The Nature of the Arrestin·Receptor Complex Determines the Ultimate Fate of the Internalized Receptor*

The vast majority of G protein-coupled receptors are desensitized by a uniform two-step mechanism: phosphorylation of an active receptor followed by arrestin binding. The arrestin·receptor complex is then internalized. Internalized receptor can be recycled back to the plasma membrane (resensitization) or targeted to lysosomes for degradation (down-regulation). The intracellular compartment where this choice is made and the molecular mechanisms involved are largely unknown. Here we used two arrestin2 mutants that bind with high affinity to phosphorylated and unphosphorylated agonist-activated β2-adrenergic receptor to manipulate the receptor-arrestin interface. We found that mutants support rapid internalization of β2-adrenergic receptor similar to wild type arrestin2. At the same time, phosphorylation-independent arrestin2 mutants facilitate receptor recycling and sharply reduce the rate of receptor loss, effectively protecting β2-adrenergic receptor from down-regulation even after very long (up to 24 h) agonist exposure. Phosphorylation-independent arrestin2 mutants dramatically reduce receptor phosphorylation in response to an agonist both in vitro and in cells. Interestingly, co-expression of high levels of β-adrenergic receptor kinase restores receptor down-regulation in the presence of mutants to the levels observed with wild type arrestin2. Our data suggest that unphosphorylated receptor internalized in complex with mutant arrestins recycles faster than phosphoreceptor and is less likely to get degraded. Thus, targeted manipulation of the characteristics of an arrestin protein that binds to a G protein-coupled receptors can dramatically change receptor trafficking and its ultimate fate in a cell.

G protein-coupled receptors (GPCRs) 1 are the largest known group of sensor proteins. There are over 1,000 members of this family that respond to a wide variety of stimuli: light, odorants, hormones, neurotransmitters, peptides, extracellular calcium, etc. (1). Activated GPCRs catalyze GDP/GTP exchange on heterotrimeric G proteins, whereupon the GTP-liganded G protein ␣-subunit and the free ␤␥-dimer modulate the activity of various effectors, including adenylyl cyclase, phospholipase C, cGMP phosphodiesterase, ion channels, etc. (1). The same active receptor conformation that interacts with G proteins is phosphorylated by GPCR-kinases (GRKs) (1). Arrestins then bind to the active phosphorylated state of the receptor (2). Arrestin binding prevents further G protein interaction (apparently, by simple steric exclusion (3)), often targeting receptors to the coated pits due to the high affinity of non-visual arrestins for various components of the internalization machinery: clathrin (4), clathrin adaptor AP2 (5), and N-ethylmaleimide-sensitive fusion protein (NSF) (6). Arrestins also couple GPCRs to alternative, G protein-independent signaling pathways (1), such as activation of Src (7), c-Jun NH 2 -terminal kinase 3 (8), cRaf-1 (9), and extracellular signal-regulated kinase (7,9).
Agonist bound to the internalized arrestin⅐receptor complex likely dissociates due to low internal pH in the endosomes (1). The loss of active receptor conformation promotes arrestin dissociation, whereupon the receptor can be dephopshorylated and recycled to the plasma membrane. The release of receptorbound arrestin, kinetically limited by the stability of the arrestin⅐receptor complex, is the first step in this process. Both non-visual arrestins preferentially bind to the phosphorylated agonist-activated form of their cognate GPCRs (10 -13). However, their binding to the phosphorylated inactive form is also relatively high (11)(12)(13), suggesting that arrestin dissociation from internalized phosphoreceptor may be rate-limiting. Recently we have constructed several structurally diverse mutants of both non-visual arrestins that bind to activated receptors in a phosphorylation-independent fashion (10,11,13). In contrast to wild type (WT) arrestin2 binding to phosphoreceptor, the binding of the arrestin2(R169E) and arrestin2(3A) phosphorylation-independent mutants to unphosphorylated receptor is strictly activation-dependent. We expected that in cells these mutants would bind primarily to the unphosphorylated receptor and that upon internalization the mutant arrestin⅐receptor complex would dissociate substantially faster than the complex of WT arrestin2 with phosphoreceptor. Here we used phosphorylation-independent mutants to examine how this change in the properties of the arrestin⅐receptor complex affects receptor trafficking in cells.
Note that in this paper we use the systematic names of arrestin proteins. The synonyms of arrestin2 are ␤-arrestin and ␤-arrestin1; arrestin3 is also called ␤-arrestin2.

EXPERIMENTAL PROCEDURES
Direct arrestin-binding assay was performed with purified reconstituted ␤2-adrenergic receptor (␤2AR) and in vitro-translated radiolabeled arrestins, as described (12). Briefly, the receptor was phosphorylated with GRK2 in the presence of 50 M of isoproterenol and washed to remove the agonist. In vitro-translated tritiated arrestins (50 fmol) were mixed in 50 mM Tris-HCl, pH 7.5, 0.5 mM MgCl 2 , 50 mM potassium acetate with phosphorylated or unphosphorylated ␤2AR (100 fmol/assay) in the presence of 50 M ␤-agonist isoproterenol or 50 M ␤-antag-onist alprenolol in a final volume of 50 l. The samples were then incubated for 35 min at 30°C, immediately cooled on ice, and loaded onto 2-ml Sepharose 2B columns equilibrated with 10 mM Tris-HCl, pH 7.5, and 100 mM NaCl. The bound arrestin eluted with the receptorcontaining membranes in the void volume (between 0.5 and 1.1 ml). Nonspecific binding determined in the presence of 0.3 g liposomes (Ͻ10% of the total binding and Ͻ1% of the arrestin present in the assay) was subtracted.
In vitro receptor phosphorylation assay was performed by incubating 450 fmol of purified reconstituted ␤2AR with 400 fmol of purified GRK2 in the presence of 50 M of isoproterenol in 10 l of 20 mM Tris-HCl, pH 7.5, 2 mM EDTA, 5 mM MgCl 2 , 50 g/ml bovine serum albumin, and 10 M cold ATP ϩ [␥-32 P]ATP (final specific activity 10 -15 dpm/fmol) with or without the indicated concentrations of Escherichia coli-expressed purified WT arrestin2 or R169E mutant (10,11) for 10 min at 30°C. After both arrestin and GRK2 were added to the receptor, the reactions were started by the addition of ATP and stopped by the addition of 5 l of SDS sample buffer. The samples were subjected to SDS-PAGE, the gels were dried, exposed, ␤2AR bands were cut out, and the incorporated radioactivity was quantified in a liquid scintillation counter.
