Agonist Occupation of an a 2A -Adrenoreceptor-G i1 a Fusion Protein Results in Activation of Both Receptor-linked and Endogenous G i Proteins COMPARISONS OF THEIR CONTRIBUTIONS TO GTPase ACTIVITY AND SIGNAL TRANSDUCTION AND ANALYSIS OF RECEPTOR-G PROTEIN ACTIVATION STOICHIOMETRY*

A fusion protein between a pertussis toxin-resistant (C351G) mutant of the a subunit of the G protein G i1 and the porcine a 2A -adrenoreceptor was stably expressed in Rat 1 fibroblasts. Agonists caused stimulation of high affinity GTPase activity, which was partially prevented by pertussis toxin treatment, demonstrating that the toxin-resistant component of the GTPase activity was derived from the receptor-fused G protein and the remainder from endogenous G i a . Half-maximal stimula- tion of the GTPase activity of endogenous G i was achieved with lower concentrations of agonist. Although the K m for GTP of the fusion protein-linked G i was lower than for the endogenous G protein, V max measurements demonstrated that adrenaline activated some 5 mol of endogenous G i /mol of fusion protein- linked G i . The isolated a 2A -adrenoreceptor could acti- vate G s ; however, the fusion protein did not. Compared with adrenaline, the efficacy of a range of partial agonists to stimulate endogenous G i a was greater than for the fusion protein-constrained C351G G i1 a . a 2A -Adreno-receptor agonists could stimulate both p44 mitogen-activated protein kinase and p70 S6 kinase and inhibit forskolin-amplified

G protein-coupled receptors (GPCRs) 1 initiate vectorial sig-nal transduction cascades via activation of heterotrimeric G proteins and the subsequent regulation of effector enzymes (1,2). These function to amplify cellular response to the presence of low concentrations of extracellular mediators. As part of this process, agonist-occupied GPCRs have the capacity to catalytically activate G proteins. Some 20 distinct G protein ␣-subunits are known, and in many cases they are highly similar in sequence. Many cells co-express a considerable number of distinct but highly related G proteins. As a novel means to examine the functional interactions of a GPCR with a single G protein, we recently generated a fusion protein between the ␣ 2A -adrenoreceptor and the ␣-subunit of G i1 (3,4). Since G i1 is a member of the subfamily of G proteins that can be modified by ADP-ribosylation catalyzed by pertussis toxin (5) and a number of these G proteins are routinely co-expressed by all cells (5), we used a modified version of G i1 ␣ (C351G G i1 ␣) that is resistant to the action of pertussis toxin (6) to generate the fusion protein. Following transient expression of the fusion protein in COS-7 cells, we were able to treat the cells with pertussis toxin to modify the endogenous G i -like G proteins and thus eliminate any potential interactions between the GPCR of the fusion protein and the endogenously expressed forms of G i (4). Despite this precaution, there was no obvious indication that the GPCR constrained within the ␣ 2A -adrenoreceptor-C351G G i1 ␣ (␣ 2A R-C351G G i1 ␣) fusion protein had significant capacity to activate endogenous G proteins as well as the fusion partner G protein in an agonist-dependent manner (3,4). However, GPCRs that have had other proteins such as green fluorescent protein linked to their C terminus (7,8) still appear to be capable of interacting with and activating cellular G proteins (7,8). We now examine interactions of the ␣ 2A R-C351G G i1 ␣ fusion protein with endogenous G proteins following stable expression in Rat 1 fibroblasts. We demonstrate that in this setting agonist-occupation of ␣ 2A R-C351G G i1 ␣ results in stimulation of the high affinity GTPase activity of both endogenous G i ␣ and the fusion protein partner G protein and that the activation of endogenous G protein families is selective. Effective regulation of downstream signaling cascades occurred only via activation of endogenous G proteins rather than via activation of the GPCR-linked G protein. Despite this, the relative contribution of the GPCR-linked and endogenous G i proteins to agonist-stimulated GTPase activity measured at V max provided a highly novel means to demonstrate that the addition of adrenaline was able to activate some 6 mol of G protein/mol of receptor. Furthermore, studies that have assumed that agonistdependent prevention of proliferation of cancer cell lines following expression of a ␤ 2 -adrenoreceptor-G s ␣ fusion protein reflects activation of the receptor-attached G protein (9) may require careful reanalysis.

Materials
All materials for tissue culture were supplied by Life Technologies, Inc. (Paisley, Strathclyde, Scotland). [ 3 H]RS-79948 -197 (90 Ci/mmol) was purchased from Amersham Pharmacia Biotech. [␥-32 P]GTP (30 Ci/mmol) was obtained from NEN Life Science Products. Pertussis toxin (240 g/ml) was purchased from Speywood. All other chemicals were from Sigma or Fisons plc and were of the highest purity available. Oligonucleotides were synthesized on a Millipore Expedite Nucleic Acids Synthesis System.

