Oncofusion

start / graph / gene-igf2bp3

Geneunrated confidence 0 of 5: unrated

IGF2BP3 — an RNA-binding protein required for MLL-Af4 leukaemogenesis in mice that CRISPR barely touches in human lines

also called IMP3 · KOC1 · VICKZ3 · insulin-like growth factor 2 mRNA-binding protein 3

IGF2BP3 is a protein that grabs hold of messenger RNA and changes how much of it survives and how it is spliced. Deleting it in mice makes MLL-Af4 leukaemia much less aggressive while barely touching normal blood, which is the shape a good target has.

The same thing, for a scientist

IGF2BP3 is overexpressed in KMT2A-translocated leukaemia and **amplifies** MLL-Af4-mediated leukaemogenesis: Igf2bp3 deletion significantly extends survival and greatly attenuates disease with minimal impact on baseline haematopoiesis, and MLL-Af4 leukaemia-initiating cells require it. Its targets include HOXA-locus and RAS-pathway transcripts, and it acts on both steady-state mRNA level and, unexpectedly, pre-mRNA splicing. **The CRISPR data do not agree**: DepMap Chronos mean −0.137 across 1,208 lines with only 1.6% of lines depleted, and in the KMT2A-rearranged index arm −0.295 lymphoid / −0.184 myeloid against −0.159 in other leukaemia lines. That tension is the entry's subject.

Go deeper
Why this entry exists, and what that says about the beat

IGF2BP3 was not found by reading the interactome; it was found by ana-literature-recall, which pointed the project's measurement outward for the first time and asked which laboratories publish in this field that the graph never cites. Rao DS at UCLA came back with six papers in the window and zero cited anywhere in 775 entries.

The Q5 interactome beat holds over 100 entries and had nothing on any IGF2BP family member. That is the instrument working as intended and it is also a comment on how the beat was built: enumeration from protein-complex membership finds proteins that bind the fusion, and misses a post-transcriptional regulator that never touches it.

The mouse genetics and the human CRISPR data disagree, and both are probably right
readoutresult
Igf2bp3 deletion, MLL-Af4 mousesignificantly increased survival, greatly attenuated disease, minimal impact on baseline haematopoiesis
DepMap Chronos, all 1,208 linesmean −0.137, 1.6% of lines below −0.5
KMT2A-rearranged, B-lymphoid (n=4)−0.295
KMT2A-rearranged, myeloid (n=7)−0.184
other leukaemia lines (n=43)−0.159

Per index line the lymphoid values cluster tightly — −0.330, −0.320, −0.285, −0.244 — with every myeloid line above −0.27. So the direction matches the published biology and the magnitude does not.

Three readings, and the entry does not choose between them. Amplifier, not requirement: the paper's own verb is amplifies, and a gene that makes a leukaemia worse without being necessary for its survival in culture is exactly what a modest Chronos looks like. Established disease versus initiation: the mouse experiment removes Igf2bp3 before and during leukaemogenesis, while CRISPR removes it from a line that has already been selected for growth without it. Species and model: MLL-Af4 in mouse is a chimeric construct, and this graph records that cord-blood and murine systems reproduce parts of this disease and not others.

What would settle it

A degron on endogenous IGF2BP3 in a human KMT2A::AFF1 line, scored on self-renewal in serial replating and on xenograft initiation rather than on bulk proliferation. CRISPR dropout over two weeks measures the wrong thing for an amplifier of leukaemia-initiating-cell function, and this graph has already made the equivalent mistake in the other direction by reading a near-zero Chronos as dispensability.

Two results that arrived a cycle after this entry, and one of them inverts the sign

This entry was written from the mouse genetics and the CRISPR data, and framed the tension between them. The Q5 RNA-layer beat then found the measurement nobody had made here: what IGF2BP3 does to the fusion transcript itself.

It binds both reciprocal t(4;11) transcripts and reduces MLL-AF4 while raising AF4-MLL. So the intuitive story — an RNA-binding protein that promotes the disease by increasing fusion output — is the opposite of the only direct measurement. Whatever IGF2BP3 contributes, it is not raising the oncoprotein.

