β-Arrestin/AP-2 Interaction in G Protein-coupled Receptor Internalization

β-Arrestins, proteins involved in the turn-off of G protein-coupled receptor (GPCR) activation, bind to the β2-adaptin subunit of the clathrin adaptor AP-2. The interaction of β2-adaptin with β-arrestin involves critical arginine residues in the C-terminal domain of β-arrestin and plays an important role in initiating clathrin-mediated endocytosis of the β2-adrenergic receptor (β2AR) (Laporte, S. A., Oakley, R. H., Holt, J. A., Barak, L. S., and Caron, M. G. (2000)J. Biol. Chem. 275, 23120–23126). However, the β-arrestin-binding site in β2-adaptin has not been identified, and little is known about the role of β-arrestin/AP-2 interaction in the endocytosis of other GPCRs. Using in vitro binding assays, we have identified two glutamate residues (Glu-849 and Glu-902) in β2-adaptin that are important in β-arrestin binding. These residues are located in the platform subdomain of the C terminus of β2-adaptin, where accessory/adapter endocytic proteins for other classes of receptors interact, distinct from the main site where clathrin interacts. The functional significance of the β-arrestin/AP-2/clathrin complex in the endocytosis of GPCRs such as the β2AR and vasopressin type II receptor was evaluated using mutant constructs of the β2-adaptin C terminus containing either the clathrin and the β-arrestin binding domains or the β-arrestin-binding domain alone. When expressed in human embryonic kidney 293 cells, both constructs acted as dominant negatives inhibiting the agonist-induced internalization of the β2AR and the vasopressin type II receptor. In addition, although the β2-adaptin construct containing both the clathrin and β-arrestin binding domains was able to block the endocytosis of transferrin receptors, a β2-adaptin construct capable of associating with β-arrestin but lacking its high affinity clathrin interaction did not interfere with transferrin receptor endocytosis. These results suggest that the interaction of β-arrestin with β2-adaptin represents a selective endocytic trigger for several members of the GPCR family.


␤-Arrestins, proteins involved in the turn-off of G protein-coupled receptor (GPCR) activation, bind to the
The functional significance of the ␤-arrestin/AP-2/clathrin complex in the endocytosis of GPCRs such as the ␤ 2 AR and vasopressin type II receptor was evaluated using mutant constructs of the ␤ 2 -adaptin C terminus containing either the clathrin and the ␤-arrestin binding domains or the ␤-arrestin-binding domain alone. When expressed in human embryonic kidney 293 cells, both constructs acted as dominant negatives inhibiting the agonist-induced internalization of the ␤ 2 AR and the vasopressin type II receptor. In addition, although the ␤ 2 -adaptin construct containing both the clathrin and ␤-arrestin binding domains was able to block the endocytosis of transferrin receptors, a ␤ 2 -adaptin construct capable of associating with ␤-arrestin but lacking its high affinity clathrin interaction did not interfere with transferrin receptor endocytosis. These results suggest that the interaction of ␤-arrestin with ␤ 2 -adaptin represents a selective endocytic trigger for several members of the GPCR family.
␤-Arrestins (␤-arrestin-1 and ␤-arrestin-2) are cytosolic proteins involved in the homologous desensitization of many Gprotein coupled receptors (GPCR) 1 (1,2). Agonist stimulation of GPCRs triggers the activation and the recruitment of specific GPCR protein kinases leading to the phosphorylation of cytosolic residues in the receptor to promote the subsequent binding of ␤-arrestin. For example, the interaction of ␤-arrestin with the phosphorylated ␤ 2 -adrenergic receptor (␤ 2 AR) prevents further coupling to its cognate G protein (i.e. G s ), thus terminating the second messenger signaling events. ␤-Arrestins, initially appreciated exclusively for their ability to desensitize agonist-activated GPCRs, are now believed to play a much more intricate role in other cellular events such as endocytosis, trafficking, and intracellular signaling of the GPCRs (2)(3)(4). For instance, the initial observation that ␤-arrestin and mutants of ␤-arrestin could modulate the internalization of ␤ 2 AR provided evidence for a role of ␤-arrestins in this process (5). Moreover, the findings that ␤-arrestins were able to associate with components of the endocytic machinery, such as clathrin and the clathrin adaptor protein AP-2, have provided an attractive mechanism to explain how ␤-arrestins could engage GPCRs in the internalization pathway (6,7). Although ␤-arrestins can associate with a great number of agonist-stimulated GPCRs at the plasma membrane to trigger their internalization, the fate of the receptor/␤-arrestin complexes differs greatly among receptors. For example, ␤-arrestin dissociates from ␤ 2 AR at or near the plasma membrane following the internalization of the receptor, whereas ␤-arrestin has been shown to traffic into endosomes with other GPCRs like the vasopressin type II receptor (V2R) and the angiotensin II type 1 receptor (AT1R) (8,9). The intracellular trafficking of ␤-arrestins with V2R and AT1R correlates with the slow recycling to the plasma membrane and resensitization of these receptors (8,10). However, for some GPCRs the intracellular trafficking of ␤-arrestins with the receptors seem to activate specific signaling pathways. For the AT1R and the proteaseactivating receptor 2, the formation of endosomal receptor/␤arrestin complexes has been shown to serve as a scaffold for the recruitment and the activation of components of the mitogenactivated protein kinase pathways (11)(12)(13).
