A DOT1L inhibitor, which does not touch the menin-KMT2A interface, removes 70% of the fusion's enrichment from its own strongest sites - two thirds of what a menin inhibitor removes - so fusion chromatin residency is partly downstream of H3K79 methylation rather than only of menin
also called Q117 SGC0946 strips the fusion · fusion residency depends on DOT1L
A drug aimed at a completely different enzyme knocks the fusion protein off its favourite DNA sites almost as effectively as the drug designed to do exactly that. Whatever holds the fusion on the chromatin, it is not only its grip on menin.
The same thing, for a scientist
SGC0946 inhibits DOT1L, the H3K79 methyltransferase. It has no described activity at the menin-KMT2A interface. In GSE260742's murine MLLAF9 line, 5 uM for 48 h, measured as a within-sample enrichment that needs no scale factor [E1]: | at the fusion's own top-0.1% bins (n = 1,228) | DMSO | **SGC0946** | VTP50469 | |---|---:|---:|---:| | enrichment over the track's own genome median | 15.9x | **6.1x** | 3.0x | | share of the track's total coverage | 1.975% | **0.589%** | 0.354% | | share ratio vs DMSO | 1.000 | **0.298** | 0.179 | | held bins still in the track's own top 0.1% | 1,228 | **431 (35.1%)** | 157 (12.8%) | | held bins losing within-sample percentile rank | - | **78.9%** | 92.3% | **So the DOT1L inhibitor achieves about 70% of the loss the menin inhibitor achieves**, on the same cells, at the same timepoint, with the same antibody. At the locus level the same ordering holds and the losses are at the same places: Meis1 0.232 (VTP 0.117), Jmjd1c 0.224 (0.167), Tsc22d2 0.095 (0.062), Erbin 0.208 (0.076), against a housekeeping floor of Gapdh 0.659 and Actb 0.836 [E2]. **This is consistent with the depositing paper rather than new to it** - their Fig. 4b box plot reports the logFC of Flag-MLLAF9 under SGC0946 versus VTP50469 with P = 3.9e-9 over 176 regions, so they measured the same thing and found the two drugs differ [E3]. What this entry adds is the **magnitude on an absolute scale-invariant footing**, which a logFC comparison between two treated arms does not give: it says how much of the fusion is left, not which drug removes more. **The mechanistic reading is that fusion chromatin residency is partly a consequence of the transcriptional state the fusion creates.** If H3K79me2 or active elongation stabilises the fusion at its own targets, then removing the mark removes the fusion, and the menin interface is one of at least two things holding it there. That makes the two drugs partially redundant at the level of occupancy while remaining non-redundant at the level of what fills the space - PRC1 in-fill here is uniform under SGC0946 and displacement-coupled under VTP50469.
The alternative reading, and why it is not excluded
SGC0946 at 5 uM for 48 h slows proliferation and begins differentiation in this model. A differentiating cell has a different chromatin landscape and a different FLAG-MLLAF9 expression level, and nothing in this measurement separates 'the fusion was displaced' from 'there is less fusion protein in the cell'. The depositing paper's own immunoblots would settle it and were not examined here.
The same objection applies with less force to VTP50469, which acts on the interface directly, but it applies. A ChIP experiment measures occupancy per unit input chromatin, not occupancy per molecule of the factor, and a drug that reduces the factor's abundance produces an identical readout to a drug that evicts it.
The cheap discriminator is in the deposit: the SGC0946 library is 11% larger in total coverage than DMSO (595.3 M against 537.6 M coverage-bases) while the VTP50469 library is within 0.7%. That is a hint about library composition and not about protein abundance, and it is the reason the share statistic rather than a fold change is used here.
