Identification of A2a adenosine receptor domains involved in selective coupling to Gs. Analysis of chimeric A1/A2a adenosine receptors.

Responses to adenosine are governed by selective activation of distinct G proteins by adenosine receptor (AR) subtypes. The A2aAR couples via Gs to adenylyl cyclase stimulation while the A1AR couples to Gi to inhibit adenylyl cyclase. To determine regions of the A2aAR that selectively couple to Gs, chimeric A1/A2aARs were expressed in Chinese hamster ovary cells and ligand binding and adenylyl cyclase activity analyzed. Replacement of the third intracellular loop of the A2aAR with that of the A1AR reduced maximal adenylyl cyclase stimulation and decreased agonist potency. Restricted chimeras indicated that the NH2-terminal portion of intracellular loop 3 was predominantly responsible for this impairment. Reciprocal chimeras composed primarily of A1AR sequence with limited A2aAR sequence substitution stimulated adenylyl cyclase and thus supported these findings. A lysine and glutamic acid residue were identified as necessary for efficient A2aAR-Gs coupling. Analysis of chimeric receptors in which sequence of intracellular loop 2 was substituted indicated that the nature of amino acids in this domain may indirectly modulate A2aAR-Gs coupling. Replacement of the cytoplasmic tail of the A2aAR with the A1AR tail did not affect adenylyl cyclase stimulation. Thus, selective activation of Gs is predominantly dictated by the NH2-terminal segment of the third intracellular loop of the A2aAR.

The endogenous release of adenosine or the administration of adenosine analogs produces a variety of physiological effects (1)(2)(3)(4). These responses result from the activation of cell surface adenosine receptors (ARs) 1 that belong to the family of GPCRs. To date, four AR subtypes, A 1 , A 2a , A 2b , and A 3 , have been cloned from a variety of species (5,6). Like other members of the GPCR family, the nature of the physiological responses to adenosine or its analogs is partly governed by the affinity of the individual AR subtypes for agonists, and perhaps more importantly, the selective coupling of the activated receptor subtype to distinct G proteins. The A 1 AR (7,8) and A 3 AR (9, 10) are coupled to G i proteins. The inhibition of adenylyl cyclase has been classically associated with A 1 AR activation (2) while a pertussis toxin-sensitive stimulation of phospholipase C has been described for cells endogenously expressing the A 3 AR (11). The A 2a AR and A 2b AR are coupled to G s activation with the resulting stimulation of adenylyl cyclase, increase in cellular cAMP levels and protein kinase A activation (2). Activation of the A 2a AR-G s signal transduction pathway appears to be involved in adenosine-induced vasodilation (12,13), inhibition of platelet aggregation (14), and modulation of neutrophil function (15,16) although the complete signal transduction pathways have yet to be delineated.
In that most cell types possess a variety of G proteins, it is assumed that the fidelity of AR to G protein signaling is dictated by structural differences in specific regions of the receptors that physically contact specific G protein ␣ subunits. With the recent recognition of the importance of G protein ␤␥ subunits in intracellular signaling, studies have also begun to focus on receptor regions which may interact with these subunits (17). The identification of regions of ARs involved in signaling to G proteins has not been reported. However, for several GPCRs such as the adrenergic receptors (18 -24) and muscarinic receptors (25)(26)(27)(28)(29)(30), mutational analysis of receptor coupling to G proteins has been extensive. These studies have delineated the involvement of multiple and probably interdependent cytoplasmic domains of the GPCRs in transmitting signals to G proteins. In particular, genetic engineering of intracellular loop 2, intracellular loop 3, and the COOH-terminal tail has demonstrated the importance of these segments. Studies employing peptides whose design is based on these cytoplasmic domains of GPCRs (31)(32)(33)(34) have supported the results of mutagenesis studies. To a lesser extent, mutations in intracellular loop 1 (35,36) and receptor transmembrane domains (37,38) have also resulted in altered G protein signaling. This extensive analysis has indicated that the precise location of amino acids of the receptor involved in G protein activation is not the same among the individual GPCRs.
In order to define the regions of the A 2a AR responsible for the selective activation of G s , the present study profiles the functional responses of a series of chimeric ARs. From the analysis of chimeric ARs designed for this study, it is found that coupling of the A 2a AR to G s is predominantly dictated by amino acids constituting the NH 2 -terminal portion of intracellular loop 3. Individual amino acids in this region that have at least a partial role in G s coupling are identified. The nature of the residues comprising the most COOH-terminal portion of intracellular loop 2 is also apparently critical. The relatively large cytoplasmic tail of the A 2a AR appears to have little importance in the selective coupling of the receptor to G s .

