Experiment: does a MEN1 resistance mutant pay a price when the drug is not there? The competition assay nobody has run
Mix cells carrying the drug-resistance mutation with ordinary leukemia cells and watch which wins when there is no drug. If the resistant ones lose, then giving the drug in pulses could keep them suppressed; if they do not, that whole idea collapses.
The same thing, for a scientist
Every evolution-informed dosing strategy — intermittent dosing, adaptive therapy, drug holidays — rests on one measurable quantity: the relative fitness of the resistant genotype in the absence of drug. For MEN1 interface mutants that quantity has never been measured. The precedents cut both ways: BRAF-mutant melanoma cells resistant through BRAF-V600E over-expression are drug-*dependent* and regress on drug withdrawal [E1], whereas BCR-ABL T315I and Y253H show no growth advantage but also no measured cost off imatinib [E2], and among BCR-ABL kinase-domain mutants the fitness effect is mutation-specific, with M351T and H396P less potent transformers than unmutated BCR-ABL while Y253F and E255K are more potent [E3]. A barcoded competition assay across the recurrent MEN1 mutants settles this for our disease.
Design
Genotypes. Isogenic KMT2A::AFF1 lines (SEM, MV4-11) carrying, at the endogenous MEN1 locus by base editing or prime editing, each recurrent interface mutant: M327I, G331R, G331D, T349M, plus a synonymous-edit control and an unedited parental line. Base editing is the right tool here because the resistance space was originally mapped by an unbiased base-editor screen 4.
Assay. Unique DNA barcodes per genotype; pooled competition at defined starting fractions (1%, 10%, 50%) in three conditions: no drug, continuous drug, and an intermittent schedule. Amplicon sequencing of barcodes at intervals over 8-12 weeks; fit selection coefficients per genotype per condition.
In vivo arm. The same pool transplanted into immunodeficient mice, with the three schedules, reading barcode fractions from serial marrow aspirates. The in vitro and in vivo selection coefficients should be reported separately and not averaged, because the niche plausibly changes the cost.
Mechanistic follow-up on any cost found. If a mutant is less fit off drug, ask why: measure menin-MLL1 binding affinity for the mutant, fusion chromatin occupancy, and HOXA9/MEIS1 output. A cost that comes from weakened menin-MLL1 binding is exploitable; a cost from a general protein-stability defect is less interesting and less likely to hold in a patient.
The number this produces, and what it decides
A selection coefficient s for each mutant off drug. Three regimes:
- s < -0.05 (real cost). Intermittent dosing and adaptive scheduling become rational, and
hyp-agency-revocation-hysteresis-pulsed-meninandhyp-adaptive-therapy-requires-a-fitness-costboth gain a foundation. - s ≈ 0 (neutral, the BCR-ABL T315I pattern). Resistant clones persist through drug holidays without expanding or contracting. Intermittent dosing is then justified by toxicity alone, and every evolution-based argument for it should be dropped.
- s > 0 (advantage). Intermittent dosing actively breeds resistance and should not be pursued. This outcome is the one that would most change practice, and it is not implausible given that some BCR-ABL kinase-domain mutants transform better than wild type 3.
Why it has not been done
The mutants were discovered in 2023 in the context of explaining clinical failure, and the field moved immediately to next-generation inhibitors and combinations. Fitness off drug is a question you only ask if you are considering not giving the drug continuously, and nobody in this field is considering that. It is a cheap experiment — one base-editing campaign, one barcoded pool — sitting behind a question nobody has posed.
| design | Barcoded pooled competition assay of base-edited endogenous MEN1 interface mutants (M327I, G331R, G331D, T349M) against unedited and synonymous-edit controls in isogenic KMT2A::AFF1 backgrounds, under no drug, continuous drug and intermittent drug, in vitro and in transplanted mice. |
|---|---|
| readout | Selection coefficient per genotype per condition from amplicon sequencing of barcode fractions over time; mechanistic follow-up on any mutant with a cost (menin-MLL1 affinity, fusion chromatin occupancy, HOXA9/MEIS1 output). |
| controls | Synonymous-edit control to separate editing burden from mutation effect; unedited parental; starting fractions varied across three orders of magnitude to detect frequency dependence; in vitro and in vivo arms reported separately rather than pooled. |
| cost tier | low |
| timeline | 6-9 months, editing and validation 3 months, competition 3 months, in vivo arm a further 3 months |
| decision value | Produces the single number on which every evolution-informed dosing proposal in this disease depends, and which currently does not exist. A neutral or positive result kills two hypotheses in this graph cleanly, which is worth as much as a positive one. |
tests hypotheses
[ "hyp-agency-revocation-hysteresis-pulsed-menin", "hyp-adaptive-therapy-requires-a-fitness-cost" ]
collaborators needed
[ "a base-editing group with endogenous-locus editing in leukemia lines", "a barcode sequencing pipeline" ]
- If any mutant is fitter than wild type off drug, then the resistant clone is not merely surviving therapy but is a better leukemia, and reserving menin inhibitors for later lines becomes even harder to defend.
