Ubiquitination of β-Arrestin Links Seven-transmembrane Receptor Endocytosis and ERK Activation*

β-Arrestin2 and its ubiquitination play crucial roles in both internalization and signaling of seven-transmembrane receptors (7TMRs). To understand the connection between ubiquitination and the endocytic and signaling functions of β-arrestin, we generated a β-arrestin2 mutant that is defective in ubiquitination (β-arrestin20K), by mutating all of the ubiquitin acceptor lysines to arginines and compared its properties with the wild type and a stably ubiquitinated β-arrestin2-ubiquitin (Ub) chimera. In vitro translated β-arrestin2 and β-arrestin20K displayed equivalent binding to recombinant β2-adrenergic receptor (β2AR) reconstituted in vesicles, whereas β-arrestin2-Ub bound ∼4-fold more. In cellular coimmunoprecipitation assays, β-arrestin20K bound nonreceptor partners, such as AP-2 and c-Raf and scaffolded phosphorylated ERK robustly but displayed weak binding to clathrin. Moreover, β-arrestin20K was recruited only transiently to activated receptors at the membrane, did not enhance receptor internalization, and decreased the amount of phosphorylated ERK assimilated into isolated β2AR complexes. Although the wild type β-arrestin2 formed ERK signaling complexes with the β2AR at the membrane, a stably ubiquitinated β-arrestin2-Ub chimera not only stabilized the ERK signalosomes but also led to their endosomal targeting. Interestingly, in cellular fractionation assays, the ubiquitination state of β-arrestin2 favors its distribution in membrane fractions, suggesting that ubiquitination increases the propensity of β-arrestin for membrane association. Our findings suggest that although β-arrestin ubiquitination is dispensable for β-arrestin cytosol to membrane translocation and its “constitutive” interactions with some cytosolic proteins, it nevertheless is a prerequisite both for the formation of tight complexes with 7TMRs in vivo and for membrane compartment interactions that are crucial for downstream endocytic and signaling processes.

The multifunctional adaptor proteins ␤-arrestins (␤-arres-tin1 and -2) were originally identified as desensitizing molecules that prevent the coupling between seven-transmembrane receptors (7TMRs) 3 and G proteins (1)(2)(3). More recently, however, it was found that ␤-arrestin binding to receptors not only stops G protein-mediated second messenger signaling but also engages several novel signaling pathways, including mitogenactivated protein kinase (MAPK) cascades (4,5). Furthermore, ␤-arrestins have also been shown to bind and regulate cell surface receptors other than 7TMRs, and their signaling has been implicated in regulating the actin cytoskeleton, chemotaxis, antiapoptosis, and metastasis (6).
␤-Arrestins serve as endocytic adaptors that bind clathrin and adaptin protein subunit 2 (AP-2) and facilitate receptor internalization via clathrin-coated vesicles (7)(8)(9). The differing affinity and trafficking patterns of GFP-␤-arrestins induced by several 7TMRs have led to the classification of receptors into two groups, Class A and Class B (10). Class A receptors (e.g. ␤ 2 -adrenergic, ␣ 1b -adrenergic, -opioid, endothelin 1A, and dopamine D1A receptors) show higher affinity for ␤-arrestin2 than ␤-arrestin1 and recruit GFP-␤arrestins only to the plasma membrane. Class B receptors (e.g. vasopressin V2, angiotensin AT1a, neurotensin1, thyrotropin-releasing hormone, and neurokinin NK-1 receptors) bind to both ␤-arrestin1 and -2 with equal affinity and cotraffic and colocalize with GFP-␤-arrestin in endocytic vesicles. Thus, complexes formed between ␤-arrestin and Class A receptors are transient and exist only at the membrane, whereas those formed between ␤-arrestin and Class B receptors are stable and persist after receptor endocytosis (10). These differential patterns of ␤-arrestin2 recruitment correlate with the amplitude of ␤-arrestin-bound phosphorylated ERK1/2 (pERK). Class B receptors, such as the angiotensin 1a and the V2 vasopressin receptors activate a ␤-arrestin-bound pool of ERK more persistently than Class A receptors, such as the ␤ 2 -adrenergic receptor (␤ 2 AR) and the ␣ 1b -adrenergic receptor (11).
␤-Arrestins also become ubiquitinated (attachment of ubiquitin (Ub) on lysine residues) upon agonist stimulation of various 7TMRs. Upon ␤ 2 AR stimulation, Mdm2 (mouse double minute2), a RING (really interesting new gene) type E3 ligase, ubiquitinates ␤-arrestin2, and this modification is required for rapid internalization of the receptor (12). The pattern of ␤-arrestin ubiquitination correlates with the stability of receptor-␤arrestin interaction (i.e. transient interaction (Class A) is associated with transient ubiquitination, and persistent interaction (Class B) is associated with sustained ubiquitination) (13,14). Exchanging the carboxyl-terminal amino acid residues of these two types of receptors reverses the patterns of ␤-arrestin trafficking as well as the time course of ubiquitination and the extent of ␤-arrestin-bound ERK activation (11,13,15). Additionally, translational fusion of ubiquitin to the C terminus of ␤-arrestin (␤-arrestin2-Ub) leads to its cotrafficking and colocalization with the ␤ 2 AR (Class A) in endocytic vesicles, thus mimicking a Class B trafficking pattern (13).
Interestingly, specific lysine residues are targeted for modification in response to agonist stimulation of a particular 7TMR. For example, angiotensin 1a receptor (AT1aR)-dependent sustained ␤-arrestin ubiquitination occurs primarily at lysines 11 and 12 in ␤-arrestin2 (16). Mutation of these lysines to arginine residues leads to the reversal of angiotensin II-stimulated ␤-arrestin ubiquitination from a sustained to a transient pattern, with a corresponding reversal of AT1aR-␤-arrestin binding from stable endosome-localized complexes to transiently associated complexes seen only at the plasma membrane.
