The stability of the G protein-coupled receptor-beta-arrestin interaction determines the mechanism and functional consequence of ERK activation.

By binding to agonist-activated G protein-coupled receptors (GPCRs), beta-arrestins mediate homologous receptor desensitization and endocytosis via clathrin-coated pits. Recent data suggest that beta-arrestins also contribute to GPCR signaling by acting as scaffolds for components of the ERK mitogen-activated protein kinase cascade. Because of these dual functions, we hypothesized that the stability of the receptor-beta-arrestin interaction might affect the mechanism and functional consequences of GPCR-stimulated ERK activation. In transfected COS-7 cells, we found that angiotensin AT1a and vasopressin V2 receptors, which form stable receptor-beta-arrestin complexes, activated a beta-arrestin-bound pool of ERK2 more efficiently than alpha 1b and beta2 adrenergic receptors, which form transient receptor-beta-arrestin complexes. We next studied chimeric receptors in which the pattern of beta-arrestin binding was reversed by exchanging the C-terminal tails of the beta2 and V2 receptors. The ability of the V2 beta 2 and beta 2V2 chimeras to activate beta-arrestin-bound ERK2 corresponded to the pattern of beta-arrestin binding, suggesting that the stability of the receptor-beta-arrestin complex determined the mechanism of ERK2 activation. Analysis of covalently cross-linked detergent lysates and cellular fractionation revealed that wild type V2 receptors generated a larger pool of cytosolic phospho-ERK1/2 and less nuclear phospho-ERK1/2 than the chimeric V2 beta 2 receptor, consistent with the cytosolic retention of beta-arrestin-bound ERK. In stably transfected HEK-293 cells, the V2 beta 2 receptor increased ERK1/2-mediated, Elk-1-driven transcription of a luciferase reporter to a greater extent than the wild type V2 receptor. Furthermore, the V2 beta 2, but not the V2 receptor, was capable of eliciting a mitogenic response. These data suggest that the C-terminal tail of a GPCR, by determining the stability of the receptor-beta-arrestin complex, controls the extent of beta-arrestin-bound ERK activation, and influences both the subcellular localization of activated ERK and the physiologic consequences of ERK activation.

Recent data from yeast two-hybrid screens and from biochemical characterization of receptor-␤-arrestin complexes have indicated that ␤-arrestins also have the ability to interact with proteins involved in signal transduction, suggesting that they may additionally function in the recruitment of signaling proteins to GPCRs (5)(6)(7). One potentially significant set of interactions is the binding of ␤-arrestins to components of the extracellular signal-regulated kinases 1 and 2 (ERK1/2) and c-Jun N-terminal kinase 3 (JNK3) mitogen-activated protein kinase cascades, which allows them to act as scaffolds for localized mitogen-activated protein kinase activation (8 -12). In KNRK cells, stimulation of the protease-activated receptor type 2 (PAR2) induces the assembly of multiprotein complexes that contain the internalized receptor, ␤-arrestin 1, Raf-1, and activated ERK1/2 (9). The formation of these complexes, which are sufficiently stable that they can be isolated by both gel filtration and immunoprecipitation, also affects the cellular distribution of activated ERK. Because ␤-arrestins are cytosolic proteins, the formation of stable complexes between ␤-arrestin and ERK leads to cytosolic retention of ERK activated by PAR2 receptors. Qualitatively similar results have been obtained for the AT1aR expressed in HEK-293 and COS-7 cells (11,12). AT1aR activation causes the formation of complexes containing the receptor, ␤-arrestin 2, and the component kinases of the ERK cascade: cRaf-1, MEK1, and ERK2. Upon receptor internalization, activated ERK2 appears in the same endosomal vesicles that contain AT1aR-␤-arrestin complexes.
Thus, the ␤-arrestins appear to play a dual role in GPCR signaling. They serve both to terminate G protein-dependent signals by precluding receptor-G protein coupling, and to confer novel signaling properties upon the receptor by acting as adapters or scaffolds for signaling proteins. Because of these divergent functions, we hypothesized that the stability of the receptor-␤-arrestin interaction might be a significant factor in determining both the mechanism of ERK activation employed by a GPCR and the functional consequences of ERK activation within the cell. To test this hypothesis, we have compared the relative efficiency with which GPCRs that form transient receptor-␤-arrestin complexes (␣1bAR and ␤2AR), and GPCRs that form stable receptor-␤-arrestin complexes (AT1aR and V2R), activate a ␤-arrestin-bound pool of ERK. We also compared the signaling properties of chimeric receptors, in which the C-terminal tail domains have been exchanged so as to alter the stability of the receptor-␤-arrestin interaction (V2␤2R and ␤2V2R). We tested whether these chimeric receptors exhibited altered activation of ␤-arrestin-bound ERK, and whether changing the receptor-␤-arrestin interaction altered the cellular distribution of activated ERK, the ability of ERK to activate transcription of an Elk-1 reporter, and the ability of the receptor to produce a mitogenic response. We find that the formation of stable receptor-␤-arrestin complexes is associated with enhanced activation of ␤-arrestin-bound ERK, cytosolic retention of active ERK, and a reduced transcriptional and mitogenic response to GPCR stimulation. Moreover, this phenotype can be reversed by exchange of the C-terminal tail domain. These data suggest that the stability of the GPCR-␤-arrestin interaction dictates the predominant mechanism of ERK activation and, thereby, the functional consequences of ERK activation following GPCR stimulation.