Endocytosis via clathrin-coated vesicles (CCVs) is one of the most common routes utilized by mammalian cells to internalize diverse classes of cargo such as extracellular molecules and membrane receptors. CCVs at the plasma membrane, also referred to as clathrin-coated pits, are composed of two main structural proteins: clathrin and the clathrin adaptor protein AP-2 (14,15). AP-2 is a heterotetrameric complex that serves the dual role of assembling clathrin into organized cage structures and acting as an adaptor to link cargo to clathrin lattices. This is achieved through the recognition of distinct signal motifs in the cytosolic domain of membrane proteins by different subunits of AP-2. For example, the 2-subunit of AP-2 recognizes tyrosine-based internalization signals within the cytosolic domains of receptors (16 -18), whereas the ␤ 2 -subunit of AP-2 (␤ 2 -adaptin) interacts with clathrin and helps to promote clathrin lattice assembly (19). The ␣-subunits of AP-2 bind dynamin (20), a GTPase that promotes budding of clathrincoated vesicles (21), and recruit other endocytic accessory proteins necessary for the formation and processing of CCVs (22). We have recently shown that the interaction of ␤-arrestin with ␤ 2 -adaptin is necessary for the targeting of the ␤ 2 AR into CCVs and have identified critical residues in ␤-arrestin that mediate this interaction (23). However, little information is available for the corresponding ␤-arrestin-binding site in ␤ 2 -adaptin. Moreover, although ␤-arrestin/AP-2 complexes have been shown to play a role in the internalization of the ␤ 2 AR, the functional significance of these complexes in clathrin-mediated endocytosis of other GPCRs remains undetermined. In this study we sought to identify critical residues within the ␤ 2subunit of AP-2 involved in ␤-arrestin binding, and have investigated the functional importance of this interaction in the internalization process of different classes of membrane receptors.

EXPERIMENTAL PROCEDURES
Materials-Isoproterenol was purchased from Research Biochemical Inc., and arginine vasopressin (AVP) was obtained from Sigma. The anti-HA 12CA5 mouse monoclonal antibody was purchased from Roche Molecular Biochemicals, anti-epsin antibody was from Santa Cruz, and anti-clathrin heavy chain antibody was from Transduction Laboratories. [ 3 H]AVP was purchased from PerkinElmer Life Sciences.
Cell Culture and Transfection-Human embryonic kidney (HEK) 293 cells were obtained from the American Type Culture Collection (ATCC). Cells were grown in Eagle's minimal essential medium with Earle's salt supplemented with 10% (v/v) heat-inactivated fetal bovine serum and gentamicin (100 g/ml). Transient transfections were performed using a modified calcium phosphate coprecipitation method as described previously (24). Twenty-four hours after transfection, cells were split into appropriate plates and experiments were performed the following day. For internalization experiments using the ␤ 2 AR or the V2R, cells were seeded at a density of 5.0 -7.5 ϫ 10 5 cells/well in 6-well plates and 2.5 ϫ 10 5 cells/well in 12-well plates, respectively.
