion relevanceIGF2BP3 binds both reciprocal t(4;11) transcripts, reducing KMT2A::AFF1 protein and increasing AFF1::KMT2A (PMID 42537480); a small-molecule inhibitor I3IN-002 disrupting IGF2BP3-RNA binding is anti-leukaemic in a syngeneic MLL-Af4 model (PMID 41307127)
unresolved tensionIGF2BP3 binding LOWERS KMT2A::AFF1 protein abundance, yet an IGF2BP3 inhibitor is anti-leukaemic; no source reconciles these, and the drug's mechanism should not be attributed to the fusion-transcript axis
shown for fusionNO for localisation - IGF2BP3 protein has never been imaged in a KMT2A-rearranged cell, although its interaction with both reciprocal t(4;11) transcripts is measured and a drug against it is in preclinical development
paralogue warningIGF2BP1, IGF2BP2 and IGF2BP3 share domain architecture and overlapping localisation; many reagents and published observations do not discriminate between them (rule 74)
patient material datafalse
bounded absence date2026-09-20
uniprot evidence codes
{
  "Nucleus": "NONE",
  "Cytoplasm": "NONE",
  "Cytoplasm, P-body": "ECO:0000269|PubMed:29476152",
  "Cytoplasm, Stress granule": "ECO:0000269|PubMed:29476152",
  "note (lamellipodia, perinuclear, plasma membrane, spermatozoa, CD44 RNP granules)": "NONE"
}
methods
[
  "immunofluorescence granule assignment (in the m6A-reader paper's figures, not read here)",
  "m6A-dependent binding and stability assays",
  "in-situ protein-RNA binding assay",
  "thermal shift and DARTS target engagement",
  "syngeneic transplantation model"
]
species and systems
[
  "human cell lines (solid tumour and leukaemia)",
  "engineered t(4;11) model systems",
  "syngeneic murine MLL-Af4 transplantation"
]
How this could bite us back
  • IGF2BP3 stabilises and STORES transcripts in granules. An inhibitor that releases stored mRNA could transiently increase translation of the very targets it is meant to suppress, and a short-timescale readout would be needed to see it.
  • If IGF2BP3 binding lowers KMT2A::AFF1 protein, an IGF2BP3 inhibitor could raise driver-fusion protein levels while still killing cells through other targets - a mechanism mismatch that would mislead biomarker selection in a first-in-human study.
  • The KH domains targeted by I3IN-002 are shared with IGF2BP1 and IGF2BP2; IGF2BP1 is required in normal development and IGF2BP2 in metabolism, so paralogue cross-reactivity is a real toxicity route and is not addressed in the published abstract.
  • Stress granules form under chemotherapy. A combination of an IGF2BP3 inhibitor with cytotoxic therapy engages a target whose localisation the cytotoxic agent is simultaneously changing, so scheduling is not neutral.
Still open
  1. Where is IGF2BP3 in a t(4;11) blast - diffuse, in P-bodies, in stress granules? A drug against this protein is in preclinical development and its compartment of action is unknown.
  2. Does chemotherapy-induced stress granule formation sequester fusion-target transcripts through IGF2BP3, protecting them from an insult and releasing them afterwards? That would be a compartment-based persistence mechanism invisible to current assays, and it is testable with granule imaging plus single-molecule FISH in a treated t(4;11) line.
  3. Why is an IGF2BP3 inhibitor anti-leukaemic if IGF2BP3 binding reduces KMT2A::AFF1 protein abundance? The two published results point opposite ways and nobody has reconciled them.
  4. Is I3IN-002 selective over IGF2BP1 and IGF2BP2? The targeted KH domains are shared, and the abstract does not report paralogue selectivity.
  5. Does IGF2BP3 have a nuclear pool in blasts? The nuclear annotation carries no evidence code anywhere, and a nuclear RNA-binding pool would act at a different step from a cytoplasmic one.
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 O00425 (human IGF2BP3, 579 aa) lists Nucleus and Cytoplasm with no evidence codes, and lists Cytoplasm/P-body and Cytoplasm/Stress granule with ECO:0000269|PubMed:29476152; an uncoded note adds lamellipodia of the leading edge, the perinuclear region, beneath the plasma membrane, the connecting piece and tail of spermatozoa, colocalisation with CD44 mRNA in RNP granules, and states that the subcytoplasmic localisation is cell specific and regulated by cell contact and growth.

