G Protein and β-Arrestin Signaling Bias at the Ghrelin Receptor

Background: The G protein coupled receptor GHSR1a mediates feeding and addictive behaviors.

Results: Mutagenesis of the second intracellular loop of GHSR1a generates biased receptors, favoring distinct signaling events.

Conclusion: Receptor conformations that support signaling bias at the wild-type receptor should exist.

Significance: Recapitulating signaling bias at GHSR1a may facilitate the identification of novel selective therapies to treat addiction.

Abstract

The G protein-coupled ghrelin receptor GHSR1a is a potential pharmacological target for treating obesity and addiction because of the critical role ghrelin plays in energy homeostasis and dopamine-dependent reward. GHSR1a enhances growth hormone release, appetite, and dopamine signaling through G_q/11, G_i/o, and G_12/13 as well as β-arrestin-based scaffolds. However, the contribution of individual G protein and β-arrestin pathways to the diverse physiological responses mediated by ghrelin remains unknown. To characterize whether a signaling bias occurs for GHSR1a, we investigated ghrelin signaling in a number of cell-based assays, including Ca²⁺ mobilization, serum response factor response element, stress fiber formation, ERK1/2 phosphorylation, and β-arrestin translocation, utilizing intracellular second loop and C-tail mutants of GHSR1a. We observed that GHSR1a and β-arrestin rapidly form metastable plasma membrane complexes following exposure to an agonist, but replacement of the GHSR1a C-tail by the tail of the vasopressin 2 receptor greatly stabilizes them, producing complexes observable on the plasma membrane and also in endocytic vesicles. Mutations of the contiguous conserved amino acids Pro-148 and Leu-149 in the GHSR1a intracellular second loop generate receptors with a strong bias to G protein and β-arrestin, respectively, supporting a role for conformation-dependent signaling bias in the wild-type receptor. Our results demonstrate more balance in GHSR1a-mediated ERK signaling from G proteins and β-arrestin but uncover an important role for β-arrestin in RhoA activation and stress fiber formation. These findings suggest an avenue for modulating drug abuse-associated changes in synaptic plasticity via GHSR1a and indicate the development of GHSR1a-biased ligands as a promising strategy for selectively targeting downstream signaling events.

Introduction

Advances in G protein-coupled receptor (GPCR)³ pharmacology over the last decade illustrate the importance of β-arrestins in regulating discrete signaling pathways with unique biological outcomes (1). The spectrum of these discrete physiological responses observed following biochemical separation of G protein from β-arrestin activity suggest that the signaling bias obtained by their pharmacological separation will provide a blueprint for designing new generations of selective therapies (2). Recent studies indicate that the growth hormone secretagogue receptor GHSR1a binds β-arrestin following treatment with the hormone ghrelin (3–5). The GHSR1a ghrelin receptor belongs to a small family of GPCRs for peptide hormones and neuropeptides (6), and, by increasing dopamine release (7–9), it mediates the reinforcement of natural rewards, such as food (10–12), and the reinforcement of chemical rewards, such as alcohol (13) and drugs of abuse (9). Therefore, modulating GHSR1a activity has attracted the attention of clinical investigators in the fields of obesity and addiction pharmacology.

Ghrelin peptide is produced mainly in the stomach and within the hypothalamus (14), whereas the GHSR1a is expressed in several brain areas, including the pituitary, hypothalamus, hippocampus, and ventral tegmental area (15, 16). GHSR1a activation in the pituitary stimulates the release of growth hormone (GH) (17, 18). Ghrelin binding to GHSR1a in hypothalamic neuropeptide Y neurons induces appetite and stimulates food intake (19, 20), and these feeding behaviors and associated rewards are controlled by ventral tegmental area and hippocampal GHSR1a. Ventral tegmental area GHSR1a controls the mesolimbic dopaminergic reward circuit (reviewed in Ref. 21), and hippocampal GHSR1a participates in the food anticipation response (22). The GHSR1a couples to G_q/11 and signals through phospholipase C (PLC) and inositol 1,4,5-trisphosphate to increase Ca²⁺ release (23). It also exhibits constitutive internalization and constitutive inositol 1,4,5-trisphosphate accumulation that can reach 50% of maximum activity levels, presumably without ligand present (3). G_i/o signaling has been also reported for GHSR1a in GTPγS assays of cell membranes overexpressing GHSR1a (24) and in lipid disc assays (25). The signaling cascades downstream of GHSR1a are not well studied, but GHSR1a exerts a proliferative effect through activation of the MAPK/ERK and IRS-1 signaling cascades in several cellular systems (26) and activates the small GTPase RhoA (27).

Detailed consequences of the interaction between GHSR1a and β-arrestin are also unclear. Here we report that GHSR1a forms transient complexes with β-arrestin-2 on the cell membrane that can be stabilized by the addition of phosphate acceptor sites in the receptor tail or destabilized by substitution of a conserved proline residue in the receptor ICL2. This interaction with β-arrestin is essential for the GHSR1a-mediated induction of RhoA signaling and actin remodeling, whereas GHSR1a-mediated ERK1/2 phosphorylation involves both β-arrestin and G protein components. Most significantly, mutating amino acid residues for the interaction with β-arrestin or G proteins in ICL2 generated signaling-biased receptors, potentially allowing future studies to determine their individual contribution to various in vivo physiological processes mediated by ghrelin.

EXPERIMENTAL PROCEDURES

Plasmids

The 3xHA-GHSR1a plasmid was purchased from the Missouri S&T cDNA Resource Center and was used to generate 3xHA-GHSR1a_P148A, 3xHA-GHSR1a_A144R, and 3xHA-GHSR1a_L149G by QuikChange site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA). 3xHA-GHSR1a_V2T was generated by replacing the last 28 C-terminal residues of GHSR1a (GFEPFSQRKLSTLKDESSRAWTESSINT) to the last 31 C-terminal residues of the vasopressin 2 receptor (CCARGRTPPSLGPQDESCTTASSSLAKDTSS). The vasopressin 2 receptor and β-arrestin-2-GFP have been described previously (28). The SRE reporter, pGL4.33[luc2P/SRE/Hygro], and the SRF-RE reporter, pGL4.34[luc2P/SRF-RE/Hygro], were purchased from Promega (Madison, WI). The mitochondrion-targeting apoaequorin expression vector (29) was a gift from Dr. Stanley Thayer (University of Minnesota).

