Modulation of Constitutive Activity and Signaling Bias of the Ghrelin Receptor by Conformational Constraint in the Second Extracellular Loop*

Background: A natural Glu for Ala variant in the ghrelin receptor extracellular loop 2 selectively eliminates constitutive signaling. Results: Computational chemistry and mutational analysis show that charged residues and metal ion sites that induce α-helix formation in ECL2 prevent constitutive signaling. Conclusion: Flexibility of ECL2 connecting TM-III and TM-V is essential for spontaneous receptor signaling. Significance: Clarification of ECL2 structural constraint is important for receptor signaling. Based on a rare, natural Glu for Ala-204(C+6) variant located six residues after the conserved Cys residue in extracellular loop 2b (ECL2b) associated with selective elimination of the high constitutive signaling of the ghrelin receptor, this loop was subjected to a detailed structure functional analysis. Introduction of Glu in different positions demonstrated that although the constitutive signaling was partly reduced when introduced in position 205(C+7) it was only totally eliminated in position 204(C+6). No charge-charge interaction partner could be identified for the Glu(C+6) variant despite mutational analysis of a number of potential partners in the extracellular loops and outer parts of the transmembrane segments. Systematic probing of position 204(C+6) with amino acid residues of different physicochemical properties indicated that a positively charged Lys surprisingly provided phenotypes similar to those of the negatively charged Glu residue. Computational chemistry analysis indicated that the propensity for the C-terminal segment of extracellular loop 2b to form an extended α-helix was increased from 15% in the wild type to 89 and 82% by introduction in position 204(C+6) of a Glu or a Lys residue, respectively. Moreover, the constitutive activity of the receptor was inhibited by Zn2+ binding in an engineered metal ion site, stabilizing an α-helical conformation of this loop segment. It is concluded that the high constitutive activity of the ghrelin receptor is dependent upon flexibility in the C-terminal segment of extracellular loop 2 and that mutations or ligand binding that constrains this segment and thereby conceivably the movements of transmembrane domain V relative to transmembrane domain III inhibits the high constitutive signaling.

Based on a rare, natural Glu for Ala-204(C؉6) variant located six residues after the conserved Cys residue in extracellular loop 2b (ECL2b) associated with selective elimination of the high constitutive signaling of the ghrelin receptor, this loop was subjected to a detailed structure functional analysis. Introduction of Glu in different positions demonstrated that although the constitutive signaling was partly reduced when introduced in position 205(C؉7) it was only totally eliminated in position 204(C؉6). No charge-charge interaction partner could be identified for the Glu(C؉6) variant despite mutational analysis of a number of potential partners in the extracellular loops and outer parts of the transmembrane segments. Systematic probing of position 204(C؉6) with amino acid residues of different physicochemical properties indicated that a positively charged Lys surprisingly provided phenotypes similar to those of the negatively charged Glu residue. Computational chemistry analysis indicated that the propensity for the C-terminal segment of extracellular loop 2b to form an extended ␣-helix was increased from 15% in the wild type to 89 and 82% by introduction in position 204(C؉6) of a Glu or a Lys residue, respectively. Moreover, the constitutive activity of the receptor was inhibited by Zn 2؉ binding in an engineered metal ion site, stabilizing an ␣-helical conformation of this loop segment. It is concluded that the high constitutive activity of the ghrelin receptor is dependent upon flexibility in the C-terminal segment of extracellular loop 2 and that mutations or ligand binding that constrains this segment and thereby conceivably the movements of transmembrane domain V relative to transmembrane domain III inhibits the high constitutive signaling.
Seven-transmembrane (7TM) 2 G-protein-coupled receptors constitute a large family of receptors that convey a multitude of extracellular signals into intracellular responses both through coupling to G-proteins and through other intracellular signaling pathways (1,2). Despite highly variable chemical properties of the ligands for 7TM receptors, it is believed that they all stabilize variations of a common overall active conformation (3)(4)(5)(6). Several highly conserved so-called microswitch residues within the transmembrane region are believed to be an integral component of the activation mechanism (7). In contrast, the loops connecting the transmembrane regions are characterized by a limited degree of structural conservation where only a disulfide bridges between the middle of extracellular loop 2 (ECL2) and the extracellular end of transmembrane domain (TM)-III is highly conserved. The conserved Cys in ECL2 divides this into an N-terminal ECL2a connecting TM-IV and -III and a C-terminal ECL2b connecting TM-III and TM-V (8).
To compare residues in ECL2 in different receptors, an index method has been proposed where the conserved Cys is the reference point, and the other residues are numbered relative to this position (9,10). Despite the lack of structural conservation of ECL2 among class A receptors, ECL2 has been suggested to serve an important role in the activation mechanism and in particular for stabilization of the inactive receptor conformation (11)(12)(13)(14). In the first described crystal structure of rhodopsin in the inactive conformation, the ECL2 formed a ␤-sheet domain, which reached deep down and immobilized the extracellular parts of the transmembrane segments (7,15). In the later crystal structures of 7TM receptors activated by soluble ligands, one of the major new observations was that the main ligand binding pocket was more open from the extracellular side where ECL2b only formed a partial lid over the ligand binding pocket between the transmembrane segments (7,16). Furthermore, the secondary structure of ECL2 differs significantly between the crystal structures obtained from the inactive conformations of the ␤-adrenergic, adenosine A2a, and CXCR4 receptors for example (17)(18)(19)(20). Even the often shorter and more structurally conserved ECL2b that connects TM-III and TM-V differs considerably. Interestingly, most antagonists trapped in the crystal structure of the inactive receptors are in direct contact with ECL2b, supporting the notion that ECL2 is important for stabilization of the inactive conformation (7,19,20).
The ghrelin receptor, which is known to be important for growth, appetite regulation, and fat accumulation (21), is characterized by very high ligand-independent signaling (22,23). The constitutive activity has recently been shown to represent an intrinsic property of the receptor protein rather than the result of influence from the cellular environment (24). Furthermore, the physiological importance of the constitutive activity has been verified for the ghrelin receptor by a naturally occurring human variant (a mutation) that selectively eliminates its constitutive activity without affecting the affinity, potency, or efficacy of the ghrelin hormone (25,26). Importantly, this mutation segregated with the development of short stature in two independent families (25). The mutation, which changed the natural Ala-204(Cϩ6) into a negatively charged Glu, was as indicated by the generic numbering located six amino acid residues after the conserved Cys in ECL2b (Fig. 1). In the present study, the molecular mechanism for the elimination of constitutive activity associated with the naturally occurring Glu for Ala mutation at position 204(Cϩ6) in ECL2b of the ghrelin receptor was characterized both by mutational analysis and by molecular modeling and computational analysis. We could not identify any interaction partner for the introduced Glu side chain; but surprisingly, we found that introduction of a positively charged residue at this position in ECL2b had a similar dampening effect on the constitutive signaling activity. The computational chemistry analysis indicated that both negatively or positively charged residues in position 204(Cϩ6) had a strong ␣-helix-inducing effect on the structure of ECL2b and engineering of a metal ion site in i and i ϩ 4 positions that upon metal ion binding stabilizes an ␣-helix, had the same inhibitory effect on the constitutive signaling. It is proposed that the free movement of TM-V relative to TM-III is crucial for the constitutive activity of the ghrelin receptor and that this is inhibited by conformational constraints such as extended helix formation.

