Extra-long Gαs Variant XLαs Protein Escapes Activation-induced Subcellular Redistribution and Is Able to Provide Sustained Signaling*

Background: Murine models suggest differences between cellular actions of Gαs and XLαs, but these are unknown. Results: XLαs, unlike Gαs, remains in the plasma membrane and can generate sustained cAMP signaling. Conclusion: The unique actions of XLαs likely stem from its strong affinity for plasma membrane. Significance: Cellular actions of XLαs have implications in cAMP signaling and diseases caused by mutations in this protein. Murine models indicate that Gαs and its extra-long variant XLαs, both of which are derived from GNAS, markedly differ regarding their cellular actions, but these differences are unknown. Here we investigated activation-induced trafficking of Gαs and XLαs, using immunofluorescence microscopy, cell fractionation, and total internal reflection fluorescence microscopy. In transfected cells, XLαs remained localized to the plasma membrane, whereas Gαs redistributed to the cytosol after activation by GTPase-inhibiting mutations, cholera toxin treatment, or G protein-coupled receptor agonists (isoproterenol or parathyroid hormone (PTH)(1–34)). Cholera toxin treatment or agonist (isoproterenol or pituitary adenylate cyclase activating peptide-27) stimulation of PC12 cells expressing Gαs and XLαs endogenously led to an increased abundance of Gαs, but not XLαs, in the soluble fraction. Mutational analyses revealed two conserved cysteines and the highly charged domain as being critically involved in the plasma membrane anchoring of XLαs. The cAMP response induced by M-PTH(1–14), a parathyroid hormone analog, terminated quickly in HEK293 cells stably expressing the type 1 PTH/PTH-related peptide receptor, whereas the response remained maximal for at least 6 min in cells that co-expressed the PTH receptor and XLαs. Although isoproterenol-induced cAMP response was not prolonged by XLαs expression, a GTPase-deficient XLαs mutant found in certain tumors and patients with fibrous dysplasia of bone and McCune-Albright syndrome generated more basal cAMP accumulation in HEK293 cells and caused more severe impairment of osteoblastic differentiation of MC3T3-E1 cells than the cognate Gαs mutant (gsp oncogene). Thus, activated XLαs and Gαs traffic differently, and this may form the basis for the differences in their cellular actions.


Murine models indicate that G␣s and its extra-long variant
XL␣s, both of which are derived from GNAS, markedly differ regarding their cellular actions, but these differences are unknown. Here we investigated activation-induced trafficking of G␣s and XL␣s, using immunofluorescence microscopy, cell fractionation, and total internal reflection fluorescence microscopy. In transfected cells, XL␣s remained localized to the plasma membrane, whereas G␣s redistributed to the cytosol after activation by GTPase-inhibiting mutations, cholera toxin treatment, or G protein-coupled receptor agonists (isoproterenol or parathyroid hormone (PTH)(1-34)). Cholera toxin treatment or agonist (isoproterenol or pituitary adenylate cyclase activating peptide-27) stimulation of PC12 cells expressing G␣s and XL␣s endogenously led to an increased abundance of G␣s, but not XL␣s, in the soluble fraction. Mutational analyses revealed two conserved cysteines and the highly charged domain as being critically involved in the plasma membrane anchoring of XL␣s. The cAMP response induced by M-PTH(1-14), a parathyroid hormone analog, terminated quickly in HEK293 cells stably expressing the type 1 PTH/PTH-related peptide receptor, whereas the response remained maximal for at least 6

min in cells that co-expressed the PTH receptor and XL␣s. Although isoproterenol-induced cAMP response was not prolonged by XL␣s expression, a GTPase-deficient XL␣s mutant found in certain tumors and patients with fibrous dysplasia of bone and McCune-Albright syndrome generated more basal cAMP accumulation in HEK293 cells and caused more severe impairment of osteoblastic differentiation of MC3T3-E1 cells than the cognate G␣s mutant (gsp oncogene). Thus, activated XL␣s and G␣s traffic differently, and this may form the basis for the differences in their cellular actions.
GNAS is a complex locus giving rise to multiple translated and nontranslated gene products (1)(2)(3)(4). Gene association and copy number variation studies have associated GNAS with the pathogenesis of multiple different complex diseases and cancers (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15). Inherited mutations within or nearby GNAS that directly impair the functions and/or the expression of its gene products are responsible for several different human diseases, including, but not limited to, pseudohypoparathyroidism, various endocrine and nonendocrine tumors, and McCune-Albright syndrome (16,17). One of the products of GNAS is the ␣-subunit of the heterotrimeric stimulatory GTP-binding protein (G␣s), a ubiquitous protein essential for the actions of many hormones, neurotransmitters, and paracrine factors (1). The GTP-bound form of the G␣s subunit transduces the activation of cell surface G protein-coupled receptors (GPCRs) 3 into intracellular signaling by stimulating a number of effectors, such as adenylyl cyclases, which catalyze the synthesis of cAMP.
GNAS also encodes a long G␣s variant, termed XL␣s ( Fig.  1A) (18), whose cellular actions remain uncertain. XL␣s uses a promoter that is distinct from the promoter of G␣s and is active on the paternal allele only, i.e. XL␣s expression is limited to a single parental allele (19). Although G␣s is encoded by GNAS exons 1-13, XL␣s uses an alternative first exon (exon XL) that splices onto exons 2-13 (19,20). Thus, G␣s and XL␣s differ in their N-terminal regions but are otherwise identical. A variant of XL␣s, termed XXL␣s, is also derived from GNAS (Fig. 1A) (21,22). The latter includes the entire coding sequence of XL␣s but extends in the N terminus, thus having about 300 additional amino acids. Like XL␣s, the cellular actions of XXL␣s are unclear.
