The pivotal menin-inhibitor paper reports that the drug does NOT globally displace MLL1 or the fusion from chromatin - menin goes, the fusion largely stays
also called MLL1n is not globally displaced by VTP-50469 · the fusion is not evicted genome-wide · ML2 fusion-only control
Everyone says menin inhibitors knock the leukaemia's fusion protein off the DNA, but the paper that launched these drugs says something different in its own results: menin comes off everywhere, while the fusion stays on the DNA almost everywhere and only leaves a handful of genes. The clearest test was in a cell line with no normal copy of the protein left, so every signal measured must be the fusion - and there too it stayed.
The same thing, for a scientist
In human KMT2A-rearranged leukaemia cell lines treated for 3 days with VTP-50469 (the preclinical analogue of revumenib), Krivtsov et al. report by ChIP-seq that the drug produces **global loss of menin binding** and, in the same experiment, that it **did not globally displace MLL1 from chromatin in MOLM13, RS4;11 and ML2 cells** [E1]. Loss of MLL1n signal is locus-restricted and the loci are named: greater than 2-fold loss at MEIS1, MEF2C and JMJD1C, and **no loss at the HOXA cluster, MYB or GAPDH** [E1]. The attribution problem that usually blunts this - the MLL1n antibody recognises wild-type MLL1 as well as the fusion - is handled inside the paper by including **ML2, an AML line carrying KMT2A::AFDN (MLL-AF6) that has lost the remaining wild-type KMT2A allele, so that all MLL1n signal is fusion**; the sites and magnitude of MLL1n loss across the genome were similar in ML2 to those in MOLM13 and RS4;11 [E1]. Genetic MEN1 knockout reproduces the pattern and its scale: CUT&RUN six days after Cas9 deletion identified **39 genes losing more than 2-fold MLL1n occupancy, 20 of them shared with the drug**, again sparing HOXA, MYB and GAPDH [E1]. Perner et al. describe the same geometry in KMT2A::AFF1 MV4;11 cells under clinical revumenib: near-complete global displacement of **menin** at 100 nM, with MLL1 loss reported from promoters with the highest degree of menin displacement (over 80%) and illustrated at MEIS1, PBX3, MEF2C, JMJD1C and the HOXA cluster as tracks [E2]. The one literature instance in which a **fusion protein** is reported evicted is the NUP98 fusion in NUP98-rearranged AML, and that too is explicitly scoped to a critical set of genes rather than genome-wide [E3]. **The consequence for this graph's vocabulary is direct.** Menin inhibitors evict the fusion from chromatin is a statement the founding pharmacology paper does not make and its own data argue against at genome scale. What is established is that the drug evicts **menin**, and that a small named set of fusion-target loci loses fusion occupancy with it.
What the ChIP-seq actually showed, readout by readout
Three human lines, 3 days of VTP-50469, ChIP-seq for menin, MLL1n, DOT1L and H3K79me2, quantified as TSS-anchored signal (windows from the TSS extending 10 kb into the gene body, 100 bp tiling) and displayed as genome-wide tornado plots of TSS +/- 3 kb 1.
| readout | result |
|---|---|
| menin | global loss across the genome, similar sites and magnitude in all three lines |
| MLL1n | not globally displaced; more than 2-fold loss only at a small named set |
| DOT1L | reduced but not lost; near-complete loss at many MLL-AF9 target genes |
| H3K79me2 | no detectable global loss; drastic loss at MEF2C, MEIS1, JMJD1C |
The supplementary tables the paper ships are Table S2 (gene occupancy of menin, MLL1n, DOT1L, H3K79me2) and Table S4, a list of genes that lose occupancy >2-fold of MLL1 and Menin. There is no table of genes that gain occupancy, in this or any other menin-inhibitor dataset - see absence-no-study-has-tested-for-gained-fusion-occupancy-under-menin-inhibition.