    What this evidence cannot show

    A database record. gene_exact:IGF2BP3 AND organism_id:9606 AND reviewed:true returned one reviewed record, gene_primary IGF2BP3, accession O00425, length 579. The evidence-code pattern is the finding: the two specific compartments are coded and the two general ones are not, which is the reverse of what an importer would assume. The note's spermatozoa and lamellipodia observations come from cell types with no relevance to leukaemia and are recorded only to show the breadth of the uncoded material.

    Evidence that argues against this (1)
    • IGF2BP3's paralogues IGF2BP1 and IGF2BP2 have overlapping localisations and highly similar domain architecture, and many reagents and many published observations do not discriminate between the three. Any compartment statement about 'IGF2BP3' from an antibody that has not been validated against the paralogues carries the rule-74 problem, and this record does not say which observations were paralogue-resolved.
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    IGF2BP1/2/3 are a distinct family of m6A readers recognising the consensus GG(m6A)C sequence in thousands of transcripts; in contrast to the decay-promoting YTHDF2, IGF2BPs promote the stability and storage of target mRNAs such as MYC in an m6A-dependent manner under normal and stress conditions, their KH domains are required for m6A recognition and are critical for their oncogenic functions.

    Recognition of RNA N6-methyladenosine by IGF2BP proteins enhances mRNA stability and translation. · Huang H, Weng H, Sun W, Qin X, Shi H, Wu H, Zhao BS, Mesquita A, Liu C, Yuan CL, Hu YC, Huttelmaier S · Nat Cell Biol 2018
    PMID 29476152DOI
    What this evidence cannot show

    The abstract describes the m6A-reader function and 'storage' under stress; the specific P-body and stress-granule assignments that UniProt attributes to this paper are in its figures rather than its abstract, and this entry read the abstract only (PubMed efetch, 2026-09-20). Cell-line systems, not leukaemic and not primary. The result concerns the IGF2BP family collectively in several of its claims, so paralogue-specific statements should not be drawn from it without checking which protein was assayed.

    Evidence that argues against this (1)
    • Whether IGF2BPs are bona fide m6A readers has been contested: the KH-domain interaction is of modest specificity and several groups report IGF2BP binding driven by sequence context rather than by the methyl mark itself. That debate does not affect the compartment claim, but it does mean a mechanistic chain running m6A to IGF2BP3 to fusion-transcript fate has a disputed link in the middle.
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In model systems of t(4;11) acute leukaemia, both reciprocal fusion transcripts KMT2A::AFF1 (MA4) and AFF1::KMT2A (A4M) are bound by IGF2BP3, resulting in reduced MA4 protein abundance but increased A4M levels; MA4 alone drives aberrantly high expression of only a few target genes while A4M counteracts this, and MA4 induces genes encoding RNA-binding proteins as a counter-regulatory response; the effect was not seen for the murinised KMT2A::mAff1 fusion, likely because it cannot recruit the SL-1 complex.

    Influence of IGF2BP3 on t(4;11) fusion mRNA and protein abundance. · Hanewald T, Benz T, Marschalek R · Leuk Res 2026
    PMID 42537480DOI
    What this evidence cannot show

    RULE 73: 'model systems' established by the authors - the abstract does not name the cell background, and the comparison arm is a murinised fusion, so this is engineered expression rather than patient-derived t(4;11) disease. A recent single-group result on reciprocal-fusion interplay, an area where this laboratory holds a distinctive position; independent replication was not located. The SL-1 explanation for the murinised-fusion difference is offered as a likelihood by the authors, not demonstrated. Read: abstract only (PubMed efetch, 2026-09-20).

