Discovery of Novel and Selective GPR17 Antagonists as Pharmacological Tools for Developing New Therapeutic Strategies in Diabetes and Obesity

Abstract

G protein coupled receptors (GPCRs) are promising targets for diabetes and obesity therapy due to their roles in metabolism and excellent potential for pharmacological manipulation. We previously reported that Gpr17 ablation in the brain-gut axis leads to improved metabolic homeostasis, suggesting GPR17 antagonism could be developed for diabetes and obesity treatment. Here, we performed high throughput screening (HTS) and identified two new GPR17 antagonists (compound 978 and 527). Both compounds antagonized downstream Gαi/o, Gαq and β-arrestin signaling with high selectivity for GPR17, but not the closely related purinergic and cysteinyl leukotriene receptors. The molecular mechanisms of antagonism were revealed through Schild analysis, structure-activity relationship (SAR) studies and homology modelling. Compound 978, a competitive antagonist against the surrogate small molecule agonist MDL29,951 (MDL), and its analog (793) attenuated GPR17 signaling and promoted glucagon-like peptide-1 (GLP-1) secretion in enteroendocrine cells. In summary, we identified selective GPR17 antagonists through HTS, which represent promising pharmacological tools for developing new therapeutic strategies in diabetes and obesity.

Introduction

Metabolic diseases represent a spectrum of disorders, including diabetes, obesity, hyperlipidemia, and cardiovascular disorders, due to the disrupted metabolic homeostasis in the human body¹. G-protein coupled receptors (GPCRs) are a large family of seven-transmembrane-domain receptors and are of particular interest for pharmaceutical intervention^2,3. Approximately 30–40% of all FDA approved drugs modulate GPCRs⁴. GPCRs in the endocrine tissues respond to metabolic hormones, nutrients, metabolites and neurotransmitters, and therefore are promising therapeutic targets for metabolic diseases⁵. For instance, the glucagon-like peptide-1 (GLP-1) receptor has emerged as one of the most successful targets for treating type 2 diabetes mellitus (T2DM) and obesity; eight GLP-1 receptor agonists have received recent approval in the US market^6,7. These new drugs, however, have high cost and adverse effects mainly associated with gastrointestinal discomfort⁶. Deorphanization efforts over the past two decades have unveiled a growing number of GPCRs as potential therapeutic targets for metabolic diseases^2,8. GPR17 is an orphan GPCR we recently identified as a potential target for intervention in diabetes and obesity^9–12. We found that GPR17 was an effector of Forkhead box protein O1 (FoxO1) in hypothalamic agouti-related peptide (AgRP) neurons, a pivotal regulator of feeding behavior and energy expenditure¹¹. Loss-of-function of Gpr17 in AgRP neurons increased central nervous system sensitivity to insulin and leptin, leading to reduced food intake and increased relative energy expenditure in animal models¹⁰. Furthermore, GPR17 was found to be expressed in the GLP-1 producing enteroendocrine cells (EECs). Inducible knockout of Gpr17 in gut epithelium led to increased secretion of GLP-1 and improved glucose tolerance¹². Beyond these animal studies, our investigations with human GPR17 revealed distinct downstream signaling profiles associated with GPR17 missense variants in patients with metabolic diseases, including diabetes and obesity¹³. Besides metabolic diseases, GPR17 also plays a critical role in regulating oligodendrocyte maturation and the process of myelination^14,15. Genetic deletion of Gpr17 in mice leads to premature myelination and transgenic mice overexpressing Gpr17 exhibit features of demyelinating disorders¹⁴. GPR17 upregulation has also been found in patients affected by multiple sclerosis¹⁴. Thus targeting GPR17 has emerged as a potential therapeutic strategy for promoting remyelination in neurodegenerative diseases such as multiple sclerosis^14,16–18.

GPR17 was initially identified from a human hippocampus cDNA library using P2Y-homologous sequences¹⁹. In the human genome, GPR17 is located on chromosome 2q21 containing 4 exons. There are four GPR17 transcript variants, which give rise to two distinct protein isoforms: the long isoform (hGPR17L) and the short isoform (hGPR17S), distinguished by an additional 28 amino acids at the N-terminus of the long isoform²⁰. In contrast, rodent Gpr17 possesses only one protein isoform, which is an ortholog of hGPR17S²⁰. GPR17 is phylogenetically located at an intermediate position between the purinergic type 2 receptors, subtype Y (P2RYs) and cysteinyl leukotriene receptors (CYSLTRs). Both cysteinyl leukotrienes and uracil nucleotides were reported to activate GPR17, leading to both inhibition of cyclic AMP (cAMP) production and calcium signaling²¹. However, the results have been challenged by other groups^22–25.

In addition to the controversy and uncertainty of endogenous ligands of GPR17, the currently available chemical probes in literature have limitations in selectivity and/or potency. For example, 3-(2-carboxyethyl)-4,6-dichloro-1H-indole-2-carboxylic acid (MDL29,951, MDL) was identified as a small molecule agonist¹⁷, which was subsequently confirmed by multiple groups including us^12,13,26. MDL was initially identified as an antagonist of N-methyl-D-aspartate (NMDA) receptors with high affinity for its glycine binding site (K_i = 140 nM)²⁷. Its activity was confirmed both in vitro and in vivo^27,28. After the discovery of MDL, selective antagonists of CYSLTRs including HAMI3379 (HAMI)¹⁸, pranlukast¹⁷, and montelukast²¹ were found to weakly antagonize the activity of GPR17. Other GPR17 antagonists with different scaffolds have been discovered and published in recent years^16,26,29,30. In our efforts to identify novel small molecule antagonists of GPR17, we developed high-throughput screening (HTS) assays, which were amenable to 1536-well plate format, and screened a collection of libraries containing ~300K small molecule compounds. Two novel compounds, namely 978 and 527, were successfully identified and validated through a battery of primary assays, counter assays, and orthogonal assays. Both compounds exhibited high selectivity for GPR17, remaining inert to P2RYs and CYSLTRs, and effectively blocked the activation of MDL-induced downstream signaling pathways of GPR17. We revealed the molecular mechanisms of antagonism through Schild analysis, structure-activity relationship (SAR) studies, and homology modelling. We were unable to use the available Cryo-EM structure of human GPR17 for modeling the ligand-receptor interactions, since the structure showcases an self-activated GPR17-Gi complex with a closed ligand binding site³¹. Furthermore, our functional studies in enteroendocrine cells showed that the newly identified GPR17 antagonist 978 and its analogs derived from SAR studies attenuated GPR17-mediated inhibition of cAMP signaling and GLP-1 secretion.

Results

Assay design, optimization, and miniaturization for HTS

GPR17 couples to Gαi/o, Gαs, Gαq, and β-arrestin signaling pathways¹⁷. β-arrestin recruitment after receptor activation was shown to be mediated via both G protein-dependent and independent mechanisms^17,32. To identify small molecule antagonists independent of signaling pathways, the PathHunter^™ β-arrestin recruitment assay (DiscoveRx/Eurofins) was chosen as the primary assay for our high-throughput screening campaign. In this technology, GPR17 (GenBank accession number NM_005291, long isoform) tagged with ProLink^™ was stably expressed in a human bone osteosarcoma epithelial (U2OS) cell line which co-expressed β-arrestin 2 protein tagged with the larger enzyme acceptor fragment of β-galactosidase. Ligand-induced receptor activation and subsequent recruitment of β-arrestin to the receptor leads to the fragment complementation of β-galactosidase, which can be monitored by chemiluminescent signal. It is a simple and robust assay with high specificity, and it is compatible for HTS²⁴. MDL, a surrogate agonist of GPR17, was used in our experiments to activate the receptor. In the PathHunter GPR17 β-arrestin recruitment assay, MDL activated GPR17 at EC₅₀=0.34 μM which was comparable to previous studies¹⁷ (Figure 1A). HAMI, which was initially developed as a cysteinyl leukotriene receptor 2 (CysLTR2) antagonist (IC₅₀=3.8 nM)²⁵, was also found to be an antagonist of GPR17 exhibiting an IC₅₀ in the single digit micromolar range¹⁸. It was used as a reference antagonist compound in our HTS campaign. Pretreatment of PathHunter GPR17 β-arrestin cells with HAMI effectively blocked the activation of receptors induced by an EC₈₀ concentration of MDL (2 μM) in a dose dependent manner with an IC₅₀ of 8.2 μM (Figure 1B). This IC₅₀ for HAMI was also comparable with previous studies¹⁸.

Figure 1.

Development of GPR17 assays for high-throughput screening. PathHunter GPR17 β-Arrestin assay was developed and optimized in 1536-well plate format and used as the primary assay for high-throughput screening. A. Dose-response curve of MDL compound in agonist mode, EC₅₀ of MDL was 0.37μM. B. Dose-response curve of HAMI compound in antagonist mode, 2μM MDL (~EC₈₀ concentration) was used to stimulate the receptor. IC₅₀ of HAMI was 8.2μM. C. Scatter plot of PathHunter GPR17 β-Arrestin assay in a 1536-well plate format. Column 1 was unstimulated control (no MDL). Column 2 was stimulated neural control (2 μM MDL). Columns 3 and 4 were the antagonist control (66.7 μM HAMI and 2 μM MDL). Column 5 to 48 were pretreated with small molecule library compounds and then stimulated with 2 μM MDL. D. Reproducibility test of PathHunter GPR17 β-arrestin recruitment assay from two runs. E. Calcium mobilization assay was optimized in 1536-well plate format and used as an orthogonal assay. Dose response curve of MDL compound in the agonist mode. EC₅₀ of MDL was 0.28μM. F. Calcium mobilization assay in antagonist mode. 1.7μM MDL (~EC₈₀ concentration) was used to stimulate the receptor. IC₅₀ of HAMI was 8.1μM. G. cAMP accumulation assay (Cisbio) was developed in 384-well plate format as a second orthogonal assay. Dose response curve of MDL compound in agonist mode. EC₅₀ of MDL was 1.9 nM. 1 μM forskolin was used to stimulate the production of cAMP. H. cAMP accumulation assay in antagonist mode. An increased concentration of HAMI was used in this experiment. 1 μM forskolin was used to stimulate the production of cAMP and 30 nM MDL was used to activate the receptor.

To enable the automated screening of small molecule libraries, the assay was miniaturized, optimized, and validated in a 1536-well microplate format. The first four columns of the 1536-well plates were reserved as the control columns. Column 1 was unstimulated control (32 wells, no MDL). Column 2 was an agonist stimulated control (32 wells, 2 μM MDL). Columns 3 and 4 were the antagonist control (32 wells each, 30 μM HAMI and 2 μM MDL). Columns 5 to 48 were pretreated with small molecule library compounds and then stimulated with 2 μM MDL. Figure 1C showed an example scatter plot of receptor activity (%) from a 1536-well DMSO control plate to validate the assay. The ratio of signal/background (S/B) was 2.6, coefficient of variation (CV) was 8.8%, and Z factor was 0.61. To assess the reproducibility, we tested the LOPAC^®1280 (LOPAC) library, a commonly used commercial benchmarking library consisting of 1280 bioactive small molecules from Millipore Sigma, for assay optimization and miniaturization, in this β-arrestin recruitment assay in two runs. The data showed acceptable reproducibility between both runs in our experiment (R²=0.73, Figure 1D).