Why the ML2 line is the load-bearing control
This graph's own attribution audit established that ChIP has no size dimension, so an N-terminal MLL antibody cannot distinguish wild-type MLL1 from the fusion, and quantitative fusion occupancy does not survive that limitation (ana-attribution-audit-chromatin-occupancy-kmt2a-aff1). ML2 sidesteps the problem by genetics rather than by reagent - it expresses KMT2A::AFDN and has lost the second KMT2A allele, so every MLL1n read is fusion 1. That makes the sites and magnitude of MLL1n loss were similar between MOLM13, RS4;11 and ML2 the most direct published statement about what happens to a KMT2A fusion protein's chromatin occupancy under a menin inhibitor. It is a similarity statement rather than a measurement of the residual fraction, which is the limit worth carrying: the paper never reports what percentage of fusion peaks survives.
What this does NOT license
- It does not show the fusion is functional where it remains. Transcription of the programme collapses; occupancy largely does not. Those two facts together are the content of
hyp-the-fusion-stays-on-chromatin-and-stops-working. - It does not exclude a redistribution on top of the retention, because no gained-peak analysis was run anywhere in this literature (
absence-no-study-has-tested-for-gained-fusion-occupancy-under-menin-inhibition). - It does not transfer cleanly to infant KMT2A::AFF1. The lines are MOLM13 (adult AML, KMT2A::MLLT3), RS4;11 (adult B-ALL, KMT2A::AFF1) and ML2 (adult AML, KMT2A::AFDN). No infant-derived line appears in any menin-inhibitor occupancy dataset (
absence-no-menin-inhibitor-occupancy-dataset-in-an-infant-derived-kmt2a-aff1-line).
| drug | VTP-50469 (preclinical analogue of revumenib/SNDX-5613) |
|---|---|
| timepoint | 3 days (VTP-50469 ChIP-seq); 6 days (MEN1 CRISPR CUT&RUN) |
| n genes losing mll1n after men1 ko | 39 |
| n shared with drug | 20 |
| geo | GSE127508 |
| shown for fusion | KMT2A::AFDN in ML2 (fusion-only line, the load-bearing control); KMT2A::AFF1 in RS4;11 and MV4;11 with a shared-epitope antibody; nothing in an infant-derived KMT2A::AFF1 line |
systems
[
{
"line": "MOLM13",
"fusion": "KMT2A::MLLT3",
"lineage": "AML",
"wt_kmt2a_allele": "retained"
},
{
"line": "RS4;11",
"fusion": "KMT2A::AFF1",
"lineage": "B-ALL",
"wt_kmt2a_allele": "retained"
},
{
"line": "ML2",
"fusion": "KMT2A::AFDN",
"lineage": "AML",
"wt_kmt2a_allele": "lost - all MLL1n signal is fusion"
},
{
"line": "MV4;11",
"fusion": "KMT2A::AFF1",
"lineage": "AML/biphenotypic",
"wt_kmt2a_allele": "retained"
}
]loci losing fusion occupancy
[ "MEIS1", "MEF2C", "JMJD1C" ]
loci retaining fusion occupancy
[ "HOXA cluster", "MYB", "GAPDH" ]
- If the fusion is largely retained on chromatin, a pharmacodynamic biomarker built on fusion occupancy will read 'no target engagement' in a responding patient, and one built on menin occupancy will read 'full engagement' in a patient whose fusion never left. Both are in use as surrogate readouts in a drug class now in paediatric trials.
- Retained occupancy is a standing reservoir for resistance: any event that restores output at a site the fusion never vacated - PRC1.1 loss, JUND liberation, a CDK4/6-reversible tumour-suppressor programme - needs no re-recruitment step and can therefore be fast.
- A retained but inactive fusion may still be doing something: it occupies the uCpG address, it shields CpG from methylation, and it excludes other CXXC readers. Removing menin without removing the fusion could leave the address blocked and the programme off, which is a different steady state from a cleared address and may behave differently on drug withdrawal.
