Small-Molecule Ligands as Potential GDNF Family Receptor Agonists.

To find out potential GDNF family receptor α1 (GFRα1) agonists, small molecules were built up by molecular fragments according to the structure-based drug design approach. Molecular docking was used to identify their binding modes to the biological target GFRα1 in GDNF-binding pocket. Thereafter, commercially available compounds based on the best predicted structures were searched from ZINC and MolPort databases (similarity ≥ 80%). Five compounds from the ZINC library were tested in phosphorylation and luciferase assays to study their ability to activate GFRα1-RET. A bidental compound with two carboxyl groups showed the highest activity in molecular modeling and biological studies. However, the relative position of these groups was important. The meta-substituted structure otherwise identical to the most active compound 2-[4-(5-carboxy-1H-1,3-benzodiazol-2-yl)phenyl]-1H-1,3-benzodiazole-5-carboxylic acid was inactive. A weaker activity was detected for a compound with a single carboxyl group, that is, 4-(1,3-benzoxazol-2-yl)benzoic acid. The substitution of the carboxyl group by the amino or acetamido group also led to the loss of the activity.

Introduction

Glial cell line-derived neurotrophic factor (GDNF) family ligands (GFLs) consist of GDNF, neurturin (NRTN), artemin (ARTN), and persephin (PSPN).[1] GDNF and NRTN have demonstrated the ability to support the survival of brain dopamine-producing neurons, thus being potential therapeutic agents for Parkinson’s disease.[2−4] The survival of sympathetic and sensory neurons is supported by ARTN, and hence it has been considered for the treatment of chronic pain.[5] However, the development of therapies for neurological diseases based on GFLs has serious problems associated with the delivery, stability, and potential side and off-target effects of these ligands.[6] Therefore, the discovery of small molecules that bind to and activate GFL receptors would have large potential for the development of new strategies against neurodegenerative disorders.[6−8] GDNF specifically binds to GDNF family receptor α1 (GFRα1), and then the complex binds to and signals through the transmembrane receptor tyrosine kinase RET. ARTN binds to GFRα3, NRTN to GFRα2, and also signal to the cells via RET.[1] Presently, structures of two scaffolds acting as the GDNF receptor agonists have been reported, XIB4035[9] and BT13[10] (Figure ). Also BT13 derivative, compound called BT18 activates RET.[11]

Figure 1

Chemical structures of XIB4035 and BT13.

The aim of the current work was to find out new low molecular weight compounds acting as GFRα1 receptor agonists using the structure-based drug-design approach and molecular docking. The potential binding site for such agonists was searched by examining the protein–protein interactions on the binding interface between GDNF and GFRα1 in the GDNF–GFRα1–RET complex. A strong binding of a small-molecule ligand in this site could lead to the conformational changes and to the RET signaling.

Results and Discussion

GDNF–GFRα1 Interface

Essential interactions at the binding interface between the GDNF and GFRα1 (i.e., hydrogen bonds and hydrophobic areas) were examined using the available crystal structures of the complex (PDB codes: 4UX8(12) and 3FUB;[13] see Target in Methods). GFRα1 was treated as a receptor and GDNF as a ligand. There are several notable regions of the ligand–receptor interaction. First, three hydrogen bonds are formed by amine group of Arg171 and Arg224 (GFRα1) with oxygens of carboxylate group of Glu61 (GDNF). Three hydrogen bonds are also formed involving amine groups and carbonyl oxygens of peptide bond between Asn162 (GFRα1) and Glu62 (GDNF), Asn162 (GFRα1) and Ser112 (GDNF), and Arg224 (GFRα1) and Leu114 (GDNF). The hydroxyl groups of Ser172 (GFRα1) and Tyr120 (GDNF) are also hydrogen bond donors to the carbonyl oxygens of peptide bonds of Asp110 and Gln227, respectively (Figure ). In addition, potential van der Waals interactions between the side groups of neighboring residues of the ligand, and the receptor can stabilize the binding complex (see Table S1 and Figure S1 in the Supporting Information). GDNF docking results are well in line with the previously published GFRα1[14,15] and GDNF[16] mutagenesis studies. Our assumption was that low-molecular-weight ligands (MW < 500) have to bind to residues of amino acids of GFRα1, therefore imitating the native ligand GDNF. According to the analysis of binding site for the GFRα1–GDNF complex, the key residues for interactions with a small molecule are Arg171 and Arg224 residues of the receptor which is in line with the previously published data.[14,15] Consequently, the ligand acting as a potential receptor agonist has to be bound in the region of Thr179, Arg171, Arg224, and Gln227, forming hydrogen bonds to Arg171, Arg224, and/or Gln227.

