β-Arrestin-dependent, G Protein-independent ERK1/2 Activation by the β2 Adrenergic Receptor*

Physiological effects of β adrenergic receptor (β2AR) stimulation have been classically shown to result from Gs-dependent adenylyl cyclase activation. Here we demonstrate a novel signaling mechanism wherein β-arrestins mediate β2AR signaling to extracellular-signal regulated kinases 1/2 (ERK 1/2) independent of G protein activation. Activation of ERK1/2 by the β2AR expressed in HEK-293 cells was resolved into two components dependent, respectively, on Gs-Gi/protein kinase A (PKA) or β-arrestins. G protein-dependent activity was rapid, peaking within 2-5 min, was quite transient, was blocked by pertussis toxin (Gi inhibitor) and H-89 (PKA inhibitor), and was insensitive to depletion of endogenous β-arrestins by siRNA. β-Arrestin-dependent activation was slower in onset (peak 5-10 min), less robust, but more sustained and showed little decrement over 30 min. It was insensitive to pertussis toxin and H-89 and sensitive to depletion of either β-arrestin1 or -2 by small interfering RNA. In Gs knock-out mouse embryonic fibroblasts, wild-type β2AR recruited β-arrestin2-green fluorescent protein and activated pertussis toxin-insensitive ERK1/2. Furthermore, a novel β2AR mutant (β2ART68F,Y132G,Y219A or β2ARTYY), rationally designed based on Evolutionary Trace analysis, was incapable of G protein activation but could recruit β-arrestins, undergo β-arrestin-dependent internalization, and activate β-arrestin-dependent ERK. Interestingly, overexpression of GRK5 or -6 increased mutant receptor phosphorylation and β-arrestin recruitment, led to the formation of stable receptor-β-arrestin complexes on endosomes, and increased agonist-stimulated phospho-ERK1/2. In contrast, GRK2, membrane translocation of which requires Gβγ release upon G protein activation, was ineffective unless it was constitutively targeted to the plasma membrane by a prenylation signal (CAAX). These findings demonstrate that the β2AR can signal to ERK via a GRK5/6-β-arrestin-dependent pathway, which is independent of G protein coupling.

The ␤2-adrenergic receptor (␤2AR) 4 is a well studied member of the large and diverse group of seven transmembrane receptors (7TMRs), which have been shown classically to exert their intracellular effects through G protein activation (1)(2)(3). Agonist stimulation of the ␤2AR leads to G s -mediated activation of adenylyl cyclase, resulting in the production of cAMP and subsequent downstream signaling events. Moreover, additional studies both in cultured cell lines and in vitro have demonstrated that, in response to agonist, the ␤2AR can undergo PKAdependent phosphorylation leading to activation of G i (a process referred to as G protein "switching"), thereby effectively changing the signaling specificity of the receptor (4).
Cessation of agonist-activated ␤2AR-G s -mediated signaling occurs via recruitment of modulatory proteins, ␤-arrestins, to the cytoplasmic surface of the receptor, a process that is enhanced by receptor phosphorylation by G protein-coupled receptor kinases (GRKs) (5). ␤-arrestin binding physically prevents receptor-G s interaction, leading to desensitization of receptor-mediated activation of G s . ␤-Arrestin binding further promotes the subsequent cytosol to membrane translocation of clathrin and adaptor protein AP-2, resulting in receptor endocytosis in clathrin-coated vesicles (6 -8).
Recent evidence has emerged, however, that, for a variety of receptors, ␤-arrestins can also function as molecular mediators of G proteinindependent signaling by acting to scaffold a variety of signaling proteins. These include small-GTP-binding proteins, as well as members of the ERK-mitogen-activated protein kinase (MAPK) signal transduction pathway, among others (5). For example, angiotensin II receptor type Ia (AT1aR) signaling following activation by angiotensin II is known to promote AT1aR coupling to G␣ q/11 (9). However, AT1aR activation by a peptide analogue of angiotensin II (termed SII), or angiotensin II activation of a mutant AT1aR (DRY 3 AAY) unable to couple to G proteins, results in activation of the MAPK cascade, which is G␣ q -independent but ␤-arrestin-dependent (10 -13). Further, these two independent signaling pathways show both distinct spatial (G q nuclear, ␤-arrestin cytosolic vesicles) and temporal (G q rapid/transient; ␤-arrestin slower and prolonged) characteristics with respect to ERK1/2 activation (14). Recent studies using siRNA have confirmed these results and have extended this general paradigm of distinct G protein/␤-arrestin signaling to the V2 vasopressin receptor (V2R) (15).
7TMRs can be functionally divided into two broad categories based upon their interaction with ␤-arrestins following agonist activation. "Class B" receptors such as both the AT1aR and V2R form a very stable interaction complex with ␤-arrestins. "Class A" receptors, including the ␤2AR, on the other hand, are known to form only transient complexes with ␤-arrestin (16). Whether ␤-arrestin-dependent ERK1/2 activation can occur via such Class A receptors is at present unknown. To investigate the potential for ␤-arrestin-dependent MAPK activation via a Class A receptor, we have used the ␤2AR to address these issues using gene silencing technology, G␣ s null cells, as well as a novel Evolutionary Trace-based mutant ␤2AR engineered to be uncoupled from G protein signaling.
