Agonist-dependent phosphorylation and desensitization of the rat A3 adenosine receptor. Evidence for a G-protein-coupled receptor kinase-mediated mechanism.

A3 adenosine receptor (A3AR) activation contributes to both the cardioprotective and antihypertensive effects of adenosine. To date, no studies have examined the mechanisms by which this receptor undergoes rapid homologous desensitization. Therefore, a functional hemagglutinin epitope-tagged A3AR has been stably expressed in Chinese hamster ovary cells, and its regulation by the AR agonist 5'-N-ethylcarboxamidoadenosine (NECA) has been studied. Cellular exposure to NECA induces rapid (t1/2 = approximately 1 min) A3AR phosphorylation on serine and threonine residues. This is associated with a functional desensitization and a 30-40% reduction in the number of high affinity agonist binding sites as determined by radioligand binding assays. Activation of second messenger-regulated kinases could not mimic the effect of NECA, suggesting a role for G-protein-coupled receptor kinases (GRKs). In vitro phosphorylation assays demonstrate that phosphorylation of agonist-occupied A3ARs is enhanced by GRK2 and that cellular pretreatment with NECA dramatically inhibits subsequent GRK2-mediated phosphorylation in vitro. Therefore, the A3AR is phosphorylated in situ by a kinase similar or identical to GRK2, and this may be involved in rapid functional desensitization of the A3AR.

A 3 adenosine receptor (A 3 AR) activation contributes to both the cardioprotective and antihypertensive effects of adenosine. To date, no studies have examined the mechanisms by which this receptor undergoes rapid homologous desensitization. Therefore, a functional hemagglutinin epitope-tagged A 3 AR has been stably expressed in Chinese hamster ovary cells, and its regulation by the AR agonist 5-N-ethylcarboxamidoadenosine (NECA) has been studied. Cellular exposure to NECA induces rapid (t 1 ⁄2 ‫؍‬ ϳ1 min) A 3 AR phosphorylation on serine and threonine residues. This is associated with a functional desensitization and a 30 -40% reduction in the number of high affinity agonist binding sites as determined by radioligand binding assays. Activation of second messenger-regulated kinases could not mimic the effect of NECA, suggesting a role for G-protein-coupled receptor kinases (GRKs). In vitro phosphorylation assays demonstrate that phosphorylation of agonist-occupied A 3 ARs is enhanced by GRK2 and that cellular pretreatment with NECA dramatically inhibits subsequent GRK2-mediated phosphorylation in vitro. Therefore, the A 3 AR is phosphorylated in situ by a kinase similar or identical to GRK2, and this may be involved in rapid functional desensitization of the A 3 AR.
The multiple physiological effects of adenosine are mediated by the activation of cell surface adenosine receptors (ARs). 1 Biochemical and molecular cloning studies have demonstrated the existence of four AR subtypes designated A 1 , A 2a , A 2b , and A 3 (1). A 3 AR cDNA clones were initially isolated from rat testis and brain libraries (2,3), but subsequently isolated cDNAs from sheep and human sources, which encode proteins with a 70% identity to the rat protein, have also been designated as A 3 ARs (4,5). Characterization of the pharmacological properties of the recombinant rat A 3 AR led to the realization that the A 3 AR is the "atypical" AR expressed in a rat mast cell-derived tumor cell line RBL-2H3 (6,7).
Despite its relatively recent discovery, A 3 AR activation has already been implicated in contributing to several important physiological effects of adenosine, including vasodilation, bronchoconstriction, and cardioprotection (8). Moreover, evidence has been presented to suggest that these effects are initiated by A 3 AR activation of mast cells, thereby leading to the release of allergic mediators that are directly responsible for the observed phenomena (8). Therefore, an understanding of how A 3 AR signaling is regulated would be a significant advance toward controlling these events.
The initial characterization of the A 3 AR expressed in RBL-2H3 cells demonstrated that agonist-stimulated calcium mobilization is subject to a rapid, homologous desensitization (6,7). However, the molecular events responsible for this effect are currently unknown. Rapid termination of signaling by G-protein-coupled receptors is typically initiated by receptor phosphorylation events catalyzed by either second messenger-activated kinases or G-protein-coupled receptor kinases (GRKs) (9). The latter constitute a growing family of proteins that specifically phosphorylate agonist-occupied receptors. To date, six such kinases, termed GRKs 1 through 6, have been cloned from mammalian sources but, with the exception of GRK1 (rhodopsin kinase), the spectrum of receptor substrates for each GRK in vivo remains unknown (10).
