Chimeric Mutagenesis of Putative G-protein Coupling Domains of the α2A-Adrenergic Receptor LOCALIZATION OF TWO REDUNDANT AND FULLY COMPETENT G

Abstract We have investigated potential Gi and Gs coupling domains within the intracellular regions of the α2AAR subtype using a series of nine chimeric mutations. The second intracellular loop (ICL2, amino acids 133-149) and the amino- and carboxyl-terminal regions of the third intracellular loop (ICL3, amino acids 218-235 and 355-371, respectively) of the cloned human α2AAR were substituted with the analogous sequence from either the Gs-coupled β2AR or the Gi-coupled serotonin type 1A receptor (5-HT1AR). Mutant and wild type α2AAR were stably expressed in Chinese hamster ovary cells and functional coupling of each receptor to Gi and Gs was assessed in membrane adenylyl cyclase assays. Substitution of 5-HT1AR sequence into ICL2 ablated coupling to Gs but not to Gi, whereas substitution of β2AR sequence significantly depressed coupling to Gi but not to Gs. Thus, the ICL2 of the α2AAR contains elements essential for both signaling pathways. Substitution of either the amino- or carboxyl-terminal segments of ICL3 with 5-HT1AR sequence ablated agonist stimulation of adenylyl cyclase activity (without affecting inhibition), suggesting that both domains are necessary for α2AAR coupling to Gs. In contrast, individual substitution of β2AR sequence into ICL3 amino or carboxyl termini had no appreciable effect on Gi coupling. Concomitant substitution of β2AR sequence into both regions substantially impaired Gi coupling, implying that each is capable of independently supporting functional coupling. Substitution of 5-HT1AR at either locus had no effect on Gi coupling. Thus, for Gs coupling, these two domains within ICL3 are both required for functional coupling. However, for Gi coupling, the α2AAR appears to have two distinct regions within ICL3 that are capable of supporting Gi coupling independently. There has been no previous elucidation of a receptor having redundant, fully competent domains for coupling to a single class of G-protein. Such duplicity of functional domains within α2AR may suggest strong evolutionary pressure to maintain Gi coupling.

Adrenergic receptors (AR) 1 are members of a superfamily of integral membrane proteins that signal to the interior of the cell through heterotrimeric guanine nucleotide binding proteins or G-proteins. The AR are divided into three classes, ␣ 1 , ␣ 2 , and ␤, which are differentiated by their relative selectivity for certain ligands, signal transduction pathways, and molecular structure. Each class of AR has been further divided into several pharmacological subtypes, each of which represents distinct receptors, as demonstrated by the cloning and recombinant expression of their respective genes from a variety of species, including human. There are three cloned human ␣ 2 AR, ␣ 2 C10, ␣ 2 C4, and ␣ 2 C2, that correspond to the pharmacologically defined subtypes ␣ 2A , ␣ 2C , and ␣ 2B , respectively (1)(2)(3). The ␣ 2 AR display a wide distribution in both peripheral (4,5) and central (5)(6)(7) tissues and mediate a variety of physiological responses (8).
The ␣ 2 AR primarily couple to the inhibitory G-protein G i , which negatively modulates the activity of the enzyme adenylyl cyclase in the production of the second messenger cAMP. In addition to the inhibition of adenylyl cyclase activity, ␣ 2 AR also have been linked to a number of other cellular signaling pathways via coupling to G i or a G i /G o class G-protein (9), activation of phospholipase C activity (10) (potentially via coupling to a G q /G ␣11 class protein (11)), and even stimulation of adenylyl cyclase activity via coupling to G s (12,13). With respect to the latter signaling pathway, increases in cAMP in response to ␣ 2 AR agonists have been observed in cerebral cortical brain slices (14) and pancreatic islet cells (15), as well as a host of transfected cell lines including CHO cells (12, 16 -20), COS-7 cells (13,18,20), HEK-293 cells (13), PC-12 cells (21), JEG-3 cells (22), and the S115 mouse mammary tumor cell line (23). In many of these cell lines, at least for the ␣ 2A AR, both receptor-mediated inhibition and stimulation of cAMP production can be readily observed and occur as a complex, biphasic response. In this regard, we have utilized CHO cells as a model system to evaluate and compare ␣ 2 AR-mediated inhibition and stimulation within a single cell type. This has proved particularly useful in some of our most recent studies in which the ability of mutated ␣ 2 AR to couple to G i and G s was examined (20).
