Coupling of β2-Adrenoceptors to XLα_s and Gα_s: A New Insight into Ligand-Induced G Protein Activation

Abstract

Gα_s and extra-large Gα_s (XLα_s) can both transduce receptor activation into intracellular cAMP generation. It is unknown, however, whether these two GNAS-locus products display distinct properties with respect to receptor coupling. Here, we show that XLα_s couples to the β2-adrenoceptor more efficiently than Gα_s. In transfected human embryonic kidney 293 cells and mouse embryonic fibroblasts null for both Gα_s and XLα_s (2B2 cells), basal cAMP accumulation mediated by XLα_s was higher than that mediated by Gα_s. Inverse agonist treatment reduced Gα_s-mediated basal activity, whereas its effect was markedly blunted on XLα_s-mediated basal activity. Rank order of ligand efficacies regarding cAMP accumulation was the same when the receptor was coupled to XLα_s or Gα_s. However, ligand-induced and XLα_s-mediated cAMP generation was higher than that mediated by Gα_s. The relatively high efficiency of XLα_s-mediated cAMP generation was conditional, disappearing with increased level of receptor expression or increased efficacy of ligand. Full-agonist responses in XLα_s- and Gα_s-expressing cells were comparable even at low receptor levels, whereas partial agonist responses became comparable only when the receptor expression was increased (>3 pmol/mg). Radioligand binding studies showed that the high-affinity component in agonist binding to β2-adrenoceptor was more pronounced in cells expressing XLα_s than those expressing Gα_s. We discuss these findings in the framework of current receptor-G protein activation models and offer an extended ternary complex model that can fully explain our observations.

Heterotrimeric G proteins, consisting of α, β, and γ subunits, constitute a large family of signaling proteins that transmit receptor signals to intracellular effectors. Upon interaction with an active receptor, G proteins undergo a conformational change that results in guanine-nucleotide exchange on the α subunit and dissociation of α and βγ subunits. Dissociated subunits interact with intracellular effectors to modulate their activity. Among others, G_s protein has specifically evolved to transmit receptor signals to the stimulation of adenylyl cyclase that leads to intracellular generation of the second messenger cAMP (for reviews, see Gilman, 1987; Hamm, 1998).

The α subunits of G_s are encoded by the complex GNAS locus on the chromosome 20q13 (Kozasa et al., 1988). This locus generates multiple products through the splicing of different alternative first exons onto a common downstream exon (exon 2). In addition, alternative splicing of exon 3 of Gα_s gene results in long and short forms of Gα_s protein (Bray et al., 1986). A recently identified product of the GNAS locus is the extra-large α_s (XLα_s) protein, in which the first exon of Gα_s is replaced by the XL-exon that encodes, in rat, 347 instead of 47 amino acids in the amino terminus of Gα_s (Kehlenbach et al., 1994). In contrast to Gα_s, which is expressed ubiquitously, XLα_s is expressed particularly in neuroendocrine tissues (Pasolli et al., 2000) and derived from the paternal allele (Hayward et al., 1998). Polymorphisms affecting the XL-exon have been shown to be associated with prolonged trauma-induced bleeding in humans (Freson et al., 2001). In addition, XLα_s knockout mice have shown poor postnatal growth and survival, suggesting an important role for XLα_s in postnatal development and adaptation (Plagge et al., 2004). Perinatal defects similar to those in XLα_s knockout mice have also been identified in two unrelated children who carried large deletions that comprised the paternal GNAS allele (Geneviève et al., 2005).

XLα_s has been shown to interact with Gβγ dimers to form a stable heterotrimeric complex and to undergo cholera toxin-induced ADP-ribosylation. It is activated by GTPγS, and upon binding of GTPγS, undergoes a conformational change similar to Gα_s (Klemke et al., 2000). It has also been shown that XLα_s couples agonist stimulation of different types of Gα_s-coupled receptors to the activation of adenylyl cyclase in transfected cells (Bastepe et al., 2002; Linglart et al., 2006). Finally, the point mutation Q548L in XLα_s (equivalent to Q227L in Gα_s) results in constitutive adenylyl cyclase stimulation (Klemke et al., 2000). Thus, consistent with the fact that XLα_s and Gα_s share identical functional domains (except their N termini), XLα_s demonstrates Gα_s-like properties. Although there have been conflicting reports about its intracellular distribution, the fact that it couples membrane receptors to adenylyl cyclase strongly suggests that XLα_s, like Gα_s, is also expressed in the plasma membrane (Kehlenbach et al., 1994; Pasolli et al., 2000; Ugur and Jones, 2000; Linglart et al., 2006). However, little is known about the signaling properties of XLα_s compared with Gα_s. Despite the functional similarities between XLα_s and Gα_s, their coupling properties to the membrane receptors may diverge because of the difference in their N termini, which has been implicated to be involved in receptor interaction and activation (Fanelli et al., 1999). Apparently, this difference does not result in receptor selectivity, because all the G_s-coupled receptors investigated to date have also been found to be able to couple to XLα_s (Bastepe et al., 2002; Linglart et al., 2006). However, the difference between the two proteins may be particularly important in terms of agonist-directed signal trafficking, where different ligands can couple the same receptor to different G proteins with diverging efficacies. Thus, a set of ligands may exhibit different order of efficacy depending on whether a particular receptor is coupled to Gα_s or XLα_s, which may have a potential pharmacological importance.

In the present study, we therefore investigated signaling properties of XLα_s in comparison with Gα_sL (Gα_s long form), by measuring its ability to mediate receptor- and ligand-dependent or -independent activation of adenylyl cyclase. We used human β2-adrenoceptor (βAR) as a prototypical G_s-coupled experimental model. The purpose of the present study was 2-fold: first, to compare the coupling properties of XLα_s and Gα_s to βAR; and second, to gain further insight into the mechanism of G protein-mediated signaling by using a system in which the same receptor is coupled to two different G proteins as a tool. The latter point is discussed in the framework of the current interpretation of the ternary complex models.