EXPERIMENTAL PROCEDURES
Materials-Vent DNA polymerase (New England Biolabs) and Taq DNA polymerase (Life Technologies) were used for the construction of all genetically engineered receptors. ZM241385 was prepared, radioiodinated, and subsequently purified by high performance liquid chromatography as recently described (39). All radiochemicals were from Du-Pont NEN. Cell culture supplies and pertussis toxin were from Life Technologies. Forskolin was from Calbiochem and R-PIA purchased from Boehringer Mannheim. NECA was a gift of Dr. R. Olsson (University of South Florida).
Mutagenesis-Sequences derived from the canine A 2a AR (40) and human A 1 AR (41) were used to construct all chimeric receptors. Chimeric receptors employed in this study are identified as A2-(majority of structure derived from the A 2a AR) or A1-(majority of structure derived from the A 1 AR) followed by the region in which the substitution of sequence from the donor receptor was made. For more restricted chimeras, an arrow (3) follows the specific amino acids of the A 2a AR that were replaced with those of the A 1 AR. For point mutations, the A 2a AR amino acid is given in single letter code followed by position number and then the residue used to make the substitution. The majority of chimeric receptors were constructed using a three-step polymerase chain reaction approach with oligonucleotides consisting of both A 2a AR and A 1 AR sequences defining the splicing regions as described previously (42). For constructs containing limited amino acid sequence, a two-step polymerase chain reaction technique was employed with a single oligonucleotide specifying the base substitutions. Constructs were sequenced with Sequenase 2.0 (U. S. Biochemical Corp.) to confirm the presence of the desired mutations.
Cell Culture and DNA Transfection-CHO cells maintained in Ham's F-12 media supplemented with 10% fetal bovine serum and penicillin (100 units/ml)/streptomycin (100 g/ml) were used for all studies. All cDNAs were subcloned into the pCMV5 expression vector (Dr. D. Russell, University of Texas Southwestern). For transient receptor expression, nearly confluent monolayers of CHO cells were transfected via a modified DEAE-dextran procedure (43) employing varying amounts (5-30 g/75-cm 2 flask depending on the construct) of receptor cDNA to obtain approximately the equivalent amount of receptor expression for functional studies. Membranes were prepared from cells approximately 72 h after transfection and employed immediately for radioligand binding and adenylyl cyclase assays.
Radioligand Binding and Adenylyl Cyclase Assays-A 75-cm 2 flask of transfected CHO cells was washed twice with 10 ml of ice-cold 10 mM Tris, 5 mM EDTA, pH 7.4, at 5°C and cells were then scraped into 6 ml of the same buffer. Cells were disrupted on ice by 20 strokes by hand in a glass homogenizer and aliquoted equally in two tubes for radioligand binding and adenylyl cyclase assays and the homogenates were centrifuged at 43,000 ϫ g for 10 min. Membrane pellets were resuspended in 50 mM HEPES, 10 mM MgCl 2 , pH 6.8, or 50 mM Tris, 10 mM MgCl 2 , 1 mM EDTA, pH 8.26, at 4°C for 125 I-ZM241385 or [ 3 H]DPCPX binding assays, respectively. All membrane preparations were then treated with 2 units/ml adenosine deaminase. Saturation and competition binding assays were performed exactly as described previously (39,42). In general, competition binding assays performed in the presence and absence of 10 M Gpp(NH)p were conducted for only those constructs displaying impaired adenylyl cyclase activity.
Assay of membrane adenylyl cyclase activity was performed via the method of Salomon (44) as described previously (45). Briefly, membrane pellets were resuspended in TNM buffer (75 mM Tris, 200 mM NaCl, 1.25 mM MgCl 2 , pH 8.12, at 5°C) and treated with 2 units/ml adenosine deaminase for 5 min at 30°C. Adenylyl cyclase assays consisted of 40 l of membrane suspension, 40 l of cyclase mixture (TNM buffer supplemented with 140 M dATP, 5 M GTP, 30 units/ml creatine kinase, 5 mM creatine phosphate, 2.2 mM dithiothreitol, 100 M papaverine, and 1.5 Ci of [␣-32 P]ATP) and 20 l of increasing concentrations of agonist or forskolin at a final concentration of 1 M. Assays were conducted at 30°C for 15 min and terminated by addition of a stop solution containing 20,000 cpm/ml [ 3 H]cAMP. Labeled cAMP was isolated by sequential chromatography over Dowex and alumina columns and quantities determined by liquid scintillation counting. Protein concentrations were determined via the Bradford assay (46).