Generation of Cell Lines Stably Expressing Receptors-Because HEK-293 cells have endogenously expressed ␤2ARs, higher expression levels of mutant receptors were necessary to correlate the effects attributable to the mutations. To maintain consistency of expression levels between different experiments, we created clonal HEK-293 cell lines with almost identical receptor expression levels for ␤2AR and ␤2AR TYY . Early passage 293 cells were transfected with 1 g of receptor plasmid, and positive clones were selected against G418 (1 mg/ml). Receptor expression levels were determined by radioligand binding (see below). When stable expression was achieved, the cells were cultured in the presence of 400 g/ml G418. Each experiment was replicated in at least two different clonal lines for both the ␤2AR and ␤2AR TYY .
cAMP Assay-Cells were plated on 12-well dishes (polylysine-Dcoated, Biocoat). Serum-deprived cells were incubated for 10 min with 1 mM 3-isobutyl-1-methylxanthine (IBMX) to inactivate phosphodiesterases. Cells were then treated with vehicle or a range of isoproterenol concentrations for 10 min at 37°C in the presence of IBMX. Forskolin treatment (10 M) for 10 min was used to measure cAMP generation due to direct activation of adenylyl cyclase. Stimulation was terminated by placing cells on ice and adding EDTA. Samples were collected by gentle scraping, boiled at 100°C for 5 min, and centrifuged for 10 min at 10,000 ϫ g. The supernatants were used in triplicate to assay for cAMP levels according to 3 H-labeled cAMP system (Amersham Biosciences) as per the manufacturer's protocol. The amount of cAMP was expressed as a percentage of that obtained with forskolin.
[ 125 I](Ϫ)Iodocyanopindolol Binding on Monolayers of Cells-Receptor expression was measured by [ 125 I](Ϫ)iodocyanopindolol radioligand binding on monolayers of cells on poly-D-lysine-coated 12-well dishes (Biocoat) in Dulbecco's modified Eagle's medium buffered with 10 mM HEPES (pH 7.5) and 5 mM MgCl 2 . Binding was performed in triplicate with 400 pM 125 I(Ϫ)iodocyanopindolol in the presence or absence of the hydrophobic antagonist propranolol (10 M, to define nonspecific binding). After incubation at 37°C for 30 min, the cells were placed on ice and washed several times with phosphate-buffered saline buffer containing calcium and magnesium. Cells were solubilized in 0.1 N NaOH and 0.1% SDS and counted for 125 I.
Internalization-FLAG epitope-tagged receptors expressed in COS-7 cells in twelve-well dishes were treated with or without agonist for 30 min in serum-free medium at 37°C. Cell-surface receptors were labeled with M2 FLAG monoclonal antibody and fluorescein isothiocyanate-conjugated goat antibody to mouse IgG as secondary antibody. Receptor internalization was quantified as the loss of cell-surface receptors, as measured by flow cytometry (Flow Cytometry Facility, Duke University).
Metabolic Labeling-HEK-293 cells stably expressing the ␤2AR or mutant receptors were incubated at 37°C for 60 min in phosphate-free minimal essential medium containing [ 32 P]P i (100 Ci/ml). After isoproterenol treatment for 5 min at 37°C, FLAG receptors were immunoprecipitated and samples separated by SDS-PAGE. Gels were dried and exposed to a PhosphorImager screen, and the 32 P incorporation was quantified.

␤-Arrestins and Adrenergic Signaling
Confocal Microscopy-HEK-293 cells stably expressing the WT, TYY, or GRK-PKA-receptors on 10-cm dishes were transiently transfected with ␤-arrestin2-GFP using FuGENE (Roche Applied Science). G s KO cells were transiently transfected with FLAG-␤2AR and ␤-arres-tin2-GFP utilizing Lipofectamine 2000 reagent. Twenty-four hours post-transfection, cells were plated on collagen-coated 35-mm glass bottom plates. On the following day, cells were starved for at least 2 h in serum-free medium prior to stimulation. After stimulation, cells were fixed with 5% formaldehyde diluted in phosphate-buffered saline containing calcium and magnesium before confocal analyses. Images of GFP fluorescence were collected using single line excitation (488 nm).
Immunoprecipitation with DSP Cross-linking-HEK-293 cells stably expressing either ␤2AR or ␤2AR TYY that are FLAG epitope-tagged were used for these experiments. Cells on 100-mm dishes were incubated in 4.0 ml of Dulbecco's phosphate-buffered saline plus 10 mM HEPES for 1 h at 37°C and subsequently stimulated with 10 M isoproterenol for 5 min. A membrane-permeable, hydrolyzable covalent cross-linker dithiobissuccinimidyl propionate (DSP, from Pierce) was added to the dishes with slow and constant agitation. After 20-min incubation at room temperature, the DSP reaction was quenched by adding Tris-HCl, pH 7.5 (final concentration of 25 mM). Cells were solubilized, and receptors were immunoprecipitated with
Evolutionary Trace-The relative importance of transmembrane sequence residues was computed from an Evolutionary Trace (21) of 129 visual opsins, 69 bioamine, 58 olfactory, and 82 chemokine receptors. Comparison to the literature revealed among the residues ranked in the top 20th percentile, a structural cluster of seven likely to participate in G protein-coupling in all rhodopsin-like receptors: Thr-68, Tyr-132, Ala-134, Tyr-219, Leu-275, Tyr-326, and Pro-330, using human-␤2AR numbering (22). These were the starting point for the rational design of the TYY mutant described below. Four of these were eliminated, because mutational data either: linked them to increased G protein activation (Leu-275) (23)(24)(25)(26) or to decreased sequestration (Tyr-326) (27,28) or was insufficient (Asn-330 and Ala-134). By contrast, the remaining three residues had mutations shown to decrease G protein signaling in multiple receptors and with no data suggesting a decrease in internalization.

