Switching menin inhibitors: four substitutions defeat all of them, three defeat only some, and the map now exists

A patient progresses on revumenib. A MEN1 interface mutation is found. There are now five menin inhibitors in clinical development with distinct chemistry, including a covalent compound. Does switching help?

The base-editor screen turns this from a guess into a lookup 1. Four substitutions at two residues — M327I, M327V, M327T and G331D — are shared across the inhibitors profiled. Three — C334R, E368K/V, V372A — are inhibitor-specific. The residues M327 and G331 are the ones that actually recur in patients: this graph's own audit of the clinical literature found the union of MEN1 substitutions observed in patients with acquired resistance to be six substitutions at four residues — S160T, M327I, M327V, G331D, G331R, T349M — across roughly twelve to sixteen patients in the whole public record.

Overlay those two lists. The substitutions patients actually get are concentrated at M327 and G331, which are in the shared class. So on present evidence the commonest clinical escape mutations defeat the class, and the compound-specific escapes are the ones the screen found but patients have not yet been reported to acquire.

The screen is paired with co-crystal structures of menin bound to each inhibitor, and the resistance mechanisms are interpreted in terms of how each compound engages the KMT2A binding pocket 1. That is the object a medicinal chemistry programme needs, and it is what makes this more than a catalogue.

It also constrains what a next-generation compound can be. If the shared substitutions cluster where every current chemotype makes contact, then a compound that survives them has to bind somewhere else — a different sub-pocket, an allosteric site, or a non-occupancy modality such as degradation. The screen therefore maps not only the escape space but the region a successor must not depend on. This graph's resistance-men1-interface-mutations entry lists that as an open question; 1 is the closest thing to an answer that exists.

In the largest series of outcomes after menin-inhibitor failure, all complete remissions occurred with venetoclax-containing regimens or with menin-inhibitor switching, and the switching arm was bleximenib to revumenib in 1 patient of 2 3. One response in two attempts is not evidence that switching works; it is evidence that switching has been tried twice.

This matters because the false-reassurance failure mode here is specific and predictable. A single published response to a switch, in a field desperate for options after menin failure, will be cited as though it established a strategy. The base-editor data say the prior should be low unless the patient's substitution is in the inhibitor-specific class — and nobody is currently genotyping to that resolution before switching.

The screen addresses MEN1 coding substitutions. It says nothing about whether a patient with non-genetic resistance — the adaptive state, PRC1.1 loss, lineage switch, ENL persistence — benefits from a different menin inhibitor. Mechanistically they should not, because all five compounds act on the same interface, and one preclinical study reports that a second menin inhibitor does not collapse the residual state where a mechanistically distinct agent does.

It also does not cover the covalent class in the way a clinician might assume. A covalent inhibitor's advantage against a substitution that weakens reversible binding is not automatic; if the substitution removes or shifts the contact geometry that positions the warhead, covalency does not rescue it. Whether the screened panel of five included a covalent compound was not established from the abstract for this entry.