Studies with transfected cell lines have shown that XL␣s can mimic the action of G␣s regarding the stimulation of cAMP generation in response to receptor activation or upon treatment with cholera toxin (CTX), which ADP-ribosylates the ␣-subunit and thus inhibits the intrinsic GTPase activity (23)(24)(25). When expressed ectopically in transgenic mice, XL␣s can also enhance cellular responses that are typically mediated by G␣s (26). However, data obtained from gene knock-out studies do not readily support a "G␣s-like" role for XL␣s at the cellular level. Mice in which XL␣s (together with XXL␣s) is ablated show poor adaptation to feeding, early postnatal lethality, and a hypermetabolic phenotype (27)(28)(29)(30), but these findings are strikingly different from and, in terms of energy and lipid metabolism, the opposite of the findings observed in G␣s knock-out mice (31,32). Thus, although XL␣s can seemingly contribute to cAMP signaling, its cellular actions are predicted to differ significantly from the cellular actions of G␣s.
Like the G␣s subunit, the XL␣s subunit is localized to the plasma membrane at the basal state (25,33). G␣s redistributes to the cytoplasmic compartments following activation (34,35), and this regulatory process is critical for limiting the activation of G␣s (36,37). However, recent studies have also shown that, in response to certain receptor agonists, the internalized G␣s protein can continue to stimulate cAMP production within endosomes (38,39). The fate of XL␣s upon activation, however, has remained unknown. A previous study using subcellular fractionation detected differences between the distributions of ADP-ribosylated forms of XL␣s and G␣s (33), and in another study, a GTPase-deficient XL␣s mutant was localized differently from the cognate G␣s mutant by immunostaining (22). Based on these findings, we hypothesized that XL␣s traffics differently from G␣s upon activation. We herein investigated this hypothesis by examining the subcellular localization of XL␣s before and after activation. Our findings revealed that XL␣s remain localized to the plasma membrane even upon activation and can thereby generate sustained signaling.

EXPERIMENTAL PROCEDURES
Expression Constructs and Mutagenesis-Construction of cDNA plasmids encoding hemagglutinin (HA) epitope-tagged G␣s, XL␣s, and XXL␣s, all in pcDNA3.1, were described previously (22,34). The truncation and point mutation constructs of G␣s, XL␣s, and XXL␣s were generated using circular wildtype cDNA plasmid as a template and specific mutagenic oligonucleotides as primers by using the QuikChange mutagenesis kit (Stratagene). cDNA encoding the XL␣s-G␣s chimera in which residues 247-318 of human XL␣s replaced Gly-2 and Cys-3 of G␣s was generated by standard methods and cloned into pcDNA3.1. All plasmid DNAs were sequenced to verify the presence of desired mutations and that they were free of aberrant random mutations. Restriction endonucleases and other enzymes for making constructs were obtained from New England Biolabs (Beverley, MA). Plasmids encoding G␣s-YFP, -G␤ 1 , and -G␥ 2 were kindly provided by Dr. Matthew Mahon (Massachusetts General Hospital and Harvard Medical School). The plasmid encoding G␣s-GFP was kindly provided by Dr. Catherine Berlot (Geisinger Health, Weis Center for Research). The plasmid encoding XL␣s-GFP was constructed by substituting, in the G␣s-GFP construct, the sequences encoding the XL domain for the sequences of G␣s derived from exon 1. The plasmid encoding the PTHR-DsRed fusion protein (PTHR-DsRed) was constructed by inserting cDNA encoding DsRed into the exon 2 encoded portion of PTHR.
Cell Culture and Transient Transfection-Cells were maintained at 37°C in a humidified atmosphere containing 5% CO 2 . HEK293, PC12, and MC3T3-E1 cells were obtained from American Type Culture Collection (ATCC; Manassas, VA). HEK293 cells were cultured in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum. PC12 cells were grown on collagen type IV-coated culture plates in ATCC-formulated F-12K medium supplemented with 15% fetal horse serum and 2.5% fetal bovine serum. Fibroblastic Gnas E2Ϫ/E2Ϫ cells null for G␣s, XL␣s, and XXL␣s, which have been described previously (24), were cultured in DMEM/F-12 medium supplemented with 5% FBS. MC3T3-E1 cells were maintained in ␣-minimal essential medium.
For determining and comparing subcellular localizations of G␣s, XL␣s, and XXL␣s and mutants thereof, and for experiments involving the quantification of cAMP levels, cells were additionally transfected with plasmids encoding G␤ 1 and G␥ 2 . HEK293 cells were transfected by using FuGENE 6 (Roche Applied Science) and Gnas E2Ϫ/E2Ϫ cells by using JetPEI DNA transfection reagent (PolyPlus transfection TM , Illkirch, France) following the protocols supplied by the manufacturer. Stimulation of cells with isoproterenol or pituitary adenylate cyclase activating peptide-27 (PACAP) was done for 20 min at 37°C. CTX stimulation was performed for 4 h at 37°C. Isoproterenol, PACAP, and CTX were purchased from Sigma. Preosteoblastic MC3T3-E1 cells were transfected by using the PolyJet transfection reagent (SignaGen Laboratories).
Cell Lysis and Subcellular Fractionation-Whole cell lysates were prepared by either 2ϫ SDS-polyacrylamide gel loading buffer or 1% Triton X-100 in a Tris-HCl (pH 7.8)-buffered solution containing protease inhibitors. For preparation of total cell membranes, cells were lysed in isotonic buffer without detergent (10 mM Tris-HCl (pH 7.8), 4 mM EDTA, and protease inhibitors) by passing 10 -15 times through a 28-gauge syringe tip on ice. After 10 min of centrifugation at 1,000 ϫ g at 4 ºC, the resulting supernatant was further centrifuged at 100,000 ϫ g (Beckman TL-100 Tabletop Ultracentrifuge, Beckman, Palo Alto, CA) for 1 h at 4 ºC. The supernatant after the second centrifugation was designated as the soluble fraction. The pellet was resuspended in a buffer containing 20 mM Hepes (pH 7.4), 0.1 M NaCl, 3 mM MgSO 4 , and 20% glycerol. The pellet obtained after the ultracentrifugation was designated as the total membrane fraction (particulate fraction). Protein concentration was determined by BCA protein assay kit (Pierce) using bovine serum albumin as standard.