In an attempt to understand the role of ubiquitination in the regulation of the endocytic and signaling functions of ␤-arrestin, we generated a ␤-arrestin2 mutant (␤-arrestin2 0K ) that is defective in ubiquitination by mutating all of the ubiquitin acceptor lysines in ␤-arrestin2 to arginines and compared it with the wild type and a stably ubiquitinated form in its ability to interact with 7TMRs and nonreceptor partners as well as its capability to facilitate receptor internalization and signaling.

EXPERIMENTAL PROCEDURES
Cell Lines, Reagents, and Plasmids-COS-7 and HEK-293 cells were obtained from the American Type Culture Collection. COS-7 cells were maintained in Dulbecco's modified Eagle's medium (Sigma) supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin and transiently transfected with Lipofectamine 2000 reagent (Invitrogen). HEK-293 cells were maintained in minimal essential medium supplemented with fetal bovine serum and transiently transfected with FuGene 6 reagent (Roche Applied Science). M2 anti-FLAG affinity-agarose beads, isoproterenol, arginine-vasopressin peptide, anti-FLAG M1 and M2 antibodies, fluorescein isothiocyanate-anti-mouse secondary IgG, and N-ethylmaleimide were from Sigma. Ubiquitin antibody FK2 was from Biomol. Monoclonal antibody 12CA5 to HA epitope was from Roche Applied Science. Alexa 594 and Alexa 633, conjugated secondary antibodies, were from Invitrogen. Horseradish peroxidase-conjugated secondary antibodies were from GE/Amersham Biosciences. Detection of active ERK was with a rabbit polyclonal anti-phospho-p44/42 MAPK (1:2000 for Western blot and 1:200 for immunostaining; Cell Signaling Technol-ogy). Total ERK was detected with anti-MAPK1/2 (1:10,000 dilution for Western blots; Millipore). A1CT, a rabbit polyclonal antibody to the ␤-arrestin1 C terminus generated in the Lefkowitz laboratory was used to detect ␤-arrestin isoforms.
Five rounds of mutations accomplished the construction of ␤-arrestin2 0K , each mutagenesis step targeting 5-7 lysine residues. We used the QuikChange multisite-directed mutagenesis kit (Stratagene) and followed the manufacturer's instructions for the design of oligonucleotides and PCR protocols. The DNA fragment encoding ␤-arrestin2 0K was later cloned into pEGFP-N1 to yield ␤-arrestin2 0K -GFP. All DNA constructs were verified by sequencing. HA-␤ 2 AR plasmid was a gift from Dr. Neil Freedman (Duke University); HA-V2R plasmid was provided by Dr. Marc Caron (Duke University). Myc-c-Raf and RFP-ERK2 have been previously reported (18).
To achieve equivalent expression of ␤-arrestin2 WT and lysine mutants (Fig. 1), we transfected cells on a 100-mm dish with 1 g of DNA for the WT and Ϫ7K, 3 g of DNA for Ϫ14K, Ϫ19K, and Ϫ26K, and 2.5 g for Ϫ31K. For the GFP-tagged plasmids (Fig. 5), 1 g was used for the WT and Ϫ7K, and 2 g was used for the rest.
To study receptor binding, the in vitro translated [ 35 S]␤-arrestins were incubated in 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 2 mM EDTA with 14.6 pmol (0.7 g) of ␤ 2 AR reconstituted in phospholipid vesicles at room temperature for 1 h. Purified GRK2 (0.5 g), 80 M ATP, 50 M isoproterenol, or 50 M propranolol were added to the reaction mixture where indicated. After the incubation period, an aliquot of the reaction was set aside to determine input levels of [ 35 S]␤-arrestins, and the remaining samples were diluted with ice-cold buffer and centrifuged at 85,000 rpm for 30 min with a bench top Optima TLX ultracentrifuge. After ultracentrifugation, the supernatants were removed, and the pellets were washed with 0.5 ml of 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, and 2 mM EDTA. The samples were centrifuged again, and the wash was repeated five times. Finally, 30 l of SDS-PAGE buffer were added to each sample, and proteins were separated by 4 -20% gel. The gels were dried, and the amounts of ␤-arrestins bound to the ␤ 2 AR were determined by autoradiography. Control experiments were performed by the same experimental procedure, except that empty vesicles were used in the place of receptorcontaining vesicles. Bands were quantified by densitometry, and the amount of each ␤-arrestin was normalized to its input levels.
Confocal Microscopy-HEK-293 cells have a favorable morphology, such that sections of cytoplasm and nucleus can be simultaneously imaged; hence, they were used in these experiments. HEK-293 cells on 10-cm dishes were transiently transfected with HA-␤ 2 AR along with ␤-arrestin2-GFP, GFP-␤-arrestin2-Ub, or ␤arrestin2 0K -GFP. Twenty-four hours post-transfection, cells were plated on collagencoated 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 PBS containing calcium and magnesium. Fixed cells were permeabilized with 0.01% Triton in PBS containing 2% bovine serum albumin for 60 min and incubated at room temperature with appropriate primary antibody. The secondary antibody incubations were done for 1 h, followed by repeated washes using PBS. Confocal images were obtained on a Zeiss LSM510 laser-scanning microscope using multitrack sequential excitation (488, 568, and 633 nm) and emission (515-540 nm, GFP; 585-615 nm, Texas Red) filter sets. Live cell GFP images were acquired using a heated (37°C) microscope stage and collected sequentially using single line excitation (488 nm).
Receptor Internalization-FLAG or HA epitope-tagged receptors expressed in HEK-293 cells in 12-well dishes were incubated with or without agonist for 30 min in serum-free medium at 37°C. Cell surface receptors were labeled with M1 FLAG or 12CA5 monoclonal antibody and with fluorescein isothiocyanate-conjugated goat antibody to mouse IgG as a secondary antibody. Receptor internalization was quantified as loss of cell surface receptors as measured by fluorescence-assisted cell sorting (Duke University flow cytometry facility).
