Census of every published chromatin-occupancy measurement made under a menin inhibitor: eight datasets, every one TSS-anchored, every one tabulating losses only
also called menin inhibitor occupancy census · what has been profiled under menin inhibition · the loss-only instrument
We listed every experiment anyone has published that maps where proteins sit on DNA after a menin inhibitor is given. There are only about eight, they all measure signal in a window around the start of genes, and every one of them reports a table of what was lost. None reports a table of what was gained.
The same thing, for a scientist
Enumerated from a bounded PubMed query (61 records, querytranslation recorded in `data`) plus the full texts of the pivotal papers. **What exists.** Eight distinct occupancy datasets under a pharmacological menin inhibitor, listed in `data.datasets`. They cover menin, MLL1n, DOT1L, H3K79me2, H3K4me3, LEDGF, UTX, MLL3/4, MOF, H3K27ac, H4K16ac, IKAROS, JUND and H2AK119ub. **What is uniform across all of them.** (i) Quantification is **anchored on transcription start sites** - TSS +/- 3 kb torpedo plots (Perner [E2], Soto-Feliciano TSS +/- 2 kb [E3]), or TSS-to-10-kb-into-gene-body windows (Krivtsov [E1]). (ii) The supplementary tables shipped are **lists of genes that LOSE occupancy**; Krivtsov's Table S4 is literally *list of genes that lose occupancy >2-fold of MLL1 and Menin* [E1]. (iii) Where a per-gene threshold is used it is a **2-fold loss** call, which is one-sided by construction. (iv) Timepoints are **late**: 3 days for the ChIP-seq, 6 days for the genetic arm, 4-6 days in the murine work [E1][E3]. There is **no sub-24-hour chromatin timepoint anywhere in the class**, so nothing in this literature can see an early redistribution that is subsequently resolved by transcriptional collapse and differentiation. (v) **Nobody spike-in normalises the fusion lane.** Krivtsov spike in Drosophila chromatin for CUT&RUN only, not for the ChIP-seq [E1]. Without calibration, a genuine global loss and a normalisation artefact are the same picture, and so are a global gain and nothing. **What this census licenses.** Strong conclusions about menin (global loss, dose-resolved, replicated across laboratories and lines). Scoped conclusions about MLL1 at promoters. **No conclusion at all about where a KMT2A fusion protein goes**, because the question was never asked of the data: the analysis was defined as differential signal at annotated TSSs, and gains were tabulated only for proteins the authors expected to gain (UTX, JUND).
The datasets
| study | drug | profiled | system | fusion | timepoint | analysis window | gains tabulated? |
|---|---|---|---|---|---|---|---|
| Krivtsov 2019 1 | VTP-50469 330 nM | menin, MLL1n, DOT1L, H3K79me2 (ChIP-seq); MLL1n (CUT&RUN, MEN1 KO) | MOLM13, RS4;11, ML2 | MLLT3, AFF1, AFDN (fusion-only line) | 3 d (drug), 6 d (KO) | TSS to +10 kb, 100 bp tiles | no - Table S4 is losses |
| Perner 2023 2 | revumenib 0.1/1/5 microM | menin, MLL1 | MV4;11, OCI-AML3, PDX3 in vivo | AFF1; NPM1c | not stated | TSS +/- 3 kb | no |
| Soto-Feliciano 2023 3 | MI-503 | menin, UTX, MLL1, MLL3/4, MOF, H3K4me1/3, H3K27ac, H4K16ac | murine MLL-AF9, human lines | MLLT3 (mouse) | 4-6 d | TSS +/- 2 kb | yes, for UTX (>13,000 new sites) |
| Aubrey 2022 4 | menin inhibitor + IMiD | IKAROS, MLL1-menin, MEIS1 | MLL-r AML | various | not stated | promoter-centric | no |
| Janssens 2021 5 | menin inhibitor | H3K4me3, H3K27me3, oncofusion targets (automated CUT&Tag, single-cell) | cell lines + patient samples | various | not stated | gene body / peak | no |
| Janssens 2024 6 | revumenib (in patient) | oncoprotein, menin, ENL (CUT&RUN) | 36 patient samples, 1 on revumenib | various | clinical course | peak-based | no - but reports ENL persistence, a retention |
| Ozyerli-Goknar 2026 7 | menin-MLL inhibitors | JUND (chromatin binding assays) | MLLr AML/ALL | various | not stated | target-gene | yes, for JUND (increased) |
| Tayari 2025 preprint 8 | SNDX-5613 +/- ORY-1001 | MLL, LEDGF, menin (CUT&RUN) | MOLM-13, MV4-11 | MLLT3, AFF1 | not stated | metaplots, tracks | no |
Two of eight tabulate gains, and both do it for a protein other than the fusion. That is the census's single most useful line: the assay can see gains, the analysis pipeline can call them, and nobody has pointed either at the oncoprotein.