Site-directed mutagenesis, chimeric receptor, and in vitro peptide studies have indicated that the intracellular portions, particularly the second and third intracellular loops, of Gprotein-coupled receptors are the regions that physically interact with G-proteins. For the ␣ 2 AR, very little is known concerning the regions within the intracellular domains of these receptors that functionally interact with G-proteins. In some of the earliest mutagenesis studies, substitution of ␣ 2 AR sequence into the ␤ 2 AR was utilized to elucidate regions within the ␤ 2 AR required for G s coupling (24 -26); however, no information concerning the specific regions involved in ␣ 2 AR-Gprotein coupling was obtained. Furthermore, subsequent to the above studies, an ␣ 2 AR-G s coupling pathway was delineated (12). In other studies, it has been shown that synthetic peptides based on relatively small portions of the intracellular regions of the human ␣ 2A AR activate purified G-proteins in vitro (27,28). And, in a competition peptide study, incubation of human platelet membranes with synthetic peptides based on the second intracellular loop and the carboxyl-terminal region of the third intracellular loop of the human ␣ 2A AR reduced high affinity agonist binding (29). None of these studies, however, directly identified a region or regions within the intact ␣ 2 AR that are required for functional G-protein coupling.
Recently, using deletion and chimeric mutagenesis of the cloned human ␣ 2A AR, we identified a discrete stretch of 11 amino acids (218 -228) in the amino-terminal region of the third intracellular loop that is required for functional G s coupling but not for G i coupling (20). This provided evidence that there are specific structural domains within ␣ 2 AR that enable the receptor to couple to G s , which are distinct and separable from the structural requirements for ␣ 2 AR-G i coupling. In the present study we have explored the second and third intracellular loop regions of the ␣ 2A AR to delineate domains responsible for functional G i coupling. We have also further assessed potential requirements for G s coupling using an expanded array of chimeric receptors. And finally, when several critical sequences were found, we considered whether they represent redundant, fully competent domains within the ␣ 2A AR for coupling to a single G-protein or whether multiple domains act together to accomplish functional coupling. Studies were undertaken using a series of nine chimeric receptors in which portions of the ␣ 2A AR were substituted with the analogous sequence from either the G s -coupled ␤ 2 AR or the G i -coupled serotonin type 1A receptor (5-HT 1A R). Mutant and wild type ␣ 2 AR were stably expressed in CHO cells, and agonist-mediated inhibition and stimulation of adenylyl cyclase activity was assessed in membrane assays.

EXPERIMENTAL PROCEDURES
Mutagenesis-The cloned human ␣ 2A AR (␣ 2 C10) in the mammalian expression vector pBC12BI was used as a template for mutagenesis (1). For mutagenesis, several regions of the ␣ 2A AR were substituted with analogous sequence from either the cloned human 5-HT 1A R or ␤ 2 AR. These regions included the amino-and carboxyl-terminal portions of the third intracellular loop and the second intracellular loop (see Fig. 1). Substitution of amino acids 218 -235 with analogous ␤ 2 AR and 5-HT 1A R sequence in the amino-terminal region of the third intracellular loop has been previously described (20). For substitutions in the carboxyl-terminal region of the third intracellular loop, a BglII/KpnI fragment of ␣ 2 C10-pBC12BI was subcloned into M13mp18, and then NheI and Mlu I restriction sites were created at positions corresponding to amino acids 352-353 and 375-376, respectively, using oligonucleotide-directed mutagenesis (30). The fragment containing the new restriction sites was ligated back into digested ␣ 2 C10-pBC12BI to create a cassette for substitution. Then ␣ 2 C10-pBC12BI containing sites for NheI and Mlu I was digested with these enzymes and then ligated with annealed, synthetic oligonucleotides that encoded either ␤ 2 AR or 5-HT 1A R sequence analogous to amino acids 355-371 of ␣ 2 C10. For second intracellular loop substitutions, a SacI fragment of ␣ 2 C10-pBC12BI was subcloned into M13mp19. Oligonucleotides encoding sequence for the analogous ␤ 2 AR or 5-HT 1A R second loop sequence with 5Ј and 3Ј tags containing complementary ␣ 2 C10 sequence directly carboxyl-and amino-terminal to the substitution were utilized to simultaneously remove and replace amino acids 133-149 of ␣ 2 C10. Mutated fragments were then ligated back into digested ␣ 2 C10-pBC12BI. Combinations of the above substitutions were constructed by splicing together restriction digest fragments containing the desired mutations. All mutations were verified by dideoxy sequencing.
Cell Culture and Transfection-CHO cells were grown in monolayers in Ham's F-12 medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin at 37°C in a 5% CO 2 atmosphere. For stable expression of mutant and wild type ␣ 2 AR, CHO cells at ϳ30% confluence were cotransfected using calcium phosphate precipitation (31) with 30 g of wild type or mutant cDNA in pBC12BI and 3 g of pSV 2 neo, which provides for G418 resistance. Transfected cells were isolated in medium containing 1 mg/ml G418. G418-resistant cells were expanded, and confluent monolayers were screened for expression of ␣ 2 AR using a [ 3 H]yohimbine binding assay as described below. Selection pressure for clonal cell lines was maintained by the addition of 80 g/ml of G418 to the above medium. For all studies, transfected cells in monolayers at ϳ90% confluence were utilized, and multiple clones expressing each receptor were studied.
