Phosphorylation and Desensitization of Human Endothelin A and B Receptors

Although endothelin-1 can elicit prolonged physiologic responses, accumulating evidence suggests that rapid desensitization affects the primary G protein-coupled receptors mediating these responses, the endothelin A and B receptors (ETA-R and ETB-R). The mechanisms by which this desensitization proceeds remain obscure, however. Because some intracellular domain sequences of the ETA-R and ETB-R differ substantially, we tested the possibility that these receptor subtypes might be differentially regulated by G protein-coupled receptor kinases (GRKs). Homologous, or receptor-specific, desensitization occurred within 4 min both in the ETA-R-expressing A10 cells and in 293 cells transfected with either the human ETA-R or ETB-R. In 293 cells, this desensitization corresponded temporally with agonist-induced phosphorylation of each receptor, assessed by receptor immunoprecipitation from 32Pi-labeled cells. Agonist-induced receptor phosphorylation was not substantially affected by PKC inhibition but was reduced 40% (p < 0.03) by GRK inhibition, effected by a dominant negative GRK2 mutant. Inhibition of agonist-induced phosphorylation abrogated agonist-induced ETA-R desensitization. Overexpression of GRK2, -5, or -6 in 293 cells augmented agonist-induced ET-R phosphorylation ∼2-fold (p < 0.02), but each kinase reduced receptor-promoted phosphoinositide hydrolysis differently. While GRK5 inhibited ET-R signaling by only ∼25%, GRK2 inhibited ET-R signaling by 80% (p < 0.01). Congruent with its superior efficacy in suppressing ET-R signaling, GRK2, but not GRK5, co-immunoprecipitated with the ET-Rs in an agonist-dependent manner. We conclude that both the ETA-R and ETB-R can be regulated indistinguishably by GRK-initiated desensitization. We propose that because of its affinity for ET-Rs demonstrated by co-immunoprecipitation, GRK2 is the most likely of the GRKs to initiate ET-R desensitization.

A variety of vital cardiovascular and developmental func-tions are mediated through the G protein 1 -coupled, heptahelical endothelin A and endothelin B receptors (ET A -R and ET B -R) (1,2). Like many G protein-coupled receptor signaling systems (3), the receptor/G q /phospholipase C system(s) activated by the ET A -R and ET B -R have demonstrated diminishing responsiveness to prolonged stimulation (4 -11). This desensitization may be particularly important in the regulation of ET-R-mediated signaling, since endothelins bind their receptors essentially irreversibly under physiological conditions (2). In several cell types or tissues, this desensitization has been characterized phenomenologically as homologous, or receptorspecific (5)(6)(7)(8)(9)(10)(11), but the molecular mechanisms effecting this desensitization remain to be elucidated.
For a growing number of G protein-coupled receptor signaling systems, phosphorylation of the receptor on serine and threonine residues appears to be an important mechanism of desensitization (3,12). This phosphorylation can be accomplished by at least two classes of protein kinases: (i) second messenger-dependent kinases, such as cAMP-dependent protein kinase or PKC isoforms, and (ii) the family of G proteincoupled receptor kinases (GRKs), which are activated by and phosphorylate agonist-occupied receptors (13,14). The GRK family comprises six cloned members, with highly homologous catalytic domains flanked by divergent N and C-terminal domains (14). In model receptor systems, GRK-mediated receptor phosphorylation appears to facilitate the binding of an inhibitory arrestin protein to the receptor, resulting in uncoupling of the receptor from its G protein in the process of homologous desensitization (12), within seconds to minutes of receptor activation (15,16). This GRK-initiated process has been shown to attenuate severely (17,18) or even to terminate (19) signaling through a variety of G protein-coupled receptors.
Like several receptors shown to be phosphorylated by GRKs (18,20,21), the human ET A -R and ET B -R each have cytoplasmic carboxyl-terminal tail domains that are rich in serine and threonine residues (22,23). Only for the ET B -R, however, does the primary structure surrounding some of these residues resemble the receptor sequences previously demonstrated to be phosphorylated by GRKs (20). Indeed, the relatively low homology between the ET A -R and ET B -R in the cytoplasmic tail domain suggests the possibility that these receptors might be regulated differently by GRKs, or regulated by different GRKs. To date, no data directly address this issue. However, experiments with cytoplasmic tail truncation mutants of the ET A -R (24) and the ET B -R (25) suggest that large portions of each receptor's cytoplasmic tail may be removed without impairing agonist-induced desensitization of endothelin-1-stimulated chloride currents (24) or calcium transients (25). To test directly whether or not homologous desensitization of the human ET A -R and ET B -R proceeds via a GRK-initiated mechanism, we used a transfected, intact cell model for assessing both receptor phosphorylation and signaling through phospholipase C. In this model system, we also overexpressed individual GRKs with the ET A -R and ET B -R to test the hypothesis that different GRKs might phosphorylate and initiate desensitization of the ET A -R and ET B -R. P]orthophosphate were procured from NEN Life Science Products. Biotinylated 12CA5 IgG and staurosporine were from Boehringer Mannheim, 12CA5 ascites was from BabCo, and the M2 monoclonal IgG 1 (both biotinylated and native) was from IBI. Horseradish peroxidase-conjugated streptavidin and antispecies IgGs were from Jackson ImmunoResearch. Protein A-Sepharose and protein G-Sepharose were obtained from Pharmacia Biotech Inc. The AG1-X8 anion exchange resin was procured from Bio-Rad. Peninsula Laboratories was the source of ET-1, and Sigma was the source of angiotensin II. The carboxyl-terminal amide of SFLLRN was synthesized on an Applied Biosystems 430A peptide synthesizer, using FastMoc™ chemistry, and the product was purified by high pressure liquid chromatography.