To examine a role of receptor phosphorylation in regulating A 3 AR signaling, a functional hemagglutinin epitope-tagged rat A 3 AR has been expressed in Chinese hamster ovary cells. We report the first visualization of a recombinant A 3 AR and the first detailed characterization of adenosine receptor phosphorylation both in situ and in vitro. Moreover, evidence is presented that supports a role for a specific GRK isoform in mediating A 3 AR phosphorylation and desensitization.

EXPERIMENTAL PROCEDURES
Materials-125 I-AB-MECA was synthesized and purified by high performance liquid chromatography as described previously (11). IB-MECA (12) was the generous gift of Dr. Kenneth Jacobson (National Institutes of Health, Bethesda, MD). Cell culture supplies were from Life Technologies Inc. Radionuclides were from DuPont NEN. Horseradish peroxidase-conjugated streptavidin and biotin LC-hydrazide were from Pierce. Cellulose-coated plastic-backed chromatography plates were from Eastman Kodak Co. Monoclonal antibody 12CA5 was from Berkeley Antibody Company. Sources of other materials have been described elsewhere (11,13).
Receptor cDNA Constructs and Expression-The six amino acid influenza hemagglutinin (HA) epitope, DVPDYA, recognized by monoclonal antibody 12CA5 (14) was inserted at both the amino and carboxyl termini of the rat A 3 AR by insertion of polymerase chain reaction products generated using the previously described pCMV5/rat A 3 AR * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. cDNA as a template (11). Correct introduction of the epitope sequences was verified by dideoxynucleotide sequencing. The resulting plasmid was termed pCMV5/HA-A 3 AR.
Chinese hamster ovary (CHO) cell lines stably expressing the epitope-tagged A 3 AR cDNA were generated by co-transfecting cells with pCMV5/HA-A 3 AR and pSV2Neo in a 20:1 ratio using a modified calcium phosphate precipitation/glycerol shock procedure previously described (15). After selection in G418, resistant colonies were isolated, expanded, and screened for receptor expression by radioligand binding using 1 nM 125 I-AB-MECA (11). Cells were propagated in T-75 flasks with Ham's F-12 medium supplemented with 10% (v/v) fetal bovine serum, penicillin (100 units/ml), and streptomycin (100 g/ml) in a 37°C humidified atmosphere containing 5% CO 2 .
Intact Cell Receptor Phosphorylation-HA-A 3 AR-expressing CHO cells were plated into 6-well dishes at a density of approximately 1 ϫ 10 6 cells/well and cultured overnight in regular media. The next day, the cells were washed twice with phosphate-free Dulbecco's modified Eagle's medium and incubated for 90 min at 37°C with the same medium supplemented with 1 unit/ml adenosine deaminase and 200 Ci/ml [ 32 P]orthophosphate. After stimulation with the indicated agonists, reactions were terminated by placing the cells on ice and washing the monolayers twice with 3 ml of ice-cold phosphate-buffered saline. Cells were scraped into 1 ml of lysis buffer (50 mM Hepes, pH 7.5, 5 mM EDTA, 10 mM sodium fluoride, 10 mM sodium phosphate, 0.1 mM phenylmethylsulfonyl fluoride, and 10 g/ml each of soybean trypsin inhibitor, leupeptin, and pepstatin A), transferred to microfuge tubes, and lysed by vigorous vortexing. Membranes were pelleted by centrifugation (14,000 ϫ g for 10 min) and solubilized by resuspension in 150 l of lysis buffer supplemented with 1% (w/v) SDS followed by brief sonication and heating to 65°C for 15 min. After chilling to 4°C, 600 l of a concentrated nonionic detergent mixture was added to the solubilized membranes such that the final mix contained 1% (v/v) Triton X-100, 0.5% (w/v) sodium deoxycholate, 0.2% (w/v) SDS, and 150 mM sodium chloride in lysis buffer. Insoluble material was removed by centrifugation (14,000 ϫ g for 10 min), and the supernatant precleared by incubation for 1 h at 4°C with protein A-Sepharose in the presence of 0.2% (w/v) IgG-free bovine serum albumin. The precleared supernatant was then incubated for 2 h at 4°C with protein A-Sepharose and 7 l of 12CA5 ascites. Immune complexes were isolated by centrifugation, washed twice with detergent buffer containing 0.2 M ammonium sulfate and once with detergent buffer alone, and eluted from the protein A-Sepharose by the addition of electrophoresis sample buffer and incubation at room temperature for 45 min. Analysis was by SDS-PAGE using 10% (w/v) polyacrylamide resolving gels and autoradiography. Quantitation of phosphorylation experiments was by excision from the dried gel and Cerenkov counting of bands of interest.