Membrane Preparation-For membrane studies, cells in monolayers were rinsed three times with phosphate-buffered saline, and membranes were prepared by scraping with a rubber policeman in ice-cold hypotonic lysis buffer (5 mM Tris-HCl, 2 mM EDTA, pH 7.4) followed by centrifugation at 42,000 ϫ g for 10 min at 4°C. Crude membrane pellets were then resuspended in the appropriate buffer for use in the assays. Protein concentration in prepared membranes was measured using the copper-bicinchoninic acid method (32) with bovine serum albumin as standard.
Radioligand Binding-Expression of mutant and wild type ␣ 2 AR was determined using a [ 3 H]yohimbine binding assays as described previously (12). Briefly, membranes from stably expressing CHO cells prepared as described above were incubated with 25 nM Adenylyl Cyclase Assays-␣ 2 AR-mediated inhibition or stimulation was determined in membranes prepared from CHO cells stably expressing mutant and wild type ␣ 2 AR using the method of Salomon et al. (33) as modified (12). Previously, we have shown that ␣ 2 AR-mediated responses in CHO cells are biphasic, consisting of both inhibitory and stimulatory components, due to the ability of ␣ 2 AR to couple to both G i and G s (12). In order to isolate either monophasic ␣ 2 AR-mediated inhibition or stimulation, transfected CHO cells were pretreated with either 20 g/ml of CTX or 500 ng/ml of PTX, respectively, for 16 h prior to membrane preparation and adenylyl cyclase assays. In all experiments, mutant and wild type ␣ 2 AR were studied at matched expression levels ranging from ϳ5-10 pmol of receptor/mg of membrane protein. Activities were determined in the presence of water (basal), 1.0 M forskolin, and 1.0 M forskolin with the indicated concentrations of the ␣ 2 AR agonist epinephrine.
Data Analysis-Data from adenylyl cyclase and agonist competition radioligand binding assays were analyzed by iterative least squares techniques (34). For determination of whether agonist competition data were best fit by a simple model (one-site curve) or a more complex model (two-site curve), F-test analysis was used with assignment of the more complex model when p Ͻ 0.05.

Materials-[ 3 H]Yohimbine (80 Ci/mmol), [␣-32 P]ATP (30 Ci/mmol), and [ 3 H
]cAMP (31 Ci/mmol) were from DuPont NEN. Cholera toxin, forskolin, and epinephrine were from Sigma. Pertussis toxin was from List Biologicals. Geneticin (G418), penicillin, and streptomycin were from Life Technologies, Inc. Ham's F-12 medium and all other tissue culture reagents were from JRH Biosciences. Fig. 1 are the regions of the intracellular domains of the ␣ 2A AR that were investigated as potential G i or G s coupling domains. Three principal regions were chosen based on studies with other G-protein-coupled receptors, primarily the ␤ 2 AR (24 -26, 35-38), and the muscarinic receptors (39,40), and studies using synthetic peptides based on regions of the human ␣ 2A AR (27)(28)(29). These included the second intracellular loop and the most amino-and carboxyl-terminal regions of the third intracellular loop region directly adjacent to transmem-brane domains V and VI, respectively. For mutagenesis, each region of the ␣ 2A AR was substituted with the analogous sequence from either the cloned human ␤ 2 AR or 5-HT 1A R. Due to the coupling pathways of the ␤ 2 AR (G s -coupled) and the 5-HT 1A R (G i -coupled), mutant ␣ 2 AR containing substitution with ␤ 2 AR sequence were primarily utilized to discern losses in ␣ 2 AR-G i coupling, and mutants containing substitution with 5-HT 1A R sequence were primarily utilized to discern losses in ␣ 2 AR-G s coupling. CHO cells were stably transfected with the cDNAs encoding mutant and wild type ␣ 2 AR, and then pharmacological characterization and functional adenylyl cyclase assays were performed in washed membranes to determine the consequences of these mutations.

Shown in
All the mutant ␣ 2 AR bound the antagonist [ 3 H]yohimbine with high affinity and displayed guanine nucleotide-sensitive high affinity agonist binding (Table I). This suggests that the mutations made did not result in global conformational changes affecting ligand binding, which might confound subsequent interpretation of functional responses. Some small but statistically significant differences in K i values were found with some of the mutants as compared with the wild type ␣ 2A AR, as indicated (Table I). There were also some differences in K L or K H values; however, the relevance of such changes in the absence of functional studies was not clear at this juncture. Receptor expression for the wild type ␣ 2A AR was 7 Ϯ 1 pmol/mg of membrane protein (n ϭ 11), and for each experiment, mutant and wild type ␣ 2 AR were studied at matched levels of receptor expression.