Figure 2

Analysis of interactions between GFRα1 and GDNF residues (PDB code: 4UX8). The amino acid residues in the interface of GFRα1 (green) and GDNF (blue) are colored as gray (carbon), blue (nitrogen), red (oxygen), and white (hydrogen). Hydrogen bonds are represented in green dashed lines.

Ligands Construction and Docking

The ligands were built up using molecular fragments that have significant docking binding to these receptor areas [see structure-based drug design (SBDD) approach in Methods]. The first group of ligand structures consisted of fragments of benzoate anion and its alkyl-substituted (Me, Et, and i-Pr) derivatives at the meta position. The length of the carbon chain (CH2) of the carboxylate substituent was also varied, with n = 0, 1, and 2 (Figure A). Thereafter, the heterocyclic derivatives were generated by the substitution of carbon atoms by up to two nitrogen atoms in the benzene ring (Figure B). The calculated binding energies for the studied 28 molecules from the first group of structures (Table S2) were in the range of −4.3 to −5.6 kcal/mol. Only one of the heterocyclic compounds (25) was able to demonstrate substantial specific binding, as illustrated in Figure A, where the nitrogen atom of this compound is hydrogen bonded to the amine group of Arg171 and has additional van der Waals interactions with residues of Ile175, Thr179, Met211, Gln227, and Thr228. All results from the molecular docking calculations, that is, the corresponding binding energies and binding modes, are given in Table S2 of the Supporting Information.

Figure 3

Structures of low-molecular-weight ligands used in the molecular docking to receptor GFRα1.

Figure 4

Calculated binding modes of compound 25 (A), compound 42 (B), compound 77 (C), and compound 86 (D) in the active site of GFRα1 (PDB code: 4UX8). The amino acid residues of GFRα1 are colored as gray (carbon), blue (nitrogen), red (oxygen), and white (hydrogen). Hydrogen bonds formed between compound and residues of GFRα1 are represented in green dashed lines.

The second group of the constructed GFRα1 receptor ligands included 46 structures, where (i) at the R2 position, hydrophobic alkyl group in the cyclic structures was substituted by hydroxyl, hydroxymethyl, amino, or aminomethyl groups (Figure C) enabling to form an additional hydrogen bond; (ii) more flexible cyclohexene (Figure D) or cyclohexanone basic structures (Figure E) were also used instead of rigid aromatic rings; and (iii) the carboxylate group was replaced by sulfonate group (Figure F), which can be strongly bound to proteins. The free energy of binding for all compounds was similar, which is in the range of −4.4 to −5.2 kcal/mol. Compounds consisting of more flexible cyclohexene or cyclohexanone rings as well as the sulfonate group did not have significantly lower binding energy. Most of the ligands of this subclass formed hydrogen bonds with the receptor by the carboxyl oxygens and/or by the added polar substituent at the R2 position, by the nitrogen atom of the aromatic cycle, and by an oxygen atom of the sulfonate group (Table S2). As an example, compound 42 has two hydrogen bonds between its carboxyl oxygens and NH2 group of Gln227 as well as between its CH2OH group and NH group of Arg171 in addition to the interactions with Ile175, Met211, Arg224, and Thr228 (Figure B).

The third group of developed GFRα1 receptor ligands was built using the following fragments: (i) flexible 5-oxocyclohex-3-ene-1-carboxylate, (ii) hydroxyl group, and (iii) hydroxyalkyl group (n of alkyl chain (CH2) = 0–4; Figure G) or (2-hydroxyalkylcyclopropyl)methyl group (n of alkyl chain (CH2) = 0–2; Figure H). The binding energies for these eight compounds ranged from −5.0 to −6.1 kcal/mol (Table S2). Figure C demonstrates the binding mode of compound 77, which forms three hydrogen bonds by its carboxyl oxygens with amine group of Arg224, by hydroxypropyl group with the amine group of Gln227, and by hydroxyl group with carbonyl oxygen of Thr228 as well as van der Waals interactions with nearby residues of Arg171, Ile175, Met211, and Val230 of GFRα1.