Plasmid Constructs-The cDNAs for the human ET A -R and ET B -R were the generous gifts of Masachi Yanagisawa (Department of Molecular Genetics, University of Texas Southwestern Medical Center, Dallas, TX) (23,26). Cassette PCR was employed to remove the endogenous signal sequence of each cDNA and to add to each cDNA an influenza virus hemagglutinin signal sequence (27) followed by either a hemagglutinin epitope recognized by the 12CA5 monoclonal antibody (21) (for the ET A -R) or the FLAG TM epitope (for the ET B -R) (27). The 5Ј (mutagenic) primers for the ET A -R and ET B -R, respectively, were 5Ј-cgcgggaagcttacc(atgaagaccatcatcgccctgagctacatcttctgcctggtgttcgcc)gacgcc[tacccctacgacgtccccgactacgcc]gataatcctgagagatacagc-3Ј and 5Ј-cgcgggaagcttacc(atgaagaccatcatcgccctgagctacatcttctgcctggtgttcgcc)[gactacaaggacgatgatgacgcc]gaggagagaggcttcccgcct-3Ј. Parentheses surround the signal sequence, brackets surround the epitope sequence, boldface type highlights the initiator methionine, and underlined type denotes nucleotides 541-561 (ET A -R) and 309 -329 (ET B -R) of the native receptor sequences. For the ET A -R, first the 1.1-kilobase pair EcoRV/Xmn I cDNA fragment and then the 292-base pair HindIII/EcoRV-cut PCR product were subcloned into pcDNA I (Invitrogen) to generate the epitope-tagged ET A -R (t-ET A -R). For the ET B -R, first the 1-kilobase pair EcoRV/XbaI cDNA/polylinker fragment (23) and then the 606-base pair HindIII/EcoRV-cut PCR product were subcloned into pcDNA I (Invitrogen) to generate the epitope-tagged ET B -R (t-ET B -R). The authenticity of the sequences amplified by PCR was verified by dideoxy sequencing.
The bovine GRK2 N-terminal minigene construct was also made by cassette PCR. The cDNA corresponding to amino acids 1-180 was amplified with a 5Ј-primer that also encoded the hemagglutinin epitope recognized by 12CA5; the resulting fragment was ligated into pCMV5 at BamHI and HindIII sites. Plasmid constructs for the 195-amino acid bovine GRK2 C-terminal domain polypeptide (28), human GRK6 (29), the hemagglutinin epitope-tagged AT 1A -R (18), bovine GRK2, the bovine GRK2 K220R dominant negative mutant, bovine GRK3, and bovine GRK5 (21) have been described previously.
Cell Culture and Transfection-Rat A10 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 20% fetal bovine serum as well as penicillin (100 units/ml) and streptomycin (100 g/ml) and incubated in 95% air, 5% CO 2 at 37°C in a humidified incubator. Human 293 cells were propagated and transfected by the calcium phosphate method as described (21).
For most t-ET A -R transfections, 10 g of receptor plasmid was used, along with 5 g of GRK, GRK2 polypeptide, or empty vector plasmid per 100-mm dish. For homologous desensitization experiments, 293 cells were transfected with 4 g of t-ET A -R and 10 g of AT 1A -R per 100-mm dish, to ensure that phosphoinositide hydrolysis mediated by the AT 1A -R exceeded that mediated by the t-ET A -R. For t-ET B -R transfections, 4 g of receptor plasmid was used, along with either 4 g of GRK, 2-6 g of empty vector, or 6 g of AT 1A -R plasmid (for a total of 10 g) per 100-mm dish. For AT 1A -R experiments, cells were transfected as described (18). The day after transfection, cells from all identically transfected plates were pooled and aliquoted, at the same density, for functional or receptor expression assay (21). With transfection on day 1, cells were assayed on day 3 or 4 (with equivalent results). Assays for Receptor Expression-t-ET A -R and t-ET B -R expression levels were assessed in pilot experiments by [ 125 I]ET-1 binding assays, modified from Sakamoto et al. (23). Assays were performed in triplicate, and data were analyzed by least squares nonlinear regression with Prism™ software (Graphpad). These assays demonstrated receptor expression levels of 1-3 pmol/mg of total cell protein for both receptor subtypes.
The relative receptor expression among cell lines co-transfected with various plasmids was routinely assessed by immunofluorescence and flow cytometry, performed as described previously (21). Typical transfection efficiencies ranged from 35 to 50% and from 45 to 70% for the t-ET A -R and t-ET B -R, respectively. To obtain a surrogate for receptor expression in each transfected cell line, the mean fluorescence/cell was multiplied by the percentage of cells staining positive for receptor. Cells co-transfected with various plasmids typically demonstrated receptor expression levels within 30% of that measured in empty vector-cotransfected control cells. Cells with receptor expression levels outside this range were not used.
Cellular Phosphoinositide Hydrolysis and Desensitization Assays-A10 and 293 cells were metabolically labeled for 18 -24 h with 2 Ci of [ 3 H]inositol/ml of labeling medium, which differed from the cell-specific growth medium only in that it contained 5% fetal bovine serum. A10 cells were plated in labeling medium at 5.3 ϫ 10 4 /cm 2 in 12-well dishes. After labeling, cells were washed with Dulbecco's phosphate-buffered saline and then exposed at 37°C to 0.2% bovine serum albumin and 20 mM LiCl in Dulbecco's phosphate-buffered saline ("PI medium"), without (basal) or with (stimulated) 100 nM ET-1 for the indicated times. Each time point comprised a single 12-well plate containing triplicate wells for both basal and stimulated cells. Reactions were stopped by the addition of an equal volume of 0.8 M perchloric acid to the wells, and total inositol phosphates were assayed by anion exchange chromatography as described (18). To facilitate comparisons between independent experiments, cellular inositol phosphates obtained from each well were normalized to the amount of intracellular 3 H counts in that well (determined by counting a 50-l aliquot of neutralized cell extract before chromatography). The resulting value, multiplied by 100, is referred to as "percent conversion of 3 H into inositol phosphates." Signaling assays with 293 cells (see Figs. 7-9) were performed in an analogous manner with the following modifications: cells were plated at 1.6 ϫ 10 5 /cm 2 in labeling medium; each 12-well plate comprised triplicate wells of each transfected cell type; and agonist stimulation transpired for 8 min. Homologous desensitization assays with 293 cells were performed with three 12-well plates manipulated in concert. During the "first period" of 3 min, cells were exposed at 37°C to PI medium without ("control plate") or with ("desensitized plates") the indicated concentrations of ET-1. PI medium was then removed from all wells. Next, during the "signaling period," cells from both control and desensitized plates were exposed for 5 min at 37°C to PI medium containing vehicle ("basal"), 100 nM ET-1, or 100 nM angiotensin II. Reactions were then terminated, as above. For desensitized cells, basal inositol phosphate values for the signaling period were obtained by terminating reactions on one of the desensitized plates at the outset of the signaling period.