To further validate the hits identified from HTS, we also developed calcium mobilization and cAMP accumulation assays as orthogonal assays using a 1321N1 cell line (human astrocytoma cell line) that stably expresses hGPR17L. A Calbryte-520 probenecid-free and wash-free calcium assay was used to detect calcium mobilization. Cells were treated with increasing concentrations of MDL and calcium influx was measured by a Fluorometric Imaging Plate Reader (FLIPR) Penta high-throughput cellular screening system. MDL activated the receptor at EC₅₀=0.28 μM (Figure 1E), and HAMI effectively blocked the receptor activation stimulated by MDL (at EC₈₀ concentration of 1.7 μM) at IC₅₀=8.1 μM (Figure 1F). GPR17-mediated cAMP signaling was measured with a homogeneous time resolved fluorescence (HTRF) cAMP assay. Cells were treated with varying concentrations of MDL and subsequently stimulated with the direct adenylyl cyclase activator forskolin (1 μM). MDL activated GPR17 at EC₅₀=1.9 nM (Figure 1G), and HAMI effectively blocked the receptor activation by MDL (at EC₈₀ concentration of 30 nM) at IC₅₀=1.5 μM (Figure 1H).

Identification and validation of antagonists of GPR17 from HTS

To identify the antagonists of GPR17, we conducted a high-throughput screening campaign following a rigorous and stringent process (Figure 2A). We screened in-house libraries consisting of a total number of 301,408 small molecule compounds in 1536-well format using the PathHunter GPR17 β-arrestin recruitment assay as the primary assay. The libraries were screened at a final concentration of compounds at 66.7 μM. Briefly, the cells were pretreated with the library compounds for 10 min and then stimulated with 2 μM MDL (EC₈₀ concentration). The activity of each compound was normalized to 66.7 μM HAMI as the positive control (−100%) and DMSO as the neutral control (0%). The activity of a compound equal or less than −50% was designated as a hit. According to this criteria, 1,378 hits were identified and further assessed in the PathHunter GPR17 β-arrestin recruitment assay at 11 doses (a series of 5-fold dilutions from 66.7 μM to 0.0068 nM). 380 hits were successfully confirmed, and the confirmation rate was 27.6% (Figure 2A). Figure 2B and 2C show the structures and dose response curves of two confirmed hit compounds, namely compounds 978 and 572.

Figure 2.

Identification of antagonists of GPR17 by high-throughput screening. A. Flow chart of high-throughput screening. PathHunter GPR17 β-arrestin assay was used as the primary assay for high-throughput screening campaign. 301,408 small molecule compounds from in-house libraries were screened at single concentration (66.7 μM) by the primary assay. 1,378 compounds, of which the max response was more than 50%, were cherry-picked as hits and tested in 11 doses in the PathHunter GPR17 β-arrestin assay. The activity of 380 compounds was confirmed by confirmatory assay. The artifacts of PathHunter assay were triaged by two counter assays, PathHunter FGFR1 functional assay and PathHunter GHSR1α β-arrestin assay. 19 of 380 hits were inactive in two counter assays and were selected for further validation in two orthogonal assays, calcium mobilization assay and cAMP accumulation assay. Two antagonists (compounds 978 and 527) were finally validated in these two orthogonal assays. B. Structure of compounds 978 and 527. C. Dose response curve of compounds 978 and 527 in PathHunter U2OS GPR17 β-arrestin assay. D. Dose response curve of compounds 978 and 527 in PathHunter U2OS FGFR1 functional assay. E. Dose response curve of compounds 978 and 527 in PathHunter U2OS GHSR1α β-arrestin assay. F. Dose response curve of compounds 978 and 527 in calcium mobilization assay (1321N1 GPR17 stable cell line). G. Dose response curve of compounds 978 and 527 in calcium mobilization assay (U2OS GPR17 β-arrestin cell line). H. Attenuation of MDL-induced inhibition of forskolin-stimulated cAMP accumulation in 1321N1 GPR17 stable cells. Cells were pretreated with 10 μM of compounds and then stimulated with 1 μM forskolin and 30 nM MDL. The date was analyzed by one-way ANOVA with Dunnett’s posttest. Compound 978 was re-purchased and subjected to purification, and compound 527 was re-synthesized by our chemist. The purified compounds were reassessed in three GPR17 assays, PathHunter GPR17 U2OS β-arrestin assay (I), calcium mobilization assay (1321N1 GPR17 stable cell line) (J), cAMP HTRF assay (1321N1 GPR17 stable cell line) (K). HAMI was used as a reference compound. The IC₅₀ of 978 was 6.6 μM (arrestin assay), 2.3 μM (calcium mobilization assay) and 6.1 μM (cAMP HTRF assay). The IC₅₀ of 527 was 33.3 μM (arrestin assay), 13 μM (calcium mobilization assay) and 13.2 μM (cAMP HTRF assay).

To triage the assay artifacts and nonselective hits, two other PathHunter assays were chosen as the counter assays. Fibroblast growth factor receptor 1 (FGFR1) is a receptor tyrosine kinase (RTK), and activation of receptor leads to the recruitment of downstream SH2 domain proteins, which was engineered in the PathHunter β-galactosidase complementation assay for detection of receptor activation. The FGFR1 functional assay was used as a counter assay to triage interference compounds that modulate FGFR1 and β-galactosidase. Growth hormone secretagogue receptor 1a (GHSR1a) is a functional GPCR for gut peptide ghrelin. GHSR1a β-arrestin recruitment assay was used as another counter assay to triage false positive hits that may induce β-arrestin recruitment non-selectively. The performance of counter assays was shown in supplemental figure 1. 19 out of 380 hits (5%) identified from the confirmatory assay were shown to be inactive in both the FGFR1 functional assay and GHSR1a β-arrestin recruitment assay (Figure 2A). The dose response curves of two compounds (978 and 572) showed no activity in the counter assays (Figures 2D&E).

As the primary screen and counter screen were conducted in the PathHunter β-arrestin recruitment assay, activation of G-protein signaling pathways by 19 hits was further evaluated to triage assay artifacts. Activation of the Gαq signaling pathway was evaluated in two GPR17 stable cell lines with different cell backgrounds (1321N1 cell line and U2OS cell line), and antagonism of the Gαi/o signaling pathway by the compounds was evaluated in the 1321N1 GPR17 stable cell line. The different cell background, different readouts, and distinct signaling pathways were expected to further triage artifacts identified from the primary assay. In these orthogonal assays, 2 out of 19 hits, compounds 978 and 572, were confirmed. Compound 978 had a tetrahydro-1H-carbazole core while compound 527 had a sulfonamido acetamide core in a dipeptide-like backbone (Figure 2B). Compounds 978 and 572 dose dependently reduced calcium signaling (Figures 2F&G). This antagonism of compound 978 and 527 was also confirmed in the short form of GPR17 where they showed dose dependent reduction of calcium response in HEK293 stably expressing hGPR17S (Supplemental figure 7). Furthermore, 10 μM compounds 978 and 572 significantly antagonized the MDL-mediated inhibition of forskolin-stimulated cAMP accumulation (Figure 2H).

Contaminants or decomposition products from library compounds are a source of pan-assay interference compounds (PAINS)³³. To exclude this possibility, compounds 978 and 527 were either re-purchased or re-synthesized, subjected to purification, and reassessed in three GPR17 assays. The antagonistic activity of these two compounds was successfully validated in the β-arrestin recruitment assay, calcium mobilization assay, and cAMP accumulation assay. As a reference, HAMI was included in the experiments. The potency of compound 978 was moderately higher than HAMI, while the potency of compound 527 was similar or lower than HAMI (Figures 2I–K). In the β-arrestin recruitment, calcium mobilization and cAMP accumulation assays, the IC₅₀ of compound 978 was 6.6 μM, 2.3 μM and 6.1 μM, respectively; the IC₅₀ of compound 527 was 33.3 μM, 13 μM and 13.2 μM, respectively; the IC₅₀ of HAMI was 14.6 μM, 21 μM and 10 μM, respectively.

Characterization of antagonists of GPR17

As GPR17 is in an intermediate position between the P2RY and CYSLTR families³⁴, the receptors from these two families were chosen to further evaluate the selectivity of compounds 978 and 527. First, we assessed the selectivity of compounds at a fixed dose in cells transiently overexpressing GPR17, CYSLTR1, CYSLTR2 and P2RY1. As expected, 10 μM of compound 978 and 527 blocked calcium signaling of GPR17 stimulated by 200 nM MDL (Figure 3A). 10 μM of compound 978 and 527 failed to block the activation of CYSLTR1 (Figure 3B), CYSLTR2 (Figure 3C), and P2RY1 (Figure 3D) stimulated by their respective ligands. In contrast, pranlukast, HAMI, and MRS2179, as the positive controls, antagonized CYSLTR1, CYSLTR2, and P2RY1, respectively (Figure 3B–D). Secondly, we tested the selectivity of 978 and 527 in a dose dependent manner in the stable cell lines. The activation of CYSLTR1, CYSLTR2, and P2RY1 was evaluated by calcium mobilization assay in CYSLTR1, CYSLTR2, and P2RY1 HEK293 stable cell lines and ACTone cAMP assay³⁵ in the P2RY12 HEK293 stable cell line. The CYSLTR1 and CYSLTR2 were activated by their endogenous ligand, leukotriene D4 (LTD4), with an EC₅₀ of 3 nM (Supplemental figure 2A) and 4.2 nM (Supplemental figure 2B) respectively. The P2RY1 was activated by its endogenous ligand, adenosine triphosphate (ATP), at EC₅₀ of 4.6 nM (Supplemental figure 2C). The EC₅₀ of 2-Methylthioadenosine diphosphate (2MeSADP) was 0.032 nM measured by ACTone cAMP assay in the P2RY12 HEK293 stable cell line (Supplemental figure 2D). The antagonist activities of 978 and 527 were evaluated in these stable cell lines by stimulating them with their respective ligands at approximately EC₈₀ concentrations. The data showed that 978 exhibited some inhibitory effect (~ 41%) on activation of CYSLTR1 receptor only at the highest assay concentration of 33 μM (Figure 3E), but not the other three receptors (Figures 3F–H). Compound 527 did not show any inhibitory effects on blocking the activation of the receptors by their respective ligands (Figure 3E–H).

Figure 3.

Selectivity of two hits against the closely related receptors of GPR17. Neuro-2a cell, a mouse neuroblastoma cell line, was transiently transfected with hGPR17L, hCysLTR1, hCysLTR2, and hP2RY1, and the antagonism of compounds against these receptors was assessed by calcium mobilization assay after stimulating with their respective agonist ligands. A. Neuro-2a cells expressing hGPR17 were pretreated with 10 μM of compounds 978, 527 and HAMI for 10 minutes before stimulated with 200 nM MDL. B. Neuro-2a cells expressing hCysLTR1 receptor were pretreated with 10 μM of compounds 978, 527 and pranlukast for 10 minutes before stimulated with 50 nM LTD4. C. Neuro-2a cells expressing hCysLTR2 receptor were pretreated with 10 μM of compounds 978, 527 and HAMI for 10 minutes before stimulated with 50 nM LTD4. D. Neuro-2a cells expressing hP2Y1 receptor were pretreated with 10 μM of compounds 978 and 527, and 300 μM MRS2179 for 10 minutes before stimulated with 8 nM 2MeSADP. Data (A-D) were analyzed by one-way ANOVA followed by Dunnett’s post hoc test where each test compound treatment condition was compared to vehicle treatment. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. Selectivity of two hits was also assessed by hCysLTR1, hCysLTR2, hP2RY1, hP2RY12 HEK293 stable cells. E. hCysLTR1 stable cell line was pretreated with the compounds for 10 min and then stimulated with 10nM LTD4. F. hCysLTR2 stable cell line was pretreated with the compounds for 10 min and then stimulated with 10nM LTD4. G. hP2RY1 stable cells were pretreated with the compounds for 10 min and then stimulated with 20nM ATP. H. hP2RY12 stable cells were treated with the compounds for 25 min and then stimulated with 0.1 nM 2MeS-ADP. The signals were detected by FlexStation. Schild regression analysis of two antagonists in β-arrestin recruitment assay. I. Dose response curve of MDL in the presence of increased concentration of compound 978 in β-arrestin recruitment assay. J. Dose response curve of MDL in the presence of increased concentration of compound 527 in β-arrestin recruitment assay. K. Schild plot of compounds 978 and 527 in β-arrestin recruitment assay. The slopes of compounds 978 and 527 were 1.15 and 0.64 respectively. The pA2 value of compounds 978 and 527 were 6.2 and 5.9 respectively.