According to the docking results on compounds from all these groups, we can conclude that (1) compounds with rigid aromatic ring(s) give somewhat lower binding energies than compounds with flexible cycles; (2) besides carboxyl functional group, hydroxyl group and its analogues as well as nitrogen atom of the heteroaromatic cycle may lead to additional hydrogen bonds with amino acid residues of GFRα1. Proceeding from these observations, the fourth group of 10 structures was constructed from fragments based on quinoline-3-carboxylate or isoquinoline-3-carboxylate, hydroxyl or hydroxymethyl group, and (2-hydroxyalkylcyclopropyl)methyl group (Figure I,J). This group showed the lowest binding energies: −5.5 to −6.4 kcal/mol (Table S2). Surprisingly, only one hydrogen bond was formed through carboxyl oxygens of ligand to the amine group of Arg224 as demonstrated in Figure D for compound 86, which has also van der Waals contacts with Arg171, Thr179, Met211, Gln227, and Thr228.

Testing of Potential GFRα1 Receptor Agonists

On the basis of the last group of structures, similar compounds were searched from ZINC[17] and MolPort[18] databases (see Table S2). Five available compounds from the ZINC library (Figure ) were initially tested in 5 and 20 μM concentrations in luciferase assay in cells expressing GFRα1–RET.[19] Those ones activating luciferase by 1.5 times and above (compounds 107 and 118) were selected for dose-dependent experiments and RET phosphorylation (RET pY) assays (see Experimental Section); the remaining compounds were considered inactive (fold induction below 1.5 at 20 μM) and excluded from further experiments. As predicted by the modeling, the bidental compound 107 with two carboxyl groups has the highest activity. Notably, a weak activity was detected for a much smaller compound 118 with a single carboxyl group in its structure. However, the relative position of these groups is important. The meta-substituted structure 108 otherwise identical to compound 107 is inactive. The substitution of the carboxyl group in compound 118 by the amino or acetamido group (compounds 119 and 120, respectively) also leads to the loss of the activity.

Figure 5

Compounds tested in luciferase assay in cells expressing GFRα1–RET.

Thus, compounds 107 and 118 were tested in luciferase assay in six concentrations (0.1, 1, 5, 10, 25, and 50 μM). Notably, compound 118 is a close derivative of the compound XIX described in the patent.[20] Application of 107 to GFRα1–RET expressing cells led to moderate activation of luciferase [one-way analysis of variance (ANOVA) F(12,39) = 33.11] in 25 μM (1526 ± 77.61 vs 514.8 ± 43.77 in control, P < 0.0001, one-way ANOVA with Dunnett’s post hoc test) and 50 μM concentrations (1963 ± 193.1 vs 514.8 ± 43.77 in control, P < 0.0001, one-way ANOVA with Dunnett’s post hoc test), application of 50 μM 118—to borderline activation of luciferase (956.3 ± 78.89 vs 514.8 ± 43.77, P = 0.0063, one-way ANOVA with Dunnett’s post hoc test) (Figure ). To confirm luciferase assay results and evaluate the ability of these compounds to stimulate RET by direct methods, the selected compounds were tested in RET pY assays using western blotting. RET was immunoprecipitated, and western blotting membranes were probed first with antibodies against phosphorylated tyrosine residues (pY) and afterward—against RET protein to evaluate protein loading. Resulting images were quantified using Visual Studio software. Compound 118 failed to activate RET pY (relative intensity 0.70 ± 0.17 in control vs 0.64 ± 0.11 in compound-treated cell lysates, N = 3). Compound 107 activated RET pY in the cells expressing GFRα1–RET but not in the cells expressing only RET (Figure A–C). Repeated measurements (RMs) ANOVA showed statistically significant differences between treatment groups in RET pY assay in both GFRα1–RET (P = 0.0020, F(1.120,5.601) = 28.31) and GFP–RET (P = 0.0219, F(1.069,4.275)) expressing cells (Figure B,C). Post hoc Dunnett’s test revealed that compound 107 increased RET pY by 45% (relative intensity 0.68 ± 0.17 in control vs 0.99 ± 0.26 in compound-treated group, P = 0.0405, RM ANOVA with Dunnett’s post hoc test) in GFRα1–RET, but not in GFP–RET expressing cells (relative intensity 0.42 ± 0.16 in control vs 0.34 ± 0.077 in compound-treated group, P = 0.693 RM ANOVA with Dunnett’s post hoc test). As expected, the positive controls (GDNF in GFRα1–RET and soluble GFRα1–GDNF complex in GFP–RET expressing cells) increased RET pY by 3.8-fold (P = 0.0042, RM ANOVA with Dunnett’s post hoc test) and 4.9-fold (P = 0.0465, RM ANOVA with Dunnett’s post hoc test) in GFRα1–RET and GFP–RET expressing cells, respectively. These results indicate that compound 107 might, similarly to GDNF protein, require the presence of GFRα1 coreceptor to stimulate RET pY.