Cell-Surface Labeling with Biotin LC-hydrazide-This was performed on cells in 6-well dishes basically as described by Lisanti et al. (16). Briefly, cell monolayers were washed with ice-cold PBS and then treated with 10 mM sodium periodate in PBS for 30 min at 4°C in the dark. After removal of the periodate solution and further washing with PBS, cells were washed once with 0.1 M sodium acetate, pH 5.5, and then incubated for 30 min at 4°C with 2 mM biotin LC-hydrazide in the same buffer; this procedure labels all cell surface carbohydrate residues with biotin. Cells were then washed prior to membrane preparation and immunoprecipitation with 12CA5 as described above. After SDS-PAGE, resolved proteins were transferred to a polyvinylidene difluoride membrane and nonspecific protein binding sites blocked by a 60-min incubation in blocking buffer (5% (w/v) skimmed milk solution in PBS containing 0.2% (v/v) Triton X-100 and 0.02% (w/v) thimerosal). The membrane was then incubated for 60 min at room temperature with 1 g/ml horseradish peroxidase-conjugated streptavidin in a high detergent skimmed milk solution. After three washes in blocking buffer and two washes in PBS, reactive proteins were visualized by an enhanced chemiluminescence protocol in accordance with the manufacturer's instructions (Renaissance, DuPont NEN).
GRK2 Purification-Recombinant bovine GRK2 was purified from Sf9 cells by previously published procedures 48 h after infection with the appropriate baculovirus construct (17)(18)(19). Using light-activated rhodopsin as a substrate, the specific activity of the purified enzyme was approximately 1 mol/min/mg protein.
In Vitro Assay of Receptor Kinase Activity-After treatment of HA-A 3 AR cells in T-75 flasks with or without agonist, incubations were terminated by placing the cells on ice, rapidly washing three times with ice-cold PBS, and scraping the cells into lysis buffer (10 mM Hepes, pH 7.5, 2 mM EDTA, 0.25 M NaCl, 10 g/ml each of soybean trypsin inhibitor and leupeptin, and 0.1 mM phenylmethylsulfonyl fluoride). After Dounce homogenization on ice (20 strokes), membranes were pelleted by centrifugation and resuspended in GRK assay buffer (25 mM Hepes, pH 7.5, 2.5 mM EDTA, and 7.5 mM MgCl 2 ) supplemented with 1 unit/ml adenosine deaminase and protease inhibitors for immediate use. Assays consisted of 40 l of membrane suspension, 40 l of kinase mixture (GRK assay buffer supplemented with 0.25 mM ATP, 0.88 mM dithiothreitol, 0.15 M okadaic acid, and 10 Ci of [␥-32 P]ATP), 10 l of vehicle or purified GRK, and 10 l of vehicle or NECA. After incubation at 30°C for 5 min, reactions were terminated by placing the tubes on ice and adding 0.5 ml of ice-cold stop solution (0.1 M sodium phosphate, pH 7.5, and 10 mM EDTA). Membranes were pelleted by centrifugation (14,000 ϫ g for 10 min), and the resulting pellets were solubilized in 1% (w/v) SDS prior to dilution in nonionic detergent buffer as described above. After centrifugation to remove insoluble material, detergent extracts were equalized by protein assay prior to immunoprecipitation with 12CA5 and analysis by SDS-PAGE and autoradiography as described above.
