Oncofusion

start / graph / mechanism-rbp-control-of-fusion-dose

Mechanismmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q6-interactome-rna-layer

RNA-binding proteins set how much KMT2A::AFF1 protein a cell makes from its own fusion transcript

also called post-transcriptional control of MLL-AF4 · fusion dose · translational control of KMT2A::AFF1

The amount of fusion protein a leukaemia cell makes is not fixed by the broken chromosome alone — proteins that grip the fusion's messenger RNA decide how much of it gets built. Two different laboratories have now found RNA-binding proteins sitting on the t(4;11) message and turning its output down.

The same thing, for a scientist

Two independent reports place RNA-binding proteins on the KMT2A::AFF1 transcript itself, both reducing fusion protein output. KHDRBS-family proteins bind AU-rich regulatory sites and stall the ribosome; synonymous mutations that remove those sites restore translation and make an otherwise poorly-transforming MLL-AF4 construct leukaemogenic in vivo [E1]. Separately, IGF2BP3 binds **both** reciprocal fusion transcripts, reducing KMT2A::AFF1 (MA4) protein abundance while increasing AFF1::KMT2A (A4M) — an effect absent for a murinised KMT2A::mAff1 construct [E2]. The mechanism implies that fusion dose is a regulated variable set after transcription, that a failure to model MLL-AF4 in mouse is partly a translational ceiling rather than a statement about the fusion's potency, and that this graph's treatment of the fusion as present-or-absent is an approximation.

Go deeper
Why a post-transcriptional layer can matter to a fusion that acts on transcription

The obvious objection to an RNA layer in this disease is that KMT2A::AFF1 is a chromatin-tethered elongation driver: it works on DNA, so why should RNA metabolism be more than downstream scenery?

The answer this mechanism gives is not 'the RNA layer is also important'. It is narrower and harder to dismiss: the RNA layer is upstream of the fusion, not downstream of it. The fusion is itself a transcript that must be translated, and the proteins that decide how efficiently that happens are RNA-binding proteins 12. Everything the fusion does on chromatin is scaled by a quantity these proteins set.

That is a different class of claim from the one the rest of this beat makes, and it is the one worth defending. A post-transcriptional regulator of a fusion's downstream targets competes with dozens of chromatin mechanisms for explanatory weight. A post-transcriptional regulator of the fusion's own dose does not compete with them; it multiplies them.

The objection this entry has to survive, stated fairly

The objection. An RNA-binding protein's abundance tracks biosynthetic demand. A fast-growing blast makes more ribosomes, more spliceosome, more m6A machinery and more IGF2BP protein than a resting lymphocyte, so any comparison of leukaemia to normal marrow will find RNA-binding proteins 'up in leukaemia', and any dependency screen in fast-growing lines will find them 'required'. Two of the three routes this beat used to enumerate genes - expression association and dependency screening - are therefore systematically biased toward this class for reasons that have nothing to do with the fusion. The IGF2BP3 literature contains exactly this shape: an association with a proliferative phenotype in B-ALL 6, which is close to circular.

Why it does not defeat this particular mechanism. The objection is an argument about correlational evidence, and the two reports here are not correlational.

  • The KHDRBS result turns on a synonymous mutation 1. The protein sequence is unchanged, the construct's expression is unchanged, and only the RNA regulatory sites differ - yet transformation appears. A biosynthetic-demand confounder cannot produce that contrast, because nothing about the cell's growth state differs between the two constructs.
  • The IGF2BP3 fusion-transcript result is differential in a way demand cannot mimic: the same protein raises one reciprocal product and lowers the other, and the effect disappears for a murinised fusion 2. A general abundance effect would move both in the same direction.

Where the objection does bite. It fully applies to the expression and prognosis literature on this class 6, and it applies to the DepMap-based arguments in this beat. This graph's own scan found no RNA-layer gene approaching menin-grade selectivity, and several of the strongest rows are pan-essential genes whose differences are floor-compressed 5. So: the dose mechanism survives the objection on its own evidence; the broader claim that the RNA layer is a distinctive vulnerability of KMT2A-rearranged disease does not, and should not be asserted.

A tension the graph should carry rather than resolve

Both reports have RNA-binding proteins reducing fusion protein output 12. Yet the most-cited result in this area is that IGF2BP3 amplifies MLL-AF4 leukaemogenesis, with Igf2bp3 deletion greatly attenuating disease 3.