Cell Culture and Transient Transfection-HEK-tsA201 cells, a clone of human embryonic kidney (HEK) 293 cells stably expressing the simian virus 40 large T antigen, were used. The cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum at 37°C in a 5% CO 2 environment. At 60 -70% confluence in T75 flasks, the cells were co-transfected with 5 g of pcDNA3-␤2AR-GFP (14) and 6 -10 g of pcDNA3-arrestin2 or its R169E or 3A mutants. In the experiments shown in Figs. 6 and 7, the cells were also co-transfected with 10 g of pcDNA3-GRK2 (14). For the experiments shown in Fig. 4, the cells were transfected with 5 g of the indicated arrestin2 construct along with 2 g of pcDNA3-␤2AR-GFP plus 2 g of hemagglutinin (HA)-␤2AR (7) in 60-mm tissue culture dishes. LipofectAMINE (Invitrogen) was used according to the manufacturer's instructions for transfection.
Receptor Internalization and Down-regulation Assays-Cells were harvested by trypsin/EDTA 48 h post-transfection. The cells were then divided and treated in suspension in PBS supplemented with 1 mM ascorbic acid (for 1 and 2 h incubation) with or without 10 M (Ϫ)isoproterenol (Sigma). At the end of the incubation, the cells were washed three times with ice-cold PBS. For 24-h incubation the cells were treated in six well plates and harvested after the wash. Total and cell surface receptor was measured with 0.2 nM [ 125 I]pindolol (1 h; 22°C) and 15 nM [ 3 H]CGP-12177 (PerkinElmer Life Sciences) (3 h; 14°C), respectively, as described (14). Nonspecific binding was determined in the presence of 10 M (Ϫ)alprenolol (Sigma) and subtracted. Internalized receptor was calculated as the difference between total and cellsurface receptor. Cells expressing 1-2.5 pmol ␤2AR per 1 mg of total protein were used.
Western Blotting-The expression of arrestins and GRK2 in each experiment was measured by quantitative Western blotting (with corresponding purified proteins as standards), as described (11). Mouse monoclonal anti-arrestin F4C1 (15) and anti-GRK 2/3 (Upstate Biotechnology) were used as the primary antibodies, and horseradish peroxidase-conjugated goat anti-mouse (Roche Molecular Biochemicals) was used as the secondary antibody. The bands were visualized with Super Signal ECL reagent (Pierce) and quantified, as described (11,16).
Quantitative Image Analysis and Statistics-To make confocal microscopy data as reliable and quantitative as the data obtained by direct measurements of arrestin and ligand binding, for each time point in every experiment we collected 25-70 images, quantified relevant signals, and statistically analyzed the results. Images collected on the confocal microscope were analyzed using C-Imaging systems software Simple 32 (Compix Inc, Cranberry Township, PA).
To analyze intracellular antibody deposition in the presence of WT arrestin2 or R169E mutant (Fig. 4), two-color confocal images of green ␤2AR-GFP and red HA-␤2AR labeled with anti-HA rhodamine-conjugated antibody were collected and saved as two-color and red singlecolor images. The membranes of the cells were carefully outlined using two-color images (green ␤2AR-GFP and red HA-␤2AR) to improve the visibility of the membrane, particularly after longer incubations. Using red single-color images converted to grayscale, the area of red fluorescence corresponding to the receptor was thresholded based on the grayscale, and the total amount of red fluorescence (anti-HA rhodamine-conjugated antibody) and intracellular red fluorescence were measured for each cell. To measure only intracellular receptor, we excluded the receptor associated with the cell membrane by disqualifying any fluorescent object that touched the membrane outline. The results were expressed for each cell as a percentage of the total cell-associated red fluorescence localized inside the cell. The data were analyzed by twoway analysis of variance with protein (WT versus R169E) and incubation time as main factors. To determine the dependence of intracellular antibody accumulation on incubation time, separate analysis of variance for each arrestin protein with incubation time as the main factor was used. Post hoc comparison and contrasts to compare individual time points were used where appropriate. For all statistical analyses, p Ͻ 0.05 was considered significant.
Internalization and Recycling Assays-Assays using reversible surface receptor biotinylation (Fig. 5) were performed essentially as described (17,18). Briefly, HEK-tsA201 cells were transfected with a plasmid encoding ␤2AR-GFP with and without plasmids encoding WT arrestin2 and R169E mutant. 24 h post-transfection the cells were re-plated in poly-D-lysine-covered 24-well plates and serum-starved overnight. The cells were washed three times with ice-cold PBS supplemented with 1 mM CaCl 2 and 0.5 mM MgCl 2 (PCM), biotinylated on ice with 6 -10 mg/ml of Sulfo-NHS-SS-Biotin (Pierce) in PCM for 40 -50 min. The cells were washed 3 times with PCM with 1 mg/ml bovine serum albumin and incubated for 2 h at 37°C in DMEM supplemented with 1 mM ascorbic acid in the presence or absence of 10 M isoproterenol with or without 20 mM NH 4 Cl (to inhibit recycling) (25). At the end of the incubation, plates were cooled on ice and washed three times with ice-cold PCM. To quantify internalized receptor, surface biotin was removed by two 15-min cycles of reduction with ice-cold 150 mM glutathione, pH 8.75, in 150 mM NaCl, followed by neutralization for 20 min with 50 mM iodoacetamide in PCM. To determine the rate of receptor recycling, the plates after agonist incubation and wash were further incubated in DMEM with 10 M antagonist alprenolol for 10 and 20 min at 37°C and then reduced and neutralized. The cells in each well were lysed at room temperature in 0.1 ml of extraction buffer (EB) (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 10 mM lysine, 2 mM benzamidine, 1 mM phenylmethylsulfonyl fluoride) supplemented with 1% SDS. The lysates were transferred to an Eppendorf tube, diluted with 0.5 ml of EB with 1% Triton X-100, and sonicated for 10 s on ice. Cell debris were pelleted by centrifugation for 20 min at 4°C. The upper 0.5 ml were transferred to an Eppendorf tube containing 15 l of UltraLink immobilized avidin gel (Pierce) and incubated overnight with gentle rolling at 4°C (the remainder was used to quantify arrestin expression by Western). To remove unbound protein, the gel was washed twice with EB with 1% Triton X-100 and twice with EB. The gel was then resuspended in 30 l of SDS sample buffer. The samples were subjected to SDS-PAGE and quantitative Western blotting using primary anti-GFP Living Colors A.v. peptide antibody (Clontech) (1:1,000) and secondary horseradish peroxidase-conjugated goat anti-rabbit antibody (Jackson ImmunoResearch Laboratories) (1:12,500). The bands were visualized with Super Signal ECL reagent (Pierce) and quantified, as described (11,16), on Fluor-S MultiImager (Bio-Rad) using Quantity One 4.2.0 software.