Phosphoamino Acid Analysis-Following SDS-PAGE, proteins were transferred to a polyvinylidene difluoride membrane. After overnight autoradiography, the region of the membrane corresponding to the phosphorylated HA-A 3 AR was excised, hydrated, and hydrolyzed at 110°C in 200 l of 5.7 M HCl for 90 min. The resulting hydrolysate was lyophilized and resuspended in chromatography buffer supplemented with phosphoamino acid standards. After spotting onto cellulose-coated plates, samples were subjected to ascending chromatography in an isobutyric acid, 0.5 M ammonium hydroxide (5:3, v:v) buffer system (20). Standards were visualized by ninhydrin staining, and 32 P-labeled amino acids were visualized by autoradiography.
Radioligand Binding and Adenylyl Cyclase Assays-Saturation binding experiments employing 125 I-AB-MECA were performed and analyzed as we have described previously (11,15). Adenylyl cyclase assays were performed exactly as described previously using IB-MECA (11). Dose-response curves were analyzed by a previously validated curve-fitting program (21).

Functional Expression of HA-A 3 AR in CHO Cells-Satura-
tion binding analysis of transfected cell membranes demonstrated that the expressed receptor bound the high affinity A 3 AR agonist radioligand 125 I-AB-MECA with K d and B max values of 2.04 Ϯ 0.38 nM and 0.89 Ϯ 0.38 pmol/mg, respectively (four experiments) (Fig. 1A). The K d value observed is similar to that exhibited by the untagged rat A 3 AR after expression in CHO cells and by the native rat A 3 AR in RBL-2H3 cells (11). Adenylyl cyclase assays demonstrated that the A 3 AR agonist IB-MECA inhibited 5 M forskolin-stimulated adenylyl cyclase activity in a dose-dependent manner (IC 50 ϭ 48.3 Ϯ 11.3 nM, three experiments) producing a maximal inhibition of 75 Ϯ 5% (three experiments) (Fig. 1B). Therefore, the addition of HAepitope sequences to the A 3 AR fails to diminish its ability to either bind agonist radioligand with high affinity or interact productively with G i -proteins to inhibit adenylyl cyclase.
To identify the HA-A 3 AR protein, we utilized the presence of three predicted sites for N-linked glycosylation within the A 3 AR sequence (3). Cell surface carbohydrate residues were covalently labeled with biotin by sequential treatment of cell monolayers with periodate and biotin LC-hydrazide. After immunoprecipitation with 12CA5 and SDS-PAGE, immunoprecipitated glycoproteins were visualized by probing blots with horseradish peroxidase-conjugated streptavidin. 12CA5 specifically immunoprecipitated a 50 -70-kDa glycoprotein from HA-A 3 AR-expressing cells but not from nontransfected CHO cells, demonstrating that this labeled protein represents the expressed HA-A 3 AR (Fig. 1C). This also demonstrates that HA-A 3 AR is appropriately processed with respect to utilization of glycosylation sites and is expressed on the cell surface, thereby rendering it accessible to agonist.
Agonist-stimulated HA-A 3 AR Phosphorylation-To address any possible role of protein kinases in regulating A 3 AR signaling, we directly examined receptor phosphorylation after exposure of transfected cells to the nonselective AR agonist NECA. Fig. 2A shows that 12CA5 immunoprecipitated a 50 -70-kDa phosphoprotein from transfected cells after pretreatment with 10 M NECA for 10 min. This protein migrated exactly with the HA-A 3 AR as identified by biotin labeling (Fig. 1C) and was not immunoprecipitated from nontransfected CHO cells (Fig. 2A). Therefore, the HA-A 3 AR protein is rapidly phosphorylated upon cellular exposure to agonist. Phosphoamino acid analysis revealed that phosphorylation was predominantly on threonine residues with some phosphoserine also being detected but no phosphotyrosine (Fig. 2B) In an attempt to identify which kinases may be responsible, transfected CHO cells were exposed to activators of several second messenger-regulated ki- nases. However, activation of protein kinase C by phorbol 12myristate 13-acetate, Ca 2ϩ -calmodulin kinases by the calcium ionophore A23187, or cyclic nucleotide-dependent kinases by forskolin and 8-bromo-cyclic GMP all failed to stimulate HA-A 3 AR phosphorylation under conditions in which NECA was effective (Fig. 2C).
