Occupancy-matched three-arm comparison of ziftomenib, revumenib and bleximenib on gamma-H2A.X resolution and HR efficiency in a KMT2A-rearranged B-lineage line, with the peptide competition as the biochemical pre-test
also called does the menin inhibitor class split on DNA repair · V371 contact experiment
Three menin drugs touch a menin amino acid that starts the protein's DNA-damage recognition patch and a fourth does not, so this experiment gives the same cells the same amount of each drug and asks whether DNA-break repair suffers differently. The drug that does not touch that spot is the built-in control, which is what makes the comparison interpretable at all.
The same thing, for a scientist
Two stages, ordered so the cheap one can kill the expensive one. **Stage 1, biochemical, one week.** Biotinylated Pep-MH (menin 371-400, sequence published by Lian et al.) pulldown of recombinant menin, pre-incubated with DMSO, ziftomenib, revumenib, DS-1594b or bleximenib at 1 and 10 uM, with the alanine-substituted Pep-Mut as the peptide control and an MBM1 peptide pulldown as the compound-activity control. Read by western densitometry, three independent pulldowns. **Stage 2, cellular, six weeks, only if stage 1 is positive.** RS4;11 and SEM (KMT2A::AFF1, human, B-lineage - the index fusion, not a myeloid surrogate), compound concentrations titrated so that HOXA9 and MEIS1 mRNA suppression at 96 h is within 20% across arms, which is the occupancy match; then (a) gamma-H2A.X foci at 0.5, 2, 6 and 24 h after 2 Gy, scored as the fraction of cells with more than 10 foci, 200 cells per arm per replicate; (b) HR and NHEJ efficiency by a DR-GFP / EJ5-GFP pair; (c) a menin-chromatin control - CUT&RUN for menin at HOXA9 - to confirm the arms are matched on the thing they are supposed to be matched on. **The class is its own control**, which is the design's whole value: any difference between bleximenib and the other three, at matched HOXA9 suppression, localises to what those three do and bleximenib does not.
Why occupancy-matched and not dose-matched
The commonest way this design fails is that the arms are matched on concentration rather than on target engagement, and a compound that is simply less potent then looks like a compound that is mechanistically different. Menin-inhibitor potencies differ across the class and the graph does not hold a clean head-to-head.
So the matching variable is a downstream pharmacodynamic readout the class is known to share - HOXA9 and MEIS1 suppression - and it is measured in the same wells, in the same experiment, as the DNA-repair readout. If the arms cannot be matched to within 20% on HOXA9, the experiment reports that and stops; an unmatched three-arm comparison is not worth running.
A second control that is not optional: menin occupancy at HOXA9 by CUT&RUN. HOXA9 mRNA is two steps downstream and can be matched by accident.
What each possible outcome means
| stage 1 | stage 2 | reading |
|---|---|---|
| ziftomenib/revumenib compete Pep-MH, bleximenib does not | repair phenotype splits the same way | the hypothesis holds; the class is not uniform and compound choice is a DNA-repair decision |
| no compound competes Pep-MH | - | stop. The rim contact does not occlude the motif and the cellular prediction has no mechanism. Record as REFUTED inside the falsifier. |
| compounds compete Pep-MH | no repair difference at matched HOXA9 | the biochemistry is real and does not reach the cell - the more interesting negative, and worth publishing as a class-safety result |
| all four compete equally | - | the competition is a pocket effect rather than a V371 effect; the structural argument was wrong and bleximenib is not a control |
A null here is worth more than most positives. A measured demonstration that four clinical menin inhibitors do not differ on genomic-instability readouts, at matched target engagement, is a class-safety statement about a drug being given to infants, and no such measurement exists.
| design | Stage 1: biotinylated Pep-MH (menin 371-400) pulldown of recombinant menin with DMSO, ziftomenib, revumenib, DS-1594b or bleximenib at 1 and 10 uM; Pep-Mut peptide control; MBM1-peptide pulldown as compound-activity control; n=3 independent pulldowns. Stage 2 (only if stage 1 positive): RS4;11 and SEM, three compound arms plus DMSO, concentrations titrated to match HOXA9 and MEIS1 mRNA suppression at 96 h to within 20%, with menin CUT&RUN at HOXA9 confirming the match; 2 Gy irradiation; n=3 biological replicates per line. |
|---|---|
| readout | Stage 1: recovered menin by western densitometry, normalised to DMSO. Stage 2: (a) fraction of cells with >10 gamma-H2A.X foci at 0.5, 2, 6 and 24 h after 2 Gy, 200 cells per arm per replicate; (b) HR and NHEJ efficiency by DR-GFP and EJ5-GFP; (c) menin occupancy at HOXA9 by CUT&RUN. |
| controls | Pep-Mut alanine-substituted peptide; DMSO; MBM1-peptide pulldown to show each compound is active in the assay; HOXA9/MEIS1 matching as the occupancy control; CUT&RUN as the engagement control; EJ5-GFP as the pathway-specificity control (the published peptide potentiates NHEJ while suppressing HR, so NHEJ should move the other way). |
| cost tier | low for stage 1, med for stage 2 |
| timeline | stage 1 one week; stage 2 six weeks if stage 1 is positive |
| system | human KMT2A::AFF1 B-lineage lines RS4;11 and SEM - the index fusion and the index lineage, chosen because every published menin DNA-repair experiment is in fibroblasts, MEFs or A549 |
| what would make it uninterpretable | failure to match HOXA9 suppression across arms to within 20%; in that case report the mismatch and do not report the repair comparison |
| shown for fusion | NOT YET - this experiment is designed in KMT2A::AFF1 precisely because no menin DNA-repair result has ever been obtained in any fusion-driven cell. |
| absence class | never-measured |
- A positive result would mean the menin inhibitor with the best leukaemia data is also an HR inhibitor in normal progenitors, and that is a late-effects problem in a cured child that no trial endpoint currently measures.
