RNA m6A (METTL3/YTHDF/YTHDC1) in KMT2A-rearranged leukemia — the empty cell that turned out not to be empty
also called m6A METTL3 MLLr · epitranscriptome KMT2A-r · METTL3-YTHDC1-CTCF axis · arcRNA m6A
Cells put small chemical tags on their RNA, and one of those tags turns out to be needed for leukemia cells to keep their DNA folded the way the cancer program requires. In KMT2A-rearranged leukemia the enzyme that writes the tag is itself switched on by KMT2A, and there is already a drug for it.
The same thing, for a scientist
We entered this beat expecting RNA modification to be an honest empty cell for KMT2A-r. It is not. Three lines converge: (1) METTL3 is over-expressed in AML relative to normal HSPCs, its depletion drives differentiation and apoptosis and delays leukemia in vivo, and m6A promotes translation of MYC, BCL2 and PTEN mRNAs in the KMT2A::MLLT3 line MOLM-13 [E1]; (2) a selective catalytic METTL3 inhibitor, STM2457, reduces AML growth, increases differentiation and apoptosis and prolongs survival in mouse models, with a co-crystal structure with METTL3-METTL14 [E2]; (3) most directly, in MLL-rearranged AML METTL3 is **transcriptionally activated by MLL**, complexes with YTHDC1 and CTCF, co-localises at promoters and enhancers, and m6A-marks architectural RNAs to stabilise R-loops and sustain CTCF-dependent TAD boundary activity — METTL3 depletion disrupts CTCF-binding sites and accessibility at MYB and RUNX1, and YTHDC1 loss compromises CTCF-dependent 3D organisation by Hi-C [E3]. The gap that remains is specific and narrow: none of this has been done in KMT2A::AFF1 infant B-ALL.
Why this belongs in a genome-and-non-coding beat at all1 deeper
Because the 2025 result is not really about mRNA translation. It says m6A on architectural RNA is part of how the 3D genome boundary is held 3. That makes the epitranscriptome a structural component of the chromatin scaffold, which is the same scaffold that genome-ctcf-cohesin-hoxa-tad-boundary shows the fusion program depends on. If that holds up, there is a chain running: MLL → METTL3 → m6A on arcRNA → YTHDC1 → R-loop → CTCF boundary → TAD → leukemic transcription. Every link in that chain is a potential intervention point and two of them (METTL3 catalytically, R-loops via RNase H) are already tractable.
It also means the cleanest KMT2A-r-relevant m6A claim is a feed-forward loop: the fusion's parent gene drives the writer that stabilises the architecture the fusion needs.
How much of this is actually KMT2A-r and how much is generic AML
Being precise, because this is where the literature gets sloppy:
- Generic AML, strong: METTL3 over-expression versus normal HSPCs; differentiation on depletion; STM2457 efficacy 12.
- KMT2A-fusion cell line, strong: the m6A/ribosome-profiling work identifying MYC, BCL2 and PTEN translational targets was done in MOLM-13, a KMT2A::MLLT3 line 1. So a KMT2A-fusion cell was the discovery system, though the paper does not claim fusion-specificity.
- MLL-rearranged AML, moderate and single-group: the METTL3-YTHDC1-CTCF architectural axis 3.
- KMT2A::AFF1 infant B-ALL: nothing. No m6A map, no METTL3 dependency, no STM2457 experiment that we could find in a t(4;11) B-lineage model.
The YTHDF readers, which the PI's framing named, are the weakest link in the KMT2A-r chain: the direct architectural evidence runs through YTHDC1 (a nuclear reader), not the cytoplasmic YTHDF1/2/3 proteins. Anyone citing 'the METTL3/YTHDF axis' in KMT2A-r should be asked for the YTHDF paper; we could not find one specific to this disease.
What the lab should do with this
This is the highest-value ready-to-run item the sweep produced, because the reagents exist and the disease-specific experiment has not been done.
- Cheap and immediate: DepMap METTL3, METTL14, YTHDC1 and WTAP dependency scores restricted to KMT2A::AFF1 lines (SEM, SEMK2, RS4;11, MV4-11) versus other leukemia lines. If KMT2A::AFF1 lines are not selectively dependent, the whole idea deflates for this disease at essentially zero cost. Written up as
ana-m6a-machinery-dependency-kmt2a-aff1. - One month, one plate: STM2457 dose-response in KMT2A::AFF1 B-ALL lines, alone and with revumenib. A synergy readout would be immediately publishable and immediately relevant.