Subcellular Fractionation-Monolayers of COS-7 cells transfected with ␤-arrestin2 or ␤-arrestin2-Ub plasmids were gently scraped and collected in PBS containing protease inhibitors and 40 mM NaCl, subjected to two freeze-thaw cycles for lysis. Samples were centrifuged at 800 ϫ g for 5 min to precipitate unlysed cells. The resulting supernatant was centrifuged at 100,000 ϫ g to separate soluble and membrane components. 40 g of each fraction was separated on SDS gels and analyzed by Western blotting.

A Ubiquitin Minus ␤-Arrestin2
Mutant-To obtain a ␤-ar-restin2 mutant that is not ubiquitinated upon 7TMR stimulation, we made conservative changes of groups of lysines to arginines, overexpressed FLAG-tagged mutants in COS-7 cells, and tested the ␤-arrestin precipitates for the ubiquitination signal induced by 1-min isoproterenol stimulation (Fig. 1, A and B). Surprisingly, elimination of such a signal required replacement of all 31 lysine residues of ␤-arrestin2 ␤-Arrestin Ubiquitination and 7TMR Signalosomes OCTOBER 5, 2007 • VOLUME 282 • NUMBER 40 Fig. 1B). When a FLAG epitope-tagged ␤-arrestin2 0K is overexpressed in HEK-293 cells, no ubiquitination smear is detected upon isoproterenol stimulation (Fig. 1C).
Although these experiments indicate that ␤-arrestin2 0K can be expressed as a properly folded protein that is isolated and detected by the epitope tag, a concern still remains whether ␤-arrestin2 0K , despite its 31 lysine to arginine changes, is a bona fide form of ␤-arrestin.
To test whether the basic folding and binding properties of ␤-arrestin2 0K are retained, we compared the binding of in vitro translated ␤-arrestin2 and ␤-arrestin2 0K to purified recombinant ␤ 2 AR reconstituted in vesicles. We also tested ␤-arres-tin2-Ub for receptor binding under the same conditions. In these in vitro assays, both WT and ␤-arrestin2 0K represent nonubiquitinated forms, and only the ␤-arrestin2-Ub chimera constitutes the ubiquitinated form. As shown in Fig. 2, A and B, ␤-arrestin2 0K bound to the ␤ 2 AR to the same extent as ␤-arres-tin2. However, the presence of a single ubiquitin moiety increased the binding by 4-fold (Fig. 2, A and B). These experiments suggest that although both nonubiquitinated forms of ␤-arrestin2 (i.e. WT and 0K) are equipotent for ␤ 2 AR binding, there is more binding between the ␤ 2 AR and the ubiquitinated form (i.e. ␤-arrestin2-Ub). When binding was performed in the presence of isoproterenol, a small increase was observed for all three ␤-arrestin forms (data not shown). We hypothesized that reconstituted ␤ 2 AR was already in an activated conformation due to the presence of zinterol in purification buffers. If so, inclusion of an antagonist could alter the observed binding. When ␤-arrestinreceptor complex formation was tested in the presence of propranolol, we found a dramatic decrease in binding for all three ␤-arrestin forms (Fig. 2, A and B), suggesting that propranolol destabilizes but does not eliminate receptor-␤-arrestin binding in these experiments. Moreover, when reconstituted receptor samples were probed with a ␤ 2 AR-specific phosphoserine antibody (serines 355 and 356), a small amount of phosphorylation was detected (Fig. 2C, top). The addition of GRK2 leads to an increase in the phosphorylation signal, and isoproterenol augments it further (Fig.  2C). We observed a comparable increase in binding above basal conditions for all three ␤-arrestin forms upon GRK2 phosphorylation and isoproterenol treatment of the reconstituted ␤ 2 AR (Fig. 2D). Collectively, these in vitro binding assays confirm that, although ubiquitinated ␤-arrestin2 forms a tight complex with the ␤ 2 AR, nonubiquitinated ␤-arrestin2 can bind reconstituted ␤ 2 AR and that the protein-protein interaction domain(s) between the receptor and ␤-arrestin2 0K is mostly unperturbed.
We next transfected HEK-293 cells stably expressing the ␤ 2 AR with either ␤-arrestin2-GFP, GFP-␤-arrestin2-Ub (stable ubiquitination), or ␤-arrestin2 0K -GFP (no ubiquitination) and examined the translocation patterns induced by isoproterenol. All three ␤-arrestin variants are uniformly distributed in the cytosol prior to agonist treatment (Fig. 4, A-C). Within 1 min of agonist stimulation, both WT and GFP-␤-arrestin2-Ub are recruited to the cell membrane and form distinct puncta, and at 30 min, GFP-␤-arrestin2-Ub is recruited to endosomal vesicles (Fig. 4B), whereas the WT remains at the membrane (Fig. 4A). As previously shown, with a Class A receptor, a stably ubiquitinated ␤-arrestin traffics into endosomes, whereas the transiently ubiquitinated WT ␤-arrestin dissociates and remains at the plasma membrane. On the other hand, agonist stimulation for 1 or 30 min does not lead to a major change in the intracellular distribution of ␤-arrestin2 0K -GFP (Fig. 4C, center panels).