Intact Cell Phosphorylation-These assays were performed as described previously (21). Briefly, cells metabolically labeled with 32 P i were stimulated with the indicated stimulus for 3 min (ET-1) or 10 min (TPA), washed, and solubilized in detergent buffer. Cells expressing the t-ET A -R were solubilized, and receptors were immunoprecipitated with 12CA5, as described (21). Cells expressing the t-ET B -R were solubilized in "M2 detergent buffer" (1% (w/v) Triton X-100, 0.05% SDS, 5 mM EDTA, 50 mM Tris-Cl, pH 8.0 (25°C), 200 mM NaCl, with phosphatase and protease inhibitors as described (21)). The t-ET B -Rs were immunoprecipitated as described (21), but with the M2 monoclonal IgG 1 and protein G-Sepharose. Immune complexes were dissociated and resolved on SDS-10% polyacrylamide gels, which were loaded with equivalent amounts of receptor protein per lane, as described (21). Autoradiogra-phy and Molecular Dynamics PhosphorImager analysis of radioactive gels were also performed as described (21).
Stoichiometry of Receptor Phosphorylation-Equilibrium biosynthetic labeling and immunoprecipitation were employed to determine the stoichiometry of ET B -R phosphorylation in intact 293 cells, as described previously (21).
Membrane GTPase Assays-293 cells plated at 1 ϫ 10 5 /cm 2 in 150-mm dishes were subsequently exposed to 100 nM ET-1 ("desensitized") or vehicle ("naive") at 37°C, in minimal essential medium containing 20 mM HEPES, pH 7.4, and 1 M okadaic acid (Calbiochem) (to inhibit possible phosphatase activity during the subsequent acid washes (30)). After 3 min, the cells were washed three times at 4°C in 50 mM glycine, pH 2.5, in a protocol shown by Chun et al. (31) to remove receptor-bound ET-1. After a further wash in Dulbecco's phosphatebuffered saline, the cells were scraped at 0°C into 10 mM Tris-Cl, 2 mM EDTA, pH 7.4 (25°C), supplemented with protease inhibitors (21). Cells were lysed with a motorized Teflon pestle, and crude membranes were pelleted at 30,000 ϫ g (4°C) for 15 min. Membranes were washed once at 4°C in 10 mM triethanolamine, pH 7.4 (30°C), supplemented with protease inhibitors (21), and finally resuspended at a protein concentration of 0.3 mg/ml in the same buffer. ET-1-stimulated hydrolysis of [␥-32 P]GTP was determined with triplicate 15-g aliquots of these membranes, as described (18). Reactions were stopped at 10 min, a time at which reaction rates were still linear.
Co-immunoprecipitation and Immunoblotting-293 cells were transfected with either an endothelin receptor or the AT 1A -R and the indicated plasmids (see Fig. 10); they were plated at 7-9 ϫ 10 4 /cm 2 in 90-mm dishes and assayed the next day. As in the inositol phosphate signaling assays, cells were washed and then exposed to PI medium for 8 min (37°C) in the absence (unstimulated) or presence (stimulated) of 100 nM ET-1 or 100 nM angiotensin II. (Albumin was omitted from the PI medium for these experiments.) PI medium was then replaced with 1 ml of cross-linking buffer (Dulbecco's phosphate-buffered saline containing 10 mM HEPES, pH 7.4 (to increase buffering capacity), 10% (v/v) Me 2 SO, and 2.5 mM of the cell-permeant homobifunctional cross-linking agent dithiobis(succinimidylpropionate) (Pierce). Cells were incubated at 25°C for 30 min, with continuous rocking. (In AT-R experiments, the cross-linking buffer on stimulated plates contained 100 nM angiotensin II.) Cross-linking buffer was then replaced with 1 ml/plate of the detergent buffer appropriate for the monoclonal IgG to be used for immunoprecipitating the receptor (see above), and receptor immunoprecipitation proceeded as described previously (18,21), but with 17 g of IgG/sample. Before the addition of IgG to the detergent-solubilized cells, aliquots were removed for protein assay and for subsequent immunoblotting to determine the approximate fraction of cellular GRK co-immunoprecipitated with the receptor. Dithiobis(succinimidylpropionate)-mediated cross-links were reduced, and immune complexes were dissociated by incubation in Laemmli buffer (32) at 37°C for 90 min. As described above for phosphorylation assays, equivalent amounts of receptor protein were resolved on SDS-10% polyacrylamide gels and then transferred to nitrocellulose by semidry blotting (32). Immunoblotting was performed with biotinylated monoclonal IgGs, either the GRK2/3-specific C5/7 or the GRK4/5/6-specific A16/17 (33), and horseradish peroxidase-conjugated streptavidin. Enhanced chemiluminescence (Amersham Corp.) was used to develop the immunoblots, which were exposed to Biomax MR film (Eastman Kodak Co.). Quantitation of band density was performed with a laser densitometer. To confirm that equivalent amounts of receptor protein had been loaded in each gel lane, immunoblots processed for detecting GRK2 or -5 were stripped (50°C for 3-4 h in 2% (w/v) SDS, 100 mM 2-mercaptoethanol, 62.5 mM Tris-Cl, pH 6.7 at 25°C) and reprobed with either biotinylated M2 IgG (for the t-ET B -R) or biotinylated 12CA5 (for the t-ET A -R and AT 1A -R).