Generation of the ␣ 2A -Adrenoreceptor-C351G G i1 ␣ Fusion Construct
A pertussis toxin-resistant C351G form of rat G i1 ␣ was generated (6) and linked to the porcine ␣ 2A -adrenoreceptor (10). Briefly, the ORF of the ␣ 2A -adrenoreceptor DNA was amplified by PCR using the following oligonucleotides: sense, 5Ј-TTGGTACCATGTATCCTTACGACGTTC-3Ј; antisense, 5Ј-AAGAATTCCATGGCGATCCGTTTCCTGTCCCCACG-GC-3Ј (restriction sites for KpnI, EcoRI, and NcoI are underlined). The PCR-amplified fragment was digested with KpnI and EcoRI and ligated to pBluescript (pBS) (Stratagene) through these restriction sites. Introduction of the NcoI site at the 3Ј-end of the ORF resulted in the C-terminal amino acid of the receptor being altered from Val to Ala and removal of the stop codon. The rat C351G G i1 ␣ cDNA contains two NcoI sites: one straddling the ATG start codon and the other 268 bp downstream from this. This 268-bp fragment was removed from C351G G i1 ␣ in pBS by digestion with NcoI, and the remaining C351G G i1 ␣ pBS cDNA was religated. The shortened cDNA was excised from pBS with EcoRI and cloned into the EcoRI site of the ␣ 2A -adrenoreceptor in pBS, adjacent to the 3Ј-end of the receptor ORF. The 268-bp fragment was then inserted between the NcoI sites at the 3Ј-end of the ␣ 2A -adrenoreceptor ORF and at the 5Ј-end of the C351G G i1 ␣ ORF. This resulted in production of an in-frame construct whereby the 3Ј-end of the ␣ 2Aadrenoreceptor ORF was exactly adjacent to the 5Ј-end of the C351G G i1 ␣ ORF. The full fusion construct was then excised from pBS with KpnI and EcoRI and ligated into the eukaryotic expression vector pcDNA3.

Cell Culture and Transfection
Generation of Stable Cell Lines Expressing the ␣ 2A -Adrenoreceptor-C351G G i1 ␣ Fusion Protein-Rat-1 fibroblasts were transfected with ␣ 2A R-C351G G i1 ␣ in pcDNA3 using DOTAP TM (Boehringer Mannheim), according to the protocol provided by the manufacturer, on cells plated in 10-cm 2 dishes at a confluency of 50 -60%. Cells were allowed to grow for 3 days before being split 1:10 in medium containing 0.7 mg/ml Geneticin sulfate (Calbiochem). Individual clones were isolated and expanded in Geneticin-containing medium.
Preparation of Membranes-Plasma membrane-containing P2 particulate fractions were prepared from cell pastes that had been stored at Ϫ80°C following harvest. Cell pellets were resuspended in 0.5 ml of 10 mM Tris-HCl, 0.1 mM EDTA, pH 7.5 (buffer A), and rupture of the cells was achieved with 50 strokes of a hand-held Teflon on-glass homogenizer followed by passage (10 times) through a 25-gauge needle. Cell lysates were centrifuged at 1000 ϫ g for 10 min in a Beckman TJ-6 centrifuge to pellet the nuclei and unbroken cells, and P2 particulate fractions were then recovered by centrifugation of the supernatant at 200,000 ϫ g for 30 min in a Beckman TL 100 bench-top ultracentrifuge using a Beckman TLA 100.2 rotor. P2 particulate fractions were resuspended in buffer A and stored at Ϫ80°C until required.
[ 3  Reactions were incubated at 30°C for 45 min, and bound ligand was separated from free by vacuum filtration through GF/C filters. The filters were washed with 3 ϫ 5 ml of assay buffer, and bound ligand was estimated by liquid scintillation spectrometry.
High Affinity GTPase Assays-These assays were performed essentially as described in Refs. 3 and 4 using [␥-32 P]GTP (0.5 M, 60,000 cpm) and varying concentrations of agonists (up to 10 M). Nonspecific GTPase was assessed by parallel assays containing 100 M GTP. In a range of experiments, GTPase activity was measured at V max as described previously (3).
Adenylyl Cyclase Activity-Intact cell adenylyl cyclase activity was measured as described by Wong (12) and Mercouris et al. (13). Briefly, cells in 12-well plates were loaded for 16 h. with 1 Ci/ml [ 3 H]adenine, in the presence or the absence of 25 ng/ml pertussis toxin. Cyclic AMP formation was stimulated by the addition of 50 M of forskolin, and regulation of forskolin-induced cAMP formation was assayed in the presence of adrenaline or UK14304 (100 M).
Reverse Transcriptase-PCR (RT-PCR)-RT-PCR was performed on RNA preparations from a range of clones expressing different levels of the ␣ 2A R-C351G G i1 ␣ fusion protein as determined from binding studies, as well as the parental Rat 1 fibroblast cell line.
The primers used were as follows: sense, CGA GAA GCG CTT CAC TTT CGTG, designed to anneal to the C-terminal region of the receptor cDNA sequence; antisense, CAC TGC TTT GTA CTG CTT ACA CTC, designed to anneal to the sequence of G i1 ␣. These should generate a 465-bp fragment from cDNA corresponding to the fusion protein. PCR conditions were as follows: one cycle of 5 min at 95°C, 1 min at 62°C, and 1 min at 72°C followed by 30 cycles of 1 min at 95°C, 1 min at 62°C, and 1 min at 72°C and one cycle of 1 min at 95°C, 1 min at 62°C, and 5 min at 72°C. The PCR was performed in the presence of 10% (v/v) dimethyl sulfoxide.