Isoproterenol-dependent, H-89/PTX-insensitive pERK Stimulated by
the ␤2AR in HEK-293 Cells-Two different ␤2AR-mediated pathways of G s -dependent ERK1/2 activation have been described in HEK-293 cells (29 -31). Both pathways involve the activity of PKA, in one case to activate downstream effectors such as the small G protein Rap and the other to phosphorylate the ␤2AR itself, thereby switching its coupling to G i proteins. To evaluate the extent of PKA dependence of ERK1/2 acti-vation in HEK-293 cells, we examined the time course of isoproterenolinduced pERK in the presence and absence of H-89, a well defined PKA inhibitor. ␤2AR is endogenously expressed in HEK-293 cells (ϳ40 fmol/mg of cellular protein) and mediates a modest and transient activation of ERK1/2 with peak signal at 5 min after the addition of 10 M isoproterenol ( Fig. 1, A and B). The level of pERK declines to 22% (of maximal response) at 15 min and returns almost to basal levels after 30 min. Stable overexpression of a FLAG-tagged ␤2AR (2 pmol) augments the overall ERK1/2 activation by 4-fold at early time points and up to 8-fold beyond 10 min of agonist treatment (Fig. 1, C and D, and data not shown). However, the overall pattern of ERK1/2 activation is identical for both endogenous and exogenous receptor expression. Under both conditions, H-89 abolishes most of the early activity but, quite surprisingly, pERK at later time points is unaffected, indicating the presence of H-89-insensitive, isoproterenol-dependent signals beyond 5 min (Fig. 1, A-D). H-89 treatment precludes ␤2AR-G i coupling due to the absence of G s /G i switching (4). However, to independently assess the role of G i in the activation of ERK at late time points, we also determined the effects of pertussis toxin on the time course of isoproterenol-stimulated pERK. Pertussis toxin preincubation of HEK-293 cells expressing the ␤2ARs had similar effects on isoproterenol-dependent ERK activation as H-89 treatment, especially at the late time points of isoproterenol stimulation. The signals beyond 5 min were resistant to pertussis toxin (Fig. 2, A-D) suggesting that the late activity occurs in the absence of G s or G i coupling.
Effects of ␤-Arrestin siRNA on ␤2AR-stimulated pERK-We hypothesized that the ERK activity at later time points that is insensitive to H-89 and PTX is mostly independent of G protein activity and might be regulated by ␤-arrestins. To test the potential role of ␤-arrestins, we analyzed the time course of isoproterenol-induced pERK after depleting cellular levels of ␤-arrestin1 or -2 by transfecting siRNA specifically directed against each isoform. In the presence of a non-targeting control siRNA, isoproterenol stimulated ERK phosphorylation in cells stably expressing the ␤2AR is identical to that observed without any siRNA transfection (compare Figs. 1C and 3A). However, both ␤-arrestin1 and -2 siRNA individually reduce the 5-min signal by ϳ50% and essentially eliminate signals beyond 10 min (Fig. 3, A and C). Minimal decrements were observed at 1 and 2 min suggesting that most of the early signals are due to G protein activation. In these experiments, we could achieve at least 90% reduction of the two isoforms by RNA interference. A representative Western blot of the levels of ␤-arrestin1 and -2 in control and siRNA-treated cells is shown in Fig. 3B. We also observed similar ␤-arrestin1/2-dependent ERK activation by the ␤2ARs upon transient expression (300 -800 fmol/mg of protein) in HEK-293 cells (data not shown). Furthermore, ␤-arrestin siRNA transfection led to complete elimination of the H-89-insensitive pERK (Fig. 3D) suggesting that most of the late activity that is PKA-independent is in fact ␤-arrestin-dependent. Our attempts to determine PTX effects in siRNA-transfected cells were unsuccessful for technical reasons. Unfortunately, the siRNAtransfected cells could not withstand the prolonged starvation conditions (16 h) used for PTX pretreatment.
An Evolutionary Trace-based Mutant ␤2AR Uncoupled from G s -We have previously shown that a mutant Angiotensin II 1a receptor (AT1aR DRY 3 AAY), which is completely uncoupled from G q proteins is nonetheless able to activate ERK1/2 in response to angiotensin stimulation, in a ␤-arrestin2-dependent manner (12). To rationally design an analogous ␤2AR mutant, we employed Evolutionary Trace analyses (see "Materials and Methods") and altered three residues, Thr-68, Tyr-132, and Tyr-219 by site-directed mutagenesis. These residues are indicated in an amino acid sequence alignment of the ␤2AR and rhodopsin (Fig. 4A). The relative positions of the corresponding residues are shown in a structural model of rhodopsin (Fig. 4B). These were mutated together to new side chains that were charge neutral, non-conservative, and not found at cognate positions  (53) indicating the corresponding sites of mutations. C, cAMP generated upon stimulation of HEK-293 cells that are transfected with either pcDNA3 (endogenous ␤ARs), ␤2AR or ␤2AR TYY plasmids. Cells were incubated with IBMX and then treated with increasing doses of isoproterenol. Isoproterenol-dependent cAMP values were normalized to forskolin-induced levels. Data represent mean values Ϯ S.E. from 6 -8 independent experiments. D, HEK-293 cells transfected with pcDNA3, FLAG-␤2AR receptor, or FLAG-␤2AR TYY were stimulated or not with 10 M isoproterenol. FLAG receptors were immunoprecipitated and probed with an antibody specific to phosphoserines 345 and 346 on the C-tail of the ␤2AR (upper panel). This is a consensus site for PKA phosphorylation. The same blot was reprobed with a ␤2AR antibody (H-20, Santa Cruz Biotechnology) as shown in the lower panel. These blots are representative of four identical experiments.
in any of the receptors traced for this study. This T68F-Y132G-Y219A mutant was predicted to drastically alter the component of G protein coupling located at the boundary between the transmembrane domain and the intracellular loops while leaving intact the cytoplasmic interaction of these loops with GRK and ␤-arrestin.