EXPERIMENTAL PROCEDURES
Molecular Biology-Receptor mutants and chimeras were constructed using the PCR overlap extension method (27,28) with Pfu polymerase (Promega). The engineered cDNAs were cloned into the eukaryotic expression vector pCMV-Tag2B (Stratagene) encoding N-terminal FLAG tag epitope within appropriate restriction endonucleases sites, i.e. BamHI and EcoRI. All constructs were verified by DNA sequencing (Eurofins MWG Operon, Ebersberg, Germany).
Inositol Phosphate Turnover Assay-COS7 cells were cultured in Dulbecco's modified Eagle's medium 1885 supplemented with 10% fetal bovine serum (FBS), 2 mM L-glutamine, and antibiotics (penicillin, streptomycin). The cells were seeded at a density of 6⅐10 6 cells/175-cm 2 flask and transiently trans-fected using the calcium phosphate precipitation method (29) with chloroquine addition. The next day, 1⅐10 5 cells in 300 l of medium/well were seeded in 24-well plates and cultured overnight with 5 Ci/ml myo-[ 3 H]inositol (American Radiolabeled Chemicals). Prior to the assay, cells were washed with Hanks' balanced salt solution (HBSS), and 0.5 ml of Hanks' balanced salt solution supplemented with 10 mM LiCl was added and incubated for 30 min at 37°C. Cells were stimulated with specific ligands, i.e. ghrelin (Polypeptide, Hillerød, Denmark), [D-Arg 1 ,D-Phe 5 ,D-Trp 7,9 ,Leu 11 ]substance P (substance P analog), motilin (Bachem AG, Bubendorf, Switzerland), and ZnCl 2 (Sigma), for 45 min at 37°C and subsequently extracted with 1 ml of 10 mM formic acid/well in a 30-min incubation on ice. The Top, the naturally occurring ghrelin receptor variant, A204(Cϩ6)E, is located in the fragment of ECL2 (denoted as ECL2b (green circles)), which connects TM-III and TM-V via the conserved disulfide bridge. The other part of ECL2 connecting TM-III to TM-IV, called ECL2a, is marked in dark red circles. The residues highly conserved among class A 7TM receptors are denoted with white letters on black circles. Bottom, a model in which ECL2 is enlarged and every residue is depicted as single letter code. The localization of the Ala-204(Cϩ6) residue is shown by the red circle. The ghrelin receptor exhibits a high level of constitutive activity that is decreased by A204(Cϩ6)E mutation. This rare variant was described to result in a clinical phenotype characterized by short stature and potentially by obesity (25).
resulting supernatant was loaded on anion exchange resin (Dowex 1X8-200; Bio-Rad) to purify the negatively charged inositol phosphates. Glycophosphatidylinositol was discarded by washing the resin with 10 ml of 60 mM sodium formate, 5 mM sodium tetraborate decahydrate, and inositol phosphates were eluted with 3 ml of 1 M ammonium formate, 100 mM formic acid. Resin was regenerated with 3 ml of 3 M ammonium formate, 100 mM formic acid and washed with 10 ml of deionized water. The eluates were mixed with 10 ml of Gold Star scintillation mixture (Meridian), and ␤-emission was determined in a liquid scintillation counter (Beckman). Each data point was determined in duplicate.
Serum-responsive Element (SRE) Gene Reporter Assay-The SRE reporter gene assay was performed on HEK293 grown in Dulbecco's modified Eagle's medium 1966 supplemented with 10% FBS, GlutaMAX TM , and antibiotics (penicillin and streptomycin). The cells were seeded in white 96-well plates at 3.5⅐10 5 cells/well, cultured overnight, and transiently transfected with 5 ng of ghrelin receptor construct and 50 ng of inducible cis-reporter pSRE-Luc plasmids (PathDetect System, Stratagene) using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. One day after transfection, the cells were stimulated with the specific ligands for 5 h. Afterward, the cell culture medium was aspirated, and the cells were washed with PBS containing Ca 2ϩ and Mg 2ϩ . SRE-induced luciferase expression was detected by adding 100 l of bivalent ion-supplemented PBS followed by 100 l of chemiluminescent luciferase substrate, i.e. steadylite plus TM (PerkinElmer Life Sciences), per well. The luminescence was measured using an EnVision multilabel plate reader (Perkin-Elmer Life Sciences).
␤-Arrestin Mobilization Assay-Ligand-induced recruitment of ␤-arrestin was studied with the PathHunter ␤-Arrestin Assay obtained from DiscoveRx (Fremont, CA) utilizing enzyme fragment complementation technology. Ghrelin receptor cDNA constructs were C-terminally fussed with ProLink 1 tag (Discov-eRx, catalog number 93-0491) encoding a small fragment of ␤-galactosidase (␤-gal). The assay was performed in the U2OS enzyme acceptor-arrestin2 parental cell line (DiscoveRx, catalog number 93-0166) stably expressing ␤-arrestin fused with ␤-gal inactive form (enzyme acceptor) cultured in minimum essential medium Eagle (Sigma-Aldrich, catalog number M5650) supplemented with 10% FBS, penicillin-streptomycin-glutamine (Invitrogen, catalog number 10378-016), and 250 g/ml hygromycin B (Invitrogen, catalog number 10687-010). The cells were transfected with DNA using FuGENE 6 transfection reagent (Roche Diagnostics, catalog number 11814443001) according to the manufacturer's protocol. The cells were incubated overnight followed by exchange of the Opti-MEM I medium with minimum essential medium Eagle without antibiotics. The ␤-arrestin recruitment assay was performed 2 days after transfection where the cells were stimulated with ghrelin for 90 min. Upon ␤-arrestin-enzyme acceptor binding to the ProLink-tagged ghrelin receptor, complementation occurs between two ␤-gal fragments, forming an active enzyme. Therefore, ␤-arrestin recruitment to the receptor was detected as ␤-gal activity using the PathHunter detection kit (Dis-coveRx, catalog number 93-0164). Chemiluminescent substrate composed of Galacton Star Substrate, Emerald II Solution, and PathHunter Cell Assay Buffer in a ratio of 1:5:19, respectively, was prepared and added to the cells (50 l/well). The luminescence signal was determined after a 60-min incubation at ambient temperature using a TopCount microplate counter (PerkinElmer Life Sciences).
Data Analysis-Dose-response curves were fitted with a nonlinear regression, variable slope equation using GraphPad Prism 5.0 (GraphPad Software). Curves fitted in a single experiment were normalized to the background signaling measured for mock-transfected cells (0%) and ghrelin receptor wild-type maximal stimulation by ghrelin (100%) in the particular assay. The normalized data points were merged and fitted to sum curves. Each sum curve was calculated from at least three independent experiments.
Expression Analysis-Cell surface expression of all studied receptor variants was assessed using ELISA. COS7 cells were transiently transfected with the N-terminally FLAG-tagged receptor constructs as described for the inositol phosphate turnover assay. One day after transfection, the cells were seeded in 96-well plate at 3.0 -3.5⅐10 5 cells/well and cultured overnight. The cells were fixed with 3.7% formaldehyde in PBS and washed thoroughly with PBS followed by blocking of unspecific binding sites with 3% nonfat dry milk in PBS (blocking buffer). Anti-FLAG antibody (Sigma, catalog number F1804) diluted 1:1000 in blocking buffer and anti-mouse IgG horseradish peroxidase (HRP)-linked anybody (Thermo Scientific, catalog number 32430) diluted 1:1250 in 1.5% nonfat dry milk in PBS were used as primary and secondary antibodies, respectively. The HRP activity was determined using a chromogenic substrate, tetramethylbenzidine (Kem-En-Tec, catalog number 4380). The reaction was terminated with 0.5 M sulfuric acid, and the color reaction product was transferred into a clear 96-well plate where an absorbance at 450 nm was determined. The mean absorbance for mock-transfected cells was subtracted as a background cutoff. The expression level was estimated as a ratio between absorbance mean values for the receptor construct and the wild-type receptor measured in the same assay.
The preliminary comparative models were subsequently refined to construct complete ghrelin receptor models including intra-and extra cellular loops using Rosetta (version 3.2.1 originally developed to address the protein folding problem and later extended to focus on protein designs, protein folding mechanisms, protein-protein interactions, and docking) (31)(32)(33)(34)(35). In brief, the refinement protocol involves 1) generation and optimization of the extracellular loops using a combination of the cyclic coordinate descent (34) and kinematic closure (35) application and 2) repacking of backbone and side chains using the relax protocol. During model refinement, a disulfide bridge between Cys-116(III:01) and Cys-198 in the ELC2 was defined; otherwise, loops were modeled ab initio. Fragment files for the individual ghrelin receptor variants were obtained from the Robetta server. Detailed explanations of the cyclic coordinate descent and kinematic closure algorithms can be found elsewhere (35,36). In brief, the goal of both algorithms is to explore the conformational space of structurally variable regions (loop) initially using a centroid representation of protein side chains and explicit backbone representation followed by a higher resolution search using all atoms and hydrogen. In the initial stage, loops are generated by a fragment buildup/insertion Monte Carlo algorithm where cyclic coordinate descent is used to close the loop at the end of the simulation. In each step of this Monte Carlo cycle, kinematic closure is used to refine the structures where all residues within the neighbor distance of a loop are repacked and then subjected to side chain rotamer trials. The backbone and side chains of all loop residues and neighboring residues are then followed by a line minimization of the loop / torsions, and the final conformation is accepted/rejected by the Metropolis criterion using the full-atom Rosetta scoring function. In the present study, we generated a total of 6000 loop models for the wild type and each of the A204E, A204K, and G208A ghrelin receptor variants. The resulting ensemble of structures for the individual receptor variants was clustered with respect to the structure of the second extracellular loop using a Rosetta cluster routine and a 1.5-Å root mean square deviation cluster threshold. Structural features of the ECL2b such as secondary structure, spatial location of mutated residues, population, and scoring energy are compared between the different receptor variants and presented under "Results."