Statistical Analysis-The t test for independent means was used to compare results for different cell types, and two-sided p values were calculated using Excel software (Microsoft), assuming equal sample variances (34). Because of considerable variability in membrane GTPase activity among experiments, paired t testing was used to compare naive with desensitized cells and to compare GRK2 K220R cells with control cells in the membrane GTPase experiments. Concentration/response curves were fitted by least squares nonlinear regression with variable slope, using Prism software (Graphpad). In the text, numerical values are given as means Ϯ S.D., whereas means Ϯ S.E. are plotted in all figures.

RESULTS
To demonstrate rapid desensitization of the ET A -R, we investigated the time course of ET-1-stimulated phosphoinositide hydrolysis in intact A10 cells, which endogenously express the ET A -R (31). Within 10 min, the A10 cell phospholipase C response to ET-1 ceased, consistent with ET A -R desensitization ( Fig. 1). When these ET-1-desensitized cells were challenged with thrombin agonist peptide, however, their phospholipase C response resumed. Thus, the ET A -R desensitization observed in A10 cells was homologous, or receptor-specific. Moreover, because the ET A -R-dependent phosphoinositide hydrolysis in A10 cells waned substantially by 6 min, this desensitization would seem to involve mechanisms other than receptor sequestration, which has been shown to follow a somewhat longer time course in these cells (31).
To examine the molecular mechanisms responsible for endothelin receptor desensitization and to facilitate direct comparison between the cloned human ET A -R and ET B -R, we employed transfected human 293 cells as a model system. To evaluate homologous desensitization in this system, we assayed ET-1-stimulated phosphoinositide hydrolysis in 293 cells co-transfected with the rat AT 1A -R and either the human t-ET A -R or t-ET B -R. A 3-min challenge with ET-1 reduced the subsequent inositol phosphate response to ET-1 in a dose-dependent manner for both the t-ET A -R and the t-ET B -R (Fig. 2). However, the initial challenge with ET-1 had no effect on the subsequent inositol phosphate response to angiotensin II. Thus, rapid homologous desensitization characterizes the signaling behavior of both the t-ET A -R and the t-ET B -R in this model 293 cell system.
Receptor sequestration attendant to the first challenge with ET-1 could not explain the desensitization findings in transfected 293 cells. Separate aliquots of cells challenged with 100 nM ET-1 and then processed for flow cytometry demonstrated that only 12 Ϯ 5% of the cell surface t-ET A -Rs and 9 Ϯ 1% of the t-ET B -Rs were internalized within 3 min. These small decrements in cell surface receptor number were modeled by transfecting cells with varying amounts of receptor plasmid, and had no effect on ET-1-stimulated phosphoinositide hydrolysis (data not shown).
To determine whether or not receptor phosphorylation was involved in the observed agonist-promoted desensitization of the t-ET A -R and t-ET B -R, we immunoprecipitated these receptors from 293 cells metabolically labeled with 32 P i . Receptors were isolated by SDS-polyacrylamide gel electrophoresis, and representative autoradiograms are presented in Figs. 3 and 4. The t-ET B -R (migrating at M r ϳ48,000 -54,000; Fig. 3) and the t-ET A -R (migrating at M r ϳ61,000 -79,000; Fig. 4) both underwent significant agonist-dependent phosphorylation in intact cells within 3 min. Thus, agonist-dependent phosphorylation correlated temporally with the agonist-dependent desensitization seen in Fig. 2. Attesting to the biochemical significance of this receptor phosphorylation, the stoichiometry of t-ET B -R phosphorylation was 1 Ϯ 0.2 mol of phosphate/mol of receptor (n ϭ 2).
Because endothelin receptor stimulation results in activa-tion of PKC isoforms, we sought to determine the extent to which PKC might participate in the rapid agonist-dependent phosphorylation of the endothelin receptors. In experiments presented in Fig. 3, 293 cells were incubated with a PKCselective concentration of staurosporine briefly and then challenged with either ET-1, to evoke agonist-induced receptor phosphorylation, or TPA, to activate PKC. Whereas staurosporine abolished the TPA-stimulated t-ET B -R phosphorylation by PKC, it had no significant effect on agonist-promoted t-ET B -R phosphorylation. Similar results were obtained with the ET A -R (data not shown). Thus, for both the t-ET A -R and the t-ET B -R, agonist-induced phosphorylation must be mediated by kinases other than PKC.
To test the possibility that agonist-promoted t-ET A -R phosphorylation is mediated by one or more of the GRKs endogenously expressed by 293 cells (21) of GRK2 cannot perform a phosphotransferase reaction (35); therefore, for any receptor substrate of GRK2, the K220R mutant can act as a competitive inhibitor of any GRK that could bind to and subsequently phosphorylate the receptor (18,21,(35)(36)(37). Characteristically, however, the GRK2 K220R mutant inhibits GRK action incompletely; even when present in 15-fold molar excess over GRK2, the K220R mutant inhibits only 60% of agonist-stimulated ␤ 2 -adrenergic receptor phosphorylation in vitro (35). Co-expression of the GRK2 K220R with the t-ET A -R in 293 cells resulted in a 40% reduction in agonistinduced t-ET A -R phosphorylation (Fig. 4). In contrast, GRK2 K220R expression had no effect on TPA-stimulated t-ET A -R phosphorylation (Fig. 4). Thus, a substantial fraction of the agonist-induced t-ET A -R phosphorylation appears to be mediated by one or more GRKs expressed endogenously in 293 cells.