To differentiate characteristics of the mutant ␤2AR TYY from those of the endogenously expressed ␤2AR in HEK-293 cells, experimental determinations were made only when the mutant receptor was overexpressed at Ͼ20to 50-fold. Under such conditions, no difference was observed between untransfected cells (endogenous receptors), and ␤2AR TYY -transfected cells (Fig. 4C) with respect to either the levels of cAMP accumulation or the half-maximal stimulating concentrations of isoproterenol. On the other hand, when the ␤2AR was expressed in these cells at levels comparable to the ␤2AR TYY (i.e. Ͼ20 -50 times in excess of endogenous receptors) cAMP response was much more robust with a greatly increased V max and reduced EC 50 for isoproterenol as compared with the endogenous receptors. The dose-response curves shown in Fig. 4C establish that ␤2AR TYY does not stimulate cAMP accumulation beyond that elicited by the endogenous ␤2AR, confirming its lack of coupling to G s .
Previous studies have demonstrated that cAMP increase and subsequent PKA activation lead to PKA-mediated phosphorylation of the ␤2AR on serine residues within the consensus motif RRSS (Ser-261 and Ser-262 in the third loop and Ser-345 and Ser-346 in the carboxyl tail) (30,32). Because ␤2AR TYY does not stimulate a cAMP response, it would be predicted not to undergo the feedback PKA phosphorylation that occurs in the ␤2AR. To test this, we employed a commercially available PKA site-specific antibody that recognizes phosphoserines (Ser-345 and Ser-346) and analyzed receptor immunoprecipitates for phosphorylation in a Western blot (Fig. 4D). A basal level of phosphorylation is detected for the ␤2AR in the absence of isoproterenol stimulation (lane 3 in the upper panel, Fig.  4D), and a 5-min agonist treatment leads to a marked increase in phosphorylation. In contrast, both basal-and agonist-induced phosphorylations are absent in the ␤2AR TYY samples (lanes 5 and 6 in the upper panel, Fig. 4D). Both the ␤2AR and ␤2AR TYY immunoprecipitates contained equal amount of receptor protein as detected by a ␤2AR-specific antibody (lower panel, Fig. 4D). These data confirm that ␤2AR TYY does not provoke feedback phosphorylation by PKA.

␤-Arrestin Binds and Functions as an Endocytic Adaptor for ␤2AR TYY -
To determine if ␤-arrestin can interact with ␤2AR TYY , we utilized confocal microscopy to visualize the translocation of ␤-arrestin2-GFP to agonist-activated receptors. HEK-293 cells stably expressing ϳ2 pmol of either ␤2AR or ␤2AR TYY receptors were transiently transfected with ␤-arrestin2-GFP. Prior to isoproterenol stimulation, ␤-arrestin is distributed uniformly in the cytosol (not shown). In cells expressing ␤2AR TYY less robust plasma membrane translocation was observed upon isoproterenol stimulation in comparison to cells harboring the ␤2AR (Fig. 5A, first two panels). In contrast, ␤-arrestin2 recruitment to ␤2AR TYY was more pronounced when compared with cells stably expressing a phosphorylation-defective ␤2AR mutant lacking all phosphorylation sites, which has virtually no ␤-arrestin binding properties (Fig. 5A, last panel).
We also determined the association of endogenous ␤-arrestins with the stably expressed receptors (␤2AR or ␤2AR TYY ) by immunoprecipitation assays performed in the presence of chemical cross-linkers (Fig.  5B). Detection of ␤-arrestins utilized an antibody that recognizes both ␤-arrestin isoforms. In these assays, no ␤-arrestin binding was observed prior to agonist treatment. Upon isoproterenol stimulation, robust ␤-arrestin2 and weak ␤-arrestin1 recruitment was seen for both ␤2AR and ␤2AR TYY . However, much less ␤-arrestin was bound to ␤2AR TYY amounting to ϳ23% of the levels coimmunoprecipitated with ␤2ARs (Fig. 5B).
It has been reported that ␤2AR internalizes poorly in COS-7 cells, which express very low levels of endogenous ␤-arrestin. Furthermore, exogenous expression of ␤-arrestin2 has been shown to enhance the isoproterenol-induced internalization of the ␤2AR in COS-7 cells (33). As seen in Fig. 5C, FLAG epitope-tagged ␤2AR TYY internalized to the same extent (ϳ10%) as the FLAG-␤2AR in COS-7 cells as measured by the disappearance of cell-surface receptors after a 30-min isoproterenol treatment. Additionally, expression of ␤-arrestin2 increased the internalization of the WT as well as the mutant by ϳ3-fold (Fig. 5C). These data suggest that the mutant receptor utilizes ␤-arrestin-dependent endocytotic mechanisms similar to the WT receptor.
ERK Activation by ␤2AR and ␤2AR TYY -We next evaluated whether ␤2AR TYY , which does not stimulate cAMP accumulation, could nonetheless stimulate cellular ERK1/2. We treated untransfected HEK-293 cells (endogenous WT receptors) and either WT or TYY stable expressers (ϳ2 pmol of receptors/mg) with a range of isoproterenol concentrations for 5 min and analyzed whole cell lysates for pERK and ERK1/2 content (Fig. 6, A and B). Even at 10 pM isoproterenol, ERK activation was detected in the ␤2AR stable cells. Peak activation was reached at ϳ10 -20 nM isoproterenol. On the other hand, 1 nM isoproterenol could only weakly activate ERK via the endogenous and ␤2AR TYY receptors. Peak activity for the endogenous receptors was reached at 100 nM isoproterenol. However, at this agonist concentration, ␤2AR TYY -mediated stimulation of pERK was much greater than that mediated by the endogenous receptors. At maximal agonist concentration, ␤2AR TYY (10 M)-mediated ERK1/2 activation was 3-to 4-fold more than what was induced by the endogenous receptors and was essentially equivalent to that observed by the ␤2AR. Thus, ␤2AR TYY can initiate robust signals via the effector pERK in the absence of any second messenger generation.