The validation and characterization of compounds 978 and 527 suggested specific and selective antagonism of GPR17, but these assays do not provide information regarding the molecular mechanisms of antagonism. Therefore, the mode of action for each compound was investigated in β-arrestin recruitment assays by evaluating the effects of varying concentrations of test compounds on dose-response curves of MDL in 1321N1-GPR17 cells. To test the effects of compounds on GPR17-mediated β-arrestin recruitment, compounds 978 and 527 were added 10 minutes prior to addition of the agonist MDL in 1321N1-hGPR17 cells. β-arrestin recruitment was measured using the Cisbio HTRF β-arrestin recruitment assay. 1 μM, 3.3 μM, and 10 μM concentrations of compounds 978 (Figure 3I) and 527 (Figure 3J) caused concentration-dependent rightward shifts of the MDL concentration-response curve for GPR17-mediated β-arrestin recruitment. Schild regression analysis for compounds 978 and 527 yielded slopes of 1.15 and 0.64, respectively (Figure 3K). They were not significantly different than one, as demonstrated in 95% confidence intervals (Table 1). This suggests a mechanism of surmountable antagonism that is consistent with a competitive interaction between these compounds and MDL at the binding site of the receptor.

Table 1.

Schild Analysis of compounds in β-arrestin recruitment assay

Compound	β-arrestin Schild Slope (95% CI)
527	0.81 (0.62 – 1.02)
978	1.13 (0.96 – 1.30)

Structure activity relationship of antagonists

After identification of compounds 978 and 527 as selective antagonists of GPR17, we engaged in exploratory structure-activity relationship (SAR) campaigns around each of the chemotypes. To that end, we developed synthetic routes that were amenable to diversification in most regions of the chemical structures. For the chemical series represented by 978, we started the synthesis with substituted phenyl hydrazines such as 1-chlorophenyl hydrazine (Figure 4A, a). These were condensed with cyclohexane-1,2-dione (Figure 4A, b) to provide a tetrahydrocarbozol-1-one (Figure 4A, c) whose ketone was then reduced using ammonium acetate and sodium cyanoborohydride to the corresponding aliphatic amine (Figure 4A, d). This amine served as a handle for the Hexafluorophosphate Azabenzotriazole Tetramethyl Uronium (HATU)-mediated coupling with diverse carboxylic acids (Figure 4A, e) to provide a range of structurally diverse tetrahydrocarbazolyl amides (Figure 4A, f). A modular synthesis was also developed for the 527 chemotype. The synthetic sequence (Figure 4B) started with the three-component coupling of thiophene-2-sulfonamide (Figure 4B, g), oxoacetic acid (Figure 4B, h), and phenyl boronic acid (Figure 4B, i) to provide central intermediate (Figure 4B, j). This has a carboxylic acid handle for HATU-mediated amide coupling with diverse anilines (Figure 4B, k).

Figure 4.

Schemes 1 (A) and 2 (B) are representative synthetic schemes to make analogs of chemotypes 978 and 572. Representative synthesized analogs exploring chemical diversity around chemotypes 978 (C) and 572 (D) are presented.

Representative synthesized analogs around the 978 chemotype are presented in Figure 4C. Their activities are reported in the calcium assay in Tables 2 and 3. The IC₅₀ is reported only when the activity at the highest concentration is greater than 75%. The efficacy is reported as a percentage; it is essentially the curve-fitted activity at the highest concentration of the assay. Table 2 shows the SAR at the amide region where a variety of electron donating groups (entries 2, 3, 4, 17) as well as withdrawing substitutions (entries 6, 7, 8, 12, 13, 15, 16) are explored. Analogs with a combination of electron withdrawing and donating groups are also shown in entries 5, 9, 10, and 11. An analog that borrows features from the chemotype represented by 527 is shown in entry 14. Only 6 analogs, including the hit 978, in Table 2 show a maximum response greater than 75% inhibition and had an IC₅₀ ≤ 15 μM.

Table 2.

978 chemotype amide SAR

Compd	NCGC ID	R	Calcium Assay IC50(μM) Avg ± SEM	% Efficacy Avg ± SEM
1	NCGC00423978		5.47 ± 0.30	109.05 ± 3.38
2	NCGC00424034			47.39 ± 3.11
3	NCGC00424042		11.4 ± 0.59	77.43 ± 3.46
4	NCGC00685956		10.69 ± 0.29	81.5 ± 1.23
5	NCGC00843520			50.58 ± 3.54
6	NCGC00843431			38.05 ± 2.73
7	NCGC00687089		4.42± 0.34	116.89 ± 6.08
8	NCGC00843419		12.65± 0.14	74.12± 2.16
9	NCGC00685742		12.61± 0.89	90.31 ± 7.04
10	NCGC00686236			52.79 ± 2.26
11	NCGC00843420			29.78 ± 1.21
12	NCGC00843412			44.34 ± 3.47
13	NCGC00685877		11.64± 1.19	100.62 ± 9.98
14	NCGC00687594		7.94± 0.49	103.89 ± 9.32
15	NCGC00687793			49.38± 2.45
16	NCGC00650884			40.37 ± 4.47
17	NCGC00650889			62.57 ± 4.84

The IC₅₀ and % efficacy are reported as Mean ± SEM. IC₅₀ is reported only when efficacy ≥75%

Table 3.

978 chemotype tetrahydrocarbazol SAR

Compd	NCGC ID	R1	R2	R3	R4	R5	R6	R7	Calcium assay IC50(μM) Avg ± SEM	% Efficacy Avg ± SEM
18	NCGC00423976	H	H	H	H	OMe	OMe	H	9.44±1.25	72.1 ± 4.2
19	NCGC00841699	Br	H	H	H	OMe	OMe	H	4.29±0.48	125.04 ± 4.30
20	NCGC00843841	Br	H	H	H	OMe	H	OMe		55.54 ± 1.74
21	NCGC00843850	Me	H	Me	H	OMe	OMe	H	12.24±0.5	87.54 ± 0.92
22	NCGC00843432	Me	H	Me	H	OMe	H	OMe		68.15 ± 1.59
23	NCGC00846053	Me	H	Me	Me	OMe	H	OMe	11.78±0.5	83.65 ± 2.39
24	NCGC00842971	CO2H	H	H	H	OMe	OMe	H	inactive
25	NCGC00841530	H	CO2H	H	H	OMe	OMe	H	inactive
26	NCGC00842489	H	H	CO2H	H	OMe	OMe	H	inactive

The IC₅₀ and % efficacy are reported as Mean ± SEM. IC₅₀ is reported only when efficacy ≥75%

Table 3 shows the SAR in the aromatic ring present within the tetrahydro carbazole core. The R₁ substituent which was chloride in hit 978 was replaced with hydrogen (18), bromide (entries 19, 20) and a methyl group (entries 21–23). In this group of compounds, the bromide 19 with an IC₅₀ ~ 4 μM had an activity that was slightly better than hit 978. A carboxylic acid group was also systematically walked around the aromatic (R₁ to R₃) ring to mimic the carboxylic acids present in MDL with the hope of interacting with polar positively charged residues in GPR17. However, these compounds (entries 24–26) were inactive in the calcium assay.

We also explored the SAR around chemotype 527; representative analogs are shown in Figure 4D and their activities in the calcium assay reported in Tables 4–6. Table 4 showcases an exploration of substitution at the central phenyl ring at the substituent shown as R. Realizing that the substitution was a stereocenter we separated the two enantiomers using chiral chromatography and found that one enantiomer was more active (IC₅₀ 12.62 ± 0.28 μM) while the other one was inactive. However, the activity of the pure enantiomer was lower than the racemic mixture (IC₅₀ 4.36 ± 0.36 μM). While this was unusual, we think the activities of the purified active enantiomer and the racemic compounds were comparable (within 3-fold of each other) and within the intrinsic variability of the assay. Because of this we decided to focus our SAR exploration on racemic compounds. We are yet to ascertain the exact configuration of these enantiomers. Entries 27–30 explore ortho substitutions, 31–33 meta substitutions, and 34–43 para substitutions. Although a variety of electron withdrawing (fluoro, chloro, ester, acid, amide), electron donating (methyl, methoxy, alkyl), and lipophilic groups were explored, none of them showed inhibitory activity that were discernibly better than the starting compound 527. Tables 5 and 6 show representative compounds that explored SAR around the western sulfonamide aromatic ring and eastern amide. Unfortunately, the structural diversity explored in these regions also did not lead to an analog with better activity than compound 527. Thus, our medicinal chemistry efforts suggested that both chemotypes might be associated with SARs that are relatively flat.

Table 4.

527 Chemotype central ring SAR

Compd	NCGC ID	R	Calcium assay IC50(μM) Avg ± SEM	% Efficacy Avg ± SEM
27	NCGC00105527		4.36 ± 0.36	94.08 ± 3.22
28	NCGC00601858		17.02 ± 1.38	78.79 ± 2.42
29	NCGC00601837		17.94 ± 0.73	117.81 ± 1.54
30	NCGC00601856		11.2 ± 0.67	122.79 ± 3.12
31	NCGC00601919		14.77 ± 0.61	111.29 ± 1.84
32	NCGC00601924		15.74 ± 1.32	99.1 ± 7.63
33	NCGC00601981		inactive
34	NCGC00601764		12.22 ± 0.79	92.96 ± 8.02
35	NCGC00601888		9.98 ± 1.54	83.41 ± 3.74
36	NCGC00601840		8.66 ± 0.54	99.22 ± 3.09
37	NCGC00601848		14.04 ± 0.76	108.56 ± 4.65
38	NCGC00602234			65.05 ± 2.61
39	NCGC00601925			66.3 ± 11.05
40	NCGC00602233		3.58 ± 0.24	136.65 ± 4.38
41	NCGC00645518			48.27 ± 3.32
42	NCGC00602145			42.2 ± 3.98
43	NCGC00601864		3.18 ± 0.43	91.88 ± 4.8

The IC₅₀ and % efficacy are reported as Mean ± SEM. IC₅₀ is reported only when efficacy ≥75%

Table 6.

527 chemotype amide SAR

Compd	NCGC ID	R	Calcium assay IC50(μM) Avg ± SEM	% Efficacy Avg ± SEM
50	NCGC00842619		4.65 ± 0.54	108.59 ± 4.46
51	NCGC00687958			54.59 ± 4.59
52	NCGC00689502		Inactive
53	NCGC00689480		11.65 ± 0.42	100.06 ± 4.85
54	NCGC00843760		8.65 ± 1.29	74.89 ± 1.88
55	NCGC00843757		11.73 ± 0.26	75.78 ± 6.38
56	NCGC00843756			64.35 ± 3.39
57	NCGC00649955			60.54 ± 7.23

The IC₅₀ and % efficacy are reported as Mean ± SEM. IC₅₀ is reported only when efficacy ≥75%

Table 5.