The foregoing experiments demonstrate a temporal correlation between agonist-induced phosphorylation and desensitization of the t-ET A -R and t-ET B -R. To correlate the extent of GRK-mediated receptor phosphorylation with the extent of desensitization, we inhibited agonist-promoted ET A -R phosphorylation in 293 cells with dominant-negative GRK2 and assessed the effect of such inhibition on ET-1-stimulated GTPase activity in membranes derived from these cells (Fig.  5A). A 3-min exposure of cells to ET-1 reduced the subsequent ET-1-stimulated membrane GTPase activity by 40%, compared with that observed in membranes derived from control cells. With cells expressing dominant-negative GRK2, however, exposure to ET-1 did not affect the subsequent ET-1-stimulated membrane GTPase activity. Thus, abrogation of desensitization appears to attend the inhibition of GRK-mediated t-ET A -R phosphorylation by GRK2 K220R. In contrast to dominantnegative GRK2-transfected cells, native GRK2-transfected cells demonstrated desensitization indistinguishable from control cells (Fig. 5B), although GRK2 overexpression resulted in augmented agonist-induced ET A -R phosphorylation (see below; Fig. 6). In 293 cells, agonist-induced ET A -R phosphorylation mediated by endogenous GRK(s) therefore appears both necessary and sufficient for ET A -R desensitization.
Which of the anatomically disseminated (12) mammalian GRKs can effect agonist-induced phosphorylation of the endothelin receptors? To address this question, we overexpressed GRK2, -5, and -6 along with each endothelin receptor in 293 cells and studied the effects of GRK overexpression on agonistinduced receptor phosphorylation. Overexpression of any of the three GRKs tested significantly increased the level of agonistinduced receptor phosphorylation (Fig. 6). For the t-ET A -R, co-overexpression of either GRK2, GRK5, or GRK6 augmented agonist-induced phosphorylation 1.6 Ϯ 0.2-, 1.7 Ϯ 0.3-, or 2.0 Ϯ 0.6-fold, respectively, over that observed in control cells (p Ͻ 0.02) (Fig. 6A). For the t-ET B -R, co-overexpression of GRK2, -5, or -6 augmented agonist-induced phosphorylation 2.6 Ϯ 0.4-,  (GRK2 K220R, ϩ). Cells labeled with 32 P i were exposed to medium without (basal) or with (stimulated) 100 nM ET-1 or 4 M TPA for 3 or 10 min, respectively, at 37°C. Receptor immunoprecipitation and analysis proceeded as for Fig. 3. Pictured in the upper panel is an autoradiogram representative of three experiments performed. (Lane 7 (None) shows the immunoprecipitate from untransfected cells.) In the lower panel, these results are plotted (means Ϯ S.E.) in terms of stimulus-dependent receptor phosphorylation: counts in receptor bands from stimulated cells minus those from unstimulated cells. For each condition, stimulus-dependent receptor phosphorylation is normalized to that seen in ET-1-treated control cells: 100 ϫ ((stimulated Ϫ basal)/ ((control stimulated with ET-1) Ϫ (control basal))). *, p Ͻ 0.03 compared with control cells. A and B), the dominant negative GRK2 K220R mutant (K220R, A), or GRK2 (B). Cells were exposed, as in Fig. 4, to medium containing vehicle without (None, naive) or with (ET-1, desensitized) 100 nM ET-1 for 3 min at 37°C. Receptor-bound ET-1 was then removed with serial acid washes, and the cells were processed for membrane preparation and GTPase assay as described under "Experimental Procedures." Membrane GTPase reactions were conducted in the absence (basal) or presence (stimulated) of 100 nM ET-1. The percentage above basal was determined for each sample as follows: 100 ϫ ((stimulated/basal) Ϫ 1). Plotted are the means Ϯ S.E. from three experiments performed in triplicate, for control and K220R cells (A), and for control and GRK2 cells (B). Basal GTPase values (fmol/g of protein/10 min) for naive and desensitized cells, respectively, were 6.7 Ϯ 4.9 and 7.7 Ϯ 4.5 (control); 6.6 Ϯ 2.3 and 4.8 Ϯ 2.2 (K220R); and 26 Ϯ 16 and 26 Ϯ 22 (GRK2). *, p Ͻ 0.03 (paired t test) compared with samples from naive cells.

FIG. 5. ET A -R desensitization assessed by membrane GTPase activity: the effect of inhibiting agonist-induced ET A -R phosphorylation with GRK2 K220R. 293 cells were transfected with plasmids encoding the t-ET A -R and empty vector (Control,
2.1 Ϯ 0.8-, or 2.4 Ϯ 0.8-fold, respectively, over that observed in control cells (p Ͻ 0.02) (Fig. 6B). Under the conditions prevailing in these cellular assays, then, the t-ET A -R and t-ET B -R each appear susceptible to agonist-promoted phosphorylation mediated by each of the three widely expressed mammalian GRKs.
To correlate GRK-mediated receptor phosphorylation with effects on receptor signaling in the intact cell, we assayed ET-1-stimulated phosphoinositide hydrolysis in cells co-expressing the ET-Rs with GRK2 and GRK5 (Fig. 7, A and B). Cells overexpressing GRK2 with the t-ET A -R or t-ET B -R produce only 16 Ϯ 4% or 22 Ϯ 5%, respectively, of the maximal inositol phosphate response observed in control cells after an 8-min challenge with ET-1 (p Ͻ 0.002). By contrast, cells overexpressing GRK5 with the t-ET A -R or t-ET B -R produce 75 Ϯ 8% or 73 Ϯ 13%, respectively, of the maximal inositol phosphate response observed in control cells (p Ͻ 0.03) (Fig. 7, A and B). Despite these consistent and distinct effects on ET-R-promoted phosphoinositide hydrolysis, overexpression of GRK2 or GRK5 failed to affect phosphoinositide hydrolysis engendered by fluoroaluminate, which stimulates G proteins directly (38) (data not shown). Thus, overexpression of either GRK2 or GRK5 appears to diminish agonist-promoted phosphoinositide hydrolysis at the level of the endothelin receptor and not at the level of even the proximate signal-transducing protein.
