Unraveling G Protein-coupled Receptor Endocytosis Pathways Using Real-time Monitoring of Agonist-promoted Interaction between β-Arrestins and AP-2*

The most widely studied pathway underlying agonist-promoted internalization of G protein-coupled receptors (GPCRs) involves β-arrestin and clathrin-coated pits. However, both β-arrestin- and clathrin-independent processes have also been reported. Classically, the endocytic routes are characterized using pharmacological inhibitors and various dominant negative mutants, resulting sometimes in conflicting results and interpretational difficulties. Here, taking advantage of the fact that β-arrestin binding to the β2 subunit of the clathrin adaptor AP-2 (β2-adaptin) is needed for the β-arrestin-mediated targeting of GPCRs to clathrin-coated pits, we developed a bioluminescence resonance energy transfer-based approach directly assessing the molecular steps involved in the endocytosis of GPCRs in living cells. For 10 of the 12 receptors tested, including some that were previously suggested to internalize via clathrin-independent pathways, agonist stimulation promoted β-arrestin 1 and 2 interaction with β2-adaptin, indicating a β-arrestin- and clathrin-dependent endocytic process. Detailed analyses of β-arrestin interactions with both the receptor and β2-adaptin also allowed us to demonstrate that recruitment of β-arrestins to the receptor and the ensuing conformational changes are the leading events preceding AP-2 engagement and subsequent clathrin-mediated endocytosis. Among the receptors tested, only the endothelin A and B receptors failed to promote interaction between β-arrestins and β2-adaptin. However, both receptors recruited β-arrestins upon agonist stimulation, suggesting a β-arrestin-dependent but clathrin-independent route of internalization for these two receptors. In addition to providing a new tool to dissect the molecular events involved in GPCR endocytosis, the bioluminescence resonance energy transfer-based β-arrestin/β2-adaptin interaction assay represents a novel biosensor to assess receptor activation.

G protein-coupled receptors (GPCRs) 7 are seven transmembrane domain receptors that constitute the largest family of cell surface proteins involved in signal transduction. In humans, it is estimated that GPCRs are encoded by ϳ800 distinct genes that control a variety of important physiological responses (1). Following agonist binding, GPCRs undergo conformational changes that regulate the activity of downstream effector systems to mediate various cellular responses. The extent and duration of GPCR signaling is tightly regulated by mechanisms that terminate the initial signaling and later re-establish the capacity of the receptors to respond to new agonist exposure. The removal of GPCRs from the cell surface, also known as internalization or endocytosis, plays an important role in these processes (2,3). For most GPCRs, rapid feedback desensitization is initiated by G protein-coupled receptor kinases that phosphorylate agonist-occupied GPCRs to create high affinity binding sites for ␤-arrestins, which in turn uncouple the receptor from its cognate G protein (reviewed in Ref. 4). ␤-Arrestins also target receptors for endocytosis by linking them to the endocytic machinery, including clathrin and the clathrin adaptor AP-2 (5)(6)(7)(8)(9)(10)(11). By controlling receptor recycling following endocytosis, ␤-arrestins have also been shown to regulate the rate of receptor resensitization (12). Based on their interaction with ␤-arrestins, GPCRs are divided into two main classes. Class A, which includes receptors such as the ␤2-adrenergic (␤2AR), endothelin A (ETAR), and V1a vasopressin (V1aR), interacts transiently with ␤-arrestins and can rapidly recycle back from the endosomes to the cell surface (13,14). Class B receptors, such as vasopressin V2 (V2R), CC-chemokine 5 (CCR5), prostaglandin EP4 (EP4R), and vasoactive intestinal peptide 1 (VIP1R), interact more stably with ␤-arrestins, leading to a complex that resides for extended periods of time into endosomes. Receptors tightly associated with ␤-arrestins in endosomes are only poorly recycled to the cell surface and eventually targeted for lysosomal degradation (13)(14)(15)(16)(17). The two classes of receptors also differ by their binding preference toward ␤-arrestin1 and ␤-arrestin2. Indeed, whereas class A receptors binds with greater affinity to ␤-arrestin2, class B receptors do not show preference between the two ␤-arrestins (13). More recently, it was suggested that some receptors may not be easily classified in class A or B. For example, the bradykinin B2 receptor (B2R) was found to internalize with ␤-arrestins into endosomes but can then dissociate from ␤-arrestin and efficiently recycle to the plasma membrane after agonist removal (18). The authors suggested that receptors with such hybrid features could be referred to as class C.
Although the roles of ␤-arrestins and clathrin-coated vesicles in GPCR endocytosis have been well characterized, alternative pathways involving non-coated vesicles, such as caveolae, or other non-clathrin and non-caveolae mediated routes, have also been described for several receptors (reviewed in Refs. 2,3,19). Also, the requirement for ␤-arrestin does not seem to be universal, and endocytosis of some GPCRs through either clathrin-coated vesicles or caveolae was proposed to be ␤-arrestin-independent (20). To characterize the various endocytic routes used by different receptors, several pharmacological and biochemical tools have been used. These include blockers that do not discriminate between clathrin-dependent or caveolaemediated endocytosis (e.g. hypertonic sucrose, low temperature, concanavalin-A, and dominant negative mutants of dynamin) as well as inhibitors that are believed to selectively inhibit clathrin-coated vesicle (e.g. monodansylcadaverine, chlorpromazine, as well as dominant negative mutants of ␤-arrestin and Eps-15) or caveolae-mediated endocytosis (e.g. filpin and nystatin) (3,19,21,22). Although these approaches have been useful, the interpretation of their data was sometimes difficult and generated some controversies on the exact mechanism underlying the endocytosis of certain GPCRs. For example, inhibition of endocytosis by dominant negative mutants of ␤-arrestins (␤-arrestin1(V53D), ␤-arrestin1-(319 -418), and ␤-arrestin2(⌬LIELD/F391A) (7,23,24)) or a dominant negative of the AP-2 interacting protein Eps15 (25) has often been used to suggest the involvement of a clathrin-mediated process. However, in the case of the ETAR, endocytosis through caveolae was also inhibited by a dominant negative mutant of ␤-arrestin (20). Similarly, inhibiting Eps15 has been recently found to prevent caveolae-mediated endocytosis of the epidermal growth factor receptor (26). Off-target effects of dominant negative mutants can also complicate interpretation of the results. For instance, whereas it was initially believed that ␤-arrestins were specific regulators for GPCRs, growing evidences demon-strate its role in the regulation of non-GPCRs membrane proteins endocytosis (27). Although this may reflect direct interaction of ␤-arrestins with non-GPCR proteins (28), it could also result from the interference with components of the endocytic machinery that could be shared by clathrin-and caveolae-mediated routes.