Regulation of p44 MAP Kinase Phosphorylation and Mobility-The phosphorylation of p44 MAP kinase was determined by an electrophoretic mobility shift assay (14). Cells were stimulated with the appropriate ligands for 5 min following maintenance in serum-free medium for 48 h and subsequently lysed at 4°C in a buffer containing 25 mM Tris-HCl, 40 mM p-nitrophenol, 25 mM NaCl, 10% (v/v) ethylene glycol, 10 M dithiothreitol, 0.2% (v/v) Nonidet P-40, 1 g/ml aprotinin, 1 mM sodium orthovanadate, 3.5 g/ml pepstatin A, and 200 M phenylmethylsulfonyl fluoride at pH 7.5. Following centrifugation of the lysed samples in a microcentrifuge (13,000 rpm, 5 min), SDS-polyacrylamide gel electrophoresis loading buffer was added to a sample of the supernatant and applied to SDS-polyacrylamide gel (10% (w/v) acrylamide, 0.063% (w/v) bisacrylamide, containing 6 M urea) following heating of the sample in a boiling water bath for 5 min. Samples were transferred from the gel to nitrocellulose and immunoblotted with an antipeptide antiserum raised against amino acids 325-345 of p44 MAP kinase (14). These gel conditions provide excellent "gel shift" of this polypeptide upon ligand regulation, a feature that is synonymous with its phosphorylation and activation (14).
p70 S6 Kinase Activity Measurements-Cells were grown to confluency in 100-mm cell culture dishes and serum-starved for approximately 16 h before treatment. After treatment with agonist for 10 min, the cells were washed twice with ice-cold phosphate-buffered saline before scraping into 1 ml of lysis buffer (55 mM Tris, pH 8.0, 4°C, 132 mM NaCl, 22 mM NaF, 1.1 mM EDTA, 5.5 mM EGTA, 11 mM sodium pyrophosphate, 33 mM 4-nitrophenyl phosphate, 1% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, 2 g/ml pepstatin A, 2 g/ml leupeptin). 20 min later, the lysates were centrifuged for 10 min at 14,000 ϫ g. The pellet was discarded, and an aliquot of supernatant was retained for Western blot analysis and protein determination. The remainder of the supernatant was precleared for 1 h with 20 l (packed volume) of protein A-agarose that had been equilibrated with lysis buffer.
Precleared lysates (250 g) were immunoprecipitated with 2.5 l of anti-p70 S6 kinase (15,16) for 2 h at 4°C before the addition to 20 l of protein A for a further 1 h. The immunoprecipitates were then washed twice with lysis buffer and twice with assay buffer (25 mM MOPS, pH 7.2, 30°C, 1 mM EDTA, 0.05% (v/v) Triton X-100, 1 mM dithiothreitol, 20 mM 4-nitrophenyl phosphate). 40 l of peptide substrate (32-mer of sequence KEAKEKRQEQIAKRRRLSSLRASTSKSGGSQK, final concentration 20 M) and 5 l of PKI were then added to the beads, and following a preincubation for 2 min at 30°C, the reaction was initiated by the addition of ATP (250 M ATP, 100 mM MgCl 2 containing 3 Ci of [␥ 32 P] ATP). The reaction was terminated by centrifuging for 10 s at 10,000 ϫ g, and 30 l of the supernatant was spotted onto P81 paper, which was dropped into 500 ml of phosphoric acid. Following five 5-min washes in phosphoric acid and a final wash in absolute alcohol, the papers were allowed to dry for approximately 30 min before transfer to scintillant and counting.
Data Analysis-Analysis of data curves was performed using the Kaleidagraph curve fitting package driven by an Apple Macintosh computer.

RESULTS
A fusion protein (␣ 2A R-C351G G i1 ␣) in which the N terminus of a pertussis toxin-resistant (C351G) mutant of the ␣-subunit of the G protein G i1 ␣ was fused to the C-terminal tail of the porcine ␣ 2A -adrenoreceptor was stably expressed in Rat 1 fibroblasts. 3  . RT-PCR of RNA derived from these cells using a primer pair in which the sense primer was designed to anneal with sequence contributed from the ␣ 2A -adrenoreceptor and the antisense primer from the sequence of G i1 ␣ allowed amplification of a 465-bp fragment (data not shown). A fragment of equivalent size was generated by PCR of the fusion protein cDNA using the same primer pair, but no equivalent product was obtained from RT-PCR of RNA from parental Rat 1 cells (data not shown).
The addition of the natural ligand adrenaline (100 M) or the synthetic agonist UK14304 (100 M) to membranes of clone RAGI 77 cells resulted in a substantial stimulation of high affinity GTPase activity (Table I). When equivalent experiments were performed on membranes derived from pertussis toxin-treated (25 ng/ml, 24 h) cells, agonist-stimulated GTP hydrolysis was still marked although reduced (Table I). This level of pertussis toxin treatment has been shown to fully attenuate functional interactions between the ␣ 2A -adrenoreceptor and endogenously expressed G i proteins (6,17) in Rat 1 cells. Membranes from a clonal cell line (TAG WT17) derived from Rat 1 cells that express some 4 pmol/mg protein of the isolated porcine ␣ 2A -adrenoreceptor produced similar levels of stimulation of high affinity GTPase upon the addition of adrenaline or UK14304 (data not shown). As such, these results demonstrated both that adrenaline and UK14304 were capable of causing guanine nucleotide exchange and subsequent hydrolysis of GTP by the GPCR-linked G protein and that the GPCR within the fusion protein also had the capacity to activate endogenous G i -like G proteins as well as the linked G protein. Trivial explanations for these data, related to potential reduction in cellular levels of the ␣ 2A R-C351G G i1 ␣ fusion protein during pertussis toxin treatment, were eliminated by performing [ 3 H]ligand binding studies (data not shown). Equivalent experiments performed on membranes of clone RAGI 17 also resulted in agonist stimulation of high affinity GTPase activity, which was partially pertussis toxin-sensitive (data not shown). However, since the level of stimulation was lower than in RAGI 77, clone 77 was used for all subsequent studies.