- If retention is real, combination partners should be chosen to act ON chromatin at the retained sites rather than to deepen menin displacement, which is already near-complete at 100 nM.
- What fraction of fusion peaks survives saturating menin inhibition? The pivotal paper states 'not globally displaced' and names the losses, but never reports the residual fraction, and that number is what any occupancy-based pharmacodynamic endpoint would have to be built on.
- Is the retained fusion at HOXA and MYB still assembled with SEC and PAFc, or is it a stripped monomer sitting on uCpG? Occupancy and complex composition have never been co-measured under drug.
- Does the ML2 result hold for a CXXC-competent index fusion in a B-lineage infant line, where the CXXC-uCpG tether is the same but the partner, the lineage and the methylome are not?
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
In MOLM13 (KMT2A::MLLT3), RS4;11 (KMT2A::AFF1) and ML2 (KMT2A::AFDN, wild-type KMT2A allele lost) human leukaemia cells treated for 3 days with VTP-50469, ChIP-seq showed global loss of menin binding to chromatin, but the authors state that treatment did not globally displace MLL1 from chromatin in any of the three lines; MLL1n loss exceeded 2-fold only at named loci including MEIS1, MEF2C and JMJD1C while HOXA genes, MYB and GAPDH retained occupancy, and CRISPR deletion of MEN1 followed by CUT&RUN identified only 39 genes losing more than 2-fold MLL1n occupancy, 20 of them shared with the drug.
A Menin-MLL Inhibitor Induces Specific Chromatin Changes and Eradicates Disease in Models of MLL-Rearranged Leukemia. · Krivtsov AV, Evans K, Gadrey JY, et al. · Cancer Cell 2019What this evidence cannot show
The ML2 attribution argument rests on cited reports that the line has lost its remaining wild-type KMT2A allele rather than on an allele assay run in this paper. Occupancy is quantified as TSS-anchored window signal rather than by de novo peak calling, and no residual-fraction statistic is given, so 'not globally displaced' is a qualitative statement about tornado plots. Single timepoint (3 days) at a single concentration in the drug arm; all three lines are adult-derived and two of three are myeloid, whereas the index disease is infant B-lineage KMT2A::AFF1 (rule 73). The paper also reports reduced menin and DOT1L protein levels under drug, so for those two factors occupancy loss and protein loss are not separated.
Evidence that argues against this (2)
- The opposite emphasis appears in a murine system: in MLL-AF9 mouse leukaemia cells, MI-503 produced a significant decrease in MLL1 chromatin enrichment alongside the UTX gain, which reads as displacement rather than retention. Species, fusion partner and drug all differ, and the two papers quantify different things (metagene enrichment versus per-gene 2-fold calls), but a reader should not treat retention as settled across systems. Soto-Feliciano YM et al. 2023
- In the one human sample profiled under revumenib, the oncoprotein was reduced to undetectable levels - loss of occupancy AND of protein, not retention. That sample was simultaneously undergoing a lymphoid-to-myeloid lineage switch, itself associated with reduced oncoprotein levels, so it is confounded; but it is the only in-patient observation and it points the other way. Janssens DH et al. 2024
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In KMT2A::AFF1 MV4;11 cells, revumenib at 100 nM produced near-complete and global displacement of menin from chromatin by ChIP-seq, and MLL1 loss was reported specifically from promoters with the highest degree of menin displacement (over 80%) rather than genome-wide, with MEN1 interface mutations blunting both.
MEN1 mutations mediate clinical resistance to menin inhibition. · Perner F, Stein EM, Wenge DV, et al. · Nature 2023What this evidence cannot show
Signal was quantified in TSS +/- 3 kb windows and displayed as torpedo plots; MACS2 peaks were called but no differential peak analysis, and in particular no gained-peak analysis, is reported. The MLL1 antibody recognises wild-type MLL1 as well as the fusion and MV4;11 retains a wild-type KMT2A allele, so the MLL1 lane is not fusion-attributable. Treatment duration for the ChIP-seq arm is not stated in the text or methods; the dose ladder is 0.1, 1 and 5 microM.