To ascertain that the overexpressed GRKs were indeed inhibiting ET-1-promoted phosphoinositide hydrolysis at the level of the endothelin receptors, we tested the 293 cell GRK overexpression system with two additional G q -coupled receptors. First, we transfected 293 cells with plasmids encoding the   FIG. 7. Attenuation of G q -coupled receptor signaling by overexpressed GRKs in intact cells. 293 cells were transfected with expression plasmids encoding the t-ET A -R (top), t-ET B -R (middle), or AT 1A -R (bottom), along with plasmids encoding either no protein (Control), GRK2, or GRK5. Cells were processed for agonist-stimulated phosphoinositide hydrolysis as described under "Experimental Procedures." Cells were stimulated with the indicated concentration of ET-1 (ET A -R and ET B -R) or angiotensin II (AT 1A -R) for 8 min. The "signal above basal" was obtained by subtracting from each data point the inositol phosphate value obtained from unstimulated control cells (basal). Each value was then divided by the signal above basal for the control cells stimulated with 100 nM agonist, and multiplied by 100 to obtain the percentage of control maximum. Plotted are the means Ϯ S.E. from at least three separate experiments performed in triplicate. In control ET A -R, ET B -R, and AT 1A -R cells, respectively, the values for percent incorporation of 3  rat type IA angiotensin II receptor (AT 1A -R), along with plasmids encoding GRK2 or GRK5. Using an identical transfection scheme, we previously demonstrated that overexpression of either GRK2 or GRK5 augments agonist-induced AT 1A -R phosphorylation to the same degree (18). As can be seen in Fig. 7C, however, overexpression of these GRKs with the AT 1A -R produces distinct yet similar effects on agonist-promoted phosphoinositide hydrolysis. Interestingly, the signaling suppression achieved by GRK5 overexpression is substantially greater with the AT 1A -R (60%) than with either endothelin receptor (ϳ25%), despite equivalent levels of GRK5 expression attained in each distinct cell line (assessed by immunoblotting; data not shown). Contrastingly, the profound signaling suppression achieved by GRK2 overexpression is similar for the AT 1A -R, ET A -R, and ET B -R (Fig. 7).
To assess the receptor specificity of GRK-induced signal dampening further, we exploited the thrombin receptors expressed endogenously by 293 cells (Fig. 8). In phosphoinositide hydrolysis assays, cells like those in Fig. 7 were stimulated either with ET-1, to stimulate the transfected ET-Rs, or with the thrombin agonist peptide SFLLRN, to stimulate the 293 cell thrombin receptors. As before, overexpression of GRK2 suppressed ET-1-stimulated signaling profoundly, and overexpression of GRK5 suppressed ET-1-stimulated signaling modestly (Fig. 8). In these same cells, however, overexpression of either GRK2 or GRK5 suppressed thrombin receptor signaling to an equivalent, modest degree (Fig. 8). Thus, whether assessed with different G q -coupled receptors in distinct cell lines or with different G q -coupled receptors within the same cell, the signal-suppressing activity of overexpressed GRK5 appears receptor-specific.
Whereas overexpressing GRK2 or GRK5 achieved equivalent levels of augmented agonist-induced receptor phosphorylation, such overexpression achieved disparate levels of signaling suppression for the t-ET A -R, t-ET B -R, and AT 1A -R. Thus, incremental receptor phosphorylation achieved by GRK overexpression failed to correlate with signaling suppression achieved by GRK overexpression. This apparent paradox prompted us to test further the mechanism by which GRK overexpression suppresses G q -coupled receptor phosphoinositide signaling in 293 cells.
To determine whether or not receptor phosphorylation is required for receptor-specific signal dampening by overexpressed GRKs, we examined the effect of the dominant-negative GRK2 on receptor-mediated phosphoinositide hydrolysis in intact 293 cells. Surprisingly, the dominant-negative (K220R) GRK2 mutant suppressed receptor signaling as effectively as the wild type GRK2 for both ET-Rs (Fig. 9, and data not shown). Moreover, by transfecting graded amounts of GRK2 and GRK2 K220R plasmid DNA in parallel cell groups, we could find no significant disparity between the wild type and mutant GRK effect on signaling at any level of GRK protein expression (data not shown). Thus, GRK-mediated receptor phosphorylation appears unimportant to the suppression of receptor signaling observed in intact 293 cells overexpressing certain GRKs.
Because GRK2 had the most profound dampening effect on endothelin receptor signaling, we investigated the ability of the noncatalytic GRK2 polypeptide domains (14) to accomplish such signal dampening. Cellular expression of the 180-amino acid GRK2 N-terminal domain polypeptide diminished t-ET A -R-dependent phosphoinositide hydrolysis, by 27 Ϯ 6% (p Ͻ 0.02). Expression of the 195-amino acid C-terminal domain polypeptide (28), however, had no effect on signaling (Fig. 9).
Both the phosphorylation-incompetent GRK2 mutant (GRK2 K220R) and the GRK2 N-terminal domain polypeptide suppressed endothelin receptor signaling. Could a simple physical association of GRK with the activated receptor, by itself, engender the observed suppression of signaling? To test this hypothesis, we immunoprecipitated the t-ET A -R, t-ET B -R, or AT 1A -R from 293 cells co-transfected with either GRK2 or GRK5 and immunoblotted these immunoprecipitates to detect the presence of GRK co-immunoprecipitated with each receptor. Data obtained with the t-ET B -R and AT 1A -R are depicted in Fig. 10.