Materials and Methods

Cell Culture. HEK293 cells were grown in DMEM supplemented with penicillin (100 U/ml), streptomycin (100 μg/ml), and 10% fetal bovine serum in a humidified atmosphere of 5% CO₂ at 37°C. For 2B2 Gnas^E2-/E2- fibroblasts, where the second exon of Gα_s or XLα_s was disrupted out, DMEM-F-12 medium supplemented with 5% fetal bovine serum was used, and cells were grown at 33°C.

Plasmid Constructs and Transfections. The point mutation (L519P) present in the original cDNA of XLα_s was corrected by replacing the P codon with the L codon at the relevant position by standard site directed mutagenesis techniques on the pcDNA3.1(+) vector (Kehlenbach et al., 1995). cDNAs encoding rat XLα_s or rat Gα_sL were cleaved from the original vectors and reinserted into pcDNA3.1-hygromycin plasmids. cDNA encoding human βAR was inserted into the pcDNA3.1-zeocin plasmid. HEK293 cells and 2B2 cells were transfected with calcium phosphate precipitation (Kingston et al., 1996) and DEAE-dextran methods (Gulick, 1997), respectively. Stable monoclones were selected using appropriate antibiotics. Protein expression levels of the selected clones or transiently transfected cells were determined by radioligand binding or Western blot analysis. HEK293 clones that overexpress βAR at a level of 30 pmol/mg membrane protein were a kind gift of Dr. Tommaso Costa (Istituto Superiore di Sanitá, Rome, Italy). Original cDNAs for human-β₂AR, rat-Gα_sL and rat-XLα_s were kind gifts of T. Costa, T. L. Z. Jones (NIH, NIDDK, Bethesda, MD), and W. B. Huttner (Max Plank Institute of Cell Biology and Genetics, Dresden, Germany), respectively.

Immunoblots and SDS-Polyacrylamide Gel Electrophoresis. Proteins were separated by SDS-polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membrane (Bio-Rad, Hercules, CA) by using standard procedures. Proteins were detected by custom-designed polyclonal antibody (produced by Pacific Immunology Corp., Ramona, CA) raised against the C-terminal decapeptide [NH₂-(Cys)-Arg-Met-His-Leu-Arg-Gln-Tyr-Glu-Leu-Leu] of Gα_s (and XLα_s), and enhanced chemiluminescence as described by the manufacturer (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Densitometric analysis of blots was carried out by using an image analysis system (Diana version 1.6; Aida version 2.43; Raytest, Straubenhardt, Germany).

Membrane Preparations and Receptor Binding Assays. Cells were pelleted at 200 g for 5 min at room temperature, resuspended in homogenization buffer [5 mM Tris-HCl, pH 7.4, and protease inhibitor mixture (Roche Diagnostics, Mannheim, Germany)], and homogenized by passing the suspension (10–15 times) through a 26-gauge syringe tip on ice. The homogenate was centrifuged at 450g for 10 min at 4°C, and the resulting supernatant was centrifuged at 100,000g for 30 min at 4°C (Optima LE-80K ultracentrifuge; Beckman Coulter, Fullerton, CA). The pellet was resuspended in a buffer containing 50 mM Tris-HCl, pH 7.4, 0.3 mg/ml dithiothreitol, 5 mM MgCl₂, and protease inhibitor mixture and then repelleted by centrifugation at 100,000g for 30 min at 4°C. The final pellet was suspended in 50 mM Tris-HCl, pH 7.4, 10 mM MgCl₂, protease inhibitor mixture, and 25% sucrose at a protein concentration of approximately 2 mg/ml and stored at -80°C.

In saturation binding assays, 0.5 to 1 μg of membrane protein was incubated with [¹²⁵I]iodocyanopindolol (100,000 dpm/well) in a total volume of 100 μl of buffer (100 mM KCl, 10 mM MgCl₂, and 50 mM Tris-HCl, pH 7.4) for 2 h at 37°C in 96-well plates. The reaction was stopped by rapid filtration through a Whatman GFB glass fiber filter (Whatman, Maidstone, UK) by using a cell harvester (Skatron Instruments, Lier, Norway). Radioactivity on the filters was counted by using a scintillation counter (MicroBeta 1450 Trilux; PerkinElmer Wallac, Turku, Finland). Nonspecific binding was determined in the presence of 1 μM cyanopindolol. Competition binding assays were conducted similarly except that varying concentrations of indicated competitor ligands and 20,000 cpm/well [¹²⁵I]iodocyanopindolol were used in the presence or absence of GTPγS (1 μM) or GPD (100 μM) + AlF (20 μM NaCl/10 mM NaF) at a final buffer volume of 200 μl. Nucleotide-induced shift in agonist binding curves was found to be more complete with GDP + AlF than with GTPγS. Therefore, we presented the results of the experiments in which GDP + AlF were used. Binding curves were analyzed by nonlinear regression of a four-parameter logistic equation or numerical solution of multisite binding equation in the presence of multiple ligands by means of an in-house Excel routine (Microsoft, Redmond, WA). Binding curves obtained in parallel experiments in the presence or absence of guanine nucleotide were analyzed by sharing receptor concentration and affinity values among binding curves. The effect of parameter sharing was tested by using F statistics.

Determination cAMP Accumulation. Cells were seeded in 96-well plates at a density of 5 to 10 × 10³ cells/well 24 h before the experiment. Two hours before the assay, cells were washed once with serum-free DMEM. Assays were conducted in a total volume of 100 μl at 37°C for 5 min. After incubating the cells with the receptor ligands for 5 min at room temperature, cAMP assay was initiated by adding 1 mM isobutylmethylxanthine and terminated by adding 100 μl of 0.2 N HCl. cAMP accumulation was determined by a radioimmunoassay as described previously (Ugur and Onaran, 1997).