For each experiment, three wells were not biotinylated (negative control), three were left unreduced after biotinylation (total biotinylated receptor, control for the efficiency of biotinylation), and three wells were reduced immediately after biotinylation (control for the efficiency of biotin removal). The amount of biotinylated ␤2AR-GFP in these samples was quantified as above. The total biotinylated receptor in each experiment was used as a reference point (100%) for the quantification of internalized receptor. Only the data from the experiments where the amount of biotinylated ␤2AR-GFP was high and the removal of the biotin was complete were used. In all experiments three wells per experimental condition were used.
In Vivo Receptor Phosphorylation-This was performed essentially as described (47). HEK 293 cells were transfected with HA-␤2AR to achieve receptor expression of 3-5 pmol/mg protein and either control pcDNA3 vector or constructs expressing WT arrestin2, arrestin2(R169E), or arrestin2(3A) mutants (Ͼ25 fmol/g of total protein, i.e. Ͼ Ͼ25-fold endogenous arrestin2). Transfected cells were replated into 12-well plates, serum-starved overnight, washed 3ϫ with 2 ml of phosphate-free DMEM (PF-DMEM), incubated in PF-DMEM for 7 min, and washed with the same medium 3ϫ with 2 ml per well. For receptor and arrestin quantification by [125]IPIN binding and Western, the cells were scraped into 50 mM Tris-Hcl, pH 8.0, 100 mM NaCl, 5 mM EGTA, 2 mM benzamidine, 1 mM phenylmethylsulfonyl fluoride, and homogenized by pipetting. For [ 32 P] labeling, the cells were incubated for 2 h in 1 ml of PF-DMEM supplemented with 0.25 mM ascorbic acid, 170 Ci of [ 32 P]phosphoric acid (PerkinElmer Life Sciences) with or without 2.5 M isoproterenol, as well as with 2.5 M isoproterenol plus 20 mM NH 4 Cl (to inhibit receptor recycling) at 37°C. Following incubation, the cells were chilled on ice, washed with ice-cold PBS, and solubilized in 0.5 ml per well of lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Igepal CA-630, 0.5% sodium deoxycholate, 0.1 mM Na 3 VO 4 , 10 mM Na 4 P 2 O 7 , 10 mM NaF, 5 mM EGTA, 3 mM benzamidine, 1 mM phenylmethylsulfonyl fluoride). Lysates were centrifuged for 20 min at 100,000 rpm (rotor TLA100, Beckman TL-100 tabletop ultracentrifuge). To reduce background, supernatants were pre-cleared by incubation with 10 l of protein G-agarose (Boehringer-Mannheim) for 3 h at 4°C. After the removal of precipitated material, the supernatants were transferred to a fresh tube and incubated for 1 h with 1 g of 3F10 high-affinity rat monoclonal anti-HA antibody (Boehringer-Mannheim) at 4°C. After the addition of 10 l of protein G-agarose beads, incubation was continued overnight (14 -15 h) at 4°C. Beads were washed twice for 20 min at 4°C with 1 ml of lysis buffer, twice with high salt buffer (lysis buffer containing 0.1% Igepal, 0.05% sodium deoxycholate, and 500 mM NaCl), and twice with the same buffer without NaCl. Immune complexes were then resuspended in 2ϫ SDS sample buffer and separated on 8% SDS-PAGE gel. After autoradiography, 32 P incorporation into receptor band was quantified on Fluor-S MultiImager (Bio-Rad) using Quantity One 4.2.0 software. The identity of the radiolabeled band with HA-␤2AR was further confirmed by Western blotting of immunoprecipitated material with HA-Tag mouse monoclonal antibody (Cell Signaling).

The Binding of Phosphorylation-independent Arrestin2 (R169E) to ␤2-adrenergic Receptor Is Strictly Activation-dependent and Inhibits Receptor Phosphorylation by GRK2-A
major conformational rearrangement of the arrestin molecule is necessary for high-affinity receptor binding (2, 19 -22). The basal (inactive) conformation of arrestin is stabilized in part by the polar core, an unusual network of buried solvent-excluded charged residues in the fulcrum of the two-domain arrestin molecule (20 -24). Receptor-attached phosphates intrude upon the polar core and upset its charge balance (20 -24). Charge reversal mutations of the main phosphate sensor in the polar core, Arg-175 in visual arrestin (16,20,23), Arg-169 in arres-tin2 (10,11), and Arg-170 in arrestin3 (13), sufficiently destabilize the polar core to allow the mutant to bind with high affinity to any activated form of the receptor, phosphorylated or not (10, 11, 13, 16, 20 -24). Although the binding of WT arres-tin2 to phosphorylated ␤2AR (P-␤2AR) is enhanced by receptor activation, its binding to inactive P-␤2AR (in the presence of an antagonist) is quite substantial (Fig. 1A). The phosphorylationindependent R169E mutant demonstrates similar high binding to active and inactive P-␤2AR. In contrast, its binding to unphosphorylated ␤2AR is strictly activation-dependent (Fig. 1B).