EXPERIMENTAL PROCEDURES
Materials-LipofectAMINE was from Invitrogen. FuGENE 6 was from Roche Molecular Diagnostics. Monoclonal M2 anti-FLAG affinity agarose was from Sigma. Monoclonal anti-hemagglutinin (HA) affinity agarose was from Covance. Anti-phospho-ERK1/2 antibody was from Cell Signaling and anti-ERK1/2 antibody was from Upstate Biotechnology. Goat polyclonal anti-lamin B antibody, rabbit polyclonal anti-actin, anti-retinoblastoma protein, anti-HA, and anti-FLAG antibodies were from Santa Cruz Biotechnology. Rhodamine-conjugated monoclonal anti-HA antiserum was from Molecular Probes. Horseradish peroxidaseconjugated donkey anti-rabbit antibody was from Amersham Biosciences. Horseradish peroxidase-conjugated rabbit anti-goat antibody was from Sigma. Rabbit polyclonal anti-␤-arrestin was prepared in the Lefkowitz laboratory.
Cell Culture and Transfection-HEK-293 cells and COS-7 cells were obtained from the American Type Culture Collection. HEK-293 cells stably expressing 1.5-2 pmol/mg of the HA-V2R or HA-V2␤2R were generated by standard procedures using zeocin selection (0.4 mg/ml). HEK-293 cells were grown in Eagle's minimal essential medium with Earle's salt supplemented with 10% fetal bovine serum and 100 g/ml gentamicin. COS-7 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 100 g/ml gentamicin. Transient transfection of HEK-293 cells was performed using FuGENE 6 according to the manufacturers instructions. Transient transfection of COS-7 cells was performed using LipofectAMINE as previously described (12). Transfected cells were incubated overnight in serum-free growth medium supplemented with 10 mM HEPES, pH 7.4, 0.1% bovine serum albumin, and 100 g/ml gentamicin prior to stimulation.
Confocal Microscopy-HEK-293 cells were used for confocal microscopy because of their favorable morphology for comparing the cytoplasmic and plasma membrane distribution of proteins. Subconfluent monolayers in 100-mm dishes were transfected with expression plasmids encoding HA-AT1aR, HA-V2R, HA-␣1bAR, HA-␤2AR, HA-V2␤2R, or HA-␤2V2R (8 g/plate) along with GFP-␤-arrestin 2 (2 g/plate). Twenty-four hours after transfection, cells were passed onto collagen-coated 35-mm glass bottom dishes and serum-starved overnight. For visualization of HA epitope-tagged receptors, cell surface receptors were stained using a 1:500 dilution of rhodamine-conjugated monoclonal anti-HA IgG in serum starving medium for 1 h at 37°C. Cells were then washed with serum starving medium, stimulated as described in figure legends, washed with phosphate-buffered saline, fixed with 10% paraformaldehyde for 30 min at room temperature, and again washed prior to examination. Confocal microscopy was performed using a Zeiss LSM510 laser scanning microscope using a Zeiss 63 ϫ 1.4 numerical aperture water immersion lens with dual line switching excitation (488 nm for GFP, 568 nm for rhodamine) and emission (515-540 nm for GFP, 590 -610 nm for rhodamine) filter sets.
Immunoprecipitation and Immunoblotting-Immunoprecipitation of FLAG epitope-tagged ␤-arrestin 2 was performed following transient transfection of COS-7 cells in 100-mm dishes. Cells were transfected with expression plasmids encoding HA-AT1aR, HA-V2R, HA-␣1bAR, HA-␤2AR, HA-V2␤2AR, or HA-␤2V2AR (2 g/plate) and GFP-ERK2 (1 g/plate) with or without plasmid encoding FLAG-␤-arrestin 2 (3 g/ plate), as indicated. Stimulation was performed as described in the figure legends. After stimulation, monolayers were washed with phosphate-buffered saline, solubilized in 0.8 ml of glycerol lysis buffer (50 mM Hepes, 50 mM NaCl, 10% (v/v) glycerol, 0.5% (v/v) Nonidet P-40, 2 mM EDTA, 100 M Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride, 4 g/ml leupeptin, 2.5 g/ml aprotinin), and clarified by centrifugation. For determination of protein expression and total cellular phospho-ERK1/2, a 50-l aliquot of each clarified whole cell lysate was removed and mixed with an equal volume of 2ϫ Laemmli sample buffer. Approximately 50 g of protein from each lysate was resolved by SDS-PAGE and transferred to polyvinylidene difluoride membrane for immunoblotting. To isolate ␤-arrestin-bound GFP-ERK2, bovine serum albumin was added to each lysate to a final concentration of 1%, and immunoprecipitation was performed using 20 l of 50% slurry of monoclonal M2 anti-FLAG affinity agarose, with constant agitation overnight at 4°C. Immune complexes were washed three times with glycerol lysis buffer and boiled in Laemmli sample buffer. Immunoprecipitated proteins were resolved by SDS-PAGE and transferred to polyvinylidene difluoride membrane for immunoblotting.
