Development of inhibitors of heterotrimeric Gα_i subunits

Abstract

Heterotrimeric G-proteins are the immediate downstream effectors of G-protein coupled receptors (GPCRs). Endogenous protein guanine nucleotide dissociation inhibitors (GDIs) like AGS3/4 and RGS12/14 function through GPR/Goloco GDI domains. Extensive characterization of GPR domain peptides indicate they function as selective GDIs for Gα_i by competing for the GPCR and Gβγ and preventing GDP release. We modified a GPR consensus peptide by testing FGF and TAT leader sequences to make the peptide cell permeable. FGF modification inhibited GDI activity while TAT preserved GDI activity. TAT-GPR suppresses G-protein coupling to the receptor and completely blocked α₂-adrenoceptor (α₂AR) mediated decreases in cAMP in HEK293 cells at 100 nM. We then sought to discover selective small molecule inhibitors for Gα_i. Molecular docking was used to identify potential molecules that bind to and stabilize the Gα_i–GDP complex by directly interacting with both Gα_i and GDP. Gα_i–GTP and Gα_q-GDP were used as a computational counter screen and Gα_q-GDP was used as a biological counter screen. Thirty-seven molecules were tested using nucleotide exchange. STD NMR assays with compound 0990, a quinazoline derivative, showed direct interaction with Gα_i. Several compounds showed Gα_i specific inhibition and were able to block α₂AR mediated regulation of cAMP. In addition to being a pharmacologic tool, GDI inhibition of Gα subunits has the advantage of circumventing the upstream component of GPCR-related signaling in cases of overstimulation by agonists, mutations, polymorphisms, and expression-related defects often seen in disease.

1. Introduction

Heterotrimeric G-proteins are a ubiquitously expressed class of proteins that transduce signals from activated G-protein coupled receptors (GPCRs). In its inactive form, the G-protein heterotrimer, composed of Gα and βγ subunits, is found at the plasma membrane with Gα in the GDP bound state. When a GPCR is activated by agonist binding, the receptor undergoes a conformational change which facilitates the exchange of GDP for GTP on the Gα subunit. Once activated, the heterotrimer Gα and βγ subunits dissociate and signal to their downstream effectors. The G-protein signal is then terminated when GTP, bound to Gα, is hydrolyzed to GDP and GDP-bound Gα re-associates with Gβγ.

There are four major subclasses of heterotrimeric alpha sub-units, Gα_i, Gα_s, Gα_q, and Gα_12/13. Of these, a large number of vital hormones such as epinephrine, dopamine, acetylcholine, somato-statin, and angiotensin all signal through the Gα_i pathway.¹ Gα_i is best described as being the inhibitory isoform of Gα that suppresses adenylate cyclase activity leading to decreased cAMP accumulation.^2–4 Other effector proteins coupled to Gα_i include c-Src, ERK1/2, phospholipase-C (PLC), and monomeric GTPases.^5–12 In recent years, GPCR signaling cascades have come to light as primary mediators of inflammatory and oncogenic signaling playing critical roles in invasion, metastasis, and uncontrolled growth.^13,14 Misregulated signaling at the level of the G-protein is also a major component of multiple cancers leading to the potential that Gα_i inhibitors could serve a therapeutic niche.^11,15,16

The principal laboratory tool for inhibiting the Gα_i subunit is pertussis toxin (PTX), an enzyme complex produced by the bacterial pathogen Bordetella pertussis. The toxin is taken into the cell via endocytosis where the active enzyme subunit, the S1 A-protomer, ADP-ribosylates the GDP-bound Gα_i subunit of heterotrimeric G-proteins.^17,18 ADP-ribosylation of the alpha subunit locks the entire heterotrimer in its inactive state and prevents activation by agonist-stimulated receptors.¹⁹ PTX blocks the mitogenic effects of GPCR activation by hormones such as epinephrine, sphingosine-1-phosphate(S1P), lysophosphatidic acid (LPA), cytokines, and serotonin.^15,20–23 As an enzymatic modifier, PTX efficacy is slow requiring overnight incubations leading to compensatory mechanisms. Alternative inhibitors would be useful when acute pharmacological study is required or the monomeric subunit only needs to be inhibited.

An alternative strategy used endogenously by cells to inhibit the Gα subunit is to prevent nucleotide exchange. Cellular proteins containing the G-protein regulatory (GPR) or GoLoco motifs bind to Gα_i family subunits and function as GDP dissociation inhibitors (GDI) by preventing nucleotide exchange and subunit activation.^24–26 GPR/Goloco motif-containing proteins in humans include RGS12/14, GPSM1 (AGS3), GPSM2 (LGN), GPSM3 (AGS4 or G18.1b), GPSM4 (PCP2), Rap1GAP, and WAVE1.^26–30 These proteins bind to and stabilize the GDP-bound inactive state of Gα_i, prevent nucleotide exchange, and block heterotrimer oligomerization.^{24,25,31–35} To date, there are no conclusive reports of an endogenous GDI for any of the other heterotrimeric alpha subunit families beyond Gα_i which underscores the critical nature of regulating the Gα_i. GPR/Goloco motif-containing proteins have been implicated clinically in organism development, addiction, and cancer where they are central in the regulation of cell division and differentiation.^36–38

