GABA(A) receptors are pentameric ligand-gated ion channels playing a critical role in the modulation of neuronal excitability. These inhibitory receptors, gated by -aminobutyric acid (GABA), can be potentiated and even directly activated by intravenous and inhalational anesthetics. Intersubunit cavities in the transmembrane domain have been consistently identified as putative binding sites by numerous experiment and simulation results. Synaptic GABA(A) receptors are predominantly found in a 2↵:2 :1 stoichiometry, with four unique inter-subunit interfaces. Experimental and computational results have suggested a perplexing specificity, given that cavity-lining residues are highly conserved, and the functional effects of general anesthetics are only weakly sensitive to most mutations of cavity residues. Here we use Molecular Dynamics simulations and thermodynamically rigorous alchemical free energy perturbation (AFEP) techniques to calculate affinities of the intravenous anesthetic propofol and the inhaled anesthetic sevoflurane to all intersubunit sites in a heteromeric GABA(A) receptor. We find that the best predictor of general anesthetic affinity for the intersubunit cavity sites is water displacement: combinations of anesthetic and binding site that displace more water molecules have higher affinities than those that displace fewer. The amount of water displacement is, in turn, a function of size of the general anesthetic, successful competition of the general anesthetic with water for the few hydrogen bonding partners in the site, and inaccessibility of the site to lipid acyl chains. The latter explains the surprisingly low affinity of GAs for the ↵ intersubunit site, which is missing a bulky methionine residue at the cavity entrance and can be occupied by acyl chains in the unbound state. Simulations also identify sevoflurane binding sites in the subunit centers and in the pore, but predict that these are lower affinity than the intersubunit sites.
SignificanceAfter over a century of research, it is established that general anesthetics interact directly with hydrophobic cavities in proteins. We still do not know why not all small hydrophobic molecules can act as general anesthetics, or why not all hydrophobic cavities bind these molecules. General anesthetics can even select among homologous sites on one critical target, the GABA(A) heteropentamer, although the origins of selectivity are unknown. Here we used rigorous free energy calculations to find that binding affinity correlates with the number of released water molecules, which in turn depends upon the lipid content of the cavity without bound anesthetic. Results suggest a mechanism that reconciles lipid-centered and protein-centered theories, and which can directly inform design of new anesthetics.