In order to isolate and study ␣ 2 AR-G i versus G s coupling in membrane adenylyl cyclase assays, transfected CHO cells were pretreated with either CTX or PTX, respectively, as described under "Experimental Procedures." Previously, we have extensively confirmed that losses in functional G-protein coupling observed using membranes prepared from transfected CHO cells pretreated with toxin are also found using untreated CHO cells (seen as a loss of one limb of the biphasic response) and using cell lines that exclusively display either ␣ 2 AR-mediated inhibition or stimulation of adenylyl cyclase activity (20). Therefore, in order to assess the loss of one or both of these coupling pathways in a single cell system and to simplify analysis with monophasic dose-response curves for inhibition or stimulation, we have utilized transfected CHO cells pretreated with the appropriate toxin.
Amino-terminal Region of the Third Intracellular Loop-Shown in Fig. 2 and summarized in Table II are  Shown is a two-dimensional schematic diagram of the proposed topology of the members of the G-protein-coupled receptor superfamily, which includes an extracellular amino terminus, an intracellular carboxyl terminus, and seven transmembrane spanning domains linked by three extracellular and three intracellular loop regions. Key portions of the intracellular regions within the human ␣ 2A AR were investigated as potential G i or G s coupling domains and were targeted by substitution with the analogous sequence from the G s -coupled ␤ 2 AR and the G i -coupled 5-HT 1A R. Substitution mutations of the ␣ 2A AR were undertaken that would replace the entire second intracellular loop, corresponding to amino acids 133-149, and the most amino-and carboxyl-terminal regions of the third intracellular loop of the ␣ 2A AR, corresponding to amino acids 218 -235 and 355-371, respectively. Substitutions of the second intracellular loop with the analogous sequence from the ␤ 2 AR or the 5-HT 1A R are denoted as ␣ 2 (␤ 2 2L) and ␣ 2 (5-HT 2L), respectively. Substitutions in the amino-and carboxyl-terminal regions of the third intracellular loop with ␤ 2 AR and 5-HT 1A R sequence are referred to as ␣ 2 (␤ 2 NT), ␣ 2 (5-HT NT), ␣ 2 (␤ 2 CT), and ␣ 2 (5-HT CT). Several additional mutations were constructed that contained combinations of the above substitutions, including substitution of the amino-and carboxyl-terminal regions of the third intracellular loop with ␤ 2 AR or 5-HT 1A R sequence (referred to as ␣ 2 (␤ 2 NT ϩ CT) and ␣ 2 (5-HT NT ϩ CT), respectively) and substitution of the second loop and the carboxyl-terminal region of the third intracellular loop with ␤ 2 AR sequence (referred to as ␣ 2 (␤ 2 CT ϩ 2L). elicited an increase of 308 Ϯ 25% over forskolin-stimulated adenylyl cyclase activity with an EC 50 of 16 Ϯ 1 M ( Fig. 2A and Table II). Following treatment with CTX, epinephrine elicited a decrease of 53 Ϯ 2% in forskolin-stimulated adenylyl cyclase activity with an EC 50 of 183 Ϯ 11 nM (Fig. 2B and Table II). In assays performed with the mutant ␣ 2 (5-HT NT) under the same conditions, no epinephrine-mediated stimulation was observed ( Fig. 2A), whereas epinephrine-mediated inhibition occurred identically to that observed for the wild type ␣ 2A AR (Fig.  2B). Similarly, the mutant ␣ 2 (␤ 2 NT) elicited epinephrine-mediated inhibition of adenylyl cyclase activity to the same extent as the wild type receptor with no loss in the maximum inhibition (67 Ϯ 2% versus 53 Ϯ 2% of forskolin-stimulated activity, respectively) and only a slight increase in the EC 50 (490 Ϯ 60 nM versus 183 Ϯ 11 nM, respectively). Thus, we considered the loss in the stimulatory pathway to be due to specific removal of G s coupling determinants in that ␣ 2 AR-G i coupling remained entirely intact. Interestingly, although ␤ 2 AR substitutions were not primarily used to observe losses in ␣ 2 AR-G s coupling and despite the importance of the amino-terminal domain of the ␤ 2 AR in ␤ 2 AR-G s coupling (35)(36)(37), the maximum epinephrine-mediated stimulation found with the mutant ␣ 2 (␤ 2 NT) was reduced by ϳ75% with a ϳ2.5-fold increase in the EC 50 as compared with the wild type receptor ( Fig. 2A and Table II). Taken together, these results indicate that the amino-terminal domain of the third intracellular loop is absolutely required for wild type functional ␣ 2 AR-G s coupling but not for ␣ 2 AR-G i coupling. Also, the data suggest that other regions of the receptor support G i coupling, because this function was not altered with the ␤ 2 AR substitution.