For experiments involving the visualization of covalently crosslinked receptor-␤-arrestin-ERK complexes, COS-7 cells in 100-mm dishes were transfected with expression plasmids encoding HA-V2R or HA-␤2V2AR (6 g/plate). Stimulations were performed at 37°C in 4.6 ml of Dulbecco's phosphate-buffered saline. Incubations were terminated by the addition of 0.4 ml of 25 mM dithiobis(succinimidylpropionate) in dimethyl sulfoxide. Monolayers were agitated gently for 30 min at room temperature, washed with Dulbecco's phosphate-buffered saline containing 50 mM Tris-HCl, pH 7.4, to neutralize unreacted dithiobis(succinimidyl propionate), and lysed in 0.5 ml of glycerol lysis buffer. 25-l aliquots of clarified cross-linked detergent lysates were mixed with an equal volume of 2ϫ Laemmli sample buffer with or without ␤-mercaptoethanol, for electrophoresis under reducing or nonreducing conditions, respectively. The remainder of each lysate was agitated overnight at 4°C with 20 l of 50% slurry of monoclonal anti-HA affinity agarose to immunoprecipitate HA-epitope-tagged receptor and any cross-linked proteins. After washing, immunoprecipitates were resolved by SDS-PAGE under reducing conditions and subjected to immunoblotting.
Quantitation of Nuclear Phospho-ERK1/2-For the separation of nuclear and extranuclear pools of endogenous ERK1/2, COS-7 cells on 100-mm plates were transfected with plasmids encoding either the HA-V2R or HA-V2␤2R (2 g/plate) and FLAG-␤-arrestin 2 (0.5 g/ plate). Serum-starved cells were stimulated with vasopressin (1 M) for 5 min. Monolayers were washed twice with ice-cold phosphate-buffered saline and collected in 2 ml of hypotonic lysis buffer (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl, 0.3% (v/v) Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 10 g/ml aprotinin). Cells were incubated on ice for 3 min to allow lysis, and an aliquot of the lysed cell suspension was removed prior to centrifugation for measurement of total phospho-ERK1/2. Lysates were then centrifugation at 500 ϫ g for 5 min to pellet nuclei. The supernatants, containing a mixture of plasma membrane, microsomal vesicles, cytoskeleton, and cytosol, represented the extranuclear fraction. Pellets containing cell nuclei were washed in lysis buffer without Nonidet P-40 and again pelleted at 500 ϫ g. Both the extranuclear and nuclear fractions were solubilized in 2ϫ Laemmli sample buffer and phospho-ERK1/2 was determined by protein immunoblotting. The purity of each fraction was verified by immunoblotting with antibodies to nuclear lamin B, retinoblastoma protein, and actin as described previously (12).
Elk-1 Luciferase Reporter Assays-ERK1/2-dependent transcription was measured using an Elk-1 driven luciferase reporter system as described previously (12). Briefly, HEK-293 cells stably expressing either HA-V2R or HA-V2␤2R receptors were plated in 100-mm dishes and transfected with GAL4-Elk-1 (1 ng/plate), pFR-luc (1 g/plate), and pRL-tk-luc (20 ng/plate). The GAL4-Elk-1 plasmid encodes a fusion protein containing the GAL4 DNA binding domain and the transactivation domain of Elk-1, pFR-luc encodes the firefly luciferase gene under the control of the GAL4 DNA binding element, and pRL-tk-luc encodes Renilla luciferase under the control of the thymidine kinase promoter. One day following transfection, the cells were passed into 6-well dishes and serum-starved overnight. Stimulation with vasopressin (1 M) was carried out for 6 h. Luciferase activities were determined using a dual luciferase assay kit (Promega). Cells were extracted and assayed sequentially for firefly and Renilla luciferase activities. 5-l aliquots of cell lysate were incubated with 50 l of luciferin reagent and luminescence was recorded for 5 s. Stop and Glo TM reagent (50 l) was added and the specific luminescence from Renilla luciferase was recorded for an additional 5 s. Firefly activities were normalized to Renilla luciferase activity.

The Stability of the GPCR-␤-Arrestin Complex Determines the Extent of Activation of a ␤-Arrestin-bound Pool of ERK-
Oakley et al. (3) have shown that most GPCRs exhibit one of two patterns of receptor-␤-arrestin interaction. Fig. 1 illustrates each of these patterns, using anti-HA rhodamine staining and GFP-␤-arrestin 2 to track the cellular localization of an HA epitope-tagged receptor and ␤-arrestin, respectively. As shown for the HA-␤2AR in Fig. 1A, in the absence of agonist, rhodamine-stained receptors (left panel, red) were confined to the plasma membrane, whereas GFP-␤-arrestin 2 (middle panel, green) was diffusely cytosolic in distribution. An indistinguishable pattern of receptor and ␤-arrestin distribution was observed in unstimulated cells expressing the ␣1bAR, AT1aR, and V2R (data not shown). As shown in Fig. 1B, 15 min of agonist treatment led to the coalescence of both the ␤2AR and ␣1bAR (left panels, red) and GFP-␤-arrestin 2 (middle panels, green) in small puncta along the plasma membrane. This redistribution reflects the recruitment of ␤-arrestin to agonist-occupied GPCR and the accumulation of GPCR-␤-arrestin complexes in clathrin-coated pits or nascent endosomes. Internalized receptors, represented by the intracellular accumulation of HA-rhodamine, did not colocalize with GFP-␤arrestin 2 (right panels), suggesting that the receptor-␤-arrestin complex dissociates at or near the plasma membrane following receptor internalization.