Receptor Sequestration-␤ 2 -Adrenegic receptor sequestration was assessed by flow cytometry as described previously (8). Transfected HEK 293 cells were incubated with or without isoproterenol (10 M) for 20 min. Sequestration of receptors was defined as the fraction of cell surface receptors that was removed from the surface after exposure to agonist. Internalization of the V2R was assessed using a previously described assay with minor modifications (8). In brief, cells expressing the V2R were incubated at 37°C in Eagle's minimal essential medium with Earle's salt containing 0.2% (w/v) of bovine serum albumin in presence of 1 nM [ 3 H]AVP for 20 min. Internalization of receptors was stopped on ice by rapidly washing the cells with ice-cold phosphatebuffered saline (PBS). Cells were washed three times with either icecold PBS to remove the unbound agonist or ice-cold acid wash buffer (0.2 N acetic acid, pH 2.6, 150 mM NaCl) to remove both the unbound AVP and the cell surface receptor-bound agonist. Sequestration of receptors was defined as the percentage of radioligand that was acidresistant after incubating cells for 20 min at 37°C. Internalization of transferrin was assessed using confocal microscopy to measure the uptake of FITC-conjugated transferrin in HEK 293 cells. Cells were incubated with FITC-transferrin for 15 min at 37°C and then fixed in PBS containing 4% paraformaldehyde. For the detection of ␤ 2 -adaptin mutants, fixed cells were stained in PBS containing 2% bovine serum albumin (w/v) and 0.2% Triton X-100 (v/v) using a primary HA antibody followed by secondary Texas red labeling. For quantification of transferrin uptake, 100 cells showing comparable expression of ␤ 2 -adaptin minigene constructs (as estimated by the intensity of the fluorescence) were analyzed. Inhibition of transferrin uptake was defined as maximal when transfected cells showed a FITC signal of less than 25% of that of adjacent cells lacking the expression of the ␤ 2 -adaptin constructs (i.e. when transfected cells showed a reduction of transferrin fluorescence signal of more than 75%). The level of transferrin and ␤ 2 -adaptin fluorescence (in intensity per pixel) was measured in different areas of the cytoplasm using the LSM 510 microscope software as described previously (25).
GST Fusion Protein Purification and Pull-down Experiments-␤ 2 -Adaptin and ␤-arrestin-1 C-terminal constructs in pGEX-5X-2 were transformed in Escherichia coli BL21-gold (DE3) cells, and GST fusion proteins were prepared as previously described (23). HEK 293 cells expressing the Flag-tagged ␤-arrestins or HA-tagged ␤ 2 -adaptin constructs were solubilized in TGH buffer (50 mM HEPES, pH 7.4, 1% (v/v) Triton X-100, 10% (v/v) glycerol, 50 mM NaCl, 1 mM EDTA) containing protease inhibitors. Cells were solubilized for 1 h at 4°C, centrifuged at 40,000 ϫ g for 30 min, and the supernatant was recovered for GST fusion protein binding assays. For experiments involving mouse whole brain extract, a similar procedure was followed for protein solubilization, with the exception that whole brain was homogenized in TE buffer (10 mM Tris-HCl, pH 7.4, 2 mM EDTA, 1.0% Triton X-100) containing protease inhibitors. Binding assays were performed using 10 g of GST fusion proteins on glutathione-Sepharose beads that were incubated for 1 h at 4°C with solubilized proteins. Beads were recovered by centrifugation and washed three times with cold TGH or TE buffer, and the protein complexes were separated by SDS-PAGE. Proteins were transferred to nitrocellulose, analyzed by Ponceau's staining to detect the integrity of the GST fusion proteins, and subjected to immunoblotting.
To confirm the fine mapping studies with ␤ 2 -adaptin fusion proteins, we used the yeast two-hybrid system to identify the ␤-arrestin binding site in ␤ 2 -adaptin. Full-length and different C-terminal constructs of ␤ 2 -adaptin were expressed with ␤-arrestin-2 in yeast, and the association between the proteins was evaluated ( Table I). Expression of full-length ␤ 2 -adaptin, ␤ 2adaptin-(592-937), or ␤ 2 -adaptin-(664 -937) fragments together with ␤-arrestin-2 was sufficient to induce yeast growth on selective media. Moreover, a fragment of ␤ 2 -adaptin C terminus encoding the last 112 residues (825-937) was also found to interact with ␤-arrestin in the same yeast-based assay. Taken together, results from both approaches indicate that the ␤-arrestin binding site is located within residues 825-937 of the ␤ 2 -adaptin C terminus.