By virtue of high-affinity binding to unphosphorylated receptor, arrestin2(R169E) can be expected to compete with receptor kinases. Indeed, in sharp contrast to WT arrestin2, purified R169E mutant effectively reduces ␤2AR phosphorylation by GRK2 in vitro (Fig. 1C). In these experiments we used 40 nM GRK2 because physiological concentrations of GRK2 in the human brain are in the range of 16 -60 nM. 2 Because both proteins compete for the same agonist-activated receptor, a substantial molar excess of R169E mutant (2.5-25-fold) is necessary for significant inhibition of receptor phosphorylation (Fig. 1C). These functional characteristics of arrestin2(R169E) suggest that upon its overexpression in cells the phosphorylation of ␤2AR by endogenous GRK2 will be attenuated and that unphosphorylated ␤2AR will form complexes with the mutant arrestin2, which are likely to dissociate upon receptor deactivation due to agonist loss in endosomes faster than the complexes of WT arrestin2 with phosphorylated ␤2AR.

Phosphorylation-independent Arrestin2 Mutant Reduces the Proportion of Internalized Receptor and Prevents Agonist-induced ␤2AR Down-regulation in HEK 293
Cells-To test how the change in arrestin2 functional characteristics due to the R169E mutation affects the trafficking of ␤2AR, we co-transfected HEK 293 cells with ␤2AR-GFP fusion (which is a fully functional ␤2AR (14)) with WT arrestin2 and R169E mutant. Both arrestins are rapidly mobilized to the plasma membrane upon agonist stimulation and can be subsequently detected co-localized with the receptor in intracellular vesicles near the membrane (Fig. 2), suggesting that the R169E mutant supports receptor internalization. However, at later time points (e.g. 60 min in Fig. 2, A and B) we found a substantially lower proportion of ␤2AR-GFP in the endocytic vesicles in the presence of the R169E mutant than with WT arrestin2. In cells expressing the R169E mutant, most of the receptor is present either on the membrane or in small endocytic vesicles close to it, and most of the R169E mutant itself appears in the cytoplasm (Fig. 2B). In all cells examined, a virtually contiguous green outline of the plasma membrane is visible after 1-2 h of agonist incubation. In contrast, in cells expressing WT arres-tin2 most of ␤2AR-GFP and detectable amounts of arrestin2 gather in large intracellular vesicles, and a substantially smaller proportion of ␤2AR-GFP is found at or near the plasma membrane, so that the green outline is no longer contiguous and often is not visible at all.
To quantify cell surface and total ␤2AR in cells expressing either form of arrestin2, we measured the binding of the hydrophylic ligand [ 3 H]CGP-12177 and the hydrophobic membrane-penetrating ligand [ 125 I]iodopindolol, respectively (4,14). Under continuous stimulation with a saturating concen-2 E. V. Gurevich, unpublished results.
tration of the ␤2AR agonist isoproterenol in the presence of WT arrestin2, an increasing proportion of the receptor is found inside the cells (i.e. is labeled with [ 125 I]iodopindolol but not with [ 3 H]CGP-12177). In contrast, only about 10% of the receptor is internalized at any time point in the presence of the R169E mutant (Fig. 3A). In our experiments the abundance of internalized receptor correlates with the rate of receptor downregulation, in agreement with a previous report (14). In the presence of the R169E mutant there is very little loss of ␤2AR even after 24 h of incubation with the agonist (Fig. 3B). In contrast, almost half of the receptor is lost with WT arrestin2 after 24 h (Fig. 3B).
Interestingly, the effect of the R169E mutant on receptor down-regulation increases with its expression level in the range of 2.7-21.2 fmol/g of total protein (i.e. 3-25-fold molar excess over endogenous WT arrestin2 in HEK 293 cells, as determined by quantitative Western blot) (F(1,27) ϭ 36.4; p ϭ 0.0001). The effect reaches its maximum at the expression level of the mutant higher than 10 fmol/g of total protein (Ͼ10-fold molar excess over endogenous). In contrast, the level of WT arrestin2 overexpression in the same range does not affect ␤2AR down-regulation (F(1,30) ϭ 1.053; p ϭ 0.313). These data support the notion that receptor trafficking in the presence of WT arrestin2 and its phosphorylation-independent mutant is fundamentally different and that a substantial molar excess of the mutant is necessary to out-compete endogenous arrestin2 (and likely endogenous receptor kinase, cf. Fig. 1C). In a separate series of experiments with WT arrestin2 and R169E mutant overexpression we found that receptor down-regulation at all time points correlates with its transport to lysosomes (visualized using LysoTracker Red (Molecular Probes)) (not shown), in accord with previous reports (14).
Both WT Arrestin2 and the R169E Mutant Support Rapid Endocytosis of ␤2AR in HEK 293 Cells-The very low proportion of internalized ␤2AR at all time points along with minimum receptor transport to lysosomes and down-regulation in the presence of arrestin2(R169E) could be explained by an impaired ability of the mutant to support receptor internalization. Alternatively, rapid ␤2AR endocytosis (as suggested by the association of the mutant with ␤2AR-GFP on the membrane and in the endocytic vesicles shown in Fig. 2) coupled with an accelerated recycling back to the plasma membrane may account for these observations. To ascertain which is the case, we took advantage of the sensitivity of the antibodyantigen interaction to pH. We reasoned that a substantial proportion of an antibody bound to the extracellular surface of ␤2AR will dissociate during the time the receptor spends in the endosomes, thus leaving evidence of its presence there even after the receptor itself moves out of this compartment. To this end, we co-expressed ␤2AR-GFP (to follow receptor localiza-tion) and HA-tagged ␤2AR with WT arrestin2 and R169E mutant. Cells were then pre-labeled with rhodamine-conjugated anti-HA antibody. The free antibody was washed away, and the cells were challenged with isoproterenol to induce receptor internalization, which was stopped at different time points. The cells were fixed, and the anti-HA antibody was visualized either directly by its own fluorescence or after signal amplifi-FIG. 2. Both WT arrestin2 and R169E mutant are rapidly mobilized by activation of ␤2AR. HEK 293 cells expressing ␤2AR-GFP and either WT ar-restin2 (A) or R169E mutant (B) were challenged with 10 M isoproterenol for the indicated periods and fixed. The receptor was visualized by its own fluorescence (green), and arrestins were visualized by immunocytochemistry with F4C1 anti-arrestin primary (15) and rhodamine-labeled secondary antibodies (red). Arrestin-receptor co-localization appears as the yellow color. The images of typical cells shown were processed using Adobe Photoshop software. cation with a rhodamine-conjugated secondary antibody. As shown in Fig. 4, most of the antibody (red) stays on the plasma membrane in unchallenged cells, whereas upon isoproterenol stimulation it accumulates in endocytic vesicles in cells expressing WT arrestin2 and R169E mutant. The amount of anti-HA antibody detected inside the cell in the absence of an agonist does not appreciably change with time (not shown).