Pertussis Toxin Treatment-To determine the effects of pertussis toxin on adenylyl cyclase activity, two 75-cm 2 flasks of CHO cells were transiently transfected with receptor constructs as described above. 24 h post-transfection, cells were detached with trypsin, pooled, and aliquoted into three 75-cm 2 flasks. The following day, a single flask was treated with 200 ng/ml pertussis toxin. Following a 24-h incubation, adenylyl cyclase activity was determined in membranes obtained from control and pertussis toxin-treated cells. The remaining flask of cells was used for receptor quantification via antagonist radioligand saturation binding.
Data Analysis-125 I-ZM241385 and [ 3 H]DPCPX saturation binding curves and NECA versus 125 I-ZM241385 competition binding data were analyzed via a computer modeling program as described previously (47). For adenylyl cyclase assays, maximum responses to agonist are reported as a percentage of the adenylyl cyclase activity induced by 1 M forskolin and agonist dose-response curves were analyzed via the computer modeling system described above to determine EC 50 values. An ANOVA was used to compare ligand binding and adenylyl cyclase parameters of all mutant receptors to those obtained for the WT A 2a AR.

RESULTS
To begin to identify regions of the A 2a AR which upon agonist binding are responsible for activation of adenylyl cyclase via G s coupling, chimeric adenosine receptors in which cytoplasmic regions of the canine A 2a AR were replaced with analogous segments of the human A 1 AR were constructed, transiently expressed in CHO cells, and pharmacologically characterized. Expression of receptor constructs was quantitated via saturation binding analysis with the antagonist radioligand, 125 I-ZM241385 (39). As certain studies have demonstrated a correlation between the level of GPCR membrane expression and agonist activity in functional assays (48 -50), approximately equivalent levels (ϳ3.0 -5.0 pmol/mg) of receptor expression were obtained for the majority of constructs by varying the amount of cDNA employed in transfections. As the expression of a limited number of constructs ranged from a low level of ϳ1.5 pmol/mg, WT A 2a AR activity was also examined at this relatively lower receptor density.
Initial chimeric receptors contained the relatively large sequence substitution of the entire carboxyl terminus tail (A2-Tail) and entire third intracellular loop (A2-IC3) of the A 2a AR with the corresponding segments of the A 1 AR (Fig. 1). The pharmacological profiles of WT A 2a AR expressed at 2 levels, A2-Tail and A2-IC3, are shown in Table I. In membranes from untransfected CHO cells, NECA produced a maximal stimulation of adenylyl cyclase that was Ͻ5% of that induced by forskolin (data not shown). The WT A 2a AR, expressed at a level of ϳ4.3 pmol/mg, responded to NECA with a maximal stimulation of adenylyl cyclase activity which was 87.6% Ϯ 11.0% of that induced by 1 M forskolin. This represented an approximate 7-fold increase in adenylyl cyclase activity above basal levels. The EC 50  ing WT A 2a AR expression by ϳ66% did not significantly affect these parameters of adenylyl cyclase stimulation. At either level, 125 I-ZM241385 displayed high affinity binding as described previously (39). Replacement of the entire carboxyl terminus tail of the A 2a AR with that of the A 1 AR (A2-Tail) did not diminish the maximal stimulation of adenylyl cyclase induced by NECA nor significantly affect the EC 50 of the agonist. Conversely, replacement of the entire third intracellular loop of the A 2a AR (A2-IC3) resulted in an ϳ75% reduction in maximal adenylyl cyclase stimulation relative to wild-type receptor. Additionally, at A2-IC3 the EC 50 of NECA increased ϳ5-fold.
To determine which regions of the third intracellular loop replacement constituting A2-IC3 may be responsible for this chimera's impaired stimulation of adenylyl cyclase, a series of more restricted chimeric receptors was created. Fig. 1 contains a sequence alignment of intracellular loop 3 of the canine A 2a AR and human A 1 AR as well as a schematic representation of the mutants focusing on the third intracellular loop of the A 2a AR. Of the four chimeric receptors studied (Table II), only A2-IC3N representing substitution of 20 amino acids at the NH 2 -terminal portion of the loop displayed diminished capacity to activate adenylyl cyclase. In response to NECA, A2-IC3N was able to mediate a maximal stimulation of adenylyl cyclase approximately 50% of that observed at the WT A 2a AR. The EC 50 for NECA (189.5 Ϯ 34.0 nM), however, was similar to that displayed by the WT A 2a AR. At receptors A2ER3 AA and A2RSTL3 QKYY that both consisted of sequence substitution in the mid-portion of the third intracellular loop, NECA induced a maximal stimulation of adenylyl cyclase similar to that at the WT A 2a AR. The potency of NECA at A2ER3 AA (EC 50 ϭ 492.7 Ϯ 75.9 nM) was slightly lower than at WT A 2a AR. Data obtained with A2-KVSAS (see below) also indicated that sequence in the mid-portion of the third intracellular loop had no role in the selectivity of G protein coupling by the A 2a AR. A2-IC3C consisting of amino acid substitutions in the COOHterminal segment of the third intracellular loop of the A 2a AR stimulated adenylyl cyclase in response to NECA in a fashion nearly identical to that of the WT A 2a AR.
