Hypothesis: menin-inhibitor effect is hysteretic — the program collapses faster than it recovers — so intermittent dosing can hold efficacy at much lower cumulative exposure
Switching the fusion protein off happens within minutes. Nobody has ever measured how long it takes to switch back on after the drug is stopped. If switching back on is slow, a child could be given the drug in pulses instead of continuously, with the same effect and far less toxicity.
The same thing, for a scientist
Revocation of an agent's authority is prospective: it stops new acts and does not unwind completed ones. Applied to menin displacement, this predicts an asymmetry in time constants. Degron removal of MLL::AF9 changes elongation within 15 minutes, with chromatin changes following [E1], so the collapse arm is fast and measured. The recovery arm — how quickly the fusion re-loads menin, re-establishes elongation and restores H3K79me2 and the HOXA9/MEIS1 output after drug washout — has never been reported for any KMT2A fusion. Bounded 2026-09-19: PubMed `menin[Title/Abstract] AND (washout OR withdrawal OR "drug holiday" OR intermittent OR rebound OR reversib*)[Title/Abstract]` returns **11** records, all read, none of which is a drug-withdrawal or recovery time course in a KMT2A-rearranged model; they are medicinal chemistry (reversible/irreversible/covalent menin ligands), reviews, and menin biology outside leukaemia. If the collapse-to-recovery ratio favours collapse, an intermittent schedule delivers equivalent time-integrated program suppression at substantially lower cumulative exposure, which in an infant is the difference between a tolerable and an intolerable drug. There is a direct oncology precedent: discontinuous vemurafenib dosing forestalled lethal resistant disease in melanoma xenografts by exploiting the fitness disadvantage of resistant cells off drug [E2]. This is our inference; no menin-inhibitor schedule study exists.
The measurement that is missing1 deeper
Everything about this hypothesis reduces to one unmeasured pair of numbers: t-collapse and t-recovery for the KMT2A::AFF1 program under menin displacement. The literature has t-collapse (about 15 minutes for elongation under complete degron removal of a different fusion) and nothing at all for t-recovery 1. The graph's own mechanism node on reversibility kinetics states explicitly that the H3K79me2 decay half-life is not reported anywhere in the indexed literature.
This is unusual: it is a cheap, unambiguous experiment that would change how a newly approved drug is dosed, and nobody appears to have done it.
Why hysteresis is plausible rather than merely convenient
The recovery arm requires re-recruitment of menin, re-loading of the SEC, resumption of elongation, re-deposition of H3K79me2 over gene bodies, and re-accumulation of HOXA9 and MEIS1 protein to a threshold. Several of those are slow, template-copying or accumulation steps. The collapse arm requires only that one protein-protein contact be occupied. Asymmetric architecture of that kind usually produces asymmetric kinetics. That is an argument from structure, not evidence, and it is the weakest joint in the hypothesis.
Two distinct clinical payoffs, which should not be conflated1 deeper
Payoff 1 — toxicity. Revumenib is dosed every 12 hours continuously and carries boxed warnings for differentiation syndrome and QTc prolongation, with differentiation syndrome in 10-20% of patients across the class 3, grade ≥3 differentiation syndrome 16.0% and grade ≥3 QTc prolongation 13.8% in AUGMENT-101 4. Hyperparathyroidism appears at a median exposure of 2.5 months. Each of these is an exposure-related toxicity in a patient who may be under one year old. Halving cumulative exposure without halving efficacy is worth a great deal on its own.
Payoff 2 — resistance. This is a separate and much more speculative claim. The vemurafenib precedent works because resistant cells are drug-dependent and lose fitness off drug 2, and the equivalent has never been shown for MEN1 interface mutants. The nearest relevant datum is from a different disease: BCR::ABL1 T315I and Y253H confer no growth advantage over wild-type in the absence of imatinib in competitive repopulation assays 5 — neutrality, not a cost. Claiming the resistance payoff without first measuring the fitness of a MEN1 mutant off drug would be exactly the kind of dressed-up reasoning this section of the graph exists to avoid. That measurement is exp-men1-mutant-fitness-cost-competition.
