The menin-inhibitor-resistant state has its own address: SMARCA4, BRD4 and CREBBP, with no MEN1 mutation anywhere

resistance-adaptive-nongenetic-menin establishes that escape without a MEN1 mutation happens and names five routes by which it might. It does not say what the resistant cell needs. That is the question a therapeutic programme actually has to answer, and 1 answers it experimentally: derive the resistant state, then screen it.

The answer — SMARCA4, BRD4, CREBBP — is not a restatement of the mechanism. It is an orthogonal handle. A cell can arrive at the MITR state by any of the routes in the adaptive entry and still be killed by a BRG1/BRM inhibitor, because the dependency is a property of the destination rather than of the path. That is a different and more useful kind of fact than a mechanism, and it is why this belongs as its own entry rather than as a fifth bullet in someone else's named_routes list.

The combination worked in four distinguishable settings 1:

• MITR cells (acquired, non-genetic resistance, derived in the laboratory)
• cells expressing the mutant menin M327I (acquired genetic resistance — the recurrent clinical interface mutation)
• patient-derived AML cells with MLL1 rearrangement that were ex vivo resistant to menin inhibitors (primary, not a line)
• patient-derived AML cells with mutant NPM1, same condition

That a single combination covers both the genetic and the non-genetic escape route matters, because the rest of this beat is a catalogue of mechanisms that each demand their own countermeasure. If a chromatin-remodeller dependency is common to resistant cells regardless of how they became resistant, then it is the most efficient place to intervene, and it is the strongest argument in this beat for a combination rather than a sequence.

The in vivo arm used revumenib itself, not only the tool compound, and reported improved survival without significant toxicity in immune-depleted mice.

MITR cells were made by repeated drug shocks with recovery in between — a pulsed-exposure protocol, not continuous selection. That is a deliberate and defensible choice, and it is closer to clinical dosing than continuous exposure. But it is still two cell lines (MV4-11, an MLL-rearranged line, and OCI-AML3, an NPM1-mutant line) in vitro, and a state reached by laboratory selection is not guaranteed to be the state a child reaches on therapy.

The primary patient-derived arm partly answers that objection, but note its exact form: those cells were ex vivo resistant, meaning intrinsically insensitive, not taken from a patient who progressed on a menin inhibitor. The study therefore shows the combination works on intrinsically resistant primary cells and on laboratory-derived acquired resistance; it does not show it works on cells from a patient who relapsed on revumenib. That sample type remains the scarcest material in the field.

Both proposed partners hit common-essential machinery. BRD4 is broadly essential across cell lines, which is precisely why BET inhibitors have struggled with therapeutic index for fifteen years. SMARCA4 and SMARCA2 are the ATPases of the SWI/SNF complex, required for normal haematopoietic stem cell function, and the same property that lets a BRG1/BRM inhibitor target AML stem cells 2 makes it a threat to normal ones.

This graph's other resistance entries already flag CDK9 and BRD4 as the agents with the worst therapeutic index in an infant. The SWI/SNF ATPases belong on that list. Nothing in 1 or 2 addresses paediatric tolerability and the xenograft toxicity statement concerns immune-depleted adult mice.