| script | analyses/q117-gained-fusion-occupancy/run117.py + locus117.py |
|---|---|
| status | run |
| result summary | SGC0946 reduces fusion enrichment at its own top sites from 15.9x to 6.1x over the track median (share ratio 0.298), about 70% of the loss and two thirds of the effect of VTP50469 (3.0x, 0.179). |
| drug | SGC0946, DOT1L methyltransferase inhibitor, 5 uM |
| comparator | VTP50469, menin-KMT2A interface inhibitor, 500 nM |
| metadata inconsistency | the FLAG sample records give treatment durations of '3 days' (DMSO, GSM8123359), '72hrs' (SGC0946, GSM8123360) and '48hrs' (VTP50469, GSM8123361), while the paper's legends say 48 h for both drugs |
| replicates | n = 1 per condition |
| shown for fusion | KMT2A::MLLT3 (MLLAF9) in a murine myeloid line, FLAG-tagged |
enrichment over own genome median — the numbers behind the chart
{
"DMSO": 15.9,
"SGC0946": 6.1,
"VTP50469": 3
}share ratio at fusion held bins
{
"SGC0946": 0.298,
"VTP50469": 0.179
}held bins retained in top 0.1pct — the numbers behind the chart
{
"SGC0946": 431,
"VTP50469": 157,
"denominator": 1228
}locus share ratio SGC over DMSO — the numbers behind the chart
{
"Tsc22d2": 0.095,
"Erbin": 0.208,
"Jmjd1c": 0.224,
"Meis1": 0.232,
"Nipbl": 0.247,
"Pds5a": 0.266,
"Epha7": 0.276,
"Six1": 0.32,
"Cdk6": 0.386,
"Hoxa10": 0.554,
"Myb": 0.568,
"Hoxa9": 0.614,
"Gapdh": 0.659,
"Actb": 0.836,
"Mef2c": 1.201
}library totals coverage bases — the numbers behind the chart
{
"FLAG_DMSO": 537618204,
"FLAG_SGC": 595298580,
"FLAG_VTP": 541500899
}- If a DOT1L inhibitor removes most of the fusion from chromatin, then the rationale for combining it with a menin inhibitor as two independent mechanisms is weaker than it looks: at the level of occupancy they overlap by two thirds. The non-overlapping part - uniform versus displacement-coupled PRC1 in-fill - is where a combination benefit would have to come from, and that is a different argument from the one usually made.
- It also complicates attribution in the clinic. A patient on a menin inhibitor whose blasts show reduced fusion occupancy has demonstrated something about their chromatin, not necessarily about target engagement at the menin interface, because at least one other perturbation produces the same readout.
- Pinometostat failed to produce durable responses as a single agent despite doing this to fusion occupancy, if the effect holds at clinical exposure. That is a warning that fusion eviction is not sufficient for clinical benefit, and therefore that an occupancy endpoint may not predict one.
- Is fusion protein abundance unchanged at these timepoints? A single immunoblot decides whether this is displacement or depletion, and the distinction changes the interpretation completely.
- Does the fusion's loss under SGC0946 track the H3K79me2 loss locus by locus? Both arms are in this series and the H3K79me2 samples were not downloaded; a per-locus correlation would test the proposed feedback directly.
- Is the same true for a clinical DOT1L inhibitor at a clinical exposure? Pinometostat's trial exposures are far below 5 uM for 48-72 h, and an occupancy effect that needs 5 uM is not obviously a property of the drug class as used.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
In murine MLLAF9 cells, SGC0946 at 5 uM for 48 h reduces FLAG-MLLAF9 enrichment at its own top-0.1% bins from 15.9-fold to 6.1-fold over the track's genome median, a share ratio of 0.298, with 431 of 1,228 held bins remaining in the top 0.1% - against 3.0-fold, 0.179 and 157 for VTP50469.
NCBI GEO GSE260742, GSM8123359 / GSM8123360 / GSM8123361 (MLLAF9 FLAG ChIP-seq, DMSO / SGC0946 5 uM / VTP50469 500 nM), re-binned to 2 kb from the deposited igvtools .tdf · NCBI Gene Expression Omnibus 2026What this evidence cannot show
n = 1 per condition. The held set is defined by the DMSO sample itself, so regression to the mean contributes an unknown share of every fall, including this one. Occupancy per unit chromatin is not occupancy per molecule: reduced fusion protein abundance would produce the same readout and was not measured here. The SGC0946 library is 11% larger in total coverage than DMSO, which the share statistic absorbs but which a fold change would not have. Timepoint note: the SGC0946 FLAG sample's record says 'treatment: 72hrs' while the VTP50469 sample says '48hrs' and the DMSO sample says '3 days' - the sample-level metadata is not internally consistent with the paper's stated 48 h, and this comparison inherits that ambiguity.