The GPR consensus peptide represents the conserved amino acids found among the eight human proteins that contain a GPR motif.²⁵ The 28 amino acid GPR motif contains 12 conserved residues arranged within a minimum 20mer consisting of charged and hydrophobic microdomains.²⁵ The core conserved residues that are required for activity corresponds to: EE-FF-LL—Q–RMDDQR.³⁵ An X-ray crystal structure of the GPR/Goloco domain of RGS14 supports hydrophobic interactions with the amino terminal residues of the peptide with a hydrophobic patch on Gα_i, and ionic interactions between the carboxy terminal residues of the peptide and Gα_i. Also of critical importance the conserved arginine in the core of the peptide has a direct interaction with the bound nucleotide. These three points of contact create an elegant and specific mechanism to create a stable four part complex of G-protein, nucleotide, Mg²⁺, and GDI. Like endogenous GPR domains, the GPR consensus peptide competes for binding of AGS3 to Gα_i, inhibits GTP binding to Gα_i, blocks receptor coupling of Gα_i, and stabilizes the GDP-bound conformation of Gα_i.^25,35

Here we report the TAT-GPR peptide development strategy and in vitro effects on cAMP. In addition to being an intriguing probe to study G-protein pharmacology, the TAT-GPR peptide also serves as a baseline and platform to develop small molecule GDIs. The biochemistry of the GPR domain allows us to exploit native mechanisms in the design of synthetic mimetics. We therefore go on to report the strategy, discovery, and validations of small molecules GDIs designed to mimic the GPR peptide interaction with Gα_i and GDI biological effects.

2. Experimental section

2.1. GPR Peptides

Peptides were synthesized and purified by Bio-Synthesis, Inc. (Lewisville, TX) and United Peptide Corp (Herndon, VA), and peptide mass verified by matrix-assisted laser desorption ionization mass spectrometry.

2.2. Cell Culture

GloSensor HEK293 cells (Promega, Inc.) were maintained at 37 °C in a 5% CO₂ atmosphere and maintained in Dulbecco's Modified Eagle Medium (ThermoScientific #SH30243.01) supplemented with 10% fetal bovine serum and 50 μg/mL hygromycin B was added for selection.

2.3. GTPγS Exchange Assay

[³⁵S]GTPγS (1250 Ci/mmol) (21.8 Ci/mmol) were purchased from Dupont/NEN (Boston, MA). GTPγS binding assays were generally conducted as described.³⁹ Gα_i (100 nM) was preincubated for 20 min at 24 °C in the presence and absence of GPR peptides. Binding assays (duplicate determinations) were initiated by addition of 0.5 μM GTPγS (4.0 × 10⁴ dpm/pmol) and incubations (total volume = 50 μL) continued 30 min at 24 °C. Reactions were terminated by rapid filtration through nitrocellulose filters (S&S BA85) with 4 × 4 ml washes of stop buffer (50 mM Tris-HCl, 5 mM MgCl₂, 1 mM EDTA, pH 7.4, 4 °C). Radioactivity bound to the filters was determined by liquid scintillation counting. Nonspecific binding was defined by 100 μM GTPγS.

2.4. Computer infrastructure

Molecular manipulations and imaging was performed on desktop computers using freeware and MOE.2011 (CCG). Computationally intensive docking was accomplished with the MUSC Computational Biology Resource Center's Cluster. The CBRC cluster consists of sixteen Dell PE1950 dual quad core nodes running RHEL 4.3 and managed by Platform Computing Open Cluster Stack software with LAVA scheduler. Initial in silico screening was performed against the neutral clean lead 280,357 compound subset of the 2009 ZINC database (http://zinc.docking.org/).^40,41

2.5. Protein targeting

X-ray coordinates came from the PDB 2OM2 file (Gαi1-GDP)⁴² and 2RGN (Gαq-GDP).⁴³ Both proteins had GDP-Mg²⁺ as well as a peptide from either RGS14 or p63RhoGEF complex respectively. PDB file 1AGR was selected as a transition state analog mimicking the activated state of Gαi1-GTP and was used as a negative control to remove compounds that bound to either state in the docking simulations. In order to better mimic the natural state of the G-protein the sulfur from the Xray GSP was replaced with an oxygen. Receptor viewing and editing was done with CHIMERA 1.4. Receptor setup was based on Dock suggested setting. The selected_spheres file was manually edited to further focus the docking procedure. The bound RGS14 peptide and p64RhoGEF as well as any waters were removed prior to docking. Before analysis or simulations, proteins were protonated at pH 7.4 and structures energy minimized with heavy atoms constrained. The positions of GDP and Mg²⁺ were used as positional seeds to set up the initial compound search.

2.6. Screening Docking Simulations

Ligand screening was performed with UCSF DOCK version 6.3.⁴⁴ DOCK was compiled from source with MPICH2.1 that in turn was compiled with INTEL C and FORTRAN compilers (v 10.1). DOCK used default GRID settings, vdw_AMBER_parm99 energy field and 1000 maximum poses per sphere. For initial screening, all ligands were evaluated in rigid mode. The ligand-receptor energy was scored with the unmodified GRID method of DOCK. Both electrostatic and vDW parameters were calculated. Molecules were ranked according to this rigid docking GRID score with 500 maximum poses per sphere. Best scoring compounds were re-docked in flexible mode, using the same parameters as the rigid docking simulations leading to 210 top predicited compounds.

Additional analysis of the high scoring hits and pharmacophore development were accomplished with MOE 2012(CCG). The top 210 compounds were then docked to Gα_i1-GDP-Mg²⁺, Gα_i1-GTP-Mg²⁺, and Gα_q-GDP-Mg²⁺. Docking was performed in screening mode where the receptor (in this case the trimeric system of G-protein, nucleotide, and magnesium) was held rigid and the ligands were allowed to flex. MOE docking used Amber12:EHT forcefield and Born solvation model. Simulations were focused on the nucleotide by selecting the magnesium ion as the ligand placement strategy. Initial placement calculated 60 poses per molecule using triangle matching placement with London dG scoring, the top 30 poses were then refined using forcefield placement and Affinity dG scoring. The 0990 focused docking simulations used the same structural pbd file and docking as in in the DOCK6 simulations. In this case, initial placement calculated 100 poses using triangle matching with London dG scoring. The top 50 poses were then refined using forcefield refinement and Affinity dG scoring.

