Discovery of a functionally selective ghrelin receptor (GHSR1a) ligand for modulating brain dopamine.

SignificanceThe modulation of growth hormone secretagogue receptor-1a (GHSR1a) signaling is a promising strategy for treating brain conditions of metabolism, aging, and addiction. GHSR1a activation results in pleiotropic physiological outcomes through distinct and pharmacologically separable G protein- and β-arrestin (βarr)-dependent signaling pathways. Thus, pathway-selective modulation can enable improved pharmacotherapeutics that can promote therapeutic efficacy while mitigating side effects. Here, we describe the discovery of a brain-penetrant small molecule, N8279 (NCATS-SM8864), that biases GHSR1a conformations toward Gαq activation and reduces aberrant dopaminergic behavior in mice. N8279 represents a promising chemical scaffold to advance the development of better treatments for GHSR1a-related brain disorders involving the pathological dysregulation of dopamine.

Ghrelin is a peptide hormone secreted from gastric cells during energy deprivation to mediate food-seeking behavior and restore physiological homeostasis (1, 2). Ghrelin exerts its effects via activation of the growth hormone secretagogue receptor-1a (GHSR1a), a G protein–coupled receptor (GPCR) (3). In the brain, the GHSR1a is expressed most highly in agouti-related protein (AgRP)/neuropeptide Y (NPY) neurons of the hypothalamic arcuate nucleus and regulates feeding, energy balance, and metabolism (4). In extrahypothalamic regions, GHSR1a is expressed predominantly in the hippocampus, where it regulates learning and cognition (5), and in dopaminergic midbrain neurons, including the mesolimbic dopamine (DA) neurons of the ventral tegmental area (VTA) and nigrostriatal DA neurons of the substantia nigra pars compacta (SNc) (6, 7). Within DAergic cells, GHSR1a acts as a DA neuromodulator through its effect(s) on neuronal firing rate and DA release probability, biochemical processes that influence locomotion, reward-seeking behavior, and cellular health/neuroprotection (6, 8, 9).

Disruptions to central nervous system (CNS) DA homeostasis can lead to psychiatric, neurological, and neurodegenerative conditions (10). While DA-directed pharmacotherapies (e.g., levodopa, DA receptor agonists/antagonists) may benefit patients initially, these approaches often produce unacceptable side effects when administered chronically (11, 12). These unwanted consequences are due, in part, to the challenges of developing selective DA receptor modulators, as well as the inability of these therapeutic strategies to fully restore DA signaling to healthy, homeostatic levels. Accordingly, endogenous DA neuromodulators, such as the ghrelin-GHSR1a system, may provide a more fine-tuned and safer pharmacological means to normalize the dysfunctional DA signaling underlying brain disorders of mood, cognition, or movement, including addiction, Alzheimer’s disease (AD), and/or Parkinson's disease (PD) (13, 14).

In peripheral tissue, GHSR1a is essential for a diverse array of physiological processes, including glucose-insulin homeostasis, gastrointestinal motility, cardiovascular health, inflammation, and tissue growth and repair (15). In contrast, GHSR1a activity in the CNS appears less well-defined and may be species- and disorder-dependent (16, 17). For instance, GHSR1a(s) expressed in blood–brain barrier (BBB)–protected regions, including the VTA and SNc, may be poorly accessible to the GHSR1a-active form of circulating ghrelin (acyl-ghrelin) and many synthetic GHSR1a agonists when administered peripherally (18). Conversely, the hypothalamus and brainstem are surrounded by fenestrated capillaries that enable circulating acyl-ghrelin to reach GHSR1a(s) expressed in these regions (17). Thus, GHSR1a-targeted neurotherapeutics for DA-based brain disorders must be 1) BBB penetrant in order to modulate hippocampal, mesolimbic, and/or nigrostriatal GHSR1a functions as well as 2) efficacious and bearing a selective therapeutic profile that minimizes on- and off-target side effects.

GHSR1a signals principally through Gαq/11, but it is also capable of engaging heterotrimeric G proteins within the Gαi/o and Gα12/13 families (19). In addition, GHSR1a elicits β-arrestin (βarr)–dependent cellular responses in a temporally and spatially distinct manner from G protein–mediated signaling, including GHSR1a desensitization, internalization, endocytic trafficking and recycling, and the activation of kinases and transcription factors (20, 21). Thus, GHSR1a-directed cellular responses vary according to the extent by which each pathway is selectively activated (ligand selectivity) and modulated downstream (pathway selectivity) (22–24). Most notably, point mutations to adjacent amino acids of the GHSR1a intracellular loop 2 (ICL2), located near the highly conserved E/DRY motif, induce functionally selective G protein or βarr signaling (20, 21). Thus, it may be possible to pharmacologically stabilize GHSR1a into a conformation that selectively activates signaling through one or more of these pathways.

In this study, we present the synthesis and characterization of a Gαq-biased GHSR1a agonist, N8279 (NCATS-SM8864), that contains a 2-carboxamide-3-benzoyl-4-chromenone backbone. In vitro and in silico analyses reveal that N8279 is functionally selective at GHSR1a, and its activity is mediated, at least in part, by extracellular loop 2 (ECL2) and related determinants in the receptor extracellular domain (ECD). In mice, N8279 readily penetrates the BBB and reaches pharmacologically active levels in brain for extended, druggable periods of time. In vivo efficacy studies show that N8279 attenuates hyperlocomotion in DA transporter (DAT) knockout (KO) and cocaine-sensitized C57BL/6J mice—both mouse models of hyperDAergia—but it does not affect novelty-related locomotor activity in inbred C57BL/6J mice under normal physiological conditions. Collectively, our findings support that functional selectivity is a promising strategy when designing GHSR1a treatments that target pathophysiological changes in CNS DA homeostasis.

Results

Discovery of a GHSR1a-Selective Small Molecule, N8279, by High-Throughput GHSR1a Screening and Structural Characterization.

To discover biased GHSR1a ligands, a cell-based, human GHSR1a/βarr1 chemiluminescent assay (DiscoverX, PathHunter) was used to screen ∼47,000 compounds from the Sytravon library and the National Center for Advancing Translational Sciences (NCATS) pharmacological collection (NPC) (). We identified 36 hits (0.09% hit rate) with activities greater than 50% of the activity shown by the full length, human acyl-ghrelin peptide (1–28, amino acids) (). Structure-cluster analysis of the hits revealed six chemical scaffolds from the 36 compounds. The hit compounds were reassessed in secondary assays for Gαq/11-dependent, intracellular Ca2+ mobilization (iCa2+) and βarr2GFP (green fluorescent protein) translocation (). From these experiments, NCGC141956 (N1956) () was selected for further characterization based on its submicromolar potency and full efficacy relative to the unbiased, small molecule GHSR1a agonist, L692,585 (L585). A further directed library screen of commercial N1956 analogs identified NCGC00136164 (N6164), which unexpectedly, was determined to be a Gαq/11-biased GHSR1a agonist relative to βarr2 translocation (). However, the activity of N1956 and N6164 could not be confirmed upon resynthesizing these molecules. A liquid chromatography-mass spectrometer (LC-MS) examination of the dimethyl sulfoxide (DMSO) aliquots used in the screening campaign disclosed impurities within N1956 and N6164 solutions, corresponding to oxidated derivatives of the 1-phenyl-chromeno-pyrrole-dione scaffold. Further characterization, investigation using NMR and MS methods (), and resynthesis of pure oxidated products resulted in the determination of the active molecule, N8279 (NCATS-SM8864) (Fig. 1 and ), which contains a 2-carboxamide-3-benzoyl-4-chromenone backbone. In solution, N8279 equilibrates between open and closed conformers, which in specific solvents and conditions can be observed by 1H NMR (). The structure of the active, open form of N8279 was confirmed by single crystal X-ray diffraction (Fig. 1 and ).

Fig. 1.