Western Blot Analysis-The particulate and soluble fractions of cell lysates were subjected to electrophoresis for separation of proteins by either 10% (G␣s) or 6% (for XL␣s and XXL␣s) SDS-PAGE. Whole cell lysates were separated by 4 -15% gradi-ent SDS-PAGE. Separated proteins were transferred onto Immobilon PVDF membranes (Millipore, Temecula, CA). After blocking with 5% nonfat milk in Tris-buffered saline, 0.1% Tween 20 for 1 h at room temperature, the blots were probed with primary antibodies, either rabbit anti-HA antibody (AbCam, Cambridge, MA) or a rabbit antibody against the C-terminal decapeptide common to G␣s, XL␣s, and XXL␣s (Millipore). The immunoblots were then reacted to goat antirabbit IgG secondary antibody conjugated to horseradish peroxidase (HRP) (Santa Cruz Biotechnology), and immunoreactive proteins were visualized using Western Lightning Plus-ECL enhanced chemiluminescence detection kit (PerkinElmer Life Sciences). Blots were stripped by using the Re-Blot Plus solution (Millipore), and subsequently immunoreacted to a polyclonal antibody against ␤-actin (Santa Cruz Biotechnology) as a gel loading control. Densitometric analysis of blots was carried out by using FluorChem SP imaging system and Alpha-easeFC software version 4.1.0 (Alpha Innotech, San Leandro, CA).
Immunocytochemistry and Confocal Microscopy-Cells were grown and transfected in collagen-coated, four-well chamber slides with cover (Nunc, Naperville, IL). Cells were washed three times with phosphate-buffered saline (PBS) and fixed with 4% paraformaldehyde in PBS for 20 min. After permeabilization with 0.1% saponin in PBS, 0.5% BSA for 15 min and subsequently blocking for 30 min with 0.1% saponin and 0.5% BSA in PBS, cells were incubated with a rabbit anti-HA antibody (Santa Cruz Biotechnology) and then incubated with Cy3conjugated anti-rabbit IgG (Amersham Biosciences). The immunoreactivity was visualized and analyzed by using a laser scanning confocal fluorescent microscope (Nikon, Tokyo, Japan).
Analysis of Protein Palmitoylation-To inhibit palmitoylation, HEK293 cells were cotransfected with expression constructs encoding XL␣s-GFP, G␤ 1 , and G␥ 2 , and immediately following transfection, these cells were treated with 25 M 2-bromo-hexadecanoic acid (2BP) or vehicle (DMSO) alone. The subcellular localization of the protein was evaluated by confocal fluorescent microscopy 24 h after transfection or by subcellular fractionation followed by Western blot 48 h after transfection. For direct determination of XL␣s palmitoylation, cells transiently expressing either wild-type XL␣s or the XL␣s-C287S,C318S mutant were metabolically labeled with 0.16 mCi/ml [ 3 H]palmitic acid ([9,10-3 H]palmitic acid (PerkinElmer Life Sciences)) in Dulbecco's modified Eagle's medium supplemented with 0.2% fatty acid-free bovine serum albumin at 37 ºC for 5 h. After washing with PBS, cells were lysed in PBS containing 5 mM EDTA, 1% Triton X-100, and a protease inhibitor mixture (Sigma). Wild-type and the mutant XL␣s were immunoprecipitated by a mouse monoclonal anti-HA antibody (Abcam). Following separation by 10% SDS-PAGE, proteins were analyzed by fluorography to detect the emission of 3 H. Briefly, gels were fixed and processed by using the EN 3 HANCE reagent (PerkinElmer Life Sciences) according to the manufacturer's instructions. Dried gels were exposed to a Kodak Bio-Max MS x-ray film for at least 6 weeks at Ϫ80 ºC. In some experiments, immunoprecipitated proteins were separated on two parallel 7.5% SDS-polyacrylamide gels. One of the gels was blotted onto an Immobilon PVDF membrane and immunoreacted to the polyclonal antibody against the G␣s/XL␣s/XXL␣s C terminus. The other gel was sliced according to the position of the immunoreactive bands in the Western blot. Radioactivity in each gel slice was counted in a Beckman scintillation counter. Background was determined by counting the radioactivity in blank gel slices.
TIRF Microscopy-HEK293 cells were transiently transfected with cDNA encoding G␣s-GFP or XL␣s-GFP, as well as PTHR-DsRed, G␤ 1 , and G␥ 2 , and 48 h later, cells were plated on Fluo-roDish (World Precision Instruments, Inc.) coated with poly-Llysine. The following day, cells were washed two times with Hanks' balanced salt solution (Invitrogen) and stimulated with 10 nM [Nle 8 ,Nle 21 ,Tyr 34 ]rPTH(1-34) (PTH(1-34)). Fluorescence measurements of single cells were performed using a total internal reflection (TIRF) objective, a motorized laser TIRF illumination unit as the TIRF microscopy condenser (Nikon), and argon laser (Nikon). GFP was excited with 488 nm laser line using 530 nm emission filter. The fluorescent images were recorded for 30 min after PTH(1-34) stimulation, and the intensity of green emission fluorescence was recorded. The intensity of green fluorescence was measured and analyzed by NIS Elements software (Nikon). To minimize photobleaching during the experiment, recording was performed at 5-s intervals in the first 5 min, 10-s intervals from 5 to 20 min, and 30-s intervals from 20 to 30 min.