To quantify anti-HA antibody accumulation inside the cell, we analyzed the internalization of the red antibody at different time points upon agonist exposure (0, 10, 30, and 60 min). To this end, 48 -71 cells per group, expressing ␤2AR-GFP and HA-␤2AR at comparable levels (as judged by green and red fluorescence, respectively), were selected in random view fields. The internalized red as the percentage of total cell-associated red was quantified and analyzed. Exposure to an agonist induced rapid internalization of the red antibody in the presence of both WT arrestin2 and R169E mutant (F(3,449) ϭ 70.0, p ϭ 0.0001) After 10 min incubation, about 50% of the total red antibody was found inside the cells (Fig. 4B). At this early time point, the R169E mutant was even more efficient than WT arrestin2 in mediating internalization of the red antibody. Over the 60-min incubation period, the percentage of the red antibody internalized in cells expressing the R169E mutant was somewhat lower than in the cells expressing WT arrestin2. In cells expressing the mutant arrestin, the percentage of internalized antibody reached a maximum after 10 min of stimulation and did not change much afterward; whereas, in cells expressing WT arrestin2, the percentage of internalized anti-body increased with time up to 60 -70% at the latest time point tested. Conceivably, the antibody brought in early on in the presence of R169E is subsequently transported to lysosomes and destroyed. We did not further investigate the reason for this difference because it is quite small and cannot account for the almost total lack of receptor down-regulation seen in the presence of the R169E mutant. These data indicate that the receptor is effectively internalized with both WT arrestin2 and R169E mutant. Interestingly, in cells expressing phosphorylation-independent arrestin2(R169E), most of the GFP-labeled receptor (green) is seen on the plasma membrane even after 1 h-long agonist stimulation, whereas in cells expressing WT arrestin2, a substantial proportion of the green receptor colocalizes with red anti-HA antibody in endocytic vesicles. These data support the idea that the rates of ␤2AR internalization with both forms of arrestin2 are similar. Thus, the major difference lies in the rapid recycling of the receptor internalized in complex with the R169E mutant.
To ascertain that this is the case using a different approach, we employed reversible surface biotinylation (17,18). In this paradigm, surface proteins including ␤2AR-GFP are labeled with a biotin derivative containing an S-S bond between the attachment point and the biotin moiety. Receptor internalization is then induced by agonist incubation for 2 h. Reduction with glutathione (which cannot penetrate the membrane) at this point removes biotin from surface receptor, so that only internalized ␤2AR-GFP remains biotinylated (Fig. 5). When the agonist is washed away and the cells are further incubated with a saturating concentration of the antagonist alprenolol for 10 min prior to reduction, the biotin moiety is also removed from ␤2AR-GFP that was recycled back to the plasma membrane, so that a smaller fraction of the receptor (inversely proportional to the rate of recycling) remains biotinylated (Fig.  5). The cells are then lysed and biotinylated ␤2AR-GFP is bound to immobilized avidin beads, washed, and subsequently eluted with SDS sample buffer. Bound biotinylated ␤2AR-GFP is then visualized and quantified by Western using anti-GFP antibody (Fig. 5). In control cells 35 Ϯ 4% of ␤2AR-GFP is internalized after 2 h of incubation with 10 M isoproterenol. Overexpression of WT arrestin2 increases the fraction of internalized receptor from 35 Ϯ 4% to 60 Ϯ 7%, (p Ͻ 0.01, as compared with control), whereas overexpression of R169E mutant significantly decreases it (to 8 Ϯ 5%; p Ͻ 0.01, as compared with control and WT arrestin2-expressing cells), in good agreement with radioligand binding data (compare Figs. 3A and 5). Interestingly, when endosome acidification and receptor recycling was prevented by 20 mM NH 4 Cl (25), essentially the same fraction of ␤2AR-GFP (67-75%) was internalized in all cases (Fig. 5), corroborating the idea that WT and mutant forms of arrestin2 equally support receptor internalization and suggesting again that the low proportion of ␤2AR-GFP found inside the cell at any given moment in the presence of R169E is due to its accelerated recycling.
To compare the rates of recycling directly, after the same 2 h incubation with isoproterenol to achieve receptor internalization, we removed the agonist and NH 4 Cl and allowed the cells to recycle internalized receptor for 10 min in the presence of 10 M antagonist alprenolol prior to the removal of surface biotin. Note that only the percentage of the receptor remaining inside the cell can be measured (Fig. 5). Thus, the difference between the amount of internalized (biotinylated) receptor after agonist incubation with and without subsequent antagonist incubation represents recycled receptor. Therefore, we calculated the percent of the recycled receptor as the ratio of this difference to the total amount of the receptor internalized. We found that the same proportion of internalized receptor (61-63%) is recycled in 10 min in control cells and in cells overexpressing WT arrestin2, even though the amount of internalized receptor was significantly higher in the latter case (Fig. 5). A somewhat smaller proportion of the large amounts of receptor internalized in the presence of NH 4 Cl is recycled in 10 min. Again, it is the same (39 -41%) in control cells and cells overexpressing WT arrestin2. In sharp contrast, over 90% of the receptor internalized in complex with arrestin2(R169E) mutant was recycled, even in cells that were forced to internalize 75% of the receptor by endosome alkalinization (Fig. 5). These results clearly demonstrate that the rate of ␤2AR-GFP recycling depends on the nature of the arrestin it is internalized with and that phosphorylation-independent arrestin2 mutant ensures unusually rapid recycling of the receptor.
To summarize, three independent lines of evidence suggest that arrestin2(R169E) promotes ␤2AR internalization as effectively as wild type arrestin2. We observed rapid mobilization of R169E mutant to the receptor in response to agonist (Fig. 2), rapid accumulation of anti-receptor antibody in endosomes in the presence of R169E (Fig. 4), as well as high receptor internalization in the presence of R169E mutant when recycling is blocked by NH 4 Cl (Fig. 5). These data are consistent with recent in vitro findings that arrestin2(R169E) interacts with clathrin and AP2 with even higher affinity than WT arrestin2 (26). Thus, very fast recycling of ␤2AR-GFP internalized in complex with R169E mutant (Fig. 5) is the main reason for the low levels of internalized ␤2AR in R169E-expressing cells at all time points during agonist incubation (Fig. 3).