N8279 is a potent agonist of GHSR1a-mediated Gαq signaling. N8279 (A) 2D structure and (B) structure determined by single crystal X-ray diffraction. (C) N8279 (1 μM) selectivity for human GHSR1a plotted versus onefold (blue line) and ≥3-fold (dotted purple line) activity above baseline. (D) Linear regression analysis of βarr-based Tango assay results for GPCRome with each point XY corresponding to a distinct receptor and its coordinates defined by X = average of replicates 1 and 2 and Y = average of replicates 3 and 4. (E) [125I]ghrelin competition binding in hGHSR1aWT-expressing HEK293/T cells (unlabeled ghrelin curve, black; N8279 curve, red). Data were normalized to vehicle conditions within each experiment and pooled data normalized to the unlabeled (cold) ghrelin Top (100%) and Bottom (0%). (F) iCa2+ in hGHSR1aWT and miAeq-expressing HEK293/N cells after treatment with ghrelin (black), MK-0677 (green), L585 (blue), or N8279 (red). Bottom and Top parameters were constrained to 0% and 100% of ghrelin (% reference); ghrelin and N8279 h > 1. (G) Ghrelin-induced iCa2+ with concomitant N8279 treatment. Data are normalized to the vehicle Emax (100%) and the image displays best-fit three- or four-parameter regressions for each condition. (H) Gαq dissociation (TRUPATH) in hGHSR1aWT-expressing HEK293/T cells. Bottom and Top parameters were constrained to 0% and 100% of ghrelin (% reference) and the h shared (P > 0.05). (I–K) Ghrelin, N8279, and L585 heat map (I); yellow—higher potency, blue—lower potency or inactivity; (pEC50) potencies (J) and max efficacies (10 μM) (K) at different G proteins derived from curves in . Statistical differences are derived from Dunnett’s multiple comparisons relative to each ligand’s Gαq response. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. All data represent the mean ± SEM from multiple independent experiments.

To determine target selectivity, N8279 activity was evaluated across “the 320 receptor, human GPCRome” by high-throughput screening with a βarr2 recruitment assay (Tango) (25). Hits were defined as ≥3-fold activation above baseline. N8279 stimulated ∼6-fold activation at GHSR1a and did not exceed ≥3-fold activation at any other GPCR (Fig. 1). The assay quality can be assessed by plotting duplicate, independent trial averages (derived from four independent wells) for each receptor as a point (X, Y). The corresponding plot of points for an ideal assay would be fit by a regression line with the slope = 1. A plot of the assay points produced a regression line with slope of 0.98 (Fig. 1) and is shown with its accompanying 99% prediction band that contains GHSR1a as the only hit. Next, the relative affinity of N8279 for the hGHSR1aWT (GHSR1a) was determined by radioligand binding with [125I]ghrelin. Initial saturation studies confirmed that [125I]ghrelin bound GHSR1a asymptotically with nanomolar affinity (). Subsequent competition binding using [125I]ghrelin at its ~Kd demonstrate that both unlabeled ghrelin and N8279 displace [125I]ghrelin from GHSR1a with high and relatively low affinity, respectively (IC50 [nM] = 2.5 and 1,300) (Fig. 1).

N8279 Is a Potent Agonist of GHSR1a-Mediated Gαq Signaling.

GHSR1a primarily couples to Gαq/11, leading to phospholipase C-β-dependent inositol trisphosphate generation and iCa2+ (20). Initial structure-activity relationship screening suggested that the N8279 precursors N1965 and N6164 may exhibit Gαq/11 bias (). To confirm this with the active congener, N8279 (), we performed concentration–response (C/R) analyses in cells stably expressing GHSR1a and an iCa2+ reporter (21). The results show that N8279 is nearly an order of magnitude (8.9-fold) more potent than the endogenous ligand ghrelin and is a full agonist (Fig. 1). N8279 was only 3.4- and 5.3-fold less potent than the high-affinity, unbiased small molecule agonists L585 and MK-0677, respectively (Fig. 1). Consistent with prior work (26), ghrelin was relatively weak at stimulating iCa2+ compared to its GHSR1a binding affinity (Fig. 1 ; Ref. 26). Conversely, the iCa2+ half-maximal effective concentration (EC50) of N8279 was 41-fold more potent than its GHSR1a binding IC50 (Fig. 1 ), suggesting possible allosteric activity (26). Only ghrelin and N8279 each had a Hill slope (h) > 1, suggesting that two or more molecules or GHSR1a binding sites are required for these ligands to elicit full efficacy in this cell system (Fig. 1).

The iCa2+ evoked by EC80 N8279 (), ghrelin (), or L585 () was competitively inhibited by the GHSR1a antagonists YIL781 and JMV2959, supporting GHSR1a-dependent effects. For each agonist, YIL781 was the more potent inhibitor (). To determine whether N8279 elicits GHSR1a-mediated iCa2+ through Gαq/11 specifically, we tested iCa2+ in Gαq/11 KO and wild-type (WT) cells (27) and confirmed that 10 μM ghrelin-, L585-, and N8279-induced iCa2+ was abolished ().

We next evaluated the effect of N8279 on ghrelin-induced iCa2+ signaling to test for ago-allosteric activity. Parenthetically, ago-allosteric agonists interact with topographically distinct receptor sites (allosteric) from the endogenous ligand (orthosteric), elicit agonist behavior on their own, and cooperatively act as a positive (PAM), negative (NAM), or silent (SAM) modulator of orthosteric ligand affinity, potency, and/or efficacy (28). As expected, N8279 displayed intrinsic GHSR1a agonism on its own (Fig. 1, left dashes). At 10 μM, N8279 produced an ∼3-fold increase in ghrelin’s potency but did not reach statistical significance (Fig. 1). N8279 additively increased ∼EC20 ghrelin (100 nM) efficacy in a concentration-dependent manner (Fig. G, right dashes/upward arrow) and marginally increased the ghrelin Emax (Fig. 1, upward arrow), supporting weak ago-PAM activity (29). To assess reciprocal cooperativity, we tested iCa2+ upon concomitant treatment of EC25 MK-0677 or EC50 ghrelin and an N8279 C/R. In the presence of these orthosteric agonists, N8279 was equipotent relative to N8279 alone; however, the h of N8279 alone (h > 1) was reduced to unity (h = 1) (). These findings indicate that N8279 evokes a complete and potent signaling response despite simultaneous occupancy of the GHSR1a orthosteric binding pocket.

To model how N8279 could co-occupy the monomeric GHSR1a with ghrelin, we employed molecular docking with an NMR-based homology model of the ghrelin-bound GHSR1a (30). Concomitant N8279 docking to ghrelin (1–17, amino acids)–bound GHSR1a suggests that the propylamine moiety of N8279 could form a strong ionic bond with a negatively charged ECL2 (Asp191) (). In this pose, N8279 was found to bind GHSR1a atop ghrelin, enabling ghrelin’s N terminus to insert into the deep orthosteric binding pocket and interact with Glu124 in transmembrane domain (TM)-III, consistent with prior models and the proposed agonist-induced activation mechanism of GHSR1a (30, 31). Thus, N8279 may exhibit state-dependent allosteric GHSR1a binding by anchoring to ECL2.

N8279 Biases GHSR1a Toward Gαq Coupling over Other Gα Subunits.

To evaluate the effect of N8279 on Gαq proximal to the GHSR1a, we employed parallel NanoBiT- (32) and bioluminescence resonance energy transfer (BRET, TRUPATH; Ref. 33)–based heterotrimeric G protein subunit dissociation approaches. In both assays, N8279 was a full agonist for Gαq activation with a potency comparable to that of iCa2+ (Fig. 1 and ). In NanoBiT-Gαq assays, N8279 and ghrelin potencies were statistically equivalent (). In BRET-Gαq assays, N8279 was 6.1- and 1.7-fold less potent than ghrelin and L585, respectively (Fig. 1). In contrast to iCa2+ assays (Fig. 1), the h was <1 for each agonist. These differences may be due to distinct G protein subunit composition/ratios and/or response sensitivities between these assays (see ). Notably, ghrelin was ∼30 to 50-fold more potent at activating Gαq proximally than eliciting downstream iCa2+ (Fig. 1 and ), consistent with a prior report (26). These findings suggest that in these cell systems, low ghrelin concentrations are sufficient to dissociate the heterotrimeric Gαq complex, whereas high ghrelin concentrations are required to fully engage downstream GHSR1a signaling, as seen with other homodimeric GPCRs (34). Alternatively, ghrelin could elicit Gαq-independent signaling that counter regulates iCa2+ in this assay. Collectively, these results demonstrate that N8279 is a potent agonist of Gαq signaling at GHSR1a.