As shown in Fig. 1C, a markedly different pattern was observed for the AT1aR and V2R. With these receptors, 15 min of agonist exposure led to removal of most of the HA-rhodaminelabeled receptors from the plasma membrane and their accumulation in large endosomal vesicles (left panels, red). An overlapping pattern of redistribution of GFP-␤-arrestin 2 was observed (middle panels, green; right panels, yellow), indicative of the formation of stable receptor-␤-arrestin complexes that remain associated throughout receptor endocytosis and sorting.
In overexpression studies, we have previously demonstrated that ␤-arrestins coprecipitate with cRaf-1, MEK1, and ERK2, and that overexpression of cRaf-1 or stimulation of AT1aRs increases the phosphorylation of a ␤-arrestin-bound pool of ERK2 (10,11), suggesting that ␤-arrestins can function as scaffolds for the component kinases of the ERK1/2 cascade. We have also found that overexpression of ␤-arrestin promotes the targeting of endogenous phospho-ERK1/2, along with the AT1aR, to endosomal vesicles and away from the nucleus, resulting in a smaller nuclear pool of activated ERK1/2 and a reduction in the Elk-1-driven transcriptional response to angiotensin II stimulation (12). Similar results have been reported for the PAR2 (9). However, both the AT1aR and PAR2 form stable receptor-␤-arrestin complexes that remain associated following internalization of the receptor. Because GPCRs differ in their ability to form stable complexes with ␤-arrestins, we hypothesized that they might also differ in the extent to which they activated ␤-arrestin-bound ERK.
To determine whether the stability of the receptor-␤-arrestin interaction affects the ability of GPCRs to activate a ␤-arrestinbound pool of ERK, COS-7 cells were transfected with HA epitope-tagged receptor, FLAG epitope-tagged ␤-arrestin 2, and GFP-tagged ERK2, and the phosphorylation state of GFP-ERK2 in FLAG-␤-arrestin 2 immunoprecipitates was determined before and after agonist stimulation. GFP-ERK2 was employed in these studies to facilitate quantitation of ERK phosphorylation in the transfected cell pool, because its slower electrophoretic mobility allows it to be easily resolved from endogenous ERK1/2. GFP-ERK2 has previously been shown to undergo agonist-stimulated phosphorylation and nuclear translocation in a manner analogous to endogenous ERK (9).
As shown in Fig. 2A, 5 min stimulation of AT1aR, V2R, and ␣1bAR expressed in COS-7 cells produced equivalent levels of whole cell GFP-ERK2 phosphorylation (top immunoblot). Under these conditions, FLAG-␤-arrestin 2 and GFP-ERK2 were constitutively associated, as shown by their coprecipitation in both the presence and absence of agonist treatment (second and third immunoblots). The ␤-arrestin-bound pool of GFP-ERK2 was minimally phosphorylated in the absence of agonist. In response to stimulation of AT1aR and V2R, phosphorylation of ␤-arrestin-bound GFP-ERK2 increased substantially. In contrast, stimulation of ␣1bAR had little effect on the phosphorylation of ␤-arrestin-bound GFP-ERK2, despite a robust increase in the level of phospho-GFP-ERK2 in the whole cell lysate (bottom immunoblot). Fig. 2B, provides a quantitative assessment of the relative extent of phosphorylation of ␤-arrestinbound GFP-ERK2 following stimulation of each of the three receptors. While no significant difference was observed between the AT1aR and V2R, the ␣1bAR was 4 -5-fold less efficient at activating the FLAG-␤-arrestin-bound pool of GFP-ERK2.
The AT1aR and ␣1bAR couple primarily to G q/11 family heterotrimeric G proteins to stimulate phosphatidylinositol hydrolysis (13), whereas the V2R couples primarily to the G sadenylyl cyclase pathway (4). As shown in Fig. 1, the AT1aR and V2R form stable receptor-␤-arrestin complexes, whereas the ␣1bAR interacts with ␤-arrestin transiently. Thus, the Phosphorylation of GFP-ERK2 in whole cell detergent lysates was determined by immunoblotting using phospho-ERK1/2-specific IgG as described (upper immunoblot). Anti-FLAG immunoprecipitates containing FLAG-␤-arrestin 2 were probed for FLAG-␤-arrestin 2 (second immunoblot), coprecipitated GFP-ERK2 (third immunoblot), and phosphorylated GFP-ERK2 (lower immunoblot). B, bar graph depicting the amount of phosphorylated GFP-ERK2 present in FLAG-␤-arrestin 2 immunoprecipitates following agonist treatment. In each experiment, the amount phospho-GFP-ERK2 in the FLAG-␤-arrestin 2 immunoprecipitate was normalized to the amount of GFP-ERK2 present. Data are expressed as a percentage of the ␤-arrestin 2-associated phospho-ERK signal observed in angiotensin II-stimulated cells. Data shown represent the mean Ϯ S.E. from four separate experiments. *, less than AT1aR, p Ͻ 0.05. ability of these receptors to activate ␤-arrestin-bound ERK2 correlated with their ability to form stable receptor-␤-arrestin complexes, rather than with the activation of a specific G protein pool.