Ligands and Inhibitors

The ghrelin peptide was purchased from Abcam (Cambridge, MA). L692,585 (L585), YIL781, GSK269962, and SL327 were purchased from Tocris Biosciences (Ellisville, MI). U73122 and pertussis toxin were from Sigma-Aldrich (St. Louis, MO), lysophosphatidic acid (LPA) was from Avanti Polar Lipid (Alabaster, AL), and JMV3002 was from Cayman Chemicals (Ann Arbor, MI).

Arrestin Translocation Assays

U2OS Cells

U2OS cells permanently expressing 3xHA-GHSR1a and β-arrestin-2-GFP were plated in 384 glass-bottomed (0.15–0.17 mm) plates at a density of 12,000 cells/well in minimum Eagle's medium with 10% FBS. The medium was replaced to clear minimum Eagle's medium with no serum for overnight starvation. For agonist assays, increasing doses of L585 (×4, in 0.1–0.2% DMSO), ghrelin, or DMSO vehicle were added to the cells for 40 min in 37 °C. The cells were washed and fixed with 1% paraformaldehyde (PFA). For the antagonist assay, the cells were preincubated with increasing doses of YIL781 and then incubated with an EC₈₀ of L585 for 40 min prior to their fixation. Images were acquired using a Zeiss Axiovert 200 M fluorescent microscope. The β-arrestin-GFP aggregates were identified by Wavelet software (30), evaluating the number and intensity of fluorescent dots. To compare the different GHSR1a mutants, U2OS cells were plated in 384-well plates and transiently transfected with β-arrestin-2-GFP and the wild-type GHSR1a or one of the mutated receptors using Lipofectamine 2000 transfection reagent (Invitrogen). The next day, the cells were stimulated with L585 and assayed as described for the stable line.

HEK293 Cells

HEK293 cells were plated on poly-d-lysine-coated (Sigma-Aldrich) 35-mm glass-bottom culture dishes (MatTek, Ashland, MA), transiently transfected with β-arrestin-GFP and either the wild-type or mutated 3xHA-GHSR1a using CaCl₂ transfection, serum-starved overnight, and stimulated with either 100 nm ghrelin or L585 for the times indicated in the figures. The treated cells were fixed with 4% PFA for 20 min, permeabilized with 0.1% Triton X-100 for 10 min, blocked in 5% BSA for 30 min, and incubated with a chicken anti-HA antibody (1:500, Abcam) for 1 h. The cells were washed, incubated with a Alexa Fluor 568 secondary antibody (Invitrogen), and imaged using a Zeiss LSM510 confocal microscope.

Internalization Assay

Internalization assays (confocal and on-cell ELISA) were performed in HEK293 cells as described in Ref. 31, with the addition of 100 nm ghrelin or L585 when cells were moved to 37 °C. Cells were starved overnight with 0% FBS and 1× serum replacement medium (Sigma-Aldrich) prior to the trafficking assay.

Ca²⁺ Mobilization Assays

Ca²⁺ mobilization was determined using a functional bioluminescence-based aequorin (Aeq) assay (29, 32). HEK293 cells permanently expressing 3xHA-GHSR1a and mitochondrial apoaequorin were trypsinized, spun down, and resuspended in 1.5 ml of assay medium which contained DMEM, 10% FBS, 10 mm HEPES, and 2.5 μm coelenterazine-h (Promega, Madison, WI). The cells were incubated for 2 h in a 37 °C rotating incubator set to 180 rpm. Following this incubation, the cells were diluted in clear minimum Eagle's medium supplemented with 10 mm HEPES and 2 mm glutamine and dispensed (40 μl, 50,000 cells/well) into increasing doses of ghrelin or L585 that were predispensed into a 96-well white flat and clear bottom microplate at a 2× concentration. Luminescence was recorded using a Mithras LB 940 instrument (Berthold Technologies, Oak Ridge, TN) for 20 s/well immediately after dispensing the cells. After all wells had been read, 80 μl of calcium lysis buffer (100 mm CaCl and 0.2% Triton-X) were dispensed to each well, and the luminescence was recorded for 5 s. To control for variations in cell numbers, the agonist-stimulated response (net Aeq) was normalized by dividing the agonist-induced response (L) by the total response value (L, the agonist-stimulated response, plus the response resulting from detergent cell lysis). For antagonist assays, the cells were first plated on a 96-well plate at a density of 50,000 cell/well and grown overnight in 10% FBS. The next day, the medium was replaced with the assay medium described above, and the cells were incubated for 2 h in a 37 °C incubator. After the 2-h incubation, the assay medium was aspirated, and the cells were treated for 30 min with increasing doses of either YIL781 or JMV3002 in 40 μl at 37 °C. The cells were then treated with the EC₈₀ concentration of L585, dispensed at 5× concentration in 10 μl/well, and the luminescence was recorded as described above. To compare the Ca²⁺ mobilization response of the different GHSR1a mutants, HEK293 cells that permanently expressed the mitochondrial apoaequorin were transiently transfected with either the wild type or one of the mutated GHSR1as in 60-mm culture dishes, collected, and subjected to the agonist assay the next day.