Taken together, the above results suggest that the impairment of adenylyl cyclase observed with the chimera A2-IC3 occurred principally due to replacement of amino acids in the NH 2 -terminal portion of the third intracellular loop. In order to further identify residues in the 20-amino acid replacement of A2-IC3N which may confer coupling to G s , more restricted chimeric receptors were created ( Fig. 1, Table III). A2-KVSAS represents a substitution of solely the distal 5 amino acids constituting the A2-IC3N chimera. At A2-KVSAS, NECA induced a maximal stimulation of adenylyl cyclase and displayed a potency similar to that at the WT A 2a AR. Based on this finding, subsequent mutagenesis focused on the first 15 amino acids of the third intracellular loop. Unfortunately, chimeric receptors and a deletion mutant constructed to further study this region did not display appreciable 125 I-ZM241385 binding nor stimulate adenylyl cyclase in response to NECA regardless of the amount of transfected cDNA. It is assumed that such proteins did not undergo the processing or folding required for proper membrane insertion or orientation. The presently employed sequence alignment of the A 1 AR and A 2a AR indicated that 4 of the 15 amino acids in this region are conserved among the receptors, thus several of the remaining nonconserved residues were targeted for point mutations (Fig. 1, Table III). For the three mutant receptors examined, amino acids of the WT  To determine if the impaired stimulation of adenylyl cyclase by those mutant receptors described above resulted from reduced affinity of the receptors for agonist, the ability of NECA to compete for 125 I-ZM241385 binding was analyzed. An additional parameter of receptor-G protein coupling, the sensitivity of agonist high affinity binding to the nonhydrolyzable guanine nucleotide Gpp(NH)p, was also examined.  1.8 nM). In the presence of Gpp(NH)p, NECA binding at A2-IC3 was best fit by a one-site model with relatively high affinity binding remaining (K i ϭ 31.4 Ϯ 6.3 nM). At A2-IC3N, NECA binding was consistently best fit to a two-site model with parameters very similar to those of the WT A 2a AR (K H ϭ 13.2 Ϯ 3.4 nM; K L ϭ 250.9 Ϯ 99.1 nM; %R H ϭ 82.6 Ϯ 6.9%). Binding was not significantly affected by Gpp(NH)p. Agonist binding to both K209N and E212Q was best fit by a single-site model with K i values of 35.2 Ϯ 3.4 and 61.9 Ϯ 18.3 nM, respectively. As shown in Table IV, the addition of Gpp(NH)p did not affect NECA binding at either of these two point mutants.
