A chromatin-occupancy readout has one dimension and the question has three: eviction, degradation and relocation all lower the signal at a target gene
also called the three-way confound · occupancy is not a fate measurement · why ChIP cannot answer where it went
Measuring how much of a protein sits on a particular gene tells you the amount there went down. It cannot tell you whether the protein moved away, was destroyed, or moved somewhere else on the DNA. Every pivotal measurement in this drug class is that one kind of measurement.
The same thing, for a scientist
Every pivotal readout for menin inhibitors is chromatin occupancy at target genes, and a fall in that number is produced identically by four distinct events. **(1) Eviction** - the protein comes off and enters the soluble pool. **(2) Degradation** - the protein is destroyed, and this is not hypothetical for this drug class: menin inhibitors shorten menin's half-life without changing MEN1 mRNA, via inhibitor-promoted menin-CHIP/Hsp70 ubiquitination, rescuable by proteasome or E1 inhibition [E1], and the pivotal paper independently observed reduced menin and DOT1L protein under drug [E2]. **(3) Relocation** - the protein moves to sites outside the measured window or outside the target-gene list. **(4) Normalisation artefact** - in depth-normalised data with no spike-in, a uniform global change is rescaled to nothing and a genuine global loss is partly rescaled away. **Three measurements disambiguate all four, and none is routinely made.** A **spike-in calibrated total** separates (4) from the rest. A **protein-level time course with proteasome rescue** separates (2). A **genome-wide gain analysis** separates (3). Each is cheap; none appears in the class [E3]. The consequence is that the field's central mechanistic statement - the drug removes the oncoprotein from chromatin - rests on a measurement that cannot distinguish removal from three other things, one of which (degradation) is documented for the drug's own target. **This is not a criticism of the occupancy assay.** ChIP and CUT&RUN answer *is the protein at this locus* extremely well, which is the right question for target engagement at a known target. It is a criticism of using that answer to support a claim about fate, which is a different question, and of a literature that describes the result with fate verbs.
Which readout answers which question
| question | readout that answers it | present in the class? |
|---|---|---|
| is the protein at this locus? | ChIP/CUT&RUN at that locus | yes, extensively |
| how much protein is bound in total? | spike-in calibrated genome-wide signal | no - spike-in used only in one CUT&RUN arm |
| is the protein still in the cell? | western, cycloheximide chase | no, for the fusion; yes, for menin |
| is the loss proteasome-dependent? | bortezomib or MG132 rescue | no, for the fusion; yes, for menin |
| did it move? | genome-wide differential including gains | no, for the fusion |
| is it still working where it is? | nascent transcription at retained sites | no - RNA-seq is steady state and at different timepoints |
| where is it in the nucleus? | fractionation or imaging of a tagged fusion | no - recorded as an open question in localization-menin-inhibition-fusion-fate |
Seven questions, one of them answered.
| shown for fusion | no - none of the disambiguating measurements has been made for a KMT2A fusion under a menin inhibitor |
|---|
confounded events
[ "eviction into the soluble pool", "proteasomal degradation", "relocation outside the measured window", "depth-normalisation rescaling of a global change" ]
disambiguating measurements
[ "spike-in calibrated genome-wide total", "protein level with cycloheximide chase and proteasome rescue", "genome-wide differential analysis including gains", "nascent transcription co-measured at retained sites" ]
measured for menin
[ "protein level", "half-life", "proteasome rescue", "global occupancy" ]
measured for the fusion
[ "occupancy at target genes" ]
- If part of the drug's action is degradation of menin rather than interface blockade, then resistance could arise through the degradation machinery - STUB1, HSPA - with a wild-type MEN1 sequence, and nobody sequences those genes in relapse samples.
- A pharmacodynamic assay that cannot distinguish target engagement from target destruction will give the same answer in two clinical situations that call for different next steps.
- The practical warning is narrow and should not be overstated: the occupancy data in this class are good data, correctly reported. It is the fate verbs attached to them that outrun the measurement.
- Is the KMT2A fusion protein itself destabilised by menin inhibitors? One western time course with a cycloheximide chase answers it, and the answer would reinterpret every occupancy result in the class.
- Would spike-in calibration change the menin conclusion as well as the fusion one? The global menin loss is reported from depth-normalised data, in which a global loss is systematically underestimated.