It sits inside a larger result from a group this graph already names as a collaborator: KHDRBS and IGF2BP family proteins control the oncogenic activity of MLL-AF4 through translational regulation, with KHDRBS repressing the transcript by ribosomal stalling at AU-rich sites. That makes fusion DOSE a regulated variable, which this graph had been treating as binary.

m6A is an RNA mark, not a chromatin mark - and the one bridge to histone methylation runs through SAM

The category statement first, because it is the thing most likely to be got wrong later. IGF2BP3 is a reader, and in this corpus "reader" almost always means a chromatin reader - a YEATS domain, a bromodomain, a PHD finger, a CXXC. IGF2BP3 is not that. IGF2BP1/2/3 are a distinct family of m6A readers that recognise the consensus GG(m6A)C on messenger RNA through their K homology (KH) domains, and - unlike the decay-promoting YTH-domain protein YTHDF2 - they promote the stability and storage of their targets, MYC among them, in an m6A-dependent manner; the KH domains are required both for m6A recognition and for the oncogenic function E9. m6A is an epitranscriptomic modification of mRNA. It is not a histone modification, it is not DNA methylation, and IGF2BP3 carries no chromatin-modifying or chromatin-binding module. This entry's epigenetic answer, on the activity question, is therefore: no histone mark, written, erased or read.

The one bridge that is real, and it is metabolic. IGF2BP3's oncogenic function reshapes metabolism: metabolic profiling shows changes in glycolytic flux and one-carbon metabolism including S-adenosylmethionine (SAM) biosynthesis, and eCLIP plus polyribosome profiling show IGF2BP3 promoting translation of MAT2B, the regulatory subunit of the methionine adenosyltransferase complex that makes SAM; IGF2BP3 then alters the level and pattern of m6A on mRNA, which the authors frame as rewriting the epitranscriptome E10. This entry already carries that paper for its metabolic content. What it does not carry is the consequence for chromatin, because SAM is not an RNA-specific cofactor - it is the universal methyl donor, used by every histone methyltransferase including DOT1L.

And in the index fusion, SAM restriction hits H3K79 specifically. Disrupting methionine/SAM metabolism, by methionine deprivation or by pharmacological inhibition of downstream metabolism, reduces overall cellular methylation potential and induces apoptosis selectively in established MLL-AF4 cell lines and MLL-AF6-expressing patient blasts but not in BCR-ABL-driven K562 cells; global histone methylation dynamics are altered with a profound loss of requisite H3K79 methylation, indicating inhibition of DOT1L function, and H3K27me3 occupancy rises at the DOT1L promoter with reduced DOT1L mRNA and protein E11. That paper was cited in zero entries of this corpus before 2026-09-21, which is notable on its own for a graph built around H3K79 methylation.

The prediction, stated as a prediction. If IGF2BP3 raises SAM through MAT2B, and if the index fusion's H3K79 methylation is unusually sensitive to the SAM pool, then IGF2BP3 loss should lower H3K79me2 at fusion target genes. Nobody has measured a histone mark after any IGF2BP3 perturbation - the SAM paper's readout is m6A, and the Met/SAM paper never mentions IGF2BP3. This is a two-paper chain with no experiment joining it, and it is recorded as inference rather than as a finding. It is also cheap to test, and it would bear directly on this entry's central unresolved tension: the mouse genetics say Igf2bp3 amplifies MLL-Af4 leukaemogenesis while human CRISPR barely moves, and a methyl-donor mechanism is the kind of graded, non-binary contribution that a two-week dropout screen is poorly designed to see.

A note on the direction of the epitranscriptomic claim. The graph should not slide from "IGF2BP3 rewrites the epitranscriptome" to "IGF2BP3 is an epigenetic regulator". The first is supported. The second imports a chromatin meaning the evidence does not carry, and in a project whose central mark is H3K79me2 that slide would be expensive.

