The RNA layer and the menin interface are separable: IGF2BP3 depletion sensitises MLL-AF4 leukaemia to menin inhibition and outperformed it alone
also called IGF2BP3 menin combination · post-transcriptional plus transcriptional targeting
Menin inhibitors work by prying the fusion protein off the genes it switches on. Removing an RNA-binding protein instead attacks the same leukaemia from a completely different side, and in mice doing both together worked better than either alone — with the RNA side, on its own, working better than the menin drug.
The same thing, for a scientist
Depletion of Igf2bp3 sensitises MLL-AF4 leukaemia to menin-MLL inhibition on cell growth and on leukaemic initiating cells in vitro; both manipulations increase differentiation in vitro and in vivo; and **IGF2BP3 knockdown had a greater effect on increasing survival and attenuating disease than pharmacological menin-MLL inhibition with MI-503 alone**, with enhanced antileukaemic effects in combination [E1]. The authors frame it as targeting leukaemia at both the transcriptional and post-transcriptional level. This is the single most decision-relevant result the RNA layer produces for this graph, because it is the only place where a non-chromatin intervention is measured head-to-head against the menin interface in the index fusion - and it is also the result most exposed to a like-for-like objection [E1 counter].
What this does and does not say about the menin-interface hypothesis
This graph's central bet is the menin-KMT2A interface. The result above is the closest thing in the RNA literature to a challenge, so it is worth being precise about its scope.
It does not refute the menin hypothesis. Menin inhibition still worked, still increased differentiation, and still combined productively 1. Nothing here says the interface is the wrong target.
It does challenge the sufficiency claim. If a post-transcriptional intervention sensitises MLL-AF4 leukaemia to menin inhibition, then menin inhibition alone leaves a reservoir of disease that a different axis reaches. This graph already holds resistance data showing menin-pocket mutations and adaptive escape; a sensitising partner on a completely different axis is the constructive version of the same observation.
The head-to-head comparison is where it is weakest. A genetic knockdown removing a protein entirely is not comparable to partial reversible occupancy by a tool compound. MI-503 is not revumenib as dosed in patients. 'IGF2BP3 knockdown beat MI-503' is a statement about two very different manipulations and should not be paraphrased as 'IGF2BP3 is a better target than menin'.
Two opposite predictions, and why they discriminate
The combination behaviour of an RNA-layer target with menin inhibition depends on whether the two act on the same programme or on different ones, and the beat now contains genes on both sides.
Different axes predict synergy. IGF2BP3 acts on transcript stability, splicing and RISC competition for its own target set 2, which need not overlap with the fusion's chromatin programme. Observed: sensitisation and enhanced combination effect 1.
The same axis predicts sub-additivity. PTBP1's CLIP targets are IKZF1, MEF2C, EZH2, SIK3 and PBX3 3 - transcripts of the KMT2A-r programme itself. If PTBP1's requirement runs through fine-tuning exactly what the fusion drives, then removing PTBP1 and removing the fusion's chromatin function should be partly redundant.
Nobody has run the PTBP1-plus-menin-inhibitor experiment. It is a single combination assay in an existing model, and its outcome would tell this graph whether the RNA layer is an orthogonal axis or a second point on the same one. That distinction decides whether RNA-layer targets belong in combination strategy at all.
The chromatin-to-RNA bridge makes the same question sharper
If H3K79me2 guides RBM15B-directed m6A deposition to start codons and thereby sets translation efficiency of oncogenic transcripts 4, then the chromatin layer and the RNA layer are not two systems but one, coupled at a specific molecular step.
Under that model, DOT1L inhibition should redistribute m6A, and a DOT1L inhibitor plus an m6A-pathway inhibitor would be redundant rather than additive. Under the IGF2BP3 model they are orthogonal and additive 1. Both models are in the same beat and both cannot be generally true, which is the sort of tension the graph should hold explicitly rather than average away.
| mechanism class | combination / axis separability |
|---|---|
| observed | IGF2BP3 depletion sensitises MLL-AF4 leukaemia to menin-MLL inhibition; knockdown alone outperformed MI-503 alone; combination enhanced |
| untested experiment | PTBP1 depletion combined with a menin inhibitor in a KMT2A-rearranged model; a single combination assay that discriminates the models |
competing models
[ "orthogonal axes - predicts synergy (IGF2BP3 case, observed)", "one coupled axis via H3K79me2-RBM15B-m6A - predicts redundancy", "same programme at two steps (PTBP1 case) - predicts sub-additivity" ]
relevant drugs
[ "drug-revumenib", "drug-i3in-002" ]
- A genetic knockdown outperforming a tool compound is not evidence that the knocked-down protein is the better target; it is evidence that complete removal beats partial occupancy.
- If the chromatin and RNA layers are coupled through H3K79me2-directed m6A, then combining a DOT1L inhibitor with an m6A-pathway inhibitor buys nothing and costs toxicity.
- Adding a second agent to a menin inhibitor in infants means adding a second toxicity profile to a population in which differentiation syndrome and QTc prolongation are already dose-limiting.
- The only head-to-head data are murine and genetic. No clinical or even pharmacological combination data exist for any RNA-layer target in this disease.
- Does PTBP1 depletion combine sub-additively with menin inhibition, as the shared-programme model predicts?