Overexpression of GRK2 Rescues ␤2AR Down-regulation in the Presence of the Phosphorylation-independent ␤-arrestin Mutant-The main mechanistic difference between WT arrestin2 and the R169E mutant is that the latter binds to unphosphorylated ␤2AR. We hypothesized that the observed difference in receptor down-regulation is the result of the formation of an arrestin complex with unphosphorylated receptor, which is highly sensitive to receptor deactivation (compare the binding of R169E mutant to inactive ␤2AR and agonist-activated ␤2AR* on Fig. 1B). If that is true, then enhanced receptor phosphorylation should abolish the difference (compare the binding of both WT arrestin2 and R169E mutant to inactive P-␤2AR and agonist-activated P-␤2AR* on Fig. 1). To test this hypothesis, we overexpressed GRK2 along with WT and mutant arrestin2, challenged the cells with isoproterenol, and measured total ␤2AR using [ 125 I]iodopindolol. In these experiments, GRK2 expression was 8 -24 fmol/g of total protein (as determined by quantitative Western blot), i.e. the concentration of the kinase in these cells was equal to or higher than that of arrestin to ensure that the mutant cannot effectively compete with the kinase for ␤2AR* (cf. Fig. 1). As before (Fig. 3B), without GRK2 we observed a significant time-dependent receptor loss with WT arrestin2 and very little down-regulation with R169E (Fig. 6). However, in the presence of overexpressed GRK2 there is a substantial ␤2AR loss with both forms of arrestin2 (Fig. 6). GRK2 overexpression facilitates down-regulation in both cases (F(1,27) ϭ 47.7; p ϭ 0.0001). The effect is most dramatic after 24 h of incubation with agonist, particu- FIG. 5. Accelerated recycling of ␤2AR-GFP internalized in the presence of arrestin2(R169E) mutant. Surface receptor was reversibly biotinylated in HEK 293 cells overexpressing ␤2AR-GFP (1-2 pmol/mg protein) alone (C) or co-expressing receptor with 10 -14 fmol/g protein of WT arrestin2 (WT) or arrestin2(R169E) mutant (R169E), as described under "Experimental Procedures." The receptor was allowed to internalize for 2 h at 37°C in the presence of 10 M ␤-agonist isoproterenol (ISO, 2 h) or 10 M isoproterenol with 20 mM NH 4 Cl (ISO, NH 4 Cl, 2 h) to inhibit recycling. In parallel plates, after the internalization with or without NH 4 Cl, the agonist was removed and the cells were allowed to recycle internalized ␤2AR-GFP for 10 min at 37°C in the presence of 10 M ␤-antagonist alprenolol (ISO, 2 h, ALP, 10 min, and ISO, NH 4 Cl, 2 h, ALP, 10 min, respectively). Cells incubated in the absence of ligands served as controls (Control). Surface biotin was then removed, the cells were lysed, and biotinylated proteins were isolated on immobilized avidin resin, eluted with SDS sample buffer, subjected to SDS-PAGE, transferred to polyvinylidene difluoride membrane, and visualized using anti-GFP antibody (top panel). The bands were quantified, and the amount of biotinylated ␤2AR-GFP was expressed as a percentage of the total biotinylated ␤2AR-GFP. Means Ϯ S.D. of 4 independent experiments each performed in triplicate are shown in the lower three panels. *, p Ͻ 0.01, as compared with WT; a, p Ͻ 0.01, as compared with C; b, p Ͻ 0.05, as compared with C. larly in cells expressing the R169E mutant, where receptor loss increases from about 14% without GRK2 to 47% in the presence of GRK2 (F(1,16) ϭ 19.1, p ϭ 0.0005). In fact, there is no significant difference in the degree of receptor loss in the presence of the two forms of arrestin2 after 24 h stimulation when GRK2 is simultaneously overexpressed (F(1,16) ϭ 2.3, p ϭ 0.15), suggesting that ␤2AR internalized in complex with arrestin2(R169E) can be successfully targeted to lysosomes and degraded. These results rule out any unanticipated differences in the ability of the two forms of arrestin2 to interact with the internalization machinery that could have affected receptor trafficking.
The Effects of a Structurally Distinct Phosphorylation-independent Arrestin2(3A) Mutant Mimic the Effects of R169E-To ascertain that phosphorylation-independent binding of R169E to the receptor rather than some unanticipated peculiarity of this particular mutant is indeed the cause of the observed change in receptor trafficking, we tested another structurally distinct arrestin2 mutant with a similar phenotype (13) (Fig.  7). For these experiments, we chose arrestin2(I386A,V387A, F388A), referred to below as arrestin2(3A). In contrast to R169E, the 3A mutation does not affect the polar core and destabilizes a different intramolecular interaction holding arrestin in its basal (inactive) conformation: the three-element interaction between N-terminal ␤-strand I, ␣-helix I, and ␤-strand XX of the C-tail (13,21,22,24,27). Similar to its visual arrestin analog (13,27) and the R169E mutant (Fig. 1), arrestin2(3A) demonstrates the ability to bind to unphosphorylated receptor in a strictly activation-dependent fashion (Fig.  7A). The expression of arrestin2(3A) in HEK 293 cells at the same level as R169E in the experiments described above results in a sharp decrease in ␤2AR down-regulation upon prolonged agonist exposure, as compared with WT arrestin2 (Fig.  7B) (F(1,8) ϭ 25.33; p ϭ 0.001). Simultaneous expression of GRK2 with this mutant (at levels equal to or higher than that of arrestin) fully rescued receptor down-regulation (Fig. 7B) (F(1,16) ϭ 22.35; p ϭ 0.0002). Thus, similar selectivity profiles of the two structurally diverse phosphorylation-independent mutants (compare Fig. 1B and Fig. 7A) translate into virtually identical effects of their expression on receptor trafficking in cells (compare Fig. 6 and Fig. 7B).