Immunocytochemistry and Confocal Microscopy. Cells, grown on glass coverslips, were washed three times with phosphate-buffered saline (PBS) and fixed with 2% paraformaldehyde (w/v) in PBS for 20 min. After permeabilization with 0.1% Triton X-100 in PBS (v/v) for 15 min and blocking for 15 min with 1% bovine serum albumin in PBS (w/v), cells were incubated with an antibody raised against carboxyl terminus decapeptide of Gα_s (see Immunoblots and SDS-Polyacrylamide Gel Electrophoresis for the specification of the antibody) at a dilution of 1:500 for 1 h, washed with PBS, and then incubated with Cy3-conjugated anti-rabbit antibody (1:1000) (Zymax, San Luis Obispo, CA) for 1 h at room temperature. Cells were then washed three times with PBS and once with distilled water, mounted with Immu-Mount reagent (Shandon, Waltham, MA), and visualized by the use of a confocal microscope (TSC SP5; Leica, Wetzlar, Germany).

Modeling and Numerical Simulations. To explain the experimental observations on a quantitative basis we used a modified ternary complex model as schematized in Fig. 1. In the classical ternary complex models, the G protein activation has been considered implicitly as equivalent to the amount of receptor-G protein complex formed, regardless of whether the receptor activation is considered explicitly or not. In the present case however, the G protein activation is considered explicitly as a binary process. Thus, the model given in Fig. 1 is a new interpretation of the well known ternary complex model that has been widely used to explain ligand behavior in different contexts. The reason for such a modification is discussed under Discussion.

Fig. 1.

Description of the equilibrium model where the ternary complex model was extended to include G protein activation (G-G^* transition) explicitly. Left, meanings of the parameters are given schematically: Three binding partners are designated as H, R, and G for ligand, receptor, and G proteins, respectively. Binding sites on the proteins are shown as black circles. Activation of G protein is symbolized as a gray area in the G protein, in which the activation reaction (^*↔0) takes place. Three reaction constants (K, M, and L) for relevant binding (or isomerization) reactions, and three allosteric couplings (α, β, and γ) that links these reactions are indicated on the picture. Among the allosteric factors, γ is a second-order effect that is transmitted between ligand binding and G protein activation reactions once the RG complex is formed. Note that ligand efficacy in the model comprises of the mixed effects of α and γ. Right, equilibrium reactions and corresponding equilibrium constants are shown. Elementary reactions are indicated with thick lines.

In the present scheme, three unconditional affinity constants, K, M, and L, govern ligand-receptor binding, receptor-G protein binding, and state transition of the G protein, respectively. Three allosteric constants, α, β and γ, depict the coupling between the following processes: 1) ligand binding to the receptor and receptor binding to the G protein (α), 2) receptor binding to G protein and G protein activation (β), and 3) ligand binding to receptor-G protein complex and G protein activation (γ). See the left panel of Fig. 1 for a schematic presentation of the affinity and allosteric constants. All these constants can be defined as follows:

Combining the definitions of the above-mentioned reaction constants with the conservation equations for the three components H, R, and G yields the following equations for the corresponding free species:

(1)

(2)

(3)

Rt, Ht, and Gt in eqs. 1, 2, 3 signify the total concentrations of the corresponding components. Given the reaction constants and total concentrations of the three components, we calculated free concentrations of the three components by solving eqs. 1, 2, 3 numerically using an algorithm that has been described previously (Costa et al., 1992; Onaran et al., 1993). This algorithm has been proved to converge to a unique solution vector for the free species (Pradines et al., 2001). Once the concentrations of the free species are thus obtained, the concentrations of all the other species can be readily calculated by using the definitions of the reaction constants given above. See Results for the choice of parameter values.

Other Procedures. The number of living cells was determined by methylthiazolyldiphenyl-tetrazolium bromide (MTT) assay as described by the manufacturer (Sigma-Aldrich, Taufkirchen, Germany). Protein concentration in the membrane preparations was determined by Bradford assay using bovine serum albumin as standard (Bradford, 1976).

All standard reagents (e.g., buffers, salts, detergents) were from Sigma-Aldrich or Thermo Fisher Scientific at appropriate purity. βAR ligands were from Tocris Bioscience (Bristol, UK). Guanine nucleotides were from Roche Diagnostics. [¹²⁵I]iodocyanopindolol and [¹²⁵I]NaI were purchased from Amersham Biosciences (Chalfont St. Giles, Buckinghamshire, UK). β-Adrenoceptor ligands were obtained from the following suppliers: Tocris Biosciences for clenbuterol, cimaterol, procaterol hydrochloride, dobutamine hydrochloride, pronethalol hydrochloride, (S)-(-)-propranolol hydrochloride, sotalol hydrochloride, ICI-188,551 hydrochloride, cyanopindolol hemifumarate, (S)-(-)-pindolol, and ICI-89,406; and Sigma-Aldrich for (-)-isoproterenol hydrochloride, (-)-adrenalin, (-)-alprenolol hydrochloride, timolol maleate, and terbutaline hemisulfate.

Results

Expression and Cellular Localization of Gα_s and XLα_s Proteins. Transfection of HEK293 or 2B2 cells with Gα_sL or XLα_s resulted in considerable overexpression of the proteins in the membrane fractions (Fig. 2A). Note that HEK293 cells endogenously express the long (52-kDa) and the short (45-kDa) forms of Gα_s but not XLα_s (94 kDa), whereas the untransfected 2B2 mouse embryonic fibroblasts express neither Gα_s nor XLα_s because of homozygous ablation of GNAS exon 2. Expression levels of Gα_sL or XLα_s were similar in the 2B2 clones selected for further experiments (Fig. 2B). Because the similarity of the expression levels has a critical importance for the interpretation of data presented in the following sections, we gave the details of the measurement procedure in Supplemental Fig. 1. cAMP response to agonist stimulation in these clones was restored by Gα_sL or XLα_s expression (Fig. 2C).