The above discussion illustrates the importance of developing new tools that will allow the direct assessment of the molecular steps involved in the endocytosis of specific GPCRs. To generate one such tool, we took advantage of the fact that, for internalization, the ␤-arrestins binding to the ␤2-adaptin subunit of the clathrin adaptor AP-2 are needed for the ␤-arrestinmediated targeting of GPCRs to clathrin-coated pits (7,8,10,29). Specifically, we designed a bioluminescence resonance energy transfer (BRET)-based assay that allows real-time monitoring of the interaction between ␤-arrestins and ␤2-adaptin as a biosensor for ␤-arrestin-promoted clathrin-mediated endocytosis. BRET is a non-radiative energy transfer that occurs between the energy donor Renilla luciferase (Rluc) and variants of the green fluorescent proteins (EYFP for BRET 1 and GFP, GFP 2 , or Tsapphire for BRET 2 ) as the energy acceptor only when the two proteins are within 100 Å from one another and situated in favorable orientation, making it an attractive assay for studying both inducible and constitutive protein-protein interactions (30,31). Thus, in response to the activation of 12 different GPCRs, previously suggested to internalize via different endocytic routes, BRET 1 was assessed between ␤-arrestin-Rluc and ␤2-adaptin-EYFP and used as an indicator of clathrin-mediated endocytosis involving ␤-arrestins. By combining the spectrally resolved BRET 1 and BRET 2 technologies, we could simultaneously assess the recruitment of ␤-arrestin to GPCR and ␤-arrestin/AP-2 interaction, which allowed us to monitor the kinetics of the two events. The data obtained herein show that some of the GPCRs previously suggested to internalize via a clathrin-independent pathway were capable of promoting ␤-arrestin interaction with AP-2. Thus, contrary to what was deduced from indirect methods, these receptors internalize via a ␤-arrestinand clathrin-dependent mechanism. Other receptors, such as ETAR and ETBR, which are capable of recruiting ␤-arrestins but were shown to internalize via caveolae, did not promote ␤-arrestin interaction with ␤2-adaptin, indicating a ␤-arrestin-dependent but clathrin-independent endocytic process. In addition to providing a new tool to dissect the molecular events involved in GPCR endocytosis, the BRET-based ␤-arrestin/␤2-adaptin interaction assay can also be used to detect constitutive receptor internalization and quantify receptor activation in a pharmacologically relevant manner.
Cell Culture, Plasmid Transfections, and Stable Cell Line Production-Unless otherwise stated, human embryonic kidney (HEK) 293T cells were cultured in high glucose DMEM supplemented with 10% fetal bovine serum, 100 units/ml penicillin-streptomycin, and 2 mM L-glutamine at 37°C in a humidified chamber at 95% air and 5% CO 2 . For transient transfec-tions in 6-well plates, 400,000 HEK293T cells were seeded and transfected the next day using Fugene6 according to manufacturer's recommendations. For transfections in 100-mm plates, cells were seeded at a density of 2.5 million cells/plate in DMEM containing 5% fetal bovine serum and transfected the next day using polyethyleneimine (PEI, 25-kDa molecular mass, linear form prepared at 1 mg/ml in sterile distilled H 2 O) at a DNA:PEI ratio of 1:3. Plasmid DNA and PEI were diluted, each in separate tubes, with 500 l of NaCl solution (150 mM). The PEI solution was then added onto the DNA solution, vortexed at maximum speed for 5 s, and incubated at room temperature for 20 min prior to addition to the cells. To generate cells stably expressing ␤2-adaptin-EYFP (HEK293T/␤2-adaptin-EYFP), transfected cells were selected with 1 mg/ml G418. Clonal cells expressing ␤2-adaptin-EYFP were obtained by limited dilution and were tested for expression of ␤2-adaptin-EYFP by fluorescence measurements and Western blotting. HEK293T cells stably expressing MycV2R (HEK293T/V2R) were generated by selection with 0.45 mg/ml G418, and a clonal cell line (41) expressing ϳ8 pmol/mg of V2R was used for transient cotransfection of ␤2-adaptin-EYFP and ␤-arrestin2-Rluc.
Total Fluorescence Measurements-To measure total fluorescence, cells were washed once with PBS-Mg (PBS containing 0.5 mM MgCl 2 ), gently detached by pipetting in PBS-Mg, and seeded at a density of ϳ100,000 cells/well in 96-well plates (white wall with clear bottom plates). All throughout this study, cell number quantification was based on measuring the A 600 of the cell suspension and correlating it to a pre-established standard curve (A 600 versus cell number). The fluorescence level was measured using a FlexStation II (Molecular Devices). For measuring fluorescence of V2R-Tsapphire, transfected cells were excited at 400 nm, and the emission was recorded at 511 nm (cut-off at 490 nm), whereas measurements of ␤2-adaptin-EYFP fluorescence were done at an excitation wavelength of 470 nm and emission wavelength of 535 nm (cut-off at 500 nm).