What a TSS-anchored analysis can and cannot miss
CpG islands are predominantly promoter-associated, so a relocation of the fusion onto previously unused unmethylated CpG would land mostly at other genes' promoters - and would therefore be inside a genome-wide TSS +/- 3 kb window set. So the window choice is not fatal on its own.
What is fatal is the statistic: every study quantifies signal at a pre-specified target-gene list or plots all TSSs as a heat map ordered by control signal. A tornado plot ordered by DMSO signal shows you what happened to the sites that were occupied. It cannot show you sites that were empty and became occupied, because those rows are at the bottom of the plot and compressed to nothing. Gains require either de novo peak calling in the treated sample or a differential test over a union peak set, and neither was run for MLL1n in any dataset in the table. Krivtsov did call MACS2 peaks for the CUT&RUN arm and Perner called MACS2 peaks with input controls - so the peaks exist in both pipelines and the comparison was simply not made.
The gene-body uCpG sites that Kerry showed the fusion spreads into are the exception: those are 4 kb or more into the gene and would fall outside a TSS +/- 3 kb window entirely.
Bounded rejections - what I read and did not use
Per the standing rule on recording rejections:
- PMID 42449441 (J Exp Clin Cancer Res 2026), the single hit for the menin-inhibitor name list AND
redistribut*[tiab]: a hyaluronic-acid liposome/mesoporous-silica nanoplatform co-delivering BMF-219 and 5-FU in gastric cancer. Redistribution refers to drug biodistribution. Off-beat. - PMID 34582559 (Heikamp, NUP98) is on-beat and is cited in
mechanism-vtp-50469-does-not-globally-displace-the-fusion-from-chromatin, not here, because it is not a KMT2A-fusion occupancy dataset. - The five
de novo-matching records (41347983, 40196700, 40055116, 36944324, 29672864) were title-checked; all use de novo in its clinical sense (de novo AML) or for de novo mutation, none for de novo peaks. - PMID 38234854 (bioRxiv 2024, KMT2A oncoproteins induce epigenetic resistance to targeted therapies) is a preprint in the 61-record set that plausibly bears on this question and was not read in full for this beat for time reasons. It is the highest-priority item for whoever picks this up: if a gained-peak analysis exists anywhere, that is the likeliest place.