To further validate our findings with the ␤2AR TYY that ERK activation can proceed in the absence of G s coupling, we tested G s null mouse embryonic fibroblasts for isoproterenol-stimulated pERK. Isoproterenol treatment of these cells did not yield any detectable cAMP accumulation (data not shown) (17). A time course of pERK stimulated in these cells by 10 M isoproterenol is shown in Fig. 7A (left panel). The ERK activity has a slow onset and peaks ϳ10 min and returns to basal levels by 30 min. This pERK is insensitive to pertussis toxin treatment. Accordingly, the ERK activation by endogenous ␤ARs in these cells, which lack G␣ s , does not proceed via G i . As a comparison, a time course of ERK activation in the presence and absence of pertussis toxin was determined in wild-type MEF cells (Fig. 7A, right panel). The pattern of

␤-Arrestins and Adrenergic Signaling
ERK activation is reminiscent of what is seen in HEK-293 cells, such that peak activity occurs between 2 and 5 min. Furthermore just as observed in HEK-293 cells, the pERK time course includes an early phase sensitive to pertussis toxin and a late phase that is not (compare with Fig. 2A). These results suggest that pertussis toxin-sensitive G i -mediated ERK activation can occur only after G␣ s coupling.
In the G s null cells a normal "Class A" pattern of ␤-arrestin2-GFP recruitment was observed with isoproterenol stimulation (Fig. 7B). Iso-proterenol stimulation for 30 min also induced up to 30% internalization of ␤2ARs (data not shown). These data provide further evidence that the absence of cognate G protein activation does not alter the ␤-arrestin binding and internalization properties of the ␤2AR. The pERK stimulated by the ␤2AR in the absence of G s is most likely mediated by ␤-arrestin. Although siRNA against murine ␤-arrestin isoforms are available (34), we were unable to achieve any reduction in ␤-arrestin levels in these cells. Hence we could not determine the effects of ␤-ar-  (ChemGenius2). B, the panels depict confocal images of ␤-arrestin2-GFP translocation to transfected ␤2ARs as stimulated by 10 M isoproterenol for 15 min in G s null cells. FIGURE 8. pERK stimulated by ␤2AR TYY is ␤-arrestin-dependent. HEK-293 cells stably expressing ␤2AR TYY were transfected with the indicated siRNA. Cells at ϳ50% confluence were serumstarved for 4 h and stimulated with 100 nM isoproterenol for the indicated times. Whole cell lysates were analyzed for pERK, ERK, and ␤-arrestin levels. Fifteen micrograms of protein was used in each sample lane. The pERK and ERK bands were quantified by densitometry and pERK was normalized to total ERK levels. The graphs (A) represent mean values Ϯ S.E. from six independent experiments. No significant difference was seen between the curves representing ␤-arrestin1 or two siRNAtreated samples. However, curves representing either ␤-arrestin1or ␤-arrestin2-depleted cells were significantly different from the control curve as analyzed by two-way ANOVA. ***, p Ͻ 0.001 by Bonferroni post test. Representative blots for ␤-arrestin, pERK, and ERK levels are depicted in panels B and C, respectively.

␤-Arrestins and Adrenergic Signaling
restin depletion on the ␤2AR-mediated pERK in G s null cells. The lower temperature required for culturing these cells (33°C) (17) probably prevented efficient siRNA uptake. On the other hand, we could demonstrate that the G s -independent pERK generated by the ␤2AR TYY was in fact ␤-arrestin-dependent (see below).
pERK Stimulated by ␤2AR TYY Is ␤-Arrestin-dependent-To determine if the ERK1/2 activation elicited by ␤2AR TYY is transduced by ␤-arrestin proteins, we next examined the effect of ␤-arrestin depletion by RNA interference on the ␤2AR TYY -mediated ERK response (Fig. 8, A and C). HEK-293 cells stably expressing ␤2AR TYY (2 pmol/mg) were transfected with control, ␤-arrestin1, or ␤-arrestin2 siRNA. Under control conditions, after isoproterenol stimulation, peak activity occurred at 5 min and decreased to 20% of maximal levels at 30 min. ␤-Arrestin1 as well as ␤-ar-restin2 siRNA dramatically decreased the 5-min signal to 15-18% and completely abolished the late activity. The reductions in pERK levels in the presence of ␤-arrestin siRNA correlate with similar amounts of reduction in ␤-arrestin levels (Fig. 8B). These data indicate that ERK activation stimulated by the ␤2AR TYY is mediated largely by ␤-arrestin isoforms.
The data presented in Figs. 1 and 3 demonstrate that ERK activity stimulated by the ␤2AR can be resolved into components mediated by either G s /PKA or ␤-arrestins 1 and 2, sensitive, respectively, to H-89 and ␤-arrestin siRNA. Because ␤2AR TYY is uncoupled from G s it would be expected that its ability to activate ERK1/2 might be mediated exclusively by ␤-arrestins. To confirm this we tested the sensitivity of ␤2AR TYY -activated ERK to H-89 (Fig. 9, A and B). These experiments are complicated by the fact that activity observed with ␤2AR TYY -expressing cells also contains a component due to the endogenous ␤2AR. As seen in Fig. 9, after isoproterenol stimulation of ␤2AR TYY -expressing cells, only ERK activity stimulated in the first few minutes is sensitive to H-89. From 10 min on, the activity is completely resistant to H-89. This early H-89-sensitive activity is likely due to PKA-dependent ERK stimulated via endogenous ␤2ARs, whereas the bulk of H-89-resistant activity is attributable to ␤2AR TYY . Thus in Fig. 9 the curve depicting pERK in the presence of control siRNA and H-89 is likely a representation of the ␤2AR TYY -stimulated activity. Consistent with this model, this H-89-resistant activity is eliminated by siRNA to either ␤-arrestin1 or ␤-arrestin2 (Fig. 9, A and B).