To test whether the loss of translocation correlates with a loss of ubiquitination due to cumulative lysine mutations, we examined isoproterenol-induced recruitment of all of the mutants shown in Fig. 1, A and B. When the GFP-tagged version of each mutant was coexpressed with HA-␤ 2 AR in HEK-293 cells, we observed normal cytosolic expression under basal conditions for all of the ␤-arres-tin2 variants (supplemental Fig. 1). However, upon 1 min of isoproterenol stimulation, a decrease in the level of recruitment was observed, correlating with the ubiquitination status of ␤-arrestin2 (Fig. 5, middle panels, and Fig. 1B). Some amount of recruitment remains even when 26 lysine residues are altered. The only mutant that is totally defective in translocation is  35 S-labeled ␤-arrestin2, ␤-arrestin2 0K , or ␤-arrestin2-Ub was incubated with either empty phospholipid vesicles or vesicles containing ␤ 2 AR. The vesicles were then precipitated by repeated centrifugation and wash cycles. The final pellet was solubilized and separated on SDS gels, and the bound ␤-arrestins were detected by autoradiography. In the indicated lanes, propranolol (50 M) was included in the reaction mixture. B, results from three separate binding experiments were quantified and shown as bar graphs. **, p Ͻ 0.001, ␤-arrestin2 or ␤-arrestin2 0K versus ␤-arrestin2-Ub, one-way analysis of variance, and Tukey's multiple comparison test; ␤-arrestin2 versus ␤-arrestin2 0K , no significant difference. C, reconstituted ␤ 2 AR samples were probed with a phosphoserine antibody (serines 355 and 356 within the ␤ 2 AR carboxyl tail) in the upper panel and with a ␤ 2 AR antibody (H-20; Santa Cruz Biotechnology, Inc., Santa Cruz, CA) in the lower panel. D, receptor-␤-arrestin binding was tested as in A. GRK2 alone or GRK2 and isoproterenol (ISO; 50 M) were added to the reaction as indicated. The bar graph is a summarization of two independent experiments performed in duplicate. For each ␤-arrestin form, basal binding in the absence of GRK2 and isoproterenol was assigned as 1. IB, immunoblot; EV, empty vesicles. OCTOBER 5, 2007 • VOLUME 282 • NUMBER 40

JOURNAL OF BIOLOGICAL CHEMISTRY 29553
␤-arrestin2 0K , where no ubiquitination sites remain. These data suggest a strong correlation between ␤-arrestin ubiquitination status and its ability to bind activated receptors at the plasma membrane.
The above experiments also suggest that eliminating ␤-arrestin ubiquitination decreases its binding affinity for receptors in vivo, and hence only unstable receptor-␤-arrestin2 0K -GFP complexes arise at the cell membrane. In the case of Class A receptors, such as the ␤ 2 AR, although receptor-␤-arrestin complexes are initially formed at the plasma membrane, these complexes are not long lived. Thus, ␤-arrestin is rapidly deubiquitinated and dissociates from the receptor, and the receptor alone traffics into endosomes. This indicates that events occurring during pit/vesicle formation (e.g. ␤-arrestin deubiquitination) can influence the stability of ␤-arrestin-receptor complexes. Possibly, the lack of ubiquitin moieties on ␤-arrestin2 0K -GFP leads to its rapid disengagement from the receptor complex.
We hypothesized that ␤-arrestin2 0K binds activated receptor at the plasma membrane but that its deficiency in ubiquitina-tion results in decreased stability of the complex as the receptor moves into pits. If this were true, then blocking the internalization of receptors should result in the retention of ␤-arrestin2 0Kreceptor complexes at the plasma membrane. Indeed, when we inhibited the internalization of either the ␤ 2 AR (Fig. 6A) or the  . Recruitment of ␤-arrestin2, ␤-arrestin2-Ub, and ␤-arrestin2 0K to activated 7TMRs. HEK-293 cells were transiently transfected with HA-␤ 2 AR and either ␤-arrestin2-GFP (A), GFP-␤-arrestin2-Ub (B), or ␤-arrestin2 0K -GFP (C). Cells were starved for 1 h in serum-free media (NS) and then stimulated with isoproterenol (iso) for the indicated times, fixed, and immunostained for the ␤ 2 AR (red). Shown are the confocal images of the receptor immunofluorescence (red) and the GFP fluorescence (green). Colocalization (yellow) of the receptor with respective ␤-arrestins is indicated in the overlay. These images are from one of three experiments with identical results.

␤-Arrestin Ubiquitination and 7TMR Signalosomes
Class B V2R (Fig. 6B) by co-expressing dynamin K44A (a classical inhibitor of endocytosis (19,20)), we trapped activated receptors as well as ␤-arrestin2 0K -GFP at the membrane. These experiments clearly indicate that the deficiency in ubiquitination does not inhibit translocation of cytosolic ␤-arrestin2 0K to activated receptors at the cell membrane but rather decreases the stability of the receptor-␤-arrestin complexes that are formed.
Role of ␤-Arrestin2 Ubiquitination in Receptor Internalization-A characteristic feature of ␤-arrestin2 is its ability to augment receptor internalization upon overexpression (9). This effect is particularly striking in COS-7 cells, which express very low levels of endogenous ␤-arrestin2 (21). To characterize the ability of ␤-arrestin2 0K to promote receptor internalization, we overexpressed it together with HA-tagged ␤ 2 AR or V2R and measured the decrease in cell surface receptors after a 30-min agonist treatment. Overexpression of ␤-arrestin2 0K does not lead to any increase in receptor internalization, whereas WT ␤-arrestin2 leads to ϳ2.5-fold increase in both ␤ 2 AR and V2R internalization (Fig. 7, A and B). Similarly, overexpression of ␤-arrestin2 0K results in no change in receptor internalization in HEK-293 cells (Fig. 7, C and D). Predictably, because of the unstable interaction of ␤-arrestin2 0K with activated receptors compared with the WT, the mutant does not have any inhibitory effect on receptor internalization in both cell types. We have previously demonstrated that the stably ubiquitinated form of ␤-arrestin2 (␤-arrestin2-Ub) enhances receptor internalization compared with the WT (13). In contrast, ␤-arrestin2 0K , which is not ubiquitinated, forms unstable complexes with activated receptors and does not support internalization of either the ␤ 2 AR or the V2R.