SRE and SRF-RE Luciferase Assays

HEK293 cells were transiently transfected with one of the luciferase reporters, SRE or SRF-RE, and either the wild type or one of the mutated GHSR1as using Lipofectamine 2000 reagent in suspension at a ratio of 1:40 receptor:reporter and plated immediately in poly-d-lysine-coated, 96-well, white, flat, and clear bottom microplates. The next day, the medium was replaced with clear minimum Eagle's medium supplemented with 10 mm HEPES, 2 mm glutamine, and serum replacement medium (Sigma-Aldrich) for overnight starvation. Cells were then incubated for 5 h with increasing doses of L585 at 37 °C and lysed with 20 μl of 1× passive lysis buffer (Promega) for 10 min at room temperature. Luminescence was recorded for 8 s immediately after dispensing 40 μl of firefly luciferin substrate. Luminescence values in the absence of GHSR1a served as background subtraction, and the values were normalized to the response of cells to stimulation with 10% FBS as a positive control. For antagonist assays, the cells were incubated with increasing doses of YIL781 for 30 min at 37 °C prior and during their 5 h incubation with the EC₈₀ value of L585. For experiments with inhibitors, the cells were incubated with either 10 μm SL327 or 5 μm GSK269962 for 30 min at 37 °C prior and during their 5 h incubation with increasing doses of L585.

Curve and Bias Analysis

All dose-response curves were analyzed by nonlinear regression using GraphPad Prism software (GraphPad, San Diego, CA) to obtain EC_max, EC₅₀, and IC₅₀ values. Mean values ± S.E. were calculated by GraphPad Prism from multiple repeats. The signaling bias curve (Fig. 11) was prepared using the GraphPad Prism software according to the mathematical model and formalism described in Ref. 33 and by comparing the parameters between the Ca²⁺ mobilization and β-arrestin translocation assays.

FIGURE 11.

GHSR1a biased signaling. Bias curves for WT and mutated GHSR1a were generated from parameters extrapolated from the dose responses of each receptor in the β-arrestin and Ca²⁺ mobilization assays (33). Positive values indicate a bias toward G_q/11, and negative values toward β-arrestin with a value of plus or minus one indicate complete bias. The ratio of the relative preference for each response is also presented on the abscissa as fold change (for example 2×). Data were plotted using the modeling function of GraphPad Prism 5.0.

Stress Fiber Formation Assay

U2OS cells permanently expressing either the wild-type GHSR1a or HA-GHSR1a-P148A were plated in 35-mm glass bottom culture dishes (MatTek). The next day, cells were treated with 10 μm cytochalasin-B (Sigma-Aldrich) in 0.5% FBS for 2.5 h to depolymerize actin filaments or left untreated. The cytochalasin-B-treated cells were washed and incubated with either 50 μm LPA or 500 nm L585 or left untreated for 30 min, washed, and fixed with 4% PFA for 20 min. To test the effect of GHSR1a antagonist, one dish with cytochalasin-B-treated cells was treated with a combination of YIL781 and L585: 5 μm YIL781 was added to the last 20 min of the cytochalasin-B treatment, and the YIL781 treatment continued during the 30 min with L585. To detect F-actin, the fixed cells were incubated with Alexa Fluor 594 phalloidin for 1 h (1:200, Invitrogen) and imaged using a Zeiss LSM510 confocal microscope. The percent of cells that assemble stress fibers under the different conditions is expressed as mean ± S.E. and analyzed by one-way analysis of variance with Bonferroni's post hoc test (GraphPad).

ERK1/2 Phosphorylation

HEK293 cells were plated in 6-well culture dishes and transiently transfected with wild-type 3xHA-GHSR1a, 3xHA-GHSR1a-P148A, or 3xHA-GHSR1a-L149G. The cells were starved overnight in 0.5% FBS and 1× serum replacement medium. Cells were either treated with 200 ng/ml pertussis toxin (Sigma-Aldrich) during the overnight starvation or treated the next day for 30 min with 5 μm U73122, or DMSO vehicle. At the end of the 30 min, the cells were stimulated with 500 nm L585 for 0, 5, 10, 20, and 30 min and lysed with a fractionation buffer (50 mm Tris (pH 7.4) and 0.3% Triton X-100 supplemented with 1× complete mini protease inhibitor mixture (Roche) and 1× phosphatase inhibitor mixture II (Alfa Aesar, Ward Hill, MA). Lysed cells were spun down for 5 min at 2300 rpm, and the supernatant was collected as the cytoplasmic fraction. The pellet (the nuclear fraction) was washed with the fractionation buffer, centrifuged again, and resuspended in the fractionation buffer. Both fractions were sonicated briefly, and the nuclear fraction was run through an insulin syringe. Both fractions were loaded on SDS gels, and the levels of ERK1/2 phosphorylation were assessed by the ratio between phosphorylated ERK1/2 detected by an anti-pERK1/2 antibody and the total amounts of ERK1/2 detected by an anti-ERK1/2 antibody (1:500, Cell Signaling Technology, Danvers, MA). The band intensities of ERK1 and ERK2 in the two cellular fractions were analyzed separately using ImageJ software and summarized (ERK1 + ERK2 for each fraction), and the ratio of pERK1/2:total ERK1/2 at time 0 was normalized to 1. Values are expressed as mean ± S.E. p values were calculated by one-way analysis of variance with Bonferroni's post hoc test (GraphPad). Lamin B was detected in the nuclear fraction only (anti-Lamin B antibody, 1:1000, Abcam).

RESULTS

β-Arrestin-2 Interaction with the GHSR1a in U2OS Cells

GPCRs fall into two classes on the basis of the strength of their ligand-mediated, phosphorylation-dependent internalization with β-arrestins. Class A GPCRs dissociate from β-arrestins at or near the plasma membrane, whereas class B ones form stable complexes that traffic into endocytic vesicles (28). We evaluated this trafficking in U2OS cells that permanently express GHSR1a and β-arrestin-2-GFP (Fig. 1A) because it reflects the formation, stability, and duration of β-arrestin signaling complexes. Stimulation of GHSR1a with its natural peptide agonist ghrelin and with the non-peptide small molecule agonist L585 (34, 35) leads to the formation of numerous β-arrestin-2·GFP/receptor complexes on or near the plasma membrane (Fig. 1B). Wavelet-based determination of the formation of receptor/β-arrestin complexes (aggregates, Fig. 1C) (30) showed that ghrelin was more potent than L585, with an EC₅₀ of 8.5 ± 1.7 nm versus 46.8 ± 16.9 nm (Fig. 1D). Interestingly, in the absence of an agonist (basal) and in contradistinction to what has been reported previously (5), β-arrestin-2-GFP remains in a diffuse cytosolic distribution with no apparent receptor-mediated constitutive translocation. L585-induced translocation was also blocked in a concentration-dependent manner by the small molecule antagonist YIL781 (36), with an IC₅₀ of 498 ± 168 nm (Fig. 1E).