Determination of cAMP Response-Basal cAMP accumulation was determined 72 h after transfection of HEK293 cells. Cells were lysed after incubation in a buffer containing 2 mM isobutyl methylxanthine (Sigma) for 15 min at 37°C. The medium was removed, and cells were lysed with 50 mM HCl. Radioimmunoassay was performed to determine the amount of cAMP. For determining the time course of cAMP generation in live cells, a FRET-based assay was used, as described previously (38). HEK293 cells stably expressing PTHR and transiently expressing a cAMP biosensor, Epac-CFP/YFP with or without XL␣s were continuously perfused with control buffer, 100 nM M-PTH(1-14) or 10 M isoproterenol; the details of the "M" substitution and the signaling properties of this PTH analog has been described previously (40,41).
Osteoblastic Differentiation-Preosteoblastic murine MC3T3-E1 cells grown in 24-well plates were transfected with various expression plasmids. Forty eight hours after transfection, growth medium was replaced (day 0) with osteogenic differentiation medium containing ascorbic acid (50 g/ml). Differentiation of cells into osteoblasts was assessed on days 0, 2, and 5 by staining for alkaline phosphatase activity, which was performed on cells fixed with 10% formalin. After washing with PBS, cells were incubated for 30 min at room temperature with a 0.1 M Tris-HCl (pH 8.5)-buffered solution containing 0.01% naphthol AS-MX phosphate (Sigma) and 0.06% Fast Blue BB salt (Sigma). RNA was also isolated on the same day by using the RNeasy mini kit (Qiagen), and first strand cDNA was synthesized by using Superscript III reverse transcriptase (Invitrogen) and random hexamer primers. To quantify the level of alkaline phosphatase transcript, real time RT-PCR using the SYBR Green reagent (Qiagen) was performed. The ␤-actin transcript was also amplified as an internal control. Forward and reverse PCR primers for amplification of the alkaline phosphatase transcript were 5Ј-AACCCAGACACAAGCATTCC-3Ј and 5Ј-CGAAGGGTCAGTCAGGTTGT-3Ј, respectively. Primers for amplification of ␤-actin were described previously (42).
Statistical Analyses-Statistical significance of difference between two sample means was determined by paired Student's t test. Statistical significance of difference among multiple sample means was determined by one-way analysis of variance. A p value of less than 0.05 was considered to be significant.

RESULTS
XL␣s Is Targeted Differently from G␣s upon Activation-To determine whether XL␣s mimics G␣s regarding activation-induced subcellular redistribution, we transfected HEK293 cells with cDNA encoding HA-tagged XL␣s or G␣s. Immunofluorescence confocal microscopy using an anti-HA antibody demonstrated that both proteins were localized to the plasma membrane at the basal state, although the staining for XL␣s appeared to be punctate (Fig. 1B). In cells treated with a saturating concentration of isoproterenol (10 M), a selective full agonist for the endogenous ␤ 2 AR (a class A GPCR; ␤ 2 AR), or with 1 g/ml cholera toxin, which directly stimulates G␣s by inhibiting its intrinsic GTPase activity, G␣s was detected at intracellular localizations, whereas XL␣s staining remained in the plasma membrane (Fig. 1B). In the same assays, the extended XL␣s variant XXL␣s behaved similarly to XL␣s with respect to subcellular localization before and after activation (supplemental Fig. S1). Using either anti-HA antibody (for G␣s) FIGURE 1. XL␣s is targeted differently from wild-type G␣s upon activation. A, depiction of the GNAS locus and the transcripts encoding G␣s, XL␣s, and XXL␣s. Exons and introns are indicated by boxes and connecting lines, respectively. Splicing pattern is shown by angled lines. Arrows indicate the origin of transcription. B and C, HEK293 cells were transiently transfected with cDNA encoding either HA-tagged G␣s or XL␣s. Forty eight hours after transfection, cells were treated with 10 Ϫ5 M isoproterenol for 20 min or with 1 g/ml CTX for 4 h, and subcellular localizations of G␣s and XL␣s were examined by immunocytochemical analysis by using the anti-HA antibody (B), and Western blot analysis by using either the anti-HA antibody or a polyclonal antibody against the common C terminus of G␣s and XL␣s. (C). D, HEK293 cells were transiently co-transfected with cDNA encoding HA-tagged XL␣s and G␣s-YFP. Forty eight hours after transfection, cells were treated with 10 Ϫ5 M isoproterenol (ISO) for 20 min or with 1 g/ml CTX for 4 h. XL␣s and G␣s-YFP were detected in the same cells by immunocytochemical analysis using the anti-HA antibody and by confocal fluorescence microscopy. E, Gnas E2Ϫ/E2Ϫ cells were transiently transfected with cDNA encoding HA-tagged G␣s-R201H, XL␣s-R543H, or XXL␣s-R844H. Two days later, immunofluorescence confocal microscopy using anti-HA antibody was employed to determine the subcellular localization of those mutants. F and G, subcellular localizations of endogenous G␣s and XL␣s in PC12 cells stimulated by 1 g/ml CTX (F), 10 M isoproterenol (ISO; G), or 100 M PACAP (G). Endogenous G␣s and XL␣s in soluble (S) and particulate (P) fractions were subjected to immunoblotting with the polyclonal antibody against G␣s and XL␣s. or a polyclonal antibody directed against the C terminus of XL␣s and G␣s (for XL␣s), Western blot analysis of particulate and soluble fractions of these transfected HEK293 cells also showed that at the basal state G␣s exists mostly within the particulate fraction and that stimulation of these cells with isoproterenol or cholera toxin results in a modest redistribution to the soluble fraction (Fig. 1C). In contrast, XL␣s immunoreactivity was detected entirely in the particulate fraction both at the basal state and upon stimulation with isoproterenol or cholera toxin (Fig. 1C). Findings similar to those observed for XL␣s were also obtained when using cells transfected with XXL␣s (supplemental Fig. S1).
