Three hours after the fusion is degraded to 2.5-7.1% of its own sites, NOT ONE of the eight complex components profiled beside it falls below 68% retention genome-wide - menin 80.0%, DOT1L 68.7%, CDK9 83.6%, AFF4 92.7%, RNA Pol II 99.6%, pS2 98.3%, pS5 86.8% - so nothing in the complex is held on chromatin by the fusion at genome scale
also called nothing leaves with the fusion · the SEC stays · eight components, none cleared · loss relative to HA never exceeds 0.33
When the leukaemia-driving fusion protein is destroyed you might expect its partners to fall off DNA with it, but across eight partners measured in the same experiment none does - they stay at two-thirds to all of their strongest spots while the fusion itself drops below a tenth. Whatever holds these proteins on DNA across the genome, it is not the fusion.
The same thing, for a scientist
**The panel, 3 h after dTAG-VHL 500 nM, top-decile summit retention against each channel's own DMSO:** | component | role | retained | loss relative to HA | |---|---|---:|---:| | **HA — the fusion** | the thing removed | **2.5% / 7.1%** | 1.00 | | DOT1L | H3K79 methyltransferase | **68.7%** | 0.33 | | MEN1 (menin) | adaptor | 80.0% | 0.21 | | CDK9 | SEC catalytic subunit | 83.6% | 0.17 | | RNA Pol II pS5 | initiation | 86.8% | 0.14 | | AFF4 | SEC scaffold | **92.7%** | 0.08 | | RNA Pol II pS2 | elongation | 98.3% | 0.02 | | RNA Pol II total | — | 99.6% | 0.00 | **Nothing crosses the pre-registered 33% 'cleared' line. Nothing comes within a factor of three of it.** The most-displaced component, DOT1L, loses a third of what the fusion loses; the least, total Pol II, loses nothing measurable. **The interpretive rule fixed in advance was: a component retained when the fusion is gone is not held on chromatin by the fusion.** Applied genome-wide, it returns a clean and slightly deflating answer — **none of them is**. The super elongation complex in particular (AFF4 92.7%, CDK9 83.6%) is essentially where it was. **AND THIS IS A STATEMENT ABOUT GENOME SCALE, NOT ABOUT FUSION TARGET GENES.** Olsen *et al.*, measuring normalised read-count fold change **restricted to their MLL::AF9 target-gene set**, report that AFF4 and CDK9 *"were also specifically lost at promoters of MLL::AF9 target genes ... with more pronounced loss at the 'highly sensitive genes'"*, and that there is **no change at non-MLL::AF9-bound genes** [E1]. Both results are correct. A genome-wide peak-presence measure is dominated by the non-target majority that the authors state does not move, so it is the wrong instrument for the target-restricted question and the right one for the global question. See `concept-q139-a-target-restricted-signal-measure-and-a-genome-wide-peak-measure-are-different-instruments`.
What survives as a claim, stated narrowly
Survives: Removing the fusion does not clear menin, DOT1L, AFF4, CDK9 or RNA Pol II from chromatin genome-wide within three hours. Every one of these proteins has an extensive fusion-independent chromatin life, and that life is the great majority of their peak calls.
Does not survive, and was never tested here: any claim about whether these components leave the fusion's own target loci. The deposited HA peak table would allow that partition — score each component's retention separately inside and outside HA-called regions — and it was not done in this beat. That is the single most valuable follow-up and it needs no new data.
Also does not survive: any quantitative claim. U2 binds throughout — a peak call is a threshold, and a component can lose most of its signal at a site and keep its call.
The one ordering that does carry information
Magnitudes here are not interpretable at n = 1. The ordering is, and it agrees with the authors' orthogonal assay: DOT1L is the most displaced component and menin is less displaced than DOT1L. Olsen et al. state it directly — "DOT1L dissociates from chromatin faster than MENIN upon MLL::AF9 degradation", with ChIP-qPCR showing ~50% DOT1L loss at 15 min and near-complete loss at 30 min, while "MENIN loss was more sluggish, with robust changes only occurring after 3 hours" 1.
That is developed in mechanism-q139-dot1l-leaves-before-menin-and-the-ordering-survives-a-change-of-instrument.
| script | analyses/what-leaves-when-the-fusion-is-degraded/run.py |
|---|---|
| status | run |
| result summary | Top-decile summit retention at 3 h: DOT1L 68.7%, MEN1 80.0%, CDK9 83.6%, pS5 86.8%, AFF4 92.7%, pS2 98.3%, Pol II total 99.6%, against HA 2.5%/7.1%. No component reaches the pre-registered cleared threshold. Loss relative to HA spans 0.00-0.33. |
| preregistration | analyses/what-leaves-when-the-fusion-is-degraded/PREREGISTRATION.md |
inputs
[ "GSE173599 seven protein DMSO/dVHL 3 h pairs plus the HA control pairs" ]
- If the SEC, menin and Pol II remain on chromatin genome-wide after the oncoprotein is gone, then a therapeutic that removes the fusion perfectly still leaves the entire elongation apparatus in place. Whatever therapeutic benefit follows must come from what the fusion was doing at its own loci, not from dismantling the machinery.
