Tyrosine 308 Is Necessary for Ligand-directed Gs Protein-biased Signaling of β2-Adrenoceptor*

Background: Ligand-specific receptor signaling is often referred to as functional selectivity or biased agonism. Results: Single amino acid substitution on β2-adrenoreceptor (Y308F) converts a ligand-specific signaling from Gs-biased to promiscuous Gs and Gi dual signaling. Conclusion: Specific ligand-receptor interaction results in receptor conformation(s) sufficient to convey biased signaling. Significance: Our work reveals a molecular mechanism for biased agonism. Interaction of a given G protein-coupled receptor to multiple different G proteins is a widespread phenomenon. For instance, β2-adrenoceptor (β2-AR) couples dually to Gs and Gi proteins. Previous studies have shown that cAMP-dependent protein kinase (PKA)-mediated phosphorylation of β2-AR causes a switch in receptor coupling from Gs to Gi. More recent studies have demonstrated that phosphorylation of β2-AR by G protein-coupled receptor kinases, particularly GRK2, markedly enhances the Gi coupling. We have previously shown that although most β2-AR agonists cause both Gs and Gi activation, (R,R′)-fenoterol preferentially activates β2-AR-Gs signaling. However, the structural basis for this functional selectivity remains elusive. Here, using docking simulation and site-directed mutagenesis, we defined Tyr-308 as the key amino acid residue on β2-AR essential for Gs-biased signaling. Following stimulation with a β2-AR-Gs-biased agonist (R,R′)-4′-aminofenoterol, the Gi disruptor pertussis toxin produced no effects on the receptor-mediated ERK phosphorylation in HEK293 cells nor on the contractile response in cardiomyocytes expressing the wild-type β2-AR. Interestingly, Y308F substitution on β2-AR enabled (R,R′)-4′-aminofenoterol to activate Gi and to produce these responses in a pertussis toxin-sensitive manner without altering β2-AR phosphorylation by PKA or G protein-coupled receptor kinases. These results indicate that, in addition to the phosphorylation status, the intrinsic structural feature of β2-AR plays a crucial role in the receptor coupling selectivity to G proteins. We conclude that specific interactions between the ligand and the Tyr-308 residue of β2-AR stabilize receptor conformations favoring the receptor-Gs protein coupling and subsequently result in Gs-biased agonism.

␤ 2 -AR phosphorylation by PKA or G protein-coupled receptor kinases. These results indicate that, in addition to the phosphorylation status, the intrinsic structural feature of ␤ 2 -AR plays a crucial role in the receptor coupling selectivity to G proteins. We conclude that specific interactions between the ligand and the Tyr-308 residue of ␤ 2 -AR stabilize receptor conformations favoring the receptor-G s protein coupling and subsequently result in G s -biased agonism.
Increasing evidence has accumulated over the past decade indicating that a G protein-coupled receptor (GPCR) 3 does not respond similarly to all agonist ligands. Ligands can initiate multiple cascades of intracellular reactions that can be mediated by G proteins or be G protein-independent. The term biased agonism initially referred to the ability of a ligand to selectively activate either a G protein-mediated event, such as stimulation of adenylyl cyclase, or activation of the G proteinindependent noncanonical ␤-arrestin-dependent signal transduction pathway (1). The emerging paradigm of biased agonism or functional selectivity suggests that binding of one ligand can stabilize receptor conformation(s) preferentially favoring recognition by a given set of signaling proteins or pathways on the intracellular side, although another ligand stabilizes a receptor state that is preferred by a different set of signaling proteins. In this manner, ligands can trigger qualitatively distinct signaling events in the cell (1)(2)(3)(4)(5)(6)(7)(8).
␤-Adrenergic receptors (␤-AR) are G s -coupled GPCRs and in fact ␤ 1 -AR couples only to G s . For ␤ 2 -AR, the prototypical member of the GPCR family, studies (9,10) have shown that agonist binding can activate both G s and G i . Our previous study has shown that fenoterol (Fen) is unique among the ␤ 2 -AR agonists in terms of ligand-induced receptor-G protein coupling selectivity (11). Although most ␤ 2 -AR agonists produce contractile responses in cardiomyocytes that can be sensitized by the G i disruptor pertussis toxin (PTX), indicating dual G s and G i coupling, the inotropic effect of Fen is PTX-insensitive, suggesting that Fen preferentially promotes ␤ 2 -AR-G s coupling. Fen contains two chiral centers in its molecule and may exist as four stereoisomers (Fig. 1). The stereoisomers of Fen and a series of Fen analogs have been synthesized (12)(13)(14). The role of ligand chirality in G protein-coupling selectivity has recently been demonstrated using these Fen derivatives (15). Specifically, we have shown that (R,RЈ)-Fen and (R,RЈ)-4Ј-methoxyfenoterol ((R,RЈ)-methoxyFen) preferentially activated G s signaling, as evidenced by the lack of PTX sensitivity of their contractile responses in cardiomyocytes and their inability to activate G i -dependent ERK signaling in HEK-293 cells. In contrast, the corresponding (S,RЈ)-isomers exhibited robust PTX sensitivity in these responses suggesting that they activated both G s and G i .