2.7. Protein expression and purification

Human Gα_i1 was expressed using a plasmid containing the Profinity eXact^® (BioRad)⁴⁵ construct of Gα_i1 inserted into pPAL7 and stably transfected into Escherichia coli BL21 cells. BL21 cells were grown at 28 °C in LB media made using 15.5 g/L of Difco™ 2× YT (ThermoScientific #244020) plus 7.5 g/L NaCl and 50 mg/L ampicillin. Gα_i1 expression was induced with 30 μM isopropyl-thiogalac-tose when OD₆₀₀ was ∼0.6. Temperature was reduced to 25 °C at induction, and induction was carried out 12-16 h. Cells were harvested at 3200g for 18 min. Gαq was expressed in Hi5 insect cells using Gαq baculovirus (generous gift of Steven Graber WVU) at 5 × MOI for 48 h.⁴⁶

Cell pellets were lysed using 100 mM NaPO₄, pH 7.2 supplemented with 150 μM GDP, 5 mM β-mercaptoethanol, Roche Inhibitor Cocktail^®, and 1 mM phenylmethylsulfonyl fluoride. Cell lysates were sonicated and then centrifuged at 100,000g for 30 min. Supernatants were removed and filtered using 0.45 μm syringe filter. Gα_i1 was then purified according to the Profinity eXact^® protocol. Gα_q was partially purified by first isolating the low speed pellet (10,000g) after and lysis followed by isolation using high (50 k) and low (30 k) molecular weight cut-off filters (Amicon Ultracel UFC805024 and UFC803024).

2.8. Real time nucleotide exchange assays

Real time nucleotide exchange assays were performed in 96-well clear flat bottom plates (Corning 3596). 600 nM purified Gα_i1 or Gα_q was incubated with inhibitor compounds in 10 mM HEPES, 100 mM NaCl,5 mM MgCl₂, 0.1 mM EDTA, pH7.4 (BODIPY Buffer) for30 min, and 250 nM of the non-hydrolyzable fluorescent BODIPY FL GTPγS (Invitrogen) was added to initiate nucleotide exchange. Final well volumes were 50 μL. Test compounds were dissolved in DMSO and final vehicle concentration was ≤1% v/v. 1 μM unlabeled GTPγS was added after 20 min to assess protein viability. Relative fluorescence units (RFU) were measured using a SpectraMax M5 (Molecular Devices) plate reader. Fluorescence was measured at approximately four minute intervals at 475 nm excitation, 525 nm emission, and 515 nm cut-off wavelengths for one hour.

2.9. Synthetic chemistry

Reactions were monitored by analytical thin-layer chromatography on EM-Science hard layer silic gel-60F-254 plates cut into 1 × 2.5 cm pieces. Visualization was effected by ultraviolet light (254 nm), followed by staining the plate and drying on a hot plate. The stain was made with 25 grams, of phosphomolybdic acid, 10 grams of cerium sulfate, 60 mL H₂SO₄, and 940 mL H₂O. The potassium permanganate stain was made with 200 mL H₂O, 1.33 grams KMnO₄, 13.33 grams of K₂CO₃, and 4 mL of 5% NaOH. Chromatography was performed by Teledyne ISCO. In reactions where water was not present by solvent, reagent or by-product, the vessels were flame-dried under a slow argon flow. A slight positive pressure of dry argon was maintained via rubber septa seal during the course of the reaction. The argon stream originated from a regulated high pressure argon tank and was used without further drying. All reactions were stirred with a Teflon-coated magnetic stir bar and stir plate. Removal of solvents was typically done using a Buchi rotary evaporator, Model number 011 connected to house vacuum line. The condenser was cooled by a Haake B3 circulator chilled to 0 °C cold thumb. All reaction solvents were purchased as anhydrous with sure seal tops. ¹H NMR spectra were recorded at 400 MHz on a Bruker spectrometer. All chemical shifts were reported from tetramethylsilane with the solvent resonance of CDCl₃ (7.27 ppm), DMSO-d₆ (2.5 ppm). Mass spectra were recorded on a Finnegan LCQ Advantage Max. Reverse phase analytical runs were done on an Agilent 1100 HPLC 95:5 H₂O w/0.1% TFA/MeCN w/0.1% TFA -100% MeCN w/0.1% TFA on a phenomenex Luna 3 micron C18 100 angstrom column 20 × 4 mm, with a flow rate of 1.0 mL/minute.

2.10. 0990CL synthesis

To a flame dried flask cooled under argon is added the amine (1.0equiv), the aryl halide (1.1 equiv), and triethylamine (1.5equiv) The mixture is then taken up in 1.4 mL of ethanol (stored over 4 angstrom mol. sieves). The vessel is sealed, and then heated to 120 °C for a period of 4 days which was determined empirically to give the optimum yield. The contents are concentrated and purified via basic alumina using dichloromethane/Methanol (97:3) ¹H NMR (CDCl₃) δ 7.92 (d, 8.28 Hz, 1H), 7.79 (m, 2H), 7.68 (m, 2H), 7.51 (m, 3H), 7.25 (m, 1H), 4.50 (s, 1H), 1.87 (s, 3H), 1.39 (s, 6H). Mass calcd for C₂₁H₂₁N₅, 343.18. Observed, 344.3 (M+1). The final yield for 0990CL synthesis was ∼40%. Congener synthesis is provided in the Supplementary methods.