The argument against, stated fairly
If the program recovers quickly, intermittent dosing gives the leukemia periodic full-function windows in which it expands, and the schedule is strictly worse. Worse still, repeated sub-saturating exposure is the classic way to breed resistance rather than delay it. The hypothesis is therefore not safe to act on from the analogy alone; it is safe to act on only after the two time constants are measured, which is why the experiment precedes any schedule proposal.
| statement | The KMT2A::AFF1 transcriptional program collapses faster under menin displacement than it recovers after drug withdrawal (t-recovery / t-collapse > 3), and consequently an intermittent menin-inhibitor schedule delivering roughly half the cumulative exposure achieves non-inferior time-integrated suppression of the HOXA9/MEIS1 program and non-inferior in vivo anti-leukemic effect compared with continuous q12h dosing. |
|---|---|
| origin | analogy |
| source analogy | analogy-law-agency-authority-revocation |
| prior probability | 0.4 |
| prior reasoning | Two claims are being made and they deserve different priors. That hysteresis exists at all (t-recovery > t-collapse) we put at about 0.7, because the recovery arm involves protein re-accumulation and mark re-deposition while the collapse arm involves only pocket occupancy. That the asymmetry is large enough to support a halved-exposure clinical schedule we put at about 0.4. That intermittent dosing additionally delays resistance we put much lower, near 0.2, because it requires MEN1 mutants to be drug-dependent and nobody has tested that. |
| decision value | Dosing schedule is the cheapest variable in the entire menin programme to change and the only one that can be changed after approval without new chemistry. If hysteresis is wide, a paediatric schedule study is immediately justified. |
predictions
[ "After washout of a menin inhibitor at saturating concentration, HOXA9 and MEIS1 transcript recovery to 50% of baseline takes substantially longer than the time to 50% suppression on drug addition, measured in the same cells.", "H3K79me2 at HOXA cluster gene bodies recovers more slowly than nascent transcription does, so chromatin lags output in both directions.", "In a KMT2A::AFF1 patient-derived xenograft, an intermittent schedule at matched cumulative dose gives non-inferior survival to continuous dosing.", "Intermittent dosing reduces the incidence or severity of differentiation syndrome, because differentiation is driven by the integral of suppression and toxicity by the peak.", "The hysteresis width differs between KMT2A::AFF1 and NPM1-mutant disease, since the NPM1-mutant program depends on the same adaptor through a different principal." ]
experiments
[ "exp-menin-pulse-chase-hysteresis", "exp-men1-mutant-fitness-cost-competition" ]
analyses
[]
falsifiers
[ "Program recovery after washout is as fast as or faster than collapse, showing continuous occupancy is required and pulsing would simply give the leukemia growth windows.", "An intermittent schedule at matched cumulative dose is inferior in survival in a patient-derived xenograft.", "Intermittent dosing accelerates the emergence of MEN1 interface mutants in a resistance-competent model, which would invert the intended benefit.", "Differentiation syndrome incidence tracks peak rather than integrated exposure, removing the main toxicity rationale." ]
bounded searches
[
{
"query": "menin[Title/Abstract] AND (washout OR withdrawal OR \"drug holiday\" OR intermittent OR rebound OR reversib*)[Title/Abstract]",
"records": 11,
"relevant": 0,
"date": "2026-09-19",
"source": "PubMed eutils esearch"
}
]- Sub-saturating exposure during the off-phase is the textbook way to select resistance. If the trough is not genuinely zero-occupancy, an intermittent schedule is a resistance-breeding schedule.
- Differentiation syndrome is a mechanism-linked toxicity that scales with how well the drug works. A schedule that reduces it may be reducing efficacy by the same mechanism, and the two cannot be separated by a toxicity endpoint alone.
- Proposing a schedule change for an approved paediatric drug on the strength of an analogy would be irresponsible. The two time constants must come first.
- What is the H3K79me2 decay half-life in a KMT2A::AFF1 cell? It is not reported anywhere in the indexed literature and it sets the floor on how long a drug holiday can be.
- Are MEN1 interface mutants drug-dependent in the way BRAF-V600E-amplified melanoma cells are? If they are, the resistance payoff is real; if they are merely neutral, as BCR-ABL T315I is, intermittent dosing buys toxicity relief and nothing more.
- Does the hysteresis width vary with breakpoint class or with HOXA-high versus HOXA-low status, which would make the optimal schedule patient-specific?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Degron-mediated degradation of MLL::AF9 changes transcriptional elongation at a core subset of target genes within 15 minutes, with loss of the active chromatin landscape occurring subsequently.
MLL::AF9 degradation induces rapid changes in transcriptional elongation and subsequent loss of an active chromatin landscape · Olsen SN, Godfrey L, Healy JP, et al. · Molecular Cell 2022What this evidence cannot show
MLL::AF9 rather than KMT2A::AFF1; complete protein removal rather than adaptor displacement; recovery kinetics after restoring the protein were not reported.
Evidence that argues against this (1)
- Fifteen-minute elongation changes with chromatin following is a DEGRON result on MLL::AF9 — complete removal of a myeloid fusion. A pulsed drug gives partial occupancy of menin, which is upstream and different, so the kinetics quoted set no bound on how long a pulse would have to be. Journal of Clinical Oncology 2025
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In two independently derived primary human melanoma xenograft models, vemurafenib-resistant tumours became drug-dependent — stopping the drug caused regression of established resistant tumours — and a discontinuous dosing strategy exploiting the fitness disadvantage of resistant cells off drug forestalled the onset of lethal drug-resistant disease.