The mechanism for the differential G protein coupling of ␤-ARs has been the major focus of various studies. It has been suggested that phosphorylation of the ␤ 2 -AR by cAMP-dependent protein kinase (PKA) or G protein-coupled receptor kinases (GRKs) promotes the receptors to couple to G i proteins (16 -20). However, some evidence argues against this perception (21,22). Thus, the molecular basis for inducing ␤ 2 -AR's coupling selectivity to different G proteins remains largely elusive.
Our recent simulation studies employing data obtained with [ 3 H]CGP-12177 as the marker ligand have identified hydrogen bond (HB) formation between the tyrosine 308 residue (Tyr-308 or Y 7.35 in Ballesteros-Weinstein numbering) in transmembrane (TM) 7 of ␤ 2 -AR and a HB acceptor at the 4Ј-position of (R,RЈ)-Fen and (R,RЈ)-methoxyFen (13,23). These preliminary results suggest that the Tyr-308 residue is essential for agonistinduced ␤ 2 -AR preferential coupling to G s protein. In this study, using site-directed mutagenesis, receptor pharmacology, and cardiomyocyte activation in conjunction with computer simulation, we demonstrated that (R,RЈ)-4Ј-aminofenoterol ((R,RЈ)-aminoFen) (Fig. 1) targets the WT ␤ 2 -AR but not the mutant receptor, ␤ 2 -AR Y308F, to G s -biased signaling. These results experimentally verify our computation-based predictions, and here we present the first evidence that interactions with an individual residue in a conformation of the ␤ 2 -AR can induce or stabilize a conformation that leads to selective coupling to a G protein subunit.

EXPERIMENTAL PROCEDURES
␤ 2 -AR Model Construction and Docking Methodology-Docking of ligands to ␤ 2 -AR models was performed as recently described (23). In brief, a crystallographic model of the ␤ 2 -AR co-crystallized with an inverse agonist carazolol (PDB entry 2RH1) was used as a docking target in simulations. The model was modified by swapping the tyrosine 308 residue into phenylalanine using Yasara to obtain the model representing the Y308F mutant. Molegro Virtual Docker software (MVD ver-sion 2010.4.0.0, Aarhus, Denmark) was used for docking simulations within the binding cavity of the target model using MolDock SE algorithm as a search engine.
Animals-Male Sprague-Dawley rats (200 -250 g) were purchased from Charles River. ␤ 2 -AR knock-out mice were generous gifts from Dr. Brian Kobilka (Stanford University Medical Center, Palo Alto, CA). Animals were housed and studied in conformance with the "Guide for the Care and Use of Laboratory Animals, Eighth Edition" from the National Institutes of Health (NOT-OD-12-020, released 2011), with institutional Animal Care and Use Committee approval.
Generation of Stable Cell Lines and Recombinant Adenoviruses-HEK293A cells were obtained from Invitrogen and maintained in DMEM supplemented with 10% FBS at 37°C in a humidified 5% CO 2 incubator. The plasmids encoding for the human ␤ 2 -AR and ␤ 2 -AR Y308F mutant were kindly provided by Dr. Brian Kobilka and Dr. Hitoshi Kurose (Kyushu University, Fukuoka, Japan), respectively. Each of these coding sequences of ␤ 2 -AR was subcloned into pcDNA3.1Ϫ vector (Invitrogen). HEK cells were transfected with the resultant plasmids using Lipofectamine 2000 reagent (Invitrogen), and stably transfected clones were selected against G418 (0.8 mg/ml). When stable expression was achieved, the cells were cultured in the presence of 0.3 mg/ml G418. Adenoviruses for human ␤ 2 -AR (adeno-␤ 2 -AR) and green fluorescent protein (adeno-GFP) have been described previously (24,25). Adenoviral expression vector carrying the ␤ 2 -AR Y308F coding sequence was generated by subcloning. Viral particles were purified from transfected HEK cells using standard viral amplification and CsCl purification methods. Viral titers were determined in dilution assays by an immunocytochemical technique using an antibody raised against ␤ 2 -AR (sc-569, Santa Cruz Biotechnology, Santa Cruz, CA).
Cardiomyocyte Isolation, Adenoviral Gene Transfer, and Contractility Measurement-Cardiomyocytes were isolated from male ␤ 2 -AR knock-out mice (2-4 months old) or Sprague-Dawley rats using a standard enzymatic technique (9,24). Mouse cardiomyocytes were seeded on laminin-coated coverslips and infected with the adenoviruses at a multiplicity of infection of 100 (24). The cells were subsequently cultured for 24 h in minimal essential medium (M1018, Sigma) supplemented with forskolin (1 M) and 2,3-butanedione monoxime (10 mM). Contractility of single cardiomyocytes was measured as described previously (9). In brief, cardiomyocytes were perfused with a buffer containing (in mM) 137 NaCl, 4.9 KCl, 1.2 MgCl 2 , 1 NaH 2 PO 4 , 1 CaCl 2 , 20 glucose, and 20 HEPES (pH 7.4) and electrically paced (0.5 Hz for rat cardiomyocytes or 1 Hz for mouse cardiomyocytes) at ambient temperature on a microscopic stage. Cell length was monitored by an optical edgetracking method using an instrument setup manufactured by IonOptix (Milton, MA). Measurements were made under steady-state conditions before and after exposure of the myo- cytes to a single dose of the agonist. In a subset of experiments, aliquots of cells were incubated with PTX (0.75 g/ml at 37°C for Ͼ3 h) to block G i signaling, as described previously (9).
cAMP Accumulation Assay-Cells cultured on poly-D-lysine-coated 12-well plates were treated with PTX (0.3 g/ml) or vehicle overnight and incubated for 10 min with the phosphodiesterase inhibitor 3-isobutyl-1-methylxanthine (1 mM) in HEPES-buffered Hanks' balanced salt solution prior to stimulation. Cells were then treated with the agonist or control vehicle for 10 min. Reactions were stopped by the addition of HCl. cAMP contents in the clarified cellular extracts were determined with an enzyme immunoassay kit (Cayman Chemical, Ann Arbor, MI) according to the manufacturer's protocol. cAMP contents were normalized with total cellular protein. Protein contents were determined by the method of Lowry (Bio-Rad).