The analysis of the chimeric A 1 /A 2a ARs described above indicated that replacement of the third intracellular loop, and in particular its NH 2 -terminal domain, of the A 2a AR with the analogous segment of the A 1 AR resulted in an impairment of NECA-stimulated adenylyl cyclase activity relative to the WT A 2a AR. This diminished response of the chimera may have resulted from either the removal of amino acids in the A 2a AR required for selective coupling to G s or due to introducing amino acid sequence from the G i -coupled A 1 AR or a combination of these factors. Upon activation by agonist, a chimeric AR containing sequence from the A 1 AR may at least partially produce a diminished adenylyl cyclase stimulation due to the activation of G i proteins leading to inhibition of adenylyl cyclase. Additionally, it has been shown that chimeric GPCRs containing intracellular domains derived from parent receptors that individually couple to distinct G proteins may demonstrate a promiscuity in signal transduction (29). In order to identify the mechanism(s) responsible for the diminished functional responses observed with A2-IC3 and A2-IC3N, NECAinduced adenylyl cyclase stimulation at these chimeras was studied in membranes derived from pertussis toxin-treated cells. If these chimeras were coupled to G i and thus upon activation by agonist produced partial inhibition of adenylyl cyclase, the stimulation of the enzyme may be increased in membranes in which pertussis toxin had abolished signaling via G i proteins.  ml ϫ 24 h). These conditions for pertussis toxin treatment completely abolished the A 1 AR-mediated inhibition of forskolin-stimulated adenylyl cyclase observed in membranes obtained from CHO cells stably expressing the human A 1 AR (data not shown). As described under "Experimental Procedures," expression levels of WT A 2a AR and A 2a /A 1 AR chimeras were equivalent between control and treated cells. At the WT A 2a AR, pertussis toxin treatment had no effect on the maximal stimulation of adenylyl cyclase by NECA. This finding indicates that even at high agonist concentrations, no dual coupling to G s and G i by the WT A 2a AR was unmasked by pertussis toxin treatment. However, for both A2-IC3 and A2-IC3N, adenylyl cyclase stimulation was enhanced by pertussis toxin treatment relative to control. At A2-IC3, pertussis toxin increased adenylyl cyclase activity by ϳ78% relative to untreated cells although stimulation remained lower than that observed at the WT A 2a AR. At A2-IC3N, pertussis toxin treatment likewise increased adenylyl cyclase stimulation by ϳ56% relative to control membranes. The maximal response induced by NECA at A2-IC3N in the presence of pertussis toxin (66.9 Ϯ 9.6%) approached that obtained with the WT A 2a AR. At WT A 2a AR, A2-IC3, and A2-IC3N, pertussis toxin had no effect on the EC 50 of NECA for the stimulation of adenylyl cyclase. Data derived following replacement of the entire third intracellular loop (A2-IC3) or solely the NH 2 -terminal portion of this loop (A2-IC3N) of the A 2a AR with the analogous regions of the A 1 AR indicate the importance of these segments in adenylyl cyclase stimulation mediated by the A 2a AR. However, it is possible that a loss of function by the A 2a AR resulting from sequence substitution may have occurred due to nonspecific disruption of receptor structure. To determine if the identified regions when placed into the A 1 AR could mediate activation of adenylyl cyclase, the reciprocal chimeric receptors were created (Fig. 1), transiently expressed in CHO cells, and adenylyl cyclase activity in response to the agonist R-PIA was analyzed. Receptor expression levels were quantitated via radioligand binding with the A 1 AR selective antagonist, [ 3 H]DPCPX. Adenylyl cyclase assays were performed with control and pertussis toxin-treated cells (Fig. 2). In the presence of pertussis toxin, any residual coupling of these chimeric receptors that are composed primarily of A 1 AR sequence to G i proteins would be eliminated and thus make stimulation of adenylyl cyclase activity more readily apparent. Membranes prepared from pertussis toxin-treated cells expressing the WT A 1 AR displayed no stimulation of adenylyl cyclase activity in response to R-PIA (data not shown). Replacement of the entire third intracellular loop of the A 1 AR with that of the A 2a AR produced a receptor, A1-IC3, that responded to R-PIA with a maximal increase in adenylyl cyclase activity that was 18.7 Ϯ 4.4% of that induced by 1 M forskolin. The EC 50 of R-PIA at A1-IC3 was 9.75 Ϯ 1.2 nM which is similar to that at the wild-type A 1 AR for the inhibition of adenylyl cyclase activity (53). In membranes prepared from pertussis toxin-treated A1-IC3 cells, maximal adenylyl cyclase stimulation by R-PIA was 27.9 Ϯ 7.4% with an EC 50  A series of chimeric ARs was also created in order to examine the role of the second intracellular loop of the A 2a AR in coupling to the stimulation of adenylyl cyclase. As shown in Fig. 3, amino acids constituting the mid-portion of intracellular loop 2 are conserved among the A 2a AR and A 1 AR, suggesting these residues are not involved in the fidelity of G protein coupling. Thus, focus of this mutational analysis to study selective G s coupling was on the NH 2 -and COOH-terminal portions of this region (Fig. 3). Results obtained with these chimeras and subsequently analyzed point mutations are shown in Table VI. The 4-amino acid substitution from the A 1 AR constituting the A2-IC2N chimera designed to study the NH 2 -terminal residues had no effect on NECA-stimulated adenylyl cyclase activity  Similarly, the more restricted chimera A2NGL3 KMV demonstrated unimpaired functional response. However, substitution of Gly-118 and Thr-119 that are the amino acids at the junction of intracellular loop 2 and transmembrane domain 4 in the A 2a AR with the analogous amino acids of the A 1 AR (proline and arginine, respectively), substantially altered adenylyl cyclase activation. At A2GT3 PR, NECA induced a relatively unimpaired maximal stimulation of adenylyl cyclase (84.5 Ϯ 12.9%) but was approximately 50-fold less potent (EC 50 ϭ 12,600 Ϯ 5190 nM) than at the WT A 2a AR. This rightward shift of the NECA dose-response curve was approximately 10-fold greater than that observed at any other construct examined in this study. Competition binding assays with A2GT3 PR also reflected altered receptor-G protein coupling (Table IV). Relative to WT A 2a AR, both the K H and K L values obtained for NECA at A2GT3 PR increased approximately 5-fold. Additionally, only 14.8 Ϯ 7.5% of the A2GT3 PR population was in the agonist high affinity state which is an ϳ75% decline relative to WT A 2a AR. In an attempt to identify an individual substitution in A2GT3 PR responsible for its impaired signaling, two single point mutations were constructed. However, at both G118P and T119R the parameters for NECA-induced adenylyl cyclase stimulation including EC 50 values were identical to those obtained at the WT A 2a AR. The construct A2GT3 AA in which Gly-118 and Thr-119 were both replaced with alanine rather than the analogous A 1 AR residues activated adenylyl cyclase in response to NECA in a fashion similar to that of the WT A 2a AR. NECA competition binding at A2GT3 AA was best fit by a one-site model and remained of relatively high affinity (K i ϭ 30.5 Ϯ 4.6 nM).

