Origins of Resistance Conferred by the R292K Neuraminidase Mutation via Molecular Dynamics and Free Energy Calculations.

Point mutations in the influenza virus enzyme neuraminidase (NA) have been reported that lead to dramatic loss of activity for known NA inhibitors including the FDA approved sialic acid mimics zanamivir and oseltamivir. A more complete understanding of the molecular basis for such resistance is a critical component toward development of improved next-generation drugs. In this study, we have used explicit solvent all-atom molecular dynamics simulations, free energy calculations (MM-GBSA), and residue-based decomposition to model binding of four ligands with NA from influenza virus subtype N9. The goal is to elucidate which structural and energetic properties change as a result of a mutation at position R292K. Computed binding free energies show strong correlation with experiment (r(2) = 0.76), and an examination of individual energy components reveal that changes in intermolecular Coulombic terms (ΔEcoul) best describe the variation in affinity with structure (r(2) = 0.93). H-bond populations also parallel the experimental ordering (r = -0.96, r(2) = 0.86) reinforcing the view that electrostatics modulate binding in this system. Notably, in every case, the simulation results correctly predict that loss of binding occurs as a result of the R292K mutation. Per-residue binding footprints reveal that changes in ΔΔEcoul for R292K-wildtype at position 292 parallel the change in experimental fold resistance energies (ΔΔGR292K-WT) with S03 < S00 < S02 < S01. The footprints also reveal that the most potent ligands have (1) less reliance on R292 for intrinsic affinity, (2) enhanced binding via residues E119, E227, and E277, and (3) flatter ΔEcoul and ΔH-bond profiles. Improved resistance for S03 appears to be a function of the ligand's larger guanidinium group which leads to an increased affinity for wildtype NA while at the same time a reduction in favorable interactions localized to R292. Overall, the computational results significantly enhance experimental observations through quantification of specific interactions which govern molecular recognition along the N9-ligand binding interface.