Role of GRKs in ␤2AR TYY Phosphorylation and Signaling-All the above findings are consistent with the formation of a receptor-␤-arrestin complex for the ␤2AR TYY analogous to the ␤2AR leading to downstream ERK activation. To dissect the roles of GRK phosphorylation in ␤-arrestin recruitment to the ␤2AR TYY and regulation of ERK signaling, we analyzed effects of coexpression of different GRK isoforms on ␤2AR TYY phosphorylation, ␤-arrestin recruitment, and ERK activation.
We first determined the agonist-induced phosphorylation of the ␤2AR and ␤2AR TYY by 32 P metabolic labeling of HEK-293 cells stably expressing the receptors at ϳ2 pmol/mg of cellular protein. As seen in Fig. 10 (A and  B), ␤2AR TYY is phosphorylated to ϳ20% of WT levels as determined by quantification of autoradiographs by PhosphorImager. ␤2AR TYY phosphorylation was not augmented by GRK2 coexpression (Fig. 10, A and B). This is not surprising, because ␤2AR TYY does not couple to G proteins and thereby release G protein ␤␥ subunits that are known to play a major role in GRK2 membrane targeting. On the other hand, expression of either GRK2 with a membrane-tethering prenylation signal (CAAX) at the carboxyl terminus, or GRK5 that is constitutively localized to the plasma membrane, resulted in a doubling of ␤2AR TYY phosphorylation (Fig. 10, A and B).
We also determined ␤2AR TYY phosphorylation on specific GRK phosphorylation sites (Ser-355 and Ser-356) with commercially available antibodies directed specifically against these phospho-serines (35). As shown in Fig. 10C (upper left panel), a weak phosphorylation signal is detected in immunoprecipitates of ␤2AR TYY upon agonist stimulation. Coexpression of GRK5 or GRK6 but not GRK2 enhances ␤2AR TYY phosphorylation at these serine residues. However, in the presence of lower concentrations of agonist (100 nM isoproterenol), only GRK6 could augment ␤2AR TYY phosphorylation at these sites (data not shown). A robust phosphorylation signal was observed in response to agonist in the ␤2AR immunoprecipitates, and the signal changed minimally with coexpression of different GRK isoforms (Fig. 10C, upper right panel). These data suggest that ␤2AR TYY is preferentially phosphorylated by GRK 5/6 isoforms in HEK-293 cells upon agonist treatment.
As demonstrated above, ␤2AR TYY phosphorylation can be augmented by GRK5/6 isoforms but not by GRK2. To assess the relative effects of GRK phosphorylation on subsequent ␤-arrestin recruitment to the ␤2AR TYY , we expressed different GRK isoforms in HEK-293 cells

␤-Arrestins and Adrenergic Signaling
stably expressing the ␤2AR TYY , immunoprecipitated the receptors after chemical cross-linking and determined the amount of bound endogenous ␤-arrestins by Western blotting. Expression of GRK2 did not cause any increase in ␤-arrestin binding beyond that observed under mock conditions, upon 5 min of isoproterenol stimulation (Fig. 11A). In contrast, both GRK5 and GRK6 markedly enhanced the recruitment of ␤-arrestin to the ␤2AR TYY (Fig. 11, A and B). In the presence of GRK5 or GRK6, ␤-arrestin binding to the receptor increased by at least 2-fold (Fig. 11B). Again, these results closely correlate with the increase in ␤2AR TYY phosphorylation observed after the expression of these GRKs (Fig. 10C). In all cases, no further increase in ␤-arrestin recruitment was seen at longer times of agonist stimulation (data not shown).
To evaluate if GRK expression altered the translocation patterns of ␤-arrestin, we performed confocal microscopy by transiently expressing ␤-arrestin2-GFP along with GRKs in HEK-293 cells stably expressing either the ␤2AR or ␤2AR TYY receptors. For the ␤2AR, GRK2 expression caused greater cytosolic "clearance" of GFP fluorescence and formation of brighter puncta at the plasma membrane after 20 min of 1 M isoproterenol (Fig. 11C, compare first and second upper panels). Quite unexpectedly, GRK5 or GRK6 led to detection of ␤-arrestin-GFP in vesicles, indicating a stable Class B-type interaction between the internalized receptor and ␤-arrestin (Fig. 11C, top row, third and fourth  panels). In some experiments, basal recruitment of ␤-arrestin was observed at the plasma membrane with GRK5/6 but not GRK2 (data not shown). In the case of ␤2AR TYY , GRK2 overexpression did not increase the efficiency of ␤-arrestin translocation. Similar to the WT receptor, GRK5 and -6 promoted the receptor-driven accumulation of ␤-arrestin in intracellular vesicles, although less robustly than in the case of the WT receptor (Fig. 11C, bottom row, third and fourth panels).