Evidence that argues against this (1)
- SGC0946 is a much longer exposure at a much higher concentration relative to its cellular potency than VTP50469 at 500 nM, and 5 uM DOT1L inhibition for two to three days produces substantial differentiation in MLL-rearranged models. A loss of fusion occupancy in a differentiating cell is not obviously the same phenomenon as a loss in a cell whose interface has just been blocked, and reading the two as comparable perturbations is the weakest step in this entry. Nature Cell Biology 2026
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
At the locus level the SGC0946 losses fall at the same places as the VTP50469 losses and are consistently smaller: Meis1 0.232 against 0.117, Jmjd1c 0.224 against 0.167, Tsc22d2 0.095 against 0.062, with a housekeeping floor of Gapdh 0.659 and Actb 0.836.
NCBI GEO GSE260742 locus table over UCSC mm10 refGene windows, analyses/q117-gained-fusion-occupancy/locus-table.json · NCBI Gene Expression Omnibus 2026What this evidence cannot show
Twenty-four hand-picked loci, n = 1 per condition, gene body minus 5 kb / plus 2 kb windows from the live UCSC mm10 refGene track. The housekeeping floor under SGC0946 (0.659, 0.836) is notably higher and less consistent than under VTP50469 (0.508, 0.571), which weakens the floor as a reference in this arm specifically.
Evidence that argues against this (1)
- That the two drugs hit the same loci is expected under any model in which those loci are simply the fusion's strongest and therefore have the most to lose. A shared rank ordering is not evidence of a shared mechanism, and the entry's mechanistic reading needs the H3K79me2 data - in the same series, not downloaded - to become more than a hypothesis. NCBI Gene Expression Omnibus 2026
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The depositing paper reports a box plot of the logFC of Flag-MLLAF9 under SGC0946 versus VTP50469 relative to DMSO with P = 3.9e-9 over 176 regions, so it measured fusion occupancy change under the DOT1L inhibitor and found the two drugs differ.
DOT1L provides transcriptional memory through PRC1.1 antagonism. · Neville D, Ferguson DT, Heikamp EB, et al. · Nature Cell Biology 2026What this evidence cannot show
Read from the figure legend in the retrieved fullTextXML on 2026-09-23. The legend reports a comparison BETWEEN the two treated arms; it does not state how much fusion occupancy remains in either. The 176 regions are those with baseline H3K79me2 RPM > 10 assigned to genes downregulated by both drugs, which is a different and narrower region set than the top-0.1% FLAG bins used here.
Evidence that argues against this (1)
- Because the paper's region set is conditioned on high baseline H3K79me2, it is selected for exactly the loci where a DOT1L inhibitor is expected to act, and its comparison is therefore not a neutral test of whether SGC0946 displaces the fusion. This entry's region set is conditioned on FLAG occupancy instead, which is neutral with respect to the DOT1L axis but selected for the fusion - neither set is unconditioned. Nature Cell Biology 2026
SGC0946 inhibits DOT1L; at 5 uM for 48 h it removes about 70 percent of the fusion's enrichment from its own top sites
- related to →Pinometostat (EPZ-5676) — the DOT1L inhibitor, and the most instructive failure in this field
the clinical DOT1L inhibitor; whether this occupancy effect holds at clinical exposure is unknown
the same held-set statistic, the other drug arm
- related to →KMT2A::MLLT3 (MLL-AF9), t(9;11)(p21.3;q23.3)
the fusion measured is FLAG-tagged MLLAF9 (KMT2A::MLLT3)
same shape of question on a different channel: Q117 measured the FLAG-tagged fusion in murine MLL-AF9, this measures MEN1 in human MOLM13