- Does IGF2BP3's sensitisation of menin inhibition survive replacement of MI-503 with revumenib at clinically achievable exposure?
- Does DOT1L inhibition redistribute m6A, which would establish the chromatin and RNA layers as one coupled system rather than two?
- Is the IGF2BP3-menin interaction mediated by fusion protein dose, given that IGF2BP3 is reported to lower KMT2A::AFF1 abundance?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Depletion of Igf2bp3 sensitises MLL-AF4 leukaemia to menin-MLL inhibition on cell growth and leukaemic initiating cells in vitro; both Igf2bp3 depletion and menin-MLL inhibition increase differentiation in vitro and in vivo; IGF2BP3 knockdown had a greater effect on increasing survival and attenuating disease than pharmacological menin-MLL inhibition with MI-503 alone, and the two showed enhanced antileukaemic effects in combination.
Targeting IGF2BP3 enhances antileukemic effects of menin-MLL inhibition in MLL-AF4 leukemia · Lin TL, et al. · Blood Adv 2024What this evidence cannot show
Abstract read; full text not read. Murine MLL-AF4 models. The comparison is genetic depletion against a tool-compound menin inhibitor, which is not like-for-like: complete protein removal outperforms partial reversible occupancy for almost any target. MI-503 is not a clinical menin inhibitor.
Evidence that argues against this (2)
- A different laboratory reports IGF2BP3 binding both reciprocal t(4;11) fusion transcripts and REDUCING KMT2A::AFF1 protein abundance. If IGF2BP3 loss raises fusion protein, the mechanism of the combination effect is not what the amplifier framing implies, and the interaction with a menin inhibitor could be operating through fusion dose rather than through an orthogonal axis. Hanewald T et al. 2026
- Both manipulations converge on increased differentiation, which is a shared endpoint rather than an orthogonal one, and differentiation is the mechanism menin inhibitors are already understood to work through. Convergent endpoints are weak evidence for separable axes. Jaiswal AK et al. 2026
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.preprint
IGF2BP3 antagonises RISC-mediated repression of oncogenic mRNAs, competing directly with AGO2-miRNA complexes for 3'UTR binding at 111 IGF2BP3-dependent AGO2 sites, with miR-181a over-expression phenocopying IGF2BP3 loss.
IGF2BP3 remodels the microRNA targeting landscape in MLL-AF4 leukemia · Kabbani LES, et al. · bioRxiv 2025What this evidence cannot show
Preprint, not peer reviewed at the time of writing; abstract read only. A third proposed mechanism for IGF2BP3 in this disease alongside the mRNA-stability/splicing operon and metabolic rerouting reports.
Evidence that argues against this (1)
- If IGF2BP3's action is competition with RISC on 3'UTRs, its target set is defined by miRNA occupancy rather than by the fusion, which would make the axis orthogonal to menin by construction and the synergy unsurprising rather than informative. Lin TL et al. 2024
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
PTBP1 CLIP-seq shows preferential binding to a subset of transcripts critical for KMT2A-rearranged AML proliferation including IKZF1, MEF2C, EZH2, SIK3 and PBX3, and PTBP1 loss disrupts the KMT2A-r-essential transcriptional programme with widespread splicing dysregulation.
PTBP1 controls oncogenic transcript processing and maintenance of acute myeloid leukemia · Cao Z, et al. · Leukemia 2026What this evidence cannot show
Abstract read; full text not read. AML. The overlap between PTBP1's targets and the menin-dependent programme is inferred from the named transcripts, not measured as an overlap statistic.
Evidence that argues against this (1)
- Sharing named target genes does not establish that two interventions act on the same rate-limiting step; PTBP1 controls splicing and processing of those transcripts while the fusion controls their transcription, and the two could be additive on the same genes. Lin TL et al. 2024
- E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
H3K79me2 guides RBM15B-directed site-selective m6A deposition in 5'UTRs and around start codons, enhancing translation efficiency of oncogenic transcripts and promoting leukaemic stem-cell self-renewal and leukaemia maintenance in MLL-rearranged leukaemia.
RBM15B recognizes H3K79me2 to guide selective m(6)A-modification of mRNA and enhance oncoprotein translation in MLL-r leukemia · Chen TQ, et al. · EMBO J 2026What this evidence cannot show
Abstract read; full text not read. 2026, not independently reproduced, and not lineage-specified in the abstract. The coupling claim is the entry's most consequential and its least established.
Evidence that argues against this (1)
- RBM15B has a Chronos mean of -0.041 with 0.2% of 1,208 DepMap lines depleted and positive index-arm differences in both lineage halves. A coupling mechanism whose central protein produces no dependency signal anywhere needs the self-renewal-assay explanation to stand, and that explanation is used repeatedly across this beat without independent support. Oncofusion knowledge graph (internal analysis) 2026
IGF2BP3 depletion sensitises MLL-AF4 leukaemia to menin-MLL inhibition
The clinical menin inhibitor this combination result bears on, though the experiment used MI-503
Tests whether the menin interface is sufficient, by adding a post-transcriptional axis alongside it
Predicted to be sub-additive with menin inhibition because its CLIP targets are KMT2A-r programme transcripts
- ← related toRBM15B — the bridge from the DOT1L mark to the RNA layer: it reads H3K79me2 and steers m6A to start codons
The H3K79me2-m6A coupling model, which predicts redundancy rather than synergy