Time courses of agonist stimulation of high affinity GTPase activity at 37°C demonstrated that adrenaline allowed hydrolysis of GTP by the GPCR-associated C351G G i1 ␣ to proceed in an essentially linear manner for at least 60 min. By contrast, the stimulated GTPase activity derived from the endogenous G i ␣ population was noted to decay at time points beyond 30 min (Fig. 1), and no further stimulation was observed at times beyond 60 min. Using a time period at which the GTPase activity of both the GPCR-linked and endogenous G proteins was functioning at an essentially maximal rate in response to adrenaline, when using 0.5 M GTP as substrate, 31.7 Ϯ 1.3% (mean Ϯ S.E., n ϭ 11) of the total stimulated GTPase activity was derived from the fusion protein-linked G protein. Concentration-response curves demonstrated that half-maximal activation of the endogenous signal was achieved with 3.6-fold lower concentrations (EC 50 for adrenaline ϭ 6.4 Ϯ 0.6 ϫ 10 Ϫ7 M) of either adrenaline or UK14304 than required to cause half-maximal activation of the receptor-linked G protein (EC 50 for adrenaline was 2.3 Ϯ 0.4 ϫ 10 Ϫ6 M) ( Fig. 2A and data not shown). Such data may appear to indicate that the agonists were less potent in catalyzing GTPase activity of the fusion protein partner. However, following transient expression in COS-7 cells of either greater potency to stimulate the high affinity GTPase activity of the fusion protein containing the wild type G protein sequence (Fig. 2B).
By performing agonist-stimulated high affinity GTPase measurements on membranes of untreated and pertussis toxintreated clone RAGI 77 cells, the efficacy of a range of ligands could be assessed. Adrenaline, noradrenaline, and ␣-methylnoradrenaline all produced a similar and maximal stimulation of high affinity GTPase activity. The other agonists tested produced less stimulation at maximally effective concentrations and functioned as partial agonists. By comparing the capacity of each ligand to stimulate high affinity GTPase activity relative to the stimulation produced by adrenaline, measurements of agonist efficacy could be obtained. The same rank order of  clonidine Ն guanabenz Ͼ oxymetazoline, was obtained for both the receptor-associated G i1 ␣ and the endogenous G i ␣ population (Fig. 3). However, when compared with adrenaline the efficacy of all of the partial agonist ligands to stimulate the endogenous G i ␣ population was greater than for stimulation of the GPCR-associated G protein (Fig. 3). Indeed, in the cases of guanabenz and oxymetazoline, there was no significant level of stimulation of the GPCR-linked G protein. The efficacy assessments for agonist stimulation of the GPCR-associated C351G G i1 ␣ in membranes of clone RAGI 77 cells were, however, almost identical to those obtained following transient expression of ␣ 2A R-C351G G i1 ␣ in COS-7 cells (18).
Maximal GTPase activity of the GPCR-linked G i1 ␣ and the endogenous G i ␣ pool was assessed by measuring adrenalinestimulated GTPase activity in membranes of untreated and pertussis toxin-treated clone RAGI 77 cells at varying concentrations of GTP. The large increase in high affinity GTPase activity caused by adrenaline in membranes of untreated cells was confirmed to represent an increase in V max without an alteration in K m for GTP (Fig. 4A). Although pertussis toxin treatment reduced V max of the basal high affinity GTPase activity (Fig. 4B), this was also produced without a measurable alteration in the measured K m for GTP (Fig. 4B). However, resolution of agonist-induced stimulation of the GTPase activity of the fusion protein-coupled and endogenous G i populations demonstrated the K m for GTP of the fusion protein-coupled G i1 to be some 2-fold lower (260 Ϯ 7 nM) than for the endogenous G i (520 Ϯ 50 nM) (mean Ϯ S.E. n ϭ 3) (Fig. 4B). Since the levels of expression of the ␣ 2A R-C351G G i1 ␣ fusion protein could be measured directly from saturation [ 3 H]antagonist binding studies, this allowed calculation of the catalytic center activity of the adrenaline-stimulated fusion protein (3) as 4.2 Ϯ 0.5 min Ϫ1 (mean Ϯ S.E., n ϭ 3). Since amounts of endogenous G protein activated by the fusion protein cannot be assessed directly, it was not possible to measure catalytic center activity. However, if this is similar to that of the fusion protein-linked G protein, then it could be estimated from the V max calculations that the adrenaline-occupied ␣ 2A R-C351G G i1 ␣ fusion protein FIG. 2. The effect of the C351G mutation in G i1 ␣ on the potency of adrenaline to stimulate high affinity GTPase activity. A, varying concentrations of adrenaline were used to stimulate high affinity GTPase activity in membranes of untreated and pertussis toxintreated clone RAGI 77 cells. The effects of agonist on endogenous G proteins (filled symbols) and the ␣ 2A R-C351G G i1 ␣-associated G protein (open symbols) were resolved as in Fig. 1 were expressed transiently in COS-7 cells. Following the addition of varying concentrations of adrenaline to membranes prepared from these cells, high affinity GTPase activity was measured. Data are means Ϯ S.E. of triplicate observations from an experiment that was representative of three performed. was capable of causing activation of some 5 mol of endogenous G i ␣/mol of receptor-linked G protein at saturating concentrations of GTP (Fig. 4).