Immunoblotting-Whole-cell lysates in lysis buffer (Cell Signaling Technology, Danvers, MA) containing protease inhibitor mixture (Roche Diagnostics) and phosphatase inhibitor mixture (EMD Millipore, Billerica, MA) were centrifuged at 15,000 ϫ g for 10 min. Clarified cell lysates (20 g) for ␤ 2 -AR detection were further denatured and treated with 125 units of peptide:N-glycosidase (New England Biolabs) for 2 h at 37°C. The samples were denatured in Laemmli sample buffer and resolved by SDS-PAGE. Phosphorylation of ERK and ␤ 2 -AR was detected by immunoblotting using the same antibodies as described previously (15).
Radioligand Binding Assay-Receptor density was determined on membranes derived from HEK stable cell lines expressing WT or mutant forms of ␤ 2 -AR as described previously (14). The B max and K d values were determined by nonlinear regression analysis using Prism 4 (GraphPad Software, San Diego).
Statistical Analysis-Results are expressed as means Ϯ S.E. Unless described otherwise, unpaired Student's t test was performed to compare the means between two groups and oneway analysis of variance for multiple group comparison fol-lowed by post hoc analysis with Bonferroni's t test. Statistical analysis and curve-fitting of the concentration-response curves were conducted using Prism 4. The curves of the cAMP assays were fitted to the sigmoid curves by nonlinear regression analysis using the four-parameter logistic model without giving any constraints. Curve-fitting of the cardiomyocyte contractility data was conducted using the same algorithms and constraints laid out in our previous study (15).

RESULTS
Role of the Aminoalkyl Substituent of (R,RЈ)-Fen on Preferential ␤ 2 -AR-G s Coupling-To define the structural features of Fen compounds contributing to selective ␤ 2 -AR-G s signaling, we have undertaken a structure-activity relationship approach. In this campaign, PTX was used to distinguish the contribution of ␤ 2 -AR-G i signaling in the agonist-stimulated inotropic effects of a collection of Fen derivatives (Fig. 1) on a cardiomyocyte contractility model. By inhibiting the G i signaling with PTX, the regulatory inhibition of adenylyl cyclase on cAMP synthesis would be decreased, and as a result the G s -stimulated contractile response would be enhanced. Four Fen derivatives ((R,RЈ)-Fen, (R,RЈ)-methoxyFen, (R,RЈ)-aminoFen, and (R,SЈ)-aminoFen) eliciting a G i -independent activation of ␤ 2 -AR (thus PTX-insensitive) were identified (Table 1 and Fig. 2B) with (R,RЈ)-aminoFen demonstrating the highest G s selectivity (as assessed by the small difference between the EC 50 values of the ϪPTX and the ϩPTX groups, Table 1). Their contractility stimulatory effects were mediated through ␤ 2 -AR because these effects could be antagonized by ICI-118,551, a specific ␤ 2 -AR antagonist (Fig. 3). (R,RЈ)-AminoFen was subsequently used as the typical G s -selective ␤ 2 -AR agonist in the rest of this investigation.
(R,RЈ)-AminoFen Selectively Activates ␤ 2 -AR-G s Signaling in Cardiomyocytes Expressing WT ␤ 2 -AR but Activates Both G s and G i in Cardiomyocytes Expressing the ␤ 2 -AR Y308F Mutant-Cardiomyocytes express both ␤ 1 -AR and ␤ 2 -AR, and robust ␤ 2 -AR-G i coupling has been demonstrated in freshly isolated adult mouse cardiomyocytes expressing endogenous ␤ 2 -AR or human ␤ 2 -AR at 200-fold over basal level (10). Hence, we employed cardiomyocytes from ␤ 2 -AR knock-out mice transduced with exogenous ␤ 2 -AR or its mutants as a physiological model to investigate the role of the ␤ 2 -AR Tyr-308 residue on ligand-directed G protein selectivity. In our recent study, we have shown that ␤ 2 -AR in adult rodent cardiomyocytes lost its coupling to G i after overnight culture, and addition of forskolin in the culture medium could maintain functional dual coupling of ␤ 2 -AR to G s and G i proteins (26). In this investigation, we first confirmed the presence of functional ␤ 2 -AR-G i coupling in ␤ 2 -AR knock-out mouse cardiomyocytes reconstituted with human ␤ 2 -AR using zinterol, a selective ␤ 2 -AR agonist (Fig. 4). In another control experiment, cultured cardiomyocytes from ␤ 2 -AR knock-out mice were infected with adeno-GFP and then subjected to (R,RЈ)-aminoFen stimulation to study the effect of this compound on stimulating ␤ 1 -AR. The results in Fig. 5A show that the ␤ 1 -AR stimulatory effect of (R,RЈ)-aminoFen was undetectable at 100 nM, was minor at 500 nM (175 Ϯ 26%), and became very substantial (about 350%) at 1 M. In subsequent contractility studies, we only tested (R,RЈ)-aminoFen up to 500 nM.