- A cost measured in a cell line may vanish in a marrow niche, so a positive in vitro result should not on its own justify a schedule change in children.
- Is fitness frequency-dependent? A mutant that is neutral at 50% may be costly at 1% if it depends on paracrine support from sensitive neighbours, which the varied starting fractions are designed to detect.
- Do the mutants differ from each other enough that class-level statements are meaningless, as they are for BCR-ABL kinase-domain mutants?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Vemurafenib-resistant melanoma xenografts became drug-dependent, so that stopping the drug caused regression of established resistant tumours, and a discontinuous dosing strategy exploiting the fitness disadvantage of resistant cells off drug forestalled 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
Resistance there arose by over-expression of the drug target, a mechanism with an obvious cost. An interface point mutation that preserves normal partner binding has no obvious reason to cost anything, so this precedent is optimistic rather than representative.
Evidence that argues against this (1)
- REFRAMES THE EXPERIMENT'S DECISION VALUE. The vemurafenib drug-addiction precedent is the strong-cost case, and this experiment is positioned as testing whether the same precondition holds for MEN1 interface mutants. But adaptive therapy does not require it: reviewing the modelling, Wodarz concludes that clinical benefit 'can be achieved even in the absence of a fitness cost for resistant cells, which broadens the applicability of adaptive therapy.' So a null result here — no measurable cost — does not rule out intermittent or adaptive menin dosing, and the experiment should not be written up as a go/no-go. What it does supply is the size of the effect, which determines the schedule, not whether there is one. Wodarz D 2021
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In competitive repopulation assays in primary murine bone marrow 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
Demonstrates neutrality, not cost. The assay design is, however, the right template for the experiment proposed here.
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
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Among five common BCR-ABL kinase-domain mutants, M351T and H396P were less potent transformers than unmutated BCR-ABL while Y253F and E255K were more potent and T315I was assay-dependent, showing that the fitness effect of a resistance mutation is mutation-specific rather than uniformly costly.
Kinase domain mutants of Bcr-Abl exhibit altered transformation potency, kinase activity, and substrate utilization, irrespective of sensitivity to imatinib · Griswold IJ, MacPartlin M, Bumm T, et al. · Molecular and Cellular Biology 2006What this evidence cannot show
A kinase, not a chromatin adaptor, and transformation potency is not the same measurement as competitive fitness. It is cited to justify testing every mutant separately rather than assuming a class effect.
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. The finding itself is that fitness effects are mutation-specific and assay-dependent, with T315I going both ways, which argues against generalising a fitness cost to a class of resistance mutation in a different protein. Nature 2023
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
MEN1 drug-interface mutations were identified in patients with acquired resistance to menin inhibition and were confirmed as a conserved resistance mechanism in xenografts and in an unbiased base-editor screen, which establishes base editing as a validated tool for constructing the mutant panel.
MEN1 mutations mediate clinical resistance to menin inhibition · Perner F, Stein EM, Wenge DV, et al. · Nature 2023What this evidence cannot show
All reported patients are adults with AML; whether the same mutant spectrum arises in infant B-ALL is unknown.
Evidence that argues against this (1)
- Prior expectation is near-neutrality, and the design should be powered for that rather than for a large cost. The interface mutations were shown to attenuate drug-target binding through structural perturbations that affect small-molecule binding 'but not the interaction with the natural ligand MLL1'. A mutation that leaves the physiological interaction intact has no obvious mechanism for a large off-drug fitness penalty, which is exactly what the BCR-ABL comparator in this entry found for T315I and Y253H. Detecting a few-percent-per-generation difference in competitive repopulation requires many more replicates and longer competition than a conventional two-arm assay, and an underpowered null would be the worst possible outcome because it would be read as 'no cost'. Perner F et al. 2023
counter-evidence pass: adaptive-therapy benefit is achievable without a fitness cost (Wodarz 2021), so a null result in this experiment does not license the entry's scheduling conclusion
measures whether resistant clones lose ground during a drug holiday
selection coefficients for M327I, G331R, G331D and T349M off drug
the selection-coefficient measurement on which the negative prediction rests