Fig. 2.

Expression of Gα_sL or XLα_s in HEK293 or 2B2 cells. A, representative Western blots of membrane preparations from transiently transfected HEK293 or permanently cloned 2B2 cells as indicated in the picture; φ, G, and X signify no transfection, transfection with Gα_sL or XLα_s, respectively. B, densitometric analysis of 2B2 cell clones that were used throughout the present experiments. Mean values (bars) were calculated from four independent membrane preparations (for each clone) by using three serial dilutions (5, 2.5, and 1.25 μg/lane) for each preparation. Lines on the bars represent S.E.M. No significant differences were found between Gα_sL and XLα_s signals in these clones (p > 0.05, as assessed by Student's t test). C, intracellular cAMP accumulation in untransfected cells, or in Gα_sL or XLα_s expressing 2B2 cell clones measured in the presence of 0.1 mM isoproterenol. cAMP response were normalized to the number of living cells in the wells as assessed by parallel MTT assays. The numerical value of the MTT signal was ∼0.3 on average. Response in transfected cells is presented as average response as we observed no significant difference between Gα_sL or XLα_s expressing clones. D, cellular distribution of Gα_sL or XLα_s in HEK293 cells that overexpress βAR. Cells were transfected with the cDNA of Gα_sL or XLα_s (as indicated in the picture), and confocal images were obtained after immunostaining by using an antibody against the carboxyl terminus of Gα_s (the same antibody was used in the immunoblotting experiments). Each picture consists of a collage of cell images chosen intentionally to show that the distribution pattern of both Gα_sL and XLα_s exhibits a cell-to-cell variation. Scale bars, 10 μm.

We used confocal microscopy and transiently transfected HEK293 cells to determine subcellular localization of XLα_s or Gα_sL. Endogenous Gα_s in HEK293 cells did not produce a detectable fluorescence signal in untransfected cells, allowing us to distinguish XLα_s (or additional Gα_sL) signal in the transfected HEK293 cells by using an antibody against the common carboxyl terminus of Gα_s and XLα_s. Localization pattern of the proteins differed considerably among cells for both XLα_s and Gα_sL (Fig. 2D): cell membrane, diffuse cytoplasmic, and perinuclear staining were all evident in both cases. Thus, in HEK293 cells we were unable to diagnose any obvious difference between the distribution patterns of Gα_sL and XLα_s as opposed to what has been reported previously (Kehlenbach et al., 1994; Uǧur and Jones 2000; Linglart et al., 2006). Despite the diffuse cytoplasmic staining in some cells we found no XLα_s in soluble fractions of the cell homogenates, which indicated that XLα_s was mostly associated with membranes. Unlike XLα_s, a small fraction of Gα_s could be found in the soluble fractions (data not shown).

Despite the obvious expression of Gα_sL or XLα_s (Fig. 2, A–C) in stably transfected 2B2 clones, we failed to obtain a good quality immunostaining for Gα_sL or XLα_s in these cells because of a high background signal that resulted apparently from the nonspecific interactions of the fluorescent antibodies with some constituents of the 2B2 cells.

Stimulation of Adenylyl Cyclase Activity. We measured cAMP production in the presence or absence of βAR ligands in intact HEK293 cells cotransfected with βAR and XLα_s or Gα_sL. As shown in Fig. 3A, cells transfected with XLα_s showed higher basal and agonist stimulated cAMP accumulation than those transfected with Gα_s, although the expression levels of each G protein α-subunit and the βAR were comparable. In Gα_sL-transfected cells, inverse agonists timolol and ICI-118,551 reduced the basal cAMP levels. In contrast, the basal cAMP level in XLα_s-transfected cells was not responsive to these inverse agonists (Fig. 3A). Thus, we asked whether the elevated basal XLαs activity was independent of receptor coupling. Increasing the level of βAR expression resulted in an increase of basal cAMP accumulation in both Gα_sL- and XLα_s-transfected cells, demonstrating that the high basal activity of XLα_s is associated, at least partly, with the receptor. At the high receptor expression levels, the basal cAMP level in XLα_s-transfected cells remained insensitive to inverse agonists (Fig. 3A, right).

Fig. 3.

Cyclic AMP responses in HEK293 (A) or 2B2 (B) cells in the absence or presence of saturating concentrations of indicated ligands. The level of βAR expression in each cell type is indicated in the picture. Gα_sL and XLα_s expression levels were comparable in these cells as explained in Fig. 2B. cAMP response were normalized to the number of living cells in the wells as assessed by parallel MTT assays. The numerical value of the MTT signal was ∼0.3 on average. Results are mean + S.E.M. values of 10 (A) or eight (B) independent quadruplicate experiments.

One interpretation for this result might be that XLα_s was partially or completely unable to distinguish between the inverse agonist-bound and the empty receptor conformations. However, a more comprehensive scenario can also explain this observation, along with the observations presented below, by a different mechanism (see Discussion).