Figure 6

Dose-dependent activation of luciferase reporter gene in the cells expressing GFRα1–RET receptor complex by compounds 107 and 118. Compounds are analyzed in quadruplicates. The results are presented as mean ± SEM. ****P < 0.0001, **P < 0.01, one-way ANOVA with the Dunnett’s post hoc test. CTR—control.

Figure 7

RET pY by compound 107 (100 μM) in MG87RET fibroblasts transiently transfected with GFRα1 (B) and GFP (C). (A) Image of western blotting analysis. Molecular weight markers (in kDa) are indicated on the left. (B,C) Quantification of western blotting data. Quantitative data are presented as mean ± SEM from 6 (B) or 5 (C) independent experiments. *P < 0.05, **P < 0.01, RM ANOVA with the Dunnett’s post hoc test. As a positive control, we used GDNF (100 ng/mL) in cell transfected with GFRα1 (B) and soluble GFRα1 (1 mg/mL) and GDNF (100 ng/mL) (a1/GDNF) in GFP-transfected cells (C). WB—western blotting, IP—immunoprecipitation.

The discrepancy between luciferase assay and RET pY assay data for compound 118 can be explained by the differences in the sensitivity of these assays. Luciferase assay is extremely sensitive and accumulates signals for the prolonged period of time that can include multiple cycles of RET and downstream signaling cascade activation, whereas RET pY assay is less sensitive and reflects activation of the receptor that is achieved at the moment of cell lysis. Another possibility is RET independent activation of luciferase in response to compound 118. Luciferase assay is designed to represent an activation of mitogen-activated protein kinase (MAPK) signaling pathway.[19] This pathway includes multiple regulatory proteins that can be targeted by compound 118.

All corresponding binding modes obtained by molecular docking are illustrated in Figure . The most active compound 107 forms hydrogen bond by its carboxyl oxygens with NH group of Arg224 and interacts with Thr180, Ala220, Gln227, and Val230 (Figure A). The another potential GFRα1 receptor agonist 118 has also one hydrogen bond between its carboxyl oxygens and amine group of Arg224 as well as van der Waals interactions with residues of Arg171, Met211, Gln227, Thr228, and Val230 (Figure C). Although inactive compounds 108 and 120 also form hydrogen bonds (Figure B,E), their molecular configuration can be different in solution and there is no binding. For instance, dipole moments μ are different for para- and meta-substituted structures 107 and 108 (μpara < μmeta), respectively.

Figure 8

Calculated binding modes of compound 107 (A), compound 108 (B), compound 118 (C), compound 119 (D), and compound 120 (E) in the active site of GFRα1 (PDB code: 4UX8). The amino acid residues of GFRα1 are colored as gray (carbon), blue (nitrogen), red (oxygen), and white (hydrogen). Hydrogen bonds formed between compound and residues of GFRα1 are represented in green dashed lines.