In an independent set of experiments, we evaluated N8279 signaling through other G proteins to compare to Gαq. We selectively tested Gα subunits that are expressed highly in midbrain DAergic neurons (35) and/or reported previously to exhibit GHSR1a coupling, including GαsS (Gαs), Gαi1, Gαi2, GαoA, Gα12, and Gα13 (19, 23, 36). Relative to Gαq, ghrelin potency was statistically equivalent for each Gαi/o and was reduced moderately for Gα12 and Gα13 (Fig. 1 and ). Ghrelin did not activate Gαs, consistent with a prior study (19). L585 displayed a similar profile, except that it had reduced potency at Gαi1 and only a statistical trend for reduced potency at Gα12. In contrast, N8279 potency was significantly reduced to concentrations >1 μM for every Gα (Fig. 1 and ), suggesting bias toward Gαq coupling. Although N8279 and L585 showed statistically equipotent activation of Gαs, their maximal efficacies (10 μM) at Gαs were markedly reduced (Fig. 1 and ). Moreover, N8279 maximum efficacy was significantly reduced at Gα12/13 compared to ghrelin and L585, as well as at Gαi2 and GαoA compared to L585 (Fig. 1 and ). For each ligand, maximal efficacy was reduced at every Gα relative to their respective effect on Gαq (Fig. 1 and ).

Together, the data in Fig. 1 support that N8279 is a potent GHSR1a agonist with functional selectivity toward Gαq. All pharmacological results (IC50, Ki, logEC50, Emax, h ± SEM) and statistical comparisons for Fig. 1 are shown in .

N8279 Recruits βarr2 to GHSR1a More Weakly than Ghrelin.

Having established that N8279 is a potent activator of Gαq signaling, we next assessed its effect on GHSR1a-mediated βarr recruitment first using a NanoBiT-based approach. Cells expressing a fixed ratio of GHSR1aLgBiT and SmBiTβarr2 were treated with ghrelin, L585, or N8279. Here, N8279 was ∼20-fold less potent than ghrelin, and it approached full agonism (Fig. 2). Conversely, L585 recruited βarr2 with moderately higher potency and equivalent efficacy to ghrelin (Fig. 2), and comparatively, N8279 was ∼43-fold less potent than L585 in this assay. Thus, N8279 is a weaker agonist of GHSR1a-βarr2 recruitment than ghrelin and L585, supporting that it exhibits functional selectivity toward Gαq over βarr coupling (Fig. 1).

Fig. 2.

N8279 is a weak activator of GHSR1a-mediated, βarr2-dependent cellular responses relative to ghrelin. (A) Peak SmBiTβarr2 recruitment (average, 0 to 5 min) to hGHSR1aLgBiT in HEK293/T cells. Data were baseline normalized within each experiment, then to the ghrelin Emax (% reference). (B) hGHSR1aLgBiT-SmBiTβarr2 saturation after treatment with ghrelin (100 nM, black) or N8279 (100 nM, light red; 200 nM, dark red). Hyperbola were fit by one-site regression to derive a Bmax (BiTmax) and Kd (BiTd), then normalized to the ghrelin BiTmax (100%). (B, Inset) Ghrelin and N8279 BiTd values derived from B and analyzed by one-way analysis of variance (ANOVA) followed by Sidak's multiple comparisons. (C) [125I]ghrelin competition binding in HEK293/T cells expressing hGHSR1aL149G. Data were normalized as in Fig. 1. (D) Maximum Venusβarr2 recruitment (over 60 min) to hGHSR1aWT or hGHSR1aL149G in HEK293/T cells. Data were baseline normalized within each experiment, then the ghrelin-WT Emax (% reference). (E) EC80 ghrelin (40 nM)–induced SmBiTβarr2 recruitment to hGHSR1aLgBiT after pretreatment (5 min) with YIL781, JMV2959, or N8279. The 100% point represents EC80 ghrelin alone and the 0% line represents baseline. (F) Representative images of vehicle-, ghrelin (100 nM)–, or N8279 (100 nM)–induced βarr2 translocation (45 min, 37 °C) in U2OS cells expressing hGHSR1aWT and βarr2GFP. (G) hGHSR1aWT internalization in HEK293/T cells (45 min, 37 °C). Data are expressed as the percentage of GHSR1a expression relative to baseline (100%), and pooled data were normalized to the ghrelin Top (100%) and Bottom (0%) (% reference). (H) bBRET-based hGHSR1aWT-RLucII internalization in HEK293/T cells with MyrPalmVenus. Data represent the average net BRET (60 min) normalized to baseline within each experiment and then to the ghrelin Top (100%) and Bottom (0%) (% reference). (I) bBRET-based hGHSR1aWT-RLucII trafficking in HEK293/T cells with 2×FYVEVenus. Data represent the average net BRET (60 min) normalized to the ghrelin Emax (% reference). (J) GHSR1a trafficking Emax over 120 min as derived from I. (K) SRF-RE–mediated transcription in HEK293/T cells expressing hGHSR1aWT. Data were normalized to the ghrelin Emax (% reference). (L) N8279 bias factor (RAi model) with β (log10) quantified using ghrelin (black/gray) or L585 (blue) as reference ligands. All data represent the mean ± SEM from multiple independent experiments. ***P < 0.001; **P < 0.01; *P < 0.05; n.s. (nonsignificant), P > 0.05.

To further test this hypothesis, cells expressing a variable ratio of GHSR1aLgBiT and SmBiTβarr2 were treated with ghrelin (100 nM) or ∼EC80 N8279 for Gαq signaling (100 to 200 nM; Fig. 1 and ). These analyses revealed that the relative affinity (BiTd) of βarr2 for GHSR1a was 2- to 2.5-fold weaker in N8279-treated cells than in ghrelin-treated cells (Fig. 2, Inset). Furthermore, in competitive binding studies with the ICL2 mutant GHSR1aL149G (Fig. 2), a βarr2-biased receptor (20), the ability of N8279 to displace [125I]ghrelin was diminished and shifted rightward by ~10-fold (IC50 > 10 μM) relative to GHSR1aWT (Fig. 1). In contrast, the IC50 of unlabeled ghrelin for GHSR1aL149G was reduced by only ∼2-fold compared to GHSR1aWT (Figs. 1 and 2). A follow-up, BRET-based GHSR1aL149G-RLucII-Venusβarr2 recruitment assay further supported distinct properties between ghrelin and N8279 at the βarr-biased GHSR1aL149G. While GHSR1aL149G reduced the Emax of ghrelin and N8279 to similar extents (∼35%), the N8279 potency was reduced by ∼5-fold, whereas the ghrelin potency was reduced by only ∼2-fold relative to GHSR1aWT (Fig. 2). Notably, interassay comparisons revealed that N8279 was ∼100-fold less potent than ghrelin in BRET-based GHSR1aWT-βarr2 recruitment assays (Fig. 2) but only ∼20-fold less potent in NanoBiT-based measurements (Fig. 2). This distinction may be due to variations in GHSR1a-βarr2 expression ratios, biosensor interaction kinetics, and/or measurement time, e.g., GHSR1a-βarr2 expression ratios (BRET = 1:15 versus NanoBiT = 1:1) and measurement durations (BRET = 60 min versus NanoBiT = 5 min). Nonetheless, these findings together suggest that N8279 stabilizes GHSR1a conformations that disfavor GHSR1a-βarr2 coupling relative to ghrelin and, reciprocally, that GHSR1a conformations preferentially supporting βarr2 coupling (GHSR1aL149G) diminish N8279-GHSR1a interaction(s).