Those are not directly contradictory - IGF2BP3 could lower fusion dose while raising the output of the fusion's targets, and the amplification paper's targets are HOXA-locus and RAS-pathway transcripts rather than the fusion itself. But it does mean one thing should stop being said: that IGF2BP3 promotes this leukaemia by increasing the fusion. The only measurement of that quantity reports the opposite sign 2.

A third mechanism has since been proposed for the same protein - competition with AGO2/RISC for 3'UTR sites, restricting miR-181a-mediated repression of oncogenic messages 4. Three mechanisms for one protein in one disease is a reason for caution rather than confidence.

What follows for experiments in this graph

Three consequences, in decreasing order of how confident the graph should be.

  1. Any conclusion drawn from a low-penetrance retroviral MLL-AF4 mouse model is ambiguous. The difficulty of modelling this fusion, long read as biology, is at least partly a translational ceiling on the construct 1. Entries citing weak murine transformation as evidence about the fusion should say which construct was used.
  2. Fusion protein per unit fusion transcript is a measurable variable that nobody in this graph measures. Patient samples with identical breakpoints could differ in it.
  3. A construct that is codon-optimised or synonymously altered is not the same molecule as the endogenous fusion, for this specific reason. That is a design constraint on any new model system built here.
Facts and figures
mechanism classpost-transcriptional / translational control
acts onthe KMT2A::AFF1 and AFF1::KMT2A transcripts themselves, not on the fusion protein's chromatin targets
directionboth reported effects REDUCE KMT2A::AFF1 protein output
key discriminating evidencea synonymous mutation in the MLL-AF4 coding sequence, which changes RNA regulatory sites without changing protein sequence or cell growth state, and which confers transformation
proteins
[
  "gene-khdrbs1",
  "gene-khdrbs2",
  "gene-khdrbs3",
  "gene-igf2bp3"
]
How this could bite us back
  • If RNA-binding proteins restrain fusion translation, an inhibitor of one of them would raise fusion dose. Several proteins in this beat are on target lists for other reasons, and the sign of their effect in t(4;11) disease has not been checked.
  • Menin inhibitors act on the fusion's chromatin function and not on its abundance. A cell that raises fusion dose post-transcriptionally has a resistance route that no menin-pocket mutation screen would detect.
  • The biosynthetic-demand confounder is not disposed of by this entry; it is confined. Every expression-based and dependency-based argument in this beat remains subject to it.
  • Codon-optimised or synonymously altered fusion constructs used to build models may be mechanistically different molecules from the endogenous fusion, which would propagate silently into every result obtained with them.
Still open
  1. What is KMT2A::AFF1 protein per unit fusion transcript in primary t(4;11) blasts, and does it vary between patients with the same breakpoint?
  2. Do the AU-rich KHDRBS sites exist in the endogenous human fusion transcript, or were they a property of the retroviral construct?
  3. If IGF2BP3 lowers fusion protein while amplifying the disease, what does it amplify - and can that be separated from fusion dose by titrating the fusion directly?
  4. Does the reciprocal product AFF1::KMT2A, which IGF2BP3 raises, contribute enough to explain the amplification without any effect on the forward fusion?
Every source behind this entry (6)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    MLL-AF4 translation is inhibited by ribosomal stalling at regulatory sites containing AU-rich sequences recognised by KHDRBS-family RNA-binding proteins; synonymous mutations disrupting KHDRBS association restore proper MLL-AF4 translation and result in leukaemic transformation, and the synonymous MLL-AF4 mutant induces leukaemia in vivo.

    PMID 36335100DOI
    What this evidence cannot show

    Abstract read; full text not read. Murine retroviral transduction. The manipulation is of the transcript's sites, not of the KHDRBS proteins, so the experiment shows the sites are load-bearing more directly than it shows any named protein is.

    Evidence that argues against this (1)
    • The regulation was characterised in a construct whose poor translation was the problem being solved. Whether the endogenous human t(4;11) transcript, transcribed from its native locus with native UTRs, carries the same ceiling is not established by this experiment. Hanewald T et al. 2026
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Both reciprocal t(4;11) fusion transcripts are bound by IGF2BP3, resulting in reduced KMT2A::AFF1 (MA4) protein abundance and increased AFF1::KMT2A (A4M) abundance; the effect is not observed for the artificial murinised KMT2A::mAff1 fusion, which the authors attribute to its inability to recruit the SL-1 complex. In the same models MA4 alone drives aberrantly high expression of only a few target genes, A4M counteracts that, and MA4 induces genes encoding RNA-binding proteins as a counter-regulatory response.