To determine if the enhancing effects of GRK5 and -6 on receptor trafficking and on ␤-arrestin recruitment to the ␤2AR TYY are associated with increased ERK1/2 activation, we treated cells stably expressing the ␤2AR TYY (ϳ1 pmol/mg) with 1 M isoproterenol for different times and analyzed the cellular lysates by Western blotting for pERK1/2. ␤2AR TYY -stimulated pERK peaked at 5 min of agonist just as for the WT receptor. However, the signal was more sustained and remained fairly stable for 20 min. pERK1/2 levels returned to basal beyond 40 min (data not shown). GRK2 overexpression did not cause significant changes in the pERK activation by ␤2AR TYY (Fig. 12, A and B). In contrast, both GRK5 and -6 significantly augmented the activity especially at earlier time points (Fig. 12, A and C). Additionally, GRK5 also markedly enhanced ERK activation beyond 5 min (Fig. 12, A and C). These data indicate that GRK5 and -6 enhance not only ␤-arrestin recruitment to ␤2AR TYY and its trafficking but also its ability to activate ERK1/2.

DISCUSSION
Our results document that isoproterenol stimulation of the ␤2AR can induce pERK signals in HEK-293 cells by at least two separate pathways: (i) H-89-sensitive, PKA-dependent G protein-mediated and (ii) H-89-insensitive, PKA-independent ␤-arrestin-dependent. Furthermore, the H-89-insensitive signals are also unaffected by pertussis toxin indicating that the lateERKactivityisindependentofG s /G i switching.Inessence,theGprotein- dependent signals display an early and transient response, whereas the ␤-arrestin-dependent signals are late and sustained. Evidently, the ␤2AR can bind ␤-arrestin and stimulate pertussis toxin-insensitive ERK in the complete absence of cognate G␣ s proteins in G s null fibroblasts. When the ␤2AR is uncoupled from G s by mutagenesis, it can still signal to ERK in response to isoproterenol in an efficient and ␤-arrestin-dependent manner.
␤-Arrestin-dependent ERK1/2 activation is a recently appreciated mechanism of signal transduction elicited by 7TMRs (5). For the AT1aR and the V2R, this function is exclusively carried out by the ␤-arrestin2 isoform while the ␤-arrestin1 isoform plays an inhibitory role (15,36). On the other hand, both ␤-arrestin1 and ␤-arrestin2 are crucial for protease-activated receptor type 2-stimulated ERK activation (37,38). Furthermore, it has been shown for the AT1aR, utilizing either a mutant receptor or a mutant agonist peptide, that ␤-arrestin2-dependent pERK1/2 signals are stably generated in the absence of G protein coupling (12). These data indicate that the adaptor protein ␤-arrestin2 not only scaffolds MAPK components but also functions as an indispensable signal transducer in response to 7TMR activation.
Both the AT1aR and the V2R are members of a class of receptors that display stable interaction with ␤-arrestins leading to cointernalization of receptor-arrestin complexes to be localized on endosomes (16). For these receptors as well as for the protease-activated receptor type 2 and the Neurokinin1 receptors, ␤-arrestin has been shown to stably associate with pERK upon receptor activation (39 -41). Hence, it may not be surprising that ␤-arrestin can act as an essential intermediate in the ERK activation pathway elicited by these receptors. On the other hand, we now show that the ␤2AR, which interacts only transiently with ␤-arrestin-GFP at the plasma membrane and does not form stable receptor-␤arrestin-GFP complexes on endosomes, can nevertheless lead to ERK activation in a ␤-arrestin-dependent manner. Additionally, ␤2ARs require both ␤-arrestin1 and -2 for efficient ERK activation, because knockdown of either ␤-arrestin isoform leads to significant inhibition of isoproterenol-stimulated pERK. In contrast, both AT1aR and V2R specifically use ␤-arrestin2 as the signaling intermediate for the ERK pathway (15,36). It remains to be determined whether ␤-arrestin1 and -2 act sequentially or simultaneously in the isoproterenol-dependent ERK activation. Alternatively, heterodimerization of both ␤-arrestin1 and -2 may be necessary for ERK activation. An interesting question remains as to how transient complex formation between a class A receptor and ␤-arrestin can still induce the longer ␤-arrestin-dependent ERK activity demonstrated in this report for the ␤2AR and by Gesty-Palmer et al. (42) for LPA receptors. One possibility is that there is a continuous reformation of these receptor-␤-arrestin complexes at steady state, even though they are short-lived.
For both the AT1aR and the V2R, ␤-arrestin2-dependent ERK activation has been shown to be insensitive to inhibitors of second messenger-dependent kinases (12,15). Similarly, ␤-arrestin-dependent pERK stimulated by the ␤2AR is completely insensitive to the PKA inhibitor, H-89 (Figs. 3D and 9). H-89 has also been reported to act as a ␤ receptor blocker and as an activator of cAMP in alternate systems (43,44). In our assays, we do find that H-89 displays antagonistic properties at lower (Ͻ0.5 M) isoproterenol concentrations at the 5-min time point (data not shown). On the other hand, H-89 did not lead to any cAMP generation in our assays.
As demonstrated by the experiments performed with ␤2AR TYY , isoproterenol-stimulated ␤-arrestin-dependent ERK1/2 activation can proceed in the total absence of G protein activation or cAMP generation. This mutant is not phosphorylated by PKA and hence is not coupled to G i . Nonetheless ␤2AR TYY does recruit ␤-arrestin upon agonist stimulation and activates pERK in a more sustained manner than the pERK generated by the wildtype ␤2AR (compare the 10-min time point of Figs. 1, 3, 8, and 9). It is possible that, in the case of WT receptor, the G protein-dependent pathway exerts a suppressive effect on the ␤-arrestin-dependent pathway. Hence, in the absence of such suppressive mechanisms, we are able to detect persistent ERK activation by the ␤2AR TYY .