RESULTS
The main signaling pathway for the ghrelin receptor is phospholipase C activation through G␣ q . Thus, to characterize the effect of substitutions in ECL2, we transiently expressed the wild-type receptor and variants of this receptor in COS7 cells and measured cell surface expression by ELISA and signaling properties mainly by determination of inositol phosphate accumulation. An important characteristic of the ghrelin receptor is its high constitutive activity, i.e. 45% of maximal ghrelin-induced stimulation (Table 1), which makes the receptor particularly suited for studying mutational effects on receptor activation mechanisms in a ligand-independent manner. The impact of the substitutions on constitutive and ligand-induced signaling was expressed relative to the signaling induced by ghrelin on the wild-type ghrelin receptor ( Table 1).

Introduction of Glu in Position 204(Cϩ6) in ECL2b
Decreases Constitutive Activity-Based on the original, natural Ala-204(Cϩ6) to Glu variant, we probed effects of the introduction of a Glu in the neighboring positions. All of these mutants were well expressed at the cell surface (Table 1 and Fig. 2B). As

TABLE 1 Mutational analysis of ghrelin receptor mutants within ECL2b
The ghrelin receptor constructs were expressed in transiently transfected COS7 cells. Pharmacological properties of each variant were assessed by inositol trisphosphate turnover assay. For each receptor mutant, both constitutive activity and ghrelin-induced efficacy (E max ) were normalized to maximal ghrelin-induced stimulation of wild-type receptor (100%) and basal inositol phosphate accumulation in mock-transfected cells (0%) performed in parallel in the same assay. Ghrelin potency (EC 50 ) was determined from the dose-response curve and compared with the wild type-specific value, indicating the -fold shift (F mut ) caused by the structural change in the receptor due to the specific point mutation. Expression of each mutation was assessed by cell surface ELISA and is given as a fraction of the wild-type receptor level. Values are shown as ϮS.E.  2D). In contrast, Glu substitution of the preceding Phe-203(Cϩ5) or the positively charged Arg-206(Cϩ8) located two residues after Ala-204(Cϩ6) had no or a minimal effect on both constitutive and ghrelin-induced signaling (Fig. 2, C and F). However, substitution of the neighboring Val-205(Cϩ7) with Glu did not eliminate but did reduce the constitutive signaling to ϳ50% of that observed in the wild-type receptor and as in position 204(Cϩ6) without affecting the maximal stimulatory effect of ghrelin (Fig. 2E). Thus, in respect to the dampening effect on constitutive signaling, position 204(Cϩ6) in ECL2b is particularly susceptible to the introduction of a negatively charged Glu residue, but to a lesser degree, this is also observed in the following position, 205(Cϩ7).

Construct
Search for a Charge-Charge Partner for Glu-204(Cϩ6)-One obvious explanation for the impact of the Ala to Glu substitution at position 204(Cϩ6) in ECL2b would be that the acidic side chain of the Glu forms a charge-charge interaction with a positively charged residue in the proximity. To address this, Ala substitution of four positively charged residues expected to be found close to Ala-204(Cϩ6), Lys-III:02(3.25), Arg-199(Cϩ1), Arg-206(Cϩ8), and Lys-VI:25(6.60), was performed (Fig. 3). Ala substitution of these positively charged residues did not in itself affect the constitutive activity of the ghrelin receptor, and importantly, it did not rescue the inhibitory effect on receptor signaling induced by the Glu substitution in position 204(Cϩ6) as judged by a series of double mutants (Table 2 and Fig. 3). Thus, we did not find any indication that the effect of Glu in position 204(Cϩ6) is due to a charge-charge interaction with residues in the local environment.
Because charge residues appeared to affect receptor signaling, we extended the systematic mutational analysis to the six potentially charged residues located in ECL2a, i.e. the loop that connects the extracellular pole of TM-IV with that of TM-III: i.e. Asp-191, Asp-194, Glu-185, Glu-187, Glu-197, and His-186. However, functional analysis of uncharged substitutions introduced in these positions revealed that none of them had any impact on either the constitutive activity or the ghrelin-induced activation of the receptor (Table 3).
Substituting ECL2b of the Ghrelin Receptor with That of the Motilin Receptor-In the ghrelin receptor subfamily of receptors, the neurotensin 2 receptor and GPR39 both display a high degree of ligand-independent signaling like the ghrelin receptor, whereas its closest relative, the motilin receptor, is totally silent in respect to constitutive signaling (23). We therefore constructed chimeras of both the ghrelin and motilin receptors where ECL2b was replaced with the corresponding sequence from the counterpart (Fig. 4). The motilin receptor was not affected at all by the introduction of ECL2b from the ghrelin receptor, and in particular, it did not become constitutively active (Fig. 4B). In contrast, introduction of ECL2b from the motilin receptor in the ghrelin receptor totally eliminated its high constitutive signaling, further supporting the notion that the structure of ECL2b is very important for the ligand-independent signaling of the ghrelin receptor (Fig. 4A). None of the chimeras gained the ability to be activated by the counterpart endogenous agonist (data not shown) or lost the ability to be activated by the endogenous ligand (Fig. 4).

TABLE 2 Mutational analysis of potential interaction partners for A204(C؉6)E mutant
Point mutations of charged residues as the potential interaction partners for Glu-204(Cϩ6) alone and in combination with A204(Cϩ6)E substitution were transiently expressed in COS7 cells. The expression of each construct was assessed by cell surface ELISA and is given as a fraction of wild-type receptor level. The effect of mutation on ghrelin receptor functional properties was assessed by inositol trisphosphate turnover assay. For each receptor mutant, both constitutive activity and ghrelin-induced efficacy (E max ) were normalized to maximal ghrelin-induced stimulation of wild-type receptor (100%) and basal inositol phosphate accumulation in mock-transfected cells (0%) performed in parallel in the same assay. Ghrelin potency (EC 50 ) was determined, and its -fold shift compared with wild type (F mut ) due to the structural change in the receptor is indicated. Results are given as ϮS.E. 8-fold to the right (EC 50 ϭ 2.6 versus 0.34 nM in the wild-type receptor) (Fig. 5C and Table 1). In fact, the constitutive signaling of the ghrelin receptor was decreased by introduction of nearly all other residues in position 204(Cϩ6) other than the natural Ala; albeit the effect of Val for example was very little (Fig. 5D and Table 1). In contrast, the inhibitory effect on the constitutive signaling was larger and significant with the larger aromatic Phe and in particular with the large aliphatic Leu residue ( Fig. 5E and Table 1). Surprisingly, substitution with the positively charged Lys residue resulted in a signaling phenotype similar to that observed with the negatively charged Glu residue, i.e. elimination of the constitutive activity (Fig. 5F). Thus, substitution of the native Ala at position 204(Cϩ6) in ECL2b not only with negatively charged but also with positively