When immunoprecipitated from unstimulated cells, both the ET B -R and the AT 1A -R appeared to be associated with a small amount of either GRK2 or GRK5. For both receptors, there was a substantial increase in receptor-associated GRK2 attendant to activation by agonist (13 Ϯ 8-fold for the ET B -R; 29 Ϯ 3-fold for the AT 1A -R). Only for the AT 1A -R, however, did the amount of receptor-associated GRK5 increase substantially with agonist stimulation (8 Ϯ 1-fold). Agonist-dependent GRK5 association with the receptor occurred exclusively for the AT 1A -R, even in 293 cells co-transfected with the ET B -R, AT 1A -R, and GRK5 (data not shown). Agonist-dependent association between an overexpressed GRK and the receptor thus appears to correlate with GRK-mediated suppression of signaling, a process that appears to be specific at the level of both the G qcoupled receptor and the GRK. DISCUSSION Because of their exceptionally high affinity for their agonists, the ET A -R and ET B -R may remain stably associated with their agonists long after initial agonist binding (2,39). The resulting persistence of the receptors' activated states makes intracellular mechanisms of receptor/effector desensitization exceedingly important for cellular homeostasis. In this investigation, we have shown for the first time that rapid, agonist-promoted desensitization of both the ET A -R and ET B -R involves phosphorylation of the receptors, and that this phosphorylation results from GRK activity in an intact, transfected cell model. Surprisingly, despite the dissimilarity of their cytoplasmic C-terminal tail domains, the ET A -R and ET B -R appear to interact with GRKs indistinguishably.
That the ET A -R and ET B -R are regulated by one or more GRKs in vivo is anatomically plausible, since the ET-Rs (40) as well as GRKs 2, 5, and 6 (12) are all expressed together, for example, in human myocardium. Moreover, the paradigm for GRK-initiated desensitization of only agonist-occupied receptors (12) is consonant with the homologous ET A -R and ET B -R desensitization observed in a multitude of experimental systems (5)(6)(7)(8)(9)(10)(11). The rapid rate of receptor/effector desensitization observed in some of these systems (4,5,(7)(8)(9)11) is also consistent with a GRK-initiated mechanism (15,16). In this study, the use of dominant-negative GRK2 to inhibit both agonistinduced phosphorylation and desensitization of the ET A -R demonstrates that these events are indeed mediated by a GRK mechanism (18,21,35,36).
This study describes three distinct aspects of the relationship between GRKs and the ET-Rs: (i) the dependence of ET-R desensitization upon GRK-mediated ET-R phosphorylation, demonstrated in experiments that use dominant-negative GRK2 to inhibit both processes; (ii) the ability of individual GRKs to phosphorylate the ET-Rs, demonstrated in experiments that show augmented agonist-induced ET-R phosphorylation in cells transfected with specific GRK expression plasmids; and (iii) the affinity of individual GRKs for the ET-Rs, suggested by GRK-specific inhibition of ET-R-dependent signaling in intact cells and demonstrated by agonist-dependent association of GRK with receptor. As we discuss below, this last point seems most informative regarding GRK substrate specificity.
The incremental ET-R phosphorylation seen in cells overexpressing a particular GRK demonstrates the ability of that GRK to phosphorylate the receptor under the conditions studied. However, this incremental phosphorylation is probably unimportant with regard to desensitization of the receptor/G protein signaling system. This latter inference is supported by our membrane GTPase data with the ET A -R. Desensitization of ET-1-stimulated GTPase was equivalent in cells that overexpressed GRK2 and in cognate control cells that expressed only endogenous levels of GRK2. We obtained similar results with the AT 1A -R previously (18). In 293 cells expressing endogenous GRKs, both the ET-R and the AT 1A -R undergo GRK-mediated phosphorylation to a stoichiometry of ϳ1 mol of phosphate/mol of receptor. For these receptors in 293 cells, then, it appears that this stoichiometry of ϳ1 suffices for maximal binding of arrestins to the receptor and hence for maximal desensitization assessed by receptor signaling in membrane preparations.
The most unexpected finding of this study emerged from examining the ability of overexpressed GRKs to inhibit ET-Rmediated phosphoinositide hydrolysis in intact 293 cells. We anticipated that cellular overexpression of GRKs would diminish ET-R-dependent phosphoinositide hydrolysis in intact cells by augmenting both agonist-promoted ET-R phosphorylation and desensitization. Surprisingly, however, GRK-mediated suppression of ET-R signaling in intact cells failed to correlate with the incremental, GRK-mediated receptor phosphorylation observed. Rather, the ability of an overexpressed GRK to suppress receptor-mediated signaling correlated with the GRK's ability to bind to the receptor in an agonist-dependent manner, as revealed by co-immunoprecipitation experiments. Bound to the GRK, the agonist-occupied receptor appears sterically hindered from stimulating its G protein and thus from signaling. As assessed by the parameters of agonist-dependent receptor binding and signal dampening, both the ET A -R and ET B -R appear to interact selectively with GRK2, as opposed to GRK5 or GRK6. Contrastingly, the AT 1A -R appears relatively nonselective with regard to GRK2 and GRK5.
Although probably attributable to our level of GRK overex- The cells were exposed to vehicle, 100 nM ET-1, or 100 nM angiotensin II for 8 min at 37°C. Cellular proteins were then crosslinked, receptors were immunoprecipitated, and immunoblotting for GRK2 (A and B) or GRK5 (C and D) was carried out as described under "Experimental Procedures." The immunoblots are each representative of three performed. Arrows designate the positions of electrophoretic migration for the GRK5 doublet (C and D; from 293 cells overexpressing GRK5, M r 66,000 -67,000) and for purified GRK2 (A and B, M r 80,000). The GRK5 signal in C represents only ϳ0.25% of the GRK5 present in the cellular extract subjected to receptor immunoprecipitation, while the GRK signals in the ϩagonist lanes of A, B, and D represent Ն1% of the GRK present in the cellular extract subjected to receptor immunoprecipitation. (In these immunoprecipitations, Ͻ5% of the total cellular receptor complement is immunoprecipitated (data not shown). Thus, the percentage of cellular GRK co-immunoprecipitated with agonistbound receptor approximates the percentage of cellular receptor immunoprecipitated.) A and C derive from the same experiment and thus from cells with equivalent ET B -R expression. Reprobing these blots for receptor (see "Experimental Procedures") confirmed that equivalent amounts of receptor protein had been loaded in lanes 1-3 of each gel (A-D). Results equivalent to those depicted for the ET B -R were obtained with the ET A -R (data not shown).