Facts and figures
uniprotO00425
length AA579
gene primaryIGF2BP3
length addedverified against live UniProt 2026-09-20; rules 66/68/69 require accession AND length_aa so verify-accessions.py can fail a swap
gene symbolIGF2BP3
gene symbol notepromoted from gene_primary 2026-09-20: verify-accessions.py reads data.gene_symbol, so a gene_primary-only entry was being length-checked or id-inferred rather than symbol-asserted
subnuclear localizationnot applicable in practice - no nuclear subcompartment is reported for IGF2BP3, and the nuclear annotation itself carries no evidence code
shown for fusionNO for localisation - IGF2BP3 has never been imaged in a KMT2A-rearranged cell, although it binds both reciprocal t(4;11) transcripts and a drug against it is in preclinical development
epigenetic readoutAN RNA-MODIFICATION READER, NOT A CHROMATIN READER - AND THE DISTINCTION IS THE POINT. IGF2BP3 reads N6-methyladenosine (m6A) on mRNA through its K HOMOLOGY (KH) domains, recognising the consensus GG(m6A)C, and IGF2BP1/2/3 promote the STABILITY and storage of their m6A-marked targets (MYC among them) rather than their decay, which is the opposite of the YTH-domain reader YTHDF2. m6A is an EPITRANSCRIPTOMIC mark on messenger RNA: it is not a histone modification and not DNA methylation, and IGF2BP3 has no SET, JmjC, HAT, HDAC, bromo, chromo, Tudor, PHD or CXXC module. This matters because 'reader' in this corpus otherwise means a CHROMATIN reader - YEATS, bromodomain, PHD, CXXC - and an entry that wrote 'IGF2BP3 is an m6A reader' next to those would be read as a chromatin protein by anyone skimming. IT WRITES, ERASES AND READS NO HISTONE MARK. ONE REAL BRIDGE TO CHROMATIN, AND IT IS METABOLIC AND UNTESTED: IGF2BP3 redirects glycolytic flux into one-carbon metabolism, promotes translation of MAT2B (the regulatory subunit of the methionine adenosyltransferase complex) by eCLIP and polyribosome profiling, and thereby raises S-ADENOSYLMETHIONINE, the methyl donor shared by m6A writers AND every histone methyltransferase. In the index fusion's own cells that matters: disrupting Met/SAM metabolism in MLL-AF4 cell lines and MLL-AF6 patient blasts - but not in BCR-ABL K562 - reduces methylation potential, kills the cells, and causes a PROFOUND LOSS OF H3K79 METHYLATION with increased H3K27me3 at the DOT1L promoter and reduced DOT1L mRNA and protein (PMID 31717699, cited in ZERO entries of this corpus before 2026-09-21). SO THE PREDICTION IS: IGF2BP3 loss should lower the SAM pool and therefore H3K79me2 at fusion targets. NOBODY HAS MEASURED A HISTONE MARK AFTER IGF2BP3 PERTURBATION - the SAM paper measures m6A, and the chromatin consequence is inference, labelled as such here. CAUSE OR CONSEQUENCE: the m6A-reading is established by domain mapping and consensus-motif recognition; the histone-methylation link is a two-paper chain with no experiment joining it.
localization
{
  "subcellular": "cytoplasmic RNA granules. UniProt O00425 codes only two compartments experimentally - Cytoplasm/P-body and Cytoplasm/Stress granule, both ECO:0000269|PubMed:29476152 - while Nucleus and general Cytoplasm carry NO evidence code, as does the note placing IGF2BP3 in lamellipodia, the perinuclear region, beneath the plasma membrane, in spermatozoa and with CD44 mRNA in RNP granules. UniProt also states the subcytoplasmic localisation is cell specific and regulated by cell contact and growth.",
  "subnuclear": "not applicable in practice - no nuclear subcompartment is reported, and the nuclear annotation itself carries no evidence code",
  "how_shown": "m6A-reader characterisation with granule assignment (PMID 29476152); no localisation readout accompanies the leukaemia work",
  "systems": "human solid-tumour and leukaemia cell lines, engineered t(4;11) model systems, syngeneic murine MLL-Af4 transplantation; no image in any KMT2A-rearranged cell",
  "why_it_matters": "IGF2BP3 binds both reciprocal t(4;11) transcripts and sets their protein ratio - reducing KMT2A::AFF1 and increasing AFF1::KMT2A (PMID 42537480) - and a small-molecule inhibitor of IGF2BP3-RNA binding, I3IN-002, is anti-leukaemic in a syngeneic MLL-Af4 model (PMID 41307127). The compartment in which all of this occurs has never been imaged in a leukaemic cell.",
  "paralogue_warning": "IGF2BP1/2/3 share domain architecture and overlapping localisation; many reagents do not discriminate between them",
  "node": "localization-igf2bp3-cytoplasmic-rnp-granules-p-bodies-stress-granules"
}
cell type
{
  "evidence_class_key": "concept-expressed-required-maintains-three-evidence-classes",
  "expressed_in": {
    "class": 1,
    "summary": "an ONCOFETAL RNA-binding protein: expressed in fetal tissue and in tumours and largely off in normal adult tissue. IGF2BP3 associates with a proliferative phenotype and with prognostic features in B-cell acute lymphoblastic leukaemia (PMID 33805930), and the IGF2BP family is already listed as a marker of cell-state-fetal-program-retention in this graph.",
    "note_on_evidence_class": "the B-ALL association is a CLASS 1 expression-and-correlation result and must not be read as a requirement. This is the gene in the worklist where the expression-for-dependency substitution is most tempting."
  },
  "required_in": {
    "class": 2,
    "kmt2a_rearranged_lines_mean_gene_effect": -0.224,
    "all_1258_lines_mean_gene_effect": -0.137,
    "pct_all_lines_depleted": 1.6,
    "floor_compressed": false,
    "myeloid_kmt2ar_mean": -0.183,
    "blineage_kmt2ar_mean": -0.295,
    "verdict": "no meaningful requirement in any screened cell type: depleted below -0.5 in 1.6% of 1,258 lines and only -0.087 selective for the KMT2A-rearranged arm. The prognostic and expression literature is much larger than the requirement evidence, and the gap between them is the finding.",
    "counter": "a small-molecule IGF2BP3 inhibitor shows anti-leukaemic activity (PMID 41307127, Haematologica 2026), which is a pharmacological requirement claim that the genetic screen does not support. Either the compound has off-target activity or acute inhibition and chronic knockout differ; the graph should not assume which."