We further tested if the above lack of effect of ␤-arrestin2 0K to enhance receptor internalization was due to altered binding to endocytic proteins, such as clathrin and AP-2. Clathrin binds ␤-arrestin directly and stoichiometrically, and ␤-arrestinclathrin binding is essential for receptor internalization via clathrin-coated vesicles (7). AP-2-␤-arrestin interaction is required for the movement of receptors to clathrin-coated pits (8). When ␤-arrestin2, ␤-arrestin2 0K , or ␤-arrestin2-Ub were immunoprecipitated from COS-7 cells transfected with HA-␤ 2 AR after 0, 1, and 10 min of isoproterenol treatment, an agonist-dependent increase in clathrin binding was observed for both WT and ␤-arrestin2-Ub but not for ␤-arrestin2 0K (Fig.  8, A and B). ␤-Arrestin2 0K displayed only a weak interaction and a decrease in binding in the presence of isoproterenol (Fig.  8, A and B). For the wild type ␤-arrestin2, we found a 5-fold increase in AP-2 binding at 1 min of agonist treatment, and this binding decreased to basal levels at 10 min (Fig. 8, A and C). A FIGURE 5. Translocation of GFP-tagged ␤-arrestin2 lysine mutants to activated ␤ 2 ARs. HEK-293 cells were transfected with HA-␤ 2 AR and the indicated ␤-arrestin2 plasmid. Cells were stimulated with isoproterenol for 1 min, fixed, and immunostained for the ␤ 2 AR. Confocal images shown represent one of three similar experiments. ␤ 2 AR detection is shown in red, ␤-arrestin fluorescence is shown in green, and overlay panels depict colocalization (yellow).  OCTOBER 5, 2007 • VOLUME 282 • NUMBER 40 similar time course of AP-2-␤-arrestin2 interaction has been previously reported (8). Surprisingly, both ␤-arrestin2-Ub and ␤-arrestin2 0K displayed robust binding to AP-2 under both basal and stimulated conditions (Fig. 8, A and C). Understandably, the weak interaction of ␤-arrestin2 0K with clathrin upon isoproterenol stimulation could be a major factor in its inability to promote receptor endocytosis. Unlike the previously reported ␤-arrestin1-(319 -418), which binds clathrin but lacks receptor interaction (22), ␤-arrestin2 0K did not act as an inhibitor of receptor internalization.

␤-Arrestin Ubiquitination and 7TMR Signalosomes
Impact of Ubiquitination on Raf and ERK Scaffolding Properties of ␤-Arrestin2-Previous studies have shown that ␤-arrestin-Raf complexes are stable, since their isolation is possible by gel filtration as well as by coimmunoprecipitation (18,23). We tested the interaction of the above three ␤-arrestin forms with the MAPK kinase kinase, c-Raf (Fig. 9A), and did not observe any differences between the WT and ␤-arrestin2 0K in their ability to bind Myc-c-Raf1. In these assays, ␤-arrestin2-Ub, however, bound more c-Raf than the WT (Fig. 9A). The amount of c-Raf in the immunoprecipitate normalized to total input levels was significantly higher with ␤-arrestin2-Ub than with the wild type, as indicated by the quantification of bands from three independent experiments (data not shown).
Previous studies have also shown that by merely coexpressing ␤-arrestin2 with an MAPK kinase kinase (such as

␤-Arrestin Ubiquitination and 7TMR Signalosomes
c-Raf or ASK-1) and a MAPK (such as ERK2 or JNK3), robust activation of MAPK could be achieved (18,24,25). This property of ␤-arrestin2 is attributed to its capacity to simultaneously bind component enzymes of a kinase cascade, thus bringing them into proximity and allowing robust phosphorylation to occur. Accordingly, cotransfection of ␤-arrestin2, c-Raf, and GFP-ERK2 could enhance precipitation of phosphorylated ERK2 with FLAG-␤-arrestin2 (18). To examine whether ␤-arrestin2 0K was capable of a similar function, we transfected COS-7 cells with WT, ␤-arrestin2 0K , or ␤-arres-tin2-Ub along with RFP-ERK2 and increasing amounts of Mycc-Raf-1. As shown in the Western blots (Fig. 9B) and the bar graphs depicting quantification of pERK in ␤-arrestin2 precip-itates (Fig. 9C), ␤-arrestin2 0K could scaffold pERK to the same extent as the WT. Interestingly, ␤-arrestin2-Ub precipitates contained 60 -80% more pERK than the WT, suggesting a greater level of kinase activation and/or a stronger interaction of ␤-arrestin2-Ub with pERK. All of the above coimmunoprecipitation data suggest that ubiquitination is not required for ␤-arrestin interaction with c-Raf and pERK and that despite the 31 lysine mutations, ␤-arrestin2 0K can interact with these ␤-arrestin partners.
Role of ␤-Arrestin2 Ubiquitination in the Formation and Subcellular Targeting of Receptor Signalosomes-We next examined the effects of ␤-arrestin2, ␤-arrestin2-Ub, and ␤-arrestin2 0K on the assembly of receptor-␤-arrestin2-ERK complexes. As depicted in Fig. 10A, a significant amount of FIGURE 10. Magnitude of ERK activity in receptor complexes correlates with ␤-arrestin ubiquitination. A, COS-7 cells were transfected with RFP-ERK2 overexpressing FLAG-␤ 2 AR (A) along with either pEGFP vector, ␤-arres-tin2-GFP, GFP-␤-arrestin2-Ub, or ␤-arrestin2 0K -GFP. After 5 min of agonist stimulation, receptors were immunoprecipitated, separated on gels, and probed for the amount of pERK content. The blots were reprobed for total ERK (second panel) followed by a second reprobe for the ␤ 2 AR (third panel). The levels of RFP-pERK2, RFP-ERK2, and the different ␤-arrestins in the cell extracts are also shown (Lysates panel). B, the bar graphs depict the quantification of pERK coprecipitated with the agonist occupied receptor. Data represent mean value Ϯ S.E. from five independent experiments. ***, ␤-arrestin2 versus ␤-arrestin2-Ub, p Ͻ 0.001; **, mock versus ␤-arrestin2 0K , p Ͻ 0.01, as determined by paired t tests.  OCTOBER 5, 2007 • VOLUME 282 • NUMBER 40 pERK was associated with ␤ 2 AR immunoprecipitates from COS-7 cells upon coexpression of ␤-arrestin2-Ub. Lesser amounts of pERK were detected upon wild type ␤-arrestin2 expression, although this amount was still higher than the pERK detected with endogenous ␤-arrestin2 in the mocktransfected samples (Fig. 10, A and B). Expression of ␤-arrestin2 0K resulted in a significant decrease in pERK in receptor complexes compared with that obtained with endogenous ␤-arrestin2 as seen in the bar graph representing the quantification of signals from five independent experiments (Fig. 10B). This decrease was not due to a decline in overall ERK activation, since the level of activation in whole cell lysates was identical in all transfection conditions.