- The honest falsifier: if METTL3 dependency in KMT2A::AFF1 lines is no greater than in unselected leukemia lines, then m6A in this disease is general leukemia biology, not fusion biology, and belongs in a different chapter.
| class | RNA modification / epitranscriptome |
|---|---|
| writer | METTL3-METTL14 (+WTAP) |
| nuclear reader with direct mllr evidence | YTHDC1 |
readers without kmt2a r evidence
[ "YTHDF1", "YTHDF2", "YTHDF3" ]
evidence tier by disease
{
"aml_generic": "strong",
"kmt2a_fusion_myeloid_cell_lines": "strong (MOLM-13, KMT2A::MLLT3)",
"mll_rearranged_aml_architecture": "moderate, single group",
"kmt2a_aff1_infant_b_all": "none identified"
}chemical matter
[
{
"compound": "STM2457",
"target": "METTL3 catalytic site",
"status": "preclinical tool with co-crystal structure",
"refs": [
"E2"
]
}
]proposed analyses
[ "ana-m6a-machinery-dependency-kmt2a-aff1" ]
translational targets reported
[ "MYC", "BCL2", "PTEN" ]
architectural targets reported
[ "MYB", "RUNX1" ]
- m6A is required for normal haematopoietic differentiation as well as leukemic self-renewal, so METTL3 inhibition in an infant risks an on-target developmental haematopoietic toxicity that mouse AML models will not reveal.
- If METTL3 is transcriptionally driven by wild-type MLL rather than by the fusion, then the dependency is a lineage property, not a fusion property, and selectivity over normal progenitors would be poor.
- A differentiation-inducing agent added to menin inhibition compounds differentiation-syndrome risk, which is already the dose-limiting on-mechanism toxicity of menin inhibitors.
- Reliance on a single unreplicated 2025 paper for the architectural mechanism is itself a risk; if it does not replicate, the chain from m6A to TAD integrity breaks and only the translational rationale survives.
- Is METTL3 or YTHDC1 a selective dependency in KMT2A::AFF1 B-lineage lines relative to unselected leukemia lines? A DepMap query answers this today and could kill the idea for this disease at no cost.
- Does STM2457 synergise with revumenib in KMT2A::AFF1 models, and if so is the synergy architectural (loop loss) or translational (MYC/BCL2 loss)?
- Is there any YTHDF-specific evidence in KMT2A-r, or is the widely cited 'METTL3/YTHDF axis' in this disease an extrapolation from other AML subtypes? We could find none.
- Because BCL2 translation is m6A-dependent in a KMT2A-fusion line, would METTL3 inhibition sensitise to venetoclax, or would it instead phenocopy and thus not add?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
METTL3 is more abundantly expressed in AML cells than in healthy haematopoietic stem/progenitor cells; its depletion induces differentiation and apoptosis and delays leukemia progression in vivo, and m6A promotes translation of MYC, BCL2 and PTEN mRNAs in the KMT2A::MLLT3 cell line MOLM-13.
The N6-methyladenosine (m6A)-forming enzyme METTL3 controls myeloid differentiation of normal hematopoietic and leukemia cells · Vu LP, Pickering BF, Cheng Y, Zaccara S, Nguyen D, Minuesa G, Chou T, Chow A, Saletore Y, MacKay M, Schulman J, Famulare C, Patel M, Klimek VM, Garrett-Bakelman FE, Melnick A, Carroll M, Mason CE, Jaffrey SR, Kharas MG · Nature Medicine 2017What this evidence cannot show
shRNA depletion rather than pharmacology for the mechanistic work; the single-nucleotide m6A/ribosome-profiling was in one cell line (MOLM-13). The paper makes a myeloid-differentiation argument and does not claim any KMT2A-fusion specificity, so using MOLM-13's genotype as evidence of fusion relevance is our inference.
