Resistance without a detectable bindingenergy penalty: distinct dynamic defects of gatekeeper and molecular-brake mutations in FGFR1–erdafitinib recognition

The backbone Cα–Cα distance widely used to report kinase molecular-brake status gives a false negative under a clinically relevant FGFR1 resistance mutation: the brake's side-chain contact is destabilized in every replicate, yet the criterion reports it intact. Acquired FGFR1 kinase-domain mutations limit the clinical durability of erdafitinib, the first approved pan-FGFR inhibitor; loss of binding affinity is the assumed mechanism. We tested two questions in a single design: nine all-atom MD trajectories of the erdafitinib-bound kinase domain (wild type, V561M and molecular-brake mutant N546K; three replicates each; 4.491 μs). First, what a fully replicated comparison can resolve: neither MM/GBSA nor MM/PBSA detected a mutation-associated change in binding free energy; the observed mutant-versus-WT ΔΔG values lie below the detection limit of the present replicate set and within the intrinsic uncertainty of the end-point methods (GB ANOVA p = 0.27; PB p = 0.054). A large, E3810-like penalty (≈5 kcal·mol-1) is excluded. Second, the brake itself—N546K destabilizes the persistent engaged state of the residue-546–Glu562 contact: occupancy falls to 16–55% across replicates, entirely below the 95–100% of WT and V561M, at unchanged Cα–Cα geometry, dwell times fall from tens of nanoseconds to sub-nanosecond flickers, and broken periods reach 19 ns. V561M leaves the brake contact intact and instead loosens the gatekeeper-flanking hydrophobic network (Leu484/Val492/Lys514), dispersing its conformational ensemble across replicates. Side-chain proximity occupancy thus emerges as an operational order parameter for the molecular brake; the kinetic corollary, altered rates at unchanged affinity, is testable by SPR/BLI and unbinding calculations.