527 chemotype sulfonamide SAR

Compd	NCGC ID	R	Calcium assay IC50(μM) Avg ± SEM	% Efficacy Avg ± SEM
44	NCGC00650036		15.19 ± 1.27	104.07 ± 4.46
45	NCGC00650047		18.15 ± 2.43	87.54 ± 12.73
46	NCGC00650052			68.9 ± 6.57
47	NCGC00650046			44.45 ± 9.79
48	NCGC00650049		17.91 ± 0.34	99.67 ± 9.28
49	NCGC00650043			65.67 ± 7.45

The IC₅₀ and % efficacy are reported as Mean ± SEM. IC₅₀ is reported only when efficacy ≥75%

Homologous model of antagonists with GPR17.

To simulate the interaction of compounds 978 and 527 with GPR17, a predicted structure of human GPR17 was modeled using CysLT2R as a template. The seven-transmembrane structure of GPR17 formed a ligand binding site on the extracellular side, which was basic (Figures 5A and 5B). The predicted pose of the known agonist MDL29,951 formed a salt bridge with ARG⁸⁷ and ARG²⁸⁰, and a hydrogen bond with TYR²⁵⁸ (Figure 5C). The phenyl ring in compound 527 was predicted to stack with the side chain of PHE³⁴, whereas the oxygen atoms in the sulfonamide and amide were predicted to form hydrogen bonds with ASN²⁷⁹ and ARG⁸⁷ (Figure 5D). Inhibitor 978 interacted with GPR17 largely through hydrophobic interactions, by burying its tricyclic group into the narrow binding pocket (Figure 5E). The residues in the ligand binding pocket are largely overlapped with those that were published by other researchers^30,36.

Figure 5.

Homogeneous model of GPR17 docking with MDL and two antagonists. The predicted structure of human GPR17 with the agonist MDL docked in the ligand binding site: (A) Cartoon presentation (B) Electrostatic potential surface (blue color indicates electron-deficient surface, and red surface is electron rich) of the seven-transmembrane structure. The predicted ligand binding poses by molecular docking for (C) MDL, (D) compound 527 and (E) compound 978. The binding modes were displayed in both 3D structural presentation (Left panel) and 2D molecular interaction maps (Right panel).

GPR17 antagonist effects on cAMP signaling and regulation of GLP-1 secretion in GLUTag cells.

GPR17 is expressed in enteroendocrine cells of the gastrointestinal tract and regulates nutrient-stimulated GLP-1 secretion and thereby modulates glucose-stimulated insulin secretion and glucose homeostasis¹². Such effects are thought to be mediated by GPR17 Gi/o coupling¹². The GLUTag cell line is a GLP-1 secreting enteroendocrine cell model that has been extensively used to investigate GPCR regulation of GLP-1 secretion^37–39. We evaluated compounds 978 and 527, as well as some analogues from SAR studies (Supplemental figure 3), for effects on GPR17-mediated cAMP signaling and GLP-1 secretion in GLUTag cells.

The putative GPR17 antagonists were first tested for effects on GPR17-mediated cAMP signaling modulation. Forskolin stimulation of hGPR17L expressing GLUTag cells elevated cAMP levels from 0.59 ± 0.04 nM to 4.7 ± 0.36 nM. Forskolin-stimulated cAMP production was inhibited by the activation of GPR17 with MDL to 3.5 ± 0.18 nM, representing a 29 ± 3% inhibition of forskolin stimulated cAMP, consistent with Gi/o coupling of GPR17 in GLUTag cells. The compounds (10 μM) were evaluated for the ability to attenuate MDL-mediated inhibition of forskolin-stimulated cAMP (Figure 6A). Compound 978 and its analogues 419 (Compd 8), 850 (Compd 21), 793 (Compd 15), and the positive control HAMI significantly attenuated the inhibition (with 100 nM MDL) of forskolin-stimulated cAMP in GLUTag cells expressing hGPR17L. While the activity of compound 527 was modest in this assay, surprisingly, its analogues 518 (Compd 41) and 480 (Compd 53) had no effect on the MDL inhibition of forskolin-stimulated cAMP (Figure 6A). The attenuating effects on cAMP inhibition appear to be specific to GPR17-mediated inhibition of cAMP, as no attenuating effects were observed for any test compounds when forskolin-stimulated cAMP was inhibited by activating endogenous somatostatin receptors (Supplemental figure 4). Therefore, 978, 419, 850, 793, 527, and HAMI demonstrated antagonist activity with respect to GPR17-mediated cAMP signaling modulation in GLUTag cells.

Figure 6.

Confirmation of GPR17 antagonist activity for cAMP signaling and GLP-1 secretion modulation in GLUTag cells. A. Evaluation of test compounds for effects on 100 nM MDL-mediated inhibition of 10 μM forskolin-stimulated cAMP in GLUTag cells expressing hGPR17L. Data were normalized to the forskolin response within each test compound treatment condition and represent mean ± SEM of three to four independent experiments performed in triplicate wells. Data were analyzed by unpaired t-test where each test compound treatment condition was compared to vehicle treatment. * p < 0.05, ** p < 0.01, *** p < 0.001. B. GLP-1 secretion assay in GLUTag cells expressing hGPR17L. Test compounds were tested for effects on 6.3 μM MDL-mediated inhibition of 10 μM forskolin and 10 μM IBMX-stimulated GLP-1 secretion in GLUTag cells expressing hGPR17L. Data were normalized to the forskolin and IBMX response within each test compound treatment condition and represent mean ± SEM of four independent experiments performed in triplicate wells. Data were analyzed by unpaired t-test where each test compound treatment condition was compared to vehicle treatment. * p < 0.05, ** p < 0.01. Compound key: 419 (Compd 8), 978 (Compd 1), 850 (Compd 21), 793 (Compd 15), 527 (Compd 27), 518 (Compd 41), 480 (Compd 53).

The compounds were also evaluated for effects on GPR17-mediated regulation of GLP-1 secretion. GLUTag cells that were transduced with adenovirus for expression of hGPR17L had 19.7 ± 2.6 pM GLP-1 in the supernatant under basal conditions. Stimulating hGPR17L expressing GLUTag cells with 10 μM forskolin and 10 μM IBMX caused over 3-fold increase of GLP-1 secretion (65.3 ± 8.0 pM vs 19.7 ± 2.6 pM at baseline) (Supplemental figure 5). Activation of hGPR17L with 6.3 μM MDL inhibited the forskolin and IBMX-stimulated GLP-1 secretion to 50.1 ± 4.6 pM (i.e., 33 ± 8% reduction), consistent with GPR17-mediated negative regulation of GLP-1 secretion (Supplemental figure 5). The compounds were evaluated for the ability to attenuate the MDL-mediated inhibition of forskolin and IBMX-stimulated GLP-1 secretion (Figure 6B). Compounds 978, 793, and HAMI significantly attenuated MDL-mediated inhibition of GLP-1 secretion in GLUTag cells expressing hGPR17L. In summary, compound 978 and its analogue 793, at 10 μM, significantly attenuated MDL-mediated inhibition of cAMP and GLP-1 secretion in GLUTag cells expressing hGPR17L. These effects are consistent with GPR17 antagonist activity and suggest potential for targeting GPR17 with small molecule antagonists to enhance GLP-1 secretion.

Discussion

In an effort to explore new chemical space for GPR17 antagonists, we screened over 300K compounds in a high throughput β-arrestin assay. Two antagonists with novel scaffolds (compounds 978 and 527) were successfully identified and validated in multiple assays. The compounds from both scaffolds effectively inhibited multiple downstream signaling pathways of GPR17 with high selectivity over other closely related receptors. Our findings provide new insights into the molecular mechanisms of GPR17 antagonism and offer valuable pharmacological tools for advancing therapeutic strategies in diabetes and obesity.

GPR17 is an orphan GPCR, and its putative endogenous ligand has not yet been conclusively validated. Phylogenetically GPR17 is in an intermediate position between the P2RY and CysLTR families (27). Initial reports indicated that endogenous ligands from both families (UDP, UDP-glucose, LTC4 and LTD4) activated GPR17 in both cAMP accumulation assay and calcium mobilization assay²¹. However, subsequent studies by other research groups have failed to confirm these findings^22,26,40. Harrington et al.²⁶ reported that oxysterols, particularly 24(S)-hydroxycholesterol (24S-HC), could be the physiological activators of GPR17. Notably, 24S-HC, the major brain cholesterol metabolite, is an agonist of liver X receptors⁴¹ and a positive allosteric modulator of NMDA receptor⁴². Analysis of the data presented in their study suggests 24S-HC behaves more like an allosteric modulator instead of an endogenous agonist²⁶. However, we were not able to confirm that 24S-HC either activated GPR17 or potentiated the activity of MDL in both cAMP accumulation assay and calcium mobilization assay (supplemental figure 6). To identify the potential lipid ligands of receptor, we cherrypicked and screened a collection of 745 lipid-like compounds from our in-house library. Unfortunately, we could not identify any compounds activating the receptor (data now shown). However, we can’t exclude the possibility that lipid-like molecules such 24S-HC can act as endogenous ligands of receptor. The authors of the study with 24S-HC have been pointed out to us that addressing poor aqueous solubility of lipids is critical in these assays and they have recommended intermediate compound dilutions in DMSO prior to addition to the assay.

It is intriguing that while endogenous ligands from closely related P2RY and CYSLTR families were unable to activate GPR17, antagonists from the CYSLTR subfamily can completely (pranlukast, HAMI and BayCysLT2) or partially (monelukast) block the MDL-stimulated receptor activation of GPR17¹⁷. In contrast, antagonists from the P2RYs family (ticagrelor and cangrelor) failed to block MDL-stimulated receptor activation^18,26. This phenomenon was also observed in our studies (data not shown). In the calcium mobilization assay, we found that pranlukast, HAMI, BayCysLT2 and zafirlukast showed full antagonist activity; montelukast and tomelukast showed partial antagonist activity, while MK571 showed no antagonist activity. The CYSLTR2 selective antagonists (HAMI and BayCysLT2) showed better antagonist activity than CYSLTR1 selective antagonists (montelukast and MK571) at GPR17. Like previous studies, we do not observe antagonist activity with P2RY12 antagonists (clopidogrel, prasugrel and ticlopidine). The cross activity of these antagonists indicates that GPR17 might be more closely related to the CYSLTR family than the P2RY family. Besides these promiscuous and weak antagonists, more potentGPR17 antagonists (I-116²⁹, UCB6651¹⁶, and PSB-22269³⁰) have been developed by the Müller group. I-116 and UCB6651 were derived from a common scaffold. UCB6651 showed nanomolar range potency in blocking GPR17 activation, while it was inert in four P2RYs (P2RY1, P2RY12, P2YR13 and P2RY14) and one CYSLTR (CYSLTR1). However, the effects of UCB6651 on CYSLTR2, a closely related receptor were not reported¹⁶. PSB-22269, an anthranilic acid derivative, has shown nanomolar range potency in a binding assay and functional assays, while selectivity data has not been reported³⁰. Compounds 978 and 527 showed high selectivity for GPR17, and no activities was observed in P2RYs (P2RY1 and P2RY12) and CYSLTRs (CYSLTR1 and CYSLTR2). We believe these novel selective scaffolds will serve as promising starting points for further optimization, potentially advancing the development of targeted therapies with improved efficacy and safety for treating diabetes and obesity.