Monitoring Protein-Protein Interaction Using BRET 1 -Approximately 18 h after transfection, cells were detached by trypsinization and seeded (ϳ50,000 cells/well) into 96-well (white wall, clear bottom) tissue culture plates previously treated with poly-D-lysine, and re-incubated at 37°C for another 18 h. On the day of the experiment, the culture medium was replaced by PBS-Mg and incubated with or without various concentrations of the tested agonist for the specified time at room temperature. To measure the BRET 1 signal, the transparent bottom of the 96-well plate was covered with a whitebacked tape adhesive (PerkinElmer Life Sciences), and the BRET 1 substrate for Rluc, coelenterazine-h, was added to all wells (5 M final concentration), followed by BRET 1 measurement on the Mithras LB940 plate reader, which allows the Monitoring ␤-Arrestin and AP-2 Interaction by BRET OCTOBER 5, 2007 • VOLUME 282 • NUMBER 40 sequential integration of signals detected in the 480 Ϯ 20 nm and 530 Ϯ 20 nm windows. The BRET 1 signal was calculated as a ratio of the light emitted by EYFP (530 Ϯ 20 nm) over the light emitted by Rluc (480 Ϯ 20 nm). Unless otherwise stated, BRET 1 ratios were corrected by subtracting the background signal detected when ␤-arrestin-Rluc constructs were expressed in the absence of the EYFP fusion protein or the tested GPCR (in the case of the stable ␤2-adaptin-EYFP cell line) to generate BRETnet values. Agonist-promoted BRET 1 was calculated by subtracting the BRET 1 ratio obtained in the absence of agonist addition from the one obtained in the presence of an agonist. In the case of kinetic measurements, coelenterazine-h was added simultaneously with the agonist, followed by BRET 1 measurements. Receptor agonists used herein were: AVP for V2R and V1aR, isoproterenol for ␤2AR, bradykinin for B2R, vasoactive intestinal peptide for VIP1R, ET1 for ETAR and ETBR, carbachol for M2R, RANTES for CCR5, recombinant C5a for C5aR, prostaglandin E2 for EP4R, and angiotensin II for AT1AR.

Concomitant Monitoring of ␤-Arrestin2
Interactions with V2R and with AP-2 by Combining BRET 1 and BRET 2 -HEK293T/␤2-adaptin-EYFP cells were seeded in 100-mm dishes at a density of 2.5 million cells per plate. The next day, cells were co-transfected with different concentrations of the V2R-Tsapphire plasmid (0.25, 0.5, and 1 g) and 0.25 g of ␤-arrestin2-Rluc plasmid. Forty-eight hours post-transfection, cells were washed with PBS-Mg and then gently detached. Total fluorescence measurements were taken in 96-well plates containing ϳ100,000 cells/well, as described above. Cells were then dispensed into 96-well white opaque plates, and BRET 1 or BRET 2 kinetic measurements were done after the co-addition of 100 nM AVP and either 5 M coelenterazine-h (Rluc substrate for BRET 1 ) or coelenterazine DeepBlueC (Rluc substrate for BRET 2 ). BRET 2 measurements were done using a modified TopCount plate reader that sequentially incorporates emission signals in the 370 -450 nm and 500 -530 nm ranges. The BRET 2 signal was determined by calculating the ratio of the light emitted by the V2R-Tsapphire (500 -530 nm) over the light emitted by the ␤-arrestin2-Rluc (370 -450 nm). Agonist-promoted BRET 2 was calculated by subtracting the BRET 2 ratio obtained in the absence from the one obtained in the presence of agonist. Agonist-promoted BRET 1 was calculated as described above.
Measurement of Cell Surface Receptor Expression by ELISA-Agonist-promoted internalization was assessed as described previously (37). Briefly, the culture medium was removed and replaced with DMEM/0.5% bovine serum albumin/20 mM HEPES in the presence or absence of the specific agonist. After 30-min incubation at 37°C, the medium was removed and cells were fixed with Tris-buffered saline/3.7% formaldehyde for 5 min at room temperature. Cell surface expression was measured using anti-HA (Santa Cruz Biotechnology, Santa Cruz, CA, for ETAR), anti-Myc (for V1aR, V2R wt, and V2R(R137H)), or anti-FLAG M2 (Sigma, for ETBR) monoclonal antibodies. The percentage of agonist-promoted receptor internalization was determined as follows: (1 Ϫ stimulated/unstimulated) ϫ 100.
␤-Arrestin Knock-down-HEK293T cells were transfected with ETAR or ETBR plasmids along with minimal amount of ␤-arrestin1-Rluc and ␤-arrestin2-Rluc and either a specific siRNA targeting both ␤-arrestin1 and ␤-arrestin2 (AAAGC-CTTCTGCGCGGAGAAT) or a control siRNA (Santa Cruz Biotechnology). 18 h post-transfection, cells were seeded in 24-well plates and incubated at 37°C for an additional 18 -24 h. siRNA efficacy for ␤-arrestins knock-down expression was assessed by measuring the luminescence obtained from the co-expressed ␤-arrestin1-Rluc and ␤-arrestin2-Rluc and expressed as the percentage of luminescence obtained in cells transfected with the control siRNA.