Next, we assessed whether N8279—as an agonist functionally selective for Gαq—could behave as a βarr2 antagonist in the presence of ghrelin. We pretreated cells expressing GHSR1aLgBiT and SmBiTβarr2 with increasing concentrations of N8279 or the antagonists YIL781 or JMV2959, followed by EC80 ghrelin. N8279 inhibited ghrelin-induced βarr2 recruitment significantly, but incompletely, in a concentration-dependent manner and was 1.7- and 2.9-fold less potent than JMV2959 and YIL781 (Fig. 2). These data suggest that N8279 stabilizes GHSR1a conformation(s) that weaken βarr2 coupling in both apo- (Fig. 2 ) and ghrelin-bound (Fig. 2) receptor states.

N8279 Reduces βarr-Dependent Cellular Responses Relative to Ghrelin.

Qualitative microscopy of U2OS cells expressing GHSR1a and βarr2GFP showed minimal response to 100 nM N8279 and displayed diffusely distributed cytosolic βarr2GFP similar to vehicle-treated cells (Fig. 2). Conversely, 100 nM ghrelin–treated cells exhibited marked accumulation of cytosolic puncta, indicative of robust βarr2GFP translocation and GHSR1a endocytosis/trafficking (Fig. 1). Though ghrelin produced a robust response, the relatively weak response produced by N8279 in these experiments may reflect differences in engagement with early (e.g., GPCR kinases; Ref. 37) or late molecular mediators of receptor endocytosis and/or endosomal trafficking (21). Thus, we evaluated ligand-induced GHSR1a endocytosis using three independent methods. First, in a quantitative, cell surface enzyme-linked immunosorbent assay (ELISA) approach, both ghrelin and N8279 stimulated GHSR1a internalization in a concentration-dependent manner (Fig. 2). However, N8279 internalization potency was reduced by 24-fold, and the efficacy was reduced modestly (∼30%) relative to ghrelin (Fig. 2). Second, we employed a bystander BRET (bBRET)–based plasma membrane sensor, MyrPalmVenus (38), and found that N8279-induced GHSR1a internalization (over 60 min) potency was reduced by 32-fold and efficacy was attenuated by ∼20% relative to ghrelin (Fig. 2). Last, the bBRET-based sensor for early endosomes (21, 39), 2×FYVEVenus (40), showed that N8279-induced GHSR1a endosomal transit was 32-fold less potent and less efficacious (∼35%) than ghrelin over 60 min posttreatment (Fig. 2). Temporal analyses revealed that N8279-stimulated GHSR1a endosomal trafficking occurred on a time course similar to ghrelin, albeit with reduced efficacy across the entire 120-min measurement (Fig. 2).

βarr2 is required for GHSR1a-mediated RhoA GTPase/ROCK signaling (41), leading to transcriptional activation and cytoskeletal rearrangement by induction of actin polymerization (20). To test whether N8279 affects these processes, we utilized the RhoA-dependent transcriptional reporter serum response factor response element (SRF-RE) (20). Here, N8279 was a full agonist with mildly increased maximal efficacy relative to ghrelin (Fig. 2). However, N8279 potency was reduced ∼10- to 15-fold relative to L585 and ghrelin, respectively. The enhanced efficacy of N8279 in these assays compared to Fig. 2 could be a time-dependent effect, in part, due to the 6-h treatment duration (see ) and/or the partial contribution(s) of Gα12/13 or especially MAPK/ERK signaling to SRF transcription (20).

To quantitatively assess N8279 bias between the Gαq and βarr2 (Fig. 2) pathways, we used the intrinsic relative activities (RAi) model as described previously (42). Relative to ghrelin, N8279 had a proximal (Gαq dissociation, Fig. 1) bias factor (β) of 0.59 (∼4-fold) and a downstream (iCa2+, Fig. 1) β of 2.63 (∼427-fold) relative to βarr2 recruitment (NanoBiT, Fig. 2) (Fig. 2 , Left). Relative to L585, N8279 had a proximal β of 1.45 (∼28-fold) and a downstream β of 1.15 (∼14-fold) (Fig. 2 , Right). Notably, assay-standardized bias calculations (i.e., Gαq-βarr2 BRET versus Gαq-βarr2 NanoBiT) using ghrelin as a reference ligand showed that N8279 had a proximal β of 1.16 (∼14-fold) and 1.36 (∼23-fold) when employing paired BRET-Gαq/BRET-βarr2 or NanoBiT-Gαq/NanoBiT-βarr2 assays, respectively (). Ultimately, these analyses together support that N8279 is a G protein–biased agonist of both proximal and downstream GHSR1a-Gαq signaling.

Collectively, the data in Fig. 2 support that N8279 is a weak agonist of GHSR1a-mediated, βarr-dependent signaling relative to ghrelin, and thus, N8279 is a G protein–biased GHSR1a agonist. All pharmacological results and statistical comparisons for Fig. 1 are shown in .

GHSR1a Mutagenesis and Molecular Docking Suggest an ECD-Dependent, Extended Binding Mode of N8279.

We next evaluated whether determinants outside the orthosteric binding pocket are required for N8279 signaling by first using a naturally occurring variant, GHSR1aA204E (Fig. 3). Substitution of glutamic acid at this ECL2 site abolishes constitutive activity and causes short stature in humans (43). However, the GHSR1aA204E mutation does not appreciably affect ghrelin binding (affinity) or ghrelin-induced Gαq signaling (potency) (43, 44), supporting that it lies outside the orthosteric binding pocket.

Fig. 3.

N8279 requires receptor sites and/or conformational states driven by the GHSR1a ECD that are distinct from ghrelin. (A) Amino acid snake plot of the hGHSR1aWT highlighting Ala204ECL2 (red) and the Ala204Glu mutation. Ghrelin-induced (B) iCa2+, (C) TRUPATH Gq dissociation, and (D) NanoBiT βarr2 recruitment at hGHSR1aWT (black) or hGHSR1aA204E (purple). N8279-induced (E) iCa2+, (F) TRUPATH Gαq dissociation, and (G) NanoBiT βarr2 recruitment at hGHSR1aWT (red) or hGHSR1aA204E (yellow). All data are normalized to the GHSR1aWT Emax. Gαq dissociation and βarr2 recruitment assays are also baseline normalized. (H) Superimposition of the ghrelin-bound model structure (blue) with the antagonist-bound X-ray crystal structure (6KO5, green). (I) Proposed GHSR1aDTP (red) and GHSR1aECD (blue) binding pockets in the ghrelin-bound model. Gln120, Glu124, Phe279, Arg283, and Phe309 constitute the canonical orthosteric GHSR1aDTP pocket; mutations to these residues cause significant loss of ghrelin and orthosteric agonist efficacy (46). Asp99, Cys198, and Asn305 constitute predicted interaction sites within the GHSR1aECD binding pocket of the ghrelin-bound model. (J) N8279 docking pose (green) in GHSR1aDTP. (K) N8279 docking pose (green) in GHSR1a ECD. Dash lines indicate hydrogen bonds (yellow), ionic interactions (pink) and π–π stacking interactions (turquoise), or halogen bonds (purple). (L) Snake plot of hGHSR1aWT with experimentally mutated residues: D99A (blue), E197A (green), R199A (orange), P200A (purple), A204E (yellow), N305A (teal), and Glu124 and Cys198 (gray, mutations not made). (M) Ghrelin and N8279 Gαq dissociation (TRUPATH) and (N) iCa2+ pEC50 and Emax at GHSR1a mutants shown in L, derived from . All data represent the mean ± SEM from multiple independent experiments.