Although the identity of the endogenous ligand of GPR17 is still controversial and elusive, the synthetic small molecule agonist MDL has been consistently validated by multiple groups including us^12,13,26. After activation by MDL, GPR17 was able to couple to major G protein and β-arrestin signaling pathways¹⁷. As downstream signaling pathways were equally activated, MDL was considered as an unbiased agonist for the receptor. In our recent study, we demonstrated functional selectivity or biased signaling of GPR17 variants identified from metabolic disease patients¹³. Eight nonsynonymous GPR17 variants showed diverse downstream signaling profiles after being activated by MDL. Compared to the wildtype receptor, the GPR17-V96M variant showed biased Gαi/o signaling, and the GPR17-D105N variant showed biased Gαi/o and β-arrestin signaling pathways. GPCR biased signaling has important implications for drug development, as targeting biased signaling can result in better therapeutic effects with fewer adverse effects⁴³. The diverse downstream signaling profiles of GPR17 variants in metabolic disorders might indicate the importance of biased signaling in disease and the possibility of developing biased ligands to probe the different downstream signaling of the receptor.

In conclusion, we identified two novel and selective antagonists (compounds 978 and 527) of GPR17 through high-throughput screening. Both antagonists blocked GPR17 downstream Gαi/o, Gαq and β-arrestin signaling with high selectivity for GPR17, but not the closely related purinergic and cysteinyl leukotriene receptors. One antagonist (978) and its analog (793) attenuated GPR17 signaling and promoted gut hormone, glucagon-like peptide-1 (GLP-1), secretion in enteroendocrine cells. Our findings suggested that antagonizing GPR17 might be a promising therapy for metabolic disorders.

Materials and Methods

Chemicals, libraries, and cell lines

The following four compounds were purchased in Tocris Bioscience (Minneapolis, MN): MDL29951 (Cat# 0467), ATP (Cat# 3245), L-692,585 (Cat# 2261), PF 05190457 (Cat# 6350). Leukotriene D4 (LTD4, Cat# 20310) and 2-Methylthioadenosine diphosphate (2MeS-ADP, Cat# 21230) were purchased from Cayman Chemical (Ann Arbor, Michigan). HAMI 3379 was purchased from MedChemExpress (Monmouth Junction, NJ). Recombinant Human FGF2 (Cat# 233-FB-010) was purchased from R&D Systems (Minneapolis, MN). FIIN-2 (Cat# S7714) was purchased from Selleck Chemicals (Houston, TX). Five in-house libraries (Sytravon, Genesis, NPC, Kinase, LOPAC) were screened. The Sytravon library is an in-house collection of small molecules characterized by chemically diverse and novel structures, with many featuring medicinal chemistry-tractable scaffolds. The Genesis library is a collection of modern chemical library emphasizing high-quality chemical starting points, sp3-enriched chemotypes, and core scaffolds that facilitate rapid acquisition and derivatization through medicinal chemistry. The NCATS Pharmaceutical Collection (NPC) is a curated library of compounds that have been approved for use by the Food and Drug Administration (FDA) and includes several approved molecules from related agencies in other countries. The Kinase inhibitor library is an in-house collection of small molecule inhibitors of kinases. The LOPAC^®1280 (LOPAC) is library of pharmacologically active compounds commercially available from Millipore Sigma; it contains drug-like molecules including many modulators of in the fields of cell signaling & neuroscience. PathHunter U2OS GPR17 β-arrestin cell line Catalog# (Catalog# 93–0633C3), PathHunter U2OS FGFR1 cell line (Catalog# 93–1090C3) and PathHunter U2OS GHSR1a β-arrestin cell line (Catalog#93–0242C3) were purchased from DiscoverX/Eurofin (Fremont, CA). 1321N1 GPR17 stable cell line (Catalog# Cr1525–3) was purchased from Multispan Inc (Hayward, CA). HEK293-hGPR17S cell line (Catalog# CB-80400–301), HEK293-hCysLTR1 cell line (Catalog# CB-80400–299), HEK293-hCysLTR2 cell line (Catalog# CB-80400–300), HEK293-hP2RY1 cell line (Catalog# CB-80400–302), and -hP2RY12 cells (Catalog# CB-80300–303) were purchased from Codex Biosolutions Inc (Gaithersburg, MD).

PathHunter β-arrestin recruitment assay

PathHunter U2OS GPR17 β-arrestin cells were maintained in the growth medium consisting of MEM (Thermo, # 11095080), 10% FBS (Thermo, # 10082) 1xPen/Strep, 1xGlutamine, 500μg/mL Geneticin (Thermo, # 10131035), and 250μg/mL Hygromycin B (Thermo, # 10687010).

For the high-throughput screen, cells were harvested with StemPro Accutase Cell Dissociation Reagent (Thermo, # A1110501), and seeded in 1536-well white flat bottom plates (Greiner, #789173-F) at the density of 500 cells/ 3μL/ well in AssayComplete Cell Plating Reagent 5 (Eurofin/DiscoverX, # 93–0563R5A). After incubated at 37°C, 5% CO₂ overnight, 20 nL/well of library compounds were added to the columns 5–48 of each plate with a pintool transfer (Kalypsis, San Diego, CA). As the positive control, 20 nL/well of 10mM HAMI 3379 (66.7 μM final concentration) was transferred to column 3 and 4 of each plate. The microplates were incubated at room temperature for 10 minutes, and then ~EC₈₀ concentration of agonist MDL29951 (final concentration 2.0 μM) was added into the microplate. After stimulation, cells were incubated 37°C, 5% CO₂ for 90 minutes. Finally, 1.5μL/ well of PathHunter detection reagent (DiscoverX# 93–0001L) was added to microplates with BioRAPTR FRD dispenser (Beckman Coulter, Brea, CA). The detection reagent was prepared by mixing Galacton Star Substrate, Emerald II solution and PathHunter buffer (supplied by the assay kit) together at 1:5:19 proportion accordingly prior to dispensing. The microplates were incubated at room temperature for 1 hour and then the luminescent signal was detected on ViewLux plate reader (PerkinElmer, Waltham, MA).

For the experiment of Schild regression analysis, 1321N1-GPR17 cells were seeded in 96-well plates. Cells were serum starved for 2 h and treated with varying concentrations of test compound and vehicle for 10 min in Stimulation Buffer 4 provided with the Cisbio HTRF Beta-Arrestin assay kit at room temperature. Cells were then treated with varying concentrations of the GPR17 agonist MDL in stimulation buffer 4 for 30 min at room temperature. The Cisbio HTRF Beta-Arrestin recruitment kit was then used to quantify the β-arrestin recruitment as specified in the manufacturer’s protocol. The HTRF signal was read using a Synergy 4 plate reader. Schild regression analysis was calculated using 4-variable concentration-response curves.

Calcium mobilization assay

1321N1 GPR17 stable cell line was maintained in the growth medium consisting of DMEM (Thermo, # 10566016), 10% FBS (Thermo, # 10082) 1xPen/Strep, 1xGlutamine, 500μg/mL puromycin (Thermo, # A1113802).

For the high-throughput screen, cells were harvested with StemPro Accutase Cell Dissociation Reagent (Thermo, # A1110501), and seeded in 1536-well μclear black microplates (Greiner, #782097) at the density of 1000 cells/ 3μL/ well in DMEM medium. After incubated at 37°C, 5% CO₂ overnight, 3 μL Calbryte^™ 520NW dye-loading solution (AAT Bioquest, # 36319) was dispensed into each well. The dye-loading solution was prepared by mixing of 1xHHBS buffer, 10X Pluronic^® F127 Plus and Calbryte^™ 520NW stock solution (supplied by the assay kit) accordingly the manual from the kit. The dye-loading plates were incubated at 37°C, 5% CO₂ for 30 minutes, and then incubated at room temperature for another 15–30 minutes. 20 nL/well of library compounds were transferred to the columns 5–48 of each plate by using Echo 525 acoustic dispenser (Beckman Coulter, CA). The microplates were incubated at room temperature for 10 minutes and then loaded on the FLIPR Penta High-Throughput Cellular Screening System (Molecular Devices, San Jose, CA). The baseline fluorescence signal was measured for 10 seconds. MDL (~ EC₈₀ concentration) was added to stimulate the cells, and then fluorescence signals were measured for 3 minutes at the 1-second interval. The relative fluorescence unit (RFU) was calculated by maximum signal – minimum signal.

For the experiment of Schild regression analysis, 1321N1-GPR17 cells were seeded into 384-well plates at a density of 7,500 cells/well. The following day, media was replaced with assay buffer (HBSS, 20 mM HEPES) and cells were loaded with Calcium 6 dye for 2 h at 37°C and 5% CO2. The cells were allowed to equilibrate to room temperature for 15 minutes and cells were incubated in test compound or vehicle for 10 minutes. Subsequently, MDL was added and plates were read every 0.5 s for 3 minutes on a Hamamatsu FDSS/μCell.

cAMP accumulation assay

hGPR17–1321N1 cells were seeded at 3,000 cells/well into white, opaque 384-well plates and grown overnight at 37°C. The next day, cells were treated with varying concentrations of compound 527 or compound 978 for 10min at room temperature. Cells were treated with 30nM MDLin stimulation buffer (3 μM forskolin/500uM IBMX) for 1 hour at room temperature. HRTF assays were performed using the homogenous time-resolved fluorescence (HTRF) HiRange cAMP detection kit (Cisbio, Bedford, MA) according to the manufacturer’s instructions. Plates were incubated at room temperature for 30 minutes and FRET signals (665 and 620 nm) were read using Biotek Synergy Neo 2 plate reader (Agilent, Santa Clara, CA). HTRF signal was calculated as the ratio of signal from the 665 nm (acceptor) and 620 nm (donor) channels.

ACTone cAMP assay

ACTOne-hP2RY12 cells (Codex Biosolutions, #CB-80300–303) were plated on a 384-well black clear plate at the density of about 12,000 cells/well in 20 ul culture medium. After incubated at 37°C, 5% CO₂ overnight, 20 ul of 1x MP dye solution (Codex BioSolutions Inc) was loaded into each well. The plate was incubated at room temperature in the dark for 2 hours. Compounds were serially diluted (5x of the final concentration) in 1x DPBS solution. 10 ul of the diluted compounds were added into each well and incubated at room temperature in the dark for 25 min. The baseline fluorescent intensity (F0) of each well was recorded on FlexStation Multi-Mode Microplate Reader (Molecular Devices, San Jose, CA). 12.5 ul of 5x stimulation solution (0.1 nM 2MeS-ADP, 125 μM Ro20–1724 and 1.5 μM Iso in 1x DPBS) was then added and incubated at room temperature in the dark for 15 min. The fluorescent intensity (Ft) of each well was recorded again on FlexStation. The ratios of Ft/F0 were used to plot the dose response curves.

Cisbio β-arrestin HRTF Assay

hGPR17–1321N1 cells were seeded at 30,000 cells/well into white, opaque 96-well plates and grown overnight. The next day, cells were serum starved then treated with varying concentrations of the stated compounds, or vehicle control, for 10min at room temperature. Cells where then treated with varying concentrations of MDL29,951 for 30min at room temperature. Cisbio manufactures protocol was then followed for the remainder of the experiment. HTRF signal was calculated as the ratio of signal from the 665 nm (acceptor) and 620 nm (donor) channels.

GLUTag cell GLP-1 secretion assay

GLUTag cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM), supplemented with 10% fetal bovine serum, and 1% penicillin-streptomycin in a humidified incubator at 37°C and 5% CO2. For adenovirus transduction, 75,000 cells/well GLUTag cells were seeded into poly-D-lysine coated 96-well plates. The next day, cells were transduced with Ad-CMV-HA-hGPR17L-EF1α-mCherry or Ad-EF1α-mCherry control adenovirus (625–1250 viral particles per cell at time of seeding) as indicated. After approximately 30 h, adenovirus-containing media was changed to serum-free growth media for overnight serum starvation. The following day, GLUTag cells were used for cAMP signaling assays or GLP-1 secretion assays.