| script | none - bounded PubMed enumeration plus full-text reading, 2026-09-21 |
|---|---|
| status | run |
| query | (revumenib[tiab] OR ziftomenib[tiab] OR bleximenib[tiab] OR enzomenib[tiab] OR VTP-50469[tiab] OR SNDX-5613[tiab] OR KO-539[tiab] OR MI-503[tiab] OR MI-3454[tiab] OR MI-2-2[tiab] OR MI-463[tiab] OR MI-538[tiab] OR JNJ-75276617[tiab] OR DSP-5336[tiab] OR DS-1594[tiab] OR BMF-219[tiab] OR "menin inhibitor"[tiab] OR "menin inhibitors"[tiab] OR "menin inhibition"[tiab] OR "menin-MLL inhibitor"[tiab] OR "menin-MLL inhibitors"[tiab]) AND chromatin[tiab] |
| records | 61 |
| n datasets | 8 |
| n tabulating gains | 2 |
| earliest chromatin timepoint in class | 3 days |
| spike in normalised fusion lane | none |
| shown for fusion | MLL1n profiled under drug in RS4;11 (KMT2A::AFF1), MOLM13 (KMT2A::MLLT3), ML2 (KMT2A::AFDN, fusion-only) and MV4;11 (KMT2A::AFF1); no infant-derived line anywhere in the census |
inputs
[ "PubMed esearch, 2026-09-21", "Europe PMC / PMC full texts of PMC5263239, PMC7227117, PMC10157896, PMC9827117, PMC6726985" ]
subqueries — the numbers behind the chart
{
"AND \"ChIP-seq\"[tiab]": 2,
"AND \"CUT&RUN\"[tiab]": 2,
"AND \"CUT&Tag\"[tiab]": 2,
"AND occupancy[tiab]": 13,
"AND genome-wide[tiab]": 7,
"AND evict*[tiab]": 4,
"AND displace*[tiab]": 5,
"AND redistribut*[tiab]": 1,
"AND relocaliz*[tiab]": 0,
"AND retarget*[tiab]": 0,
"AND \"new binding sites\"[tiab]": 0,
"AND \"gain of occupancy\"[tiab]": 0
}gains tabulated for
[ "UTX/KDM6A", "JUND" ]
datasets
[ "Krivtsov 2019 PMID 31821784", "Perner 2023 PMID 36922589", "Soto-Feliciano 2023 PMID 36264143", "Aubrey 2022 PMID 35534777", "Janssens 2021 PMID 34663924", "Janssens 2024 PMID 39472576", "Ozyerli-Goknar 2026 PMID 42371796", "Tayari 2025 preprint PMID 41279447" ]
- A census that finds no gain analysis is not proof that no gain occurred; it is a statement about what the field asked. Any intervention argument built on 'the fusion is gone' should be treated as unsupported rather than refuted.
- Because every dataset is depth-normalised rather than spike-in calibrated, the class's headline pharmacodynamic claim - global menin loss - is quantitatively uncertain in the direction that flatters the drug, and a correlative endpoint built on it could be misleading in a paediatric trial where sample volume forbids repeating the assay.
- Two of eight studies looked for gains and both found one. The base rate for finding a gain when you look is, so far, two out of two.
- What is the residual fusion occupancy fraction at saturating drug? Every dataset in this census could report it and none does, and it is the number a pharmacodynamic endpoint would need.
- Does anything change in the first six hours? The earliest chromatin timepoint in the entire class is 3 days, by which time differentiation has begun and any transient redistribution has been overwritten.
- Would spike-in-calibrated profiling change the menin result too? Every 'global loss' in this census is reported from read-depth-normalised data, in which a true global loss is partly rescaled away - so the menin result may be an underestimate and the MLL1 retention may be an overestimate in the same direction.
- Does the bioRxiv preprint PMID 38234854 on KMT2A oncoprotein-induced epigenetic resistance contain a gained-peak analysis? Not read for this beat; it is the most likely place for a counter-example.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Krivtsov et al. profiled menin, MLL1n, DOT1L and H3K79me2 by ChIP-seq after 3 days of VTP-50469 in MOLM13, RS4;11 and ML2, and MLL1n by CUT&RUN 6 days after MEN1 CRISPR deletion in MOLM13; signal was quantified in windows from the TSS extending 10 kb into the gene body with 100 bp tiling, and the supplementary tables report gene occupancy and a list of genes that lose occupancy more than 2-fold, with no corresponding gain list.
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 methods and supplementary-table titles were read from the full text of PMC7227117; the underlying tables themselves were not downloaded, so a gain column inside Table S2 cannot be excluded from the text alone - the paper's prose and the Table S4 title both describe losses. One drug concentration, one timepoint for the drug arm, adult-derived lines, two of three myeloid.