Carboxyl-terminal Region of the Third Intracellular Loop-Similar to what was found with substitution in the aminoterminal region, the mutant ␣ 2 (5-HT CT) displayed markedly reduced functional ␣ 2 AR-G s coupling as compared with the wild type ␣ 2A AR with a 36 Ϯ 5% versus 308 Ϯ 25% increase over forskolin-stimulated adenylyl cyclase activity, respectively ( Fig. 2A). Again, these losses appeared to be specific in that epinephrine-mediated inhibition was fully intact with this mutation (Fig. 2B and Table II). For ␣ 2 (␤ 2 CT), epinephrine-mediated inhibition occurred with no change in the EC 50 and, although statistically significant, only a slight reduction in the maximum decrease in forskolin-stimulated adenylyl cyclase activity (42 Ϯ 1% versus 53 Ϯ 2% with wild type ␣ 2A AR). Thus, even with substitution of the ␤ 2 AR sequence in the carboxyl terminus of the third intracellular loop, the ␣ 2 AR functionally couples to G i with the wild type phenotype. This was also what was found when ␤ 2 AR sequence was substituted in the amino terminus of the third intracellular loop. That is, ␣ 2 AR-G i coupling was not affected. This suggested that either these two regions are not important in G i coupling or that perhaps either domain of the ␣ 2A AR could independently provide for functional coupling to G i . On the other hand, G s coupling was ablated by the 5-HT 1A R substitution in the carboxyl terminus. Because this also occurred with 5-HT 1A R substitution in the amino terminus, these results suggested that both regions are required for G s coupling and do not function independently. Finally, substitution of the carboxyl-terminal region with ␤ 2 AR sequence entirely supported ␣ 2 AR-G s coupling, which was not an unexpected finding because this region of the ␤ 2 AR has been shown to be a G s coupling domain (25,26,36,38).
Second Intracellular Loop-The second intracellular loop was found to be necessary for both the stimulatory and inhibitory pathways. For the mutant ␣ 2 (␤ 2 2L), a substantial loss in epinephrine-mediated inhibition was observed with a ϳ27-fold increase in the EC 50 and a ϳ34% reduction in the maximum inhibition as compared with the wild type ␣ 2A AR (Fig. 3A and Table II). For the mutant ␣ 2 (5-HT 2L), epinephrine-mediated stimulation was entirely absent (Fig. 3B and Table II). Interestingly, for the mutant ␣ 2 (5-HT 2L) epinephrine-mediated inhibition occurred to a similar extent as the wild type receptor (Fig. 3A). The mutant ␣ 2 (␤ 2 2L) did not display a loss of ␣ 2 ARmediated stimulation; in fact, epinephrine-mediated stimulation was augmented as compared with the wild type ␣ 2A AR (Fig. 3B). These latter data suggest that the second intracellular loop regions of the 5-HT 1A R and the ␤ 2 AR are capable of specifically supporting their respective pathways, inhibition and stimulation of adenylyl cyclase activity, within the context of the ␣ 2A AR. By inference, within the second intracellular loop of the wild type ␣ 2A AR, the necessary elements for both pathways are present.
Combination Mutations-Because the single ␤ 2 AR substitu-

FIG. 2. Effects of substitution in the amino-and carboxyl-terminal regions of the third intracellular loop on ␣ 2 AR-G i and G s coupling.