Residue Tyr-308 of ␤ 2 -AR Is Necessary for Ligand-directed G s -biased ␤ 2 -AR Signaling-It has been demonstrated in HEK cells that ␤ 2 -AR agonists trigger an acute increase in ERK phosphorylation, which peaks at 5 min, and this effect is mediated in part by a G i -dependent mechanism (16,27). Furthermore, both G s -and G i -mediated ␤ 2 -AR activation can lead to ERK phosphorylation (16,22,27). Cell lines stably expressing WT ␤ 2 -AR (HEK-␤ 2 -AR cells) and ␤ 2 -AR Y308F mutant (HEK-␤ 2 -AR Y308F cells) were established from HEK293A cells. The levels of ␤ 2 -AR in these cell lines were 4033 Ϯ 826 and 2300 Ϯ 80 fmol/mg protein, respectively, as assayed by radioligand binding, whereas the level of ␤ 2 -AR in the parental cells was 30 -40 fmol/mg (21). Next, we studied the G protein pathways responsible for phospho-ERK (p-ERK) induction by (R,RЈ)-aminoFen and ISO in these cell lines. The sensitivity of agonist-induced ERK phosphorylation toward PTX was used to indicate G i activation.
Stimulation with ISO increased p-ERK by about 6-fold in HEK-␤ 2 -AR cells (Fig. 6, A and B) and HEK-␤ 2 -AR Y308F cells (Fig. 6, C and D). This activation of p-ERK was mediated by a combination of G s -and G i -dependent pathways as demonstrated by the decreases in maximal ERK phosphorylation in the PTX-treated groups. In contrast, in HEK-␤ 2 -AR cells, (R,RЈ)-aminoFen induced ERK phosphorylation in a PTXinsensitive manner (Fig. 6, A and B). Importantly, the increase in phosphorylation of ERK in response to (R,RЈ)-aminoFen exhibited a robust PTX sensitivity in HEK-␤ 2 -AR Y308F cells, and this G i -dependent effect appeared to be positively correlated with the concentration of (R,RЈ)-aminoFen (Fig. 6, C and D).
We also measured cAMP accumulation in HEK cells stably expressing ␤ 2 -AR and its Y308F mutant. Although (R,RЈ)-

Role of Tyr-308 of ␤ 2 -Adrenoceptor on G s -biased Signaling
aminoFen produced a PTX-insensitive cAMP response in HEK-␤ 2 -AR cells (Fig. 7A), the concentration-response profile shifted upwards in response to PTX treatment in HEK-␤ 2 -AR Y308F cells (Fig. 7B). Efficacy data ( Table 2) show that PTX significantly increased the E max value (from 89 Ϯ 1 to 117 Ϯ 2, p Ͻ 0.01) without altering the logEC 50 value of the cAMP response of (R,RЈ)-aminoFen in HEK-␤ 2 -AR Y308F cells (from Ϫ8.05 Ϯ 0.01 to Ϫ8.03 Ϯ 0.02, p ϭ 0.75). We did not observe sensitivity of cAMP response to PTX when HEK-␤ 2 -AR cells were stimulated with ISO (data not shown), thus corroborating a previous experiment (17). This may be due to an inherent limitation of the assay. In contrast, PTX caused an increase in the E max of the ISO-stimulated cAMP response (from 109 Ϯ 4 to 126 Ϯ 5, p Ͻ 0.05, Table 2) in HEK-␤ 2 -AR Y308F cells, mirroring a similar observation in cells expressing the ␤ 2 -AR D-4 mutant, a receptor phenotype having a reduced G s -coupling, and an increased G i -coupling (17). Taken together, these results demonstrate that the ␤ 2 -AR-Y308 residue is necessary for the G s -biased ␤ 2 -AR signaling and that Y308F mutation fully restored ␤ 2 -AR-G i signaling in response to (R,RЈ)-aminoFen, a G s -selective ␤ 2 -AR agonist, in cardiomyocytes and in HEK cells.