MD Simulations of GFRα1–Ligand Complexes

To specify the nature of the ligand–protein interactions, the molecular mechanics/Poisson–Boltzmann surface area (MM/PBSA)[21] binding energy calculations were carried out using data from molecular dynamics simulations. In MM/PBSA, the free energy of a state (ligand or protein) is estimated from the following sumwhere the first three terms are standard molecular mechanics energy terms from bonded (bond, angle, and dihedral), electrostatic, and van der Waals interactions. Gpol and Gnp are the polar and nonpolar contributions to the solvation free energies, respectively. Gpol is typically obtained by solving the Poisson–Boltzmann equation or by using the generalized Born model (giving the MM/GBSA approach), whereas the nonpolar term is estimated from a linear relation to the solvent accessible surface area. The last term in the above equation is the absolute temperature, T, multiplied by the entropy, S, estimated by a normal-mode analysis of the vibrational frequencies. The results for the lowest binding energy states for the complexes of compounds 107, 108, and 118 with GFRα1 at the GDNF–GFRα1 interface are given in Table . The binding free energy for the active para-substituted compound 107 is by 19.0 kcal/mol and by 9.4 kcal/mol lower than that for the inactive meta-substituted compound 108 and for compound 118 with weaker activity, respectively.

Table 1

Binding Free Energies of the GFRα1–Ligand Complexes Calculated Using the MM/PBSA Methoda

	107	108	118
energy term	GDNF–GFRα1 interface	GDNF–GFRα1 interface	GDNF–GFRα1 interface
ΔEvdW	–37.01	–23.49	–23.04
ΔEel	–137.78	–145.68	–96.96
ΔEpol	137.94	151.83	97.82
ΔEnp	–19.37	–10.77	–16.79
ΔEbond	–4.98	–5.29	–5.35
ΔGbind	–53.38	–34.35	–44.02

All energies in kcal/mol.

The binding mode of compound 107 was also confirmed by the counterpart molecular dynamics modeling. The simulations carried out at the GFRα1 interface to GDNF indicated strong hydrogen bonding of the carboxylate groups of the ligand to the receptor residues Arg224 and Thr180 (Figures and S2). This binding is complemented by the water-assisted binding of the nitrogen atoms of the benzimidazole rings to residues Glu223 and Glu227.

Figure 9

Calculated binding mode of compound 107 by the molecular dynamics simulations.

It should be noted that this compound is very different from the earlier reported active compounds, that is, XIB4035[9] and BT13.[10] XIB4035 is unable to activate RET in the absence of endogenous ligand, that is, GFL,[22] and BT13 is the RET agonist that is able to signal in the absence and presence of GFRα coreceptors. Therefore, compound 107 seems to represent a new scaffold that can be further optimized to develop efficient GFRα1 agonists. Although weak biological activity of 107 makes the evaluation of its interaction with GFRα1 difficult, biological data indicate that it might be the first compound activating GDNF receptor system via GFRα1 coreceptor in the absence of endogenous GFL. Thus, further biochemical and cellular testing of such compounds have large biomedical interest.

Methods

Structure-Based Drug Design

When the three-dimensional structure of biomolecular target is known, the SBDD approach can be applied in the drug discovery and development process.[23,24] The task is to design small molecules (receptor agonists) that will fit to the binding pocket of the target (hydrophobic surfaces, hydrogen bonding sites, etc.), and the binding affinity will be predicted by a fast approximate docking program. Ligands are built up by molecular fragments using “linkers” or “scaffolds”, if necessary, within the constraints of the binding site in the case of de novo method. The compounds constructed this way have to correspond to Lipinski’s rule of five.[25]

Target

We proceeded from two available crystal structures on the GDNF–GFRα1 complex downloaded from Protein Data Bank. The hybrid structural model of reconstituted mammalian GDNF–GFRα1–RET complex (i.e., RET ternary complex; code: 4UX8) had been derived from electron microscopy and low-angle X-ray scattering data with a resolution of 24.0 Å.[12] The protein consists of chain A and chain B (RET; residues 29-508), chain C and chain E (GFRα1; residues 6-348), and chain D and chain F (GDNF; residues 42-134). The crystal structure of the GDNF–GFRα1 tetrameric complex (code: 3FUB) was measured by X-ray diffraction with a resolution of 2.35 Å.[13] The asymmetric unit contains two chains of GFRα1 (chain A, residues 150-348; chain C, residues 150-348) and two chains of GDNF (chain B, residues 40-134; chain D, residues 32-134).