Consistent with prior studies (43, 44), GHSR1aA204E showed no basal iCa2+ activity (Fig. 3 ); surface expression was reduced by ∼50% (); it had minimal-to-no effect on ghrelin-stimulated Gαq dissociation and iCa2+ (Fig. 3 ); and ghrelin-stimulated βarr2 recruitment efficacy, but not potency, was reduced (Fig. 3; Ref. 44). In contrast, N8279-induced iCa2+ potency was reduced by 6.5-fold, while full agonism was retained (Fig. 3). Furthermore, N8279-induced Gαq dissociation signaling was reduced dramatically such that the C/R curve did not saturate, supporting that the N8279 potency is blunted by >20-fold and the maximal efficacy decreased by ∼45% (Fig. 3). The effect magnitude discrepancy between measurements of iCa2+ and Gα dissociation likely reflects signal amplification differences between the assays. N8279-induced βarr2 recruitment potency at GHSR1aA204E was similarly diminished (>20-fold) and did not reach saturation, with a maximal efficacy comparable to ghrelin (∼35%; Fig. 3). Thus, relative to ghrelin, N8279 signaling requires distinct ECL2 sites and/or ECD-dependent conformational states.

Next, we used the NMR-based homology model of the ghrelin-bound GHSR1a () (30) to simulate N8279-GHSR1a binding. We deprioritized the antagonist-bound GHSR1a crystal structure (45) (Fig. 3 and ) because it better models the inactive GHSR1a conformation (45). Docking N8279 with ghrelin removed discloses two potential binding modes for N8279 within the apo-GHSR1a (Fig. 3). Both modes display strong ionic interactions between N8279’s propylamine moiety and specific acidic (negatively charged) GHSR1a residues. In one mode (Fig. 3, red), N8279's terminal tertiary amine group forms a salt bridge with the conserved TMIII residue Glu124, located within the deep transmembrane pocket (GHSR1aDTP). In the second mode (Fig. 3, blue), N8279's propylamine moiety forms a salt bridge with Asp99 toward the top of TMII, enabling an extended binding mode into the ECD (GHSR1aECD) or extracellular vestibule, including the extracellular end of TMVII and ECL2. Notably, the superficial residue Asp99 is too distant from Glu124 (12.7 Å) for N8279 to interact with both sites simultaneously (Fig. 3). Thus, both docking models suggest that N8279 binds GHSR1a via ionic interactions with spatially distinct anchor residues. In GHSR1aECD, N8279’s methoxy-aromatic and amide moieties form hydrogen bonds with or adjacent to potential allosteric sites based on prior mutagenesis by others (44, 46), including Asn305 on TMVII and Glu197, Arg199, or Pro200 in ECL2 (Fig. 3). Here, N8279’s amide group forms an H+ bond with Cys198, a highly conserved GPCR residue that constrains ECL2 flexibility (Fig. 3) (47). Notably, N8279 had comparable docking scores within both potential binding pockets: GHSR1aECD (−6.732) and GHSR1aDTP (−6.767) (Fig. 3 ).

To test our model, we made point mutations to predicted N8279-GHSR1a interaction sites or residue clusters (Fig. 3). Given the evidence for critical ECL2-dependent contributions to N8279 signaling (Fig. 3 ), as well as the absolute requirement of the GHSR1aDTP anchor residue, Glu124, for GHSR1a function/activation (30, 45, 46), we prioritized the GHSR1aECD-N8279 docking model for mutagenesis. Alanine substitution to the putative anchor residue Asp99 (Fig. 3, blue) abolished N8279 signaling (Fig. 3 and ). However, the surface expression of this was reduced markedly (), as shown previously (45). Mutations were not made to Cys198ECL2 (Fig. 3, light blue) because it precludes GPCR stability and ligand binding (48). Instead, we made alanine substitutions to three adjacent, putative allosteric (44, 46) and/or structurally integral (30) ECL2 residues: Glu197, Arg199, and Pro200 (Fig. 3, blue). We also made an alanine substitution to Asn305, which is located at the extracellular end of TMVII (Fig. 3, blue) and thus is considered within the ECD.

Relative to the WT receptor, the surface expression of the GHSR1aE197A was reduced moderately, the GHSR1aR199A was comparable, and the GHSR1aP200A was increased mildly (). N8279 potency at the GHSR1aE197A was diminished by ∼3- to 10-fold in iCa2+ and Gαq dissociation, respectively, whereas the N8279 Emax was reduced in iCa2+, but not Gαq dissociation assays (Fig. 3 and ). The GHSR1aR199A mutation did not affect N8279 potency in either assay, but its Emax was reduced mildly in iCa2+ assays (Fig. 3 and ). N8279 potency and Emax were reduced dramatically at the GHSR1aP200A in both Gαq dissociation and iCa2+ assays (Fig. 3 and ). Grouped analysis of the N8279 potency and Emax at GHSR1aA204E relative to other mutants supported markedly decreased N8279-induced Gαq signaling at this ECL2 residue (Fig. 3 ), despite its location being outside of the putative GHSR1aECD binding pocket (Fig. 3 ). Alanine substitution to Asn305 did not affect surface expression relative to the WT receptor (), although N8279-Gαq dissociation was reduced dramatically and N8279-iCa2+ potency was reduced moderately (6.5-fold). Surprisingly, the N8279-iCa2+ Emax was elevated at GHSR1aN305A (Fig. 3 and ). Generally, any observed differences between N8279-induced Gαq dissociation or iCa2+ at these mutants could be due to distinctions in assay kinetics or signal amplification, GPCR-transducer expression ratios, and/or the involvement of other transducers (e.g., Gα11, Gαi/o, Gβγ) in the iCa2+ response (49, 50). Together, these mutagenesis findings support that N8279 signaling requires GHSR1a sites within and/or conformational states determined by the ECD, especially in ECL2.

For comparison, we evaluated ghrelin-stimulated Gαq dissociation and iCa2+ at each GHSR1aECD mutant. Ghrelin signaling was reduced markedly at GHSR1aD99A in both assays (Fig. 3 and ), supporting previous findings (45). However, ghrelin potency at the GHSR1aD99A was similar to GHSR1aWT in Gαq dissociation assays (Fig. 3 and ). Mutations to Glu197, Arg199, or Pro200 did not significantly affect ghrelin-stimulated Gαq dissociation or iCa2+ potency (Fig. 3 and ), consistent with prior findings (44, 46). Ghrelin’s Emax was unaffected at GHSR1aR199A and GHSR1aP200A, but it was reduced moderately at GHSR1aE197A in iCa2+ assays (Fig. 3 and ), consistent with its reduced expression (). Furthermore, grouped analyses (derived from Fig. 3 ) supported that ghrelin potency is unaffected at the GHSR1aA204E mutant and its Emax is decreased only in Gαq dissociation assays (Fig. 3 ), consistent with reduced expression of this mutant (). Ghrelin-Gαq dissociation potency was reduced at GHSR1aN305A, albeit to a much lesser extent than N8279 (Fig. 3 and ). Nonetheless, ghrelin-iCa2+ potency was not significantly affected by the N305A mutation (Fig. 3 and ). The ghrelin Emax at GHSR1aN305A was increased in both assays (Fig. 3 and ), similar to that seen for N8279-induced iCa2+ (Fig. 3 and ). Collectively, these findings demonstrate that Glu197, Pro200, Ala204, and, in part, Asn305 are critical and specific GHSR1aECD sites for N8279 relative to ghrelin.

Collectively, the data in Fig. 3 support that N8279 signaling requires molecular determinants within GHSR1aECD and particularly ECL2. All pharmacological results and statistical comparisons for Fig. 1 are shown in .

N8279 Is Brain-Penetrant and Attenuates DA-Driven Behavior.

Pharmacokinetic studies with intravenous (IV; 1 mg/kg), oral gavage (PO; 5 mg/kg), and intraperitoneal (IP; 5 mg/kg) administration in C57BL/6 mice reveal  a PO bioavailability of 7% and IP bioavailability of 27% (). Significantly, IP administration of N8279 (5 mg/kg) delivered pharmacologically relevant levels (∼200 nM) in brain within 15 min, reaching peak concentrations (Cmax) of 259 nM at 2 h, followed by a slow decline and elimination by 24 h (Fig. 4 and ). In the brain, N8279 has a half-life (t1/2) of 6.6 to 11 h after IP and PO administration, maintaining levels above its Gαq/iCa2+ EC50 (∼35 nM, Fig. 1 ) for an extended period (>7 h) with a brain/plasma ratio for N8279 (IP) in the range of 0.6 to 0.9:1 (Fig. 4 and ). In summary, N8279 achieves rapid, sustained, and pharmacologically relevant concentrations in mouse brain following systemic administration.