GLUTag cells were transduced with adenovirus as specified above and then tested for cAMP signaling. Media was removed and cells were washed with cAMP stimulation buffer (Hanks balanced salt solution (HBSS), 20 mM HEPES, pH 7.4) and then incubated in cAMP stimulation buffer for 30 min at 37°C, 5% CO2. Cells were pre-treated with antagonist for 15 min prior to 30 min treatment with agonist and forskolin in the presence of 500 μM 3-isobutyl-1-methylxanthine (IBMX) at 37°C and 5% CO2. After treatment, stimulation buffer was decanted, and cAMP lysis buffer (50 mM HEPES, 10 mM CaCl2, 0.35% Triton X-100, and 500 μM IBMX) was added and shaken at 700 rpm for 1 h. Lysate was transferred to a 384-well plate and the cAMP levels were quantified using the Cisbio HTRF cAMP (Gi kit) according to the manufacturer’s protocol.

GLUTag cells were transduced with adenovirus as specified above and then tested for GLP-1 secretion. Media was removed and cells were washed two times with GLP-1 secretion buffer (138 mM NaCl, 4.5 mM KCl, 4.2 mM NaHCO3, 2.5 mM CaCl2, 1.2 mM MgCl2, 10 mM HEPES, 1 mM glucose, and 0.1% bovine serum albumin, pH 7.4). Cells were stimulated for 2 h with indicated treatments diluted in GLP-1 secretion buffer for a total volume of 200 μl/well at 37°C and 5% CO2. Plates were then centrifuged at 300xg for 1 min at 4°C and 75 μl of the supernatant was removed and frozen at −80°C until subsequent GLP-1 quantification. Total GLP-1 was quantified using the Meso Scale Discovery (Rockville, MD) Total GLP-1 V-PLEX kit (K1503PD) according to manufacturer’s instructions.

HTS Data Analysis

Analysis of activity of compounds was performed as previously described⁴⁴. Briefly, raw plate read for each compound was first normalized relative to the HAMI control wells and DMSO control wells as follows: % Activity = ((RLU_compound – RLU_DMSO)/(RLU_DMSO – RLU_HAMI)) × 100, where RLU_compound denotes the compound well values, RLU_HAMI denotes the median value of the HAMI control wells, and RLU_DMSO denotes the median values of the DMSO-only wells, and then corrected by applying a NCGC in-house pattern correction algorithm using compound-free control plates (i.e., DMSO-only plates) at the beginning and end of the compound plate stack⁴⁵. For dose-response curves in confirmatory assay, counter assays and orthogonal assays of each compound were fitted to a four-parameter Hill equation⁴⁶ yielding concentrations of half-maximal activity (IC₅₀) and maximal response (efficacy) values. Compounds were designated as Class 1–4 according to the type of dose–response curve observed^44,47. Curve classes are heuristic measures of data confidence, classifying concentration–responses based on efficacy, the number of data points observed above background activity, and the quality of fit. Compounds with −1.1, −1.2, −2.1, −2.2, −3.0 were considered active. Class 4 compounds were considered inactive. Compounds with other curve classes were deemed inconclusive. Active compounds were cherrypicked for follow up experiments.

Homology modeling and molecular docking:

The homology model of human GPR17 was constructed using the molecular Operating Environment (MOE) by Chemical Computing Group (https://www.chemcomp.com). The crystal structure of the human cysteinyl leukotriene receptor 2 (CysLT2R, PDB ID: 6rz6) was employed as the template for homology modeling of GPR17, since both the AlphaFold (https://alphafold.ebi.ac.uk/ version Nov. 2022) predicted and Cryo-EM (PDB ID: 7Y89) structures of GPR17 adopted a self-activated conformation, where the orthosteric ligand binding site was occupied by the extra-cellular loops 2. CysLT2R and GLP17 share 34.5% of identical residues. Molecular docking was performed by using the Dock module in the MOE using the force field of Amber:EHT. The protein structure was allowed to be a certain degree of flexibility by the Induced Fit algorithm.

General Methods for Chemistry:

All air- or moisture-sensitive reactions were performed under positive pressure of nitrogen with oven-dried glassware. Anhydrous solvents such as dichloromethane, N,N-dimethylformamide (DMF), acetonitrile, methanol and triethylamine were purchased from Sigma-Aldrich. Preparative purification was performed on a Waters semi-preparative HPLC system. The column used was a Phenomenex Luna C18 (5 micron, 30 × 75 mm) at a flow rate of 45 mL/min. Purification was performed using Method Acidic Standard Gradient (10:90 MeCN with 0.1% TFA/ deionized water with 0.1% TFA ramped to 100% deionized water with 0.1% TFA) and Method Basic Standard Gradient (10:90 MeCN with 0.1% NH₄OH/ deionized water with 0.1% NH₄OH ramped to 100% deionized water with 0.1% NH₄OH) over 8 minutes unless otherwise noted. Fraction collection was triggered by UV detection (220 nM). Analytical analysis was performed on an Agilent LC/MS (Agilent Technologies, Santa Clara, CA). Purity analysis was determined using a 7 minute gradient of 4% to 100% acetonitrile (containing 0.025% trifluoroacetic acid) in water (containing 0.05% trifluoroacetic acid) with an 8 minute run time at a flow rate of 1 mL/min. A Phenomenex Luna C18 column (3 micron, 3 × 75 mm) was used at a temperature of 50 °C using an Agilent Diode Array Detector.

Mass determination was performed using an Agilent 6130 mass spectrometer with electrospray ionization in the positive mode. ¹H NMR spectra were recorded on Varian 400 MHz spectrometers. Chemical shifts are reported in ppm with non-deuterated solvent (DMSO-d5 peak at 2.50 ppm) as internal standard for DMSO-d6 solutions. All of the analogs tested in the biological assays have a purity greater than 95% based on LCMS analysis. High resolution mass spectrometry was recorded on Agilent 6210 Time-of-Flight LC/MS system. Confirmation of molecular formulae was accomplished using electrospray ionization in the positive mode with the Agilent Masshunter software (version B.02).

Synthesis of 978 and related analogs

Method 1 Step 1: 8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-one (Figure 4A, c):

A mixture of a substituted phenylhydrazine (Figure 4A, a) (1.0 eq), 1,2-cyclohexanedione (Figure 4A, b) (2.0 mmol), and water (50 mL) was stirred at room temperature (23 °C) in a 250 mL round bottom flask. After 16 h, 10 mL of aqueous 1 M HCl was added and the reaction was heated to reflux. After 12 h, the reaction was refrigerated to induce crystallization. The resulting solid was filtered and air dried to provide the crude product. The desired product was purified by flash column chromatography to afford 8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-one (Figure 4A, c).

Step 2: 8-Chloro-2,3,4,9-tetrahydro-1H-carbazol-1-amine (Figure 4A, d):

Solid NH₄OAc (5.0 mmol) was added to a stirring solution of 8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-one (Figure 4A, c) (0.5 mmol) dissolved in 20 mL solvent grade CH₃OH, and this mixture was allowed to stir at room temperature (23 °C) for 3–6 h. Upon complete consumption of starting material, as determined by thin layer chromatography, solid NaCNBH₃ (2.5 mmol) was added and the temperature was raised to 60 °C. After stirring 12–16 h, the reaction was cooled to room temperature (23 °C) and treated with aqueous 1 M HCl. The mixture was extracted with ethyl acetate, and the combined organic layers were washed with brine and dried over anhydrous Na₂SO₄, filtered, and concentrated under reduced pressure to afford the crude product. The desired product was purified by flash column chromatography to afford 8-Chloro-2,3,4,9-tetrahydro-1H-carbazol-1-amine (Figure 4A, d).¹H NMR (400 MHz, DMSO-d₆) δ 10.92 (s, 1H), 7.31 (d, J = 7.8 Hz, 1H), 7.06 (d, J = 7.6 Hz, 1H), 6.92 (t, J = 7.7 Hz, 1H), 3.95 (dd, J = 7.2, 5.5 Hz, 1H), 2.56 (t, J = 6.1 Hz, 2H), 2.26 (s, 2H), 2.11 – 1.86 (m, 1H), 1.77 – 1.59 (m, 1H), 1.53 (dddd, J = 12.4, 9.7, 7.1, 2.1 Hz, 1H).

Method 2 (General Procedure): N-(8-Chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3,4-dimethoxybenzamide (Table 2, Compound 1 or Figure 4A, f):

To a solution of 8-Chloro-2,3,4,9-tetrahydro-1H-carbazol-1-amine (Figure 4A, d) (1.0 eq) in DMF (Volume: 2 ml) was added HATU (2-(3H-[1,2,3]triazolo[4,5-b]pyridin-3-yl)-1,1,3,3-tetramethylisouronium hexafluorophosphate (2.0 eq) and N-ethyl-N-isopropylpropan-2-amine (3.0 eq) and 3,4-dimethoxybenzoic acid (1.0 eq). The reaction mixture was stirred for 12 hrs at room temperature and diluted with water and extracted with 3 × 10 mL DCM. The organic layer was washed with brine, dried and concentrated. The crude mixture was diluted with DMSO and purified by reverse phase chromatography (Method Acidic Standard Gradient) to N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3,4-dimethoxybenzamide; LCMS: m/z (M+H)⁺ = 386; ¹H NMR (400 MHz, dmso) δ 11.15 (s, 1H), 8.64 (d, J = 7.5 Hz, 1H), 7.60 – 7.50 (m, 2H), 7.39 (d, J = 7.8 Hz, 1H), 7.10 (dd, J = 7.6, 1.0 Hz, 1H), 7.01 – 6.92 (m, 2H), 5.31 (q, J = 5.3 Hz, 1H), 3.77 (d, J = 4.9 Hz, 6H), 2.71 (dt, J = 15.4, 5.0 Hz, 2H), 2.66 – 2.56 (m, 2H), 2.03 – 1.85 (m, 2H).

N-(8-Chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3,4,5-trimethoxybenzamide (Table 2, Compound 2):

This compound was prepared from Method 2 using 3,4,5-trimethoxybenzoic acid to afford N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3,4,5-trimethoxybenzamide. LCMS: m/z (M+H)⁺ = 415; ¹H NMR (400 MHz, dmso) δ 11.18 (s, 1H), 8.74 (d, J = 7.6 Hz, 1H), 7.39 (d, J = 7.8 Hz, 1H), 7.28 (s, 2H), 7.11 (dd, J = 7.6, 1.0 Hz, 1H), 6.96 (t, J = 7.7 Hz, 1H), 5.31 (dd, J = 7.5, 4.6 Hz, 1H), 3.79 (s, 6H), 3.68 (s, 3H), 2.72 (dt, J = 15.5, 4.9 Hz, 2H), 2.67 – 2.57 (m, 2H), 2.06 – 1.95 (m, 1H), 1.98 – 1.77 (m, 1H).

N-(8-Chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)benzo[d][1,3]dioxole-5-carboxamide (Table 2, Compound 3):

This compound was prepared from Method 2 using benzo[d][1,3]dioxole-5-carboxylic acid to afford N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)benzo[d][1,3]dioxole-5-carboxamide LCMS: m/z (M+H)⁺ = 369; 1H NMR (400 MHz, dmso) δ 11.13 (s, 1H), 8.60 (d, J = 7.5 Hz, 1H), 7.52 (dd, J = 8.2, 1.7 Hz, 1H), 7.47 (d, J = 1.8 Hz, 1H), 7.39 (d, J = 7.8 Hz, 1H), 7.10 (dd, J = 7.6, 1.0 Hz, 1H), 7.00 – 6.91 (m, 2H), 6.06 (s, 2H), 5.27 (dd, J = 7.4, 4.8 Hz, 1H), 2.65 (dtd, J = 21.3, 15.2, 5.3 Hz, 2H), 2.05 – 1.73 (m, 4H).