The activity of the ␤2AR TYY mutant raises several structural issues. First, three simultaneous point mutations at Thr-68, Tyr-132, and Tyr-219 eliminate the G protein interaction site but apparently preserve the sites for GRK and ␤-arrestin binding. One possibility is that these mutations directly disturb G protein binding at the transmembrane-cytoplasmic boundary but leave the ligand-dependent conformational switch intact so that binding sites in the cytoplasmic part of the loops are also intact. In addition, the G protein interaction could also be disrupted allosterically. In that case, the activated ␤2AR TYY cytoplasmic loops would be in an intermediate state sufficient for some, but not all, interactions. Second, the rational, ET-based design of the ␤2AR TYY has implications for the engineering of other 7TMRs and proteins. Here, the choice of residues to target for mutation was based on an ET analysis of diverse rhodopsin-like receptors to identify functionally important residues common to all. The choice of which side chains to mutate aimed to exclude known substitutions at cognate residues among the same diverse selection of 7TMRs. Neither strategy is specific to the adrenergic receptor. Future studies will test whether cognate mutations in other rhodopsin-like 7TMRs will produce similar results.
A striking feature of ␤2AR TYY is the apparent paradox of weak receptor phosphorylation, combined with moderate ␤-arrestin recruitment and the robust ERK activation. How can a receptor that binds relatively little ␤-arrestin still robustly engage a ␤-arrestin-dependent signaling pathway? Several prior and current observations serve to explain these phenomena. Recent RNA interference studies show that GRK2 mediates most of the phosphorylation and desensitization of 7TMRs. At least a decade of research demonstrates that agonist-induced ␤2AR phosphorylation is carried out by GRK2 and is facilitated by G protein ␤␥ subunits under both in vitro and in vivo conditions. Moreover, membrane targeting of GRK2/3 requires their binding to the G protein ␤␥ subunits (45)(46)(47)(48)(49)(50). As seen in Fig. 10, GRK2 coexpression does not augment ␤2AR TYY phosphorylation, but the membrane targeted GRK2-CAAX does so. Taken together, we conclude that, due to the absence of activated ␤␥ subunits, GRK2 is unable to phosphorylate the ␤2AR TYY . What are the consequences of this? First, as also shown for the AT1aR and V2R, in the absence of GRK2 phosphorylation, a dramatic reduction is observed in the amount of ␤-arrestin recruited to the ␤2AR (Fig.  5). For the AT1aR, it was also demonstrated that depletion of GRK2, actually increased, whereas overexpression of GRK2 decreased, ␤-arres-tin2-dependent ERK activation (13). Overall, ␤-arrestin recruited to GRK2 phosphorylated receptors appears not to be conformationally competent for engaging the ERK cascade (13). Consequently, the absence of GRK2-mediated phosphorylation should actually facilitate ␤-arrestin-dependent ERK activation by the ␤2AR TYY .
As seen in our 32 P metabolic labeling experiments, ␤2AR TYY is nonetheless phosphorylated to ϳ20% compared with the WT receptors. This phosphorylation is presumably mediated by the plasma membrane-associated GRK5/6 enzymes. Previous work utilizing inactivating monoclonal antibodies for GRK5/6 and recent RNA interference data indicate that in the case of several receptors, these enzymes are responsible for ϳ20% of agonist-induced receptor phosphorylation (15,18,51). Furthermore, GRK5/6 expression is crucial for ␤-arrestin-dependent ERK activation by both the AT1aR and the V2R (13,15). Consistent with these findings, we find that both GRK5 and -6 but not GRK2 augment ␤2AR TYY phosphorylation, ␤-arrestin recruitment, and ␤-arrestin-dependent ERK activation (Figs. 10 -12). Future studies utilizing siRNAs directed against various GRK isoforms should help to decipher the exact contribution of isoform specific phosphorylation of the ␤2AR and ␤2AR TYY in receptor desensitization, ␤-arrestin binding, and ERK activation.
An interesting and unexpected finding was the conversion of a Class A trafficking pattern of the ␤2AR and ␤2AR TYY (␤-arrestin at the plasma membrane) to a Class B type (␤-arrestin in endosomes) upon coexpression of GRK5/6 but not GRK2. Additionally, overexpression of GRK5 and 6 but not GRK2 augments isoproterenol stimulated ERK activation by the ␤2AR TYY especially beyond the 5-min time point when most of the signal is ␤-arrestin-dependent. Although GRK2 can improve both receptor phosphorylation and ␤-arrestin recruitment to the ␤2AR, GRK5/6 serve to create a stable receptor-␤-arrestin complex. These data are consistent with the hypothesis of a "barcode" resulting from the phosphorylation of unique residues on the receptor by a specific GRK isoform. Thus, it is proposed that GRK2 phosphorylates a set of preferred residues on a 7TMR such as the AT1aR. When ␤-arrestin binds to a receptor so phosphorylated, it likely undergoes a conformational change facilitating interaction with the endocytic but not the signal transduction machinery. In contrast, GRK5/6 phosphorylate unique residues on the receptor, which now impart signaling functions to the bound ␤-arrestin. Details of what molecular mechanisms regulate these phosphorylation events and specify which kinases phosphorylate a particular receptor in a particular situation remain to be determined.
Previous studies have shown that ␤2ARs utilize several G protein-dependent pathways to activate ERK in a wide variety of cells, including HEK-293 cells and cardiac myocytes (2,3,30,31,52). The current study clearlydemonstratestheexistenceofGprotein-independent, ␤-arrestindependent ERK activation stimulated by the ␤2AR in HEK-293 cells. Future studies aimed at identifying specific agonists that activate only the ␤-arrestin-dependent pathway and the characterization of ␤2AR TYY "knock-in" mice should facilitate a greater understanding of the physiological significance of this newly identified pathway.