Fig. 4.

N8279 is brain-penetrant and attenuates aberrant DAergic behavior in mice. (A) Analysis of brain (red) and plasma [N8279] (orange) over 24 h in C57BL/6 mice treated with N8279 (5 mg/kg, IP). N8279 Cmax at ∼2 h = 123 ng/mL (259 nM). N8279 (IP) brain half-life (t1/2) = 6.6 h, plasma t1/2 = 3.8 h. (B) Spontaneous hyperlocomotion in DAT KO mice. 30 min (gray box) acclimation prior to injection (black arrow) of N8279 (2.5, 5, or 10 mg/kg, IP) or vehicle (5% DMSO, saline). Horizontal locomotion was monitored for 120 min postinjection, and beam breaks were collected in 5 min bins. Results are presented as mean ± SEM. N8279-treated DAT KO mice had reduced locomotion relative to vehicle-treated controls. Postinjection, repeated measures ANOVA (RMANOVA), time: [F(7.6,411.8) = 14.7, P < 0.0001], dose: [F(3,54) = 3.4, P = 0.022], time × dose interaction: [F(69,1242) = 1.2, P = 0.118]. No baseline differences were detected between groups. Baseline RMANOVA, time: [F(2.4,131.0) = 3.0, P = 0.039], dose: [F(3,54) = 0.06, P = 0.979], time × dose interaction: [F(15,270) = 1.4, P = 0.141]. n = 12 to 20 mice per group. (C) One-way ANOVA for total area under the curve (AUC, 35 to 150 min) derived from B. Treatment: [F(3,55) = 6.50, P = 0.0008]. Dunnett’s multiple comparisons revealed an effect of 2.5, 5, and 10 mg/kg N8279 relative to vehicle control (0 mg/kg). *P < 0.05, ***P < 0.001 versus vehicle. (D) Cocaine-induced behavioral sensitization in C57BL6/J mice: experimental design (Left) and locomotion (Right). Postinjection results are presented as percentage of baseline activities because locomotion was low in the N8279 + vehicle group on all days. Within-group analyses (significance denoted by +) relative to day 1 showed that the vehicle + cocaine group had increased locomotion on days 2 to 5 and day 11, whereas the N8279 + cocaine group had increased locomotion only on day 5. Between-group comparisons (significance denoted by *) showed that cocaine-induced sensitization was higher on day 4 and on challenge day 11 in the vehicle + cocaine compared to the N8279 + cocaine group. RMANOVA: day [F(5,145) = 8.797, P < 0.001], treatment [F(2,29) = 32,523, P < 0.001], day × treatment [F(10,145) = 3.215, P < 0.001]. */+P < 0.05, ***/+++P < 0.001. n = 9 to 12 mice/group.

To evaluate the effect of N8279 on DA-modulated behavior in vivo, we first used DAT KO mice, which have constitutively elevated extracellular DA levels and consequently spontaneous hyperactivity in a novel open field (51). After a 30-min acclimation period, male and female DAT KO mice were administered vehicle or pharmacologically relevant, brain-penetrant doses of N8279 (Fig. 4; 2.5, 5, or 10 mg/kg, IP), and returned to the open field with locomotion monitored for an additional 120 min. Each dose of N8279 reduced overall hyperlocomotion in DAT KO mice relative to vehicle controls (Fig. 4 ). Parallel control experiments with inbred male and female C57BL/6J mice indicated that N8279 does not affect novelty-induced open-field locomotion ().

We next assessed cocaine-induced behavioral sensitization in male and female C57BL/6J mice following subchronic (8-d) administration of the vehicle or N8279 (5 mg/kg, IP) in the home cage. Subsequently, mice were given the same treatments, and this was followed with injection (IP) of vehicle or cocaine (20 mg/kg) in the open field once a day for 5 consecutive days (Fig. 4 , Left). A 5-d hiatus (washout) was imposed, and behavioral sensitization was assessed the next day by giving (IP) vehicle or cocaine (challenge). The cumulative results showed that postinjection locomotor activities were low in the N8279 + vehicle group and were significantly reduced from the cocaine-treated mice across all days (Fig. 4). By comparison, motor activities were stimulated acutely (day 1) to similar extents in the vehicle + cocaine and N8279 + cocaine mice (Fig. 4 , Right). Locomotion was increased from day 1 though each day to day 5 in the vehicle + cocaine group, whereas significantly enhanced activity was observed only on day 5 in the N8279 + cocaine mice. Moreover, at challenge (day 11), locomotor activity was augmented relative to day 1 in the vehicle + cocaine mice, whereby no change in activity was evident in the N8279 + cocaine mice. Thus, N8279 both delayed the appearance of sensitization across days and abrogated the expression of sensitization following washout (challenge, day 11).

Collectively, these results indicate that pharmacologically relevant and brain-penetrant levels of N8279 ameliorate aberrant DA-mediated behavior in two mouse models of persistently disrupted DA homeostasis.

Discussion

In this study, we disclose a chemotype of small molecule GHSR1a modulators with functional selectivity toward Gαq signaling. Our results support that the lead compound from this chemical series, N8279 (NCATS-SM8864), stabilizes conformational states that drive the apo- and ghrelin-bound GHSR1a toward Gαq coupling over other G proteins and βarr-dependent cellular responses. Importantly, N8279 has excellent brain penetrance and exhibits salutary effects on DA-induced behavior in both genetic and pharmacological mouse models of disrupted DA homeostasis.

Collectively, our findings indicate that N8279-induced GHSR1a signaling originates, at least in part, from an extended binding mode into the extracellular vestibule and/or conformational constraints imposed by the GHSR1aECD, especially ECL2. This notion is supported most strongly by significantly reduced N8279 potency upon mutation of specific ECL2 residues, including Ala204, Pro200, and to a lesser extent Glu197. Critically, these effects are distinct from the endogenous GHSR1a ligand, ghrelin. Although N8279 is not predicted to interact directly with these residues in our in silico models, they are adjacent or proximal to predicted interaction sites within our GHSR1aECD binding pocket model. Thus, these mutations may result in GHSR1a conformation(s) that are less accessible for N8279 binding and/or less capable of stabilizing a signaling-competent N8279-GHSR1a complex via perturbation of the ECD structure and/or local disruption(s) to proximal interaction sites. Together, these findings suggest that N8279 signaling at the apo-GHSR1a is most likely mediated by a bitopic, extended binding mode that is conformationally dependent upon the GHSR1aECD. Additionally, N8279-GHSR1aECD interaction(s) may be state dependent, given that concomitant docking of N8279 and ghrelin peptide (1–17, amino acids) (30) predicts a strong ionic interaction of N8279 with ECL2 when at the ghrelin-occupied GHSR1a (). Pharmacologically, these binding properties may contribute to N8279’s ability to modestly enhance, or at least noncompetitively retain, Gαq signaling efficacy at the ghrelin-bound GHSR1a (Fig. 1; ). Alternatively, these effects could arise from asymmetric interactions between ghrelin, N8279, and GHSR1a homodimers within allosteric and/or orthosteric binding pockets. This notion is consistent with ghrelin- and N8279-induced iCa2+ exhibiting an h > 1 at the apo-GHSR1a (Fig. 1) and N8279 having an h of unity at the orthosteric agonist-bound GHSR1a (Fig. 1; ). Indeed, GHSR1a homodimerizes (52) and ago-allosteric GHSR1a agonists (e.g., L-692,429) elicit similar effects to N8279 via bitopic or bimodal state-dependent interaction(s) with the “allosteric” or “orthosteric” protomers of the GHSR1a dimer (26).

