Protein-protein interaction-based high throughput screening for adenylyl cyclase 1 inhibitors: Design, implementation, and discovery of a novel chemotype

Abstract

Genetic and preclinical studies have implicated adenylyl cyclase 1 (AC1) as a potential target for the treatment of chronic inflammatory pain. AC1 activity is increased following inflammatory pain stimuli and AC1 knockout mice show a marked reduction in responses to inflammatory pain. Previous drug discovery efforts have centered around the inhibition of AC1 activity in cell-based assays. In the present study, we used an in vitro approach focused on inhibition of the protein-protein interaction (PPI) between Ca²⁺/calmodulin (CaM) and AC1, an interaction that is required for activation of AC1. We developed a novel fluorescence polarization (FP) assay focused on the PPI between an AC1 peptide and CaM and used this assay to screen over 23,000 compounds for inhibitors of the AC1-CaM PPI. Next, we used a cellular NanoBiT assay to validate 21 FP hits for inhibition of the AC1-CaM PPI in a cellular context with full-length proteins. Based on efficacy, potency, and selectivity for AC1, hits 12, 13, 15, 18, 20, and 21 were prioritized. We then tested these compounds for inhibition of AC1 activity in cyclic AMP (cAMP) accumulation assays, using HEK293 cells stably expressing AC1. Hit 15 contained a dithiophene scaffold and was of particular interest because it shared structural similarities with our recently reported benzamide series of AC1 inhibitors. We next tested a small set of 13 compounds containing the dithiophene scaffold for structure-activity relationship studies. Although many compounds were non-selective, we observed trends for tuning AC1/AC8 selectivity based on heterocycle type and substituents. Having an ethyl on the central thiophene caused the scaffold to be more selective for AC8. Cyclization of the alkyl substituent fused to the thiophene significantly reduced activity and also shifted selectivity toward AC8. Notably, combining the fused cyclohexane-thiophene ring system with a morpholine heterocycle significantly increased potency at both AC1 and AC8. Through designing a novel FP screen and NanoBiT assay, and evaluating hits in cAMP accumulation assays, we have discovered a novel, potent, dithiophene scaffold for inhibition of the AC1- and AC8-CaM PPI. We also report the most potent fully efficacious inhibitor of AC8 activity known to-date.

Keywords: Ca2+; adenylyl cyclase; cAMP signaling; calmodulin (CAM); drug discovery; high throughput screen (HTS); inflammatory pain.

FIGURE 1

AC1/CaM FP Assay development. (A) 100 nM AC1-C1b (Blue), AC8-NT (Red), or AC8-C2b (Green) peptide with increasing concentrations of GST-CaM. Saturation is reached at ∼100 nM GST-CaM. (B) 100 nM AC1 peptide and 100 nM GST-CaM with increasing concentrations of EGTA. (C) CRC for CDZ with 100 nM AC1-C1b and 100 nM GST-CaM, with 1% DMSO in the buffer. The AC1/CaM IC₅₀ value for CDZ was ∼15 μM and the IC₉₀ was ∼100 μM, which was used in the final assay. Calmidazolium CAS number: 57265–65–3. All data represents n = 3 experiments ±SEM.

FIGURE 2

Pilot screen individual library results. All pilot screen assays used concentrations of AC1 and CaM shown in Table 1. The dotted line representing the vehicle is beneath the cluster of compounds screened (black circles). 100 μM CDZ is shown as the dotted line labeled as positive control. Polarization values correspond to well positions for plates within the library tested. (A) FDA-approved library, (B) MicroSource Spectrum (MMSP) library, (C) ChemBridge DIVERSet library, (D) ChemDiv CNS library.

FIGURE 3

Concentration response curves (A) and structures (B) for FDA-approved and MMSP library hits. Each hit was used to generate CRCs in the AC1-C1b/CaM (blue circles), AC8-Nt/CaM (red squares), and AC8-C2b/CaM (green triangles) FP assays. (B) The structures for each hit are shown with the hit number and the given compound name. Data is normalized to vehicle (100% maximum polarization) and 100 μM CDZ (0% polarization). All data represents n = 3 experiments ±SE.M.

FIGURE 4

Concentration response curves (A) and structures (B) for the ChemDiv library hits. Each hit was used to generate CRCs in the AC1-C1b/CaM (blue circles), AC8-Nt/CaM (red squares), and AC8-C2b/CaM (green triangles) FP assays. (B) The structures for each hit are shown with the hit number and the given compound name. Data is normalized to vehicle (100% Maximum polarization) and 100 μM CDZ (0% polarization). All data represents n = 3 experiments ±SEM.

FIGURE 5

Concentration response curves for AC1/CaM pilot screen hits in NanoBiT assay and CellTiter Glo-2.0 cell viability assay. NanoBiT results are shown as blue circles for AC1/CaM, green squares for AC8/CaM, and red squares for cell viability (CellTiter Glo-2.0) assay results. The baseline corrected AUC values are normalized to vehicle in the presence of thapsigargin (100%) or vehicle pre-treated with BAPTA-AM (0%) for the NanoBiT assay. The cell viability data is normalized to the cells treated with vehicle (100%) or cell free wells with vehicle added (0%). Data represents n = 3 experiments ±SEM.

FIGURE 6

NMR spectra for CaM in the presence or absence of hit compounds. (A–E) ¹H,¹⁵N-HSQC for 100 μM ¹⁵N-CaM with 10 mM Ca²⁺ in the absence or presence (red) of hit compounds at 5EQ molar excess over CaM. (A) Hit 12 (B) Hit 13 (C) Hit 15 (D) Hit 18 (E) Hit 21. Buffer: 100 mM KCl, 20 mM HEPES pH7.4, 10 mM CaCl₂, 10% D₂O, 5% DMSO, Temperature: 298 K.

FIGURE 7

cAMP accumulation CRCs of CDI compounds at AC1 and AC8. cAMP accumulation data of select CDI compounds 15 (A), 24 (B), 25 (C), 32 (D), in AC1- or AC8-HEK Δ3/6 KO cells. Points in representative summary curve displays the mean ± SEM of the percent of 3 µM A23187 inhibition from the averages of three independent experiments (N = 3).