FIGURE 1.

GHSR1a-mediated β-arrestin-2 translocation. A, U2OS cells. B and C, permanent expression of β-arrestin-2-GFP in the cytosol and 3xHA-GHSR1a on the membrane of untreated U2OS cells. Shown are high content (B) and processed (C) images of cells treated with ghrelin, L585, or vehicle control (basal). D and E, dose responses of ghrelin, L585 (D), and YIL781 (E). The numbers of cytoplasmic aggregates per area were calculated using Wavelet software. EC₅₀ and IC₅₀ values were calculated from three independent experiments. F, HEK293 cells were transiently transfected with β-arrestin-2-GFP and HA-GHSR1a, serum-starved for 16 h, and treated for 1 h with 100 nm L585 or ghrelin or left untreated. After the treatment, the cells were washed and fixed, and the receptor was detected using an anti-HA antibody (red). Colocalization is shown in yellow.

β-Arrestin-2 Interaction with the GHSR1a in HEK293 Cells

A specific cluster of phosphorylated serine/threonine residues in the C terminus of class B GPCRs enables them to form stable complexes with β-arrestin during endocytosis (37). These complexes are readily apparent by confocal microscopy because they distinctly label the membranes of early endosomes but are absent from their interior (38, 39), and we previously identified the location of these clusters for representative GPCRs by C-tail mutagenesis (37, 39). To verify our U2OS observations concerning the class A or B status of the GHSR1a, we used a HEK293 cell model. Similar to our finding for U2OS cells, β-arrestin-2 does not stably interact with GHSR1a in the absence of a ligand (Fig. 1F), and even a 1-h incubation of GHSR1a with 100 nm ghrelin or L585 leads to only plasma membrane colocalization with β-arrestin-2 (Fig. 1F). The absence of well demarcated β-arrestin-GFP endocytic vesicles indicates that GHSR1 is a class A GPCR and lacks the complement of phosphorylation sites necessary for more stable β-arrestin-2 interactions. Interestingly, under basal non-stimulated conditions, 3xHA-tagged GHSR1a can be detected by immunofluorescence not only on the membrane of HEK cells but also in β-adaptin-containing intracellular vesicles (Fig. 2). These results agree with previous reports of clathrin-dependent constitutive internalization of GHSR1a (40, 41).

FIGURE 2.

Clathrin-mediated internalization of GHSR1a. HEK293 cells that stably express adaptin-GFP were transiently transfected with HA-GHSR1a and serum-starved overnight. Live cells were incubated with an anti-HA antibody for 45 min at 4 °C to prevent internalization and then moved to 37 °C for 40 min with or without ghrelin. Cells were fixed and incubated with a secondary antibody to detect the HA-tagged receptor (red). Arrows indicate colocalization of GHSR1a and adaptin-GFP (yellow).

Identification of GHSR1a C-Tail Amino Acids Regulating β-Arrestin-2 Binding

Because GHSR1a is a class A GPCR, we next asked the question: what are the structural determinants that impart this property? We replaced the GHSR1a C-tail after amino acid Leu-322 with the last 31 residues of the class B vasopressin receptor 2 (GHSR1a-V2T, Fig. 3) (28). Using an aequorin mitochondrial-based assay, we measured bioluminescence differences reflecting G_q/11 mediated Ca²⁺ mobilization between the wild type and the chimeric receptors in HEK293 cells (29, 32). Interestingly, L585 was more potent than ghrelin in stimulating Ca²⁺, with EC₅₀ values of 2.7 ± 1.2 nm and 93.5 ± 11.6 nm, respectively (Fig. 4A). This order of potency is consistent with reports on similar small molecule agonists of the ghrelin receptor (4), but it is the reverse of what we found for the β-arrestin translocation assay. As expected, the reported ghrelin antagonists YIL781 and JMV3002 inhibited EC₈₀ L585 stimulation, with IC₅₀ values of 46.5 ± 11.5 nm and 2.5 ± 1.3 nm, respectively (Fig. 4B). The PLC inhibitor U73122 dose-dependently blocked the L585-induced Ca²⁺ response (Fig. 4C), indicating classic PLC-mediated G_q/11 signaling. GHSR1a-V2T is essentially indistinguishable from GHSR1a in potency and efficacy determinations of Ca²⁺ mobilization, and both receptors have very low constitutive G_q/11 activity (Fig. 4D). β-Arrestin-2-GFP also translocates equally well to both plasma membrane GHSR1a and GHSR1a-V2T after 2 min of stimulation with L585 (Fig. 4, E and F), but 25 min of stimulation distinguishes the two receptors. Extended incubation leads to the colocalization of β-arrestin-2-GFP and GHSR1a-V2T in endocytic vesicles (Fig. 4F), suggesting that adding phosphorylation sites to the C terminus of GHST1a converts it to a class B subtype.

FIGURE 3.

GHSR1a mutagenesis. Schematic of GHSR1a intracellular loops and the C-tail. The C-tail of the vasopressin 2 receptor (V2T), used to generate GHSR1a_V2T, is presented below the GHSR1a C-tail. Serine/threonine phosphorylation sites in both receptors are shown in light blue. The three mutated residues in ICL2 are shown in yellow.

FIGURE 4.