Characterization of NECA-stimulated HA-A 3 AR Phosphorylation-The ability of NECA to stimulate HA-A 3 AR phosphorylation was dose-dependent (Fig. 3A). Modelling of data pooled from three experiments produced an EC 50 of 0.16 Ϯ 0.04 M (Fig. 3A). Time course experiments revealed that HA-A 3 AR phosphorylation was extremely rapid. Phosphorylation was detectable at the first time point examined (15 s) and was halfmaximal by 60 s (Fig. 3B). The response was maximal by 4 min and was sustained for at least 20 min in the presence of agonist (Fig. 3B).
Effect of NECA Pretreatment on HA-A 3 AR Function-To determine whether HA-A 3 AR phosphorylation was associated with changes in receptor function, transfected cells were treated with 10 M NECA for 10 min prior to membrane preparation for radioligand binding. These experiments demon- FIG. 2. Agonist-dependent phosphorylation of HA-A 3 AR. A, nontransfected and HA-A 3 AR-expressing CHO cells were metabolically labeled with [ 32 P]orthophosphate, exposed to 10 M NECA or vehicle for 10 min at 37°C, and then lysed for solubilization and immunoprecipitation with 12CA5. Immunoprecipitates were analyzed by SDS-PAGE and autoradiography as described under "Experimental Procedures." B, transfected cells were treated with agonist and immunoprecipitated with 12CA5 as described for A. Following SDS-PAGE, proteins were transferred to a polyvinylidene difluoride membrane, and the region corresponding to the phosphorylated HA-A 3 AR was excised for phosphoamino acid analysis as described under "Experimental Procedures." The migration of ninhydrin-stained phosphoamino acid standards in this TLC buffer system is indicated. C, HA-A 3 AR phosphorylation in response to various stimuli. After labeling with [ 32 P]orthophosphoric acid, transfected cells were incubated for 10 min at 37°C in the absence of any ligand (None), 10 M NECA, 100 nM phorbol 12-myristate 13acetate (PMA), 10 M calcium ionophore A23187 in the presence of 1.8 mM calcium chloride (A23187), 100 M forskolin (Forskolin), and 100 M 8-bromo-cyclic GMP (8BrcGMP). Membranes were then prepared for solubilization and immunoprecipitation with 12CA5 as described under "Experimental Procedures."

FIG. 3. Concentration and time dependences of agonist-stimulated HA-A 3 AR phosphorylation.
A, 32 P-labeled transfected CHO cells were incubated for 10 min at 37°C with the indicated concentrations of NECA. Membranes were then prepared for solubilization and immunoprecipitation with 12CA5 as described under "Experimental Procedures." Quantitative analysis is from data pooled from three such experiments. B, 32 P-labeled transfected CHO cells were incubated for the indicated times at 37°C with 10 M NECA. Membranes were then prepared for solubilization and immunoprecipitation with 12CA5 as described under "Experimental Procedures." Quantitative analysis is from data pooled from three such experiments. strated that agonist pretreatment resulted in a 34 Ϯ 10% reduction in the B max for 125 I-AB-MECA binding (p Ͻ 0.05, four experiments) compared with untreated controls without significantly changing the K d (control, 2.04 Ϯ 0.38 nM, versus treated, 2.54 Ϯ 0.10 nM, four experiments) (Fig. 4A). This reduction in agonist binding sites was associated with a desensitization of HA-A 3 AR function, as manifested by an 8-fold increase in the IC 50 value for IB-MECA-mediated inhibition of forskolin-stimulated adenylyl cyclase activity (Fig. 4B, Table I). Therefore, exposure of transfected cells to agonist under conditions that promote receptor phosphorylation results in a loss of high affinity agonist binding sites and a functional desensitization.