Our functional evidence supports the notion that N8279 stabilizes a conformation favoring the Gαq-bound conformation of GHSR1a over other transducers (Gαi/o, Gα12/13, and βarr2) relative to ghrelin (Figs. 1 and 2). Conversely, N8279 displacement of ghrelin binding is blunted at the βarr-biased ICL2 variant, GHSR1aL149G, suggesting that the βarr2-coupled GHSR1a renders N8279 binding at the extracellular face of the receptor less accessible (Fig. 2). Thus, N8279 pretreatment-induced partial GHSR1aWT-βarr2 antagonism may depend on kinetic and/or allosteric mechanism(s). Together, these observations of bidirectional allostery are consistent with the principle of reciprocity underlying ligand–GPCR–transducer and allosteric coupling (i.e., ternary complex model) (53, 54). Indeed, ECL2 is an established determinant of GHSR1a (44) and, more generally, GPCR bias (55, 56). Several biased ligands elicit their effects through extended binding modes involving the cognate GPCR’s ECD (especially ECL2), whereby superficial ligand–receptor interactions preferentially stabilize conformations that allosterically propagate to the intracellular receptor face to influence transducer coupling and thereby functional selectivity (57). Some examples of these biased ligand–GPCR pairs include the dopamine D2 receptor (D2R) (55), the serotonin 2B (5-HT2BR) (58, 59), the muscarinic type 2 (M2R) (60, 61), the β-adrenergic receptors (β1AR, β2AR) (62), and the glucagon-like peptide-1 (GLP-1) receptor (63). Notably, GHSR1a-containing tissues likely express different levels of signaling proteins (e.g., Gα subunits, GRKs); thus, “system bias” (64, 65) may be expected (Figs. 1 and 2) to play a significant role in determining N8279-induced GHSR1a functional selectivity in vivo.

Pharmacokinetic studies indicate that N8279 was able to sustain biologically relevant levels in brain for extended periods of time following IP administration of a low dose (5 mg/kg). Critically, in vivo efficacy studies revealed that acute administration of N8279 at pharmacologically relevant doses attenuated hyperlocomotion in both genetic and pharmacological models of hyperDAergia, recapitulating effects achieved by GHSR1a antagonists (22, 66, 67). Previously, we reported that the antagonist YIL781 reduced hyperlocomotion in cocaine-sensitized WT but not DA neuron–specific βarr2 KO mice (22), suggesting that βarr2 inhibition was required for GHSR1a antagonists to blunt cocaine-induced neuroadaptations. Additionally, GHSR1a-induced βarr signaling and, more generally, Gα12/13 signaling (68, 69) are required for RhoA-dependent actin remodeling—a process integral to neuroplasticity in DA neurons (70, 71). Together, these prior findings suggest that βarr2 might be necessary for the proaddictive effects of GHSR1a activation, particularly during DA plasticity–dependent reward learning (e.g., behavioral sensitization). Here, we show that N8279 reduced novelty-related hyperlocomotion in DAT KO but not in inbred C57BL6/J mice (Fig. 3 ) and depressed behavioral sensitization in cocaine-sensitized C57BL/6J mice. Thus, it appears that reductions in GHSR1a-βarr2 signaling may preferentially blunt the development and/or expression of behaviors mediated by sensitized neurocircuits (e.g., mesolimbic, nigrostriatal) that have undergone DA-dependent plasticity, whereas GHSR1a regulation of metabolic homeostasis (e.g., feeding, glucose/insulin homeostasis, growth hormone secretion) and DA cell neuroprotection are generally considered G protein mediated (9, 23, 72, 73). Thus, an agonist functionally selective for Gαq should bias GHSR1a away from βarr or alternative Gα signaling and thus may retain the antiaddictive effects of GHSR1a antagonists while preserving, normalizing, or augmenting Gαq-dependent endocrine homeostatic and/or neuroprotective GHSR1a functions. Moreover, reductions in GHSR1a-mediated βarr signaling (e.g., desensitization, down-regulation) by a G protein–biased agonist should, in principle, minimize tolerance and retain therapeutic efficacy with chronic administration.

To elucidate these hypotheses adequately, future studies will require the development of additional pharmacological tools with distinct signaling properties and, ideally, transgenic animal models that recapitulate both pathway-selective and complete bias in vivo. Fundamentally, if biased signaling can produce desired physiological outcomes with specificity, then biased ligands should provide efficacy with reduced side effects and thereby generate superior drugs. This point has significant theoretical and practical implications for the future of pharmaceutical healthcare, as described recently for the new US Food and Drug Administration (FDA)-approved opioid drug, Olinvyk (74, 75). Importantly, further addressing such issues with the appropriate tools and technologies should reveal whether efficacy-based drug development strategies can be more successful than mechanism-based approaches, as suggested in recent efforts toward drug repurposing across several areas of medicine (76–78).

In summary, we present a functionally selective small molecule GHSR1a ligand that displays unique and favorable pharmacokinetics and pharmacodynamics (PK/PD). With rapidly expanding advances in molecular docking and dynamics, large-scale in silico compound library screening, and the ever-increasing availability of crystal and cryogenic electron microscopy (cryo-EM) structures, GPCR-directed drug discovery/optimization efforts have become more efficient and effective. N8279 is currently an early lead candidate that can undergo further optimization within this paradigm. Nevertheless, N8279 and/or its analogs may serve as an important structural scaffold for rationally designing safer and more effective GHSR1a-selective treatments to treat DAergic brain diseases, including PD, AD, and/or addictions.

Materials and Methods

Chemicals and Compounds.

All chemicals and reagents were purchased from MilliporeSigma or Bio-Techne Corporation, unless indicated otherwise. For details, see .

Animals.

All animal studies were performed in accordance with the NIH Guidelines for Animal Care and Use of Laboratory Animals and under protocols approved by the Duke University Animal Care and Use Committee (ACUC) or NIH Division of Veterinary Resources ACUC. Male C57BL/6 mice were used for pharmacokinetic studies and were purchased from Charles River Laboratories. DAT KO mice (51) were backcrossed for >10 generations onto a C57BL/6J (Jackson Laboratory) genetic background. C57BL/6J mice were used also in the novelty and behavioral sensitization experiments. Age- and sex-matched littermate mice between 2 and 6 mo of age were used for all behavioral experiments. Mice were bred and maintained on a standard 12:12 h light:dark cycle, socially housed, and supplied with standard laboratory chow and water ad libitum, except during testing.

NCGC00538279 (N8279) Synthesis.

For detailed chemical synthesis procedures of N8279 (NCATS-SM8864), see .

Cell Culture and Transfections.

U2OS, human embryonic kidney (HEK)-293/T, and HEK293/N cells were obtained from the American Type Culture Collection and cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% (vol/vol) of fetal bovine serum and 1× antibiotic-antimycotic solution (100 IU-1 penicillin, 100 μg/mL streptomycin, and 250 ng/mL amphotericin B; MilliporeSigma). HEK293/S Gαq/11 KO and its parental WT line were characterized previously (27, 79) and were a generous gift from Dr. Asuka Inoue (Tohoku University). U2OS cells stably expressing 3×HA-hGHSR1aWT and GFP-tagged βarr2 and HEK293/N cells stably expressing 3×HA-hGHSR1aWT and the luminescent Ca2+ sensor mitochondrial-Aequorin (miAeq) were made by the Caron laboratory and have been described (20, 21). All cell lines were grown in a humidified incubator at 37 °C (5% CO2). All transient transfections were performed using a standard calcium phosphate method.

Plasmids.

All plasmid constructs were purchased commercially or received as a generous gift from collaborative investigators. For details on plasmid acquisition and cloning procedures, see . Mutagenesis was performed with the QuikChange site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA).

High-Throughput and Directed Library Compound Screening.

Quantitative high-throughput screening.