Confocal Microscopy-HEK293T cells were transfected with ␤-arrestin2-GFP and either ETAR or ETBR constructs and treated with ET1 (10 nM) for the indicated time. ␤-Arrestin2-GFP images were collected on a Zeiss LSM-510 Meta laser scanning microscope.

RESULTS
Generation and Expression of ␤2-Adaptin-EYFP-This study was designed to monitor agonist-induced interaction between ␤-arrestins and AP-2 in living cells, thus, providing a sensitive detection for the early events of ␤-arrestin-promoted clathrinmediated endocytosis of GPCRs. For this, we used a ␤2-adaptin-EYFP fusion construct as a suitable BRET 1 partner (EYFP as energy acceptor) for ␤-arrestins fused to the energy donor Rluc (␤-arrestin2-Rluc or ␤-arrestin1-Rluc). Western blot analyses of HEK293T cells transfected with the ␤2-adaptin-EYFP construct confirmed that the protein was efficiently expressed. Indeed, a band migrating at ϳ130 kDa, corresponding to the expected molecular mass of the ␤2-adaptin-EYFP fusion protein was recognized by antibodies raised against either ␤2-adaptin (Fig. 1A, lane 2) or GFP (that recognizes EYFP; Fig.  1A, lane 5). A lower band of ϳ105 kDa, corresponding to the molecular weight of the endogenous ␤2-adaptin, was recognized only by the anti-␤2-adaptin antibody (compare lanes 1, 2, and 5 in Fig. 1A), in both transfected and untransfected cells. Transfection of wild-type ␤2-adaptin increased the immunoreactivity of the 105-kDa band (Fig. 1A, lane 3), confirming that this band corresponds to the native protein. To confirm the ability of ␤2-adaptin-EYFP to associate with its functional partners (43), co-immunoprecipitation experiments were carried out. As shown in Fig. 1B, immunoprecipitation of the fusion protein, using anti-GFP antibody, led to co-sedimentation of ␣-adaptin and clathrin, indicating that the fusion of ␤2-adaptin with EYFP did not prevent its association with the native AP-2/ clathrin complex (Fig. 1B).
Monitoring Agonist-promoted Interaction between ␤2-Adaptin/EYFP and ␤-Arrestin-Rluc in Living Cells by BRET 1 -To determine whether GPCR-promoted interaction between ␤-arrestin and ␤2-adaptin could be detected by BRET in living cells, we initially measured BRET 1 between ␤-arrestin2-Rluc and ␤2-adaptin-EYFP that were transiently co-transfected into HEK293T/V2R cells. The BRET 1 signal was measured in the absence and presence of increasing concentrations of the V2R agonist AVP. Fig. 1C shows that AVP promoted a concentration-dependent increase in BRET 1 , indicative of an association between ␤-arrestin2-Rluc and ␤2-adaptin-EYFP. The EC 50 calculated for this agonist-promoted interaction was 1.2 nM, a value in good agreement with the published affinity of AVP for V2R (38). These data validate the use of the ␤2-adaptin-EYFP/ ␤-arrestin-Rluc pair as a biosensor to monitor the early event in the receptor-promoted formation of the endocytic complex involving ␤-arrestin and AP-2.
To develop a robust system that allows monitoring of ␤-arrestin/AP-2 interaction by distinct receptors, we generated a HEK293T/␤2-adaptin-EYFP cells in which constant amounts of GPCR and ␤-arrestin-Rluc constructs could then be co-transfected. Twelve GPCRs were selected for this study ( Table 1). The GPCRs were chosen to include receptors belonging to classes A, B, and C, as defined by their ␤-arrestin interaction profile (13,18). The list also included receptors that were previously proposed to undergo clathrin-dependent or -independent pathways. Agonist activation of VIP1R, V2R, V1aR, B2R, ␤2AR, CCR5, EP4, C5aR, AT1aR, and M2R resulted in an increase in agonist-promoted BRET 1 , indicative of association between ␤-arrestin2-Rluc and ␤2-adaptin-EYFP. This effect was selective, because agonist-promoted BRET 1 was detected only when GPCRs were incubated with their respective agonist and not with irrelevant agonists (data not shown). The time and the concentration dependence of the agonist-promoted association are illustrated in Fig. 2 (A and B) for representative receptors and the parameters (t1 ⁄ 2 , EC 50 , and maximal BRET 1 values) of the curves obtained for all tested receptors are presented in Table 1. Although the t1 ⁄ 2 varied from 1.7 to 8.9 min between individual receptors, in all cases the BRET 1 signal reached a plateau after 12 min of stimulation with saturating concentra-

kinetics of association and comparison of generated EC 50 with agonist affinity values obtained from other established assays
Receptor classification with respect to interaction with ␤-arrestins is indicated under "class" (Refs. [13][14][15][16][17]51). The half-time (t1 ⁄ 2 ) of the maximal association between ␤-arrestin2-Rluc and ␤2-adaptin-EYFP in response to treatment with agonists (100 M for M2R and 100 nM for all other receptors) was calculated from non-linear regression analysis of BRET 1 association kinetics curves. BRET 1 max represents the agonist-promoted BRET 1 ratio obtained at maximum agonist stimulation. The BRET 1 EC 50 values were determined from non-linear regression of dose-response BRET 1 curves. BRET 1 EC 50 values were compared to previously published affinity values for the same agonists. Data are the mean Ϯ S.E. of 3-5 independent experiments.

Receptor
Agonist V1aR. This indicates that the greater affinity of ␤-arrestin2 to interact with class B receptors can be reflected in the BRETbased ␤-arrestin/AP-2 interaction assay ( Fig. 2C and inset).