Switching the C tail of GHSR1a to that of the vasopressin 2 receptor increases GHSR1a affinity to β-arrestin-2. A–C, Aeq-based Ca²⁺ release in HEK293 cells that permanently express Aeq and GHSR1a. A, ghrelin and L585 dose responses. The EC₅₀ values are 93 nm for ghrelin and 2.7 nm for L585. B, YIL781 and JMV3002 dose responses. Cells were pretreated with increasing concentrations of the indicated antagonists, stimulated with 50 nm L585 in the presence of the antagonist, and subjected to the Aeq assay. The IC₅₀ values are 47 nm for YIL781 and 2.5 nm for JMV3002. C, L585 dose response under PLC inhibition with U73122. D, HEK293 cells that permanently express Aeq were transiently transfected with either WT HA-GHSR1a or GHSR1a-V2T, treated with increasing doses of L585, and subjected to the Aeq assay. The EC_max of GHSR1a-WT was normalized to 100% (n = 3 for all dose responses). E and F, HEK293 cells were transiently transfected with β-arrestin-2-GFP and either WT GHSR1a (E) or GHSR1a-V2T (F), serum-starved for 16 h, and treated with 100 nm ghrelin for 2 min or 25 min or left untreated (basal). After the treatment, the cells were washed and fixed, and the receptor was detected using an anti-HA antibody (red; colocalization is shown in yellow in the bottom panel). The inset in F is a magnification of several cytoplasmic vesicles containing both GHSR1a-V2T and β-arrestin-2-GFP. β-arr, β-arrestin. The scale bar in the inset represents 3.5 μm.

GHSR1a Intracellular Loop 2 Regulation of β-Arrestin Complex Stability

We have shown previously that a conserved proline in ICL2, located nine amino acids distal to the start of the highly conserved E/DRY motif, regulates β-arrestin-mediated trafficking of multiple GPCRs (42). A recent study indicates that the β-adrenergic receptor complex with G_s requires an ICL2 alanine four positions upstream of the proline (43). Another conserved site in ICL2 shown to be important for G protein coupling is a hydrophobic residue located one position after the proline (Fig. 3) (42, 44). Therefore, we hypothesized that the corresponding proline 148, alanine 144, or leucine 149 of GHSR1a might play important roles in β-arrestin-2 and G_q/11 signaling. To test our hypothesis, we substituted these residues and studied the signaling from the mutated receptors. First, we tested the surface expression of the mutated receptors as well as the rates of their constitutive internalization by on-cell ELISA. We found that substitution of Pro-148 by alanine (GHSR1a_P148A, Fig. 3) results in 60% surface expression of the mutated receptor compared with the WT (Fig. 5, A and B), whereas substitution of Ala-144 by arginine (GHSR1a_A144R) results in 100% surface expression, and substitution of Leu-149 by glycine results in 150% surface expression (Fig. 5, A and B) because of relatively lower rates of constitutive internalization compared with the WT receptor (Fig. 5C). However, the changes in surface expression and internalization rates were found to be statistically insignificant (p values > 0.05). Surprisingly, the substitution of A144R does not affect either β-arrestin-2 translocation (Figs. 6A and 7A) or G_q/11 activity (Fig. 7B). In contrast, P148A dramatically reduces β-arrestin-2 trafficking (Figs. 6B and 7A) without diminishing GHSR1a-mediated Ca²⁺ mobilization (Fig. 7B). In fact, despite a reduction of 40% in its surface expression compared with the wild type (Fig. 5), GHSR1a_P148A demonstrates an efficacy of 150% that of the wild-type receptor for Ca²⁺ responsiveness, probably because of reduced β-arrestin-dependent desensitization of G_q/11 signaling (Fig. 7B). To further investigate the importance of Pro-148, we tested the translocation of β-arrestin-2 to GHSR1a-V2T_P148A and observed a similar lack of an agonist-mediated β-arrestin-2 response (Fig. 6C). Upon substitution of Leu-149 to glycine, we found that GHSR1a_L149G recruits β-arrestin-2 (Fig. 6D) with lower potency than the wild-type receptor (EC₅₀ values of 89.2 ± 20.1 nm versus 5.1 ± 1.8 nm) but with similar efficiency (Fig. 7A). Interestingly, GHSR1a_L149G lacks G_q/11 activity (Fig. 7B), even though it is highly expressed on the surface (Fig. 5), indicating that position 149 regulates signaling bias. Notably, all GHSR1a variants display comparable and statistically indistinguishable internalization rates in the presence of L585 (p > 0.05, Fig. 7C). To test whether L585 works similarly as the endogenous agonist ghrelin, we tested the responses of the three GHSR1a mutants to ghrelin in the β-arrestin-2 translocation and Ca²⁺ assays and found that ghrelin and L585 lead to similar bias profiles (Fig. 7, D and E).

FIGURE 5.

GHSR1a surface expression and constitutive internalization. HEK293 cells were transiently transfected with the indicated variant of GHSR1a, pulsed with a primary anti-HA antibody on ice, chased at 37 °C, fixed, and then stained with a 680 secondary antibody without permeabilization to assess the fraction of the receptor pulsed that remained on the surface following the chase. Cells were imaged on a LiCOR Odyssey (n = 4 independent experiments). A, stained wells from a representative experiment performed in quadruplicates. B, surface expression values of the different variants at time 0 and normalized to the WT receptor. C, internalization rates expressed as percent of receptor that remains on the cell surface at the indicated time point. Values are normalized to the receptor on the cell surface at time 0 for each construct. The changes in B and C are statistically insignificant (p > 0.05).

FIGURE 6.

Agonist-dependent β-arrestin translocation to GHSR1a-A144R and GHSR1a-L149G but not GHSR1a-P148A. HEK293 cells were transiently transfected with β-arrestin-2-GFP and either HA-GHSR1a-A144R (A), HA-GHSR1a-P148A (B), HA-GHSR1a-V2T-P148A (C), or GHSR1a-L149G (D); serum-starved for 16 h; and treated with 100 nm ghrelin for 2 or 25 min or left untreated (basal). After the treatment, the cells were washed and fixed, and the receptor was detected using an anti-HA antibody (red; colocalization is shown in yellow in the bottom rows). The images represent three independent experiments. Because the cells were transiently transfected, only cells with surface expression of 3xHA-GHSR1a were imaged and analyzed.

