Experiment: does chemotherapy itself install menin-inhibitor resistance? The two-arm controlled test, plus the serial-sample ladder
Split one leukemia sample in two, give one half chemotherapy in mice, leave the other alone for the same length of time, then test both with a menin inhibitor. If only the treated half has become resistant, treatment is the cause.
The same thing, for a scientist
Two arms. Arm 1 is the controlled causal test that the existing observation lacks: an isogenic KMT2A::AFF1 patient-derived xenograft split and passaged in parallel, one branch through a clinically representative chemotherapy course in vivo and one held untreated for the same interval, with menin-inhibitor sensitivity, transcriptome and MEN1 genotype measured on both at the same time point. Arm 2 is the observational ladder: ex vivo menin-inhibitor sensitivity across matched serial samples from the same patients, regressed on cumulative prior therapy. Arm 1 establishes or retires causation; Arm 2 establishes the dose-response and calibrates the impairment index.
Arm 1 — the controlled test
Material. At least three independent KMT2A::AFF1 PDX lines, ideally including infant-derived lines, each expanded and split.
Exposure. One branch receives an in vivo chemotherapy course modelled on the Interfant-06 backbone (a vincristine, corticosteroid, anthracycline and cytarabine-containing regimen at murine-tolerable equivalents); a second branch receives a chromatin-directed agent; a third is held untreated. All branches are harvested at the same absolute time point from split, so elapsed time is matched by construction.
Readout. Ex vivo menin-inhibitor dose-response (IC50 and maximal effect), RNA-seq for the shared non-genetic escape signature, MEN1 amplicon sequencing to exclude genetic escape, and fusion chromatin occupancy on a subset.
The point of the design. Elapsed time, clonal drift and sampling selection are all matched between branches. The only difference is exposure. If sensitivities are equal, the causal claim in hyp-therapy-installed-impairment-index is retired, and the observational association in the literature is confounding. That is a clean, publishable negative.
Arm 2 — the serial-sample ladder
Ex vivo menin-inhibitor sensitivity on every available matched serial sample set from KMT2A-rearranged patients — diagnosis, first relapse, subsequent relapses — from paediatric biobanks. Regress log IC50 on cumulative prior therapy (lines, cumulative anthracycline dose in doxorubicin equivalents, nucleoside-analogue exposure, prior chromatin agent, prior immunotherapy, prior transplant), with patient as a random effect and time from diagnosis as a covariate so that therapy and time are separated as far as observational data allows.
This arm cannot establish causation and should not be written as though it does. It calibrates the index and provides the effect size that Arm 1 either confirms or contradicts.
Feasibility and the honest obstacles
Blast yield from infant marrows at relapse is the binding constraint on ex vivo dose-response, and it is the reason a low-input or PDX-expanded assay format is specified rather than direct primary testing. PDX expansion itself selects, which must be acknowledged and, where possible, controlled by comparing PDX-expanded and direct-primary IC50 on the subset where both are possible.
Arm 1 requires murine dosing equivalents for a multi-agent paediatric backbone, which is non-trivial and is the main reason this experiment has not been done. A simplified two-agent exposure would be a defensible first version if full-backbone dosing proves impractical, provided the simplification is stated rather than glossed.
| design | Arm 1: isogenic KMT2A::AFF1 PDX split into chemotherapy-exposed, chromatin-agent-exposed and time-matched untreated branches, harvested simultaneously. Arm 2: ex vivo menin-inhibitor dose-response across matched serial patient samples, regressed on cumulative prior therapy with patient random effects and time from diagnosis as a covariate. |
|---|---|
| readout | Ex vivo menin-inhibitor IC50 and maximal effect; RNA-seq scored for the shared non-genetic escape signature; MEN1 amplicon sequencing to exclude genetic escape; fusion chromatin occupancy on a subset. |
| controls | Time-matched untreated branch from the same starting material (the control the published observation lacks); MEN1 sequencing to exclude genetic escape as the explanation; PDX-expanded versus direct-primary IC50 comparison where both are possible, to bound the selection introduced by expansion; at least three independent PDX lines. |
| cost tier | med |
| timeline | Arm 1, 9-12 months including PDX expansion and in vivo dosing. Arm 2, 4-6 months given biobank access, and can run first. |
| go no go | If Arm 1 shows no sensitivity difference between exposed and time-matched untreated branches across three lines, report the negative and retire the causal claim rather than pursuing Arm 2 as though it were causal. |
| decision value | Determines whether cytotoxic bridging before a menin inhibitor is harmful and whether the case for frontline use rests on more than fitness, both of which are live protocol questions. |
| why nobody has done it | It requires a treated-and-untreated parallel PDX design that costs roughly twice a standard resistance study and answers a question about sequencing rather than about mechanism, which is nobody's primary grant aim. |
tests hypotheses
[ "hyp-therapy-installed-impairment-index" ]
collaborators needed
[ "a paediatric leukemia biobank with matched serial samples", "a PDX facility able to deliver multi-agent in vivo chemotherapy schedules" ]
- A positive result implies current standard sequencing is destroying the value of a newly approved drug, which is a strong claim to rest on three PDX lines; independent replication should be arranged before publication rather than after.