N-(8-Chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-4-methoxy-3-methylbenzamide (Table 2, Compound 4):

This compound was prepared from Method 2 using 4-methoxy-3-methylbenzoic acid to afford N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-4-methoxy-3-methylbenzamide. LCMS: m/z (M+H)⁺ = 369; ¹H NMR (400 MHz, cdcl3) δ 8.85 (s, 1H), 7.68 (d, J = 2.2 Hz, 1H), 7.59 (d, J = 8.2 Hz, 1H), 7.47 – 7.36 (m, 2H), 7.17 (d, J = 7.6 Hz, 1H), 7.02 (t, J = 7.8 Hz, 1H), 6.33 (d, J = 7.2 Hz, 1H), 5.38 (q, J = 5.7 Hz, 1H), 3.79 (s, 3H), 2.85 – 2.67 (m, 2H), 2.44 (s, 3H), 2.31 (td, J = 9.7, 5.7 Hz, 2H), 1.99–1.96 (m 2H).

N-(8-Chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-4-methoxy-3-(methylsulfonyl)benzamide (Table 2, Compound 5):

This compound was prepared from Method 2 using 4-methoxy-3-(methylsulfonyl)benzoic acid to afford N-(8-Chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-4-methoxy-3-(methylsulfonyl)benzamide. LCMS: m/z (M+H)⁺ = 432; ¹H NMR (400 MHz, cdcl3) δ 9.24 (s, 1H), 8.05 (d, J = 7.1 Hz, 1H), 7.94 (dd, J = 8.3, 6.8 Hz, 1H), 7.90 – 7.81 (m, 2H), 7.33 (d, J = 7.8 Hz, 1H), 7.08 (d, J = 7.6 Hz, 1H), 6.94 (t, J = 7.7 Hz, 1H), 5.36 (q, J = 6.1 Hz, 1H), 3.79 (s, 3H), 3.18 (s, 3H), 2.69 (q, J = 6.7 Hz, 2H), 2.04 – 1.85 (m, 4H).

N-(8-Chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3-fluoro-4-(methylsulfonyl)benzamide (Table 2, Compound 6):

This compound was prepared from Method 2 using 3-fluoro-4-(methylsulfonyl)benzoic acid to afford N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3-fluoro-4-(methylsulfonyl)benzamide. LCMS: m/z (M+H)⁺ = 421; ¹H NMR (400 MHz, cdcl3) δ 9.24 (s, 1H), 8.05 (d, J = 7.1 Hz, 1H), 7.94 (dd, J = 8.3, 6.8 Hz, 1H), 7.90 – 7.81 (m, 2H), 7.33 (d, J = 7.8 Hz, 1H), 7.08 (d, J = 7.6 Hz, 1H), 6.94 (t, J = 7.7 Hz, 1H), 5.36 (q, J = 6.1 Hz, 1H), 3.18 (s, 3H), 2.69 (q, J = 6.7 Hz, 2H), 2.04 – 1.85 (m, 4H).

N-(8-Chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-4-(methylsulfonyl)-3-nitrobenzamide (Table 2, Compound 7):

This compound was prepared from Method 2 using 4-(methylsulfonyl)-3-nitrobenzoic acid to afford N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-4-(methylsulfonyl)-3-nitrobenzamide. LCMS: m/z (M+H)⁺ = 448; ¹H NMR (400 MHz, cdcl3) δ 8.32 (d, J = 1.6 Hz, 1H), 8.26 – 8.15 (m, 2H), 7.34 (dd, J = 7.8, 1.0 Hz, 1H), 7.10 (dd, J = 7.7, 1.0 Hz, 1H), 6.95 (t, J = 7.7 Hz, 1H), 5.35 (t, J = 5.3 Hz, 1H), 4.68 (s, 1H), 3.39 (s, 3H), 2.79 – 2.60 (m, 2H), 2.20 (ddt, J = 12.9, 8.7, 4.2 Hz, 2H), 2.04 – 1.87 (m, 2H).

N-(8-Chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3,4-dinitrobenzamide (Table 2, Compound 8):

This compound was prepared from Method 2 using 3,4-dinitrobenzoic acid to afford N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3,4-dinitrobenzamide. LCMS: m/z (M+H)⁺ = 415; ¹H NMR (400 MHz, cdcl₃) δ 10.79 (s, 1H), 8.79 (s, 1H), 8.55 (d, J = 2.2 Hz, 1H), 8.09 (dd, J = 8.8, 2.2 Hz, 1H), 7.40 (d, J = 7.8 Hz, 1H), 7.17 (d, J = 7.6 Hz, 1H), 7.02 (t, J = 7.8 Hz, 1H), 6.37 (d, J = 7.1 Hz, 1H), 5.40 (d, J = 6.3 Hz, 1H), 2.87 – 2.68 (m, 2H), 2.31 (dt, J = 13.9, 4.3 Hz, 2H), 2.02–1.97 (m, 2H).

N-(8-Chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-4-fluoro-3-methoxybenzamide (Table 2, Compound 9):

This compound was prepared from Method 2 using 4-fluoro-3-methoxybenzoic acid to afford N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-4-fluoro-3-methoxybenzamide. LCMS: m/z (M+H)⁺ = 373; ¹H NMR (400 MHz, cdcl3) δ 7.49 (dd, J = 8.2, 2.1 Hz, 1H), 7.34 (dd, J = 7.8, 1.0 Hz, 1H), 7.29 – 7.21 (m, 1H), 7.13 – 6.99 (m, 2H), 6.95 (t, J = 7.7 Hz, 1H), 5.33 (t, J = 5.2 Hz, 1H), 3.89 (s, 3H), 2.79 – 2.60 (m, 2H), 2.26 – 2.12 (m, 2H), 1.95 (dq, J = 14.0, 5.4 Hz, 2H).

4-Bromo-N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3-methylbenzamide (Table 2, Compound 10):

This compound was prepared from Method 2 using 4-bromo-3-methylbenzoic acid to afford 4-bromo-N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3-methylbenzamide. LCMS: m/z (M+H)⁺ = 419; ¹H NMR (400 MHz, cdcl3) δ 8.85 (s, 1H), 7.68 (d, J = 2.2 Hz, 1H), 7.59 (d, J = 8.2 Hz, 1H), 7.47 – 7.36 (m, 2H), 7.17 (d, J = 7.6 Hz, 1H), 7.02 (t, J = 7.8 Hz, 1H), 6.33 (d, J = 7.2 Hz, 1H), 5.38 (q, J = 5.7 Hz, 1H), 2.85 – 2.67 (m, 2H), 2.44 (s, 3H), 2.31 (td, J = 9.7, 5.7 Hz, 2H), 1.99–1.96 (m 2H).

N-(8-Chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-2-methyl-4-nitrobenzamide (Table 2, Compound 11):

This compound was prepared from Method 2 using 2-methyl-4-nitrobenzoic acidto afford N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-2-methyl-4-nitrobenzamide. LCMS: m/z (M+H)⁺ = 384; ¹H NMR (400 MHz, cdcl3) δ 9.66 (d, J = 7.1 Hz, 1H), 8.02 – 7.72 (m, 1H), 7.45 (t, J = 8.5 Hz, 1H), 7.27 – 7.15 (m, 2H), 7.03 – 6.71 (m, 2H), 5.18 (d, J = 7.8 Hz, 1H), 4.54 (s, 1H), 2.53 (s, 3H), 2.42 – 2.28 (m, 4H), 1.79 (d, J = 28.9 Hz, 2H).

N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-2,6-difluoro-4-nitrobenzamide (Table 2, Compound 12):

This compound was prepared from Method 2 using 2,6-difluoro-4-nitrobenzoic acid to afford N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-2,6-difluoro-4-nitrobenzamide. LCMS: m/z (M+H)⁺ = 406; ¹H NMR (400 MHz, cdcl3) δ 8.66 (s, 1H), 7.58 – 7.50 (m, 2H), 7.41 (d, J = 7.9 Hz, 1H), 7.19 (d, J = 7.6 Hz, 1H), 7.04 (t, J = 7.8 Hz, 1H), 6.39 (d, J = 7.2 Hz, 1H), 5.39 (q, J = 5.7 Hz, 1H), 2.86 – 2.67 (m, 2H), 2.31 (td, J = 10.7, 5.7 Hz, 2H), 2.00 (ddt, J = 30.7, 10.3, 5.9 Hz, 2H).

N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-4-fluoro-3-nitrobenzamide (Table 2, Compound 13):

This compound was prepared from Method 2 using 4-fluoro-3-nitrobenzoic acid to afford N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-4-fluoro-3-nitrobenzamide. LCMS: m/z (M+H)⁺ = 388; 1H NMR (400 MHz, cdcl3) δ 8.65 – 8.60 (m, 1H), 7.92 – 7.84 (m, 2H), 7.79 (dd, J = 8.2, 2.2 Hz, 1H), 7.74 – 7.67 (m, 1H), 7.58 – 7.50 (m, 2H), 5.39 (q, J = 5.7 Hz, 1H), 2.84–2.65 (m, 4H), 2.56–2.30 (m, 2H).

N-(8-Chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-2-phenyl-2-(thiophene-2-sulfonamido)acetamide (Table 2, Compound 14):

This compound was prepared from Method 2 using 2-phenyl-2-(thiophene-2-sulfonamido)acetic acid to afford N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-2-phenyl-2-(thiophene-2-sulfonamido)acetamide. LCMS: m/z (M+H)⁺ = 500.1; ¹H NMR (400 MHz, dmso) δ 10.89 (s, 1H), 10.72 (s, 1H), 8.80 (dd, J = 20.1, 9.4 Hz, 1H), 8.67 (d, J = 7.1 Hz, 1H), 7.85 (dd, J = 5.0, 1.4 Hz, 1H), 7.61 (dd, J = 5.0, 1.4 Hz, 1H), 7.48 (dd, J = 3.7, 1.4 Hz, 1H), 7.43 – 7.30 (m, 2H), 7.30 – 7.05 (m, 2H), 6.96 (dt, J = 12.7, 7.7 Hz, 2H), 6.86 (dd, J = 5.0, 3.7 Hz, 1H), 4.88 – 4.81 (m, 1H), 2.67 (t, J = 17.0 Hz, 2H), 2.52 (d, J = 7.1 Hz, 2H), 1.71 (dd, J = 17.1, 6.9 Hz, 2H).

4-((8-Chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)carbamoyl)benzoic acid (Table 2, Compound 15):

This compound was prepared from Method 2 using 4-(methoxycarbonyl)benzoic acid followed by hydrolysis using NaOH in MeOH:H₂O at rt to afford 4-((8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)carbamoyl)benzoic acid. LCMS: m/z (M+H)⁺ = 369; ¹H NMR (400 MHz, dmso) δ 13.14 (s, 1H), 11.20 (s, 1H), 8.96 (d, J = 7.6 Hz, 1H), 8.04 – 7.92 (m, 4H), 7.39 (d, J = 7.8 Hz, 1H), 7.11 (dd, J = 7.6, 0.9 Hz, 1H), 6.96 (t, J = 7.7 Hz, 1H), 5.37 – 5.29 (m, 1H), 2.75 – 2.57 (m, 2H), 1.98 – 1.85 (m, 4H)

4-Bromo-N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)benzamide (Table 2, Compound 16):

This compound was prepared from Method 2 using 4-bromobenzoic acid to afford 4-bromo-N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)benzamide. LCMS: m/z (M+H)⁺ = 404; ¹H NMR (400 MHz, dmso) δ 11.17 (s, 1H), 8.86 (d, J = 7.6 Hz, 1H), 7.90 – 7.82 (m, 2H), 7.68 – 7.60 (m, 2H), 7.39 (d, J = 7.8 Hz, 1H), 7.10 (dd, J = 7.6, 1.0 Hz, 1H), 6.96 (t, J = 7.7 Hz, 1H), 5.33 – 5.25 (m, 1H), 2.74 – 2.57 (m, 2H), 2.04 – 1.75 (m, 4H).