Modelling vemurafenib resistance in melanoma reveals a strategy to forestall drug resistance · Das Thakur M, Salangsang F, Landman AS, et al. · Nature 2013What this evidence cannot show
A different disease, a different target class, and a kinase rather than a chromatin adaptor. Drug dependency of resistant cells is the mechanism that makes intermittent dosing work there and has not been demonstrated for any menin-inhibitor-resistant cell.
Evidence that argues against this (1)
- No leukemia example of intermittent targeted dosing forestalling resistance could be found in a dedicated search, so the precedent stands alone and outside haematology. Das Thakur M et al. 2013
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.review
Across the menin-inhibitor class, composite complete remission rates are 20-35% and overall response 45-65% in heavily pretreated KMT2A-rearranged and NPM1-mutant AML, with on-target differentiation syndrome in 10-20% of patients, differing QTc and cytopenia profiles between agents, and acquired MEN1 mutations described in 39% of post-revumenib relapses but not identified after ziftomenib or bleximenib.
The promise of menin inhibitors: from approval to triplet regimens · Thakur RK, Wang ES · Hematology, American Society of Hematology Education Program 2025What this evidence cannot show
A review, not primary data. The 39% figure and the compound-specific claim should be traced to their primary sources before being relied on for any decision.
Evidence that argues against this (1)
- Class-level ranges of 20-35% CR and 45-65% overall response are aggregated across agents, diseases and lines of therapy, so they smooth over exactly the variation a scheduling hypothesis needs. Differentiation syndrome at 10-20% is the on-mechanism toxicity and is the constraint a pulsed schedule would have to respect. Journal of Clinical Oncology 2025
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In the registration-enabling phase 2 of AUGMENT-101, revumenib was given every 12 hours continuously in 28-day cycles to 94 patients (median age 37 years, range 1.3-75); grade 3 or worse adverse events included febrile neutropenia 37.2%, differentiation syndrome 16.0% and QTc prolongation 13.8%, and among 57 efficacy-evaluable patients CR+CRh was 22.8% with an overall response rate of 63.2%.
Menin Inhibition With Revumenib for KMT2A-Rearranged Relapsed or Refractory Acute Leukemia (AUGMENT-101) · Issa GC, Aldoss I, Thirman MJ, et al. · Journal of Clinical Oncology 2024What this evidence cannot show
Adult-dominated despite enrolling from 30 days of age; no schedule comparison was performed, and the continuous q12h schedule was inherited from phase 1 rather than optimised.
Evidence that argues against this (1)
- Continuous twelve-hourly dosing is what was tested, so the trial provides no comparator for intermittent schedules and its toxicity profile cannot be transposed. The age range down to 1.3 years also means the paediatric experience inside it is small. U.S. National Library of Medicine 2026
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In direct competitive repopulation assays in primary murine bone marrow in vitro and in a murine CML transplantation model, BCR-ABL T315I and Y253H conferred no growth advantage over unmutated BCR-ABL in the absence of imatinib, while imatinib clearly selected for the mutants.
The Bcr-Abl mutations T315I and Y253H do not confer a growth advantage in the absence of imatinib · Miething C, Feihl S, Mugler C, et al. · Leukemia 2006What this evidence cannot show
This shows absence of advantage, not presence of a cost, and it is a kinase mutation in a different disease. Quoting it as evidence of a fitness cost would be an overstatement.
Evidence that argues against this (1)
- BCR-ABL is a constitutively active KINASE and MEN1 encodes a scaffold; a kinase-domain mutation changes catalysis and drug contact at the same site, whereas a MEN1 interface mutation attenuates drug binding while preserving the natural ligand. There is no reason the fitness arithmetic should transfer. The competitive repopulation assays also measure fitness of transduced murine haematopoietic cells, which is not the fitness of an established human leukaemic clone under treatment. Nature 2023
generated by prospective revocation
- supports →Reversibility and kinetics — does the program collapse when the fusion is removed, and how fast?
supplies the missing recovery arm to the measured collapse kinetics
proposes an intermittent alternative to the continuous q12h schedule
- related to →MEN1 drug-interface mutations — the first proof that a chromatin drug can drive Darwinian escape in patients
the resistance payoff depends on whether MEN1 mutants are drug-dependent, which is untested
measures collapse and recovery half-times across five molecular layers
measures whether resistant clones lose ground during a drug holiday
denies intermittent dosing its evolutionary justification, leaving only the toxicity argument
- ← related toTurn menin off for a defined window, then back on: the missing experiment that would bound reversibility
a transcriptional signature that fails to normalise after menin restoration would be a hysteresis result in endocrine tissue
conditional counter-lead: recovery is faster than collapse in most of the assumed space
degradation on eviction is a mechanism for the hysteresis pulsed dosing requires