  },
  "context": "expression in human B-ALL; requirement measured only in cell lines",
  "graph_links": [
    "cell-type-pro-b",
    "cell-type-fetal-liver-hsc"
  ]
}
cell state
{
  "maintains": {
    "class": 3,
    "states": [
      "cell-state-fetal-program-retention"
    ],
    "claim": "IGF2BP family members are markers of the retained fetal programme in infant leukaemia; a B-cell developmental gene regulatory network including IGF2BP3 is activated in infant AML (PMID 34793513), which is a state claim about an ONTOGENY rather than about a lineage.",
    "pmids": [
      "34793513",
      "33805930"
    ],
    "strength": "marker-level, not mechanism-level: no experiment removes IGF2BP3 and asks whether the fetal programme collapses"
  },
  "bounded_query": {
    "date": "2026-09-20",
    "tool": "NCBI esearch db=pubmed",
    "query": "(IGF2BP3[tiab] OR \"IMP3\"[tiab] OR \"IMP-3\"[tiab] OR \"KOC\"[tiab] OR \"VICKZ3\"[tiab]) AND <cell-state clause> AND <leukaemia clause>",
    "count": 4,
    "on_topic_after_title_check": 1,
    "controls": {
      "gene_clause_alone": 2374,
      "state_clause_alone": 96743
    },
    "reading": "IGF2BP3 has never been shown to MAINTAIN a defined cell state in leukaemia; it has been shown to MARK one. Zero mechanism-level results in titles and abstracts; supporting figures not searched."
  }
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -0.224,
  "chronos_mean_other_leukaemia": -0.1595,
  "difference": -0.0645,
  "pan_essentiality_all_lines": -0.1371,
  "selectivity_rank": 2521,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
  "how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}
How this could bite us back
  • IGF2BP3 has two close paralogues, IGF2BP1 and IGF2BP2, sharing domain architecture and substrate preference. Any inhibitor question begins with paralogue selectivity, and single-gene CRISPR would under-report all three if they are redundant.
  • IMP-family proteins are oncofetal and re-expressed across many cancers, so normal-tissue expression has to be established before an inhibitor is contemplated.
  • The mouse deletion reports minimal impact on baseline haematopoiesis, measured UNSTRESSED - the condition under which this graph's menin work found nothing until competitive transplantation was applied.
  • An amplifier of leukaemia-initiating-cell function is the target class least likely to show benefit on the bulk-response endpoints trials actually measure.
  • A small-molecule IGF2BP3 inhibitor is in development and anti-leukaemic activity is reported. If part of its mechanism runs through the SAM pool rather than through m6A reading, then its pharmacology is methyl-donor depletion - which hits every methyltransferase in the cell, including those in normal haematopoietic and neural tissue, and which in an infant overlaps the folate and methionine dependencies of normal development. That is a categorically different safety profile from a selective RNA-binding-protein blocker, and it would not be detected by any assay that only measures m6A.
Still open
  1. Does IGF2BP3 dependency in human KMT2A::AFF1 lines appear on a self-renewal or xenograft-initiation readout, where a two-week CRISPR dropout would not see it?
  2. Are IGF2BP1 and IGF2BP2 co-dependent in the same lines? If the family is redundant, single-gene CRISPR would under-report all three.
  3. Does IGF2BP3 expression in KMT2A-rearranged patient samples separate infant from non-infant disease, given that its published targets include the HOXA locus and infant t(4;11) is HOXA-low?
  4. The two published directions have to be reconciled: IGF2BP3 amplifies MLL-Af4 leukaemogenesis in mouse and lowers MLL-AF4 abundance in human cells. A dose-response - fusion output against transformation - would tell whether the relationship is non-monotonic.
  5. Does the menin-inhibitor combination effect survive with revumenib rather than MI-503, and at an occupancy matched to what a patient achieves?
  6. Does IGF2BP3 loss lower H3K79me2 at fusion target genes? The chain is: IGF2BP3 promotes MAT2B translation, MAT2B supports SAM production, and SAM restriction in MLL-AF4 cells causes a profound loss of H3K79 methylation. No experiment joins the two ends - the SAM paper measures m6A and the Met/SAM paper never mentions IGF2BP3 - so this is a prediction, and it is one immunoblot plus a ChIP away from an answer. A positive result would also supply the graded, methyl-donor-mediated mechanism that could explain why the mouse genetics call IGF2BP3 an amplifier while a two-week human CRISPR dropout barely moves.
  7. Is the reported H3K27me3 increase at the DOT1L promoter under methionine restriction reconcilable with the H3K27me3 LOSS at fusion targets under direct DOT1L inhibition? The two point in opposite directions in the same disease, one at the enzyme's own promoter and one at its targets, and no source read here addresses both.
  8. Does IGF2BP3 perturbation change m6A and histone methylation to the same extent? If SAM is the shared bottleneck then the two should move together, and if m6A moves alone the epitranscriptomic framing is the right one. This is the experiment that decides whether IGF2BP3 belongs in this project's epigenetic layer at all or only in its RNA layer.
Every source behind this entry (11)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    IGF2BP3 is overexpressed in MLL-translocated leukaemia and strongly amplifies MLL-Af4-mediated leukaemogenesis. Deletion of Igf2bp3 significantly increases survival of mice with MLL-Af4-driven leukaemia and greatly attenuates disease with minimal impact on baseline haematopoiesis; MLL-Af4 leukaemia-initiating cells require Igf2bp3 for their function. IGF2BP3-targeted transcripts include HOXA-locus and RAS-pathway genes, and targeting regulates both steady-state mRNA levels and pre-mRNA splicing.