N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-4-methylbenzamide (Table 2, Compound 17):

This compound was prepared from Method 2 using 4-methylbenzoic acid to afford N-(8-chloro-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-4-methylbenzamide. LCMS: m/z (M+H)⁺ = 339; ¹H NMR (400 MHz, dmso) δ 11.13 (s, 1H), 8.67 (d, J = 7.6 Hz, 1H), 7.85 – 7.79 (m, 2H), 7.39 (d, J = 7.8 Hz, 1H), 7.22 (d, J = 8.0 Hz, 2H), 7.10 (dd, J = 7.6, 0.9 Hz, 1H), 6.96 (t, J = 7.7 Hz, 1H), 5.30 (d, J = 6.7 Hz, 1H), 2.65 – 2.56 (m, 2H), 2.33 (s, 3H), 2.00 – 1.87 (m, 2H), 1.79 (d, J = 12.2 Hz, 2H).

3,4-Dimethoxy-N-(2,3,4,9-tetrahydro-1H-carbazol-1-yl)benzamide (Table 3, Compound 18):

This compound was prepared from Method 2 using 2,3,4,9-tetrahydro-1H-carbazol-1-amine and 3,4-dimethoxybenzoic acid to afford 3,4-dimethoxy-N-(2,3,4,9-tetrahydro-1H-carbazol-1-yl)benzamide. LCMS: m/z (M+H)⁺ = 351; ¹H NMR (400 MHz, dmso) δ 10.71 (s, 1H), 8.64 (d, J = 8.1 Hz, 1H), 7.61 – 7.51 (m, 2H), 7.51 – 7.35 (m, 1H), 7.30 – 7.24 (m, 1H), 7.05 – 6.94 (m, 2H), 6.93 (ddd, J = 7.9, 7.1, 1.1 Hz, 1H), 5.33 (q, J = 5.1 Hz, 1H), 3.78 (d, J = 5.0 Hz, 6H), 2.66 (dt, J = 6.5, 3.2 Hz, 2H), 2.02 (tt, J = 8.8, 4.3 Hz, 2H), 1.92 – 1.75 (m, 2H).

N-(8-Bromo-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3,4-dimethoxybenzamide (Table 3, Compound 19):

This compound was prepared from Method 2 using 8-bromo-2,3,4,9-tetrahydro-1H-carbazol-1-amine and 3,4-dimethoxybenzoic acid to afford N-(8-bromo-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3,4-dimethoxybenzamide. LCMS: m/z (M+H)⁺ = 429; ¹H NMR (400 MHz, cdcl3) δ 8.85 (s, 1H), 7.48 – 7.41 (m, 2H), 7.35 – 7.24 (m, 2H), 6.97 (t, J = 7.7 Hz, 1H), 6.85 (d, J = 8.4 Hz, 1H), 6.31 (d, J = 7.2 Hz, 1H), 5.39 (q, J = 5.8 Hz, 1H), 3.94 (d, J = 14.0 Hz, 6H), 2.85 – 2.67 (m, 2H), 2.31 (td, J = 9.2, 5.6 Hz, 2H), 1.99 (tt, J = 6.7, 4.1 Hz, 2H).

N-(8-bromo-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3,5-dimethoxybenzamide (Table 3, Compound 20):

This compound was prepared from Method 2 using 8-bromo-2,3,4,9-tetrahydro-1H-carbazol-1-amine and 3,5-dimethoxybenzoic acid to afford N-(8-bromo-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3,5-dimethoxybenzamide. LCMS: m/z (M+H)⁺ = 429; ¹H NMR (400 MHz, dmso) δ 11.01 (s, 1H), 8.74 (d, J = 7.4 Hz, 1H), 7.43 (d, J = 7.8 Hz, 1H), 7.25 (d, J = 7.6 Hz, 1H), 7.09 (d, J = 2.3 Hz, 2H), 6.91 (t, J = 7.7 Hz, 1H), 6.60 (t, J = 2.3 Hz, 1H), 5.33 – 5.25 (m, 1H), 3.75 (s, 6H), 2.75 – 2.65 (m, 2H), 2.60 (d, J = 15.5 Hz, 2H), 2.02 – 1.85 (m, 2H).

N-(6,8-Dimethyl-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3,4-dimethoxybenzamide (Table 3, Compound 21):

This compound was prepared from Method 2 using 5,8-dimethyl-2,3,4,9-tetrahydro-1H-carbazol-1-amine and 3,4-dimethoxybenzoic acid to afford N-(6,8-dimethyl-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3,4-dimethoxybenzamide. LCMS: m/z (M+H)⁺ = 379; ¹H NMR (400 MHz, dmso) δ 10.52 (s, 1H), 8.61 (d, J = 7.8 Hz, 1H), 7.82 (s, 1H), 7.64 – 7.51 (m, 1H), 7.51 – 7.41 (m, 1H), 6.65 (s, 1H), 5.33 – 5.25 (m, 1H), 3.77 (dd, J = 4.5, 1.8 Hz, 6H), 2.65–2.63 (m, 2H), 2.62 – 2.49 (m, 2H), 2.34 (s, 3H), 2.30 (s, 3H), 1.90 – 1.76 (m, 2H).

N-(6,8-Dimethyl-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3,5-dimethoxybenzamide (Table 3, Compound 22):

This compound was prepared from Method 2 using 5,8-dimethyl-2,3,4,9-tetrahydro-1H-carbazol-1-amine and 3,5-dimethoxybenzoic acid to afford N-(6,8-dimethyl-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3,5-dimethoxybenzamide. LCMS: m/z (M+H)⁺ = 379; ¹H NMR (400 MHz, cdcl3) δ 9.55 (s, 1H), 8.34 (dd, J = 28.8, 4.7 Hz, 1H), 8.05 (dd, J = 8.8, 2.3 Hz, 1H), 7.17 (d, J = 7.8 Hz, 1H), 6.93 (dd, J = 26.5, 8.2 Hz, 1H), 6.77 (t, J = 7.7 Hz, 1H), 5.20 (d, J = 6.5 Hz, 1H), 3.80 (s, 6H), 2.53 (d, J = 8.0 Hz, 2H), 2.38 (s, 6H), 2.04 – 1.61 (m, 4H).

N-(5,8-dimethyl-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3,4-dimethoxybenzamide (Table 3, Compound 23):

This compound was prepared from Method 2 using 5,8-dimethyl-2,3,4,9-tetrahydro-1H-carbazol-1-amine and 3,4-dimethoxybenzoic acid to afford N-(5,8-dimethyl-2,3,4,9-tetrahydro-1H-carbazol-1-yl)-3,4-dimethoxybenzamide. LCMS: m/z (M+H)⁺ = 379; ¹H NMR (400 MHz, dmso) δ 10.57 (s, 1H), 8.59 (d, J = 7.7 Hz, 1H), 7.60 – 7.50 (m, 1H), 6.97 (d, J = 8.4 Hz, 1H), 6.66 (d, J = 7.1 Hz, 1H), 6.55 (d, J = 7.1 Hz, 1H), 5.29 (d, J = 6.7 Hz, 1H), 3.77 (d, J = 5.0 Hz, 6H), 2.88 (dd, J = 15.7, 7.0 Hz, 2H), 2.52 (s, 3H), 2.32 (s, 3H), 1.94 (d, J = 7.1 Hz, 2H), 1.82 (ddd, J = 18.2, 9.0, 3.5 Hz, 2H).

1-(3,4-Dimethoxybenzamido)-2,3,4,9-tetrahydro-1H-carbazole-8-carboxylic acid (Table 3, Compound 24):

This compound was prepared from Method 2 using 1-amino-2,3,4,9-tetrahydro-1H-carbazole-8-carboxylic acid and 3,4-dimethoxybenzoic acid to afford 1-(3,4-dimethoxybenzamido)-2,3,4,9-tetrahydro-1H-carbazole-8-carboxylic acid. LCMS: m/z (M+H)⁺ = 395; ¹H NMR (400 MHz, dmso) δ 10.57 (s, 1H), 8.59 (d, J = 7.7 Hz, 1H), 7.60 – 7.50 (m, 2H), 6.97 (d, J = 8.4 Hz, 1H), 6.66 (d, J = 7.1 Hz, 1H), 6.55 (d, J = 7.1 Hz, 1H), 5.29 (d, J = 6.7 Hz, 1H), 3.85 (d, J = 4.0 Hz, 3H), 3.77 (d, J = 5.0 Hz, 3H), 2.88 (dd, J = 15.7, 7.0 Hz, 2H), 1.94 (d, J = 7.1 Hz, 2H), 1.82 (ddd, J = 18.2, 9.0, 3.5 Hz, 2H).

1-(3,4-Dimethoxybenzamido)-2,3,4,9-tetrahydro-1H-carbazole-7-carboxylic acid (Table 3, Compound 25):

This compound was prepared from Method 2 using 1-amino-2,3,4,9-tetrahydro-1H-carbazole-7-carboxylic acid and 3,4-dimethoxybenzoic acid to afford 1-(3,4-dimethoxybenzamido)-2,3,4,9-tetrahydro-1H-carbazole-7-carboxylic acid. LCMS: m/z (M+H)⁺ = 395; ¹H NMR (400 MHz, cdcl3) δ 9.29 (s, 1H), 7.78 (d, J = 7.5 Hz, 1H), 7.54 (d, J = 8.1 Hz, 1H), 7.44 (d, J = 2.0 Hz, 1H), 7.29 (d, J = 2.1 Hz, 1H), 7.17 (t, J = 7.8 Hz, 1H), 6.87 (d, J = 8.4 Hz, 1H), 6.35 (d, J = 6.8 Hz, 1H), 5.37 (d, J = 6.3 Hz, 1H), 3.85 (s, 6H), 3.00 (d, J = 17.1 Hz, 2H), 2.38 – 2.30 (m, 2H), 1.96–1.94 (m, 2H).

1-(3,4-Dimethoxybenzamido)-2,3,4,9-tetrahydro-1H-carbazole-6-carboxylic acid (Table 3, Compound 26):

This compound was prepared from Method 2 using 1-amino-2,3,4,9-tetrahydro-1H-carbazole-6-carboxylic acid and 3,4-dimethoxybenzoic acid to afford 1-(3,4-dimethoxybenzamido)-2,3,4,9-tetrahydro-1H-carbazole-6-carboxylic acid. LCMS: m/z (M+H)⁺ = 395; ¹H NMR (400 MHz, cdcl3) δ 8.20 – 8.03 (m, 1H), 7.60 – 7.46 (m, 3H), 7.26 (s, 1H), 6.84 (dd, J = 8.2, 2.7 Hz, 1H), 5.38 (d, J = 6.8 Hz, 1H), 3.91 (s, 6H), 2.08 – 1.81 (m, 4H), 1.86–1.84 (m, 2H).

Highlights.

• Identified two novel small molecule antagonists of GPR17 through high-throughput screening in a β-arrestin assay.

• The compounds inhibited multiple downstream signaling pathways of GPR17 with high selectivity over other closely related receptors.

• Initial SAR exploration by medicinal chemistry indicated relatively flat SAR for both chemotypes.

• Compound 978 and its analog 793 attenuated GPR17 signaling and increased glucagon-like peptide-1 (GLP-1) secretion inhibited by the GPR17 agonist MDL29,951.