DISCUSSION
Through the analysis of a series of chimeric adenosine receptors composed of human A 1 AR and canine A 2a AR sequence, the present study has identified regions of the A 2a AR required for efficient coupling to G s and thus the stimulation of adenylyl cyclase. In the A 2a AR, intracellular loop 3 and in particular the NH 2 -terminal region of this domain appears to have the predominant role in conferring selective receptor coupling to G s with the nature of amino acids in the most COOH-terminal portion of intracellular loop 2 also being important. This is the first description of a structural analysis of AR-G protein signal transduction. This study focused on determinants of the selective coupling of the A 2a AR to G s . Thus, data generated from chimeras that did not demonstrate impaired adenylyl cyclase stimulation do not indicate that the targeted regions have no role in G protein signaling but rather suggest they are not involved in maintaining the fidelity of specifically G s activation. Such regions which may be conserved among certain receptor subtypes may be involved in a general activation of G protein ␣ subunits or perhaps involved in contact with G protein ␤␥ subunits.
The structure of the A 2a AR is interesting in that it possesses a cytoplasmic tail approximately 80 amino acid residues longer than any other cloned AR including the similarly G s -coupled A 2b AR. However, this tail does not appear to be involved in selective G s signaling by the A 2a AR as its complete replacement with that of the A 1 AR (A2-Tail) in this study did not impair NECA-induced adenylyl cyclase stimulation. In a study of chimeric ␤ 2 /␣ 2 -adrenergic receptors, Liggett and co-workers (20) found the proximal portion of the ␤ 2 -adrenergic receptor cytoplasmic tail to have a partial role in mediating the stimulation of adenylyl cyclase in response to isoproterenol. Alternative splicing of the cytoplasmic tail of the EP3 prostaglandin E receptor was reported to produce receptor subtypes that differentially couple to G proteins including G s (54). Conversely, studies of receptor chimeras generated from endothelin receptor subtypes (55) as well as vasopressin receptor subtypes (56) have shown that solely the replacement of the cytoplasmic tail of the G s -coupled subtype with that of the G i -coupled subtype resulted in no impairment of adenylyl cyclase stimulation relative to wild-type receptor. Even with the data presented in this study, the function of the relatively large cytoplasmic tail of the A 2a AR has yet to be defined.
In contrast to the results obtained for the cytoplasmic tail, intracellular loop 3 of the A 2a AR appears to be critical for full activation of G s and the resulting adenylyl cyclase stimulation. Replacement of the entire loop of the A 2a AR with that of the A 1 AR resulted in an ϳ75% decrease in the maximal adenylyl cyclase response elicited by NECA as well as a 5-fold shift to the right of the agonist dose-response curve. In a large part although not completely, the diminished maximal adenylyl cyclase response observed at A2-IC3 appears to arise as a result of replacement of amino acid sequence in the NH 2 -terminal portion of the loop. The effect of replacement of NH 2 -terminal residues in intracellular loop 3 (ϳ50% decrease in maximal stimulation) did not fully mimic the response observed with the entire loop substitution. However, analysis of 4 chimeras targeting remaining portions of intracellular loop 3 suggested little or no role for additional sequence in this cytoplasmic segment in selective coupling of the A 2a AR to G s . It is possible that the lack of detection of an impaired adenylyl cyclase response may have arisen due to the design of the chimeras A2-KVSAS, A2ER3 AA, A2RSTL3 QKYY, and A2-IC3C in that creation of chimeras with slightly altered boundaries of sequence substitution may have signaled differently. Conversely, the NH 2 -terminal portion of intracellular loop 3 may indeed be the primary point of contact with G s and this region may possibly be presented to the G protein in different conformations in chimeras A2-IC3 and A2-IC3N as a result of the differences in the adjacent sequence constituting the remainder of the loop.