Overexpression of Both Phosphorylation-independent Arres-tin2 Mutants Suppresses Agonist-induced Increase in ␤2AR
Phosphorylation in Vivo-To ascertain that cells expressing high levels of arrestin2(R169E) and arrestin2(3A) mutants actually internalize unphosphorylated ␤2AR, we compared isoproterenol-induced in vivo phosphorylation of HA-␤2AR in cells with no exogenous arrestin and in those expressing Ͼ 20 fmol/g protein of WT or phosphorylation-independent forms of arrestin2 (Fig. 8). In these experiments we used a lower concentration of the agonist isoproterenol to minimize the contribution of receptor phosphorylation by protein kinase A (1). The cells were treated with isoproterenol for two hours to maximize internalization (cf. Figs. 3 and 5). In parallel, the cells were also treated with agonist in the presence of 20 mM NH 4 Cl to make sure that the great majority of the receptor is internalized (cf. Fig. 5), so that the overall HA-␤2AR phosphorylation level measured by this method primarily reflects that of the internalized receptor. We found that isoproterenol treatment significantly increases 32 P incorporation into HA-␤2AR without ar-restin2 overexpression (i.e. with endogenous WT arrestin) and upon overexpression of WT arrestin2 (Fig. 8, B and C). The increase in receptor phosphorylation remains essentially the same when receptor recycling is inhibited by 20 mM NH 4 Cl under continuous stimulation. In sharp contrast, virtually no isoproterenol-induced increase in HA-␤2AR phosphorylation was observed in cells expressing either of the phosphorylationindependent forms (Fig. 8). In the presence of R169E or 3A mutants the level of receptor phosphorylation did not increase even in the presence of isoproterenol and 20 mM NH 4 Cl (Fig. 8), conditions that result in the internalization of over 70% of ␤2AR (Fig. 5). We found no statistically significant differences in isoproterenol-induced HA-␤2AR phosphorylation between cells expressing receptor only and receptor ϩ WT arrestin2, suggesting that HA-␤2AR phosphorylation depends on the functional properties of the arrestin present, but not on its expression level per se. No difference in receptor phosphorylation was found between cells expressing arrestin2(R169E) and arrestin2(3A), indicating that their common ability to bind to an active receptor in a phosphorylation-independent fashion is responsible for the suppression of agonist-induced receptor phosphorylation in living cells. These data indicate that in the presence of high levels of phosphorylation-independent mutant forms of arrestin2, ␤2AR is predominantly internalized in the unphosphorylated state, which apparently results in its accelerated recycling (Fig. 5) and minimal down-regulation (Figs. 3, 6, and 7).
Collectively, these data strongly suggest that the ability of the R169E and 3A mutants to bind agonist-activated unphosphorylated ␤2AR is responsible for the dramatically different fate of the internalized receptor observed in our experiments. The binding of these mutants to ␤2AR is strictly dependent on receptor activation (compare the binding of both mutants to inactive ␤2AR and agonist-activated ␤2AR* on Figs. 1B and 7A). We believe that the consequent higher sensitivity of R169E⅐␤2AR and 3A⅐␤2AR complexes to receptor deactivation (ligand dissociation in endosomes), as compared with the WT arrestin2⅐P-␤2AR complex (Figs. 1 and 7), is likely to play an important role in the change in receptor flow.

DISCUSSION
Receptor sorting in endocytic pathways is one of the fundamental issues in cell biology. This problem has two distinct aspects. First, the cell can simultaneously direct different receptors to different recycling and/or degradative pathways (28 -32). This appears to be accomplished via selective interaction of some receptors with certain components of the sorting machinery (reviewed in Ref. 1). Second, cells apparently change the trafficking and ultimate fate of the same receptor under different conditions. Here we are attempting to address this latter aspect using extensively characterized ␤2AR and phosphorylation-independent arrestin2 mutants R169E and 3A that effectively desensitize this receptor (11,13) and fundamentally change the properties of the arrestin⅐receptor complex ( Figs. 1  and 7).
A wealth of experimental evidence indicates that agonistactivated GRK-phosphorylated ␤2AR is internalized via the coated pits in complex with arrestin (1,4,5,7,14,29,33,34) ( Fig. 2A). This is a relatively stable complex with high agonist affinity (10,14), in which bound arrestin appears to shield receptor-attached phosphates (35). Low internal pH in the endosomes likely induces agonist dissociation. Ensuing receptor return into the inactive state facilitates arrestin release, whereupon receptor-attached phosphates become accessible to protein phosphatases. Inhibitors of protein phosphatases prevent effective receptor recycling (25,33), suggesting that only dephosphorylated receptor is transported back to the plasma membrane. This is a dynamic process, as the receptor undergoes repetitive endocytosis and recycling in the continuous presence of an agonist (33,34).
Short-term agonist exposure usually does not lead to any appreciable receptor down-regulation, whereas persistent GPCR stimulation results in a progressive loss of the receptor due to its degradation in the lysosomes (1, 14, and references therein). An increase in the time of ␤2AR intracellular "tenure" was observed after 18 h of agonist treatment (36). However, it is not clear whether any additional sorting mechanisms are turned on by prolonged stimulation, or the apparent receptor diversion from the recycling to the degradative pathway is just a cumulative effect of the same cycling mechanisms operating for a long time.
Certain GPCRs tend to recycle rapidly, whereas others stay internalized for a long time (30,32,37). The presence of clusters of serines and threonines phosphorylated by GRKs in the C termini of several receptors has been shown to determine how long arrestin stays in the complex with the receptor and how rapidly the receptor resensitizes (32,34). Different receptors also demonstrate varying preferences for the two nonvisual arrestins (12,(37)(38)(39), which, in their turn, differentially regulate trafficking even of the same receptor (35,(37)(38)(39). Certain mutations in GPCR C termini or an exchange of these elements between GPCRs often switch their arrestin preferences and/or trafficking patterns (30, 32, 38 -40). All these experimental approaches have been extensively used to delineate the mechanisms of receptor trafficking and substantially improved our understanding of differential sorting of different GPCRs in the same cell (14,32,(37)(38)(39)(40). However, evidence that the two non-visual arrestins as well as the C-terminal regions of the receptors themselves interact with numerous other partners (reviewed in Refs. 1 and 31) is accumulating rapidly. These additional interactions are likely to be affected by the above mentioned experimental manipulations, which makes these approaches unsuitable for the elucidation of mechanisms that change the fate of the same receptor in the same cell under different conditions.