We then co-transfected HEK293 cells with cDNA encoding HA-tagged XL␣s or XXL␣s and a YFP-labeled G␣s subunit (G␣s-YFP). At the basal state, immunocytochemical analysis using the anti-HA antibody and fluorescence confocal microscopy revealed both G␣s-YFP and XL␣s at the plasma membrane, but after stimulation by isoproterenol or cholera toxin, only G␣s-YFP became cytoplasmic (Fig. 1D). In the same cells, immunostaining for XL␣s remained at the plasma membrane (Fig. 1D). The subcellular localization of XXL␣s relative to that of G␣s-YFP was identical to that of XL␣s. It remained plasma membrane bound upon activation as opposed to G␣s-YFP, which internalized when stimulated by isoproterenol or cholera toxin (supplemental Fig. S1).
To rule out that the activation-induced differences between the subcellular localizations of G␣s and XL␣s reflect excessive amounts of total G␣s expression in transfected HEK293 cells, which endogenously express G␣s (but not XL␣s) at readily detectable levels, we employed mouse embryonic fibroblasts that genetically lack endogenous G␣s, XL␣s, and XXL␣s (Gnas E2Ϫ/E2Ϫ cells) expression (24). We transfected these cells with cDNA encoding a GTPase-deficient, constitutively active mutant of either G␣s (R201H) or XL␣s (R543H), which was used as a substitute for receptor-or cholera toxin-induced activation. Immunostaining of these cells with the anti-HA antibody also showed that G␣s-R201H is localized to the cytoplasm, whereas constitutively active XL␣s-R543H, as well as the cognate XXL␣s mutant (R844H), is localized to the plasma membrane (Fig. 1E). Moreover, we compared the subcellular distribution of G␣s and XL␣s in PC12 cells, a rat pheochromocytoma cell line that expresses both XL␣s and G␣s endogenously. Western blot analysis of nonstimulated PC12 cells using the C-terminal G␣s/XL␣s antibody showed that whereas G␣s is localized mostly, but not exclusively, to the particulate fraction, XL␣s is entirely confined to the particulate fraction (Fig. 1F). Following treatment of PC12 cells with cholera toxin, the abundance of G␣s shifted partly toward the soluble fraction, whereas XL␣s continued to be localized exclusively in the particulate fraction (Fig. 1F). In addition, stimulation of PC12 cells with either isoproterenol or PACAP, which bind their respective endogenously expressed G␣s-coupled receptors, also resulted in a modest shift of G␣s immunoreactivity from the particulate to the soluble fraction, whereas these receptor agonists failed to alter the association of XL␣s with the particulate fraction (Fig.  1G). These findings indicated that XL␣s (as well as XXL␣s), unlike G␣s, is not subject to internalization upon activation.
XL␣s can functionally couple to PTHR (24,25), which belongs to the family of class B GPCRs. To determine whether the differences observed between the subcellular localizations of G␣s and XL␣s also exist following activation of PTHR, we transiently coexpressed PTHR and either G␣s-GFP or XL␣s-GFP in HEK293 cells, which were then stimulated with PTH(1-34) (10 nM) and monitored for 30 min by TIRFM. The changes in the intensity of green fluorescence were strikingly different between cells expressing G␣s-GFP and those expressing XL␣s-GFP. The fluorescence intensity diminished by about 50% at the end of the 30-min time course, with a half-life of ϳ5 min, in cells expressing G␣s-GFP (Fig. 2, A and C), whereas it remained constant in cells expressing XL␣s-GFP (Fig. 2, B and C) (supplemental videos). These findings confirmed that XL␣s remains attached to the plasma membrane despite being activated.
Conserved Cysteine Residues and a Region Comprising HCD (but not PRR) Confer Strong Plasma Membrane Avidity to XL␣s and XXL␣s-To determine the structural features of XL␣s and XXL␣s that prevent activation-induced internalization, we generated a series of N-terminally truncated mutants, and we compared the subcellular localization of each mutant. The truncations started with deletion of the N-terminal residues, including the PRR, and gradually omitted the HCD and the residues before the putative G␤␥ interaction domain (corresponding to the exon 1-encoded portion of G␣s) (Fig. 3A). The subcellular localization of each truncated mutant was determined in transiently transfected HEK293 cells by using Western blot analyses using the anti-HA antibody following subcellular fractionation and confocal immunofluorescence microscopy. Truncation of the N-terminal residues, including PRR, but not HCD did not affect subcellular localization, as this mutant (NpostPRR) was detected almost exclusively in the particulate fraction (Fig. 3, B and C). However, further removal of the region extending from the C-terminal end of PRR to the C-terminal end of HCD augmented the percentage of soluble protein from 3 Ϯ 0.5 to 25 Ϯ 0.5% of total (NpostPRR versus NpostHCD; p Ͻ 0.001) (Fig. 3, B and C). With more truncation of the XL domain sequences, increasingly more recombinant proteins were detected in the soluble fraction. Comparing these results with those obtained from the truncation mutants carrying point mutations that are analogous to G␣s-R201H (i.e. GTPase inhibiting), we observed a significant difference in the localization of the mutant missing not only PRR and HCD but also both of the conserved cysteines (NpostCys). Whereas 29 Ϯ 0.8% of total NpostCys was detected in the soluble fraction, 48 Ϯ 0.3% of its GTPase-deficient form was in the same fraction (p Ͻ 0.001; Fig. 3, B and C). This finding, however, was not confirmed by confocal immunofluorescence microscopy, which detected both of these mutants in the cytoplasm (Fig.  3D). This discrepancy likely reflected localization of NpostCys in intracellular vesicles. Overall, these studies did not reveal clear differences between the native and GTPase-deficient forms of the truncation mutants and thus prevented us from identifying a distinct domain(s) responsible solely for keeping activated XL␣s and XXL␣s proteins in the plasma membrane.