In Vitro Phosphorylation of HA-A 3 AR by Recombinant GRK2-Two observations were consistent with the hypothesis that one or more GRKs was responsible for HA-A 3 AR phosphorylation in situ. Firstly, phosphorylation was dependent upon the presence of agonist but not activation of second messengerregulated kinases (Fig. 2C). Secondly, agonist treatment resulted in a reduction in high affinity agonist binding and a desensitization of HA-A 3 AR function (Fig. 4). Therefore, to directly assess the ability of HA-A 3 AR to act as a substrate for GRKs, in vitro phosphorylation experiments were performed using membranes from transfected cells and a purified recombinant GRK isoform. The effects of GRK2 on HA-A 3 AR phosphorylation were examined because this protein has been extensively studied, is ubiquitously expressed, and has been shown to phosphorylate G-protein-coupled receptors from sev-eral different receptor families (10). In the absence of any added GRK, membranes from transfected cells exhibited an endogenous agonist-dependent receptor kinase activity (Fig. 5). The addition of GRK2 increased the level of agonist-dependent phosphorylation by 2.4 Ϯ 0.3-fold over that observed in its absence (three experiments, Fig. 5). No such phosphoprotein was immunoprecipitated from nontransfected cells, demonstrating that the observed protein is the HA-A 3 AR (Fig. 5).
Effect of NECA Pretreatment in Situ on GRK2-stimulated Phosphorylation in Vitro-To determine whether the agoniststimulated phosphorylation of HA-A 3 AR was indeed due to a GRK, in vitro phosphorylation experiments were performed on membranes from cells that had been pretreated with NECA under conditions that produced optimal HA-A 3 AR phosphorylation in situ (10 M NECA, 10 min). Compared with membranes from cells treated without agonist, NECA pretreatment resulted in a 60% reduction in the ability of both endogenous GRK activity and exogenously added GRK2 to stimulate HA-A 3 AR phosphorylation in the presence of NECA in vitro (Fig. 6,  A and B). Therefore, it appears that NECA pretreatment of intact cells results in the phosphorylation of HA-A 3 AR in situ on at least some of the residues that are phosphorylated by GRK2 in vitro. Also, agonist pretreatment resulted in the appearance of an agonist-independent component of HA-A 3 AR phosphorylation, which was more pronounced in reactions containing GRK2 (Fig. 6, A and B). This phenomenon was not simply a reflection of residual agonist carried over from the cellular pretreatment, because it could be diminished in a time-dependent manner with replacement of the agonist-containing media and washing of cell monolayers prior to harvest (data not shown). Therefore, it may reflect an increased susceptibility to GRK-mediated phosphorylation induced by agonist pretreatment. The presence of agonist in vitro did not significantly enhance HA-A 3 AR phosphorylation any further in these membranes (Fig. 6, A and B). DISCUSSION Despite the growing appreciation of the contribution of A 3 AR activation toward mediating some of the physiological effects of adenosine, only limited information is available on how A 3 AR function is regulated. We have recently demonstrated that chronic exposure of A 3 AR-expressing CHO cells to the agonist NECA induces a functional desensitization that is associated with the specific down-regulation of G i ␣-3 and G-protein ␤-subunits (13). However, the time-course of G-protein down-regulation (t 1 ⁄2 ϭ ϳ6 h) would suggest that although this event may be an important adaptive mechanism to prolonged agonist exposure, it cannot account for the rapid functional desensitization observed for the native rat A 3 AR in response to acute agonist treatment (6, 7). By examining the effects of agonist  Table I.