Adenylyl cyclase activities were determined in membranes prepared from CHO cells stably expressing wild type the ␣ 2A AR and the mutants ␣ 2 (␤ 2 NT), ␣ 2 (5-HT NT), ␣ 2 (␤ 2 CT), and ␣ 2 (5-HT CT), which had been pretreated with either CTX or PTX to isolate G i or G s coupling, respectively. Shown are the R max values for both epinephrine-mediated stimulation or G s coupling (% increase over forskolin-stimulated activity, panel A) and epinephrine-mediated inhibition or G i coupling (% decrease from forskolin-stimulated activity, panel B) obtained with each of the mutants as compared with the wild type ␣ 2A AR. EC 50 values for epinephrine-mediated inhibition and stimulation for each mutant were not markedly different from the wild type ␣ 2A AR (see Table II). Forskolin-stimulated activities (pmol cAMP/min/mg) for G i coupling conditions were: wild type ␣ 2A AR, 72 Ϯ 5; ␣ 2 (␤ 2 NT), 61 Ϯ 15; ␣ 2 (5-HT NT), 40 Ϯ 9; ␣ 2 (␤ 2 CT), 66 Ϯ 24; and ␣ 2 (5-HT CT), 58 Ϯ 7. Forskolinstimulated activities (pmol cAMP/min/mg) for G s coupling conditions were: wild type ␣ 2A AR, 14 Ϯ 2; ␣ 2 (␤ 2 NT), 13 Ϯ 1; ␣ 2 (5-HT NT), 11 Ϯ 1; ␣ 2 (␤ 2 CT), 11 Ϯ 3; and ␣ 2 (5-HT CT), 55 Ϯ 11. Shown are the mean Ϯ S.E. of three to five experiments performed. *, p Ͻ 0.02 as compared with the wild type ␣ 2A AR. tions in the third intracellular loop had little or no effect on G i coupling and substitution in the second intracellular loop regions did not entirely eliminate it, we constructed and expressed two combinations of the original ␤ 2 AR mutants to examine the effects of substitution of multiple regions on ␣ 2 ARmediated inhibition of adenylyl cyclase activity: one in which both the amino-and carboxyl-terminal regions of the third intracellular loop were substituted with the analogous sequences from the ␤ 2 AR, referred to as ␣ 2 (␤ 2 NT ϩ CT) and one in which the carboxyl-terminal region of the third intracellular loop and the second intracellular loop were substituted with the analogous sequences from the ␤ 2 AR, referred to as ␣ 2 (␤ 2 CT ϩ 2L). The results from adenylyl cyclase assays performed on CTX-pretreated membranes from CHO cells expressing ␣ 2 (␤ 2 NT ϩ CT) and ␣ 2 (␤ 2 CT ϩ 2L) as well as those from the single ␤ 2 AR substitution mutations are shown in Fig. 4A. For the mutant ␣ 2 (␤ 2 NT ϩ CT), epinephrine-mediated inhibition was significantly reduced with a ϳ64-fold increase in the EC 50 as compared with the wild type ␣ 2A AR (Fig. 4A and Table II). This loss in ␣ 2 AR-mediated inhibition was most likely due to removal of specific G-protein coupling domains in that similar studies using a complementary mutant in which these two regions were substituted with 5-HT 1A R sequence (␣ 2 (5-HT NT ϩ CT)) demonstrated full preservation of the inhibitory pathway (Table II). Thus, although individual substitution of either the amino-or carboxyl-terminal side of the third intracellular loop does not affect functional ␣ 2 AR-G i coupling, substitution of both sides substantially reduced G i coupling. These findings suggest that both of the third intracellular loop regions are important components in ␣ 2 AR-G i coupling; however, it appears that each region can efficiently serve to fully support functional G i coupling domain in the absence of the other.
Because the ␣ 2 (␤ 2 2L) mutant displayed a loss of G i coupling, we wondered whether a combination mutation that included one of the third intracellular loop regions would be even more dysfunctional. This turned out to be the case, because the mutant ␣ 2 (␤ 2 CT ϩ 2L) showed the most profound loss in ␣ 2 AR-G i coupling with a ϳ1600-fold increase in the EC 50 for epinephrine-mediated inhibition of forskolinstimulated adenylyl cyclase activity as compared with the wild type receptor (Fig. 4A and Table II). In fact, no epinephrine-mediated inhibition of adenylyl cyclase activity was detected with this mutant except in the presence of the highest agonist concentrations in the assay. Thus, if the degree of agonist-mediated inhibition is compared between all of the ␤ 2 AR-substituted mutations at a single concentration of agonist (30 M) that elicits maximum inhibition with the wild FIG. 3. Effects of substitution in the second intracellular loop on ␣ 2 AR-G i and G s coupling. Adenylyl cyclase activities were determined in membranes prepared from CHO cells stably expressing the wild type ␣ 2A AR and the mutants ␣ 2 (␤ 2 2L) and ␣ 2 (5-HT 2L), which had been pretreated with either CTX or PTX to isolate ␣ 2 AR-G i coupling (panel A) or G s coupling (panel B), respectively, as described under "Experimental Procedures." The results are expressed as either the percentage of decrease or the percentage of increase from forskolinstimulated activity. Forskolin-stimulated activities (pmol cAMP/min/ mg) for G i coupling conditions were: wild type ␣ 2A AR, 72 Ϯ 5; ␣ 2 (␤ 2 2L), 67 Ϯ 9; and ␣ 2 (5-HT 2L), 63 Ϯ 9. Forskolin-stimulated activities (pmol cAMP/min/mg) for G s coupling conditions were: wild type ␣ 2A AR, 14 Ϯ 2; ␣ 2 (␤ 2 2L), 21 Ϯ 2; and ␣ 2 (5-HT 2L), 16 Ϯ 2. Shown are the means Ϯ S.E. from three to four experiments performed.