To observe pure XLα_s response to βAR and to avoid possible interference of endogenous Gα_s, we used the Gα_s-deficient 2B2 cells. As expected, cAMP production of these cells was insensitive to β-adrenergic stimulation. Transfection of these cells with XLα_s or Gα_sL restored the cAMP response, because a small amount of βAR (∼100 fmol/mg) is expressed endogenously (Fig. 2C). We nevertheless stably overexpressed βAR along with XLα_s or Gα_s for examining basal receptor activity, which was otherwise undetectable. Selected clonal cells expressed comparable levels of Gα_sL and XLα_s (Fig. 2B). At ∼1 pmol/mg βAR, the basal cAMP level in XLα_s-transfected cells was unaltered in response to inverse agonists (Fig. 3B, left). When the receptor expression level was increased to ∼5 pmol/mg, an inverse agonist effect of ICI-118,551 emerged in XLα_s-transfected cells as well (Fig. 3B, right), but the magnitude of this effect was small compared with that observed in Gα_sL-transfected cells; ICI-118,551-induced inhibition of basal activity was 65% in Gα_sL but was 30% in XLα_s-transfected cells. These results confirmed the above-mentioned observation that XLα_s-transfected cells were more resistant to inverse agonist effects than Gα_s-transfected cells.

The observation that inverse agonist-induced responses of Gα_sL and XLα_s were different from one another suggested that variation in receptor state that can be induced by different ligands, is also perceived differently by these two proteins. Thus, we systematically screened a set of βAR ligands with a broad spectrum of efficacy for their ability to stimulate cAMP accumulation in Gα_sL- or XLα_s-transfected cells. Figure 4, A and B, shows cAMP responses of Gα_sL- or XLα_s-transfected 2B2 cells that overexpress βAR at a level of ∼1or ∼5 pmol/mg. Overall, relative intrinsic activity of each ligand observed in XLα_s-transfected cells was comparable with that in Gα_sL-expressing cells. Although the basal cAMP accumulation was relatively high in XLα_s-transfected cells (consistent with results presented above), the difference between XLα_s and Gα_sL was less evident upon agonist stimulation. At ∼1 pmol/mg receptor expression, maximal cAMP accumulation was similar in XLα_s- and Gα_sL-transfected cells for strong, but not for partial, agonists (Fig. 4A). When the receptor expression was increased to ∼5 pmol/mg, the similarity between XLα_s- and Gα_s-transfected cells in terms of maximal cAMP accumulation was also observed for the partial agonist dobutamine (Fig. 4B). We were unable to further increase the receptor expression in 2B2 cells. Thus, to address the question as to whether higher receptor expression levels would result in similar levels of XLα_s- and Gα_s-mediated maximal cAMP accumulation even in response to agonists with lower efficacy than dobutamine, we used HEK293 cell clones that express βAR at a level of 30 pmol/mg. At this receptor expression level, the difference between XLα_s- and Gα_sL-mediated cAMP accumulation responses to agonists was no longer detectable: even those agents with very low intrinsic activity, starting from propranolol, which displayed a weak agonistic effect at this level of receptor expression, were able to stimulate XLα_s and Gα_sL equally well (Fig. 4C). This phenomenon is best seen when the responses are shown on normalized scales (Fig. 4D). In Fig. 4D, three categories of ligands are identifiable: 1) ICI-118,551, timolol, and sotalol, for which XLα_s is more efficient than Gα_sL in mediating cAMP production; 2) from propranolol to pindolol, for which XLα_s and Gα_sL are equally efficient; and 3) from partial agonist dobutamine to full agonist clenbuterol, for which overexpression of XLα_s or Gα_s did not increase the response any more than that obtained in HEK293 cells that overexpress βAR only. Together, these results suggested that XLα_s is intrinsically more efficient than Gα_sL in mediating receptor-induced cAMP accumulation, but this phenomenon is observed when the efficacy of ligand and/or the receptor expression level is relatively low.

Fig. 4.

Cyclic AMP responses in the absence or presence of saturating concentrations of indicated ligands in the cells that express comparable amount of Gα_sL or XLα_s but varying level of βAR: 2B2 cells expressing 1 pmol/mg βAR (A), 2B2 cells expressing 5 pmol/mg βAR (B), and HEK293 cells expressing 30 pmol/mg βAR (C). Ligand concentrations were 10^-6 or 10^-4 M depending on ligand's affinity to βAR. Results are mean + S.E.M. values of three to four independent quadruplicate experiment. cAMP response were normalized to the number of living cells in the wells as assessed by parallel MTT assays. Significant differences in A to C are shown with the asterisk (^*), as assessed by Student's test. Two cases were considered as noise: marginal difference (p = 0.06) in sotalol in C and significant difference in clenbuterol in A (p < 0.05). D, increase in cAMP response upon transfection of HEK293 cells (that express 30 pmol/mg βAR) with Gα_sL or XLα_s was given as -fold increase in cAMP response relative to vector transfected cells that express the same amount of βAR in the absence or presence of indicated ligands. The plot is constructed by dividing the data given in C by the response observed in vector transfected cells. Three groups of ligands are indicated in the picture (see text). E, cAMP responses in 2B2 cells that express Gα_sL were plotted against cAMP responses in 2B2 cells that express XLα_s in the absence or presence of following ligands that include inverse agonists and very weak partial agonists: 1, ICI-118,551; 2, timolol; 3, sotalol; 4, propranolol; 5, ICI-89,406; 6, no ligand; 7, pronethalol; 8, cyanopindolol; 9, pindolol; and 10, alprenolol (data are from B). Data are fitted with two straight lines having different slopes. Slopes of the lines differ by a factor of 3.

A more detailed analysis of the response pattern of weak and inverse agonists is given in Fig. 4E for intermediate level of receptor expression, in which the discrepancy between Gα_sL and XLα_s was best seen. In Fig. 4E, maximal cAMP accumulations in the presence of indicated ligands in Gα_sL-transfected 2B2 cells are plotted against those that were obtained in XLα_s-transfected cells, in which it is evident that XLα_s mediates (partial) agonist responses better than Gα_sL but is relatively insensitive to inverse agonist-induced inhibition of basal activity as assessed by the slopes of the lines in Fig. 4E.