    The RNA-binding protein IGF2BP3 is critical for MLL-AF4-mediated leukemogenesis · Tran TM, Philipp J, Bassi JS, et al. · Leukemia 2022
    PMID 34321607DOI
    What this evidence cannot show

    Abstract read; full text not read. Murine MLL-Af4 with a chimeric construct, and the deletion is germline or conditional from before disease rather than pharmacological in established leukaemia. 'Amplifies' is the paper's own verb and is weaker than 'requires'.

    Evidence that argues against this (2)
    • The human CRISPR data do not reproduce a strong requirement: DepMap Chronos mean −0.137 across 1,208 lines with 1.6% of lines depleted, and only −0.295 in the four B-lymphoid KMT2A-rearranged lines against −0.159 in other leukaemia lines. The direction agrees with the mouse and the magnitude does not. DepMap / Open Targets (internal analysis) 2026
    • Minimal impact on baseline haematopoiesis is the therapeutically attractive half and is measured in an unstressed mouse. This graph records the equivalent result for menin — preserved steady-state progenitors with a long-term HSC defect revealed only under competitive transplantation — so an unstressed baseline is the condition least likely to find the toxicity that matters. Oncofusion knowledge graph (internal analysis) 2026
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.analysis

    DepMap CRISPR (Chronos) gives IGF2BP3 a mean gene effect of −0.137 across 1,208 lines with 1.6% of lines depleted below −0.5. Within the KMT2A-rearranged index arm resolved from Cellosaurus by DepMap accession, the four B-lymphoid lines average −0.295 (range −0.330 to −0.244) and the seven myeloid lines average −0.184 (range −0.262 to −0.124), against −0.159 across 43 other leukaemia lines.