Following stable expression in Rat 1 fibroblasts, G i -linked receptors including the ␣ 2A -adrenoreceptor (19,20) and the ␦-opioid receptor (14) have been shown to be able to mediate the phosphorylation and activation of the extracellularly regulated kinase (ERK) MAP kinases in response to agonist ligands. The addition of UK14304 (100 M, 7. 5 min) to serum-deprived clone RAGI 77 cells resulted in the phosphorylation of virtually the entire cellular complement of p44 MAP kinase as measured by a reduced mobility of the protein through SDS-polyacrylamide gel electrophoresis (Fig. 5). By contrast, UK14304 was unable to cause phosphorylation of p44 MAP kinase in pertussis toxintreated clone RAGI 77 cells (Fig. 5). This was not due to an alteration in the time frame of p44 MAP kinase phosphorylation and dephosphorylation (data not shown). As anticipated from previous studies (14), both epidermal growth factor (10 nM, 7.5 min) (Fig. 5) and platelet-derived growth factor (10 nM, 7.5 min) (data not shown) were able to cause effective phosphorylation of p44 MAP kinase in both untreated and pertussis toxin-treated cells. The ␣ 2A -adrenoreceptor is also able to stimulate the activity of p70 S6 kinase in Rat 1 cells by a pathway that is independent of activation of the ERK MAP kinases (15,21). UK14304 (100 M, 10 min) produced robust activation of p70 S6 kinase in serum-deprived RAGI 77 cells that could subsequently phosphorylate a peptide substrate (Fig. 6). The same was true of PDGF (10 nM, 10 min). As with p44 MAP kinase phosphorylation, this capacity of UK14304, but not that of PDGF, was completely attenuated following pertussis toxin treatment (Fig. 6).
Expression of the ␣ 2A -adrenoreceptor in Rat 1 cells also allows agonist-mediated inhibition of forskolin-amplified adenylyl cyclase activity in membrane preparations (17). Following stable expression of ␣ 2A R-C351G G i1 ␣ in these cells, UK14304 and adrenaline were able to mediate inhibition of forskolin-amplified adenylyl cyclase (Fig. 7). However, following pertussis toxin treatment, this inhibition was no longer observed (Fig. 7). The ␣ 2A -adrenoreceptor has been demonstrated to have the capacity to interact with the stimulatory G protein G s as well as the G i -like G proteins to regulate adenylyl cyclase activity (22)(23)(24). In most circumstances, the inhibitory effect is predominant, but following pertussis toxin treatment of cells the activation of G s ␣ and thence adenylyl cyclase can often be unmasked (22)(23)(24). In Rat 1 cells stably expressing the isolated ␣ 2A -adrenoreceptor, neither adrenaline nor UK14304 produced marked effects on forskolin-amplified adenylyl cyclase activity in intact cells (Fig. 8). This may reflect the combination of capacity of the receptor to activate both G i and G s . In support of this concept, following pertussis toxin treatment to prevent interaction of the receptor with G i , both agonists produced a strong activation of forskolin (50 M)-amplified adenylyl cyclase activity (Fig. 8). By contrast, the agonistmediated inhibition of forskolin-amplified adenylyl cyclase activity in intact, untreated clone RAGI 77 cells (Figs. 7 and 8) was not converted into a stimulation following pertussis toxin treatment (Fig. 8); indeed, under these conditions no effect of agonist was observed. As anticipated, forskolin-amplified adenylyl cyclase activity was not modified by adrenaline or UK14304 in either untreated or pertussis toxin-treated parental Rat 1 fibroblasts (Fig. 8). Such data are compatible with an inability of the agonist-occupied ␣ 2A R-C351G G i1 ␣ fusion protein to interact effectively with endogenous G s despite its clear activation of endogenous G i . DISCUSSION Herein, we demonstrate for the first time the capacity of a GPCR constrained within a GPCR-G protein ␣-subunit fusion protein to interact with and activate G proteins other than that physically linked to it. Furthermore, this interaction was selective between G protein families, indeed displaying greater selectivity than the isolated receptor. Using this system, the ability of the natural agonist adrenaline to stimulate high affinity GTPase activity, and the capacity of the GPCR-G protein fusion protein to be considered as an agonist-activated enzyme that contains and retains both GPCR and G protein function, we show that the porcine ␣ 2A -adrenoreceptor has the capacity to activate some 6 mol of G i protein/mol at saturating concentrations of GTP. Such measurement could not have been obtained in native cell membranes without the use of the GPCR-G protein fusion protein. Given that the measured K m for GTP was some 0.4 M, with intracellular concentrations of GTP in the region of 10 M and the measured stoichiometry of G protein activation being assessed at GTPase V max , these estimates are likely to closely reflect the in vivo situation. We have also examined the capacity of the GPCR to activate downstream signal transduction via both the activated fusion protein-linked and endogenous G proteins.