The GPCR C-terminal Tail Regulates the Activation of ␤-Arrestin-bound ERK-The binding of ␤-arrestin to an agonistoccupied GPCR is dependent upon GRK-mediated phosphorylation of serine and threonine residues located within the C-terminal tail of the receptor (1,2). Previous studies have demonstrated that the structure of the C-terminal tail is also the principal determinant of the stability of the receptor-␤arrestin interaction (3,4,14). As shown in Fig. 3A, a chimeric GPCR composed of the V2R, a stable ␤-arrestin binder, substituted with the C terminus of the ␤2AR, a transient ␤-arrestin binder (V2␤2R), exhibits transient ␤-arrestin binding like the ␤2AR. Instead of accumulating in endosomes along with the receptor, GFP-␤-arrestin 2 colocalizes with the V2␤2R only at the plasma membrane. As shown in Fig. 3B, the converse is true of a chimeric ␤2AR receptor containing the V2R C terminus (␤2V2R), where the characteristic ␤2AR pattern of receptor-␤-arrestin complex formation only at the plasma membrane is converted into one of colocalized receptor and ␤-arrestin in endosomes.
If the ability of a GPCR to activate ␤-arrestin-bound ERK correlates with the formation of stable receptor-␤-arrestin complexes, then one might expect that exchanging the C terminus of receptors that differ in their pattern of ␤-arrestin binding would reverse the pattern of ERK activation. To test this hypothesis, we compared the ability of the V2R and ␤2AR to activate ␤-arrestin-bound GFP-ERK2, with that of the chimeric V2␤2R and ␤2V2R. As shown in Fig. 4A, 5 min stimulation of COS-7 cells expressing V2R and V2␤2R produced equivalent activation of GFP-ERK2 measured in the whole cell lysate. However, when the phosphorylation state of ␤-arrestin-bound GFP-ERK2 was compared, the V2R and V2␤2R differed significantly. As shown quantitatively in Fig. 4B, the V2R was about three times more effective than the V2␤2R at mediating the activation of ␤-arrestin-bound GFP-ERK. Fig. 4C depicts the results of an analogous experiment performed with the ␤2AR and ␤2V2R. Unlike the V2R and V2␤2R, the wild type ␤2AR was a relatively poor activator of ERK2 in COS-7 cells. Substitution of the V2R tail, which causes the ␤2V2R chimera to bind ␤-arrestin 2 tightly, produced a receptor capable of eliciting a substantially greater increase in GFP-ERK2 phosphorylation, even when ERK phosphorylation was measured in the whole cell lysate. Associated with this increase in overall GFP-ERK2 phosphorylation, we observed a 2-fold increase in isoproterenol-stimulated phospho-ERK2 bound to ␤-arrestin for the ␤2V2R. This result, shown quantitatively in Fig. 4D, suggests that the increase in whole cell ERK phosphorylation seen with the V2␤2R reflects increased utilization of a ␤-arrestin scaffold by the chimeric receptor.

␤-Arrestin Binding Affects the Cellular Distribution of Endogenous Phospho-ERK1/2 following Stimulation of the Wild
Type V2R and Chimeric V2␤2R-One reported consequence of the formation of ␤-arrestin-ERK complexes is the cytosolic retention of ␤-arrestin-bound ERK (9,11,12). Whereas the preceding data suggest that some fraction of the ERK activated in response to GPCR stimulation is associated with ␤-arrestin, and that the stability of the GPCR-␤-arrestin complex affects the efficiency with which ␤-arrestin-bound ERK is activated, they do not address whether the ␤-arrestin-bound ERK pool represents a large enough fraction of the total cellular pool of GPCR-activated ERK to affect its overall cellular distribution.
To address this question, we first analyzed the distribution of endogenous phospho-ERK1/2 in whole cell lysates and receptor immunoprecipitates following covalent cross-linking with the membrane-permeable reversible cross-linker, dithiobis(succinimidyl propionate). COS-7 cells transiently expressing HAtagged V2 or V2␤2 receptors were stimulated for varying times prior to cross-linking and detergent lysis. As shown in Fig. 5A, when these lysates were resolved by SDS-PAGE under nonreducing conditions, which preserve the covalent cross-links, agonist stimulation resulted in the appearance of a heterogenous population of high apparent molecular weight bands that immunoblotted for HA-tagged receptor, ␤-arrestin, ERK1/2, and phospho-ERK1/2. When the same samples were resolved under reducing conditions, which break the cross-links, these high molecular weight complexes were absent. As shown in Fig. 5B, when receptor immunoprecipitates from cross-linked lysates were resolved under reducing conditions, we observed agonistdependent coprecipitation of endogenous ␤-arrestin and phospho-ERK with the epitope-tagged receptor. Thus, the migration of receptor, ␤-arrestin, and ERK as high molecular weight species in cross-linked whole cell detergent lysates reflected, as least in part, the agonist-induced formation of receptor-␤-arrestin-ERK complexes. To directly compare the effect of ␤-arrestin binding on the distribution of V2R-activated endogenous ERK1/2, we measured the fraction of the total phospho-ERK1/2 in cross-linked whole cell lysates that was present in complexes with apparent molecular weight Ͼ100,000 following stimulation of V2R and V2␤2R. As shown in Fig. 5C (upper immunoblot and bar graph), ϳ75% of the phospho-ERK1/2 generated by the V2R was present in these complexes, compared with ϳ35% of that generated by the V2␤2R. Conversely, about 20% of the phospho-ERK1/2 migrated at 42-44 kDa when the V2R was stimulated, compared with 55% for the V2␤2R. In these experiments, the total amount of phospho-ERK1/2 generated by the two receptors was indistinguishable, as demonstrated when the cross-linked lysates were electrophoresed under reducing conditions (lower immunoblot). As the V2R and V2␤2R differ only in the stability of ␤-arrestin binding (3), these data suggest that stable ␤-arrestin binding is associated with the enhanced formation of high molecular weight complexes containing the receptor, ␤-arrestin, and ERK, and that these complexes contain a significant fraction of the phospho-ERK generated by V2R.