Although ␤-arrestin2 has been studied more extensively than other non-visual arrestins, ␤-arrestin1 has also been shown to interact with AP-2 upon activation of certain GPCRs (7, 10). Thus, we assessed the ability of the BRET-based assay to FIGURE 2. Kinetics and dose dependence of agonist-promoted interaction between ␤2-adaptin-EYFP and ␤-arrestin-Rluc. HEK293T/␤2-adaptin-EYFP cells were co-transfected with ␤-arrestin2-Rluc and the indicated GPCRs. A, real-time BRET 1 measurements were taken at regular intervals for the indicated time period directly after the addition of coelenterazine-h and receptor specific agonists. B, cells were stimulated with the indicated concentrations of the receptor-specific agonists for 20 min at room temperature followed by BRET 1 measurements. For the agonist list, see Table 1. C, cells were stimulated with 100 nM AVP for 20 min followed by BRET 1 measurement. Inset, total cell surface expression of each receptor was assessed by cell surface ELISA assay using anti-Myc antibody. detect this interaction on a representative subset of the GPCRs tested above. As shown in Fig. 4, agonist activation of the V1aR, ␤2AR, B2R, V2R, and VIP1R resulted in an increase in BRET 1 between ␤-arrestin1-Rluc and ␤2-adaptin-EYFP, confirming that these receptors can promote the interaction between AP-2 and either ␤-arrestin1 or -2. Interestingly, the extent of signal generated by ␤-arrestin1 and ␤-arrestin2 was identical when promoted by the maximal activation of V2R and VIP1R, two prototypical class B receptors that display identical ability to interact with the two ␤-arrestins (13). In contrast, maximal activation of the class A receptors, V1aR and ␤2AR, led to weaker signals with ␤-arrestin1 than ␤-arrestin2, consistent with their reported lower propensity to associate stably with ␤-arrestin1 (13). For the B2R, although its propensity to colocalize in endosomes with ␤-arrestin2 (18) suggest a class B receptor, the weaker signal observed between ␤2-adaptin-EYFP and ␤-arrestin1-Rluc as compared with ␤-arrestin2-Rluc would be more consistent with a class A. This hybrid behavior is consistent with the observation by Simaan et al. (18) that, in contrast with prototypical class B receptors, the B2R once in the endosomes can dissociate from ␤-arrestin2 upon removal of the agonist, allowing the recycling of the receptor to the cell surface. The direct comparison of the interaction profile between ␤2-adaptin and the two ␤-arrestins promoted by different receptors is made possible by the fact that energy donor/ acceptor ratio (Rluc/EYFP) can be maintained constant independently of the receptor being expressed. In that respect, the present assay provides an advantage over the direct monitoring of ␤-arrestins to the receptor, because the identities of the BRET partners used are the same, independently of the receptor being studied. Thus, the ␤-arrestin-Rluc/␤2-adaptin-EYFP BRET assay provides a useful tool to classify GPCRs according to their profile of interaction with the ␤-arrestin/AP-2 complex.
Among the receptors tested, only two, ETAR and ETBR, failed to promote the interaction between either ␤-arrestin2-Rluc or ␤-arrestin1-Rluc and ␤2-adaptin-EYFP ( Figs. 2A and 4). No BRET 1 signal could be detected even when agonist stimulation was maintained for 60 min (data not shown). To determine whether the lack of signal could result from the inability of these receptors to recruit ␤-arrestin, we used Double-Brilliance-␤arrestin (Rluc-␤-arrestin2-EYFP), a biosensor that allows monitoring of conformational changes of ␤-arrestin2 by intramolecular BRET 1 upon translocation to the receptor (42). As can be seen in Fig. 5A, agonist (ET1) activation of HEK293T cells co-transfected with ETAR or ETBR and Double-Brilliance-␤arrestin induced an increase in intramolecular BRET 1 , indicating that these receptors recruited ␤-arrestin2 but failed to promote interaction between ␤-arrestin and AP-2. In addition, ET1-promoted translocation of ␤-arrestin2 to the plasma membrane was observed by fluorescence microscopy in HEK293T cells transfected with ETAR or ETBR (Fig. 5B). This is consistent with a previous report indicating that ETAR recruits ␤-arrestin2 following agonist activation (45). Also consistent with previous observations, ET1 stimulation led to the endocytosis of both ETAR and ETBR, as evidenced by the decrease in the receptor cell surface expression detected by ELISA following agonist treatment (Fig. 5C). Moreover, involvement of ␤-arrestins in ETAR and ETBR agonist-induced internalization could be demonstrated by interfering with ␤-arrestins expression by the use of siRNAs targeting ␤-arrestin1 and ␤-arrestin2. Indeed, transfection of these siRNAs, which led to a significant reduction in both ␤-arrestin1 and ␤-arres-tin2 levels, significantly blocked ETAR and ETBR agonist-promoted internalization (Fig. 5C). Selective knock-down of ␤-ar-restin2 led to a similar inhibition of the endocytosis (data not shown), indicating a predominant role for ␤-arrestin2. Taken together these data indicate that ETAR and ETBR can be endocytosed by a ␤-arrestin-dependent but AP-2-independent pathway, suggesting a non-clathrin-mediated endocytosis for these receptors.