These data, however, made it clear that the region comprising the two conserved cysteines has an important role in the plasma membrane targeting of XL␣s and XXL␣s at the basal state. Cysteines are targets for palmitoylation, and palmitoylation is a critical post-translational modification for membrane attachment (43). To determine whether these cysteine residues were required for the membrane localization of XL␣s, the two cysteine residues were mutated into serines individually or together (i.e. XL␣s-C287S, XL␣s-C318S, and XL␣s-C287S,C318S)). Immunofluorescence confocal microscopy demonstrated that, although mutation of each cysteine alone did not affect the membrane targeting, mutations of both cysteines to serines resulted in a marked reduction of plasma membrane attachment (Fig. 4A). When analogous mutations were introduced into full-length XXL␣s (i.e. XXL␣s-C589S, XXL␣s-C619S, XXL␣s-C589S,C619S), similar results were obtained (supplemental Fig. S2). Additional introduction of the GTPase-inhibiting mutation analogous to G␣s-R201H into these Cys-to-Ser mutants also yielded similar results ( Fig. 4A; supplemental Fig. S2). Western blots using the anti-G␣s C-terminal antibody verified these findings when analyzing soluble and particulate fractions of cell lysates ( Fig. 4B; supplemental Fig. S2). These findings indicated that at least one of these conserved cysteines is necessary for the plasma membrane targeting of XL␣s and XXL␣s. We then tested the effect of 2BP, an inhibitor of protein palmitoylation, on the subcellular localization of XL␣s. Treatment of cells with 2BP resulted in retardation of XL␣s in the cytoplasm, as determined by fluorescence microscopy of fixed HEK293 cells transiently expressing XL␣s-GFP (Fig. 4C) and Western blot analysis of lysates from HEK293 cells transiently expressing either native XL␣s or XL␣s-GFP (Fig. 4D). In addition, based on metabolic labeling experiments using radiolabeled palmitic acid, wild-type XL␣s, but not the XL␣s-C287S,C318S mutant, was palmitoylated in HEK293 cells transiently expressing these proteins (Fig. 4E). Together, these results indicated that protein palmitoylation plays a critical role in the plasma membrane targeting of XL␣s (and XXL␣s) at the steady state.
Our truncation experiments indicated that a 72-amino acid segment of XL␣s extending from the C-terminal end of PRR to  NOVEMBER 4, 2011 • VOLUME 286 • NUMBER 44 the second conserved cysteine (residues 247-318 according to XL␣s) is critical with respect to plasma membrane targeting. Cys-318 is homologous to G␣s Cys-3, which has been shown to undergo palmitoylation and to be important for subcellular targeting (44,45). We therefore generated an XL␣s-G␣s chimera in which the 72-amino acid region of XL␣s replaced Gly-2 and Cys-3 of G␣s (Fig. 5A). In transfected HEK293 cells, the XL␣s-G␣s chimera was localized to the plasma membrane. Importantly, it remained localized to the plasma membrane upon activation by cholera toxin treatment or by introduction of a GTPase inhibiting mutation analogous to G␣s-R201H (Fig. 5, B and C), thus indicating that the structural features sufficient to anchor XL␣s (and XXL␣s) in the plasma membrane, even upon activation, are included in this 72-amino acid segment including HCD and the conserved cysteine residues.

XL␣s Escapes Activation-induced Subcellular Redistribution
XL␣s Extends the Duration of cAMP Response Induced by PTHR Activation-Activation-induced subcellular redistribution of G␣s serves as a regulatory mechanism that limits the time-course of cAMP generation at the plasma membrane (36,38). We therefore reasoned that the absence of XL␣s internalization could prolong the duration of cAMP generation in response to receptor activation. We addressed this hypothesis by using a Förster resonance energy transfer (FRET)-based reporter that permits real time recording of cAMP production in live cells (46). As an agonist, we chose a PTH analog, M-PTH (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14), which has been shown to induce a short lived cAMP response (41). As expected, a rapid change in the FRET signal was observed upon the addition of 100 nM M-PTH (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14) onto HEK293 cells expressing the PTHR alone or in combination with XL␣s and G␤ 1 ␥ 2 (Fig. 6). After ligand washout, the FRET signal returned back to its initial level in control cells,  A, diagram of the XL␣s-G␣s chimera protein, with the 72-amino acid segment of XL␣s spanning the two cysteines to replace the N-terminal 2 residues of G␣s. PRR, proline-rich region; HCD, highly charged domain; ␤␥, putative G␤␥ interaction domain. Asterisks indicate the conserved cysteine residues. B, immunocytochemical analysis using the anti-HA antibody was performed to study the subcellular distribution of the XL␣s-G␣s chimera in HEK293 cells transfected with cDNA encoding HA-tagged XL␣s-G␣s chimera or a constitutively active (CA) form of the chimera carrying a GTPase inhibiting mutant analogous to R201H. CTX was also used to stimulate the native form of the chimera. C, Western blot analysis using the anti-HA antibody for comparing the subcellular localizations of G␣s, the XL␣s-G␣s chimera, and the GTPase-deficient form of the chimera transiently expressed in HEK293 cells. S, soluble fraction; P, particular fraction.
whereas it remained at the maximal level at least for 6 min after washout in XL␣s-expressing cells (Fig. 6). Thus, XL␣s extended the duration of the cAMP response induced by the typically short acting agonist M-PTH (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14). Although this finding was consistent with the absence of XL␣s internalization upon activation, a similar effect was not observed when the same cells were stimulated by isoproterenol. The kinetic profile of the cAMP response elicited by isoproterenol in cells expressing XL␣s appeared indistinguishable from that in control cells (Fig.  6), indicating that the ability of XL␣s to prolong agonist-induced cAMP response is receptor-specific.