To further examine the effect of ␤-arrestin binding on the cellular distribution of active ERK1/2, we compared the fraction of total cellular phospho-ERK1/2 that translocated to the nucleus following stimulation of the V2R and V2␤2R. In these experiments, COS-7 cells transiently expressing HA-tagged V2 or V2␤2 receptors were stimulated for 5 min prior to isolation of cell nuclei by differential centrifugation. As shown in Fig. 6A, we found that with the V2R less than 10% of the total cellular phospho-ERK1/2 was present within the nucleus 5 min after stimulation, compared with about 25% with the V2␤2 chimera. Conversely, as shown in Fig. 6B, the amount of phospho-ERK1/2 present in the extranuclear fraction, representing plasma membrane, microsomes, cytoskeleton, and cytosol, was proportionally greater for the V2R than the V2␤2R. At this time point, the total amount of phospho-ERK1/2 in the V2 and V2␤2 receptor-expressing cells was indistinguishable. Thus, the data obtained by measuring the fraction of active ERK1/2 in high molecular weight complexes (Fig. 5) and those obtained by measuring the nuclear fraction of phospho-ERK1/2 are complementary, and suggest that stimulation of the wild type V2R generates more phospho-ERK1/2 in large multiprotein complexes and less phospho-ERK1/2 in the nucleus than stimulation of the V2␤2 receptor.
These data are also consistent with comparisons of the fraction of ␤-arrestin-associated phospho-ERK1/2 formed in response to stimulation of the wild type PAR2 receptor and a mutant receptor lacking GRK phosphorylation sites that does not bind ␤-arrestin (9). Using a gel filtration approach, these authors found that greater than 80% of the phospho-ERK formed in response to wild type receptor activation coeluted with the receptor, Raf-1, and ␤-arrestin, whereas less than 2% of the phospho-ERK coeluted with the mutant receptor. These authors also found that the wild type receptor activated a predominantly non-nuclear pool of ERK, whereas the mutant predominantly activated nuclear phospho-ERK. The presence of a nuclear export signal in ␤-arrestin 2, which was recently shown to account for the nuclear exclusion of ␤-arrestin 2-bound JNK3 in cells overexpressing ␤-arrestin 2 FIG. 5. Analysis of the distribution of endogenous ␤-arrestin, ERK1/2, and phospho-ERK1/2 in covalently cross-linked cell lysates following stimulation of wild type V2R and chimeric V2␤2R. COS-7 cells expressing either HA-V2R or HA-V2␤2R were stimulated with vasopressin (1 M) for the indicated times prior to covalent cross-linking with dithiobis(succinimidyl propionate) and preparation of detergent lysates. A, cross-linked detergent lysates from stimulated V2R-expressing cells were resolved by SDS-PAGE under reducing (␤-ME, mercaptoethanol) or nonreducing conditions. Identical immunoblots were probed for HA-epitope (left panel), ␤-arrestin (second panel), ERK1/2 (third panel), and phospho-ERK1/2 (right panel) as described. The representative immunoblots shown are from one of six separate experiments. B, HA-V2R was immunoprecipitated from cross-linked detergent lysates and resolved by SDS-PAGE under reducing conditions. Identical immunoblots were probed for HA epitope (left panel), and coprecipitated endogenous ␤-arrestin (center panel) and phospho-ERK1/2 (right panel). The representative immunoblots shown are from one of three separate experiments. C, cross-linked detergent lysates from stimulated V2R-and V2␤2R-expressing cells were resolved by SDS-PAGE under reducing (lower immunoblot) or nonreducing (upper immunoblot) conditions. The percent of the total phospho-ERK1/2 signal present in each stimulated lane that migrated with an apparent molecular weight of 42,000 -44,000 or Ͼ100,000 was quantified by scanning densitometry. The bar graph compares the distribution of phospho-ERK1/2 between pools of different apparent molecular weight, following V2R and V2␤2R stimulation. Data shown represent the mean Ϯ S.E. from three to six separate experiments.
FIG. 6. Effect of exchanging the C terminus of the V2R and ␤2AR on nuclear translocation of activated endogenous ERK1/2 following agonist stimulation. COS-7 cells were transfected with plasmids for either HA-V2R or HA-V2␤2R. Serum-starved cells were stimulated with vasopressin (1 M) for 5 min, cell nuclei were isolated as described, and the phospho-ERK1/2 content of the nuclear and extranuclear fractions was determined by immunoblotting. A, phospho-ERK1/2 content of the whole cell lysate collected prior to fractionation (upper immunoblot) and the nuclear fraction (lower immunoblot) following stimulation of V2R and V2␤2R. B, phospho-ERK1/2 content of whole cell lysate collected prior to fractionation (upper immunoblot) and the extranuclear fraction (lower immunoblot) following stimulation of V2R and V2␤2R. In the bar graphs, data are expressed as the percentage of the total cellular phospho-ERK1/2 pool present in the nuclear or extranuclear fraction. Data shown represent the mean Ϯ S.E. of four independent experiments. *, greater or less than V2R, p Ͻ 0.05. (15), may account for the effect of ␤-arrestin binding on ERK distribution.