    Evidence that argues against this (1)
    • The direction of the effect is awkward for a simple oncogenic reading: IGF2BP3 REDUCES KMT2A::AFF1 protein while increasing the reciprocal. If IGF2BP3 lowers the abundance of the driver fusion protein, then inhibiting IGF2BP3 should raise it - which is the opposite of what an anti-leukaemic IGF2BP3 inhibitor is assumed to do, and the two results in this entry are not reconciled by any source.
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    A biochemical screen and cell-based counter-screen identified I3IN-002, the most potent reported IGF2BP3 inhibitor, which disrupts in-situ binding of IGF2BP3 to target mRNA, alters IGF2BP3-dependent gene expression, inhibits growth with altered cell cycle and increased apoptosis in multiple leukaemia cell lines, is tolerated intraperitoneally in mice, and shows potent anti-leukaemic activity in a syngeneic transplantation model of MLL-Af4 leukaemia, with cell-free and cellular thermal shift and DARTS assays supporting on-target activity.

    A small molecule inhibitor of RNA-binding protein IGF2BP3 shows anti-leukemic activity. · Jaiswal AK, Scherer GM, Thaxton ML, Sorrentino JP, Yuen C, Mehta MM, Sharma G, Lin TL, Tran TM, Cohen A, Ritter AJ, Kotecha RS · Haematologica 2026
    PMID 41307127DOI
    What this evidence cannot show

    RULE 73: a syngeneic MURINE MLL-Af4 transplantation model plus human leukaemia cell lines; not human t(4;11) infant B-ALL and not patient-derived xenograft. A first-in-class tool compound - 'the most potent reported' is a statement about a sparse field, and no potency figure is given in the abstract. Selectivity against IGF2BP1 and IGF2BP2, which share the KH domains being targeted, is not stated in the abstract. No compartment or localisation readout accompanies the in-situ binding assay. Read: abstract only (PubMed efetch, 2026-09-20).

    Evidence that argues against this (1)
    • This sits in unresolved tension with E3. If IGF2BP3 binding LOWERS KMT2A::AFF1 protein abundance, inhibiting IGF2BP3 should increase the driver fusion protein, yet the inhibitor is anti-leukaemic. Either the anti-leukaemic activity runs through other IGF2BP3 targets (MYC and the broader oncogenic mRNA programme are the obvious candidates), or one of the two results does not generalise. No source reconciles them, and this graph should not assume the IGF2BP3-fusion-transcript axis is the mechanism of the drug.
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    Bounded absence, measured 2026-09-20: no study located reports the subcellular localisation of IGF2BP3 protein - diffuse cytoplasm, P-bodies, stress granules or nucleus - in a KMT2A-rearranged leukaemia cell.

    What this evidence cannot show

    QUERIES: (1) `IGF2BP3[tiab] AND (localization[tiab] OR granule[tiab] OR "stress granule"[tiab] OR cytoplasmic[tiab] OR immunofluorescence[tiab])` = 64 records; the first 15 titles were checked and are dominated by m6A-stabilisation studies in solid tumours. (2) `IGF2BP3[tiab] AND (leukemia[tiab] OR leukaemia[tiab] OR MLL[tiab] OR KMT2A[tiab] OR B-ALL[tiab])` = 36 records; the first 15 titles were checked and none is a localisation study. RULE 67: both shapes return large non-zero counts. NAME LIST: IGF2BP3 only; the aliases IMP3, IMP-3, KOC, KOC1 and VICKZ3 were NOT searched, and IMP3 in particular is a heavily used alias that also collides with the U3 snoRNP protein IMP3 - a rule-70-class collision. This bound is therefore weaker than it looks and is graded moderate. RULE 80 (added mid-beat): this count is bounded by TITLES AND ABSTRACTS, not by papers. Localisation panels are very often supporting figures whose abstracts never mention them, so a PubMed zero here means 'no paper ANNOUNCES this measurement', not 'no such figure exists'. Supporting figures and supplementary material were NOT searched.

    Evidence that argues against this (1)
    • The unsearched alias IMP3 is the specific weakness: much of the older IGF2BP3 cell-biology literature publishes under IMP3 or KOC, and the collision with the unrelated U3 snoRNP IMP3 means a naive search would return noise rather than nothing. A future pass should run the alias list with title-checking before this absence is relied on.
Connected to