Effects of Agonist-induced Receptor Phosphorylation on G Protein-coupling Selectivity of ␤ 2 -AR-It is well established that agonist stimulation of ␤ 2 -AR leads to phosphorylation of the receptor by PKA and GRK, with important implications on receptor desensitization (28). In addition, phosphorylation of ␤ 2 -AR at the GRK or the PKA sites has been suggested to be necessary for ␤ 2 -AR-G i protein coupling in naive cells (16,17), cardiomyocytes (18 -20), and in vivo hearts (20). Therefore, we investigated the agonist-stimulated receptor phosphorylation in HEK cells expressing either the WT ␤ 2 -AR or the ␤ 2 -AR Y308F mutant using phosphosite-specific antibodies (29). Stimulation with ISO (1 M) for 5 min, a treatment time period reportedly leading to near-maximal receptor phosphorylation responses (30 -33), increased the phosphorylation of ␤ 2 -AR and ␤ 2 -AR Y308F at Ser-262 (PKA-site) to about 5-fold of basal (Fig. 8, A and B). Similarly, (R,RЈ)-aminoFen (1 M) produced the same maximal responses in both HEK-␤ 2 -AR cells and HEK-␤ 2 -AR Y308F cells (Fig. 8, A and B). Treatment with ISO FIGURE 4. Addition of forskolin reconstitutes functional coupling of ␤ 2 -AR to G i protein in cultured ␤ 2 -AR knock-out mouse cardiomyocytes induced with human ␤ 2 -AR. Cardiomyocytes from ␤ 2 -AR knock-out mice were infected with adeno-GFP (white bar) or adeno-␤ 2 -AR (black bars) and cultured for 24 h in the presence or absence of forskolin (1 M) and/or PTX (0.75 g/ml) as indicated. Cells were transferred to a perfusion chamber, electrically paced, and subjected to stimulation with zinterol (0.2 M, a concentration without an inotropic effect in freshly isolated cardiomyocytes from WT mice, see Fig. 1A in Ref. 26). Steady-state contractility was measured. Data (mean Ϯ S.E., n ϭ 10 -15 cells from 5 to 9 hearts for each data point) are expressed as percentages of the basal contractility. *, p Ͻ 0.05. Zinterol (0.2 M) did not increase contractility in cells infected with adeno-GFP demonstrating no ␤ 1 -AR stimulatory effect at this concentration. In cells infected with adeno-␤ 2 -AR and cultured in the absence of forskolin, the inotropic response produced by zinterol stimulation was the result of a pure ␤ 2 -AR-G smediated effect because ␤ 2 -AR and G i proteins were functionally uncoupled. In cells infected with adeno-␤ 2 -AR in the presence of forskolin, the coupling of ␤ 2 -AR to G i protein was reestablished. Therefore, the cardiomyocytes were unresponsive to zinterol as if they were freshly isolated WT ␤ 2 -AR ϩ cells when ␤ 2 -AR-G i coupling was intact. In cells infected with adeno-␤ 2 -AR in the presence of forskolin and PTX, the coupling of ␤ 2 -AR to G i protein still occurred, but G i had lost its function and could no longer negatively regulate ␤ 2 -AR-G s activation by zinterol.  also increased the phosphorylation of ␤ 2 -AR at Ser-355,356 (GRK sites) by about 17-fold in these cell lines (Fig. 8, C and D). Because it has been reported that phosphorylation of the ␤ 2 -AR mediated by GRK depends on a high concentration of agonists (27)(28)(29)(30)(31), we subsequently performed the receptor phosphorylation assay at higher concentrations of (R,RЈ)-aminoFen (namely 100-and 1000-fold of EC 50 cAMP concentrations, corresponding to a near-saturating and a saturating concentration of the agonist for receptor stimulation, respectively, refer to  Table 2) (Fig. 8, C and D).
In HEK-␤ 2 -AR cells, increased phosphorylation of the ␤ 2 -AR at the PKA site could be observed 5 min after stimulation with ISO or (R,RЈ)-aminoFen at 1 M (Fig. 8, A and B). Similarly, (R,RЈ)-aminoFen (0.2 and 2 M) and ISO (1 M) also increased the phosphorylation of ␤ 2 -AR at the GRK sites (Fig. 8, C and D). The same treatment with (R,RЈ)-aminoFen (10 Ϫ8 to 10 Ϫ6 M) increased the p-ERK level by about 4-fold via activating G s but not G i (Fig. 6, A and B). Treatment with ISO after disrupting the activity of G i with PTX also increased the p-ERK level by about 4-fold, which could be further increased to 6-fold in the absence of PTX (Fig. 6, A and B). Thus, both (R,RЈ)-aminoFen and ISO can induce the phosphorylation of ERK and ␤ 2 -AR via a G s -dependent pathway, but only ISO can activate ERK through a ␤ 2 -AR-G i signaling pathway. These results clearly demonstrate that (R,RЈ)-aminoFen and ISO produced similar effects in triggering phosphorylation of ␤ 2 -AR at the PKA sites or the GRK sites, although they exhibited diverse G protein selectivity in HEK stable cell lines expressing the WT ␤ 2 -AR. In addition, Y308F substitution on the ␤ 2 -AR caused a qualitative change in the G protein selectivity of (R,RЈ)-aminoFen from being exclusively G s -activating (Fig. 6, A and B) to dually G s /G i -activating (Fig. 6, C and D) without significantly affecting its activities in eliciting receptor phosphorylation by both PKA and GRKs (p Ͼ 0.05, Fig. 8, B and D). Thus, phosphorylation of ␤ 2 -AR at its PKA or GRK sites is insufficient to trigger the receptor coupling to G i proteins.