    What this evidence cannot show

    Eleven index lines, four of them lymphoid, so every figure carries a wide interval. Reported as raw means rather than as a scaled statistic, because this project has measured that comparator-SD scaling ranks narrowness rather than effect size.

    Evidence that argues against this (2)
    • A two-week CRISPR dropout is the wrong assay for an amplifier of leukaemia-initiating-cell function. The lines have already been selected for growth in culture, which selects against dependence on a self-renewal amplifier, so a modest Chronos is compatible with a large effect on disease initiation. Tran TM et al. 2022
    • The lymphoid-versus-myeloid difference is 0.11 across eleven lines split four and seven, which is not a result. It is recorded because the direction matches the published biology, and it should not be quoted as selectivity. Oncofusion knowledge graph (internal analysis) 2026
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    IGF2BP3 associates with a proliferative phenotype and with prognostic features in B-cell acute lymphoblastic leukaemia.

    PMID 33805930DOI
    What this evidence cannot show

    Abstract-level. An association with proliferation and prognosis in B-ALL generally, not stratified by KMT2A rearrangement.

    Evidence that argues against this (1)
    • Association with a proliferative phenotype is close to circular for an RNA-binding protein whose expression tracks biosynthetic demand: fast-growing blasts express more of the machinery that makes them fast. Prognostic association in unstratified B-ALL also cannot separate a KMT2A effect from the known poor prognosis of the subtypes IGF2BP3 is highest in. Oncofusion knowledge graph (internal analysis) 2026
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    IGF2BP3 redirects glycolytic flux toward one-carbon metabolism and RNA methylation, and an m6A/IGF2BP3 axis drives serine biosynthesis supporting AML stemness and a metabolic vulnerability.

    PMID 41015030DOI
    What this evidence cannot show

    Two independent 2025 reports, both abstract-level here, in AML rather than in KMT2A-rearranged B-lymphoid disease. They describe a metabolic mechanism distinct from the post-transcriptional operon the MLL-Af4 paper describes.

    Evidence that argues against this (2)
    • Two mechanisms for one protein in the same disease is a reason for caution rather than for confidence: a post-transcriptional operon governing survival and proliferation, and a metabolic rerouting into one-carbon flux, are different claims about what IGF2BP3 does, and neither paper tests the other's model. Nature Communications 2025
    • Both metabolic reports are in AML stemness. The index disease here is infant KMT2A::AFF1 B-lymphoid leukaemia, and this graph has repeatedly had to strike myeloid mechanisms imported into it. Oncofusion knowledge graph (internal analysis) 2026
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProtKB O00425 (IF2B3_HUMAN), insulin-like growth factor 2 mRNA-binding protein 3, is 579 amino acids; the accession was resolved by gene_exact query with the primary gene name asserted as IGF2BP3.

    What this evidence cannot show

    The primary-symbol assertion matters here: this project has logged six instances of a gene_exact query returning a different protein that lists the queried symbol as a synonym.

    Evidence that argues against this (1)
    • A UniProt header gives accession, symbol and length and nothing about function in this disease. IGF2BP3 also has two close paralogues, IGF2BP1 and IGF2BP2, which share domain architecture and substrate preference, so any inhibitor question begins with paralogue selectivity that the accession does not address. UniProt Consortium 2026
  6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.analysis

    This entry exists because a literature-recall audit, not a complex-membership enumeration, surfaced it: Rao DS returned six papers in a 2020-2026 PubMed window over five stated field queries with zero cited anywhere in the graph's 775 entries, and the Q5 interactome beat held no IGF2BP-family entry.

    What this evidence cannot show

    The audit's recall figures are relative to five queries someone chose. What it establishes here is narrow and checkable: no entry cited this laboratory, and no entry mentioned this gene.

    Evidence that argues against this (1)
    • One recovered gene does not validate the instrument. The same run listed 65 laboratories with three or more papers and nothing cited, and most of them are artefacts of queries broad enough to catch renal cell carcinoma and NSCLC. The list is a reading queue whose precision is unmeasured. Oncofusion knowledge graph (internal analysis) 2026
  7. E7moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    IGF2BP3 binds both reciprocal t(4;11) transcripts and REDUCES MLL-AF4 mRNA and protein abundance while raising AF4-MLL, which is the opposite sign to the assumption that an RNA-binding protein promotes this disease by increasing fusion output.