␤-Arrestin Ubiquitination and 7TMR Signalosomes
In general, ␤-arrestin-mediated ERK signals are retained in the cytosol and are prevented from entering the nucleus. To determine if ␤-arrestin ubiquitination plays a role in the subcellular localization of agonist-stimulated pERK, we performed confocal immunofluorescence microscopy and examined the relative distribution of agonist-activated receptors, ␤-arrestins, and pERK. An antibody that specifically recognizes Thr 202 / Tyr 204 -phosphorylated ERK1/2 was employed to detect activated endogenous ERK. If ubiquitination of ␤-arrestin2 indeed plays a role in determining the spatial distribution of active ERK, then differences should be observed in the cellular distribution of pERK stimulated in the presence of ␤-arrestin2-Ub versus ␤-arrestin2 0K .
As depicted in Fig. 11A, unstimulated cells show a uniform cytosolic distribution of ␤-arrestin2-GFP (green), a membrane distribution of HA-␤ 2 AR (blue), and a negligible amount of pERK (red). When the cells were stimulated for 5 min with isoproterenol, ␤-arrestin2 redistributed to the cell membrane to colocalize with the activated receptors. A robust increase in the level of pERK was observed in both cytosol and nucleus along with a clearly demarcated pERK signal on ␤-arrestinstudded cell membranes (Fig. 11A, second row). After 30 min of isoproterenol, the ␤ 2 ARs were visualized in intracellular vesi- FIGURE 11. Subcellular distribution of 7TMR-stimulated pERK in HEK-293 cells. HEK-293 cells expressing HA-␤ 2 AR, along with either ␤-arrestin2-GFP (A), GFP-␤-arrestin2-Ub (B), or ␤-arrestin2 0K -GFP (C) were stimulated with isoproterenol for 0, 5, or 30 min or PMA for 30 min (A), fixed, permeabilized, and labeled with a rabbit polyclonal anti-phospho-p44/42 MAPK antibody followed by Alexa633-conjugated secondary antibody. Following this, receptors were labeled for HA epitope with the monoclonal antibody, 12CA5, followed by Alexa594-conjugated secondary antibody. Confocal images were collected using sequential line excitation filters (488, 568, and 633 nm) and emission filter sets at 505-550 nm for GFP detection (green), 585 nm for HA-␤ 2 AR (blue), and 650 nm for pERK (red) detection. Data represent similar results obtained from three independent experiments. NS, not stimulated.
cles. ␤-Arrestins are not localized to these vesicular structures but are retained at the cell membrane. A small percentage of receptors persist at the membrane, which most likely represent recycled and/or noninternalized receptors. After 30 min of isoproterenol treatment, a negligible amount of pERK was detected (Fig. 11A, third row). As a comparison, representative cells overexpressing both HA-␤ 2 AR and ␤-arrestin2-GFP treated with phorbol 12-myristate 13-acetate (PMA) are shown in the bottom row of Fig. 11A. PMA stimulation leads to robust activation of ERK, which is distributed in both cytoplasm and nucleus. PMA stimulation does not lead to either ␤ 2 AR internalization or ␤-arrestin2 translocation.
The results of similar experiments performed with GFP-␤-arrestin2-Ub and HA-␤ 2 ARs are shown in Fig. 11B. Under unstimulated conditions, the subcellular distributions are identical to what is observed with the WT ␤-arrestin2. Quite strikingly, at 5 min of stimulation, a distinct and robust ERK activation is observed at the cell membrane coinciding with the distinct membrane recruitment of ␤-arrestin2-Ub. Although a majority of the cells (ϳ80%) displayed such distribution at the cell membrane, some cells did have small vesicles in the vicinity of the cell membrane, which contained ␤-arrestin2-Ub, ␤ 2 AR, and pERK as shown in Fig. 11B (second row). Surprisingly, at the 5 min time point, unlike the case of WT ␤-arrestin2 expression, little active ERK was distributed in the nucleus with ␤-arres-tin2-Ub overexpression. We do not know the exact mechanism by which this occurs, but possibly, ␤-arrestin2-Ub can simultaneously promote ␤-arrestin-dependent cytosolic ERK and curb the G protein ERK pathway, leading to less nuclear ERK.
After 30 min of isoproterenol treatment, a dramatic redistribution of ␤-arrestin2-Ub, ␤ 2 AR, and pERK was seen in intracellular vesicles. These data clearly indicate that a stably ubiquitinated ␤-arrestin can remain associated with a Class A receptor (i.e. ␤ 2 AR) and target activated ERK to early endosomes, resulting in a pool of pERK complexed with internalized receptors.
In the absence of agonist, ␤-arrestin2 0K -GFP is mainly cytoplasmic, with HA-␤ 2 AR at the plasma membrane and very little active ERK (Fig. 11C, top row). After 5 min of isoproterenol stimulation, a robust activation of ERK occurs, which is seen distributed in both cytoplasmic and nuclear compartments. However, none of this active ERK is localized with ␤-arrestin2 0K . Possibly, much of this activity is G protein-mediated and is excluded from receptor complexes, since less pERK is complexed with the ␤ 2 AR in the presence of ␤-arrestin2 0K (see Fig. 10). At 30 min, levels of pERK decreased but were not abolished (Fig. 11C, bottom row). This situation contrasts with what is observed with the stably ubiquitinated ␤-arrestin2-Ub (Fig. 11B), where pERK signals are stabilized and localized on endosomal vesicles at 30 min of isoproterenol stimulation. As seen in the 30 min panels of Fig.  11C, ␤ 2 AR internalized into endosomes, which is consistent with our internalization data (Fig. 7, A-D), which indicate the inability of ␤-arrestin2 0K to inhibit receptor internalization.