The impaired adenylyl cyclase response observed with A2-IC3 and A2-IC3N relative to WT A 2a AR appeared to result from both the removal of sequence responsible for G s coupling as well as the introduction of A 1 AR sequence that promotes G i activation. Following pertussis toxin treatment, both A2-IC3 and A2-IC3N demonstrated enhanced adenylyl cyclase stimulation suggesting a dual coupling of these chimeras to G s and G i . This response was not observed with the WT A 2a AR. No mutation studies of A 1 AR-G i coupling have been reported, thus it is not known which regions of the A 1 AR are involved in G i coupling. Although generalizations may not be made with certainty, the third intracellular loop of other G i -coupled receptors has been implicated in G i activation (27,32,34,55).
The "gain of function" displayed by the two reciprocal chimeric receptors, A1-IC3 and A1-IC3N, strongly support the notion that the third intracellular loop and in particular the NH 2 -terminal portion of this region of the A 2a AR interacts with G s resulting in adenylyl cyclase stimulation. Both A1-IC3 and A1-IC3N were shown to stimulate adenylyl cyclase to a level approximately 25% that of the wild-type A 2a AR. This level of activity of A1-IC3N was observed only upon pertussis toxin treatment of the cells. This requirement likely arises from the presence in A1-IC3N of substantial amounts of A 1 AR sequence, particularly in the unmodified regions of the third intracellular loop, that retain the ability to productively couple to G i . Despite the degree of A 1 AR sequence present in both A1-IC3 and A1-IC3N, the potency of R-PIA at these chimeras is identical to that at the wild-type A 1 AR (53) indicating a highly efficient coupling to G s .
In agreement with the present results with the A 2a AR, several studies with other GPCRs have detailed the importance of the NH 2 -terminal portion of intracellular loop 3 in selectivity of G protein coupling. This region appears to have a predominant role in signaling by ␤-adrenergic receptor subtypes (21,25,29), the ␣ 1 -adrenergic receptor (22) and several muscarinic receptor subtypes (26,27,30,31,57), although the latter two receptor families are not coupled to G s . Additionally, an 11-amino acid cassette in the NH 2 -terminal region of intracellular loop 3 of the ␣ 2A -adrenergic receptor has been identified as apparently being responsible for the G s -coupled component of this receptor's dual signaling activity (24). Mutagenesis studies of other G s -coupled receptors have also identified the third intracellular loop as being crucial in signal transduction although more restricted structural analysis of this region was not performed (55,56).
Results with the chimeric receptors described above suggest that 15 amino acids constituting the proximal portion of intracellular loop 3 of the A 2a AR are crucial in coupling of the receptor to G s . Of these 15 residues, 4 are conserved between the A 2a AR and A 1 AR and several others represent apparently conservative substitutions. Targeting several of the remaining nonconserved residues resulted in the identification of single amino acids which appear to have a role in G s coupling. Replacement of lysine and glutamic acid at positions 209 and 212, respectively, of the A 2a AR each resulted in significant decreases in the potency of NECA for the stimulation of adenylyl cyclase. These residues are probably not solely responsible for G s coupling by the A 2a AR as the effects observed with the larger chimeric substitutions were not reproduced by the individual point mutations. It is probable that multiple amino acids in a specific conformation are required for the most efficacious coupling to G s . As with all mutagenesis studies of this nature, the precise role of amino acid(s) may not be unequivocally assigned as certain responses may occur due to indirect effects on overall protein architecture.
The nature of the impairment of adenylyl cyclase stimulation observed upon distinct mutations of intracellular loop 3 of the WT A 2a AR varied. The most profound disruption of functional response was observed with substitution of the entire loop as maximal adenylyl cyclase response as well as potency of NECA were both substantially affected. However, the more restricted chimera A2-IC3N responded to NECA with a diminished maximal response although the EC 50 of NECA was unaffected. Conversely, point mutations K209N and E212Q both responded to agonist with intact maximal adenylyl cyclase stimulation but with a decrease in NECA potency. Thus, it is possible that multiple amino acids of the WT A 2a AR must be responsible for full activation of G s (as judged by maximal adenylyl cyclase response) whereas individual amino acid mutations may disrupt the efficiency of the coupling. It has been shown that certain single point mutations in intracellular loop 3 of the M 1 muscarinic receptor disrupted receptor function to a greater extent than larger amino acid replacements that contained the same residue substitution (58). Thus, the context of the mutation may influence the functional response.