Mutational analysis of ghrelin receptor mutants within ECL2a
The ghrelin receptor constructs were expressed in transiently transfected COS7 cells. Pharmacological properties of each variant were assessed by inositol trisphosphate turnover assay. For each receptor mutant, both constitutive activity and ghrelin-induced efficacy (E max ) were normalized to maximal ghrelin-induced stimulation of wild-type receptor (100%) and basal inositol phosphate accumulation in mock-transfected cells (0%) performed in parallel in the same assay. Ghrelin potency (EC 50 ) was determined from the dose-response curve and compared with the wild type-specific value, indicating the -fold shift (F mut ) caused by the structural change in the receptor due to the specific point mutation. Expression of each mutation was assessed by cell surface ELISA and is given as a fraction of wild-type receptor level. Values are shown as ϮS.E.  charged residues strongly decreases or even eliminates the constitutive signaling of the ghrelin receptor. Biased Signaling Introduced by Substitutions of Ala-204(Cϩ6)-Although G␣ q is generally accepted as the main signaling pathway for the ghrelin receptor, it has recently been advocated that activation of SRE-induced gene transcription through G␣ 12/13 -driven activation of Rho A kinase is a signaling pathway of particular importance in relation to feeding behav-ior (37). In addition, other signaling cascades may be activated through ␤-arrestin coupling. As observed for G␣ q signaling, the ghrelin receptor displays 33 and 41% constitutive activity though SRE-induced transcriptional activation and arrestin recruitment as determined in transiently transfected HEK293 and U2OS cells, respectively ( Fig. 6 and Table 4).

Construct
The SRE-induced signaling was in many ways affected in a rather similar way as that observed for G␣ q signaling by the

TABLE 4 Mutational analysis of ghrelin receptor mutants in SRE-luciferase reporter assay and ␤-arrestin mobilization assays
Wild-type and mutant ghrelin receptors were transiently expressed in HEK293 and U2OS cells for their functional assessment in SRE-luciferase (luc) reporter and ␤-arrestin mobilization assays, respectively. The ghrelin dose-response curves were analyzed in respect to the constitutive activity of each variant, and efficacy (E max ) is given as the ratio of WT maximal stimulation level. A comparison of ghrelin potency (EC 50 ) in certain mutant and wild-type receptors determined as the -fold shift (F mut ) indicates the impact of the side chain substitution on receptor structure. Results are given as ϮS.E. substitutions in ECL2b, albeit in several cases to a larger degree. That is, substitutions of Ala-204(Cϩ6) with the charged residue Glu, Asp, or Lys resulted in complete loss of constitutive SRE signaling but decreased the maximal achieved efficacy induced by ghrelin, although the efficacy measured as an increase above the basal level was rather similar to that observed in the wild-type receptor ( Fig. 6A and Table 4). Substitution with a non-charged residue, i.e. Phe or Val, inhibited the constitutive signaling more seriously than observed through G␣ q , i.e. down to 3.5-6.7%, whereas the maximal efficacy of ghrelin was almost comparable with the efficacy observed in the wild-type receptor ( Fig. 6A and Table 4). In respect to ␤-arrestin mobilization, not only the constitutive but also the ghrelin-induced arrestin mobilization was eliminated by substitutions with charged residues of Ala-204(Cϩ6) (Fig. 6B and Table 4). Substitutions with Phe or Val had almost the same effect, although a small response to ghrelin was observed in the Phe-204(Cϩ6) mutant form (Fig. 6B and Table 4). In summary, introduction of charged residues at position Ala-204(Cϩ6) eliminates the constitutive signaling of the ghrelin receptor through all measured pathways but makes ghrelin act as a biased agonist that signals almost normally through the G␣ q and G␣ 12/13 pathways without being able to mobilize arrestin.