We also determined the kinetics of ERK phosphorylation in HEK-293 cells expressing the ␤ 2 AR (1 pmol/mg of cellular protein) upon transfection of vector, ␤-arrestin2 WT, ␤-arrestin2 0K , or ␤-arrestin2-Ub. As shown in Fig. 12, expression of ␤-arrestin2-Ub significantly increased ERK activity at 20 min of isoproterenol treatment, ␤-arrestin2 led to a modest augmentation, and ␤-arrestin2 0K had no effect over mock conditions (Fig. 12B). Previous studies have demonstrated that later ERK activity induced by 7TMRs is actually ␤-arrestin-mediated (reviewed in Ref. 26). These results further support the idea that ␤-arrestin ubiquitination status underlies some aspects of ␤-arrestin-dependent signaling.
Ubiquitination Favors ␤-Arrestin Distribution in Membrane Compartments-The ␤-arrestin isoforms are mainly cytosolic proteins and are translocated to the plasma membrane upon 7TMR activation. Thus far, no lipid modifications in ␤-arrestins favoring macromolecular membrane interactions have been identified. One well accepted mechanism that keeps them in a membrane environment is their binding to phosphorylated domains of receptors (5). Our current and previous results indicate that ubiquitination could be an important factor HEK-293 cells stably expressing the ␤ 2 AR were transiently transfected with vector, ␤-arrestin2, ␤-arrestin2-Ub, or ␤-arrestin2 0K . Monolayers of cells in 12-well dishes were stimulated with isoproterenol (10 M) for the indicated times, and whole cell lysates were analyzed for ERK phosphorylation by Western blotting. A, bar graphs representing the quantification of pERK bands normalized to ERK levels obtained from 4 -6 experiments. Maximal signal from an individual experiment was used as 100%. ***, p Ͻ 0.001; **, p Ͻ 0.01, two-way analysis of variance, Bonferroni post-tests, all compared with vector only condition. B, representative Western blots for phospho-ERK in the top panels. The same blots were stripped (Restore stripping buffer; Pierce) probed for ERK1/2 levels (middle), and restripped and reprobed for ␤-arrestin levels (bottom).
that determines the longevity of ␤-arrestin interactions with receptors leading to colocalization on endosomal vesicles. Interestingly, when we analyzed the distribution of the ubiquitinated form of ␤-arrestin by subcellular fractionation, we found that the ubiquitination status of ␤-arrestin favors its partitioning to membrane fractions. When COS-7 cells expressing either ␤-arrestin2 or ␤-arrestin2-Ub were lysed in a detergent free low salt buffer (40 mM NaCl) and the soluble and insoluble fractions were further separated by differential centrifugation, nonubiquitinated ␤-arrestins were mainly cytosolic. Most of the exogenously expressed ␤-arrestin2 as well as YFP-␤-arrestin2 was detectable in the soluble fraction (Fig. 13). The YFP-␤-arrestin2 band in the membrane fraction with a slightly slower mobility is unreactive to ubiquitin antibodies, such as FK2, P4D1, and FK1, and its identity remains to be elucidated. On the other hand, ubiquitinated ␤-arrestin2 was distributed mostly in the insoluble membrane fractions (Fig. 13, A and B). As seen in Fig. 4B, ␤-arrestin2-Ub appears to be uniformly distributed in the cytosol in an undisturbed cell. Accordingly, the membrane fractionation of ␤-arrestin2-Ub is not due to its pres-ence in inclusion bodies but rather due to its affinity for membrane components. These results suggest that ubiquitination increases the propensity of ␤-arrestin for membrane association, thus favoring prolonged localization of ␤-arrestin in membrane microdomains. Although ubiquitination is dispensable for ␤-arrestin interactions with cytososlic partners, it may be necessary to facilitate the formation of functional 7TMR-␤-arrestin endocytic and signaling complexes in a membrane environment.
A consequence of agonist stimulation of several 7TMRs is the ubiquitination of ␤-arrestins, which is required for rapid receptor internalization (12). Ubiquitination is a post-translational modification that was originally described in the context of regulated destruction of many proteins by the proteasomal machinery (32). However, recent years have witnessed the discovery of a plethora of nonproteasomal roles of ubiquitin (33,34), and ubiquitination has been shown to play an important role in the lysosomal degradation of 7TMRs (35). Moreover, monoubiquitination of adaptor proteins, such as Eps15, and mono-/multiubiquitination of cell surface receptors are implicated in endocytosis, whereas polyubiquitination of the adaptor protein TRAF6 is suggested to be crucial for triggering NF-B signaling pathways (36 -38). ␤-Arrestin functions as both an endocytic and a signaling adaptor for 7TMRs and bears a functional analogy to proteins such as Eps15 and TRAF6. Currently, the nature of ␤-arrestin2 ubiquitination as to being poly-or mono-ubiquitination is unknown. Nevertheless, ␤-arrestin ubiquitination facilitates both receptor internalization and MAPK activation.