We produced an ␣ 2A -adrenoreceptor-G i1 ␣ fusion protein by simply ligating the 5Ј-end of the G protein cDNA to the 3Ј-end of the receptor cDNA. The open reading frame so produced is expressed as a single, intact polypeptide (3, 4). Following tran-FIG. 3. Partial agonists display higher efficacy to activate endogenous G i proteins than ␣ 2A R-C351G G i1 ␣. The capacity of a range of agonists at the ␣ 2A -adrenoreceptor to stimulate the high affinity GTPase activity of endogenous G i -like proteins (filled bars) and ␣ 2A R-C351G G i1 ␣ (hatched bars) in membranes of clone RAGI 77 cells was assessed as in Fig. 1. All results were then compared relative to natural ligand adrenaline (1). The ligands were ␣-methylnoradrenaline (2), noradrenaline (3), UK14304 (4), BHT933 (5), dexmeditomidine (6), xylazine (7), clonidine (8), guanabenz (9), and oxymetazoline (10). Results represent mean Ϯ S.E. from three independent experiments. sient expression of this construct in COS-7 cells, we have demonstrated that the addition of an ␣ 2 -adrenoreceptor agonist to membranes of these cells results in guanine nucleotide exchange and its subsequent hydrolysis by the receptor-linked G protein (3,4). To prove this latter point conclusively, we generated the fusion protein using a pertussis toxin-resistant form of G i1 ␣ in which cysteine 351, which is the normal acceptor site for pertussis toxin-catalyzed ADP-ribosylation, was converted to glycine. Prior pertussis toxin treatment of the COS-7 cells resulted in modification of all of the endogenous G i -like G proteins (6). Since this modification prevents functional interactions between GPCRs and G i family G proteins (5), such treatment determined that the observed agonist regulation of high affinity GTPase activity could not reflect activation of the endogenous G i proteins. Moreover, since agonist stimulation of high affinity GTPase activity in COS-7 cell membranes following expression of the isolated ␣ 2A -adrenoreceptor was completely blocked by pertussis toxin treatment (6), any measured activity could not reflect interactions of the receptor with other endogenously expressed G proteins, such as G s ␣, with which the ␣ 2A -adrenoreceptor has been reported to interact (22)(23)(24) (also see below). Pertussis toxin treatment of COS-7 cells expressing ␣ 2A R-C351G G i1 ␣ resulted in little diminution of agonist signal (3). These data were consistent with interpretations centered on ideas that either the GPCR in the fusion protein could only contact its partner G protein or that the cellular distribution of the ␣ 2A R-C351G G i1 ␣ was completely distinct from the bulk of the endogenously expressed G protein.
It is certainly worth noting in this regard that a number of studies on GPCR expression on COS cells have indicated that much of the expressed protein is not targeted appropriately to the plasma membrane (e.g. see Ref. 25).
Simple attachment of another protein to the C-terminal tail of a GPCR is not inherently sufficient to prevent interaction of the GPCR with endogenous G proteins, as noted in studies in which GPCR-green fluorescent protein fusion proteins have the capacity to regulate effector function in an agonist-dependent manner (7,8). Therefore, to directly assess possible interactions of the agonist-occupied GPCR within the ␣ 2A R-C351G G i1 ␣ fusion protein with endogenously expressed G proteins, as FIG. 4. V max for stimulation of endogenous G i -like G proteins by adrenaline is higher than for stimulation of ␣ 2A R-C351G G i1 ␣, but the K m for GTP is lower. Basal high affinity GTPase activity (open symbols) and its regulation by adrenaline (filled symbols) was measured at varying concentrations of GTP in membranes of untreated clone RAGI 77 cells. The data from a representative experiment of three performed are presented as an Eadie-Hofstee transformation (A). Equivalent experiments were performed on membranes from pertussis toxin-treated cells. Using the strategy of Fig. 1, basal high affinity GTPase and its stimulation due to activation of endogenous G i -like proteins (triangles) and well as the GPCR-associated G protein, it was necessary to express the fusion construct stably. The cell line selected for these studies was Rat 1 fibroblasts because we have previously stably expressed the porcine and human ␣ 2A -adrenoreceptors in these cells and shown them to be capable of allowing agonistdependent stimulation of high affinity GTPase activity (17), inhibition of membrane-delineated forskolin-amplified adenylyl cyclase activity (17), and phosphorylation and activation of both the ERK MAP kinases (18 -20) and p70 S6 kinase (15,21).
Following stable transfection of ␣ 2A R-C351G G i1 ␣, two clones were examined by RT-PCR to confirm expression. Using a primer pair selected to amplify sequence spanning the site of GPCR and G protein fusion we demonstrated expression of mRNA encoding the fusion protein in these clones but not in parental cells. Furthermore, 3 H-antagonist binding studies demonstrated the fusion protein to be expressed at high levels. The addition of adrenaline to membranes of cells stably expressing ␣ 2A R-C351G G i1 ␣ resulted in a robust stimulation of high affinity GTPase activity. This level of activity was greater per copy of the fusion protein than we had observed in transient transfection studies (3) (Fig. 4A), suggesting either that stable expression improved cellular targeting and function of the fusion protein or that G proteins other than the fusion partner were becoming activated. Following pertussis toxin treatment of the cells, the capacity of adrenaline to stimulate high affinity GTPase activity was reduced significantly. Since pertussis toxin treatment did not reduce levels of expression of ␣ 2A R-C351G G i1 ␣, this confirmed that the GPCR within the fusion protein was capable of interacting with and activating endogenous G i -like G proteins as well as the C351G G i1 ␣ linked to the receptor. This is the first demonstration of such a capacity; thus, to explore it in detail, we devised a series of approaches to examine similarities and differences in the regulation of these interactions. In the only other reported construct of a GPCR-G protein fusion protein, between the ␤ 2 -adrenoreceptor and G s ␣, its functionality was demonstrated by examining the ability of agonist to stimulate adenylyl cyclase activity following expression of the fusion protein in S49 cyc Ϫ cells (26). Since S49 cyc Ϫ cells lack endogenous G s ␣, this system was clearly not appropriate to address the types of questions posed herein.