- COMPENSATION: components that never left have nothing to re-recruit. A residual pool of fusion - or a related wild-type complex - re-engaging chromatin where AFF4, CDK9 and menin are already sitting is a faster route back than rebuilding a complex from scratch, and is a plausible shape for rebound after drug withdrawal.
- THE OBVIOUS NEXT ANALYSIS, needing no new data: partition each component's peaks by whether they fall inside an HA-called region in the matched DMSO arm, and score retention separately inside and outside. That converts a genome-wide null into the target-restricted measurement the biology actually asks for, using peak tables already downloaded.
- Does any component leave at 24 h or 4 d? Only HA, H3K4me3, H3K27ac, H3K79me2 and H3K9ac have later timepoints; every protein channel stops at 3 h, so the possibility that the complex disassembles slowly is untested.
- AFF4 at 92.7% and CDK9 at 83.6% differ by nine points. With no replicate and two different sequencing batches, is that a real difference in how tightly the two SEC subunits are held, or batch? Nothing here can say.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
AFF4 and CDK9 are specifically lost at promoters of MLL::AF9 target genes 3 h after degradation, with more pronounced loss at highly sensitive genes and no change at non-MLL::AF9-bound genes; RNA Pol II and pSer2 are lost through gene bodies and pSer5 modestly reduced at the TSS of target genes.
MLL::AF9 degradation induces rapid changes in transcriptional elongation and subsequent loss of an active chromatin landscape. · Olsen SN, Godfrey L, Healy JP, Choi YA, Kai Y, Hatton C, Perner F, Haarer EL, Nabet B, Yuan GC, Armstrong SA · Molecular Cell 2022What this evidence cannot show
Read in full text at PMC9044330. The authors' measure is normalised read-count fold change restricted to a pre-defined target-gene set at TSS -1kb/+3kb or gene bodies, compared by Mann-Whitney; it is neither genome-wide nor presence/absence. The authors themselves note cross-reactivity in the pSer2 RNA Pol II antibody (their Figure S4G), which weakens the pS2 arm specifically.
Evidence that argues against this (1)
- This IS the counter-evidence to the genome-wide null, and it comes from the same experiment: at the loci that matter, the SEC components and Pol II do leave. A reader who takes 'nothing leaves with the fusion' as a locus-level statement will reach the opposite of the truth. Olsen SN et al. 2022
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset
Paired DMSO/dTAG-VHL 3 h ChIP peak tables exist in GSE173599 for AFF4, CDK9, DOT1L, MEN1, RNA Pol II, RNA Pol II pS2 and RNA Pol II pS5, each pair within a single sequencing batch.
GSE173599, samples GSM5271643/4 (AFF4), GSM5271645/6 (CDK9), GSM5271647/9 (DOT1L), GSM5271682/4 (MEN1), GSM5271688/9 (RNA Pol II), GSM5271690/1 (pS2), GSM5271692/3 (pS5) · Olsen SN, Armstrong SA, et al. · NCBI Gene Expression Omnibus 2022What this evidence cannot show
n = 1 per arm throughout, no replicates, no spike-in, no noise floor measurable for any of these pairs and none imported. The seven pairs come from four different sequencing batches and some are hg38-mapped while others are not, so cross-component ranking crosses batches and is suggestive only; within-pair comparison is clean. Each channel's retention is scored against its own DMSO, so no cross-channel normalisation is involved.
Evidence that argues against this (1)
- The pre-registered depth-confound rule (X-LIVE) fires on the MEN1 and pS5 arms, whose dVHL libraries are shallower than their DMSO. For a RETENTION finding a shallower treated library biases retention downward, so these two figures are lower bounds - but the rule is recorded as firing rather than waved through. Oncofusion knowledge graph 2026
AFF4 retains 92.7% of its strongest peak calls genome-wide 3 h after fusion degradation - loss relative to the fusion's own 0.08. The SEC scaffold's genome-wide residence does not require the fusion, though the authors report AFF4 loss restricted to fusion target promoters.
CDK9 retains 83.6% of its strongest peak calls genome-wide 3 h after fusion degradation, loss relative to the fusion 0.17 - the SEC catalytic subunit stays where it is at genome scale.
DOT1L retains 68.7% and is the MOST displaced component of the eight profiled - loss relative to the fusion 0.33 - consistent with the authors' report that DOT1L dissociates faster than menin.
P-TEFb's catalytic subunit CDK9 is retained at 83.6% of its strongest sites when the fusion is removed, so P-TEFb's genome-wide chromatin residence is not fusion-dependent at 3 h.
Required to reconcile this genome-wide null with the authors' report that AFF4, CDK9 and Pol II ARE specifically lost at MLL::AF9 target genes. Without the denominator distinction this entry would read as contradicting the paper.
The peak-presence half of the three-instrument comparison: DOT1L 68.7% retained against menin 80.0%, the same rank the authors' ChIP-qPCR and target-restricted read counts give.
- ← related toPeak-call retention tracks peak strength, not fusion occupancy, and that inverts the answer
the genome-wide peak instrument behind that entry is dominated by peak strength; every channel shows the same raw stratum contrast, including ones the authors report as unchanged