Agonist-independent Association between ␤-Arrestin and ␤2-Adaptin Reveals Receptor Constitutive Internalization-Recently, mutations leading to constitutive endocytosis in the absence of ligand stimulation have been described for several GPCRs (46). Among those, substitution of arginine 137 by a histidine residue within the V2R has been found to promote the constitutive recruitment of ␤-arrestin to the receptor and the ensuing endocytosis of the complex (35,47). To investigate whether constitutive internalization of GPCRs also involves ␤-arrestin/AP-2 interaction, we performed BRET 1 experiments on HEK293T/␤2-adaptin-EYFP cells transiently co-transfected with ␤-arrestin2-Rluc and either the wild-type or a constitutively internalizing V2R mutant, V2R(R137H). In the absence of agonist stimulation, V2R(R137H) promoted a slightly higher basal BRET 1 signal between ␤-arrestin2-Rluc and ␤2-adaptin-EYFP than the wild-type V2R, and this, despite a lower cell surface expression, is consistent with its constitutive endocytosis (Fig.  6). Co-expression of the dominant negative mutant of dynamin I, dynI(K44A), that prevents the pinching off of endocytic vesicles from the plasma membrane further revealed the difference in BRET 1 signal promoted by V2R(R137H) versus WT (Fig. 6). The increase in the V2R(R137H)-promoted BRET 1 observed in the presence of dynI(K44A) most likely results from the accu-mulation of the receptor in the clathrin-coated pits at the cell surface (as illustrated by the increase in cell surface receptor detected by ELISA, Fig.  6, inset) where the constitutive formation of the V2R(R137H)/␤-arrestin2/ AP-2 complex can be detected. This is consistent with the role of dynamin that promotes severing of the clathrin-coated pits but is not involved in the assembly of the early endocytic machinery. The observation, that the BRET 1 signal promoted by V2R(R137H) in the presence of dynI(K44A) is of the same magnitude as that observed following agonist activation of the wild-type receptor by AVP, suggests that the R137H mutation can promote the formation of the receptor/␤-arrestin/AP-2 complex as efficiently as the activated wild-type receptor. Taken together, these results confirm that constitutive endocytosis of V2R(R137H) and agonist-promoted internalization of the wild-type receptor involve the same initial steps of the clathrin-dependent pathway involving the interaction between ␤-arrestin and AP-2.
Receptor Activity-independent Recruitment of ␤-Arrestin2 Promotes Interaction between ␤-Arres-tin2 and AP-2-Because it was previously shown that agonist stimulation of the ␣1A-adrenergic receptor (48) could directly promote the recruitment of AP-2 (through its subunit) to the receptor, one could envisage that the interaction between ␤-arrestin and AP-2 could follow their independent recruitment to the activated receptor. Alternatively, the interaction between ␤-arrestin and AP-2 may reflect a direct recruitment of AP-2 to ␤-arrestin following the translocation of ␤-arrestin to the receptor. To distinguish between these two possibilities, we assessed whether the recruitment of ␤-arrestin to the receptor, independently of agonist-promoted receptor activation, is sufficient to promote the interaction between ␤-arrestin and AP-2. For this purpose, we took advantage of a cyclophilin-based approach developed on the ability of a synthetic bivalent heterodimerizer compound, FIGURE 5. Agonist-induced ␤-arrestin2 recruitment to ETAR and ETBR and receptor internalization. A, HEK293T cells were co-transfected with ETAR or ETBR and Rluc-␤-arrestin2-EYFP (Double-Brilliance), and BRET 1 was measured after 20 min of agonist (ϩET1, 10 nM) or vehicle treatment (Ctrl). B, HEK293T cells were transfected with ␤-arrestin2-GFP and either ETAR or ETBR and treated with 10 nM of ET1 for the indicated time period. ␤-Arrestin2-GFP translocation to the plasma membrane was observed using confocal fluorescence microscopy. C, HEK293T cells were transfected with HA-ETAR or FLAG-ETBR and either a control siRNA (Ctrl) or a siRNA targeting both ␤-arrestin1 and ␤-arrestin2 (siRNA-␤-arrestins). Surface expression of each receptor was assessed by ELISA, in the absence or presence of 10 nM ET1, using anti-Ha (for ETAR) or anti-FLAG (for ETBR) antibodies, and the percentage of agonistpromoted receptor internalization was determined. The effect of siRNA on ␤-arrestin1 and ␤-arrestin2 expression levels was assessed by measuring the reduction of luminescence activity originating from low level of co-transfected ␤-arrestin1-Rluc and ␤-arrestin2-Rluc. Data are the mean Ϯ S.E. (A and C) of four independent experiments or are representative (B) of two independent experiments. AP-21967, to function as a high affinity adaptor to link the cyclophilin FRB and FKBP protein domains. The fusion of FRB to the N terminus of ␤-arrestin2-Rluc and FKBP to the C terminus of V2R was previously shown to allow the recruitment of ␤-arrestin to the receptor upon AP-21967 treatment, in the absence of AVP. Such agonist-independent ␤-arrestin recruitment did not result in G protein activation but induced the internalization of the V2R-FKBP (34). Fusions with the cyclophilin domains were also shown not to affect the functionality of either ␤-arrestin2 or V2R (34). As shown in Fig. 7, treatment with AP-21967 induced a concentration-dependent BRET 1 between ␤2-adaptin-EYFP and FRB-␤-arrestin2-Rluc in cells co-expressing V2R-FKBP. No such BRET 1 was observed in cells expressing V2R and ␤-arrestin2-Rluc not fused to FKBP and FRB, respectively. These data clearly indicate that recruitment of ␤-arrestin to the receptor, even in the absence of agonist activation, is sufficient to induce ␤-arrestin/AP-2 association. It has previously been suggested that the interaction between a phosphate sensor within ␤-arrestin and the phosphorylated tail of the receptor contributes to the conformational rearrangements leading to the engagement of AP-2 (7). Given that the agonist-independent recruitment of ␤-arrestin promoted by AP-21967 should not promote receptor phosphorylation, our data suggest that ␤-arrestin conformational rearrangements promoted by its interaction with the receptor, even in the absence of the phosphorylation, may be sufficient to trigger the interaction between ␤-arrestin and AP-2.