- An inadequate chemotherapy exposure in Arm 1 would produce a false negative that closes off an important question, so the exposure must be pharmacodynamically validated (marrow response, blast kill) and not assumed.
- Can a murine multi-agent exposure be made clinically representative enough that a negative result is believable, or would a negative simply mean the exposure was inadequate?
- Is the installed state, if it exists, reversible after a therapy-free interval in the PDX, which would convert the impairment index from cumulative to time-dependent?
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Heavily pretreated KMT2A::AFF1 ALL samples were much less menin-inhibitor-sensitive than earlier samples from the same patients despite no prior menin-inhibitor exposure, with sustained on-target transcriptional effect and with RAS-pathway and TP53 co-mutations individually insufficient to induce resistance in vitro.
Upfront menin-inhibitor resistance in multiply pretreated leukemias · Mahdavi L, Bernt KM, et al. · Experimental Hematology 2025What this evidence cannot show
Two patients and no untreated control arm; this experiment exists to supply the control that observation cannot.
Evidence that argues against this (1)
- Clinical data bound how large the effect this ladder is designed to measure can be. In SAVE, relapsed/refractory patients with a median of two prior lines, 52% of them venetoclax-exposed, achieved composite complete remission of 71% and CR/CRh of 60% on a revumenib-containing triplet, with MRD negativity in 80% of responders. If cumulative prior therapy monotonically destroyed menin sensitivity, that result should not exist. The ladder may still detect a graded ex vivo effect, but the design should pre-specify how it will distinguish a real but small gradient from the two-patient anecdote it is built on, and should not be powered on the assumption of a large effect. Issa GC et al. 2026
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Paediatric KMT2A-rearranged ALL and AML cells show distinct, measurable dose-dependent responses to menin inhibition in vitro, with ALL showing rapid dose-dependent cell death, establishing that ex vivo menin-inhibitor dose-response is a workable readout in paediatric material.
Distinct Responses to Menin Inhibition and Synergy with DOT1L Inhibition in KMT2A-Rearranged Acute Lymphoblastic and Myeloid Leukemia · Adriaanse FRS, Zwaan CM, Stam RW, et al. · International Journal of Molecular Sciences 2024What this evidence cannot show
Cell lines and models rather than primary infant blasts at relapse, which is the harder material.
Evidence that argues against this (1)
- Ex vivo dose-response is measured on cells removed from the marrow, where survival depends on culture conditions that differ between ALL and AML samples. A difference in the rate of cell death between the two therefore mixes intrinsic sensitivity with how well each survives ex vivo, which a sensitivity ladder would inherit as a systematic offset. Oncofusion knowledge graph (internal analysis) 2026
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Interfant-06 defines the standard multi-agent chemotherapy backbone for infant KMT2A-rearranged ALL against which a clinically representative murine exposure should be modelled, with 651 patients treated and 6-year event-free survival of 46.1%.
Outcome of Infants Younger Than 1 Year With Acute Lymphoblastic Leukemia Treated With the Interfant-06 Protocol: Results From an International Phase III Randomized Study · Pieters R, De Lorenzo P, Ancliffe P, et al. · Journal of Clinical Oncology 2019What this evidence cannot show
Translating a multi-agent paediatric backbone into murine equivalents is approximate, and the approximation is the main technical risk in Arm 1.
Evidence that argues against this (1)
- A serial-sample ladder built from banked patient material is conditioned on survival to each resampling point, which is the immortal-time structure Suissa describes: later rungs of the ladder are occupied only by patients who lived long enough to be sampled again, so any association between 'more prior therapy' and a biological property is confounded by the selection that produced the sample. The ladder is an observational design wearing an experimental name. It should be analysed with landmark or time-dependent methods, and the matched in vivo chemotherapy-exposure arm the sibling hypothesis describes remains the only design that answers the causal question. Suissa S 2008
the ladder is conditioned on survival to resampling, so it inherits immortal-time/selection structure and cannot answer the causal question the hypothesis poses
controlled treated-versus-time-matched-untreated PDX arms plus a serial-sample dose-response ladder
- related to →Pinometostat (EPZ-5676) — the DOT1L inhibitor, and the most instructive failure in this field
the chromatin-agent exposure branch tests whether prior epigenetic therapy impairs more per unit than cytotoxics