The Transcriptional Activity of GPCR-activated ERK1/2 Is Regulated by the Stability of the Receptor-␤-Arrestin Interaction-The preceding data demonstrate that the ability of a GPCR to bind stably to ␤-arrestin correlates with increased activation of ␤-arrestin-bound of ERK1/2, and decreased nuclear translocation of phospho-ERK1/2. To determine whether the functional consequences of ERK activation are affected by the stability of the receptor-␤-arrestin interaction at endogenous levels of ␤-arrestin and ERK1/2 expression, we employed stably transfected HEK-293 cell lines expressing the V2R and V2␤2R. These cell lines exhibit comparable levels of cAMP production in response to vasopressin stimulation, but differ in the stability of the receptor-␤-arrestin interaction and the pat-tern of receptor internalization, dephosphorylation, and recycling. Fig. 7A depicts the time course of vasopressin-stimulated ERK1/2 phosphorylation over 4 h in the V2R-and V2␤2Rexpressing HEK-293 cells. At short time points, up to ϳ10 min, the responses were indistinguishable. Although the overall level of ERK1/2 phosphorylation declined dramatically after 30 min for both receptors, the level of persistent ERK1/2 phosphorylation was modestly, but significantly, greater for the V2␤2R than the V2R. Previous work has shown replacement of the V2R tail with that of the ␤2AR increases the rate of receptor dephosphorylation and recycling after endocytosis (3). Thus, this persistent signal may represent the effect of ␤-arrestin binding on the rate of receptor recycling and the steady state level of undesensitized receptors on the plasma membrane in the continuous presence of agonist. Fig. 7B compares the ability of V2R and V2␤2R to induce transcription of an Elk1-driven luciferase reporter. This response is dependent on activation of endogenous ERK1/2, because it is completely eliminated by incubation with the MEK inhibitor, PD98059 (data not shown). As shown, the chimeric receptor produced a transcriptional response that was ϳ8-fold greater than the wild type receptor. The response to stimulation of endogenous epidermal growth factor (EGF) receptors was comparable between the two cell lines.
Compared with the V2R, activation of V2␤2R resulted in greater nuclear translocation of activated ERK1/2 (Fig. 6), more persistent ERK1/2 activation, and enhanced ERK1/2-dependent transcription (Fig. 7). Because nuclear translocation of ERK1/2 is required for growth factor-stimulated mitogenesis (16), we compared the ability of the V2R and V2␤2R to stimulate DNA synthesis in the two cell lines. As shown in Fig. 8, the V2R, despite its ability to mediate robust activation of ERK1/2, failed to elicit a detectable increase in [ 3 H]thymidine incorporation into DNA. In contrast, the V2␤2R was weakly, but significantly, mitogenic, eliciting a response comparable with refeeding with 2% fetal calf serum. Because these two receptors differ only in their C termini, these data suggest that the stability of the receptor-␤-arrestin interaction influences the mitogenic potential of the V2R. It is likely that this effect is because of both binding of ERK1/2 to ␤-arrestin leading to its retention in the cytosol, and to ␤-arrestin-dependent removal of receptors from the cell surface leading to a diminished steady state level of ERK activation.

DISCUSSION
The binding of ␤-arrestins to agonist-occupied GPCRs serves two broad functions. By uncoupling the receptor from heterotrimeric G proteins and targeting it for endocytosis, ␤-arrestin binding is central to the processes of GPCR desensitization and sequestration that lead to rapid termination of G protein-dependent signals. At the same time, ␤-arrestins function as adapter proteins, bringing various signaling molecules into complex with the receptor and initiating additional ␤-arrestindependent signaling events. The finding that ␤-arrestins can interact directly with potential enzymatic effectors such as Src family tyrosine kinases (8,(17)(18)(19) and ubiquitin ligases (20), as well as components of the ERK1/2 (8,9,11,12) and JNK3 (10,15,21) mitogen-activated protein kinase modules, suggests that ␤-arrestins may serve in a variety of signaling roles.
Activation of the ERK cascade by heptahelical receptors is a complex process that can simultaneously involve multiple mechanisms, and in which the predominant mechanism varies depending on both the GPCR and the cellular context in which the receptor is expressed (22,23). For example, in S49 lymphoma cells, ␤2AR-mediated ERK1/2 activation occurs via a G s -, adenylyl cyclase-, and PKA-dependent pathway that causes activation of the small GTPase, Rap1 (24,25), whereas in fibroblasts, cardiomyocytes, and pancreatic acinar cells ERK1/2 activation is achieved largely through a Ras-dependent mechanism involving activation of pertussis toxin-sensitive G proteins and "transactivation" of EGF receptors (26 -29). Other receptors, notably the AT1aR, and the PAR2 and neurokinin NK-1 receptors, appear to utilize ␤-arrestins as scaffolds to a significant extent (8,9,11,12), although AT1aRs are also clearly capable of mediating ERK activation through cross-talk with EGF or platelet-derived growth factor receptors (30 -32). We have previously shown that overexpression of ␤-arrestins in COS-7 cells enhances AT1aR-mediated ERK1/2 activation, but leads to cytosolic retention of ERK and a diminished transcriptional response (11,12).