Docking Simulation on ␤ 2 -AR-To reveal the molecular interactions important for the ligand-directed G s -biased agonism of ␤ 2 -AR, a molecular model of (R,RЈ)-aminoFen was docked to the crystal model of the ␤ 2 -AR-binding site as well as to a model representing the Y308F mutant receptor using the same procedures as described previously (23). As shown in Fig.  9, A and C, the ligand molecule can be fitted reasonably well into the binding sites of both ␤ 2 -AR conformations (PDB entry 2RH1 for carazolol-bound and PDB entry 3SN6 for BI-167107and G s protein-bound, respectively) with the resorcinol ring of the ligand pointed in the direction of S203 5.42 and S207 5.46 , the ␤-OH and the secondary amine groups interacted with D113 3.32 and N312 7.39 residues. Notably, a HB interaction between the 4Ј-amino group of the ligand and the hydroxyl

TABLE 2 E max % and logEC 50 values of the (R,R)-aminoFen-and ISO-induced cAMP responses in HEK cell lines stably expressing ␤ 2 -AR WT or ␤ 2 -AR Y308F
Calculations of logEC 50 and E max values were based on concentration-response profiles for the compound-stimulated cAMP production. E max values are expressed as percentages of the E max response of the ␤ 2 -AR WT ϪPTX group. Means Ϯ S.E., n ϭ 3. *, p Ͻ 0.05; **, p Ͻ 0.01 versus ϪPTX group. group of Y308 7.35 was found probable in the carazolol-bound ␤ 2 -AR model (Fig. 9A). When (R,RЈ)-aminoFen was docked to the modified model of the Y308F mutant receptor (Fig. 9B), the position of the molecule within the binding site was very similar, and the above-mentioned interactions still occurred with the exception of the HB created by the 4Ј-amino moiety. In effect, the 4Ј-aminobenzyl ring assumed a slightly different position than during docking to the WT receptor model.

DISCUSSION
The current data (Figs. [5][6][7][8] suggest that the G s -selective signaling depends on specific interactions between the agonist and the ␤ 2 -AR-Y308 residue, and induction of receptor phosphorylation alone does not necessarily lead to a switching of the receptor coupling from G s to G i as once proposed (16). These results are consistent with those reported in our previous study on cardiomyocytes (15) in which the stereoisomers of Fen and methoxyFen possessing different G protein selectivity induced similar phosphorylation of ␤ 2 -AR at the PKA sites. Because (R,RЈ)-Fen, (R,RЈ)-methoxyFen, and (R,RЈ)-aminoFen are full agonists of ␤ 2 -AR capable of inducing receptor phosphorylation just as ISO (Fig. 8) (15), the only explanation for their preferential G s selectivity would be their exceptional abilities in stabilizing a receptor conformation favoring receptor-G s protein interaction. As the emerging paradigm of functional selectivity suggests, ligands can perturb a GPCR to attain "ensembles" of multiple conformations, and each of these conformations is capable of activating a distinct set of signaling events (1)(2)(3)(4)(5)(6)(7)(8). Therefore, one possibility is that the binding of the G s -selective agonists causes the ␤ 2 -AR to assume or stabilize conformations leading to G s protein coupling. ␤ 2 -AR in such conformation(s) regardless of its phosphorylation status interacts strongly with G s protein, prohibiting receptor-G i protein interaction from taking place. Furthermore, if the ␤ 2 -AR-Y308 residue is mutated, the G s selectivity of these agonists will be lost (Figs. 5 and 6), suggesting that mutation of this residue affects the receptor conformations stabilized by these agonists. The latter conformations, possibly resembling the ISO-bound receptor conformations, are G i protein-permissive. Y308 7.35 has been implicated for the high affinity binding of ␤ 2 -AR-selective agonists (34). Importantly, our recent docking simulation study (23) supported the initial determination (13,14) that in inverse agonist-associated conformations, such as the S 1 and S 2 conformations described by Kim et al. (35), Y308 7.35 may interact directly with the 4Ј-hydroxyl, 4Ј-methoxy, or 4Ј-amino group of the (R,RЈ)-Fen derivatives through hydrogen bonding. This hypothesis was confirmed in this study (Fig. 9A).
Because the human ␤ 2 -AR and the rat ␤ 2 -AR share a high sequence homology particularly at the TM regions encompassing the ligand binding pocket (data not shown), the data derived from the rat cardiomyocyte study (Fig. 2 and Table 1) could be interpreted in the light of the structural insight gained from the human ␤ 2 -AR models. As shown in Fig. 2, A, C, and D, and Table 1, the R,RЈ-isomers of Fen derivatives containing the following substituents on the aminoalkyl portion, phenyl (PhFen), 1-naphthylfenoterol, 2-naphthylfenoterol, ethylfenoterol, and 4Ј-methoxy-1-naphthyl (MNFen), produced PTX-sensitive contractile responses in cardiomyocytes. These results suggest that these compounds activate both G s and G i pathways of ␤ 2 -AR. In contrast, the positive inotropic effects of (R,RЈ)Fen, (R,RЈ)-methoxyFen, (R,RЈ)-aminoFen, and (R,SЈ)-aminoFen were PTX-insensitive (Table 1 and Fig. 2B), indicating that they selectively activate ␤ 2 -AR-G s signaling. Together, these data illustrate the structural features of a Fen compound for liganddirected selective ␤ 2 -AR-G s signaling as follows: (i) a benzyl rather than a naphthyl moiety on its aminoalkyl substituent; (ii) a 4Ј-oxygen or a 4Ј-nitrogen moiety on this aromatic substituent; and (iii) a mandatory R-configuration on the chiral center of the ␤-OH group and a preferred R-configuration on the second chiral center.