Both raw crystal structures were corrected, and hydrogen atoms were automatically added to the protein using Schrödinger’s Protein Preparation Wizard of Maestro 10.7.[26] AutoDockTools (ADT)[27] 1.5.6 was used to identify the binding interface between the chains of GDNF and GFRα1. As hydrogen bonds and van der Waals interactions were largely identical in both crystal structures, thus, chains C and D of GDNF–GFRα1–RET complex (code: 4UX8) was used for the further study. Water molecules were removed from the crystal structure.

Small-Molecule Library

The initial data set for virtual fragment-based docking screening was constructed by molecular fragments in a way that compounds would be bound to the biological target GFRα1 similarly to GDNF. Commercially available compounds based on these structures and using Tanimoto similarity coefficient (≥80%) were searched from ZINC[17] and MolPort[18] databases. The two-dimensional chemical structures of ligands were converted into three-dimensional structures and preoptimized by molecular mechanics MM+ field using HyperChem 8.0.[28] The “Online SMILES Translator and Structure File Generator”[29] program was used to create pdb files for molecular docking procedure.

Molecular Docking

AutoDock Vina 1.1.2[30] was used for the docking studies to find out binding modes and binding energies of ligands to the receptor. The number of rotatable bonds of ligand was set by default by ADT.[27] However, if the number was greater than 6, then some of rotatable bonds was made as nonrotatable, otherwise calculations can be inaccurate.[31] The active binding site on GFRα1, obtained by the removal of GDNF (chain D), was surrounded with a grid box sized 30 × 30 × 30 points with a spacing of 1.000 Å. The settings used for the iterated local search global optimizer based on mutation and local optimization steps, accepted or rejected with a Metropolis criterion in Vina, were nine modes, one central processing unit, and an energy range of 1 kcal/mol. Other settings were used as default.

Molecular Dynamics

The molecular dynamics simulations were carried out using Desmond simulation package of Schrödinger LLC.[32] The NPT ensemble with the temperature of 300 K and a pressure of 1 bar was applied in all runs. The simulation lengths were 10 and 50 ns with a relaxation time of 1 ps. The OPLS_2005 force field parameters were used in all simulations.[33] The long-range electrostatic interactions were calculated using the particle mesh Ewald method.[34] The cutoff radius in Coulomb’ interactions was 9.0 Å. The water molecules were described using simple point charge model.[35] The Martyna–Tuckerman–Klein chain coupling scheme[36] with a coupling constant of 2.0 ps was used for the pressure control and the Nosé–Hoover chain coupling scheme[36] for the temperature control. Nonbonded forces were calculated using an r-RESPA integrator where the short-range forces were updated every step and the long-range forces were updated every three steps. The trajectories were saved at 4.8 ps intervals for analysis. The behavior and interactions between the ligands and protein were analyzed using the Simulation Interaction Diagram tool implemented in Desmond molecular dynamics package. The stability of molecular dynamics simulations was monitored by looking on the root mean square deviation of the ligand and protein atom positions in time.

Experimental Section

Compounds

Experimentally studied compounds were purchased from MolPort Inc.[18]

Proteins

GDNF was obtained from Icosagen Ltd. (Cat# P-103-100).

Plasmids

Full-length human GFRα1 cDNA[19] subcloned in pCDNA6 (Invitrogen), full-length human RET (long isoform) in pCR3.1 (Invitrogen),[37] enhanced GFP-expressing vector—pEGFP-N1 (Clontech, Cat# 6085-1, discontinued), and PathDetect Elk-1 system (Stratagene) to detect MAPK activation.

Cell Lines

MG87RET murine fibroblasts stably transfected with RET proto-oncogene.[16] Reporter gene systems to detect MAPK activation: MG87RET stably transfected with PathDetect Elk-1 and GFRα1 or empty vector.[19]

