Abstract

Benzodiazepines exert their anxiolytic, anticonvulsant, muscle relaxant, and sedative-hypnotic properties by allosterically enhancing the action of GABA at GABA_A receptors via their benzodiazepine binding site. After 50 years of clinical use, the molecular basis of this interaction still is not known as all attempts for structural resolution failed so far. In the absence of a crystal structure, protein homology modeling and molecular dockings are the only source for structure-based binding mode hypotheses for benzodiazepines. But two obstacles make this undertaking extraordinary challenging. First, the homology models were quite uncertain due to the low target-to-template sequence identity. And secondly, although standard docking tools are capable of reproducing the correct binding mode, they regularly fail to select this out of the produced pose list. In this thesis, the obstacles were tackled in a two step approach. Primarily, the model uncertainty was faced by an explorative step, where nine benzodiazepines were docked into an array of α1β2γ2 GABA_A homology models, considering flexible sidechains and keeping the 100 best scored poses per ligand per model. Consequently in the selection step, exhaustive implementations of various validation sources in an orchestrated and integrative manner were necessary to filter the gigantic pose space. A key criterion in this filtering process was the common binding mode (CBM) hypothesis, which assumes that diazepam and its close structural analogues exhibit a common binding mode within the binding site. Therefore, ligand poses were clustered according to their common scaffold, which led to three common binding mode geometries, CBM I–III. Then, the integrative qualities of CBM came into effect by incorporating experimental information from nine benzodiazepines for CBM I-III validation. The evaluation clearly demonstrated that CBM I is convincingly supported by a large variety of structural, computational and experimental evidence. The CBM I ligandreceptor complex was then used in structure-based virtual screening runs and led to the successful discovery of three novel, experimentally validated, benzodiazepine binding site ligand classes. The structural models were also used to investigate mechanistic receptor features and finally identified α1Y168 to be important for the mechanistic coupling from benzodiazepine binding to GABA_A channel modulation. In a broader view, the gained structural models for α1β2γ2 GABA_A receptor subtypes that will ultimately improve our understanding of the structural determinants for subtype selectivity of GABA_A receptors, leading to drugs excluding subtypes known to be responsible for addicting. Finally, the underlying computational workflow can stimulate other groups to integrate various information pieces in binding mode elucidation of structurally unresolved membrane proteins.