Distinctions in the process of receptor coupling to G protein are also indicated by a comparison of the parameters of functional response to those of agonist binding. Despite the impaired stimulation of adenylyl cyclase at A2-IC3 and A2-IC3N, these chimeras did not display a marked loss of high affinity agonist binding relative to the WT A 2a AR. In competition binding assays with A2-IC3 and A2-IC3N, both the agonist high affinity state and percentage of receptors in this population were not decreased relative to the WT A 2a AR. Treatment with Gpp(NH)p did not modulate agonist binding to the WT A 2a AR or A2-IC3N but did result in NECA binding to A2-IC3 being best fit to a one-site model. Point mutations in the third intracellular loop of the A 2a AR that produced a loss of NECA potency in adenylyl cyclase assays (K209N and E212Q) also altered receptor-G s coupling as determined in competition binding assays. NECA binding to both K209N and E212Q was best described by a single site model and was not sensitive to  Fig. 3. 125 I-ZM241385 binding and adenylyl cyclase assays were performed as described under "Experimental Procedures." All values represent mean Ϯ S.E. with the number of experiments given in parentheses following identification of receptor construct. Statistical analysis was applied to K d values obtained for 125 I-ZM241385 and the parameters of NECA-induced adenylyl cyclase activity. Values for WT A 2a AR are shown for comparison. guanine nucleotide treatment. Interestingly, the differences between the K i for NECA at K209N and E212Q relative to its K H at the WT A 2a AR, ϳ3and 5-fold respectively, are similar to the shifts in the EC 50 of NECA observed in adenylyl cyclase assays with these mutant receptors.
Finally, this study examined the role of the second intracellular loop in signaling by the A 2a AR. Focus was on the COOHterminal portion of intracellular loop 2 as it is possible this segment may be in close proximity to the NH 2 -terminal of intracellular loop 3 in the membrane embedded receptor. Additionally, this region has been implicated in mutagenesis studies of receptor-G protein coupling (25,29). It was found that tandem replacement of Gly-118 and Thr-119 at the junction of intracellular loop 2 and transmembrane domain 4 with the analogous residues of the A 1 AR greatly disrupted coupling of the A 2a AR to G s . The effect of this substitution must be profound as not only was NECA-induced adenylyl cyclase activation disrupted but affinity of the agonist for receptor decreased as did the percentage of receptors in the high affinity state. Gly-118 and Thr-119 of the A 2a AR may not be directly involved in G s activation and it is possible that their substitution with proline and arginine, respectively, may produce conformational changes in intracellular loop 2 and disrupt G protein coupling at other regions of the receptor. Several observations support this hypothesis. First, the individual replacement of either glycine or threonine with proline and arginine, respectively, did not affect receptor function relative to the WT A 2a AR. Second, replacement of these residues with alanines also did not disrupt receptor stimulation of adenylyl cyclase or high affinity NECA binding. Substitution with proline, an amino acid that disrupts ␣-helical structure, in combination with the threonine replacement may alter overall protein conformation of the A 2a AR although this amino acid combination is that present in the WT A 1 AR. Indeed, the proline and arginine substitution was also present in the A2-IC2C chimera that demonstrated unimpaired adenylyl cyclase stimulation. As discussed above, this finding again indicates the importance of the context in which an amino acid(s) substitution is made in affecting functional responses.
In summary, the coupling of the A 2a AR to G s is predominantly dictated by sequence in intracellular loop 3 of the receptor. In particular, the NH 2 -terminal region of the third intracellular loop of the A 2a AR appears to be responsible for the fidelity of G protein coupling. Lysine and glutamic acid residues in this region have significant roles in the efficiency of A 2a AR-G s coupling. The analysis of multiple chimeric A 1 / A 2a ARs did not suggest a significant role for other cytoplasmic domains of the A 2a AR in selective activation of G s . However, the nature of amino acids constituting the COOH-terminal region of intracellular loop 2 is critical as it may affect conformation of this domain. Future mutagenesis studies directed at examining sequence substitutions and point mutations made in combination may more precisely define structural requirements of A 2a AR-G s coupling.