Association Kinetics between V2R/␤-Arrestin2 versus ␤-Ar-restin2/␤2-Adaptin as Measured by Combining BRET 1 and BRET 2 -From the above discussion, one would predict that ␤-arrestin translocation to the receptor would precede ␤-arrestin/AP-2 interaction. To directly test this prediction, we took advantage of the availability of two different coelenterazine substrates whose oxidation by Rluc results in resolved light emission spectra, allowing differential energy transfer to two different derivatives of GFP (49). Oxidation of DeepBlueC by Rluc (used in BRET 2 ) results in a maximal light emission at 400 nm, thus allowing efficient energy transfer to GFP variants such as GFP 2 or Tsapphire, whose maximum excitation wavelengths are ϳ400 nm, but permits only marginal energy transfer to EYFP (maximum excitation wavelength: ϳ511 nm). Conversely, Rluc-catalyzed oxidation of coelenterazine-h (used in BRET 1 ), which emits light with a peak emission at ϳ480 nm, allows energy transfer to EYFP but fails to excite Tsapphire. Therefore, it is possible to combine BRET 1 and BRET 2 in the same cells to follow two different protein-protein interaction events (49). For this, HEK293T/␤2-adaptin-EYFP cells transiently co-transfected with V2R-Tsapphire and ␤-arrestin2-Rluc were stimulated with AVP. BRET 1 and BRET 2 were then used to monitor kinetics of interaction for the ␤-arrestin2-Rluc/␤2-adaptin-EYFP and ␤-arres-tin2-Rluc/V2R-Tsapphire pairs, respectively. As shown in Fig.  8, the apparent association of ␤-arrestin2 to the receptor (t1 ⁄ 2 : 4.2 min) is faster than its interaction with AP-2 (t1 ⁄ 2 : 10 min). Because association kinetics may vary with different expression levels of the studied interactors, we further calculated the association halftimes (t1 ⁄ 2 ) from BRET 1 and BRET 2 under conditions where we varied the expression levels of V2R-Tsapphire in the presence of fixed levels of ␤-arrestin2-Rluc and ␤2-adaptin-EYFP (Table 2). Under all conditions, AVP stimulation resulted in ␤-arrestin2-Rluc/V2R-Tsapphire association that was in average 2.5 Ϯ 0.1 times faster than the interaction between ␤-arrestin2-Rluc and ␤2-adaptin-EYFP. Taken together, these results are consistent with the notion that the formation of the receptor/␤-arrestin/AP-2 complex is initiated by the recruitment of ␤-arrestin to the receptor, resulting in conformational changes leading to the association with AP-2.

DISCUSSION
This study was aimed at the development of a live cell approach that makes possible direct detection of the early events leading to ␤-arrestinand clathrin-dependent internalization of GPCRs. For this purpose, we developed a BRETbased biosensor to monitor the agonist-induced interaction between ␤-arrestin-Rluc and ␤2-adaptin-EYFP, a needed step for the ␤-arrestin-mediated targeting of GPCRs to clathrincoated pits for internalization (7,8,10).
Out of the 12 different GPCRs tested, only the ETAR and ETBR were unable to promote ␤-arrestin association with AP-2 upon agonist-stimulated receptor internalization, indicating an alternative endocytic route for these two receptors. Among the receptors promoting ␤-arrestin/␤2-adaptin interactions, those displaying lower and higher affinity for ␤-arrestins (class A and B, respectively) could be readily differentiated based on the strength of the BRET 1 signals detected between ␤2-adaptin-EYFP and both of ␤-arrestin2-Rluc and ␤-arrestin1-Rluc. The comparison of the kinetics of association of ␤-arrestin with the receptor and AP-2, made possible by combining BRET 1 and BRET 2 in living cells, respectively, indicated that ␤-arrestin association with the activated receptors precedes and probably triggers the recruitment of AP-2 to the ␤-arrestin/receptor complex. Consistent with this notion, the recruitment of ␤-arrestin to a constitutively internalizing mutant of the V2R or forcing its binding to the wild-type receptor in the absence of agonist stimulation (using a cyclophilin-based approach) was sufficient for promoting the interaction of AP-2 with the complex.
The list of receptors tested in this study included GPCRs that were previously proposed to undergo clathrin-dependent or -independent pathways. Agonist activation of VIP1R, V2R, V1aR, B2R, ␤2AR, CCR5, EP4R, M2R, AT1aR, and C5aR resulted in an increase in agonist-promoted BRET 1 between ␤-arrestin and ␤2-adaptin, indicative that internalization of these receptors occurs via a ␤-arrestinand clathrin-dependent pathway. Studies using pharmacological and biochemical inhibitors previously reported that VIP1R, V2R, V1aR, B2R, ␤2AR, CCR5, and EP4R undergo clathrin-dependent endocytosis (6, 17, 50 -54). However, the internalization route previously proposed for M2R, AT1aR, and C5aR were more controversial. Indeed, it has been previously reported that the internalization of the M2R is atypical in being both non-clathrin and non-caveolae mediated, because it was not significantly affected by the co-expression of dominant negative mutants of ␤-arrestin1 (␤-arrestin1(V53D)), dynamin (dynI(K44A)), or the heavy chain of clathrin (Hub) (55,56). These results were surprising, considering that treatment with hypertonic sucrose, a classic inhibitor of clathrin-mediated endocytosis (57), was reported to strongly inhibit M2R internalization (55). Recently, by monitoring the trafficking of an M2R fused to GFP, it was reported that the M2R appeared to internalize initially via a clathrin-independent pathway but then quickly merged with the clathrin-endocytic pathway at the level of early endosomes (58). Our BRET 1 data clearly support the involvement of the clathrin pathway in M2R internalization and demonstrate that the M2R can promote the formation of the ␤-arrestin/AP-2 complex involved in the early steps of clathrin-mediated endocytosis. For the AT1aR and C5aR, dominant negative mutants of ␤-arrestin1 (␤-arrestin1(V53D) or ␤-arrestin2-(319 -418)) and dynamin (dynI(K44A)) were initially reported not to inhibit receptor internalization (59,60). However, later studies showed that the internalization of both AT1aR and C5aR can be inhibited by dominant negative mutants of both dynamin and ␤-arrestins (51,61,62). The results obtained in the present study using the direct BRET 1 -based monitoring of the ␤-arrestin/ AP-2 interaction are consistent with the latter studies, indicating that AT1aR and C5aR can undergo ␤-arrestinand clathrinmediated endocytosis.
