Training in computational techniques & virtual compound screening
At the beginning of this project, the researcher received training in computational methods, in particular in the development of receptor homology models, the preparation of compound libraries for virtual screening, and the analysis of large data sets using RStudio. Subsequently, active and inactive state homology models of GPR3 were developed and their validity was tested validity by virtually screening known GPR3 ligands (AF64394 and two structural analogs) embedded in libraries of decoy molecules. Based on receiver operating characteristic (ROC) curve analysis, one active and one inactive state model were selected. Next, a screen of about 100,000 molecules was conducted and about 500 docking poses were visually inspected and 93 molecules were purchased for in vitro testing.
Development of live cell GPR3 assays & teaching of other lab members
A conformational GPR3 biosensor based on bioluminescence resonance energy transfer (BRET) was successfully developed in work package 2. This biosensor allows for monitoring GPR3 conformational changes in a 96-well microplate format with high sensitivity and robustness, as demonstrated by a high Z-factor of approximately 0.8 (Figure 1). In addition to this conformational readout, intracellular signaling readouts using established sensors for Gs activity and cAMP accumulation were developed and verified using known GPR3 ligands, DPI and AF64394. In the course of this work package.
In-vitro testing and pharmacological profiling of virtual screening hits
First, all 93 compounds selected from the virtual screening were tested at a single concentration using the GPR3 conformational biosensor. Three molecules induced significant changes in GPR3 conformation, and 14 chemical analogs of these molecules were purchased from commercial suppliers. Of these 14 molecules, three additional compounds induced significant changes in GPR3 conformation, yielding a total of 6 active GPR3 ligands. To check for GPR3 selectivity, all six compounds were tested as a negative control on the conformational biosensor beta2-AR. From these tests, three molecules – hereafter referred to as VH1/2/3 – were selected as GPR3 hit ligands for further optimization efforts. VH1/2/3 were also tested for effects on GPR3/6/12-mediated cAMP accumulation. VH1 reduced GPR3- and GPR6-mediated cAMP signaling, while VH2 selectively blocked GPR3-mediated cAMP production. VH3 had no effect on cAMP accumulation downstream of GPR3/6/12.
Finding affinity- and efficacy-optimized ligands through a 2nd round of screening
Two independent ligand-based approaches were implemented to obtain GPR3 ligands with improved ligand potency. First, a virtual screening for shape, 3D and functional group mimetics of VH1/2/3 was performed using the ROCS software from Openeye. Approximately 500,000 molecules, pre-filtered based on physicochemical similarity to VH1/2/3, were screened for similarity to the query molecules and a total of 19 new molecules were purchased and tested with the GPR3 conformational biosensor. Unfortunately, none of the molecules tested showed GPR3-specific activity in our assay. The second approach was based on classical chemical derivatization of a hit molecule. Here, a total of 40 new molecules was synthesized and tested using the GPR3 conformational biosensor. All molecules were first tested at two different concentrations and 20 compounds were selected for full concentration-response curves. Three molecules showed enhanced GPR3-specific activity, with the analog named MR20 having the highest potency at GPR3. MR20 induced conformational changes in GPR3 with a potency of 5 µM and inhibited GPR3-mediated Gs activation with an EC50 of 300 nM (Figure 2), representing a significant improvement in ligand potency compared to the parent molecule VH1. Compared to VH1, MR20 exhibited reduced inverse agonist potency, and neither our computational models nor the structure-activity relationship data sets enabled us to understand the attenuated potency.