Evidence that argues against this (1)
- Higher expression in AML than in healthy HSPCs does not make the dependency selective: the same enzyme controls myeloid differentiation of normal haematopoietic cells, so depletion phenotypes in leukaemia and in normal progenitors point the same way. The MYC, BCL2 and PTEN translation result is in MOLM-13, a single KMT2A::MLLT3 line, and MYC and BCL2 are the two transcripts a global translation perturbation is most likely to move. Nature Medicine 2017
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
STM2457 is a selective catalytic METTL3 inhibitor, co-crystallised with METTL3-METTL14, that reduces AML growth, increases differentiation and apoptosis, selectively lowers m6A on leukaemogenic mRNAs and prolongs survival in multiple mouse AML models including targeting of AML stem cell subpopulations.
Small-molecule inhibition of METTL3 as a strategy against myeloid leukaemia · Yankova E, Blackaby W, Albertella M, Rak J, De Braekeleer E, Tsagkogeorga G, Pilka ES, Aspris D, Leggate D, Hendrick AG, Webster NA, Andrews B, Fosbeary R, Guest P, Irigoyen N, Eleftheriou M, Gozdecka M, Dias JML, Bannister AJ, Vick B, Jeremias I, Vassiliou GS, Rausch O, Tzelepis K, Kouzarides T · Nature 2021What this evidence cannot show
Myeloid models only; no B-lineage KMT2A::AFF1 model reported. Authors include the developing company (Storm Therapeutics), so independent replication matters. m6A is required in normal HSPCs too, so the therapeutic index in an infant is unestablished.
Evidence that argues against this (1)
- A co-crystallised selective catalytic inhibitor establishes chemistry, not therapeutic window. "Selectively lowers m6A on leukaemogenic mRNAs" is a statement about which transcripts move most in a global methylation readout, which is the expected behaviour of the most heavily methylated and most rapidly turned-over messages rather than evidence of target selection. Nature Medicine 2017
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In MLL-rearranged AML, METTL3 is transcriptionally activated by MLL and forms a complex with YTHDC1 and CTCF that co-localises at promoters and enhancers; METTL3-mediated m6A stabilises architectural RNAs and R-loops to maintain CTCF-dependent TAD boundary activity, with METTL3 depletion disrupting CTCF-binding sites and accessibility at MYB and RUNX1 and YTHDC1 loss compromising CTCF-dependent 3D organisation.
The METTL3-YTHDC1 axis mediates architectural RNA m6A modification to modulate the integrity of chromatin TADs in MLLr+ AML genome · Fu R, Yu W, Zhao R, Yang H, Cao B, Dai C, Qianjiang H, Xia Y, Wang L, Lu J, Zhou R, Shi A, Lou H, Zhang X, Luo H · Molecular Cancer 2025What this evidence cannot show
Single recent study, MLLr AML cell lines, multi-omics correlation carrying a heavy mechanistic load; senior authorship overlaps the HOTTIP/CTCF-boundary group, so it is not independent of that body of work. MALAT1 is the exemplar arcRNA, and MALAT1's own functional necessity is contested in other systems. Not replicated.
Evidence that argues against this (1)
- The claim that m6A machinery is required for CTCF boundary function sits uneasily with the fact that CTCF/cohesin loop formation is well explained by loop extrusion without an RNA-modification requirement; a generic RNA-dependence of CTCF binding has been argued and disputed in the wider literature, and BRD4-degradation work shows loops are robust to loss of several factors previously thought essential. Crump NT et al. 2021
DRIP-seq in MLLr AML (PMID 41430607) is the only direct R-loop map in this disease and both entries rest on it
- ← related toCTCF/cohesin boundaries and the HOXA topological domain — the 3D scaffold the fusion program runs on
Method counter (Smolka 2021, PMID 33830170; Miller 2022, PMID 35758606): the arcRNA/R-loop arm of the METTL3-YTHDC1 result rests on S9.6 DRIP-seq, an assay whose antibody cross-reacts with rRNA and whose calls diverge starkly from dRNH-based methods; no RNase H1/T1 control is reported
Correction (Fu 2025, PMID 41430607): this hypothesis previously asserted that no R-loop map of any KMT2A-rearranged leukaemia existed; DRIP-seq in MLL-rearranged AML cells is published in that paper, so the claim of absence has been narrowed to KMT2A::AFF1 B-ALL and primary infant samples
both propose that genome architecture rather than sequence is the affected layer