Summary of EC 50 and R max values for agonist-mediated inhibition and stimulation of adenylyl cyclase activity with mutant and wild type ␣ 2 AR
Summarized are the results from membrane adenylyl cyclase experiments using mutant and wild type ␣ 2 AR as described in Figs. 3-5. The results are expressed as the mean Ϯ S.E. R max values for inhibition and stimulation of adenylyl cyclase activity are expressed as the maximal percentage of decrease in forskolin-stimulated activity and the maximal percentage of increase over forskolin-stimulated activity, respectively. ND, not determined; NA, not applicable. type ␣ 2A AR, the pattern shown in Fig. 4B becomes evident. As shown, individual substitution of either the amino-or carboxyl-terminal regions of the third intracellular loop only minimally affected, if at all, the level of ␣ 2 AR-mediated inhibition. Equally significant losses in G i coupling are found with substitution of the second intracellular loop, alone, or substitution of both of the third intracellular loop regions together (Ͼ80% loss in the extent of inhibition of adenylyl cyclase activity as compared with the wild type ␣ 2A AR). Substitution of the second intracellular loop and one of the carboxyl-terminal side of the third intracellular loop together further reduces functional ␣ 2 AR-G i coupling to a complete loss of agonist-promoted inhibition of adenylyl cyclase activity at this concentration (Fig. 4B). DISCUSSION These studies have elucidated several key regions of the ␣ 2A AR required for functional G-protein coupling. Although a number of mutagenesis studies have been carried out with G s -coupled receptors (24 -26, 35, 37, 38, 41-45), only a few have been reported with G i -coupled receptors (39,40,46), and these have been limited in scope. None of these studies have examined coupling domains within a receptor, such as the ␣ 2 AR, that is dually coupled to G s and G i . For the second intracellular loop, we found that substitution with ␤ 2 AR sequence markedly perturbed G i coupling, and substitution with 5-HT 1A R sequence similarly affected G s coupling. Thus, within the second intracellular loop of the wild type ␣ 2A AR, domains supportive of both G s and G i coupling are present. This is consistent with studies of other G-protein-coupled receptors that indicate a critical role of the second intracellular loop for G-protein coupling/specificity (44,45), although its necessity for both an efficient (G i ) and a somewhat less efficient (G s ) pathway in a dually coupled receptor could not be predicted from such studies.
Within the third intracellular loop, we found evidence for two G i coupling domains, one in the amino terminus and the other in the carboxyl terminus. When one of these regions is removed by substitution with ␤ 2 AR, the mutant receptor appears to be minimally affected in its coupling efficiency to G i . When both are substituted, G i coupling is markedly impaired. This suggests that there are redundant, fully competent G i coupling domains within the ␣ 2 AR third intracellular loop. An alternative explanation would be that the ␤ 2 AR sequences used for substitution support, to some extent, G i coupling. We do not believe that this is tenable based on several observations. First, we have recently shown (20) that a deletion ␣ 2A AR mutant, where 11 amino acids of the amino terminus of the third loop were deleted, still had the capacity to couple to G i , presumably via the other domain in the carboxyl terminus. Secondly, several of the ␤ 2 AR substitution mutants reported here displayed efficient coupling to G i (equivalent to the wild type ␣ 2A AR), which would imply that the substituted sequences, if they support G i coupling, are highly efficient in doing so. Studies with wild type and mutant ␤ 2 AR have not, however, indicated any detectable G i coupling with wild type receptor (20,26). An entirely different set of findings were obtained when G s coupling was examined. For this pathway, substitution of 5-HT 1A R sequence in either the amino or carboxyl terminus resulted in a loss of G s coupling. This implies, in contrast to what was found for G i coupling, that G s coupling requires domains within both the amino and carboxyl terminus. Redundant G s coupling domains within the ␣ 2A AR, then, appear unlikely.