pression, GRK-dependent dampening of receptor-mediated signaling nonetheless reveals a novel perspective on GRK substrate specificity. We showed that GRK-dependent suppression of signaling was receptor-specific by controlling for GRK expression in distinct cell lines expressing different receptors and by stimulating different receptors within the same GRK-overexpressing cell. Since this GRK-dependent signal dampening correlates with agonist-promoted binding of the GRK to the receptor, it must reflect the affinity of a given GRK for the activated form of a particular receptor. This receptor affinity is likely to be very important for GRK-mediated receptor phosphorylation in untransfected cells, which express physiologic (i.e. very low (41)) levels of both receptor and GRK. Applying this reasoning to our results with the ET-Rs, we would expect GRK2 to be the most active of the GRKs tested in desensitizing the ET-Rs in vivo.
Why does a GRK's affinity for the ET-Rs, manifested by co-immunoprecipitation, fail to correlate with the observed degree of agonist-induced phosphorylation in our 293 cell overexpression system? Two explanations seem relevant. First, ET-R phosphorylation is facilitated by the high expression levels of both receptor and GRK in the transfected 293 cells (20 -50-fold above those seen physiologically). By mass action, even GRKs with relatively low affinity for the agonist-occupied receptor may phosphorylate it in this system, although they may not do so when enzyme and substrate are present at physiological concentrations. Second, ET-R phosphorylation may be bridled by other intracellular proteins (particularly the arrestins), which may compete with even the high-affinity GRK(s) for receptor binding. In this regard, it is notable that phosphorylation of the receptor has been shown both to decrease the affinity of GRK binding (42) and to increase the affinity of arrestin isoform binding to the receptor (43).
GRK-mediated phosphorylation of the ET-Rs does not, of course, preclude a role for other kinases in the phenomenon of homologous ET-R desensitization. Since they are activated by the ET-Rs (2), PKC isoforms would seem likely candidates for regulating the ET-Rs. In support of this idea, both the ET A -R (five sites) (22) and the ET B -R (seven sites) (23) possess numerous serines and threonines that occur within consensus sequences for PKC-mediated phosphorylation (44), and we have shown that PKC can clearly phosphorylate the ET-Rs in our experimental system. Despite these data, PKC appears to play no role in the rapid, agonist-induced ET-R phosphorylation we have studied here. Indeed, second messenger-dependent kinases, probably because of the multitude of their potential substrates, have been shown to effect receptor phosphorylation or/and desensitization at rates considerably slower than that evinced for GRKs (16,18). In other cell systems or at later time points after agonist stimulation, however, it would not be surprising to observe a more prominent role for PKC isoforms in ET-R desensitization.
The site or sites of GRK phosphorylation on the ET-Rs remain speculative. Suggesting as yet unidentified diversity in GRK phosphorylation sequences, neither ET-R (22, 23) possesses serine or threonine residues residing in sequences conforming to those defined for the best characterized GRK substrates (20). By analogy with other receptors shown to be GRK substrates, the molecular architecture of the ET-Rs suggests that the ET-R cytoplasmic tail domains would most likely serve as sites for GRK phosphorylation (20,22,23,45). Previous mutational studies have contended, though, that the C-terminal cytoplasmic tail of neither the ET A -R (24) nor the ET B -R (25) is important for agonist-promoted receptor desensitization. Inferences drawn from these studies regarding ET-R desensitization and its potential mechanisms should be cautious, how-ever. "ET-R desensitization" in both of these studies was monitored by signals considerably downstream from the ET-Rs themselves (calcium-activated chloride currents in Xenopus oocytes and cytosolic calcium transients in Ltk Ϫ cells, respectively). Results from these downstream signaling mechanisms may be confounded by rapid desensitization of the inositol trisphosphate receptor (46) and perhaps the chloride channels, too. Furthermore, complete ET B -R desensitization appeared to be achieved with very low fractional receptor occupancy and in the setting of considerable heterologous desensitization (25). Such an observation conflicts strongly with the behavior of a GRKinitiated mechanism (47). Even if these mutational studies are correct, however, GRK phosphorylation sites on the ET-Rs may reside in the C tail remnants proximal to the truncation sites or in other cytoplasmic receptor domains (22,23).
In light of the rapid desensitization of the ET-R/G q /phospholipase C signaling system by GRK-initiated (and perhaps other) mechanisms, how can we explain the prolonged physiologic action of ET-1 administered pharmacologically (48)? Whereas ET-R-promoted phospholipase C activity desensitizes rapidly, ET-R-promoted phospholipase D activity (2) may desensitize much more slowly, as has been observed for the AT 1A -R (18,49). Furthermore, the attenuating effect of nitric oxide on ET-1-stimulated signaling (39), which may facilitate ET-1 signal termination in vivo, may be unphysiologically deficient in experimental systems demonstrating sustained ET-1-promoted activity (1,48).
In investigating the role of GRKs in the homologous desensitization of the ET A -R and ET B -R, we have employed a cotransfected cell technique used by our laboratory (21) and subsequently others (37, 50 -52) to show the effect of overexpressed GRKs on agonist-induced receptor phosphorylation and signaling. Our current study reveals how overexpressed GRKs, by binding to agonist-occupied receptors, can inhibit receptor signaling in a manner that may manifest GRK substrate specificity previously unappreciated. Using the approach employed with the ET-Rs and the AT 1A -R, future studies should substantially expand our understanding of GRK substrate specificity.