FIGURE 7.

The importance of Ala-144, Pro-148, and Leu-149 for β-arrestin-2 translocation and Ca²⁺ mobilization. A, β-arrestin translocation. U2OS cells were transiently transfected with β-arrestin-2-GFP and either GHSR1a WT or one of the three mutants (A144R, P148A, or L149G), serum-starved for 16 h, treated with increasing doses of L585 for 1 h, fixed, imaged, and analyzed using Wavelet software (n = 3). B, Ca²⁺ mobilization. HEK293 cells that permanently express Aeq were transiently transfected with the indicated variant of GHSR1a. Shown are the dose responses of L585 in the Aeq assay (n = 4–5). The EC_max of GHSR1a WT was normalized to 100%. C, receptor internalization. HEK293 cells were transiently transfected with the indicated variant of GHSR1a, pulsed with a primary anti-HA antibody on ice, chased with L585 at 37 °C, fixed, and then stained with a 680 secondary antibody without permeabilization to assess the fraction of the receptor pulsed that remained on the surface following the chase. Cells were imaged on a LiCOR Odyssey, and the values were normalized to the receptor on the cell surface at time 0 for each construct (n = 4). D and E, β-arrestin translocation (D) and Ca²⁺ mobilization (E) in response to ghrelin. Cells expressing GHSR1a WT or one of the three mutants were treated with increasing doses of ghrelin as in A and B, respectively (n = 3).

β-Arrestin Role in GHSR1a-mediated RhoA Activation

G_q/11 and G_i/o-coupled receptors are known to activate the MAPK/ERK signaling pathway, whereas those that are coupled to G_12/13 activate the RhoA/ROCK pathway. GHSR1a activates both the MAPK/ERK and RhoA/ROCK signaling pathways (45). However, the relative contributions of G_12/13, G_q/11, G_i/o, and β-arrestins are undefined. G_q/11 and G_i/o both mediate Ca²⁺ release, and G_q/11 signaling leads to ERK activation (45), but it is unclear whether G_i/o coupling to GHSR1a has the same effect. G_q/11 and G_12/13, but not G_i/o, activate SRE transcription downstream of GHSR1a (27), but the contribution of β-arrestin is, again, unknown. To measure this signaling, we employed two bioluminescence-based luciferase reporter assays, an SRE assay for MAPK/ERK1/2 activation and an SRF-RE assay, a modified SRE for RhoA/ROCK activation (46, 47). SRE responds to ternary complex factor-dependent MAPK/ERK pathway, whereas SRF-RE, a mutant form of SRE lacking a ternary complex factor binding domain, responds to an SRF-dependent and ternary complex factor-independent pathway, such as RhoA activation (48). In HEK293 cells containing GHSR1a, L585 activates both SRE and SRF-RE reporters (Fig. 8, A and B), and the YIL781 GHSR1a antagonist inhibits the reporter responses to L585 (Fig. 8, C and D). We also assessed reporter assay selectivity using the MEK1/2 inhibitor SL327 (49) and the ROCK inhibitor GSK269962 (50). Both MEK1/2 and ROCK inhibitors block the GHSR1a-mediated SRE response, reducing the efficacy and the potency of L585 (Fig. 8A). These results suggest that this SRE activity stems from a combination of MAPK/ERK1/2 and RhoA/ROCK signaling cascades, and this assay cannot be used to distinguish between them, whereas the GHSR1a-mediated SRF-RE response is predominantly inhibited by GSK269962 (Fig. 8B). Moreover, the SRF-RE response is not G_i/o-mediated, being pertussis toxin-independent (Fig. 8E). Taken together, these data indicate that the SRF-RE is a specific measure of RhoA/ROCK activity. Compared with the WT receptor, GHSR1a_P148A and GHSR1a_L149G demonstrate attenuated SRF-RE responses, with 58 and 32% efficacies and EC₅₀ values of 10.0 ± 2.0 nm and 8.1 ± 2.4 nm compared with 2.9 ± 1.4 nm of the WT (Fig. 8F), suggesting that GHSR1a-mediated RhoA signaling has a significant β-arrestin component that remains in the functionally unbiased GHSR1a _A144R (Fig. 8F).

FIGURE 8.

The importance of Ala-144, Pro-148, and Leu-149 for downstream transcriptional activity. HEK293 cells were transiently transfected with the WT GHSR1a and the SRE-luciferase (Luc, A) or SRF-RE-luciferase (B) reporters, serum-starved for 16 h, and pretreated for 30 min with either 10 μm MEK inhibitor (SL327) or 5 μm ROCK inhibitor (GSK269962). The response of the luciferase reporter to increasing doses of L585 in the presence of the indicated inhibitors was measured (n = 5). C and D, cells transiently transfected with the WT GHSR1a and SRE-luciferase (C) or the SRF-RE-luciferase (D) reporters were starved, pretreated with increasing doses of the YIL781 antagonist, and stimulated with the EC₈₀ of L585 in the presence of YIL781. EC₅₀ and IC₅₀ values were calculated from at least three independent experiments. E, HEK293 cells were transiently transfected with the WT GHSR1a and the SRF-RE-luciferase reporter, treated overnight with pertussis toxin (PTX) under starvation conditions, and then the response of the luciferase reporter to L585 was measured (n = 3). F, HEK293 cells were transiently transfected with the indicated GHSR1a variant and the SRF-RE-reporter and serum-starved, and then the response of the reporter to increasing doses of L585 was measured (n = 3–6).