    PMID 42537480DOI
    What this evidence cannot show

    Brought in by the Q5 RNA-layer beat and verified against a fetched esummary record. The reciprocal AF4-MLL product is expressed in only about half of t(4;11) cases, so a result reported as a ratio between the two has a denominator that is absent in many patients.

    Evidence that argues against this (2)
    • Lowering the fusion and promoting the leukaemia are not contradictory if the relationship is non-monotonic, which the KHDRBS work implies: too little fusion fails to transform and too much is toxic or triggers differentiation, so a regulator that holds dose in a window could do either depending on where the cell starts. The measurement constrains the mechanism and does not settle the direction. Nature Communications 2022
    • This is one laboratory's measurement in unspecified material, arriving against mouse genetics from a different laboratory that scored a different endpoint. Neither has been reproduced by the other, and this graph has repeatedly found that which model was used decides what a result about this fusion means. Oncofusion knowledge graph (internal analysis) 2026
  8. E8moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Targeting IGF2BP3 enhances the antileukaemic effects of menin-MLL inhibition in MLL-AF4 leukaemia, with Igf2bp3 depletion sensitising the leukaemia to menin inhibition and knockdown alone reported to outperform the menin-MLL inhibitor MI-503 alone on survival.

    PMID 38048400DOI
    What this evidence cannot show

    Genetic depletion against a tool compound is not a like-for-like comparison: knockdown is sustained and near-complete, MI-503 gives partial reversible occupancy, and MI-503 is not revumenib. The comparison bounds the tool compound's performance in that model, not menin inhibition as an intervention.

    Evidence that argues against this (2)
    • A knockdown outperforming a tool compound is the expected result of the comparison rather than a finding about the target, and this graph has had to make the same correction for several genetic-versus-pharmacological contrasts. The clinical measurement of menin inhibition is a complete remission rate of roughly a quarter in relapsed KMT2A-rearranged disease, which no mouse comparison bounds. Journal of Clinical Oncology 2025
    • It does challenge SUFFICIENCY, which is the part worth keeping. If removing an RNA-binding protein adds to menin inhibition, then the menin interface is not the only route to the programme in that model — and if RNA-binding proteins set fusion dose, a cell could raise fusion output on an axis no menin-pocket mutation screen monitors. Nature Communications 2022
  9. E9established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    IGF2BP3 IS AN RNA-MODIFICATION READER, NOT A CHROMATIN READER: IGF2BP1/2/3 are a distinct family of m6A readers that target thousands of mRNA transcripts by recognising the consensus GG(m6A)C sequence, and in contrast to the mRNA-decay-promoting YTH-domain protein YTHDF2 they promote the stability and storage of their target mRNAs including MYC in an m6A-dependent manner under normal and stress conditions; the K homology (KH) domains of IGF2BPs are required for m6A recognition and are critical for their oncogenic functions.

    PMID 29476152DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read in this tranche, and this paper is already cited twice in this corpus (gene-igf2bp2, localization-igf2bp3-cytoplasmic-rnp-granules-p-bodies-stress-granules), so it is recorded here specifically to fix the RNA-versus-chromatin category for IGF2BP3, which neither of those entries does. The paper treats IGF2BP1/2/3 as a family; family-level conclusions are not automatically member-level, and this entry's own open questions already flag possible IGF2BP1/2 redundancy. Not a leukaemia system and not a KMT2A-fusion system. The m6A-dependence is established by motif recognition and KH-domain mapping rather than by a fusion-context experiment.

    Evidence that argues against this (1)
    • Reader identity does not fix direction of effect, and in this disease the direction inverts. The IGF2BP family's characterised behaviour is to STABILISE m6A-marked targets, which predicts that IGF2BP3 raises the transcripts it binds - yet the only direct measurement on the fusion transcript itself shows IGF2BP3 binding both reciprocal t(4;11) transcripts and REDUCING MLL-AF4 mRNA and protein while RAISING AF4-MLL. This entry already records that inversion. So a stabilising m6A reader is, on the index fusion's own transcript, a destabilising influence, and the family-level mechanism cannot be used to predict what IGF2BP3 does here. Hanewald T et al. 2026
  10. E10moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    THE METABOLIC ROUTE FROM IGF2BP3 TO A METHYL DONOR: metabolic profiling of IGF2BP3 function shows changes in glycolytic flux and one-carbon metabolism including the biosynthesis of S-adenosylmethionine, a key substrate for methylation reactions in the cell; enhanced crosslinking immunoprecipitation (eCLIP) and polyribosome profiling show that IGF2BP3 promotes translation of MAT2B, the regulatory subunit of the methionine adenosyltransferase complex involved in SAM production; and IGF2BP3 promotes and alters the level and pattern of m6A modifications on mRNA.