Constitutive XL␣s Activity Is Markedly More Effective Than Constitutive G␣s Activity-GTPase inhibiting mutations of GNAS cause certain endocrine and nonendocrine tumors and McCune-Albright syndrome (16,17). Because these mutations can affect both G␣s and XL␣s and lead to constitutive cAMP production (23,47), we tested whether a GTPase-deficient XL␣s mutant (R543H) would be more effective in mediating basal cAMP accumulation than the cognate G␣s mutant (R201H). In transfected HEK293 cells, XL␣s-R543H showed significantly (p Ͻ 0.001) higher basal cAMP accumulation compared with G␣s-R201H (Fig. 7A). Basal cAMP accumulation was also significantly (p Ͻ 0.05) higher in cells transiently expressing the GTPase-deficient version of the XL␣s-G␣s chimera than G␣s-R201H, consistent with the plasma membrane localization of the former; however, the amount of cAMP accumulated in cells expressing XL␣s-R543H was still significantly higher (p Ͻ 0.05) than that in cells expressing the GTPase-deficient XL␣s-G␣s chimera (Fig. 7A). In contrast, the XL␣s mutant carrying both C287S,C318S substitutions and an analogous GTPase-deficient mutation, which showed poor plasma membrane localization (see Fig. 4), displayed much lower basal activity than any of the other constitutively active proteins (Fig.  7A). Western blot analyses using whole cell lysates and anti-HA antibody showed that the immunoreactivity for each of these proteins was comparable with one another, indicating that the observed differences in cAMP accumulation were unlikely to reflect total expression levels (Fig. 7B).
G␣s mutants carrying GTPase inhibiting mutations inhibit differentiation of osteoblasts, as seen in patients with fibrous dysplasia of bone (48,49). We thus compared the effects of G␣s-R201H and XL␣s-R543H expression on preosteoblastic MC3T3-E1 cells, in which the cAMP signaling pathway inhibits osteoblastic differentiation (50). When transfected MC3T3-E1 cells were grown under osteogenic conditions for 5 days, transient expression of G␣s-R201H significantly impaired osteoblastic differentiation, as judged by significantly lower alkaline phosphatase mRNA levels in these cells compared with control cells transfected with empty vector (Fig. 8A). XL␣s-R543H expression also impaired osteoblastic differentiation of MC3T3-E1 cells, and in fact, almost no increase in alkaline phosphatase mRNA expression was detected in cells transfected with cDNA encoding XL␣s-R543H (Fig. 8A). Similarly, cells transiently expressing the GTPase-deficient XL␣s-G␣s chimera failed to show a significant increase in alkaline phosphatase mRNA levels under osteogenic conditions (Fig. 8A).  These findings regarding alkaline phosphatase mRNA levels were corroborated by experiments measuring alkaline phosphatase activity. At day 5, although all transfected cells displayed weaker staining than control cells transfected with the empty vector, the staining of cells transiently expressing XL␣s-R543H or the GTPase-deficient XL␣s-G␣s chimera appeared to be nearly absent (Fig. 8B). These findings are consistent with the above results showing that activated XL␣s can continue to signal, and thus, its cellular effects can be more prolonged than the cellular effects of activated G␣s.

DISCUSSION
XL␣s is a variant of G␣s that can mediate receptor-activated stimulation of adenylyl cyclases, but the study of mouse models in which XL␣s is ablated indicate that the cellular actions of this protein differ importantly from those of G␣s. In this study we revealed a marked difference between these proteins, which entailed their subcellular localization following activation. Our results showed that XL␣s and its N-terminally extended variant XXL␣s, unlike G␣s, are not subject to activation-induced subcellular internalization. In response to the activation of most GPCRs, G␣s traffics away from the plasma membrane via an endocytic pathway, and this mechanism limits continuous generation of cAMP at the plasma membrane (36 -38, 51, 52). Because XL␣s lacks activation-induced internalization, it induces cAMP generation in a sustained manner, and mutational inhibition of GTPase activity results in markedly stronger constitutive activity for XL␣s than for G␣s.
Recent studies have shown that isoproterenol-stimulated cAMP accumulation mediated by XL␣s is higher than that mediated by G␣s in the presence of phosphodiesterase inhibitors (47,53). These findings can now be explained, at least in part, by the sustained association of activated XL␣s with the plasma membrane. However, additional mechanisms may also be involved, because in our study the degree of basal cAMP accumulation mediated by the constitutively active XL␣s mutant was significantly higher than that mediated through the constitutively active XL␣s-G␣s chimera (see Fig. 7), which was indistinguishable from XL␣s regarding the degree of plasma membrane association. These additional mechanisms are likely to be at the level of adenylyl cyclase stimulation rather than involving nucleotide exchange or GTPase activity, given that the differences between G␣s and XL␣s regarding cAMP production were also observed in a GTPase-deficient state. Studies have shown that G␣s is associated with lipid rafts and that this localization hinders its ability to fully stimulate adenylyl cyclase (52,54). It is thus possible that XL␣s is excluded from lipid rafts, which would be consistent with the potent adenylyl cyclase stimulation by this protein. Localization of XL␣s within different membrane microdomains needs to be investigated in future studies.