TABLE I Agonist-induced desensitization of HA-A 3 AR-mediated inhibition of
forskolin-stimulated adenylyl cyclase activity HA-A 3 AR-expressing CHO cells were treated with or without 10 M NECA in the presence of 1 unit/ml adenosine deaminase for 10 min at 37°C. Incubations were terminated by placing the flasks on ice and washing the monolayers extensively in ice-cold buffer prior to membrane preparation. Assay of adenylyl cyclase activity was performed in the presence of 5 M forskolin and increasing concentrations of the A 3 AR agonist IB-MECA as described under "Experimental Procedures." The data are presented as the means Ϯ S.E. for three experiments. exposure in CHO cells expressing a recombinant epitopetagged A 3 AR, the current study now provides a testable working model with which to explain the phenomenon of rapid A 3 AR desensitization. The validity of the model system we have chosen is proven by two observations. Firstly, both native and recombinant epitope-tagged A 3 ARs undergo a rapid functional desensitization ( Fig. 4B and Refs. 6 and 7). Secondly, in membranes from transfected CHO cells (Fig. 4A) and RBL-2H3 cells, 2 this desensitization is associated with a 30 -40% reduction in the number of agonist binding sites recognized by 125 I-AB-MECA in radioligand binding assays. Therefore, it is likely that similar adaptive processes operate to diminish A 3 AR signaling in CHO and RBL-2H3 cells. Several lines of evidence support a role for G-protein-coupled receptor kinase involvement in A 3 AR desensitization. Firstly, the A 3 AR is rapidly phosphorylated in an agonist-dependent manner, and this cannot be mimicked by simple activation of second messenger-regulated kinases alone. Moreover, the EC 50 value for NECA-stimulated A 3 AR phosphorylation (ϳ0.15 M) is essentially the same as its K i value for displacing 125 I-AB-MECA binding from the A 3 AR (11). Such a correlation between extents of receptor occupancy and phosphorylation coupled with the lack of any effect of second messenger-regulated kinases strongly suggest that one or more GRK isoforms is responsible for A 3 AR phosphorylation, because these kinases specifically phosphorylate agonist-occupied receptors (9,10). A specific role for GRK2 or a related kinase is suggested from the in vitro phosphorylation experiments, which demonstrated that GRK2 was capable of enhancing the agonist-dependent A 3 AR phosphorylation observed in isolated membranes. Although it is theoretically possible that GRK2 mediates its stimulatory effect on A 3 AR phosphorylation in vitro indirectly via interaction with nonreceptor proteins present in the membrane preparation, this is unlikely considering the high degree of substrate specificity that GRKs exhibit toward G-protein-coupled receptors (17-19, 22, 23). Therefore, it is most likely that the agonist-occupied A 3 AR is directly phosphorylated by GRK2 in vitro. More importantly, pretreatment of transfected cells with agonist reduced the subsequent level of GRK2-stimulated, agonist-dependent A 3 AR phosphorylation observed in vitro.
The simplest explanation for this phenomenon would be that agonist pretreatment induces A 3 AR phosphorylation in situ on some of these residues by GRK2 or a very similar kinase such that they are not available for subsequent phosphorylation in vitro. However, interpretation of these results is complicated by the observation that agonist pretreatment in situ results in an agonist-independent component of endogenous kinase-and GRK2-mediated receptor phosphorylation in vitro. Because this effect is reversible in a time-dependent manner with agonist washout, it cannot be simply due to the carry-over of residual NECA from the cellular treatment. Because agonist pretreatment would result in A 3 AR activation and a resulting dissociation of G i -proteins in the proximity of the receptor, it is possible that localized release of ␤␥-subunits near the receptor may be sufficient to translocate sufficient GRK2 such that in the absence of agonist some receptor phosphorylation can occur. The ability of ␤␥-subunits alone to stimulate low level GRK2-mediated phosphorylation of antagonist-occupied ␤ 2 -adrenergic receptors in vitro has been previously described (24). Alternatively, it is possible that agonist-dependent phosphorylation of the receptor on specific residues in situ primes the receptor for subsequent agonist-independent phosphorylation in vitro. Such a sequential model of protein phosphorylation, whereby phosphorylation at specific residues is required in order to observe phosphorylation at other sites, has been well described for the kinases that control the enzymes involved in 2 T. M. Palmer, unpublished data.

FIG. 6. Effect of NECA pretreatment in situ on GRK2-stimulated phosphorylation in vitro.
A, transfected cells were treated with 1 unit/ml adenosine deaminase in the absence (Vehicle) or presence of 10 M NECA for 10 min at 37°C. Membranes were then prepared for in vitro phosphorylation assays with the indicated additions, as described under "Experimental Procedures" followed by immunoprecipitation with 12CA5 and analysis by SDS-PAGE and autoradiography. B is a quantitative analysis of three such experiments. Phosphorylation is expressed relative to that observed in membranes from control cells in the presence of NECA but in the absence of any added GRK2 (set at 100%).

FIG. 5. GRK-stimulated agonist-dependent HA-A 3 AR phosphorylation in vitro.