Ghrelin receptor variant
Computational Chemical Analysis of the Secondary Structure of ECL2b-As no charge-charge partner could be identified for Glu-204(Cϩ6) and because introduction of a positively charged Lys residue had a similar effect as the negatively charged Glu residue, we decided to analyze the effects of these substitutions on the secondary structure of ECL2b. ECL2b varies greatly both in length and secondary structure among the available high resolution x-ray structures of the receptors (17-20, 38). Thus, to investigate the potential structural differences of the ECL2b variants, we generated ensembles of complete models of the individual receptor variants using the de novo structure prediction and protein design package Rosetta (31)(32)(33)(34)(35). Because of the great variance in loop structures in the published crystal structures, we made the Rosetta-based structure predictions unbiased by previously published structures and only used the positions of the transmembrane segments guided by the published crystal structures to generate reliable models of both ghrelin receptors. As described under "Experimental Procedures," an exhaustive set of 6000 ECL2b models was generated for the wild-type receptor and the A204E and A204K receptor variants, respectively. The resulting ensembles of extracellular loop 2b conformations were analyzed using conformational clustering, and the best scored models as defined by the Rosetta full-atom scoring function were selected as representative receptor models for the individual clusters (Fig. 7).
For the wild-type receptor and for the A204(Cϩ6)E and A204(Cϩ6)K receptor variants, we identified eight, four, and six different clusters of ECL2b structures, respectively (Fig. 7). The conformational energies for the most populated clusters were in general comparable (Table 5). In both the wild-type ghrelin receptor and the mutant forms of this receptor, either random coil or ␣-helical structures of variable length were observed for ECL2b. In the wild-type receptor, four clusters of random coil structures were observed that together comprised a total of 24.1% of all the structures (Fig. 7A). In the four other clusters, some degree of ␣-helical structure was observed of which the most populated cluster corresponding to 49.2% of the receptor population displayed one and a half ␣-helical turns in the "N-terminal" part of the ECL2b sequence, i.e. corresponding to residues from Pro-200(Cϩ2) to Arg-206(Cϩ8) (Fig. 7A). In the wild-type receptor, a cluster corresponding to 15.5% of the receptor population displayed an extended ␣-helix corresponding to almost the full length of ECL2b (Fig. 7A). In contrast, in the A204(Cϩ6)E variant of the ghrelin receptor, an ␣-helix extending all the way from Pro-200(Cϩ2) to Lys-209(V: 01) was observed in two of the four clusters corresponding to a total of 88.7% of the population of ECL2b structures (Fig. 7B). Similarly, in the A204(Cϩ6)K mutant, extended ␣-helical structures were observed in three of the six clusters of ECL2b structures corresponding to a total of 82.4% of the receptor structures (68.9 ϩ 13.5%) (Fig. 7C). In all the different clusters that were characterized by extended helical structures, the side chain of either Glu-204(Cϩ6) or Lys-204(Cϩ6) pointed "upward," i.e. toward the extracellular solvent. In conclusion, the cluster analysis of the Rosetta modeling of the ghrelin receptor indicates that, in contrast to the wild-type receptor, ECL2b of the Glu-204(Cϩ6) and Lys-204(Cϩ6) mutant forms of the receptor has a very high propensity to adopt an ␣-helical conformation spanning from the extracellular end of TM-III to TM-V.
Introduction of ␣-Helix-stabilizing Metal Ion Site in ECL2b-If the extended ␣-helical structure of ECL2b observed in both the Glu-204(Cϩ6) and Lys-204(Cϩ6) mutant forms of the ghrelin receptor is involved in the elimination of its high constitutive activity, then it should be possible to obtain a similar phenotype by stabilizing such an ␣-helical structure in ECL2b by metal ion site engineering through introduction of His residues in i and i ϩ 4 positions, which are ideal for an intrahelical zinc site, as done previously at the extracellular pole of TM-V in the NK1 and opioid receptors (39,40).
Accordingly, His residues were introduced corresponding to positions 205(Cϩ7) and 209(V:01) in ECL2b of the ghrelin receptor (Fig. 8A). The expression level of the double His mutant was decreased to 60% of the wild-type receptor (Fig.  8B), and the constitutive signaling was accordingly decreased to ϳ10% of the maximal ghrelin-induced signaling (Fig. 8C). However, the ghrelin-induced activation was not affected with respect to either potency or efficacy (Fig. 8C). In the His-205(Cϩ7),His-209(V:01) mutant form Zn 2ϩ acted as an antagonist by inhibiting the signaling induced by 1 nM ghrelin with a potency of 2.9 M (Fig. 8C, open squares). In comparison, the signaling induced by 1 nM ghrelin was not inhibited by Zn 2ϩ in the wild-type ghrelin receptor (Fig. 8C, dotted line with open  circles). Importantly, Zn 2ϩ also acted as an inverse agonist in the His-205(Cϩ7),His-209(V:01) mutant form as it inhibited the basal, ligand-independent signaling with an IC 50 value of 0.17 M (Fig. 8D, open squares), whereas zinc ions had no effect on the basal signaling of the wild-type receptor (Fig. 8D, dotted  line with open circles). It is concluded that stabilization of a helical structure in ECL2 by chelation of Zn 2ϩ in an intrahelical metal ion site inhibits both spontaneous and ghrelin-induced receptor signaling.
Substitution of the Helix-breaking Gly-208(Cϩ10) at the End of ECL2b-In view of the apparent importance of the secondary structure of ECL2b for the constitutive signaling of the ghrelin receptor, we returned to the potential helix-breaking property of Gly-208(Cϩ10) located at the C-terminal part of the ECL2b (Fig. 9A). To facilitate ␣-helix formation in this region, Gly-208 was substituted with Ala, which resulted in a significant reduc-

TABLE 5
Characteristics of clusters of the ghrelin receptor models generated using Rosetta package Rosetta score energy units are derived from empirical scoring function which has been shown to correlate to free energy.

Ghrelin receptor variant
Cluster no. tion in the constitutive activity from 45 to 11% (Fig. 9B). The potency for ghrelin was not affected, but the E max was reduced, although the ghrelin-induced increment in efficacy was similar to that observed in the wild-type receptor (Fig. 9B). Analysis of the G208(Cϩ10)A mutant by Rosetta simulations performed as described above revealed nine clusters of ECL2b conformations that could be further clustered into the three types of conformations shown in Fig. 9C. The most populated single cluster representing 35.5% of the structures displayed a long ␣-helix extending over the whole ECL2b similar to those observed in the majority of conformations in the Glu-204(Cϩ6) and Lys-204(Cϩ6) mutant forms (Fig. 7, B and C). As described above, this type of extended ␣-helix was only observed in 15.5% of the wild-type receptors (Fig. 7A). Thus, removal of the potential helix-breaking Gly-208(Cϩ10) results in decreased constitutive activity and an increased propensity to form a long ␣-helix throughout ECL2b.