Phosphorylation-independent mutants of non-visual arrestins provide a unique tool for selective manipulation of the arrestin-receptor interface without introducing any collateral changes in the process. According to the widely accepted model of GPCR de-and re-sensitization (1, 31), internalized receptor exists in three main forms: arrestin⅐phosphoreceptor complex, free phosphoreceptor, and partially and fully dephosphorylated receptor (Fig. 9). Because high-affinity arrestin binding requires receptor multi-phosphorylation (2,24,(41)(42)(43), free phosphoreceptor is heterogeneous as the result of the variation in its original phosphorylation level and its progressive dephosphorylation. Experimental evidence suggests that only fully dephosphorylated receptor is recycled (25,33), which makes perfect sense biologically. There are two distinct types of conceivable mechanisms directing internalized receptor to the lysosomes. Free phosphoreceptor and/or receptor⅐arrestin complex may be specifically recognized and transported there. Arrestin-dependent ubiquitination of ␤2AR that enhances receptor degradation (44) is a recently described example of a mechanism of this type. Alternatively, there may be a steadystate "nonspecific" transport of proteins from endosomes to lysosomes, the net result of which (on the background of selective recycling of dephosphorylated receptor) is "preferential" trafficking of phosphoreceptor and/or arrestin⅐receptor complex to the lysosomes by virtue of pure kinetics. In our view, the simplicity of the latter mechanism and its obvious applicability to the majority of GPCRs makes it rather attractive (Fig. 9). Obviously, receptor-and/or functional state-specific and purely kinetic mechanisms are by no means mutually exclusive. Most likely, an intricate interplay of several mechanisms of different types functioning side by side underlies the remarkable variety of GPCR trafficking patterns observed experimentally in various cells.
The introduction of a phosphorylation-independent arrestin mutant into cells can be expected to change the rates of several steps in the sequence of events. First, the majority of ␤2AR in complex with mutant arrestin will be unphosphorylated for two reasons: 1) its binding does not depend on receptor phosphorylation (Figs. 1B and 7A), and 2) it competes with GRK2 (Fig.  1C). Accordingly, we found that overexpression of phosphorylation-independent forms of arrestin2 (either R169E or 3A) precludes agonist-induced increase in ␤2AR phosphorylation in living cells even under conditions where the majority of the receptor is internalized (Fig. 8). Second, the life span of the receptor⅐arrestin complex is likely to decrease, because the loss of active receptor conformation dramatically reduces the affinity of the mutant for unphosphorylated ␤2AR (Figs. 1B and 7A), whereas deactivation of phosphorylated ␤2AR has only a moderate effect on WT arrestin2 binding (Fig. 1A). Third, upon arrestin dissociation most of the receptor will emerge in unphosphorylated (i.e. recycling-competent) form. Thus, in the pool of internalized ␤2AR the proportion of receptor in complex with arrestin and that of free phosphoreceptor will be sharply reduced, whereas the proportion of unphosphorylated receptor ready to be transported to the plasma membrane will increase dramatically (Fig. 9). The relatively short half-life of arrestin complex with unphosphorylated receptor is also likely to result in reduced receptor ubiquitination (44). In short, such a shift in the distribution of functional forms of the internalized receptor can only shorten the tenure of the receptor in the endocytic compartment, facilitate its recycling, and make its degradation less probable (Figs. 3, 6, and 7), regardless of the mechanism of receptor targeting to lysosomes that the cell actually employs.
Conversely, overexpression of GRK2 enhances receptor phosphorylation, thereby stabilizing the complex of either arrestin with receptor (Figs. 1, A and B, and 7A) and slowing down subsequent receptor dephosphorylation. Both effects prolong receptor tenure in endosomes and reduce the proportion of recycling-competent fully dephosphorylated receptor, thus increasing time-dependent receptor down-regulation in the presence of both WT and mutant arrestin2, as observed (Figs. 6 and 7).
Thus, phosphorylation-independent arrestin2 mutants bind to and induce the internalization of unphosphorylated ␤2AR (Figs. 1, 7, and 8), thereby shortening and simplifying the chain of events necessary to bring internalized receptor into the recycling-competent state (Fig. 9). This greatly reduces intracellular receptor transit time, as evidenced by the combination of its effective internalization (Figs. 4 and 5) and low proportion of internalized receptor at any given moment (Figs. 3A and 5). Indeed, unphosphorylated receptor internalized in complex with arrestin2(R169E) recycles substantially faster than phosphoreceptor internalized with WT arrestin2 (Figs. 5 and 8). An interesting net result of this accelerated cycling is minimum receptor loss even after very long agonist exposure (Figs. 3B, 6, and 7). It has been suggested that excessive receptor phosphorylation, internalization, and down-regulation significantly contribute to the development of congestive heart failure (45) and nephrogenic diabetes insipidus (46). The ability of phosphorylation-independent mutants of non-visual arrestins to "protect" receptors from phosphorylation, prolonged internalization, and down-regulation in conjunction with their demonstrated effectiveness in promoting short-term desensitization (11,13) makes these arrestins promising tools for therapeutic intervention in pathological processes of this kind.
Different elements in non-visual arrestins are involved in their binding to GPCRs and numerous non-receptor partners. FIG. 9. A model of receptor cycling in the presence of WT and phosphorylation-independent arrestin. A, multiple receptor phosphorylation is required for high-affinity binding of WT arrestin2. The arrestin⅐phosphoreceptor complex is then internalized. Low internal pH in the endosomes facilitates agonist dissociation. Consequent return of the receptor into its inactive state induces arrestin dissociation. The emerging multi-phosphorylated receptor goes through sequential dephosphorylation steps before the fully dephosphorylated, recycling-competent form is produced. B, constitutively active mutants (R169E and 3A) bind to unphosphorylated active receptor. The dissociation of the arrestin⅐receptor complex in endosomes immediately yields recycling-competent receptor, thereby reducing the probability of its transport to lysosomes and degradation.
As a result, these interactions can be independently manipulated by appropriate mutations (1, 2, 4 -13). We believe that "custom-designed" mutant arrestins with unusual functional characteristics represent a novel class of tools for studies of various facets of GPCR signaling and trafficking, as well as for gene therapy of various disorders associated with hereditary errors in these processes.