G_i/o, G_q/11, and β-Arrestin Components of GHSR1a-mediated ERK1/2 Activation

Studies suggest a β-arrestin-independent mechanism of GHSR1a ERK1/2 activation because dominant negative β-arrestins that prevent GHSR1a endocytosis in clathrin-coated pits do not block GHSR1a-mediated ERK1/2 phosphorylation (51). We treated HEK293 cells expressing GHSR1a with L585 in the absence and presence of pertussis toxin or the PLC inhibitor U73122 and assessed the phosphorylation of cytosolic and nuclear ERK fractions. L585 treatment induced rapid (5 min) 4.3 ± 0.6-fold and 5.2 ± 0.8-fold increases in phosphorylated ERK1/2 in the cytoplasm and nucleus, respectively (p < 0.05, Fig. 9, A–C), consistent with other studies (51). The response in the cytoplasm was reduced to 50% after preincubation with pertussis toxin or U73122. The response in the nucleus was reduced to 33% (p < 0.05) and 61% (p = 0.09) following preincubation with pertussis toxin and U73122, respectively, suggesting that GHSR1a-mediated ERK1/2 phosphorylation is mediated by both G_i/o and G_q/11. Moreover, G_i/o appears more critical than G_q for nuclear ERK1/2 induction.

FIGURE 9.

GHSR1a-mediated ERK1/2 phosphorylation. HEK293 cells were transiently transfected with the indicated variant of GHSR1a, starved, and treated with 200 ng/ml PTX, 5 μm U73122, or DMSO control. Cells were stimulated with 500 nm L585 and lysed after 0, 5, 10, 20, and 30 min. The nuclear and cytoplasmic fractions were separated and loaded on SDS gels. Phosphorylated (p)ERK1/2 and total ERK antibodies were used to measure ERK1/2 activation. The band intensities of ERK1 and ERK2 were analyzed separately and summarized. The summarized pERK1/2/total ERK1/2 is presented as a ratio of time 0 before L585 stimulation (A, nucleus; B, cytoplasm; n = 4). *, p < 0.05 comparing the 5-min time point to time 0; #, p < 0.05 comparing DMSO to pertussis toxin and U73122 at 5 min (with the exception of p = 0.09 for the WT receptor in A, comparing DMSO to U73122, one-way analysis of variance). C–E, representative gels from four experiments with the WT (C), P148A (D) and L149G (E) receptors. The Lamin B antibody was used to labeled nuclear proteins.

We further tested the ERK response to L585 for the G protein-biased receptor GHSR1a_P148A. In cells expressing GHSR1a_P148A, L585 produced a rapid 3.1 ± 0.8-fold cytoplasmic induction and a 5.5 ± 1.5-fold nuclear induction of phosphorylated ERK1/2 (p < 0.05, Fig. 9, A, B, and D). The ERK response of GHSR1a-P148A in the presence of U73122 or pertussis toxin is similar to that of the wild-type receptor. However, pertussis toxin treatment results in a delay of 5 min in the induction of the response (Fig. 9, A, B, and D). This suggests an alternative, non-G_i/o pathway that can mediate ERK signaling for GHSR1a-P148A. In cells that express the β-arrestin-biased GHSR1a_L149G, we observed, in comparison with the wild type GHSR1a, 49% and 11% reductions in ERK1/2 activation in the cytoplasmic and nuclear fractions, respectively (2.2 ± 0.1-fold induction in the cytoplasm and 4.6 ± 0.8-fold induction in the nucleus at 5 min, p < 0.05). GHSR1a_L149G-mediated ERK phosphorylation in the cytoplasm is not sensitive to either pertussis toxin or the PLC inhibitor U73122, and, in the nucleus, ERK phosphorylation is only slightly affected by U73122 (all p values > 0.05, Fig. 9, A, B, and E). These results support the notion that GHSR1a_L149G is a β-arrestin-biased receptor.

β-Arrestin-dependent GHSR1a signaling Induces Stress Fibers

Activation of RhoA by LPA in starved fibroblasts results in actin cytoskeletal reorganization and the formation of focal adhesions and stress fibers (52). To explore whether GHSR1a can mediate stress fiber formation and whether β-arrestin plays a role in this process, we measured F-actin reorganization in U2OS cells overexpressing β-arrestin-2 and either the wild-type GHSR1a or GHSR1a-P148A. In both cell lines, with no treatment, over 60% of the cells have actin stress fibers. Following a 2.5-h treatment with 10 μm cytochalasin-B in serum starvation medium stress fibers are disrupted in 50% of the wild type and in 60% of the cells stably expressing GHSR1a-P148A (Fig 10). Stimulating the cytochalasin-B pretreated cells with 50 μm LPA for 30 min leads to the reformation of stress fibers in both the WT and GHSR1a-P148A lines. In agreement with a role for β-arrestin in GHSR1a-mediated RhoA activation, 0.5 μm L585 stimulates stress fiber reassembly in WT GHSR1a cells (Fig. 10, A and C) but not in the GHSR1a_P148A line (Fig. 10, B and D). This agonist-induced increased in stress fibers is suppressed by 5 μm of the antagonist YIL781. These findings confirm that GHSR1a-mediated RhoA activation can lead to actin reorganization and that the interaction between GHSR1a and β-arrestin is necessary for this to occur.

FIGURE 10.

GHSR1a interaction with β-arrestin-2 is important for GHSR1a-mediated stress fiber formation. U2OS cells that permanently express βarrestin-2-GFP and either WT GHSR1a (A) or GHSR1a-P148A (B) were left untreated (basal) or treated with 10 μm cytochalasin-B in 0.5% FBS for 2.5 h. After 2.5 h, the treated cells were washed and then left unstimulated or stimulated with 50 μm LPA, 500 nm L585, or a combination of 500 nm L585 and 5 μm YIL781 in 0.5% FBS for 30 min (YIL781 incubation was initiated 30 min prior to L585 and maintained during the stimulation). Cells were fixed and stained with 568-phalloidin. C and D, bar graphs represent the percent of cells that assemble stress fibers under the indicated conditions. C, *, p < 0.05 (n = 5, one-way analysis of variance with Bonferroni's post hoc test). The differences between the reciprocal treatments in D are statistically insignificant.