2.11. STD NMR

Samples for STD NMR experiments were prepared with 34 μM protein and 500 μM ligand. STD NMR data were collected at 298 K on a Bruker Avance III 600 MHz NMR spectrometer equipped with a 5 mm cryogenically-cooled QCI-inverse probe and using a standard STD pulse sequence^47,48 with 30 ms 8.3 kHz spin lock to minimize background protein resonances. Solvent suppression was achieved using the excitation sculpting scheme.⁴⁹ Saturation of the protein signals was performed using a train of 10, 20 or 59 selective 56 dB Gaussian pulses of 50 ms duration, totaling a 0.51, 1.02 or 3.01 s saturation time. The on-resonance frequency was set up at –0.5 ppm, and the off-resonance one was applied to –17.16 ppm. STD spectra were acquired with a total of 2048 transients in addition to 8 dummy scans. Longitudinal T1 relaxation times were determined from an inversion recovery experiment processed and analyzed using Topspin 3.1. STD enhancements at t_sat = 3 s compensated for perturbations from differential ligand proton T₁-relaxation rates were determined following the procedure outlined by Kemper et al.⁵⁰

2.12. Transfections

For the cAMP GloSensor HEK293 cells, 1 μg of α₂ adrenergic receptor plasmid DNA was added to a final volume of 500 μl of OptiMEM in tube ‘A’, and 2.5 μl of lipofectamine were added to a final volume of 500 μl of OptiMEM in tube ‘B’. Tubes A and B were vortexed briefly and incubated at room temperature for 5 min. Following incubation, tube B was added to tube A, inverted twice to mix, and incubated at room temperature for 40 min. The 1 mL mixture of DNA, lipofectamine, and OptiMEM was added to a 15 mL conical tube containing 1 mL pre-warmed OptiMEM and 3 mL of serum-free MEM. Following medium aspiration and 1 × PBS wash, the 5 mL mixture was added to each 10 cm plate and incubated overnight.

2.13. cAMP measurements

Cyclic AMP was measured in GloSensor cAMP HEK293 cell line (Promega, Inc.) in 96-well clear bottom plates using a FLIPR^TETRA.⁵¹ For PTX treated wells, PTX was added 24 h prior to the assay to a final concentration of 100 ng/mL. The day of the assay (72 h after transfection), cells were pre-equilibrated in GloSensor cAMP reagent and pre-incubated with test compounds for 20 min prior to initiating the assay. Forskolin and/or UK14304 were added simultaneously and relative luminescence units (RLU) were recorded every 2 s for 600 reads.

3. Results

3.1. The permeable GPR peptide

The GPR peptide is a unique and practical way to study the biochemical properties of the GPR motif, Gα_i, and GPCR function in vitro and in vivo. However, a major obstacle to observing the GPR peptide effects within a cell is the inherent lack of peptide permeability through cell membranes. In order to facilitate the entry of the GPR peptide into cells, membrane permeable amino acid sequence tags were explored (Fig. 1A). Numerous membrane permeable sequence tags allow passage of usually impermeable proteins and peptides through the cell membrane.⁵² Two chemically distinct sequences were chosen for comparison. A basic peptide sequence derived from the TAT (aa 48-59, NP_057853) (GRKKRRQRRRPP) protein from HIV.^53,54 TAT fusions (TAT-SIRK) to a Gβγ binding peptide were able to dissociate Gαiβγ heterotrimers while stimulating ERK, JNK, PLC and Ca²⁺ release.^32,55 We also explored a hydrophobic signal-sequence based peptide derived from the FGF related oncogene K-FGF (7-21, NP_001998) (AAVALL-PAVLLALLA) found in humans.⁵⁶ The permeable leader sequences were fused to the amino terminus of the consensus GPR peptide to create a synthetic, cell-permeable GPR peptide.

Figure 1.

The permeable GPR peptides and their effect of on GTPγS binding to purified Gα_i. (A) Amino acid sequences of GPR peptides. GPR motif consensus residues are colored in red and membrane permeable tags are colored in blue. (B and C) GTPγ35S (500 nM) binding to purified Gα_i1 (100 nM) was measured in the absence and presence of peptides. Data are expressed as the percent of specific binding (∼5 pmol) observed in the absence of peptide and represent the mean ± SD derived from two experiments performed in duplicate. The TAT-GPR data in B has been previously published.³⁸

The inhibition of GTPγS binding to purified G-protein by the modified GPR peptides was then compared to the unmodified GPR peptide (Fig 1B). A major concern was the possibility that modification of the GPR peptide would perturb the secondary structure of the peptide, rendering it less active. Both of the tagged peptides inhibited GTPγS binding to purified Gα_i with slightly lower potency than observed with the unmodified GPR consensus peptide. FGF-GPR peptide modification exhibited a half-log right shift in IC₅₀ for Gα_i while the TAT-GPR peptide exhibited an apparent affinity similar to the unmodified GPR peptide. TAT modification remained more potent than the FGF modification and, therefore, the TAT-GPR peptide was chosen a better candidate for further studies.

We then assessed the intracellular effects of Gα_i GDIs on 2nd messenger cAMP levels using the α₂AR (Fig. 2). Gα_i1 associates with the α₂AR in multiple cellular backgrounds as well as in vivo.^57–59 Note that due to a presumed deficiency of specificity of the GPR motif across Gα_i isoforms, it is not specifically known if Gα_i1 versus Gα_i2 (or both) is coupling to the α₂AR within our system. α₂AR expression was optimized to prevent G-protein switching to Gα_s in presence of PTX, as demonstrated by the absence of enhanced forskolin (Fsk) induced cAMP signal in presence of PTX. Cells were pre-treated with vehicle, PTX (100 ng/mL, 24 h), or TAT-GPR (0.1 μM, 1.0 μM and 10.0 μM, 20 min). After pre-treatment with GPR peptides, cells were treated with 250 nM Fsk and 10 μM of the α₂AR selective agonist UK14304. Stimulation with UK14304 reduced the Fsk-stimulated cAMP, and this effect was suppressed by PTX. TAT-GPR was very effective at preventing the Gα_i mediated inhibition of cAMP stimulated by UK14304, resulting in cAMP levels that were similar to Fsk alone. Also of interest, TAT-GPR but not PTX, was able to raise basal cAMP levels compared to vehicle treatment (p = 0.009 between vehicle and GPR basal).

