In Silico Screening and In Vitro Evaluation of GSK-3β Type I and II Inhibitors: Potential Treatment for Alzheimer’s Disease?

Abstract

Alzheimer’s disease is a neurodegenerative disorder, and the most common form of dementia. The disease is becoming increasingly prevalent in our aging population, and as there are no effective methods of treating the condition, there is an increasing need for the development of new treatments. In this project, the in silico design, screening and in vitro validation of a selection of potential type I and type II inhibitors of GSK-3β as a possible pathway to a new treatment for Alzheimer’s disease is presented. The screening of compounds for possible type I activity was carried out using a newly designed docking consensus scoring method, first employing a benchmarking database to assess various different combinations of docking algorithms to be used in a virtual screening of 157,238 compounds. The final consensus scoring method was a Simple Sum Rank combination of Glide-SP and –XP, AutoDock Vina and GOLD ASP, chosen based on its superior statistical metrics produced for EF (24), EF’ (33) and BEDROC α=160.9 (0.345), all of which were higher than those of the individual programs. Pharmacophore models were applied to improve the overall accuracy of the results. Once the virtual screening had finished 10 diverse final compounds were selected fit for in vitro validation, based on a wide variety of different protein-ligand interactions. The best type I inhibitor compound 6 (ZINC000072152229) with an IC50 of 24.69 ± 0.73 μM. For the investigation of type II inhibition of GSK-3β, two different type II DFG-out models were developed for the virtual screening of a natural product database (27,286 Compounds). The first model was created using DOLPHIN docking, involving the deletion of 5 residues (201-205) to create the type II consistent binding site, and the second model was designed using a combination of Prime loop refinement, induced fit docking and molecular dynamics. Both models were validated using a selection of analogues of type II ligands with known experimental inhibition data. Both of the models produced experimentally consistent data, which indicated they were both accurate at predicting type II binding. Once the type II virtual screening was completed, the resultant ranks of the compounds for each model were combined into a simple sum consensus score, and 20 compounds were selected for biological validation. The three type II ligands that performed the best in the biological validation were Sorafenib (one of the known type II inhibitors) with an IC50 of 32.64 ± 0.76 µM, compound 2 (ZINC000008299930) with 26.96 ± 1.77 µM and compound 4 (ZINC000008297322) with 9.75 ± 2.2 µM, a promising result for the first ever screening of human GSK-3β for type II inhibition, validating such an approach in the future.