Taken together, the conflicting results obtained in some studies illustrate the difficulties of interpretations associated with the exclusive use of dominant negative mutants to define the endocytic route used by specific receptors. The expression level of the mutant used and the endogenous levels of the proteins involved in the pathway are among the factors that can explain the contradictory results obtained. In this context, the BRET-based assays, directly monitoring the interactions between components of specific endocytic pathway, offer several advantages. For instance, the behavior of each receptor can be examined in an identical cellular background where the expression level of the components assayed can be selected and monitored in the course of the assay by measuring fluorescence and luminescence levels. Also, the biosensor system used to monitor the endocytic process is kept constant, independently of the receptor studied, allowing direct comparison between receptors that are not modified. The assay is thus a direct readout of the ability of each receptor to engage specific components of the endocytic machinery and to evaluate the potential of the receptor to use specific internalization pathways in a given cell type.
As discussed above, the detection of receptor promoted BRET 1 between ␤-arrestin and ␤2-adaptin is a clear demonstration that this receptor can be internalized via clathrin-coated endosomes. However, a lack of BRET 1 , although ruling out a ␤-arrestin-mediated clathrin-dependent process, does not exclude the possibility of a ␤-arrestin-independent/clathrin-dependent or a ␤-arrestinmediated/clathrin-independent pathway. ␤-Arrestin-independent/clathrin-dependent endocytosis of GPCRs has been previously suggested for the HCMV-encoded chemokine US28 and 5-hydroxytryptamine 2A receptors (63,64). Interestingly, direct interaction between a subunit of AP-2 (2adaptin) and both the ␣1b-adrenergic receptor and the protease-activated receptor 1 (48,72) has been documented. Whether such direct interactions between AP-2 and certain GPCRs could be responsible for the ␤-arrestin-independent clathrin-mediated endocytosis remains to be investigated.
For ETAR and ETBR, our data strongly suggest a ␤-arrestindependent but clathrin-independent endocytic pathway. This is consistent with previous data suggesting an internalization process involving both caveolae and ␤-arrestin for ETAR (20). In addition to providing additional support to the hypothesis that ␤-arrestin could be involved in nonclathrin-mediated endocytosis, these results reinforce the notion that inhibition of endocytosis by ␤-arrestin dominant negative mutants cannot be used as a marker of clathrin-dependent internalization.
When considering the ␤-arrestin/clathrin-dependent pathway, our data are consistent with the notion that the recruitment of ␤-arrestin to the receptor and the ensuing conformational changes are needed for the subsequent association with AP-2 leading to clathrin-mediated endocytosis. In particular, BRET 1 and BRET 2 clearly indicated that the ␤-arrestin/AP-2 complexes did not exist in the absence of receptor activation and that the recruitment of ␤-arrestin to the receptor precedes the association of ␤2-adaptin to the complex. Consistent with this notion, the recruitment of ␤-arrestin to the V2R using the cyclophilin-based recruitment system was sufficient to promote ␤-arrestin/AP-2 interaction and the consequent endocytosis, suggesting that GPCR/␤-arrestin interaction may be the triggering signal for the recruitment of AP-2 to the complex. However, for ETAR and ETBR receptors, recruitment of ␤-arrestin was not sufficient for directing these receptors to the AP-2-and clathrindependent endocytic pathway. Thus, it is likely that AP-2 association with ␤-arrestin is determined by specific motifs or conformational rearrangements that are intrinsic to specific GPCR/␤-arrestin complexes.
In addition to allowing detailed analysis of the molecular events involved in GPCR endocytosis, the BRET-based assay developed herein provides a robust tool to monitor the activation of receptors that internalize via the ␤arrestinand clathrin-dependent process. Interestingly, the calculated EC 50 values for the agonist-promoted interaction between ␤-arrestins and ␤2-adaptin correlated very well with the agonist affinities reported for GPCR functional assays. The ␤-arrestin/␤2-adaptin BRET assay also presents the advantage of being independent of the G protein-coupling preference of the receptor (G i/o -, G s -, and G q/11 -coupled receptors) and does not require modification of the receptor, thus offering a universal functional assay for native GPCRs that internalize via the ␤-arrestinand clathrin-mediated pathway. Finally, the present study opens the way to the development of similar BRET-based assays directly assessing the interaction between other partners of the endocytic machineries that will provide useful tools to dissect alternative pathways for GPCR endocytosis. The different receptor expression levels were determined by fluorescence measurements (excitation 400 nm, emission 511 nm). BRET 1 ratio represents the interaction between ␤2-adaptin-EYFP and ␤-arrestin2-Rluc, while BRET 2 ratio corresponds to the recruitment of ␤-arrestin2-Rluc to V2R-Tsapphire. Apparent t1 ⁄ 2 (half-time) of maximal association is indicated. Data are the means Ϯ S.E. of three independent experiments.