Figure 2.

Effect of Gα_i inhibitors PTX and TAT-GPR on α₂AR/Gα_i mediated cAMP suppression. cAMP levels were measured in HEK293 GloSensor cells transfected with 0.2 μg/mL α₂AR receptor. Cells were pre-treated prior to the experiment with 100 ng/mL PTX (24 h) or 0.1 μM and 10 μM TAT-GPR peptide (GPR, 20 min). Cells were subsequently treated with 250 nM Forskolin (Fsk) or Fsk and 10 μM of the α₂AR agonist UK14304. **t Test between vehicle and TAT-GPR was significant with p = 0.009. Data was derived from triplicate determinates from three independent experiments.

These data demonstrate TAT-GPR can selectively block Gαi regulation of GPCR mediated adenylyl cyclase activity. In the only in vivo experiment using GDIs to date, the TAT-GPR peptide replicates the phenotype of chronic cocaine-treatment by inducing altered craving, locomotor behavior, and increased glutamate transmission when TAT-GPR is infused into the prefrontal cortex of rats.^38,60,61 Although TAT-GPR is very useful as a pharmacologic probe, as a peptide it has several liabilities in development as a potential therapeutic. The GPR consensus peptide has a molecular weight of 3.2 kDa and a pI of 4.1. TAT-GPR is a 40mer with a mass of 4847 Da, a pI of 10.9, and an instability index of 108. We therefore set out to discover small molecule GDIs.

3.2. Identification of small molecule heterotrimeric GDIs

Studies with the GPR peptide and GPR/Goloco domain proteins have taught us that the GDI interaction minimally only needs three points of contact between the GDI and G-protein. Therefore, a small molecule GDI drug only needs to bind the nucleotide and the protein, and each additional interaction is a gain for affinity and selectivity. The initial phase of small molecule identification was performed via computational docking using the X-ray crystal structure of Gα_i1-GDP, the ZINC small molecule database of 280,000 molecules, and DOCK6.3.^40,42,62 The intent was to identify molecules that interacted with both the protein and embedded nucleotide and therefore inhibit nucleotide exchange. This docking study used the inactive conformation crystal structure of Gα_i1-GDP (PDB: 2OM2)⁴² and docked small molecules to a region proximal to the nucleotide binding pocket (Fig. 3A and B). The crystal structure of Gα_i–GTP (active conformation; PDB: 1AGR) was used as a counter screen to ensure subunit state selectivity of the identified potential inhibitor molecules.⁶³ The crystal structure of Gα_q-GDP (PDB: 2RGN)⁴³ was used as another counter screen to filter virtual hits and ensure subunit isoform selectivity. Only 10 compounds out of 280,000 interacted with Gα_q-GDP with a predicted binding energy <–1000 kcal. After ranking and filtering, a total of 210 molecules emerged from the flexible docking after passing thresholds of −1000 kcal and represented a virtual hit rate of ∼0.08% (Fig. 3C). We further validated the DOCK6 based simulations using MOE as an alternate simulation methodology. We redocked the top 210 compounds to the top 210 compounds to Gα_i1-GDP-Mg²⁺, Gα_i1-GTP-Mg²⁺, and Gα_q-GDP-Mg²⁺ (Fig 3D). The large scatter and high (poor) scores for Gα_i1-GTP and Gα_q-GDP indicated that DOCK6 and MOE were in agreement to enrich for compounds that reproducibly interacted with the Gα_i1-GDP nucleotide site.

Figure 3.

Computational protein docking for the discovery of G-protein inhibitors. (A) Grid and placement spheres of Gα_i used for site targeted docking simulations. (B) Ribbon diagram Gα_i1–GDP–Mg²⁺. Ribbons are colored according to chain termini. (C) Top 210 compounds (from 280,000) rank ordered from ZINC neutral DOCK6 docking to Gα_i1–GDP–Mg²⁺. (D) MOE based re-docking of top 210 compounds to Gα_i1–GDP–Mg²⁺, Gα_i1–GTP–Mg²⁺, and Gα_q–GDP–Mg²⁺. Data shows scores for 30 poses per molecule.

The specific Gα_i1-GDP hits were rank ordered according to binding energy and grouped into 65 molecularly similar chemical clusters using a Tanimoto similarity coefficient (Tc) of 60%.⁶⁴ Fifteen representative molecules representing one-quarter of the possible clusters were selected to sample diverse chemical space. These compounds were then validated for their ability to regulate Gα_i and Gα_q.

3.3. Validation of small molecule Gα_i GDIs

Real time nucleotide exchange assays were performed on small molecules obtained from the docking screen using the non-hydrolyzable fluorescent nucleotide analog BODIPY FL GTPγS.^65,66 To improve sensitivity of the assay, we modified the experimental wavelengths by using an excitation of 475 nm and monitored emission at 525 nm with a 515 nm cut-off. Although this modification only left a 10 nm observation window (515 to 525 nm) and decreased the overall signal intensity, we found that it improved the ability to search for low affinity interactions while decreasing potential fluorescence interference.