DISCUSSION

Recent advances in the understanding of addiction underlie strategies calling for selective pharmacological intervention of addictive behaviors (53). The ghrelin receptor GHSR1a has attracted increased interest because of an association with dopamine-dependent reward, and potential ghrelin receptor compounds useful for this purpose would be small druggable molecules that function to attenuate the reinforcing properties of drugs of abuse without affecting GH release, body weight, and food intake. Ligands demonstrating a signaling bias between G proteins or β-arrestin have potential therapeutically benefits (reviewed in Ref. 54). The physiological advantages of pathway-selective GHSR1a targeting include more efficacious addiction treatments with reduced side effects. Our investigation into the G protein and β-arrestin signaling bias for GHSR1a was undertaken with this mind.

The biochemical basis of biased signaling may reside in the stabilization of only those conformations that are capable of promoting the desired subset of signaling pathways (54). Therefore, determining that a receptor is even capable of assuming conformations favoring either one of the G protein and β-arrestin signaling pathways strongly supports the search for small molecules that could recapitulate this behavior. In short, GPCRs showing a signaling bias when mutated in key determinants should be more likely to have biased ligands identifiable by high throughput screens, and, in our case we mutated specific residues in GHSR1a ICL2 on the basis of previous studies, crystal structure determinations, and evolutionary trace arguments for conserved binding motifs (42, 43, 55). An analysis of GHSR1a-mediated G_q/11 and β-arrestin responses (33) (Fig. 11) showed the importance of residue Pro-148 to β-arrestin-mediated signaling. In agreement with that of several other GPCRs (42), its mutation yields a complete bias of the GHSR1a toward G_q/11 signaling. Mutating the adjacent residue Leu-149 yields the opposite, generating a complete bias toward β-arrestin. The bias plot shown in Fig. 11 indicates a complete bias of the two mutants in all tested doses, ranging from low picomolar to high micromolar of the agonist. Given the physical proximity of these two residues, these opposite outcomes are somewhat intriguing.

The proline in ICL2 is central to a 10-amino acid segment that begins with the E/DRY motif and is believed to create a conformational pocket for β-arrestin and G protein binding (42). One would expect that mutating Pro-148 will lead to a local conformational change, interfering with the structure of neighboring residues, which may consequently lose their activity as well. Indeed, this is the case with the serotonin 2C receptor, which loses the ability to interact with both β-arrestin and G proteins (42). Our previous modeling of the ICL2 of the rhodopsin receptor, with a proline-to-alanine mutation, indicates that the secondary structure of rhodopsin because of the mutation is relatively unaffected at proximal downstream residues in comparison with upstream ones (42). Together, these data suggest that proline 148 and leucine 149 may serve as hot spots for β-arrestin and G_q/11 binding, respectively, rather than simply mediating structural conformations involving adjacent residues. Surprisingly, we found that mutating either proline 148 or leucine 149 did not affect the agonist-induced internalization of GHSR1a. This could suggest that β-arrestin is not necessary for this process. However, because agonist does induce β-arrestin recruitment, arrestins probably do have a distinct role in determining the evolution of the trafficking/signaling properties of the receptor complex. Identifying a receptor mutant that lacks an ability to constitutively internalize will enable the determination of the contribution of non-arrestin-mediated internalization when arrestins are activated. The G_q/11-coupled neurotensin receptor 1 (NTR1) is closely related to GHSR1a. Recently, a β-arrestin-biased ligand, ML314, was identified for NTR1, utilizing a β-arrestin translocation assay in a high throughput screen (56, 57). This brain penetrant small molecule is a full agonist in the β-arrestin assay but lacks the ability to induce G_q/11 signaling (57). Our identification of biased ghrelin receptors and the close evolutionary relationship between GHSR1a and NTR1 (6) suggest that small molecule biased ligands of GHSR1a that discriminate between β-arrestin and G protein signaling will eventually be identified. Although a few ligands have already been described for the ghrelin receptor, exhibiting some signaling bias, none has been associated with β-arrestin. The ghrelin peptide mimetic agonist wfw-Isn-NH₂ supports receptor internalization, G_q/11 signaling, and ERK1/2 phosphorylation but not SRE-mediated transcriptional activity or food intake (27). GSK161443, a small molecule ghrelin antagonist of calcium mobilization and inositol phosphate turnover (58), acts in vivo as an antagonist of GH release, an inducer of food intake, and a promoter of increased body weight (59, 60), but the mechanistic and physiological basis for this has not been determined. Other described small molecule non-peptide agonists (34, 61) and antagonists (36, 62, 63) of GHSR1a have been developed to treat GH deficiency, cachexia, and obesity, and, in a few cases, GH release was discriminated from food intake (59, 62, 63).

ERK1/2 is required for the proliferative effect of ghrelin in intestinal epithelial and pancreatic endothelial cells (64–66), and G protein activation and β-arrestin scaffolding are both associated with enhanced cell growth (1, 64). Our findings from the ICL2 mutants indicate that GHSR1a-mediated ERK1/2 phosphorylation results from a combination of G protein and β-arrestin signaling because ERK phosphorylation occurs upon activation of either the β-arrestin-impaired GHSR1a_P148A or the G_q/11- and G_i/o-impaired GHSR1a_L149G. With the P148A mutation, a reduction in the ability to desensitize G protein could potentially support ERK activity, resulting in an underestimation of the direct role of β-arrestin scaffolding in maintaining this pathway. Such a feedback mechanism, however, could make the ERK signaling cascade less sensitive to changes in G protein or β-arrestin concentrations.

The L149G mutation reduced RhoA-dependent reporter activity, suggesting a role for G_12/13 and G_q/11 but not G_i/o, given that this response is not sensitive to pertussis toxin. We found proline 148 to be critical to actin remodeling by using a RhoA-dependent reporter and measuring actin stress fiber formation. It is already known that GHSR1a mediates reward and the reinforcing properties of drugs of abuse (9), and the effects of ghrelin on spine densities in the hippocampus have already been reported to regulate learning and memory performance (65), probably through activation of the PI3K and cAMP signaling pathways (66, 67). Therefore, it will be informative to test whether GHSR1a has a similar effect on RhoA-mediated actin remodeling in neurological model systems, especially whether it contributes to the changes in spine morphology observed in the nucleus accumbens after repeated treatments with cocaine (68).