Quantitative high-throughput screening (qHTS) was performed against ∼47,000 compounds from the Sytravon library and the NPC using the PathHunter U2OS GHSR1a-βarr1 cells and the βarr Assay Kit (DiscoverX). Prior to the screening, the assay was miniaturized to a 1,536-well format and optimized in terms of signal-to-background window (S/B), Z factor, and potency of ghrelin control. The initial assay validation was performed with Library of Pharmacologically Active Compounds (Sigma-Aldrich) to confirm plate-to-plate reproducibility of parameters, hit rate identification, etc. For qHTS, two doses of the compounds—11 and 57 μM—were used to measure GHSR1a activation using a fully automated robotic screening system (Kalypsys). Briefly, 1.2 × 103 cells were seeded with MultiDrop Combi dispenser (Thermo Fisher Scientific) into white solid-bottom tissue culture–treated 1,536-well plates (Aurora Microplates) in 3 μL of AssayComplete Cell Plating 5 Reagent (DiscoverX) and cultured overnight. Next, 1 μL per well of 1 μM ghrelin diluted in Hanks' balanced salt solution (HBSS) + 10 mM Hepes (HH buffer) was added to one column of the plate, while all other wells were dispensed with 1 μL per well of HH buffer for matching the assay conditions. Subsequently, the libraries’ compounds in DMSO solution were transferred from the source plates to the assay plates at 23 nL per well. The plates were incubated for 90 min, followed by addition of 1.5 μL per well PathHunter Detection Reagent prepared according to the manufacturer’s instructions. After 60 min of incubation at ambient temperature, the luminescent signal was measured on a ViewLux uHTS Microplate Imager (Perkin-Elmer) with 20 s exposure. Quality of the screening was evaluated based on the median characteristics of Z factor and S/B, which were 0.53 and ∼4.5-fold, respectively. Hit detection window parameters were calculated based on results obtained from the vehicle control (EC0) and ghrelin as the positive control (EC100 = 250 nM final concentration) conditions. A cutoff of >60% activation by either the lower or higher compound’s dose was used to select primary hits. A follow-up PathHunter U2OS GHSR1a βarr1 assay was performed on 145 selected primary compounds. They were retested at seven doses in the range [57 μM to 3.7 nM], applying the same protocol as for qHTS. Thirty-six hits were selected based on curve response class 1, 2, or 3 for further validation.

X-Ray Diffraction.

Single crystal X-ray diffraction studies were conducted on a Bruker Kappa Photon II CPAD diffractometer equipped with Cu Kα radiation (λ = 1.54178 Å). Crystals of the subject compound were grown by dissolving ∼1 mg of sample in 350 μL of 90:10 dichloroethane/methanol solution, which was then vapor diffused with Pentane over several days. A 0.267 × 0.243 × 0.228 mm piece of a colorless block was mounted on a Cryoloop with Paratone oil. Data were collected in a nitrogen gas stream at 285 K using ϕ and ϖ scans. The crystal-to-detector distance was 40 mm using variable exposure time (20 to 90s) depending on θ with a scan width of 2.0°. Data collection was 99.5% complete to 59.009° in θ (0.90 Å). A total of 58,959 reflections were collected, covering the indices −17 ≤ h ≤ 17, −16 ≤ k ≤ 16, and −24 ≤ l ≤ 24. A total of 6,947 reflections were found to be symmetry independent, with an Rint of 0.0596. Indexing and unit cell refinement indicated a primitive, monoclinic lattice. The space group was found to be P21/c. The data were integrated using the Bruker SAINT software program and scaled using the SADABS software program. Solution by direct methods (SHELXT) produced a complete phasing model for refinement. All nonhydrogen atoms were refined anisotropically by full-matrix least-squares (SHELXL-2014). All carbon bonded hydrogen atoms were placed using a riding model. Their positions were constrained relative to their parent atom using the appropriate HFIX command in SHELXL-2014. All other hydrogen atoms (H-bonding) were located in the difference map. Their relative positions were restrained using DFIX commands and their thermals freely refined. Crystallographic data are summarized in .

GHSR1a Radioligand Binding Assays.

[125I]ghrelin-GHSR1a binding assays were performed as described (45), with modifications outlined in .

Gαq-Dependent Intracellular Ca2+ Mobilization.

iCa2+ assays were performed as described (21), with modifications described in .

βarr2GFP Translocation.

βarr2GFP translocation assays were performed as described (20), with modifications described in .

NanoBiT-Based Gαq Dissociation and βarr2 Recruitment Assays.

GαqLgBiT dissociation assays were performed as described (32), with modifications described in . SmBiTβarr2 recruitment assays were performed using a modified βarr2 recruitment protocol as described previously (27). Detailed procedures for both NanoBiT assays are described in the .

BRET-Based Gαq Dissociation and βarr2 Recruitment Assays.

GαRLuc8-GγGFP2 dissociation assays (TRUPATH) were performed as originally described (33). GHSR1aWT-RLucII and GHSR1aL149G-RLucII-Venusβarr2 recruitment assays were performed as described previously (21). Detailed procedures for both BRET assays are described in the .

Chemiluminescent Fixed-Cell ELISA.

Quantitative, fixed-cell ELISAs were performed as described (80), but with modification for chemiluminescent detection of surface-expressed GHSR1a. A detailed procedure can be found in the .

Bystander BRET.

MyrPalmVenus and 2×FYVEVenus bBRET assays were performed as described (38, 40), with modifications described in .

SRF-RE Transcriptional Activity.

SRF-RE transcriptional activation assays were performed as described (20), with modifications described in .

Molecular Docking.

Molecular docking studies were performed using the Glide and Maestro user interface (Release 2019–4, Schrodinger LLC) as described (81, 82). The model structure of ghrelin-bound GHSR1a (30) and the X-ray crystal structure of antagonist-bound GHSR1a (45) were used to represent the active and inactive state of GHSR1a, respectively. The Protein Preparation Wizard function was used to assign bond orders, add hydrogen atoms, and remove water molecules that did not participate in interactions. The GHSR1a models were subjected to energy minimization using the OPLS3 algorithm (A Force Field Providing Broad Coverage of Drug-like Small Molecules and Proteins). A receptor grid box of 30 × 30 × 30 Å3 with a default inner box (10 × 10 × 10 Å3) was centered on the ligand binding pocket. The ligand structures were generated and prepared using the LigPrep function with the OPLS3 force field. Flexible ligand docking was performed using the “standard precision” Glide algorithm, and after the postdocking minimization, the pose with the best docking score was evaluated.

Pharmacokinetic Analysis.

Male C57BL/6 mice (n = 3/time point) were administered N8279 at 1 mg/kg IV, 5 mg/kg PO, and 5 mg/kg IP. Plasma, brain, and liver samples were collected over 24 h. N8279 concentrations in plasma, brain, and liver homogenates were determined by LC-MS/MS. The mean concentration from three animals at each time point was used in the pharmacokinetic (PK) analysis. PK parameters were calculated with Phoenix WinNonlin Software (Ver. 8.0, Certara).

Novelty-Induced Locomotor Activity in DAT KO and Inbred C57BL/6J Mice.

Open-field locomotor activity in mice was performed as described (66), with modifications described in .

Cocaine Sensitization.

Male and female C57BL6/J mice (Jackson Labs) were administered (IP) vehicle or N8279 (5 mg/kg) subchronically for 8 consecutive days (once per day) (Fig. 4). Subsequently, mice were placed into the open field (Accuscan Instruments) for 30 min; they were removed, injected (IP) with vehicle or N8279, returned to the open field for 30 min, then given (IP) vehicle or cocaine (20 mg/kg) and returned to the open field for 120 min (Fig. 4). This procedure was repeated once per day for 5 consecutive days. A drug-free hiatus in the home cage was imposed for 5 d (washout), and then on day 11, the mice were challenged (IP) with vehicle or cocaine (20 mg/kg) to test for behavioral sensitization. Since the injections of vehicle and N8279 occurred 30 min prior to cocaine administration, this time period was taken as baseline activity. Since this baseline locomotor activity declined across days in the N8279 group, the results are presented as percent change from baseline activities.

Statistics.

All data are presented as the mean ± SEM derived from multiple independent experiments or animals. For binding and signaling assays, ≥2 technical replicates were included in each experiment. These data were plotted and analyzed in GraphPad Prism version 9.0 with a statistical significance threshold defined as P < 0.05. Nonlinear regression parameters and best-fit models for all C/R data were determined statistically by an extra sum-of-squares F-test. The behavioral data were analyzed by the appropriate ANOVA followed by a post hoc multiple comparisons test.