We tested the 15 compounds in the primary Gα_i1 and Gα_q exchange assay (Table 1). The exchange assay demonstrated a concentration-responsive reduction in fluorescence for several compounds compared to vehicle control reactions (Fig. 4A and B & Fig. S1). Four compounds showed selectivity for Gα_i1 over Gα_q. Selectivity was best demonstrated by compounds 0990 and 4630 which demonstrated selectivity for Gα_i and resulted in a maximum endpoint fluorescence reduction of 38 ± 8% and 34 ± 3%, respectively (Fig. 4C & Fig. S1). Compounds 8005, 8770, and 4799 were selective for Gα_q with 8005 showing a maximum reduction of endpoint fluorescence of 27 ± 7% respectively (Fig. 4D & Fig. S1). Compounds 2967, 6715, and 1026 were not selective and showed inhibition of nucleotide exchange with both Gα_i1 and Gα_q (Table 1). The remainder of the compounds did not affect nucleotide exchange activity. Compounds that showed selectivity for Gα_i1 and also a reduction in end-point fluorescence greater than 10% were retained for further testing. This threshold criterion was chosen in order to explore greater chemical space and chemotypes in the early phase of screening. We also re-synthesized 0990 to confirm structure and purity for all further experiments (Supplementary methods, in-house synthesized compound is specifically named 0990CL).

Table 1. Primary screening for GDIs through chemical diversity sampling.

			% Inhibition
CAS #	ID		Gαi	Gαq
864389-67-3	4630		38	<10
907989-49-5	9827		35	<10
511514-03-7	0990		25	<10
849019-55-2	1217		30	<10
301353-19-5	7136		13	<10
749837-97-6	7507		<10	<10
1070463-59-0	8005		<10	27
284029-69-2	1254		<10	–
1020944-29-9	1174		<10	–
1184184-00-6	2967		34	50
1244948-00-2	6715		38	35
312609-02-2	1026		18	40
111015-95-3	8770		<10	32
712346-46-8	4799		<10	25

Figure 4.

Primary validation of inhibitors by Gα_i versus Gα_q real-time fluorescence nucleotide exchange assay. (A and B) For each experiment 600 nM Gα_i or Gα_q was preincubated with vehicle (grey and red) or 300 μM (green), 100 μM (purple), or 30 μM (olive) test compound 0990CL for 30 min prior to the addition of BODIPY FL GTPγS, and nucleotide exchange was monitored via fluorescence for 60 min. Background signals were determined using BODIPY FL GTPγS alone and is represented in black. Unlabeled GTPγS introduced to control reactions after 20 min (red) resulted in a decrease in fluorescence through time and demonstrates the nucleotide exchange viability of G-protein. (C and D) Endpoint results (60 min) of the kinetic assays depicted in A and B for Gα_i and Gα_q for compounds 4630, 8005, and 0990CL on Gα_i and Gα_q. Data in C–F was derived from triplicate determinations from three independent experiments.

To confirm binding of 0990CL to Gα_i in solution we used saturation transfer difference (STD) NMR. STD NMR is a robust technique allowing direct measurement of ligand binding to target proteins through ligand signals that appear in STD experiments due to saturation transfer from the target protein while in the ligand's bound state.^47,67 STD NMR confirmed that compound 0990CL interacted with purified Gα_i1 in its GDP-bound state which we used in these experiments (Fig. 5). Interestingly, our NMR analysis revealed that compound 0990CL converted to the tautomer state shown in Figure 5A in aqueous solution (H₂O/DMSO-d₆; 90:10) exhibiting a deprotonated nitrogen N1′ and a sp³-hybridized carbon C3′. Compound 8005, which showed no inhibition of Gα_i nucleotide exchange, was used as a negative control. Figure 5B and D show the aliphatic and aromatic regions of reference spectra of a mixture of 0990CL and Gα_i1 in the presence of excess GDP. The STD NMR spectra recorded with varying saturation times (t_sat = 1 and 3 s) revealed several points of contact on 0990CL when bound to Gα_i1 (Fig. 5B and D). The relative STD enhancements for individual 0990CL protons reflect the proximity to the binding site of Gα_i1. However, the T₁-relaxation of individual protons and exchange kinetics can further complicate the analysis of the binding epitope.^50,68 Therefore, we explicitly accounted for differential longitudinal relaxation times of 0990CL protons ranging from 2.284 for the aromatic H4″ to 0.422s measured for the aliphatic 2′-Me (Fig 5A).⁵⁰ Normalized STD enhancements (setting one H3′ proton to 100%) at a saturation time of t_sat = 3 s are summarized in Figure 5C. The largest STD effects can be observed for the two aliphatic H3′ protons followed by the two 2′-Me and the 4′-Me groups. In the aromatic region, more moderate interactions of the compound with Gα_i were also observed for both aromatic ring systems involving protons 2′ through 6′ and H5, 6, 7 and 8. Inspection of mixtures of Gα_i1 in its GDP-bound state together with 0990CL and 8005 revealed that all of the signals of the 8005 negative control were absent in the difference spectrum indicating that this compound does not bind Gα_i1 (Fig. S2).

Figure 5.

Saturation transfer difference (STD) NMR of Gα_i1 and 0990CL. (A) Chemical graph depicting inhibitor compound 0990CL, proton numbering scheme for STD NMR and T₁-relaxation times (italic). (B) Enlargement of the aliphatic region of the reference and STD NMR spectrum of 0990CL with Gα_i. Increases in STD signals are observed with saturation times (τ_sat) of 1 s (blue) and 3 s (red) and 0990CL proton assignments are indicated. (C) Tautomer Gα_i inhibitor 0990 docked with Gα_i GDP. T₁-compensated STD effects observed using τ_sat = 3 s were normalized to H3′ which was set to 100%. (D) Enlargement of the aromatic region of the reference and STD NMR spectrum of 0990CL with Gα_i. Increases in STD signals are observed with saturation times (τ_sat) of 1 s (blue) and 3 s (red) and 0990CL and H8(GDP)-proton assignments are indicated.