Evidence that argues against this (1)
- The paper's own abstract uses the phrase 'inhibitor-induced eviction of menin and MLL1 from chromatin', which is broader than the result in the body and is the sentence most often quoted downstream, including in this graph. The body scopes MLL1 loss to the most menin-depleted promoters. Perner F et al. 2023
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In NUP98-rearranged AML models, VTP-50469 evicts MLL1 and the NUP98 fusion protein itself from chromatin at a critical set of genes essential for the malignant phenotype - the only published report of a fusion oncoprotein being evicted by a menin inhibitor, and it is explicitly locus-restricted.
The menin-MLL1 interaction is a molecular dependency in NUP98-rearranged AML. · Heikamp EB, Henrich JA, Perner F, et al. · Blood 2022What this evidence cannot show
The evicted fusion is a NUP98 fusion, not a KMT2A fusion, and the systems are mouse leukaemia cell lines driven by NUP98-HOXA9 and NUP98-JARID1A plus PDX models. The NUP98 fusion carries no CXXC domain and no menin-binding motif of its own; it depends on menin-MLL1 indirectly, so its eviction geometry need not transfer to a KMT2A fusion that binds menin directly and holds chromatin through CXXC-uCpG and LEDGF as well.
Evidence that argues against this (1)
- Because the NUP98 fusion reaches chromatin through wild-type MLL1 rather than through its own menin contact, its eviction is the expected result and says little about a KMT2A fusion, which has independent CXXC-uCpG and LEDGF tethers. Treating it as evidence that menin inhibitors evict fusions generalises from the case with the fewest alternative tethers. Kerry J et al. 2017
answers that entry's first open question with a number: ~18% residual enrichment, 12.8% top-0.1% membership, in murine FLAG-MLLAF9 - while also showing the loss at the strongest sites is large
re-measures that entry's named locus list in an independent deposit with an epitope-tagged fusion: MEIS1 and JMJD1C reproduce, MEF2C and MYB do not
the existing claim that VTP-50469 does not globally displace the fusion is correct about MLL1-N-reactive material (share ratio 0.7996-0.9069) and is silent about the fusion; this entry supplies the factor needed to read it
the existing claim rests on an MLL1-N antibody and is therefore a claim about MLL1-N-reactive material
the mechanism entry's central observation reproduces in a second line and is measured here as a property of a shared-epitope reagent rather than of the fusion
confirms the menin half of Krivtsov's geometry in an independent deposit: menin goes at HOXA9/HOXA10 and MYB where MLL1n stays
menin absent and MLL1n retained at HOXA9/HOXA10 and MYB is the same parting of the two channels the pivotal paper named
A partial, site-selective H3K4me3 effect is consistent with VTP-50469 not globally displacing the fusion
the clinical compound's preclinical analogue VTP-50469 leaves the fusion on chromatin at most sites
menin is globally displaced in the same experiment in which the fusion is not
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
profiled in RS4;11 and MV4;11 with a shared-epitope antibody; retention observed
- related to →KMT2A::AFDN (MLL-AF6), t(6;11)(q27;q23)
ML2 is the fusion-only attribution control in which all MLL1n signal is this fusion
- related to →KMT2A::MLLT3 (MLL-AF9), t(9;11)(p21.3;q23.3)
MOLM13 arm of the same ChIP-seq experiment
supplies the genetics-based attribution case (ML2) that the audit identified as missing from antibody-based approaches
the residual set that hypothesis predicts is already reported, and it is the majority rather than a subset
that entry's framing presupposes the fusion is evicted; the founding paper reports it is not globally displaced
the measurement behind the vocabulary correction
the named loss set is the exception inside the global retention
places the retention finding in the context of what the instrument could and could not see
would convert a qualitative 'not globally displaced' into a number