The kinetics of ␤-arrestin ubiquitination and deubiquitination appear to determine the stability and duration of ␤-arrestin-receptor interactions, which in turn determine its trafficking pattern (13). Our studies indicate that stable ␤-arrestin-receptor interaction leading to cotrafficking of receptors and ␤-arrestins into endosomes not only results in sustained ubiquitination but also in the enhanced activation of ERK. ␤-Arrestin ubiquitination also plays an important role in promoting receptor internalization (12,13). Are these cellular processes (namely receptor internalization, ␤-arrestin trafficking, MAPK activation, and ␤-arrestin ubiquitination) independent or related events? Could ␤-arrestin ubiquitination serve as a locus of control for these various pathways? To understand the integration of ubiquitination into the traffick- Twenty-four hours post-transfections, cells were collected in a detergent-free buffer and lysed by freeze-thawing the samples, and cytosolic and membrane fractions were separated by differential centrifugation (see "Experimental Procedures"). An equal protein amount of each type of sample separated on a gradient gel was then immunoblotted with a ␤-arrestin antibody. The respective protein bands are indicated in the panels. #, a modified form of YFP-␤-arrestin2; *, these bands are mostly due to the presence of internal methionines. B, bar graphs depict quantification of bands representing respective populations (C, cytosolic; M, membrane) of the indicated ␤-arrestin types. In each case, the cytosolic amount was arbitrarily assigned as 1. The data represent mean Ϯ S.E. from three independent experiments.
ing and signaling functions of ␤-arrestins, we sought to compare wild type ␤-arrestin2 with a nonubiquitinated form as well as a stably ubiquitinated form. To generate ␤-arrestin2 totally defective in ubiquitination, we had to replace all of the 31 lysines within ␤-arrestin2 with arginine residues. Sometimes it takes only one mutation to generate a completely misfolded protein, so it is a concern to study a protein with 31 lysinearginine changes. However, we believe that the conservative nature of the introduced changes allowed ␤-arrestin2 0K to be functional in both in vitro assays and our protein-protein interaction studies. For the most part, ␤-arrestin2 0K behaved like the wild type ␤-arrestin2, since its binding to recombinant ␤ 2 AR in vitro and to c-Raf and ERK was unchanged from the wild type. Interestingly, ␤-arrestin2 0K bound AP-2 much more robustly than the wild type, whereas its interactions with clathrin and the ␤ 2 AR were impaired in vivo. Previous studies have shown that certain arginine residues in ␤-arrestin2 (Arg 394 and Arg 396 ) are involved in AP-2 binding (39). It is possible that by introducing 31 arginines in the place of lysines in ␤-arrestin2 0K , we introduced additional AP-2 binding sites leading to a more robust interaction. Although the AP-2 binding domain in the wild type ␤-arrestin2 is exposed only after a receptorinduced conformational change, it is possible that for the ␤-arrestin2 0K , some of the arginines could present an interaction domain constitutively.
Although ␤-arrestin2 0K is capable of equivalent protein-protein interactions with the ␤ 2 AR as the nonubiquitinated wild type in vitro, in a cellular context, it shows impairment in binding, since unlike the wild type, it cannot be ubiquitinated at the proper site(s). ␤-Arrestin2 0K did not support internalization of either ␤ 2 AR or V2R, confirming that ubiquitination of ␤-arrestin is crucial for its role in promoting receptor endocytosis (Fig.  7). Additionally, this mutant was only transiently recruited to the plasma membrane upon stimulation of the ␤ 2 AR or the V2R (Fig. 6). In contrast, a ␤-arrestin-Ub chimera that remains stably ubiquitinated can enhance ␤ 2 AR internalization (13) and is stably recruited to endosomal compartments with the ␤ 2 AR (Fig. 4B). Although both ubiquitinated and nonubiquitinated forms of ␤-arrestin can form complexes with pERK ( Fig. 9), only the ubiquitinated form is capable of this function in a receptor complex (Figs. 10 and 11). In other words, ␤-arrestin ubiquitination plays a central role in stabilizing kinase activity in receptor signalosomes. However, whether ␤-arrestin ubiquitination acts as a "trigger mechanism" for activating the c-Raf-MEK1-ERK2 cascade remains to be determined. Thus far, TRAF6 autoubiquitination is the only example in which the adaptor protein ubiquitination initiates kinase signaling (38).
One important consequence of ␤-arrestin-dependent MAPK activation is the compartmentalization of the signals. Thus, AT1aR-stimulated pJNK3 scaffolded by ␤-arrestin2 is retained on endocytic vesicles (25). Similarly, AT1aR-stimulated ERK is concentrated on endosomes, which are also the destination for internalized receptor-␤-arrestin complexes (18). As evidenced by our microscopy experiments, this subcellular location of MAPKs is indeed directed by the ubiquitination status of ␤-ar-restin2 (Fig. 11, A-C). Thus, a stably ubiquitinated ␤-arrestin not only confers a Class B trafficking pattern on a Class A recep-tor but also leads to changes in the compartmentalization of pERK (Fig. 11B).
Our cell fractionation experiments indicate that ubiquitin moieties on ␤-arrestin somehow favor its membrane distribution. We currently do not know the exact mechanism by which ubiquitin chains favor membrane interactions that facilitate subsequent signalosome formation. It is possible that although the concave domains of ␤-arrestin interact with the phosphorylated domains of the receptor, ubiquitin chains elsewhere on ␤-arrestin help its retention in a membrane environment. Furthermore, the ubiquitinated domains of ␤-arrestin could favor a tighter interaction with c-Raf, allowing robust kinase activation at the membrane. This activity is then passed down through the cascade, leading to ERK phosphorylation within the ␤-arrestin scaffold.
Our data suggest that ␤-arrestin and the ERK protein bound to it have to remain at the membrane for an optimal duration before receptor signalosomes are formed. Moreover, a subsequent stable interaction of receptor and ␤-arrestin is required for sustaining and targeting this activity to subcellular compartments. Interestingly, previous studies have shown that the membrane recruitment of ␤-arrestin2 itself can lead to ERK signaling and that a 7TMR-␤-arrestin1 chimeric protein can act as a constitutive signalosome (44,45). Our findings indicate that ␤-arrestin ubiquitination controls not only receptor trafficking but the nature, stability, and subcellular localization of active ERK signals. It seems overwhelmingly likely that this regulation also extends to the nature of the ERK substrates and hence of the cellular consequences of receptor mediated activation of ERK. Seen in this light, ␤-arrestin ubiquitination may be viewed as the "glue" that holds the receptor "signalosome" together and directs the ultimate destination of its cellular journey. It will be of interest to determine which other signaling pathways may be regulated in this way.