Ligand Binding. The above-mentioned observation that XLα_s mediates βAR signaling more efficiently than Gα_sL (at least conditionally) suggests that XLα_s couples better to βAR and/or stimulates adenylyl cyclase more efficiently than does Gα_sL. Although the findings presented above (Fig. 4) are consistent with the former possibility, more direct evidence for this possibility could be obtained by using a ligand binding strategy; the efficiency of receptor-G protein coupling should affect agonist binding pattern, whereas that of G protein-effector coupling is not expected to have a consequence on ligand binding. Therefore, we analyzed agonist binding affinity of βAR in 2B2 cells expressing XLα_s or Gα_sL in the presence or absence of guanine nucleotides. This setting can be considered as an experimental paradigm in the framework of ternary complex-like models to reveal the efficiency of receptor-G protein coupling. As shown in Fig. 5 and Table 1, in the presence of GDP + AlF, isoproterenol bound to βAR with low affinity both in XLα_s- and Gα_sL-expressing cell membranes. In the absence of the guanine nucleotide, the binding isotherm fit to a double-site binding model in both cases. In fact, the low-affinity values estimated from the double-site fit in the absence of the guanine nucleotide were consistent with those estimated in the presence of the guanine nucleotide. Increasing the receptor expression reduced the proportion of high-affinity binding sites and nucleotide-induced shift without affecting the binding affinities (Table 1), which is consistent with the predictions of the ternary complex model. The proportion of high-affinity binding sites was higher in XLα_s-expressing cells than in Gα_sL-expressing cells, and the nucleotide-induced shifts in isoproterenol binding were more pronounced in membrane preparations from XLα_s-expressing cells. Finally, binding affinity for high-affinity binding sites was higher in XLα_s-expressing than in Gα_sL-expressing cells. We obtained similar results in HEK293 cells that overexpress βAR and XLα_s or Gα_sL (Supplemental Fig. 2). Combined, these results show that βAR-G protein coupling is more efficient in the case of XLα_s than Gα_sL.

Fig. 5.

Competition binding curves ([¹²⁵I]-iodocyanopindolol versus isoproterenol) obtained in the membranes from 2B2 cell clones that express comparable amount of Gα_sL or XLα_s and indicated amounts of βAR. Binding curves were obtained in the absence or presence of GDP (10 μM) +AlCl₃ (20 μM) + NaF (10 mM) as indicated in the picture. Binding of hot ligand is given as a fraction of total receptor (Rt) in each case. Solid curves are nonlinear regressions of numerically solved competition binding equations for two binding sites. Areas between binding curves obtained in the absence or presence of nucleotide (ΔAUC) are given in the pictures as a measure of nucleotide-induced shift in binding. See Table 1 for estimated parameters. Data are mean ± S.E.M. values of three independent quadruplicate experiments.

TABLE 1.

Estimated parameters of isoproterenol binding eurves in 2B2 cells that express β AR receptors at two different levels, and XLα_s or Gα_sL at similar levels Parameters are estimated by the regression of numerical solution of competition binding equations assuming a single dissociation constant of 40 pM for [¹²⁵I]iodocyanopindolol. Affinities for isoproterenol were estimated simultaneously by sharing the parameters between two curves obtained in the presence or absence of GDP; AlF (100 μ M; 10 mM). This procedure did not cause a significant worsening of residual variance compared with independent estimation of the parameters (p > 0.05 as assessed by F statistics). Percentage of contributions of high- and low-affinity components are indicated as R_H and R_L and corresponding affinities as K_H and K_L, respectively (see Fig. 5 for the biding curves).

Addition	Parameter	Low βAR		High βAR
		GαsL	XLαs	GαsL	XLαs
	log(KL)	6.3 ± 0.06	6.5 ± 0.06	6.3 ± 0.05	6.4 ± 0.03
	log(KH)	8.1 ± 0.12	8.9 ± 0.05	8.2 ± 0.05	8.7 ± 0.05
GDP + AlF	RL (%)	97 ± 0.2	91 ± 3	95 ± 1	90 ± 4
	RH (%)	3 ± 0.2	9 ± 3	5 ± 1	10 ± 4
No nucleotide	RL (%)	59 ± 5	49 ± 5	78 ± 3	59 ± 3
	RH (%)	41 ± 5	51 ± 5	22 ± 3	41 ± 3
	βAR (pmol/mg)	0.4 ± 0.02	1.0 ± 0.04	4.9 ± 0.53	3.5 ± 0.22

Numerical Simulations. Numerical simulations that are made by using the scheme in Fig. 1 are shown in Fig. 6. Gα_s and XLα_s are simulated as possessing different affinities for the receptor (M = 5 × 10⁹ M^-1 for Gα_s and M = 3 × 10¹¹ M^-1 for XLα_s). Ligands were simulated as follows: inverse agonist (α< 1, γ= 1), neutral ligand (α= 1, γ= 1), and agonist (α> 1, γ>1). For agonism, values of α and γ were chosen equal for simplicity and changed together to simulate partial and full agonists. The value of γ for inverse agonists was set to 1 (see below). For a particular ligand type, values of α and γ were set constant when simulating different types of G proteins. The following parameters were constant for different G protein and ligand types: β= 20, K = 10⁷ M^-1; [G_total] = 10^-10 M; and L = [G^*]/[G] = 0.01, yielding a very low spontaneous G protein activity in the absence of receptor intervention. The first row of Fig. 6A shows the predicted G protein activation depending on receptor density in the absence or saturating presence of three kinds of ligands: an inverse agonist, a partial agonist, or a full agonist. Note that the activation induced by the partial agonist differs for the two G proteins only at low receptor concentrations ([R_t] ≅10^-10 M). This difference disappears at high receptor concentrations ([R_t] ≅10^-9 M), and the activation of the two G proteins by the partial agonist becomes equal at a level below the maximal response of the system (compare the saturation levels obtained in the case of partial and full agonist in the first row of Fig. 6A). This prediction is consistent with the experimental observations presented in Fig. 4, A–C. The second row of Fig. 6A shows the formation of RG complex, which can be considered as what the traditional interpretation of the ternary complex would predict for activation in corresponding situations. In the latter case, as opposed to the former case, the system is predicted to be maximally active when the difference between the two G proteins disappears (compare the saturation levels obtained in the case of partial and full agonist in the second row of Fig. 6A), which was not the case experimentally. The assumption that inverse agonists, as opposed to agonists, do not affect G protein activation directly (γ = 1) leads to the discontinuity observed in the experiments (compare Figs. 6B and 4E). Interestingly, the present model also predicts that the inverse agonist effect should have a maximum (more pronounced for Gα_s than XLα_s) depending on receptor expression, and this was observed experimentally (Fig. 6, C and D). Although not demonstrated numerically in the present report, the entire scenario is compatible with the binding patterns observed in Fig. 5.