Expression of XL␣s resulted in a prolonged cAMP response after PTHR but not the ␤ 2 AR stimulation (see Fig. 6). This finding could reflect the differences between the mechanisms underlying the desensitization of these two receptors. The inability of activated XL␣s to extend isoproterenol-induced cAMP response, despite remaining in the plasma membrane (see Fig. 1), could reflect receptor phosphorylation and arrestin binding, which are well characterized mechanisms desensitizing ␤ 2 AR (55-57). PTHR, however, appears to utilize distinct mechanisms of activation and desensitization. Unlike ␤ 2 AR, PTHR is able to generate prolonged cAMP signaling in response to certain agonists (38,41), and arrestins, contrary to their actions on ␤ 2 AR, enhance the cAMP signaling induced by PTH (58). These differences may thus explain why the strong plasma membrane localization of XL␣s yields a prolonged cAMP response to PTH but not to isoproterenol. Alternatively, PTHR and ␤ 2 AR might interact differently with XL␣s, such that the mechanisms desensitizing PTHR, regardless of being different from those desensitizing ␤ 2 AR, fail to overcome the effect of activated XL␣s localization in the plasma membrane. These possibilities, which may have significant implications in GPCR signaling and the relative roles of G␣s and XL␣s, remain to be addressed.
Previous studies had originally indicated that rat XL␣s is localized to the trans-Golgi network (18), and six cysteine residues within the XL domain have been previously identified as being critical for this subcellular localization at the basal state (59). More recent studies have established that both rat and human XL␣s are localized to the plasma membrane (25,33). Only two of the six cysteines shown to be important for the subcellular localization of rat XL␣s are conserved in human XL␣s, which is used in this study, and our findings show that at least one of those conserved cysteines is absolutely required for plasma membrane targeting. Cysteine residues are predicted to be substrates for palmitoylation. Indeed, rat XL␣s has been shown to be palmitoylated (33), and the cysteine residues within the cysteine-rich region appeared to be required for this modification (59). Our present results demonstrate that human XL␣s is also palmitoylated. Because this lipid modification is required for the insertion of G␣s to the plasma membrane (60), it is likely to be required for the plasma membrane targeting of XL␣s as well. Our results obtained with the Cys-to-Ser mutants and the palmitoylation inhibitor 2BP are consistent with this prediction.
According to our results obtained from truncation mutants, plasma membrane targeting of XL␣s requires not only one of the two conserved cysteines in the XL domain but also the region between PRR and the C-terminal end of HCD (see Fig.  4), which consists of a highly acidic N-terminal portion and a highly basic C-terminal portion. Polybasic amino acid regions are important for membrane targeting of other proteins, such as Ras and Rho family of small GTPases (61), and indeed, a recent study identified an N-terminal polybasic region within G␣s as another signal for plasma membrane targeting (62). Similarly, the corresponding region in XL␣s, i.e. the C-terminal end of the XL domain, includes multiple basic residues, and these residues, in addition to those within HCD, may contribute to the plasma membrane targeting of XL␣s at the steady state. The polybasic regions within HCD could also play a role in preventing the activationinduced subcellular redistribution of XL␣s, because replacing several N-terminal residues of G␣s with a polybasic region from G protein-coupled receptor kinase 5 has inhibited the cytosolic redistribution of G␣s upon activation (36).
It is clear that XL␣s can mimic G␣s regarding cAMP production both in transfected cells (23)(24)(25) and in transgenic mice (26). In fact, our results, together with recent data (47,53), show that it can even be more effective than G␣s. Then, why do data from G␣s knock-out mouse models (31,32) indicate that endogenous XL␣s is unable to compensate for G␣s ablation? There could be two possible reasons. First, because activated G␣s can also act on effector molecules that are localized intracellularly (63,64), and because it can apparently signal from internalized vesicles (38,39), it is conceivable that the defects caused by G␣s ablation result primarily from the loss of its actions that occur through intracellular effectors/mechanisms. This hypothesis is certainly consistent with the inability of XL␣s to substitute for G␣s despite being a strong stimulator of cAMP production at the plasma membrane. Second, it is possible that the internalization of G␣s upon activation is required for effective cAMP signaling. For example, the G␣s-specific regulator of G protein signaling RGS-PX1 is localized to endosomes (65), and it has been suggested that the termination of G␣s activation and thereby reassembly of the G protein heterotrimer is achieved at this subcellular site (36). Conversely, it is important to note that the presence of endogenous G␣s expression in XL␣s knock-out mice is also insufficient to prevent the phenotypes observed in this model (27,28). Therefore, it appears that XL␣s and G␣s have nonredundant contributions to cAMP signaling, perhaps by having spatially and temporally different expression profiles or by mediating cAMP signaling in response to different types of stimuli. In addition, it is likely that XL␣s interacts with unique effectors at the plasma membrane and thereby trigger signaling pathways that differ entirely from the cAMP signaling pathway.
GTPase inhibiting mutations that lead to constitutive XL␣s activity are found in patients with certain endocrine and nonendocrine tumors, fibrous dysplasia of bone, and McCune-Albright syndrome (47,66). Our findings predict that by enhancing the basal levels of cAMP and/or as-yet-undefined second messengers, constitutive XL␣s signaling at the plasma membrane is likely to contribute to the pathogenesis of these disorders. For example, XL␣s is normally expressed in undifferentiated skeletal progenitors (47,67), and GTPase-deficient XL␣s mutants may thus play a role in inhibiting the differentiation of these cells into normal bone.
Recent studies have identified the GNAS locus as one of few genes whose expression and/or copy numbers are increased in various cancers (14,15,68). Given that XL␣s activity is regulated less rigorously than G␣s activity, at least with respect to cAMP production, it is tempting to speculate that elevation of XL␣s levels plays an important role in the development of some of these cancers. Consistent with that hypothesis, XL␣s expression is normally limited to the paternal GNAS allele, and it is known that overexpression of paternally expressed gene products, such as insulinlike growth factor 2 (69), can lead to neoplasia.
In summary, our results show that XL␣s traffics differently from G␣s upon activation and thereby is able to extend cAMP signaling at the plasma membrane. The unique cellular actions of XL␣s and its variant XXL␣s remain unknown, and their strong association with the plasma membrane may form the basis for these unique actions.