3.4. Structure–activity relationship survey of 0990CL

After binding confirmation by NMR, compound 0990CL was selected for further validation. Six 0990CL congener compounds were subsequently purchased to explore substituents on the aniline and phenyl rings of a quinazoline scaffold (Supp. Table 1). Compound 9585, with a methyl group at positions 6 and 5′ and compound 9586, with a methyl at positions 6 and 3′, both displayed >50% inhibition of nucleotide exchange with Gα_i. Compound 9587, with a chloride at 5′ and a methyl at position 6, displayed only weak inhibition of nucleotide exchange on Gα_i (∼16%). These results suggest that hydrophobic interactions on the aniline ring enhance exchange inhibition. Compound 9253, with a carbonyl at the position 2′, a methyl at 4′, and methyl groups at positions 4 and 6, proved ineffective at influencing nucleotide exchange. Also ineffective was compound 8358 with methyl groups at the 4 and 4′ position. These two results emphasize the need for the phenyl substituent at position 4 for exchange inhibition. Compound 6784, with a hydroxyl at the 5′ and a phenyl and methyl at positions 4 and 6, respectively, also showed no activity towards nucleotide exchange inhibition. These data indicate the requirement for hydrophobic substituents on the aniline ring and the phenyl ring at position 4.

Our data thus far confirmed the quinazoline scaffold as amendable to GDI activity. We went further and synthesized and tested ten additional congeners or fragments based on compound 0990CL (Supp. Table 1). Compound 2397, with a chloride at the 3′ position and a phenyl at position 4, and compound 2395, with a phenyl at position 4 both resulted in >30% inhibition of nucleotide exchange on Gα_i. Compounds 2353 and compound 2369 resulted in 23% and 18% inhibition of Gα_i nucleotide exchange, respectively, suggesting that the quinazoline scaffold enhances exchange inhibition. Compound 2410, with only a single nitrogen on the aniline ring and phenyl at position 4; and 2437, with methyl groups at positions 3′ and 5′ and a phenyl at position 4 were both inactive toward nucleotide exchange. These results suggest that both nitrogens are likely required for compound interaction. Compound 2471, with carbonyl linker connecting the quinazoline and the aniline ring, was inactive indicating that the phenyl-quinazoline-aniline core structure enhances compound interaction. Compound 2621, with a methoxy group at position 3′ and a phenyl at position 4, was inactive. Other inactive non-quinazoline type compounds included 2355, with just a portion the nitrogen trio of the quinazoline-aniline structure; and 8910, exploring this same feature, emphasize the need for the basic phenyl-quinazo-line-aniline root structure.

3.5. Molecular docking study of 0990CL

To further analyze the potential properties of the inhibitor/G-protein interaction, we performed molecular docking simulations on compound 0990CL. The top pose of interest for 0990CL interacted with both Arg¹⁷⁸ and Val¹⁹⁹ (Fig. S4). Energy minimization of the system which allowed the G-protein and nucleotide to flex had no effect on the final pose. Other high probability interactions were the central secondary amine of 0990CL donating a proton to Glu⁴³ or Gln⁷⁹ and H-arene interactions with Gln⁷⁹ or Lys¹⁸⁰. As we had hoped, these interactions mimic much of the biochemistry of the GPR/Goloco interactions with a mix of hydrophobic and ionic interactions. These poses also indicate that the unmodified phenyl is not planer with the rest of the molecule. Also the central secondary amine is a flexible hinge allowing the overall molecule to be convex. These features help to create a molecule with three dimensional topology to present both hydrophobic and proton donating features.

Through our synthetic, informatics, and structural characterization we observed tautomerization of 0990CL. Specifically, the 3′ carbon (Fig. 5A and C and Fig. S4) on the terminal pyrimidine is capable of switching between alternate sp² or sp³ hybridization. Tautomer T1 was appealing as our earlier pose predictions indicated the potential for the Arg¹⁷⁸ guanidinium group donating a hydrogen bond. However, our STD NMR experiments indicated the sp³ tautomer was the primary interacting form in our experiments. To further understand the consequences of tautomerization of 0990CL for interaction with Gα_i, we tested the two tautomers in molecular simulations (Fig. S5). Using the same GDP site as used in the primary virtual screening, both 0990CL tautomers were docked to Gα_i. Docking scores for the top 25 poses between the two were nearly identical. Further investigating the poses, we recognized that the poses were flipped with respect to each other and the GDP binding site. Tautomer T1 consistently had the trimethylpyrmidine buried while T2 had the phenyl portion of the quinazoline buried (Fig. S5C). While both tautomers were able to form effective bonds with Val¹⁷⁹ in the simulations, T2 was able to form additional bonds with GDP. Lastly, we took the top poses for each tautomer and calculated the potential energy for the complex and individual components (Fig. S5D). The poses predicted from the docking simulations gave the lowest energies while swapping the tautomers gave poorer energies. Interestingly, T1 in the T2 pose was very unfavorable in terms of energy, indicating T2 in its native pose is the highest affinity interaction possible.