Fig. 6.

Numerical simulations by using the model given in Fig. 1. A, G protein activation (i.e., G^* formation) in the presence or absence of saturating concentrations of indicated type of ligands is given in the first row as a fraction of total G (i.e., [G^* + RG^* + HRG^*]/[total G]), depending on receptor concentration. Fractional formation of RG complex in corresponding situations is given in the second row as indicated in the picture (see text for parameter values). B, G protein activity in the presence (or absence) of inverse agonists and weak partial agonists were simulated for XLα_s or Gα_s according to the model described in Fig. 1 at a constant receptor concentration (R_total = 2 × 10^-10 M). Activity (i.e., total G^*/total G) calculated for Gα_s was plotted against the one that was calculated for XLα_s in the presence of different ligands (represented by each dot in the picture). XLα_s and Gα_s were simulated as in A (i.e., with different M values for R). Ligands were simulated as follows: Inverse agonists (α= 0.125, 0.25, and 0.5; γ= 1), neutral antagonist (α= 1; γ= 1), and partial agonists (α=γ= 1.15, 1.32, 1.52, 1.75, 2.01, and 2.31). Values of all other parameters are the same as in A. The position of neutral ligand is indicated with two dotted lines in the picture. Simulated results were in good agreement with the experimental data given in Fig. 4E. C, relative response of an inverse agonist (i.e., activity in the presence of ligand/basal activity) was simulated for XLα_s or Gα_s at varying receptor concentrations as indicated in the picture. XLα_s and Gα_s were simulated as in A and B. Inverse agonist was simulated with the parameters α= 0.1 and γ= 1. D, experimentally observed relative response to inverse agonist ICI-118,551 (i.e., cyclase activity in the presence of ICI-118,551/basal cyclase activity) at three different receptor concentrations in XLα_s-orGα_sL-expressing cells as indicated in the picture. Data in the picture were calculated from the data given Fig. 4, A to C, for ICI-118,551. The simulation in C is in good agreement with the experimental data.

Discussion

We investigated receptor coupling properties of Gα_s and its variant XLα_s, revealing that XLα_s couples βAR signaling to adenylyl cyclase more efficiently than Gα_sL. The difference between XLα_s and Gα_sL was apparently because of the difference between coupling abilities of these G proteins to the receptor.

Gα_s is required for numerous agonist responses. Unlike Gα_s, which is ubiquitous, XLα_s seems to be more abundant in neuroendocrine tissues and brain (Kehlenbach et al., 1994; Pasolli et al., 2000), although XLα_s transcript has been detected in many different tissues (Hayward et al., 1998; Plagge et al., 2004). The phenotypes observed from mice in which either XLα_s or Gα_s is knocked out alone suggest that these two proteins have markedly different physiological roles (Plagge et al., 2004; Chen et al., 2005; Germain-Lee et al., 2005). Although the unique cellular roles of XLα_s remain to be defined, it has been clearly shown that XLα_s can mediate cyclase stimulation in response to receptor activation (Klemke et al., 2000; Bastepe et al., 2002; Linglart et al., 2006). Our results now verify these findings and suggest furthermore that XLα_s may be an important contributor of cAMP signaling, even in tissues where XLα_s levels are markedly lower than Gα_s levels. Consistent with this prediction, XLα_s mRNA is markedly lower than Gα_s mRNA in growth plate chondrocytes, but XLα_s ablation together with the ablation of one Gα_s copy (paternal Gnas exon 2 disruption) results in a more severe phenotype, i.e., premature chondrocyte hypertrophy, than the ablation of one Gα_s copy alone (maternal Gnas exon 2 disruption) (Bastepe et al., 2004). cAMP signaling is involved in a majority of cellular responses, and the superiority of XLα_s over Gα_s in terms of receptor coupling and cAMP generation may thus have important implications in physiology and diseases. Naturally occurring GNAS mutations, with the exception of those located in exon 1, are predicted to affect not only Gα_s but also XLα_s. The changes in XLα_s activity can be involved in the pathogenesis of diseases caused by these mutations, such as various endocrine and nonendocrine tumors (activating) or Albright's hereditary osteodystrophy (inactivating).