DISCUSSION
The naturally occurring variant of the ghrelin receptor, which is associated with short stature in humans, involves substitution of the native small, aliphatic Ala in position 204(Cϩ6) with a large, negatively charged Glu residue. The considerable difference in the physiochemical properties of this substitution was expected to result in some kind of specific interaction, for example a charge-charge interaction, that could be responsible for the observed selective loss of constitutive signaling believed to be responsible for the clinical phenotype (25,26). However, the rather extensive mutational analysis of the entire ECL2b, selected residues in ECL2a as well as residues in the upper parts of the transmembrane segments, failed to identify any interaction partner for the Glu variant. In contrast, it was found that also substitutions at other positions of ECL2b affected the constitutive signaling of the receptor, although the molecular phenotype with selective elimination of constitutive activity combined with normal ghrelin-induced signaling was only obtained with substitutions in position 204(Cϩ6). The notion that alterations of the secondary structure of ECL2b were responsible for the changes in receptor signaling was precipitated by the highly surprising observation that a positively charged Lys residue could mimic the effect of the negatively charged Glu residue at this position. This notion was confirmed by computational chemical analysis, which demonstrated that substitutions eliminating constitutive signaling were associated with a strongly increased propensity of ECL2b to form an extended ␣-helix. This was further substantiated by the engineering of an ␣-helixstabilizing metal ion site into ECL2b that decreased receptor signaling upon zinc binding. In summary, we found that structural constraint in ECL2b decreasing the flexibility of TM-V relative to TM-III decreases the constitutive activity of the ghrelin receptor, indicating that free movement of TM-V relative to TM-III is required for the high constitutive activity of this receptor. Interestingly, the fact that these substitutions in general did not affect the potency and efficacy of the endogenous agonist (ghrelin) indicates that ghrelin stabilizes an active conformation independently of ␣-helix constraints in ECL2b. ECL2 X-ray Structures-ECL2 varies considerably among class A 7TM receptors both in respect to length and primary structure (8,41), and the many recently characterized x-ray structures have revealed that the secondary and tertiary structures of ECL2 are also highly variable (7,9,16,42). The length and structure of ECL2a especially vary from ␣-helical and random coil to forming a ␤-sheet and in some cases having an internal disulfide bridges or in some receptors being joined to ECL3 or even to the N-terminal segment by disulfide bridges (9,18,43,44). However, ECL2b, which bridges from the highly conserved disulfide bridge at the extracellular end of TM-III to TM-V, is often shorter, forming the actual lid over the main ligand binding pocket, and is in many cases directly involved in ligand binding (7,9). In the available x-ray structures, ECL2b is in most cases basically an extended strand connecting TM-III and TM-V (17,18,38). However, in the A2a receptor and the M3 muscarinic receptor, ECL2b is longer, and the central part folds into an ␣-helical structure (20,45). Thus, the structure of ECL2b also varies between receptor types, but rather similar folds have been observed in closely related receptors such as the ␤ 1 -and ␤ 2 -adrenergic receptors and theand -opioid receptors (17,18,46,47). In the two different receptors, the ␤ 2 -adrenergic receptor and the A2a receptor, for which x-ray structures of both an active and an inactive structure have been described, no differences have been observed in the extracellular domains. It is nevertheless assumed that the extracellular loops of 7TMs are relatively adaptable or flexible and that they may change structure for example upon ligand binding and upon mutational substitutions like the ECL2b variants of the ghrelin receptor.
Computational Chemical Analysis of ECL2b Structure in the Ghrelin Receptor-No x-ray structure is yet available for the ghrelin receptor; therefore, molecular modeling is consequently required for studying its structure. However, even with the multiple high resolution structures available today and with protein structure prediction tools such as Rosetta, which has a strong track record in the field in general, it is not yet possible to reliably predict the structures of for example extracellular loops, which are long and structurally divergent (48). Nevertheless, it is possible to accurately refine and model loop structures ab initio if the loop length is 12 residues or shorter (36). Accordingly, reliable modeling of ECL2b in itself should be possible in many cases, but it is still seriously hampered by the fact that this loop interacts with the other often longer loops and with the often ill defined N-terminal extension, complicating matters.
Thus, in the present study, we instead performed an alternative computational chemical analysis of the presumed structures of ECL2b of the wild-type ghrelin receptor and the Glu-204(Cϩ6) and Lys-204(Cϩ6) mutant forms. Importantly, we did not aim at modeling a specific single predicted or "optimized" loop conformation for the receptors. Instead, we performed exhaustive extracellular loop simulations using the de novo structure prediction and protein design package Rosetta to generate a large set of likely loop structures. Subsequently, this canonical ensemble of structures was subjected to advanced cluster analysis to approximate the expected variability and distribution of actual molecular shapes. Through this approach, we obtained evidence that the substitutions in posi-tion 204(Cϩ6) dramatically changed the apparent propensity of ECL2b to adopt an extended ␣-helical structure from approximately 15% in the wild type to 70% in the Lys-204(Cϩ6) mutant and almost 90% in the original naturally occurring variant Glu-204(Cϩ6), the mutant displaying the most severe signaling phenotype, i.e. totally eliminated constitutive activity.
ECL2 as a Regulator of Receptor Signaling-Our overall analysis indicates that introduction of charged residues in particular at position 204(Cϩ6) induces the formation of a long, rigid ␣-helical structure in ECL2b, which conceivably constrains the free movement of TM-V relative to TM-III, and it is proposed that this change in secondary structure is responsible for the observed decrease or elimination of the high ligand-independent receptor signaling. Our data indicate an important function of ECL2 in the regulation of receptor signaling. Despite the low degree of sequence homology in ECL2 within the family of 7TM receptors, it has been suggested previously that this loop could be an important regulator of receptor activation (9). In the C5a receptor, random saturation mutagenesis identified several mutations that generated constitutive activity, suggesting ECL2 as a negative regulator of receptor activation (11). This observation is consistent with solid state NMR data of rhodopsin in which ECL2 is displaced from the retinal binding pocket upon light-induced activation (49). Similarly, NMR analysis of the ␤-adrenergic receptor suggests that essential movements occur in ECL2 during receptor activation, and CD analysis in the 5HT4a receptor suggests that conformational changes occur around the conserved disulfide bridge upon activation (13,14). Such biochemical and in particular biophysical studies illustrate the notion that although x-ray structures provide essential basic information about structure we need many other types of studies to understand function and that apparently rather divergent structures such as extracellular loops may still be important for function.