3.6. Effect of GDIs on cellular cAMP

We also assessed the cellular effects of small molecule Gα_i GDIs on cAMP levels using the α₂AR as a cellular model (Fig. 6 and Fig. S3). 0990CL was able to partially restore cAMP levels following UK14304 treatment. Restoration was greatest at the lowest concentration of 0990CL (100 nM), which was able to restore ∼31% of the reduction of cAMP elicited by Fsk and UK14304. At higher concentrations of 0990CL (10 μM), cytotoxic effects of the compound began to negatively impact cell viability. The negative control compound 2355 showed no effect upon cAMP levels following treatment with Fsk and UK14304 which demonstrates our inhibitors likely do not interfere with the ability of UK14304 to activate the α₂AR. Also of note, 8770 and 2967 were able to suppress cAMP signaling though α₂AR in a concentration-dependent manner. Compounds 8770 and 2967 demonstrated a lack of toxicity at 10 μM indicated by measured max efficacy of Fsk in the presence of UK14304. The fact 8770 did not inhibit Gα_i1 in our nucleotide exchange assay yet shows in vitro efficacy highlights the necessity to conduct multiple assays to fully characterize compounds as inhibitors in any drug screening platform.

Figure 6.

Effect of small molecule Gα inhibitors on α₂AR/Gα_i mediated cAMP suppression. cAMP levels were measured in HEK293 GloSensor cell line was transfected with 0.2 lg/mL α₂AR receptor. Cells were pre-treated prior to the experiment with 100 ng/mL PTX (24 h), 0.1, and 1 μM 0990CL (20 min) or 0.1, 1, and 10 μM 8770, 2967, and 2355 (20 min). Cells were subsequently treated with 250 nM Fsk or Fsk and 10 μM of the α₂AR agonist UK14304. Data was derived from triplicate determinates from three independent experiments.

4. Discussion

Heterotrimeric G-proteins are central regulators of cellular communication, and potent and selective inhibitors can serve multiple purposes ranging from being a pharmacologic probe to experimental therapeutics. As G-proteins are signaling amplifiers, they are an ideal target for therapeutic intervention and diagnosis. Here we presented the rationale, techniques, and preliminary compounds aimed at developing selective GDIs of Gα subunits.

The TAT-GPR peptide and our small molecule GDIs were designed to stabilize the GDP bound confirmation of Gα_i by blocking nucleotide release and also interacting G-protein isoform specific residues around the nucleotide binding pocket. We determined that appending a TAT leader sequence was able to make the GPR peptide cell permeable without suppressing nucleotide exchange activity. Tagging with FGF was able to permeabilize the GPR peptide, but inhibited nucleotide exchange activity. We then determined that compared to PTX, TAT-GPR was very potent at preventing UK14304 stimulated decreases in cAMP.

Using computational docking we probed the Gα_i–GDP binding using 280,000 diverse compounds. We also used Gα_i–GTP and Gα_q–GDP as computational counter-screens. From these simulations, we determined 210 potential Gα_i–GDP interacting compounds. We desired to find compounds that stabilized the GDP-bound confirmation of Gα_i and therefore directly measured nucleotide exchange. The standard nucleotide exchange assay was modified by optimizing the emission bandpass to avoid fluorescence interference from test compounds. Fifteen compounds were tested in the first phase of validation and eight were active for Gα_i, although four were promiscuous and inhibited nucleotide exchange on Gα_i and Gα_q. 0990CL, 8770 and 2967 small molecular Gα_i inhibitors showed varying degrees of efficacy to increase cAMP. From these data, a quinazoline scaffold emerged as tractable for Gα_i inhibition. The structure–activity relationship (SAR) of the seven quinazoline derivatives indicated the critical need of both hydrophobic interactions provided by compound methyl groups and hydrogen bond donors provided by compound amine groups. We confirmed compound interaction by STD NMR. Data from the exchange assay, SAR, NMR, and docking indicate that the 4′-Me moiety, and to a lesser degree, the 2′-Me moiety, from 0990CL are involved in specific hydrophobic interactions. There are also significant compound interactions in the aromatic region and implicate the pyrimidine secondary amine participates in hydrogen bonding. The compound 0990CL SAR did not indicate an obvious rout to substantial improvement, and future characterization of hits such as 9595 and 9586 might provide better compounds in the long run. However, these compounds do demonstrate the feasibility of small molecule heterotrimeric Gα GDIs.

GDIs are well suited to pursue receptor pharmacology of Gα_i versus Gβγ signaling and their specific roles in cell signaling and disease.⁶⁹ A unique aspect of Gα–GDIs is their ability to selectively inhibit Gα with the potential to stimulate Gβγ signaling. This facet, along with Gβγ directed pharmacologic tools and PTX, will allow addressing the contribution of Gα, Gβγ, and Gαβγ signaling in living cells without alterations in protein stoichiometry. Several pharmacologic regulators of Gβγ have been developed including the SIGK peptide which dissociates the heterotrimer creating free Gα and Gβγ.^70,71 The phosducin peptide inhibits Gβγ but allows Gα to signal.⁷² With these tools, pharmacologic regulation of both receptor-dependent and -independent signaling of individual heterotrimeric subunits can provide a unique window into this major signaling axis. Therapeutically, Gα subunit inhibition by GDI has the advantage of circumventing the need to directly address the upstream component of GPCR-related signaling in cases of mutations, polymorphisms, and expression-related defects often seen in disease.

Abbreviations

AC

adenylyl cyclase

α₂AR

α₂-adrenoceptor

cAMP

cyclic adenine monophosphate

Fsk

forskolin

GDI

guanine nucleotide dissociation inhibitor

GDP

guanine diphosphate

GPCRs

G-protein coupled receptor

GPR

G-protein regulatory

GTP

guanine triphosphate

GTPγS

guanosine 5′-O-[gamma-thio]triphosphate

NMR

nuclear magnetic resonance

PTX

pertussis toxin

PLC

phospholipase-C

SAR

structure-activity relationship

STD

saturation transfer difference

Tc

Tanimoto coefficient