    PMID 42537480DOI
    What this evidence cannot show

    Abstract read; full text not read. Model systems built for the purpose rather than patient material, and the reciprocal-product biology is a contested area in its own right. The SL-1 explanation for the murinised construct's behaviour is an interpretation offered in the abstract, not a demonstrated step.

    Evidence that argues against this (2)
    • This finding is in direct tension with the dominant reading of IGF2BP3 as an amplifier of MLL-AF4 leukaemogenesis, which does not report the fusion's own protein going down. Neither group tests the other's model, and the two are from different laboratories using different systems. Tran TM et al. 2022
    • That MA4 induces RNA-binding protein genes as a counter-regulatory response is the cleanest possible statement of the confounder this mechanism has to survive: if the fusion itself drives RBP expression, then finding RBPs elevated in fusion-positive cells is a consequence of the fusion and not evidence that they are a distinct vulnerability. Hanewald T et al. 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    IGF2BP3 is over-expressed in MLL-translocated leukaemia and amplifies MLL-Af4-mediated leukaemogenesis: Igf2bp3 deletion significantly increases survival and greatly attenuates disease with minimal impact on baseline haematopoiesis, and IGF2BP3 targets include HOXA-locus and RAS-pathway transcripts, with effects on both steady-state mRNA level and pre-mRNA splicing.

    PMID 34321607DOI
    What this evidence cannot show

    Abstract read; full text not read. Murine MLL-Af4 with a chimeric construct; deletion precedes disease rather than treating established leukaemia. 'Amplifies' is the paper's own verb and is weaker than 'requires'.

    Evidence that argues against this (1)
    • The direct measurement of IGF2BP3's effect on fusion protein abundance reports a reduction in KMT2A::AFF1, so whatever amplification IGF2BP3 provides is not occurring through raising fusion dose. Hanewald T et al. 2026
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.preprint

    IGF2BP3 antagonises RISC-mediated repression of oncogenic mRNAs: miR-eCLIP of AGO2 in IGF2BP3-knockout versus control B-ALL lines identifies IGF2BP3-dependent AGO2 binding at 111 3'UTRs, miR-181a targeting increases on knockout, miR-181a over-expression phenocopies IGF2BP3 loss, and biochemical assays show direct competition between IGF2BP3 and AGO2-miRNA complexes for 3'UTR binding.

    PMID 41279410DOI
    What this evidence cannot show

    Preprint, not peer reviewed at the time of writing. Abstract read only. A third proposed mechanism for the same protein in the same disease, alongside the mRNA-stability/splicing operon and the metabolic rerouting reports.

    Evidence that argues against this (1)
    • Having three mechanisms for one protein in one disease from three groups - a post-transcriptional operon, a one-carbon metabolic rerouting, and RISC competition - is a reason for caution. None of the three papers tests the others' models, and a protein that binds thousands of transcripts will produce a defensible mechanism from whichever assay is applied. Sharma G et al. 2025
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.analysis

    A 76-gene RNA-metabolism panel scored on DepMap Chronos across 11 KMT2A-rearranged versus 43 other leukaemia lines, lineage-split, contains no gene approaching menin-grade selectivity; the best is HNRNPA2B1 at -0.122 lymphoid / -0.275 myeloid against MEN1 at -0.871 / -0.863, and 16 panel genes are floor-compressed at 90-100% of lines depleted.

    What this evidence cannot show

    Eleven index lines, four of them lymphoid. A dropout screen cannot see a dependency that acts on initiation or self-renewal, and cannot see a negative regulator at all, so the scan constrains the 'distinctive vulnerability' claim rather than the dose mechanism.

    Evidence that argues against this (1)
    • This scan is the strongest single argument against the broad version of this beat's thesis, and it is cited here for that purpose. It does not touch the dose mechanism, which rests on a synonymous-mutant experiment that no dependency screen could reproduce. Okuda H et al. 2022
  6. E6moderate 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; association in B-ALL generally, not stratified by KMT2A rearrangement.

    Evidence that argues against this (1)
    • Cited here as the confounder rather than as support. An association between an RNA-binding protein and proliferation is close to circular: fast-growing blasts express more of the machinery that makes them fast, so this observation would be expected under the null in which IGF2BP3 does nothing specific to this fusion. Hanewald T et al. 2026
Connected to