Luciferase Assays

To identify compounds activating GFLs receptors and check their ability to activate intracellular signaling via RET, we used previously developed reporter-gene-based system (MG87 murine fibroblast stably transfected with PathDetect Elk-1, GFRα1, and RET).[19] The day before the experiment, the reporter cells were plated into 96-well plates (PerkinElmer) at 175 000 to 200 000 cell/mL density in Dulbecco’s Modified Eagle’s medium (DMEM), 10% fetal bovine serum, 100 μg/mL Normocin (InvivoGen, Cat# ant-nr-1), 1% dimethyl sulfoxide (DMSO), and 15 mM Hepes pH 7.2. The next day compounds or proteins under study were applied to the cells in desirable concentrations. The following day, the cells were lysed and luciferase activity was measured using the neolite luciferase detection reagent (PerkinElmer, Cat# 6016711). The luminescence was measured using a plate luminometer (Optima FW, Thermo Scientific). At first compounds were tested in triplicates in two concentrations (5 and 20 μM). Compounds activating luciferase in initial screen by 1.5-fold or above at least in one concentration were further subjected to the analysis in 6 concentrations (0.1, 1, 5, 10, 25, and 50 μM). Dose-dependent studies were made in quadruplicates.

RET Phosphorylation Assay

MG87RET cells were plated on 35 mm tissue culture dishes 2 days before the experiment to achieve 90–95% of confluency of the cells at the day of stimulation with tested substances. The next day, cells were transfected with 4 μg/well of GFRα1- or GFP-expressing plasmid using Lipofectamine 2000 (Invitrogen) for DNA delivery as described by the manufacturer. On the third day, cells were starved in serum-free DMEM containing 15 mM Hepes, pH 7.2, and 1% DMSO for 4 h and stimulated with compounds or GDNF. Then, cells were washed once with ice-cold phosphate-buffered saline containing 1 mM Na3VO4 and 1 mM NaF and lysed on ice in 0.5 mL per well of RIPA-modified buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM ethylenediaminetetraacetic acid (EDTA), 1% NP-40, 1% TX-100, 10% glycerol, EDTA-free protease inhibitor cocktail (Roche), 1 mM Na3VO4, 2.5 mg/mL of sodium deoxycholate, 1 mM NaF). Plates were incubated at +4 °C on the horizontal shaker for 30 min with vigorous shaking; lysates were collected to the Eppendorf tubes and centrifuged for 10 min at +4 °C at 13 000 rpm to precipitate cell debris.

To immunoprecipitate RET cell lysates were incubated overnight at +4 °C on the round rotator in the presence of 1 μg/mL of goat anti-RET C-20 antibodies (Santa Cruz Biotechnology Cat# sc-1290, RRID:AB_631316) and magnetic beads conjugated with protein G (Dynabeads, Thermo Fisher Scientific, Cat# 10004D). Beads were washed three times with 1× tris-buffered saline (TBS) with 1% triton X-100; bound proteins were eluted by 100 μL of 2× Laemmli loading buffer, resolved on 7.5% sodium dodecyl sulfate polyacrylamide gel and then transferred onto a nitrocellulose membrane. Membrane was blocked for 15 min at room temperature with TBS-T (1× TBS containing 0.15% Tween 20) containing 10% skimmed milk and probed with antiphosphotyrosine antibodies (Millipore, Cat# 05-321, RRID:AB_309678) diluted 1:1500 in TBS-T with 3% skimmed milk for 2 h at room temperature. The membranes were washed three times for 5 min in TBS-T and incubated in the 1:1000 solution of secondary antimouse antibodies conjugated with HRP (DAKO, Cat# P0447) diluted in TBS-T containing 3% skimmed milk for 45 min at room temperature. Membranes were washed with TBS-T for 4 × 10 min. Stained bands were visualized with a Pierce ECL western blotting substrate (Cat# 32106) or SuperSignal ELISA Femto substrate (Pierce, Cat# 37075) using LAS-3000 imaging software. To confirm, equal loading membranes were stripped and reprobed with anti-RET C-20 antibodies (1:500) diluted in TBS-T containing 3% skimmed milk. To detect C-20, we used secondary antigoat antibodies conjugated with HRP (1:1500, DAKO, Cat# P0449).

Quantification of RET Phosphorylation

Quantification of RET pY images was done using Image Studio 5.2 software. Intensities of the bands corresponding to phosphorylated surface form of RET (MW = 170 kDa) and total surface form of RET were first normalized to their areas and intensity/area values for RET pY band then were normalized to the intensity/area values of RET band. The images from 3 to 6 independent experiments were quantified.

Statistical Analysis

Quantitative data were analyzed by one-way or RMs ANOVA with the Dunnett’s post hoc test to determine the significance of the differences. P values below 0.05 were considered statistically significant.