Evidence that argues against this (1)
- Raw and processed data are public in GEO GSE127508, so the absence of a gain analysis in the paper is not an absence of data - the gain question is answerable by reanalysis and is costed in ana-reanalysis-of-ml2-mll1n-chip-for-gained-peaks-under-vtp-50469. Krivtsov AV et al. 2019
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Perner et al. assessed genome-wide chromatin occupancy of menin and MLL1 by ChIP-seq in MV4;11 and OCI-AML3 under a revumenib dose ladder, quantifying signal in TSS +/- 3 kb windows with MACS2 peak calling against input, and reported losses only.
MEN1 mutations mediate clinical resistance to menin inhibition. · Perner F, Stein EM, Wenge DV, et al. · Nature 2023What this evidence cannot show
Read from the full text of PMC10157896. Treatment duration for the ChIP-seq arm is not stated. Shared-epitope MLL1 antibody in lines that retain a wild-type KMT2A allele. The reported statistic is a Mann-Whitney comparison of TSS occupancy distributions, which is inherently two-sided but was applied to target-gene sets rather than to a genome-wide union peak set.
Evidence that argues against this (1)
- Because MACS2 peaks were called in every condition, this dataset could in principle be re-interrogated for gains without new sequencing; the absence is analytic, not experimental, which is a weaker absence than a missing assay. Perner F et al. 2023
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Soto-Feliciano et al. showed that MI-503 treatment of MLL-AF9 leukaemia cells produced a substantial increase in UTX chromatin association with more than 13,000 additional binding sites beyond those seen at baseline, predominantly at promoters, and that this was not explained by increased UTX expression or protein stability - demonstrating that the standard ChIP-seq pipeline in this field detects, calls and quantifies gained sites when it is asked to.
A Molecular Switch between Mammalian MLL Complexes Dictates Response to Menin-MLL Inhibition. · Soto-Feliciano YM, Sánchez-Rivera FJ, Perner F, et al. · Cancer Discov 2023What this evidence cannot show
The gain is for UTX in a murine MLL-AF9 system treated with the tool compound MI-503, not for a KMT2A fusion protein and not with a clinical-stage inhibitor. Promoter-centric analysis (TSS +/- 2 kb). The 13,000 figure comes from supplementary panels described in the body text.
Evidence that argues against this (1)
- A gain detected for UTX does not guarantee a gain would be detectable for the fusion at the same sensitivity: UTX starts from very few peaks so the dynamic range for a gain is large, whereas the fusion starts from several thousand peaks and a redistribution of a fraction of them would be a smaller relative change against a higher background. Kerry J et al. 2017
- E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
IKAROS was shown to co-occupy chromatin extensively with MLL1-MENIN and MEIS1 in KMT2A-rearranged AML, and combined IKAROS degradation with MENIN inhibition disrupted the leukaemogenic transcriptional network - a promoter-centric occupancy study under menin inhibition whose reported output is loss of the network.
IKAROS and MENIN coordinate therapeutically actionable leukemogenic gene expression in MLL-r acute myeloid leukemia. · Aubrey BJ, Cutler JA, Bourgeois W, et al. · Nat Cancer 2022What this evidence cannot show
Abstract-level reading only for this census entry; the full text was not read, so the analysis window and any gain analysis are inferred from the abstract's description and could be wrong. Included in the census for completeness of the enumeration rather than as a source for the loss-only claim.
Evidence that argues against this (1)
- An abstract-level entry in a census of analysis methods is exactly the reading depth that rule 80 warns about; if this study does contain a gained-site analysis it would weaken the census's uniformity claim, and the claim should be re-checked by anyone acting on it. Soto-Feliciano YM et al. 2023
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Automated and single-cell CUT&Tag mapping of KMT2A oncofusion target sites found that aberrant enrichment of H3K4me3 in gene bodies is sensitive to menin inhibitors, establishing CUT&Tag as a working chromatin readout under the drug in cell lines and patient samples.
Automated CUT&Tag profiling of chromatin heterogeneity in mixed-lineage leukemia. · Janssens DH, Meers MP, Wu SJ, et al. · Nat Genet 2021What this evidence cannot show
Abstract-level for the menin-inhibitor arm. The menin-inhibitor sensitivity reported is of a histone mark, not of oncoprotein occupancy, and no timepoint or concentration is given in the abstract.