TM-V to TM-III Interface and Movements in Receptor Activation-Biochemical and biophysical studies and a recent series of x-ray structures have predicted movements at the intracellular side of the receptor during activation (50 -52). In contrast to the large changes occurring at the intracellular face of the receptor where the transmembrane segments move apart to give room for the G-protein, more subtle movements have been observed in the extracellular parts of the receptors that even differ from receptor to receptor (13,16,(53)(54)(55). However, in the context of the present study, which points to the importance of free movement of TM-V relative to TM-III for at least the spontaneous receptor activation, the conformational changes occurring in the ␤ 2 -adrenergic receptor may be particularly relevant. Thus, binding of the ␤-adrenergic agonist and the formation of the important hydrogen bonds with Ser residues located at the extracellular end of TM-V are associated with an approximately 2-Å inward movement of this part of TM-V (5). This could indicate that an inward movement of the extracellular end of TM-V at least in some receptors is associated with receptor activation, which is in agreement with the data of the present study suggesting that constraining the free movement of TM-V relative to TM-III inhibits spontaneous receptor signaling (7,39,56,57).
The interface between TM-III and TM-V (connected by ECL2b) has in various receptors been suggested to be important for the activation process based on mutational studies. Because of the presence of the conserved Pro-V: 16(5.50), the backbone carbonyl of the residue in position V:12(5.46) is not bound in the intrahelical hydrogen bond network, and this free carbonyl group, which for example in rhodopsin points almost "horizontally" away from TM-V toward TM-III, appears to make important interactions with residues in TM-III. For example, in the histamine H1 receptor, an interaction between Thr-III:13 (3.37) and the free backbone carbonyl of Asn-V:12(5.46) has been shown to stabilize the receptor in an inactive conformation (58). In rhodopsin, Glu-III:13(3.37) forms a similar backbone interaction with TM-V that contributes to keeping the receptor in the inactive conformation (15,59).
Interestingly, a small subfamily of class A 7TM receptors, i.e. the melanocortin receptors, the cannabinoid receptors, and the sphingosine 1-phosphate receptors, lack the otherwise highly conserved disulfide bridge between the "middle" of ECL2 and the extracellular end of TM-III (60 -63). Therefore, these receptors lack an ECL2b loop altogether, and there are no structural constraints between the extracellular end of TM-V and TM-III. In these receptors, ECL2 directly connects TM-IV and TM-V and is either very short or functionally shortened by an intraloop disulfide bridge as recently demonstrated in the x-ray structure of the sphingosine 1-phosphate 1 receptor (44). One notion of the present study is that the lack of conformational constraint between TM-V and TM-III is associated with high constitutive activity as in the wild-type ghrelin receptor. This is indeed the case because a number of receptors lacking ECL2b, such as the CB1, MC4, and sphingosine 1-phosphate 1 receptors, are highly constitutively active (44,(61)(62)(63).
Inhibiting Receptor Signaling by Strengthening the Helical Propensity of TM-V-In the present study, it is the helical propensity of ECL2b and in particular of the C-terminal segment of ECL2b at the interface between ECL2b and TM-V that is increased either through introduction of a charged residue at position Cϩ6 or through engineering of an i and i ϩ 4 metal ion site into ECL2b that silenced the constitutive activity of the receptor. Previously, the flexibility at the top of TM-V has been constrained by strengthening the ␣-helix structure by engineering of a metal ion site in i and i ϩ 4 positions at the very end of TM-V between positions V:01(5.35) and V:05(5.39) in several different receptors, i.e. NK1, ORF74, and the ghrelin receptor (22,40,64). In these receptors, it was also shown that the ligand-induced activation was prevented by the metal ion site, which is expected because the site is engineered into the presumed binding pocket of the receptor. Importantly, in ORF74 and in the ghrelin receptor, the ligand-independent high constitutive activity was also decreased by zinc binding, i.e. by strengthening of the ␣-helical structure at the extracellular end of TM-V (22,40). These data together with the data from the present study indicate that constraint of the flexibility around the extracellular end of TM-V has a strong negative impact on the constitutive receptor signaling in particular.
Lack of Involvement of ECL2 in Ghrelin Receptor Ligand Binding-The extracellular loops and in particular ECL2b have been described previously to be involved in both orthosteric and allosteric ligand binding (65)(66)(67)(68)(69)(70)(71). In particular, in receptors for peptides, glycoprotein hormones, and chemokines, major interaction sites for the endogenous ligands have been described to be present in the extracellular loops and most often in ECL2 (30,68,72,73). Previously, an aromatic residue in position Cϩ4 in other 7TM receptors has been observed to modulate the ligand-induced activation, suggesting a conserved role in the activation mechanism (10). However, in the ghrelin receptor, substitution of Phe (located in the neighboring position Phe-203(Cϩ5)) did not affect the ghrelin-induced signaling or ligand-independent signaling. In the ghrelin receptor where the 28-amino acid-long endogenous agonist has only been described to be dependent on only a few residues located in the upper part of the major ligand binding pocket, i.e. Glu-III:09(3.33) and Phe-VI:23(5.58) and to a lesser degree Glu-197 located in ECL2a just before the Cys residue forming the conserved disulfide bridge to TM-III (Fig. 1). Consequently, we had expected that the ghrelin-induced signaling would be affected by several of the experiments performed in this study in which the entire length of ECL2b was carefully studied by several substitutions in each position and all the charged residues in the rest of ECL2 were substituted with non-charged residues. However, the ghrelin-induced signaling was affected to a surprisingly low degree in respect to both potency and efficacy when measuring G␣ q -mediated inositol phosphate accumulation. In contrast, when we measured ␤-arrestin recruitment, the ghrelin-induced efficacy was strongly decreased. This indicates that the ECL2b influences activation in a pathway-biased manner. Because arrestin mobilization not only prevents G-proteincoupled signaling but also mediates novel signaling pathway (2), it is possible that this molecular phenotype is actually of clinical importance (25).
In conclusion, we have demonstrated a critical role of the secondary structure of ECL2 in regulation of the constitutive activity of the ghrelin receptor. Both modeling of ECL2b with charged residues introduced in position Cϩ6 and engineering of a metal ion site suggest that formation of ␣-helix in the ECL2b will silence the high basal signaling of the ghrelin receptor. In contrast, the ghrelin-induced activation was unaffected when evaluated as G␣ q coupling, but when measuring ␤-arrestin, the efficacy of ghrelin was strongly reduced.