In vitro phosphorylations were performed as described under "Experimental Procedures" using membranes from nontransfected and transfected CHO cells incubated with or without 10 M NECA in the presence or absence of 50 nM GRK2 at 30°C for 5 min as indicated. Following the addition of stop solution, membranes were pelleted for solubilization and immunoprecipitation with 12CA5. Analysis was by SDS-PAGE and autoradiography. This is one of multiple experiments. glycogen metabolism (25). Moreover, recent experiments using synthetic peptide substrates (22), as well as a glutathione Stransferase fusion protein containing the COOH-terminal sequence of the fMet-Leu-Phe receptor (26), have indicated that GRK2 may utilize such a sequential phosphorylation mechanism.
Cellular pretreatment with agonist does not completely abolish subsequent A 3 AR phosphorylation in vitro. Therefore, it seems likely that the A 3 AR may contain multiple sites for GRK phosphorylation with only a subset of these being utilized in CHO cells in situ such that incubation with GRK2 in vitro results in the phosphorylation of some of the remaining sites. This is not an unexpected finding because studies of GRKmediated phosphorylation of rhodopsin and the ␤ 2 -adrenergic receptor have demonstrated that although high phosphorylation stoichiometries are reported in vitro, the stoichiometries exhibited by receptors in situ are much lower (27,28).
Although A 3 AR phosphorylation is associated with the onset of functional desensitization, we cannot directly determine whether the phosphorylated receptor is impaired in its ability to signal downstream. This will require the purification of the receptor and its reconstitution with inhibitory G-proteins in order to directly measure the receptor's signaling capacity in the absence of any other components. It is possible that additional factors, such as arrestins, are required to elicit a profound desensitization in response to GRK phosphorylation, as has been shown for rhodopsin and the human ␤ 2 -adrenergic receptor (29).
Unfortunately, we were unable to determine the stoichiometry of A 3 AR phosphorylation. This will be possible if either an A 3 AR antagonist radioligand is developed or purified preparations of HA-A 3 AR become available to allow us to accurately measure receptor content in 12CA5 immunoprecipitates. However, a knowledge of the A 3 AR sequence (3) and the substrate preference for GRK2 (22,23) can allow us to make some predictions as to where the phosphorylation sites for GRK2 may reside. Studies utilizing synthetic peptide substrates have demonstrated that GRK2 prefers to phosphorylate serine and threonine residues flanked on their NH 2 -terminal side by acidic amino acids and/or phosphoamino acids (22,23). Inspection of the A 3 AR sequence shows that the region of the COOHterminal domain immediately following the predicted palmitoylation site is particularly enriched in such residues (sequence reads as 305 CQTSDSLDSNLEQTTE). Hence this may be a useful domain within which to begin mutagenesis experiments to determine which residues are phosphorylated by receptor kinases in situ.
Perhaps more interestingly, the identification of GRK2 as an A 3 AR kinase in vitro, and possibly in the intact cell, as well as the association of receptor phosphorylation with desensitization, suggests that A 3 AR responsiveness in a given cell type may be controlled by overexpressing recombinant wild type and mutant forms of this kinase. The potential utility of such an approach has already been demonstrated for recombinant thrombin receptors co-expressed with GRK3 in Xenopus oocytes (30) and native ␤ 2 -adrenergic receptors in a human bronchial epithelial cell line after expression of a catalytically inactive GRK2 mutant that could function as a dominant negative ␤ 2 -adrenergic receptor kinase (31). Therefore it would be of interest to determine the effects of expressing wild type and mutant GRK2 on the functioning of the native A 3 AR in RBL-2H3 cells and, given the synergism exhibited between the re-sponses elicited via the A 3 AR and IgE receptor in these cells (6), whether expression would affect cellular responses to antigen.
In conclusion, we have demonstrated for the first time that the A 3 AR is rapidly phosphorylated in response to agonist exposure and that this event is associated with the onset of functional desensitization. Furthermore, in vitro phosphorylation studies strongly support a role for GRK2, or a kinase of similar substrate specificity, in mediating this phosphorylation, and these results suggest a potential strategy with which to manipulate A 3 AR signaling in the intact cell. Experiments to determine the usefulness of this approach and to identify the sites of A 3 AR phosphorylation are currently underway.