Contrasting binding architectures underlying the asymmetric M327I resistance of menin inhibitors

The interaction between menin and mixed lineage leukemia fusion proteins represents a critical therapeutic target. The clinical inhibitors revumenib and ziftomenib display divergent resistance profiles against the M327I mutation, which frequently compromises revumenib efficacy while ziftomenib maintains potency. Molecular dynamics simulations of a complete $2\times 2$ comparison matrix (WT/M327I $\times$ Revumenib/Ziftomenib, three independent 500 ns replicates per system) suggest that this asymmetry arises from contrasting binding architectures. Revumenib depends on an intrinsically fragile Trp346 hydrogen bond anchor that is disrupted in multiple independent trajectories, including both M327I and WT backgrounds. Ziftomenib distributes binding energy across a redundant hydrophobic network with stable Trp346 engagement across all replicates and minimal sensitivity to position-327 perturbation. Energy decomposition shows the substantial Ile327 contribution originates from forced close-range contact, whereas the minimal Met327 contribution in ziftomenib complexes aligns with distributed network architecture. The contrasting architectures may inform the design of therapeutics with improved mutational tolerance.

Twelve independent 500 ns molecular dynamics simulations (four systems $\times$ three replicates) were performed using OpenMM with AMBER ff14SB for protein and OpenFF Sage for ligands. Systems were solvated with TIP3P water. Binding free energies and per-residue contributions were estimated via MM/GBSA with the GB OBC2 implicit solvent model. Principal component analysis and dynamic cross-correlation matrices characterized conformational sampling.