Evidence that argues against this (1)
- This paper's oncofusion-target mapping is the methodological ancestor of the 2024 patient study; if any dataset in the field is positioned to answer the gain question for the oncoprotein it is this platform, which makes its absence from the gain literature a choice of question rather than a technical limit. Janssens DH et al. 2024
- E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In the only human sample profiled under revumenib, oncoprotein and menin were reduced to undetectable levels while ENL persisted on numerous oncoprotein-target loci - the single in-patient chromatin observation in this literature, and it reports a persistence rather than a gain.
MLL oncoprotein levels influence leukemia lineage identities. · Janssens DH, Duran M, Otto DJ, et al. · Nat Commun 2024What this evidence cannot show
n = 1 sample, confounded by a concurrent lymphoid-to-myeloid lineage switch that is itself associated with reduced oncoprotein levels. Undetectable oncoprotein signal in a CUT&RUN conflates loss of chromatin binding with loss of protein. No pre-treatment paired sample under drug at an early timepoint.
Evidence that argues against this (1)
- ENL persistence at loci the oncoprotein has vacated is itself evidence that a chromatin platform can be maintained without the fusion, which argues against reading residual target-gene activity as residual fusion occupancy. Krivtsov AV et al. 2019
- E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Menin-MLL inhibitor treatment increased JUND chromatin occupancy, upregulating JUND target genes and contributing to resistance - a reported gain of occupancy under the drug, for a transcription factor that competes for the same menin pocket.
Menin-MLL inhibitors enhance JUND activity in MLLr leukemic cells contributing to tumorigenesis and therapy resistance. · Ozyerli-Goknar E, Nizamuddin S, Fetsch V, et al. · Blood 2026What this evidence cannot show
Chromatin binding assays are described at abstract level without a stated genome-wide differential method, cells, concentration or timepoint; the quantitative mass spectrometry arm is the paper's primary evidence. JUND binds menin through a motif in the MLL1/2 pocket, so its liberation is a direct consequence of pocket occupancy and does not generalise to proteins that reach chromatin by other routes.
Evidence that argues against this (1)
- JUND gaining chromatin is the expected consequence of freeing it from a menin complex, so it is a weaker demonstration that the drug reshuffles chromatin than the UTX result, where the gaining complex has no direct competition for the drug pocket. Soto-Feliciano YM et al. 2023
- E8weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.preprint
A 2025 preprint profiled MLL, LEDGF and menin by CUT&RUN in MOLM-13 and MV4-11 under SNDX-5613 and reported decreased MLL and LEDGF chromatin binding at target loci from metaplots and representative tracks, with no quantified occupancy change, no dose-response and no gain analysis.
Co-targeting menin and LSD1 dismantles oncogenic programs and restores differentiation in MLL-rearranged AML. · Tayari MM, Dos Santos HG, Beckedorff F, et al. · bioRxiv 2025What this evidence cannot show
Preprint, not peer reviewed. The menin arm is a comparator within an LSD1-combination study, and the LSD1 inhibitor alone reduced MLL and LEDGF occupancy comparably, so the menin arm is not the study's focus. Reported from metaplots and tracks rather than from a differential statistic.
Evidence that argues against this (1)
- This preprint's direction - MLL occupancy decreased - is in tension with Krivtsov's 'not globally displaced' in two of the same cell backgrounds. Metaplots at target loci and genome-wide tornado plots are different measurements and can both be right, which is itself an argument that the field needs a residual-fraction statistic rather than more pictures. Krivtsov AV et al. 2019
enumerates every chromatin measurement made under this drug and its class
included in the 21-term name list bounding the enumeration
included in the 21-term name list bounding the enumeration
the enumeration is the denominator for the bounded absence
the same enumeration supplies the system list
places the retention finding in the context of what the instrument could and could not see
the LEDGF preprint is dataset 8 of 8 in this census