- Can a soluble-versus-chromatin fractionation under drug substitute for imaging? It is cheaper than a tagged line and would separate eviction from degradation on its own, and this graph already proposes a five-fraction blot for exactly this.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Menin inhibitors reduce menin protein levels and shorten menin protein half-life in MLL-fusion-expressing leukaemia cells with no change in MEN1 mRNA; proteasome or E1 ligase inhibition rescues the degradation, and mechanistically the inhibitors promote menin's interaction with the Hsp70-associated ubiquitin ligase CHIP with increased menin ubiquitination - and proteasome inhibition also rescued the drug-induced reduction in H3K4me3 and HOXA9 expression.
Disruption of the menin-MLL interaction triggers menin protein degradation via ubiquitin-proteasome pathway. · Wu Y, Doepner M, Hojnacki T, et al. · Am J Cancer Res 2019What this evidence cannot show
MI-503 in MV4;11 and THP-1; a tool compound rather than a clinical agent, and the mRNA and protein comparisons are western and qPCR rather than quantitative proteomics. Am J Cancer Res carries no DOI on the PubMed record; the article is PMC6726985. The equivalent measurement has never been made for a KMT2A fusion protein.
Evidence that argues against this (1)
- That proteasome inhibition rescues the drug's transcriptional effect implies degradation is part of the mechanism of action, not a side effect - which if it generalises to the clinical compounds means 'interface inhibitor' understates what these drugs do, and means occupancy loss is downstream of protein loss rather than a separate event. Krivtsov AV et al. 2019
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The pivotal VTP-50469 study independently observed a consistent decrease in DOT1L and menin protein levels under drug treatment, which the authors note is consistent with their ChIP-seq results - so in that dataset occupancy loss and protein loss are confounded for two of the four profiled factors.
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
Observed by western blot (Figure S3E) over up to 9 days; no cycloheximide chase and no proteasome rescue in that paper, so the mechanism of the protein decrease is not established there. No corresponding measurement of the fusion protein's level is reported.
Evidence that argues against this (1)
- A decrease in menin protein under drug is expected from the CHIP/Hsp70 degradation mechanism and does not by itself impugn the occupancy result; the point is that the two cannot be separated in a ChIP experiment, not that the ChIP is wrong. Wu Y et al. 2019
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
No study in the class reports a spike-in calibrated fusion occupancy total, a fusion protein-level time course under drug, or a genome-wide gain analysis for the fusion; the earliest chromatin timepoint anywhere in the class is 3 days.
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
This is the census conclusion from ana-what-menin-inhibition-has-actually-been-measured-on-chromatin, bounded by a 61-record PubMed query plus five full texts; two of the eight datasets were read at abstract level only, and one preprint in the set was not read at all.
Evidence that argues against this (1)
- Spike-in normalisation was used in the CUT&RUN arm of the pivotal study (Drosophila chromatin), so the technique is in the field's hands and its absence from the ChIP-seq is a choice rather than a capability gap; a reanalysis using the spiked CUT&RUN arm could partially recover a calibrated total for the genetic perturbation. Krivtsov AV et al. 2019
retained occupancy at HOXA9 says nothing about whether HOXA9 is transcribed, and the depositing paper reports its expression falling
a second axis on which an occupancy measurement is underdetermined: not only eviction versus degradation versus relocation, but which molecule was measured
a share statistic is blind to a uniform global loss, so 'MLL1-N stays' is a claim about distribution and not amount
Adds a fourth thing occupancy cannot separate, and it is instrumental rather than biological: a 1 kb bin-overlap retention figure cannot separate a protein still being at a site from MACS calling a wider, blurrier peak where the signal collapsed.
Same family of limitation, one layer up: that entry is about what an occupancy measurement cannot distinguish; this is about two occupancy measurements of the same data not being the same quantity at all.
the vocabulary outruns the measurement because the measurement has one dimension
designed specifically around the three-way confound that entry describes
menin's own degradation under the drug is the documented instance of the confound
supplies the measurement-theoretic reason that hypothesis cannot be settled by occupancy data
formalises why that entry's question is unanswered by the existing readouts
applies to every pharmacodynamic inference made about this drug class
the surrogate failure modes follow from the measurement's single dimension