There are some notable differences between our results and those of Okamoto and Nishimoto where synthetic peptides based on ␣ 2A AR sequences corresponding to amino acids 131-148 (second intracellular loop region), 218 -229 (amino-terminal region of the third intracellular loop), and 356 -371 (carboxyl-terminal region of the third intracellular loop) were studied for their ability to stimulate GTP␥S binding to the purified G-proteins G i , G o , and G s in vitro (27,28). Each peptide displayed a specific pattern of GTP␥S stimulation with each of the G-proteins examined. The peptide representing the second intracellular loop of the ␣ 2A AR specifically stimulated GTP␥S binding to G s with high efficacy but did not stimulate GTP␥S binding to G i or G o . The peptide representing amino-terminal domain of the third intracellular loop stimulated binding to all three G-proteins similarly with low efficacy, whereas the peptide representing the carboxyl-terminal domain stimulated binding to G i and G o with similar efficacies and to G s with a significantly lower efficacy. Our findings with the intact recep- FIG. 4. Effects of combined ␤ 2 AR substitutions on ␣ 2 AR-G i coupling. A, shown are the results from adenylyl cyclase assays performed using membranes prepared from CTX-treated CHO cells stably expressing wild type ␣ 2A AR, combination mutants ␣ 2 (␤ 2 NT ϩ CT), and ␣ 2 (␤ 2 CT ϩ 2L) and mutants containing the single ␤ 2 AR substitutions as indicated. The results are expressed as the percentage of decrease from forskolin-stimulated adenylyl cyclase activity. Forskolin-stimulated activities (pmol cAMP/min/mg) for G i coupling conditions were: ␣ 2 (␤ 2 NT ϩ CT), 105 Ϯ 6 and ␣ 2 (␤ 2 CT ϩ 2L), 84 Ϯ 6. Forskolinstimulated activities (pmol cAMP/min/mg) for G s coupling conditions were: ␣ 2 (␤ 2 NT ϩ CT), 20 Ϯ 5 and ␣ 2 (␤ 2 CT ϩ 2L), 30 Ϯ 4. Forskolinstimulated adenylyl cyclase activities for wild type ␣ 2A AR and other mutants are reported in the legends for Figs. 2 and 3. Shown is the mean Ϯ S.E. from three to four experiments performed. B, comparison of the loss in G i coupling between the ␤ 2 AR substitution mutations shown in panel A expressed as the percentage of decrease in forskolinstimulated adenylyl cyclase activity obtained at a single concentration of agonist (30 M epinephrine) that elicited the maximum inhibitory response for the wild type ␣ 2A AR. *, p Ͻ 0.02 as compared with wild type ␣ 2A AR. tor provide strong support for a role of the second intracellular loop in ␣ 2 AR-G s coupling. However, in contrast to the peptidebased studies, we found that this region is also critically involved in functional ␣ 2 AR-G i coupling as well (Figs. 3 and 4 and  Table II). Furthermore, despite the low efficacy of GTP␥S binding reported for the amino-and carboxyl-terminal peptides for G s coupling, we found that these regions, like the second intracellular loop, were both critically important for functional stimulation of adenylyl cyclase activity by ␣ 2 AR. In agreement with the peptide studies, our data demonstrate that both the amino and carboxyl terminus of the third intracellular loop support ␣ 2 AR-G i coupling.
The most intriguing finding in our studies is that either the amino or the carboxyl terminus of the third intracellular loop can be substituted and the inhibitory pathway remains intact, yet clearly these regions are important in ␣ 2 AR-G i coupling because dual substitution greatly impairs the inhibitory response as does substitution of one of these regions in addition to the second intracellular loop. As mentioned above, this could perhaps be explained by the presence of structural elements within both regions that can equally support coupling to G i , thereby endowing the ␣ 2 AR with redundant G i coupling domains. That there may be redundant coupling domains to a single G-protein within a receptor has not been previously proposed. There are potential advantages for a receptor to have evolved in this manner. In the case of the ␣ 2 AR, redundant domains for G i coupling might serve the maintain a certain degree of function despite regulatory processes that occur via covalent modification of the receptor by kinase-mediated phosphorylation and/or association with regulatory molecules. That is, the extent of desensitization may be limited by the presence of two G i coupling domains. Although it is difficult to make comparisons, this notion is supported by the fact that agonistpromoted desensitization of recombinantly expressed ␤ 2 AR (47) is greater than that observed with the ␣ 2A AR expressed in the same cell line (48). Also, redundant domains may serve as a protective mechanism in the case of mutagenic events, which might introduce detrimental changes in one or the other region. It is also possible that structural features within these two domains of the ␣ 2 AR may also allow for the receptor to simultaneously couple to similar G-proteins, such as the separate G i isoforms. This cannot be addressed in the current study, because CHO cells express predominantly G i3 (49).
In summary, these studies have provided several novel insights into the structural elements that govern functional ␣ 2 AR-G-protein coupling. We have identified three key domains involved in both G i and G s coupling of ␣ 2 AR, which include the second intracellular loop and the amino-and carboxyl-terminal regions of the third intracellular loop. In addition, our results suggest that these regions serve distinct roles for G i versus G s coupling. For G s coupling, each of these regions is absolutely required. In contrast, for G i coupling, the second intracellular loop is absolutely required in addition to one of either the amino-or carboxyl-terminal region of the third intracellular loop. Each of the latter two regions appears to be capable of fully supporting ␣ 2 AR-G i coupling in the absence of the other and, in this manner, represent redundant G-protein coupling domains, each of which can independently function along with the second intracellular loop to provide for ␣ 2 AR-G i coupling.