In a series of studies, divergent signaling abilities of Gα_sL and Gα_sS (splice variants of Gα_s) have been reported, in which the difference was attributed to a higher rate of dissociation of GDP from Gα_sL than Gα_sS (Seifert et al., 1998; Wenzel-Seifert et al., 2001, 2002). In these studies, receptor-G protein fusions were used to obtain 1:1 stoichiometry of receptor/G protein, which provided a good model for investigating the coupling efficiency between receptor and G protein. In the present work, however, βAR-XLαs fusion protein did not function in 2B2 cells, but it mediated agonist-induced cyclase activation in HEK293 cells (Supplemental Figs. 3 and 4). Interactions of receptor or G protein in the fusion protein with their nonfused partners in the cell membrane have been reported (Burt et al., 1998; Molinari et al., 2003). Hence, the observed discrepancy between HEK293 and 2B2 cells that express βAR-XLα_s fusion protein can be interpreted as follows: the receptor in the fusion construct interacts fruitfully with endogenously expressed Gα_s proteins, which is present in HEK293 but not in 2B2 cells. Hence, no response is observed in 2B2 cells as no intra- or interfusion interaction can actually occur between receptor and XLα_s (Supplemental Fig. 3 for a schematic representation of the idea). Therefore, we were unable to compare XLα_s and Gα_sL in the fusion model. Thus, receptor/G protein stoichiometry was variable (but controlled) in our experiments; this variability, in contrast, eventually proved to be an advantage for the present case (see below).

Strikingly, the observed difference between XLα_s- and Gα_sL was conditional. It disappeared with increasing ligand efficacy, threshold of which was dependent on the expression level of the receptor (Fig. 4). At first sight, this observation implies a saturation effect in the receptor-G protein coupling. However, cyclase activities mediated by XLα_s or Gα_sL became comparable at a level far below the full agonist-induced maximal cyclase activity observable in the cells (Fig. 4, B and C). Hence, this pattern requires further considerations in the framework of the interpretation of ternary complex models that have been used successfully to explain ligand behavior. What is not compatible with this framework is the following: in a system where a receptor couples with different efficiencies to two different G proteins, which in turn, transmit the receptor signal to a unique effector with the same efficiency, the models predict that the ligand-induced effector activations that are mediated by these two G proteins cannot be equal when the stimulated response is below the maximal response of the system. Hence, the models, in their currently interpreted forms, cannot predict the observed equalization of Gα_s- and XLα_s-mediated responses below the maximal level of adenylyl cyclase activation that is achievable in the presence of full βAR agonists. This “defect”, which is not an intrinsic property of the ternary complex models, stems from the assumption that receptor-bound G protein is fully active and can easily be eliminated by assuming that formation of receptor-G protein complex is not necessarily equivalent to full activation of G protein and that agonist binding to receptor can have further “activating” effects on the G protein via conformational changes when receptor and G proteins are bound to each other. Such a scenario has already been suggested for rhodopsin-Gt interaction (Fanelli and Dell'Orco, 2005). Consequences of this assumption become obvious when we introduce G protein activation into the ternary complex model as a simple two-state process (G-active or G-inactive; Fig. 1). Inclusion of a G protein activation step in the ternary interaction scheme inevitably divides the ligand efficacy into two parts: ability of ligand to modify R-G binding (governed by the allosteric constant α in Fig. 1) and to modify G protein activation (i.e., G-G^* transition) in the RG complex (governed by the allosteric constant γ in Fig. 1). On the basis of these fundamentals, the observed behavior of XLα_s and Gα_s in mediating βAR signal can be explained almost entirely by making the following additional assumptions (Fig. 6); XLα_s and Gα_sL differ only in their unconditional tendency to bind receptor (M_XLαs > M_GαsL); efficacy of ligands on βAR does not depend on the identity of G protein to which the receptor is coupled (at least in the case of XLα_s and Gα_sL); spontaneous tendency of RG complex to get activated is relatively low (L << 1 and 1 <β); efficacy of neutral or agonist ligands are evenly distributed over α and γ (for the sake of simplicity); and finally, inverse agonists have no direct effect on G protein activation (i.e., γ= 1) but reduces RG binding (i.e., α< 1). The last assumption is rather speculative but required to explain the inverse agonist effect on βAR-XLα_s coupling compared with Gαs_L. In this framework, the basic mechanism and the scenario that explains the advantage of XLα_s over Gα_sL depending on ligand efficacy and receptor expression can be stated as follows. The origin of difference between the two G proteins is their diverging affinities for the receptor, which is operative (and observable) only when RG interaction is not saturated. The saturation depends on the combined effect of receptor expression and ligand efficacy. Once RG saturation occurs, the difference between XLα_s and Gα_sL disappears, because ligands are assumed to have γ values that do not depend on the kind of G protein. However, even at this saturation point weak agonists do not necessarily induce maximal response of the system (i.e., G-G^* conversion in RG complex may not be complete upon agonist binding depending on the γ value of the agonist) (Fig. 6). This scenario also supports a long debated idea that inactive receptor and G protein tend to form an RG complex that does not necessarily lead to G protein activation, which actually occurs upon activation of receptor in the complex. Accordingly, agonist-induced (or spontaneous) receptor activation directly transmits an activating conformational signal to a precoupled G protein without necessarily affecting the stability of the RG complex. Several lines of indirect evidence support this statement (e.g., Fanelli and Dell'Orco, 2005).

In summary, the present study showed that βAR can couple to XLα_s better than Gα_s and that the rank order of ligand efficacies does not change when it is coupled to XLα_s. Thus, despite the differences observed between Gα_s and XLα_s signaling, no ligand-dependent divergence is predicted to occur between XLα_s and Gα_s in transmitting βAR signal to adenylyl cyclase. Nevertheless, βAR signaling becomes relatively resistant to inverse agonists when the receptor is coupled to XLα_s, which may have pharmacological implications depending on the distribution of βAR-XLα_s coupling in the body. This may justify further studies on βAR-XLα_s coupling (or receptor-XLα_s coupling in general) in physiological integrity. Finally, the experimental system where the two kinds of G proteins are coupled to the same receptor with different efficiencies enabled us to reevaluate the interpretation of the ternary complex model, especially when it is used to explain or understand G protein activation.