To examine the relative contribution to the total agoniststimulated GTPase activity of endogenous G i compared with C351G G i1 ␣ of the fusion protein, such assays were performed for 30 min, a period over which both enzymatic activities were proceeding in a maximal and linear fashion, in membranes of untreated and pertussis toxin-treated cells. 68.3 Ϯ 2.6% of the total signal in response to adrenaline was produced via endogenous G i proteins and the remainder from the fusion protein. These experiments were performed at a single (0.5 M) subsaturating concentration of GTP, as is the norm for studies on agonist regulation of GTPase activity. Since the ␣ 2A R-C351G G i1 ␣ fusion protein can be viewed as an agonist-regulated enzyme, the examination of GTPase activity with variation in GTP concentration allowed measurements of V max for the GPCR-associated and endogenous G i protein and direct assess-  1 and 4), UK14304 (100 M) (lanes 2 and 5), or epidermal growth factor (10 nM) (lanes 3 and 6) was added for 7.5 min. The cells were also treated with either vehicle (lanes 1-3) or pertussis toxin (25 ng/ml) (lanes 4 -6) for 24 h before the addition of receptor agonist. Cell lysates were resolved by SDS-polyacrylamide gel electrophoresis containing 6 M urea, and the presence of p44 MAP kinase was detected immunologically (13). Phosphorylation and activation of this polypeptide is correlated with its reduced mobility through this gel system.  1 and 4), PDGF (10 nM) (bars 2 and 5), or UK14304 (100 M) (bars 3 and 6) was added for 10 min. The cells were also treated with either vehicle (bars 1-3) or pertussis toxin (25 ng/ml) (bars 4 -6) for 24 h before the addition of receptor agonist. p70 S6 kinase was immunoprecipitated as described (14,15), and activity was measured by the capacity of the immunoprecipitated samples to cause phosphorylation of a substrate peptide. Three independent experiments produced similar data. ment for the K m for GTP as substrate. Interestingly, by resolving the contributions of the fusion protein-associated and endogenous G proteins to total agonist-induced signal by analyzing results in membranes of untreated and pertussis toxin-treated cells, we noted that the fusion protein partner G protein displayed higher affinity for GTP than the endogenous G proteins. Again, this is a unique set of observations. Since the levels of expression of the GPCR-linked C351G G i1 ␣ are known from the 3 H-ligand binding studies this allowed estimates of catalytic center activity (turnover number) of this protein. The measured V max of the endogenous G protein population cannot be determined in this way, but assuming a similar catalytic center activity this would predict that the adrenaline-occupied ␣ 2A R-C351G G i1 ␣ fusion protein can activate some 5 mol of endogenous G i ␣ protein/mol. As noted above, this measurement of G protein activation stoichiometry could not have been obtained in native membranes without use of the GPCR-G protein fusion strategy.
A range of ligands with agonist activity at the ␣ 2A -adrenoreceptor are routinely considered as partial agonists compared with adrenaline. By examining the rate of GTPase activity produced by these ligands in membranes from untreated and pertussis toxin-treated clone RAGI 77 cells, this was observed to be true for activation of both the endogenous and GPCRlinked G proteins. The rank order of efficacy was the same for the various compounds at the two G protein pools, but, interestingly, in comparison with the effects of adrenaline and the other ligands that acted as full agonists, the partial agonists all displayed greater efficacy to activate the endogenous G i -like G proteins. The basis for this difference is unclear and may relate to constraints imposed by the fusion or the C351G mutation that was necessary to distinguish agonist activation of these two G protein pools. Concentration-response curves for stimulation of high affinity GTPase of the endogenous and receptorlinked G i ␣ demonstrated a 3-fold higher affinity for activation of the endogenous G protein pool. This difference, however, is likely to reflect differences in affinity of the agonist-occupied receptor to activate wild type and the C351G G protein. We have demonstrated previously that following independent coexpression of the ␣ 2A -adrenoreceptor with either wild type G i1 ␣ or C351G G i1 ␣, some 10-fold higher levels of the agonist UK14304 were required to stimulate the mutant G protein (6). Furthermore, in the present studies we have compared the capacity of agonist to stimulate the high affinity GTPase of ␣ 2A R-C351G G i1 ␣ and a fusion protein in which the receptor is linked to the wild type G i1 ␣ sequence (␣ 2A R-C351 G i1 ␣) following their transient expression in COS-7 cells. Some 5-fold higher levels of agonist were required to produce the same stimulation of the fusion protein containing the C351G mutation. This should not be considered inherently surprising, since the C-terminal tail of the G protein comprises a key contact site for a receptor, and the addition of ADP-ribose to the cysteine residue by pertussis toxin results in attenuation of receptor-G protein interactions (5).