Given their ability both to dampen receptor-G protein coupling and to act as scaffolds for ERK activation, ␤-arrestins are uniquely positioned to influence the balance between these different mechanisms. Furthermore, because the mechanism used to activate ERK influences its spatial distribution and transcriptional activity, the binding of ␤-arrestin and receptor might dictate the cellular response to ERK activation. In a number of systems, the proliferative response to GPCR stimulation involves cross-talk between GPCRs and EGF receptors (31,33). In contrast, ␤-arrestin-dependent ERK activation does not appear to lead to proliferative signaling. Wild type PAR2 receptors, which mediate ␤-arrestin-dependent activation of a predominantly cytosolic pool of ERK1/2 in KNRK cells, do not stimulate [ 3 H]thymidine incorporation or cell replication (9). As we have shown, altering the stability of the V2R receptor-␤-arrestin interaction by replacing the C-terminal tail of the V2R with that of the ␤2AR was sufficient to confer mitogenic potential on a receptor that, in this system, did not detectably stimulate [ 3 H]thymidine incorporation (Fig. 8). Fig. 9 depicts this conceptual role of ␤-arrestins in regulating the mechanism and functional consequences of GPCR-stimulated ERK activation. Activation of ERK1/2 via classical G protein-dependent mechanisms, potentially including transactivation of receptor tyrosine kinases, leads to nuclear translocation of ERK1/2, activation of ERK1/2-dependent transcription and a mitogenic response. In contrast, activation of ERK1/2 in a ␤-arrestin-bound pool leads to localized ERK1/2 activity, with less nuclear phospho-ERK1/2 and little or no mitogenic potential. Significantly, ␤-arrestin binding to the receptor not only confers the ability to activate ␤-arrestinbound ERK1/2, it also attenuates signaling via G protein-dependent pathways and determines the rate of receptor recycling. As we have shown, altering the stability of the receptor-␤-arrestin interaction affects not only the efficiency of GPCRmediated activation of a ␤-arrestin-bound ERK pool, but also the balance between the nuclear and cytosolic phospho-ERK1/2 and the steady-state level of ERK1/2 activity during prolonged stimulation. This shift undoubtedly reflects the role of ␤-arrestins in receptor desensitization, sequestration, and recycling as much as it does the enhanced utilization of ␤-arrestin scaffolds by receptors that bind stably to ␤-arrestin.
The extent to which "G protein-dependent" and "␤-arrestindependent" ERK activation are genuinely independent processes is unclear. In murine fibroblasts derived from ␤-arrestin 1/␤-arrestin 2 homozygous null embryos, lysophosphatidic acid-stimulated ERK1/2 activation is sensitive to inhibitors of the EGF receptor tyrosine kinase, suggesting that GPCR-stimulated EGF receptor transactivation can occur in a ␤-arrestin null background. 2 On the other hand, ␤-arrestin recruitment requires GRK-mediated receptor phosphorylation, and receptor phosphorylation by GRK2 or GRK3 requires G␤␥ subunitmediated membrane translocation of the kinase. Thus, one might expect that ␤-arrestin-dependent ERK activation would not be G protein-independent. Rather, it would represent a 2 T. Kohout and R. J. Lefkowitz, unpublished observations. FIG. 9. Proposed model of the effect of ␤-arrestin binding on ERK1/2 activation and function. The binding of agonist to a GPCR initially leads to ERK activation via G protein-dependent pathways. The binding of ␤-arrestin simultaneously inhibits G protein-dependent ERK1/2 activation, by inducing homologous receptor desensitization and sequestration, and initiates the activation of a ␤-arrestin-bound pool of phospho-ERK1/2. For GPCRs that form stable receptor-␤-arrestin complexes, activation of the ␤-arrestin-dependent pathway is more pronounced, leading to the formation of a functionally distinct pool of phospho-ERK1/2. shift in the mechanism of ERK activation coincident with homologous desensitization that leads to the localized activation of a functionally distinct pool of ERK1/2. Interestingly, however, Seta et al. (34) have recently reported that in Chinese hamster ovary cells, a mutant AT1aR that is markedly impaired in G protein coupling is still fully competent to induce ERK1/2 activation, but that ERK activated by the mutant receptor is retained in the cytosol and is transcriptionally inactive (34). While these authors do not implicate ␤-arrestins in the putatively "G protein-independent" activation of ERK, their data do suggest the existence of one or more mechanisms of ERK activation that do not require G protein activation, and that, like the ␤-arrestin-dependent activation of ERK, leads to localized ERK activation outside the of the cell nucleus.
Our data demonstrate the significant impact of the receptor-␤-arrestin interaction on the mechanism and functional consequences of ERK activation. The activation of ERK bound to ␤-arrestin, which is favored in the setting of a stable receptor-␤-arrestin interaction, limits nuclear translocation of ERK and attenuates Elk-1-driven transcription. However, any additional functions of ␤-arrestin-bound ERK1/2 remain to be discovered. ERK1/2 are known to phosphorylate multiple plasma membrane, cytoplasmic, and cytoskeletal substrates (16), including several proteins involved in heptahelical receptor signaling, such as ␤-arrestin 1 (35), GRK2 (36,37), and the G␣interacting protein GAIP (38). Thus, one role of ␤-arrestin-ERK complexes could be to target ERK1/2 to substrates involved in the regulation of GPCR signaling or intracellular trafficking. Alternatively, ␤-arrestin-bound ERK1/2 may phosphorylate other cytosolic kinases involved in transcriptional regulation, such as p90 RSK (39), which in turn relay signals to the nucleus. In such a model, transcriptional events mediated directly by the nuclear pool of ERK1/2 would be attenuated, whereas alternate pathways of ERK-dependent transcription would persist, resulting in an altered pattern of GPCR-stimulated transcription.