The fact that (R,RЈ)-PhFen, but not (R,RЈ)-Fen, (R,RЈ)-methoxyFen, and (R,RЈ)-aminoFen, produced a PTX-sensitive inotropic effect (Fig. 2, A and B, and Table 1) suggests that the 4Ј-(N/O) moieties in these (R,RЈ)-Fen derivatives are indispensable for agonist-induced preferential G s activation. If either the phenyl hydroxyl group of Tyr-308 or the 4Ј-(N/O) moiety is lost, the HB between the ligand and the 7.35 residue (Fig. 9A) will not exist, and promiscuous G s and G i dual signaling rather than G s -biased signaling will be induced (Figs. 5C, 6C, and 7B). This indicates that specific interaction between Tyr-308 and the 4Ј-(N/O)-benzyl moiety promotes preferential receptor-G s protein coupling. Our simulation study (23) has also shown that (R,RЈ)-Fen derivatives with naphthyl moieties interact not by hydrogen bonding with Tyr-308 but rather byinteractions with the other aromatic residues in the ligand binding pocket (Fig. 9D). The opposite is true for compounds with 4Ј-(N/O)benzyl moieties (23). The dominance of theinteractions with (R,RЈ)-MNFen binding, irrespective of the presence of a potentially hydrogen bonding 4Ј-methoxy moiety, is associated with dual G s and G i protein coupling of ␤ 2 -AR (Fig. 2D). These key features in the ligand-receptor interaction make (R,RЈ)-MNFen a superior negative model compound as compared with ISO or the (S,RЈ)-isomers in the study of G s -biased signaling, and the functional data with (R,RЈ)-MNFen stimulation (Fig. 2D) also point to the same conclusion. Thus, based on our simulated docking study and experimental evidence, we conclude that HB interactions between the 4Ј-(N/O)-benzyl moiety of the (R,RЈ)Fen derivatives and the ␤ 2 -AR-Y308 residue play an important role on ligand-directed ␤ 2 -AR-G s signaling.
The structural features of Fen derivatives for preferential G s selectivity and receptor subtype selectivity have both similarities and differences. Our initial studies (13,14) have shown that both stereochemistry and the aminoalkyl substituent play essential roles on the ␤ 2 -AR subtype selectivity of the Fen compounds. Although an R-configuration, hydrogen bonding with Tyr-308, andinteractions with aromatic residues in the ligand binding pocket can all contribute to high ligand binding affinity and increased selectivity to ␤ 2 -AR (14), only the R-configuration and hydrogen bonding with Tyr-308 correlate with FIGURE 9. Binding poses of (R,R)-aminoFen and (R,R)-MNFen docked to the ␤ 2 -AR models. A, docking of (R,RЈ)-aminoFen to the carazolol-bound ␤ 2 -AR model (PDB entry 2RH1), and B, to the ␤ 2 -AR Y308F mutant. The 4Ј-amino group of (R,RЈ)-aminoFen forms a HB with the Tyr-308 residue of ␤ 2 -AR (green arrow), and the interaction is lost in the ␤ 2 -AR Y308F mutant. C, docking of (R,RЈ)-aminoFen to a BI-167107 and G s protein-bound conformation of ␤ 2 -AR (PDB entry 3SN6). The location of the docked molecule highly resembles the orientation of co-crystallized agonist, BI-167107, as both molecules share significant structural similarities. The agonists form a network of analogous HB interactions with the receptor residues Ser-203, Ser-207, Asn-312, and Asp-113. An additional interaction can be observed between the 4Ј-amino moiety of the ligand and Lys-305 residue. D, docking of (R,RЈ)-MNFen to a carazolol-bound conformation of ␤ 2 -AR (PDB entry 2RH1). Aromatic residues in the ligand binding pocket able to forminteractions with the naphthyl moiety of the ligand are shown. Ligand molecule is rendered in atom-type color-coded stick mode, and five essential TM helices of target ␤ 2 -AR model are shown and colored as follows: TM3, magenta; TM4, green; TM5, red; TM6, yellow; and TM7, blue. G s selectivity. Theinteractions, however, likely have a detrimental effect on G s selectivity of the Fen derivatives. It is conceivable that the above-mentioned molecular interactions would impact the receptor conformational ensembles stabilized by different ␤ 2 -AR agonists and subsequently result in the differential G protein-coupling selectivity.
Recent studies have identified H 6.55 to be a major determinant of ligand-biased signaling in dopaminergic D 2L receptors (36). A follow-up study has further characterized the receptor conformations involved in D 2L receptor functional selectivity (37). Briefly, a model of functional selectivity for D 2L receptor has been proposed in which TM6 represents a rotatory switch in response to the binding of different functionally selective agonists. In this model, if H 6.55 in TM6 rotates toward S 5.43 on TM5, the resultant ligand-stimulated receptor conformation will favor the activation of the arachidonic acid pathway. Conversely, if H 6.55 rotates toward TM7 and interacts with Y 7.35 , the ligand-stimulated receptor conformation will lead to a signaling bias toward cAMP/MAPK activation.
A very similar mechanism occurs in ␤ 2 -AR. Y308 7.35 is known to form a HB with a neighboring residue N293 6.55 in TM6, and this specific interaction (additionally shown on Fig. 9A) remains intact in all reported crystallographic structures of the receptor. Because the 4Ј-(N/O) moiety of the G s -selective ligand involves the hydroxyl group of Tyr-308 in another HB interaction, a competition occurs between the Tyr-308 -Asn-293 interaction and Tyr-308 -ligand interaction. It is therefore postulated that the naturally occurring Tyr-308 -Asn-293 HB is disrupted when the G s -selective ligand binds to the receptor. The Tyr-308 -Asn-293 interaction bridges the upper parts of TM6 and TM7. Breaking this interaction during the receptor conformational transition might be a key phenomenon leading to specific activation of the ␤ 2 -AR to a form favoring selective G s protein coupling.