␣ 2 -Adrenoreceptors are classical examples of GPCRs that mediate inhibition of adenylyl cyclase. Like many other G ilinked GPCRs, when expressed in appropriate locations, they are also able to initiate cascades that result in the phosphorylation and activation of the ERK family of MAP kinases and of p70 S6 kinase. We have previously demonstrated that following expression in Rat 1 fibroblasts the ␣ 2A -adrenoreceptor is able to produce each of these effects in a pertussis toxin-sensitive manner (21). Following expression of the ␣ 2A R-C351G G i1 ␣ fusion protein in these cells forskolin-amplified intact cell adenylyl cyclase activity was reduced by ␣ 2 -adrenoreceptor agonists. However, this was virtually abolished following treatment with pertussis toxin, demonstrating the effect to be due to activation of the endogenous G i population (Fig. 7). As anticipated from studies with expressed isolated ␣ 2A -adrenoreceptors, in clone RAGI 77, agonists were able to cause the phosphorylation and activation of both the p44 MAP kinase and p70 S6 kinase. However, following pertussis toxin treatment, there was no ability of ␣ 2 -adrenoreceptor agonists to activate either of these kinase cascades, although, as anticipated, epidermal growth factor and PDGF acted as strong activators of both kinases in both untreated and pertussis toxin-treated cells (Figs. 5 and 6). It was further noteworthy that the stably expressed ␣ 2A R-C351G G i1 ␣ fusion protein did not have the capacity to produce an enhancement of forskolin-stimulated adenylyl cyclase activity in pertussis toxin-treated intact clone RAGI 77 cells. A number of studies have noted the capacity of agonists at ␣ 2 -adrenoreceptors to stimulate adenylyl cyclase activity in defined circumstances (22)(23)(24). This is often effectively observed following pertussis toxin treatment of cells to eliminate interactions of the receptor with G i -like G proteins (23,24). Comparisons of the capacity in this regard of the isolated porcine ␣ 2A -adrenoreceptor and the ␣ 2A R-C351G G i1 ␣ fusion protein were also illuminating (Fig. 8). Although ␣ 2A R-C351G G i1 ␣ produced strong agonist-mediated inhibition of adenylyl cyclase in intact untreated clone RAGI 77 cells, the isolated receptor had little effect in these conditions. By contrast, following pertussis toxin treatment of the cells, although ␣ 2A R-C351G G i1 ␣ now failed to modulate forskolin-amplified adenylyl cyclase activity, the addition of agonist to cells expressing the isolated receptor resulted in a substantial increase in activity. The most obvious interpretation of these results is that although ␣ 2A R-C351G G i1 ␣ has the capacity to activate endogenous G i in response to agonist, it does not interact effectively with endogenous G s . This is in contrast to the isolated receptor, which in untreated cells presumably activates a combination of G i and G s to produce little net effect on forskolinstimulated adenylyl cyclase and a strong stimulation in pertussis toxin-treated cells. These are again unique observations and provide rather unexpected differences between the isolated receptor and the fusion protein.
Previous studies with a ␤ 2 -adrenoreceptor-G s ␣ fusion protein (26) demonstrated agonist-mediated stimulation of adenylyl cyclase following stable expression of the fusion protein in S49 cyc Ϫ cells, which lack expression of endogenous G s ␣. It might be argued that the observed effect was due to activation of the endogenous population of ␤ 2 -adrenoreceptors present in S49 lymphoma cells now being able to utilize the provided fusion protein-attached G protein, as it is able to do when provided with exogenous G s ␣ in cyc Ϫ reconstitution assays. This appears unlikely, however, due to the high potency of agonist following expression of the fusion protein. Furthermore, even if it were an effect of the endogenous receptor, the stimulation of adenylyl cyclase would still reflect the capacity of the tethered G protein to access and activate adenylyl cyclase. The minimal capacity of the ␣ 2A R-C351G G i1 ␣ fusion protein to mediate the inhibition of forskolin-amplified adenylyl cyclase via the physically linked G i when the fusion proteinactivated endogenous G i can do so may then reflect limited spatial opportunities of the fusion protein-linked G protein or the relative G protein activation ratios that in the fusion protein must, by definition, be limited to mol/mol of agonist-occupied receptor. Certainly the ␣ 2A -adrenoreceptor has a short C-terminal tail compared with the ␤ 2 -adrenoreceptor, and little is known about the cellular distribution of the polypeptides of G protein-linked signaling cascades relative to one another (27,28). It was perhaps even more surprising that the stably expressed ␣ 2A R-C351G G i1 ␣ fusion protein showed no capacity to mediate activation of the signaling cascades leading to activation of the ERK MAP kinases and p70 S6 kinase, although activation of endogenous G proteins by the fusion protein did result in kinase stimulation. Although the early steps of GPCRmediated activation of p70 S6 kinase are still unclear, a vast amount of work has examined the requirements for GPCRmediated activation of the ERK MAP kinases. The current consensus is that a key early component is the ␤␥ complex released by activation of G i -like G proteins (29 -31), although the relative importance of other proteins including the isoforms of Shc (32) and a recently identified 100-kDa protein (33) is still a matter of debate (34,35). The ␣ 2A R-C351G G i1 ␣ fusion protein interacts with ␤␥ complex provided as a combination of ␤ 1 and ␥ 2 subunits as measured by an enhancement of agoniststimulated GTPase activity following their co-expression (3,4). Furthermore, interaction of the ␤ 2 -adrenoreceptor-G s ␣ fusion protein with ␤␥ has also been postulated based on the capacity of that fusion protein to act as a substrate for cholera toxincatalyzed ADP-ribosylation only when provided with ␤␥ complex (26). In both cases, this is despite the known role of the tethered G protein N terminus in ␤␥ interactions (36,37). Furthermore, we have demonstrated that relatively low levels of receptor occupancy, and thus low levels of G protein activation compared with the maximal levels that can be obtained, are sufficient to fully activate both kinases in Rat 1 fibroblasts (14 -16). If these kinase cascades simply required release of ␤␥ complex for their initiation, they would have been expected to function in both untreated and pertussis toxin-treated cells, even if the concentration-response curve for the fusion protein partner G protein might have been anticipated to lie to the right of that for response via endogenous G i ␣. Since they do not, the reasons for this will require further examination and may provide new insights into the early steps in these pathways.
The generation of fusion proteins between GPCRs and G protein ␣-subunits is beginning to provide new avenues to examine specificity in both the biochemistry and pharmacology of cell signaling systems and should provide approaches to study the catalytic activation of G proteins by a GPCR. As such, we predict that fusions of this form will become widely used. However, the current studies demonstrate that due attention must be given to whether an observed activity results from the activation of the G protein partner of the fusion protein or from activation of endogenous G proteins.