    IGF2BP3 redirects glycolytic flux to promote one-carbon metabolism and RNA methylation. · Sharma G, Gutierrez M, Jones AE, et al. · Cell Rep 2025
    PMID 41015030DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read in this tranche. Already cited on this entry as E4, and recorded again here because E4's framing is metabolic while the epigenetic significance - that SAM is the methyl donor shared with every histone methyltransferase - is a separate and unstated consequence. SEPARATELY, A FLAG ON THE EXISTING ENTRY: E4's claim text includes the clause 'an m6A/IGF2BP3 axis drives serine biosynthesis supporting AML stemness and a metabolic vulnerability', and that clause does not appear in this PMID's abstract, which describes glycolytic flux, one-carbon metabolism, MAT2B, SAM and m6A without mentioning AML stemness. Either the full text supports it or it is spliced from a second source; scripts/verify-claims.py is the instrument for that and it was not run against this item in this tranche. The measured output here is m6A, NOT a histone mark: nothing in this paper measures histone methylation.

    Evidence that argues against this (1)
    • The SAM pool is a shared resource, which cuts against attributing IGF2BP3's effects to the epitranscriptome specifically. In MLL-AF4 cell lines and MLL-AF6 patient blasts, reducing methylation potential by methionine deprivation or downstream pharmacological inhibition causes a profound loss of H3K79 methylation and raises H3K27me3 at the DOT1L promoter - so in the index fusion's own cells the SAM pool is demonstrably rate-limiting for a HISTONE mark. If IGF2BP3 raises SAM via MAT2B, then some of what is currently read as an m6A phenotype could be a histone-methylation phenotype, and the two papers have never been run in the same experiment. This is an alternative explanation, not a refutation. Barve A et al. 2019
  11. E11strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    IN THE INDEX FUSION'S CELLS THE METHYL-DONOR POOL IS RATE-LIMITING FOR H3K79: disruption of methionine/S-adenosylmethionine metabolism, by either methionine deprivation or pharmacological inhibition of downstream metabolism, reduced overall cellular methylation potential, reduced relative cell numbers and induced apoptosis selectively in established MLL-AF4 cell lines and MLL-AF6-expressing patient blasts but not in BCR-ABL-driven K562 cells; global histone methylation dynamics were altered with a profound loss of requisite H3K79 methylation indicating inhibition of DOT1L function, and relative occupancy of H3K27me3 was increased at the DOT1L promoter with reduced DOT1L mRNA and protein expression.

    PMID 31717699DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read in this tranche. NOT CITED anywhere in this corpus before 2026-09-21, verified with scripts/check-citation-reuse.py - which is itself worth noting for a graph organised around H3K79 methylation. The MLL-AF4 arm is established cell lines, not primary infant B-lineage material; the patient-blast arm is MLL-AF6, a different partner. K562 is a single comparator, so 'selective' rests on one negative control line. Methionine deprivation is a blunt perturbation that lowers every SAM-dependent methylation simultaneously, so the profound H3K79 loss is a sensitivity ranking rather than a specific mechanism - and the parallel fall in DOT1L mRNA and protein means some of the H3K79 loss is loss of the enzyme rather than loss of its substrate, which the entry should not conflate. This paper says nothing about IGF2BP3.

    Evidence that argues against this (1)
    • Global histone methylation dynamics under methyl-donor restriction are not equivalent to a locus-resolved dependency, and the calibrated measurement in this disease shows why that matters. By ICeChIP-seq, depleting H3K79me2 pharmacologically produces PARADOXICAL INCREASES of H3K4me3 and loss of H3K27me3 at fusion targets - marks moving in opposite directions under a single perturbation. A bulk 'methylation potential' readout cannot see that, so 'H3K79 is the most sensitive mark to SAM restriction' is a statement about bulk abundance and not about which loci lose function. It also makes the reported H3K27me3 INCREASE at the DOT1L promoter a result that runs opposite to the H3K27me3 loss seen under direct DOT1L inhibition, and neither paper reconciles them. Richter WF et al. 2021
Connected to