Using acetylcholine M 2 receptor as a model, Bock et al. (38) have designed "dualsteric" agonists to study the role of allosteric vestibules on G protein activation. The allosteric vestibule is located at the entrance of the orthosteric binding cavity of many class A GPCRs and has been implicated for ligand binding (39). Acetylcholine is known to activate G i and G s signaling of the M 2 receptor. The authors have found that dualsteric agonists (such as iper-6-phth and iper-6-naph) exhibited a G i over G s signaling bias compared with acetylcholine and their parent compound Iperoxo, an orthosteric muscarinic agonist. Mutagenic studies have identified the M 2 -W422 7.35 residue located at the allosteric vestibule to be critical for both G s and G i protein activation, with the gain in dualsteric probe efficacy for G i activation in the allosteric mutant. Interestingly, W422 7.35 and Y177 5.32 in extracellular loop 2 of the M 2 receptor and the analogous Y308 7.35 and F193 5.32 in ␤ 2 -AR that line the passage to the orthosteric binding cavity have been suggested to undergo a conformational rearrangement during receptor activation (39,40). The authors implied from their findings that spatial rearrangement of this passage is critical for receptor movements required for appropriate unfolding of the intracellular domain region for G protein coupling.
The two previous studies and this study indicate the important role of the 7.35 residue on GPCR conformational transition leading to G protein activation. Notably, here we provide the direct evidence to pinpoint the role of this residue on functional selectivity and illustrate this point with an "extreme" form of signaling bias (in nominal terms of with or without PTX sensitivity rather than in ratiometric terms of a biased factor) in a physiological context of adult cardiomyocytes. Further studies are needed to determine whether this deduction could be generalized in a broader sense, such as to other class A aminergic GPCRs.
Our study design necessitates the investigation of a single aspect of the ␤ 2 -AR agonists, specifically their differential selectivity to G s and G i proteins. From a chemical biology perspective, however, the ␤ 2 -AR is only one of the many possible in vivo targets of the Fen compounds. Indeed, in complex biological systems such as the adult cardiomyocytes, a compound is likely to produce its effects via interactions with multiple cellular proteins. Therefore, it is unsurprising to find that high concentrations of (R,RЈ)-aminoFen also stimulate ␤ 1 -AR (Fig. 5A) given that only 1 out of the 15 amino acids that constitute the ligand binding pocket differs between ␤ 1 -AR and ␤ 2 -AR (41). Consistently, our previous binding affinity data (13) have also shown that the subtype selectivity of ␤ 2 -AR relative to ␤ 1 -AR in terms of K i ␤ 1 -AR/K i ␤ 2 -AR ratio was 9 for (R,RЈ)-aminoFen. Interestingly, it is the 7.35 residue (␤ 2 -AR-Y308) that is different, and the corresponding residue is a phenylalanine or Phe-359 in ␤ 1 -AR, incidentally the same mutation characterized in this study. However, mutation on ␤ 2 -AR to convert the amino acid residues in its ligand binding pocket to that resembling ␤ 1 -AR produced a dissimilar function of increased receptor-G i protein coupling in the ␤ 2 -AR Y308F mutant, yet it is known that ␤ 1 -AR does not normally couple to G i (10). Thus, it is not the amino acid residues themselves but rather their different interactions with the ligand and the resultant conformational changes (42) that determine the diverse selectivity of the ␤-AR subtypes or mutants to different G proteins. As a cautionary note, this interpretation is only confined to the very first step of G protein coupling at the receptor level, without taking into account other intracellular mechanisms such as phosphorylation, internalization, G protein abundance, and subcellular compartmentation.
In addition, cross-talk between different receptor-mediated signaling pathways is very common when a given ligand simultaneously stimulates two or more receptors. We observed this cross-talk of signals between ␤ 1 -AR and ␤ 2 -AR in cardiomyocytes stimulated by 500 nM (R,RЈ)-aminoFen as exemplified by a higher contractile response in the ␤ 2 -AR Y308F mutant-expressing cells (Fig. 5C) versus the WT ␤ 2 -AR-expressing cells (Fig. 5B) in the PTX-treated groups. The detailed mechanism of the cross-talk between the ␤ 1 -and ␤ 2 -AR signals is beyond the scope of this study, although an elaborated discussion can be found in Zhang et al. (43). This example illustrates that no single assay or approach can adequately elucidate the complex pharmacology of a compound.
In conclusion, this study has identified an amino acid residue in ␤ 2 -AR necessary for functional selectivity. Mutation of this residue causes a G s -selective agonist to gain the ability to activate G i when it binds to the ␤ 2 -AR. We also provide, for the first time, functional data confirming the identification of the ligand-receptor interactions important for G s -biased signaling in ␤ 2 -AR. Advances in structural biological techniques (42, Role of Tyr-308 of ␤ 2 -Adrenoceptor on G s -biased Signaling 44 -48) will ultimately unravel how these interactions during ligand binding translate into receptor conformation(s) for selective coupling to different G proteins. This investigation has elucidated the molecular basis of G s -biased agonism in ␤ 2 -AR, and this is one step closer to structure-based design of signaling pathway-specific drugs.