Multiscale Free-Energy Methods for Protonation-Coupled Light-Responsive Binding of Ionizable Photoswitchable eDHFR Inhibitors.

Photoswitchable ligands enable photocontrol of biomolecular activity by binding to targets in an isomer-dependent, light-responsive manner. Recent developments in ionizable photoswitchable ligands greatly expand their applications but introduce a major design challenge: light-responsive binding can depend on isomeric form, chemical substitution, and binding-induced shifts in protonation equilibria. These effects are tightly coupled, subtle in magnitude, and difficult to predict. Consequently, few computational methods have been developed and systematically benchmarked for quantitatively predicting them. Here, we establish a multiscale free energy method and benchmark it against experimental data for a series of recently developed photoswitchable inhibitors of Escherichia coli dihydrofolate reductase (eDHFR), a crucial target in photopharmacology. Constant pH replica-exchange molecular dynamics and quantum mechanics/molecular mechanics umbrella sampling simulations quantitatively characterize the ligand's protonation-state change upon binding to the eDHFR's active site. Thermodynamic integration (TI) simulations using alternative alchemical pathways, thermodynamic cycles, and protonation-state assignments were evaluated for predicting light-responsive affinity differentials and substituent effects. Direct cis-to-trans transformations with explicit treatment of environment-dependent protonation states best reproduce experimental trends. Compound-to-compound pathways are less reliable because force-field inaccuracies introduce large pKa errors that are difficult to correct when protonation/deprotonation processes implicitly enter the thermodynamic cycle. TI simulations that ignore binding-induced protonation-state changes fail to consistently reproduce experimental trends. Protein-ligand and ligand-water interaction analyses further reveal the energetic and structural origins of isomer-dependent binding. This study presents a multiscale free-energy framework applied to ionizable photoswitches targeting eDHFR, and highlights the importance of explicitly accounting for binding-induced changes in protonation state when predicting light-responsive binding behavior.