- A negative result at matched occupancy would be used to argue the class is interchangeable, which it is not on other axes - the compounds differ in QTc, in differentiation syndrome rates and in CYP interactions, and a clean DNA-repair null should not be allowed to license switching on those grounds.
- The experiment cannot distinguish a drug effect on the menin-H2A.X contact from a drug effect on any other fingers-lobe function, because the compounds differ at five rim residues and not only at V371. A positive result localises the effect to the rim, not to the motif.
- Is bleximenib obtainable for academic use? The whole design depends on the within-class negative control being available, and if it is not, DS-1594b and revumenib against each other is a much weaker comparison because both contact V371.
- Should the comparison include a menin degrader arm? A degrader removes the fingers lobe entirely and should show the maximal phenotype, which would calibrate how much of it a rim contact could possibly produce. `risk-a-menin-degrader-is-not-a-pocket-inhibitor-for-the-hr-scaffold` argues the same point.
- Is 2 Gy the right dose in a B-lineage leukaemia line? These lines are radiosensitive and a dose that kills confounds foci scoring; a pilot survival curve is needed before the foci experiment.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
Three of the four crystallised clinical menin inhibitors contact menin V371 at 2.3-4.3 A while bleximenib's nearest approach is 10.72 A, and bleximenib places 23.8% of its 4.5 A footprint in menin's fingers lobe against 43.5-47.6% for the other three, which is what makes it a within-class negative control.
analyses/q36-menin-residue-axis/run.py - per-chain-pair ligand contact footprints on RCSB 7UJ4, 8VA6, 9WKU and 8IG0, run 2026-09-22 · RCSB PDB, computed 2026What this evidence cannot show
One crystal form per compound and rim contacts are the least reproducible part of a structure. A structural difference is a reason to run an experiment, not a result; nothing here says the contact matters.
Evidence that argues against this (1)
- All four compounds share a 15-residue contact core and all displace KMT2A in cells, so the measured pharmacology of the class is uniform and the non-uniformity is confined to five rim residues with no phenotype attached. Huang J et al. 2012
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The reagents this experiment needs already exist and are published: a biotinylated 30-mer covering menin 371-400 (Pep-MH) that precipitates H2A.X and blocks the menin-H2A.X interaction, an alanine-substituted control peptide (Pep-Mut), and the demonstration that menin delta-371-400 fails to rescue HR in Men1-knockdown cells while wild-type menin does.
The tumor suppressor menin is a key scaffold mediator for homologous recombination repair. · Lian A, Zhu X, Jin B, Gu G, Xu B, Zhang L, et al. · Nucleic Acids Research 2026What this evidence cannot show
The exact Pep-Mut substitution positions are in Supplementary Fig. S3C, which was not retrieved, so the control peptide would have to be re-derived or requested from the authors. The published system is A549 and Men1-knockout MEFs, not a KMT2A-rearranged B-lineage line, so the HR rescue result is being transferred into a system where it has not been shown - which is itself part of what this experiment establishes.
Evidence that argues against this (1)
- The isolated peptide binds wild-type H2A.X and the S139E phosphomimetic equally, so the motif is not phospho-selective in vitro; a competition result with a compound therefore says nothing about the damage-induced form of the interaction, which is the one that matters in an irradiated cell. Lian A et al. 2026
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
No title or abstract in PubMed pairs bleximenib, under the spellings bleximenib, JNJ-75276617 and JNJ75276617, with H2AX, "H2A.X", "DNA repair", "DNA damage" or "homologous recombination"; the compound-name operand control returns 20 records and the readout operands return 6,971, 784 and more, so the zero is not an instrument failure.
PubMed E-utilities esearch via scripts/pubmed_safe.py with query translations read back, 2026-09-22 · NCBI PubMed 2026What this evidence cannot show
Titles and abstracts only. A gamma-H2A.X western in a supplementary figure of a bleximenib paper is invisible to this bound (brief rule 80), and with only 20 bleximenib records in total the literature is small enough that reading all 20 would be cheap and was not done here. Counts move as PubMed indexes.
Evidence that argues against this (1)
- An absence of published comparison is an absence of testing, not a negative result - the experiment may well have been run inside the sponsor and not reported, which would make this `measured-not-reported` rather than `never-measured`. Lian A et al. 2026
the dose-matched three-arm comparison that would settle it
arm 2
arm 3
- related to →H2AX - filed as a gene and not as a mark, and the reason that decision changes what the graph can say
gamma-H2A.X foci are the primary readout