The crystal structure of the ear domain of ␤ 2 -adaptin has recently been solved (28). Analysis of the structure reveals the presence of two regions: an N-terminal subdomain (residues 705-825) and a C-terminal or "platform" subdomain (residues 826 -937) (Fig. 3). Our results indicate that the ␤-arrestinbinding site is located within the platform subdomain of ␤ 2adaptin. We have recently shown that this interaction involves two arginine residues within the ␤-arrestin C terminus (Arg- ␤-Arrestin/AP-2 Interaction in GPCR Internalization 394 and Arg-396) (23). We hypothesized that the positively charged guanidino group of the arginine residues in ␤-arrestin might form an ionic bond with the carboxyl group of acidic residues in ␤ 2 -adaptin (i.e. aspartate or glutamate). Initial attempts at defining the ␤-arrestin binding site by replacing pairs or triplets of consecutive acidic residues in the ␤ 2 -adaptin C-terminal domain yielded only limited information on the residues involved. Indeed, most of the ␤ 2 -adaptin mutants by themselves were trans-activating in the yeast two hybrid system, with the exception of the ␤ 2 -adaptin-(825-937) (E828A, D829A, E833A) and ␤ 2 -adaptin-(664 -937) (E922A, D932A) constructs, which did not show any trans-activation when expressed by themselves and were both found to still interact with ␤-arrestin-2 (data not shown).
Alternative strategies to map the ␤-arrestin-binding site in ␤ 2 -adaptin were considered. We took advantage of the solved structure of ␤ 2 -adaptin C terminus to identify neighboring acidic residues in the platform subdomain that might participate in this interaction (28). This region of ␤ 2 -adaptin contains 13 acidic residues, and analysis of their relative position in the ␤ 2 -adaptin platform reveals the presence of a candidate pair: residue Glu-849, located in the first ␤-sheet strand between the ␣1 and ␣2 helices; and residue Glu-902, located in the fourth ␤-sheet strand (Fig. 3). Although these residues are distant with respect to the primary sequence, structurally and molecularly the side chains of Glu-849 and Glu-902 residues are in close proximity to one another (Ϸ6 -10 Å). These two charged residues are located within a hydrophobic pocket in the center of the ␤ 2 -adaptin platform subdomain (28). The location of electrostatically charged residues such as Glu-849 and Glu-902 in this polar patch of ␤ 2 -adaptin is reminiscent of other proteinprotein binding domains (29), and might provide both the specificity and strength required for ␤-arrestin interaction. We first tested whether individually substituting residues Glu-849 and Glu-902 for alanine would affect the association of ␤ 2 -adaptin with ␤-arrestin. GAL4 fusion proteins of the full-length ␤ 2adaptin or E849A and E902A mutants were co-transformed in yeast with ␤-arrestin-2, and the interaction between the ␤ 2adaptin proteins and ␤-arrestin was assessed (Fig. 4). Results show that, although full-length ␤ 2 -adaptin interacts with ␤-arrestin, replacement of glutamic acid residues 849 and 902 by alanine in ␤ 2 -adaptin impaired the association of both mutants with ␤-arrestin. To confirm that residues Glu-849 and Glu-902 were involved in the binding of ␤-arrestin, HA-tagged ␤ 2adaptin-(664 -937) constructs containing wild type or mutant amino acids at positions 849 and 902 were generated and expressed in HEK 293 cells. Cytosolic extracts from transfected cells were incubated with GST alone, or with a GST fusion protein of the C terminus of ␤-arrestin-1 (GST-␤arr-CT), which contains both the ␤ 2 -adaptin and the clathrin binding domain (23). The presence of wild type or mutant ␤ 2 -adaptin proteins and clathrin in the affinity-purified complex was revealed by immunoblotting with either anti-HA or anti-clathrin heavy chain antibodies. Results show that both wild type ␤ 2 -adaptin-(664 -937) and clathrin associated with GST-␤arr-CT, but did not associate with GST alone (Fig. 5). When GST-␤arr-CT was incubated with cytosol from cells expressing the E849A or E902A ␤ 2 -adaptin-(664 -937) mutants, ␤-arrestin binding was eliminated. However, under the same conditions, clathrin binding was unaffected.
The polar residue tyrosine 888 (Tyr-888) in the surfaceexposed hydrophobic patch of the ␤ 2 -adaptin platform shares a hydrogen bound with the residue Glu-902 and is therefore another good candidate to regulate the formation of a ␤ 2 -adaptin/␤-arrestin complex (Fig. 3). Substitution of residue Tyr-888 for alanine in ␤ 2 -adaptin-(664 -937) (Y888A) resulted in a loss in the ability of the GST-␤arr-CT to complex with ␤ 2 -adaptin (Fig. 5). These results suggest that the site for ␤-arrestin binding is contained within the platform subdomain of ␤ 2 -adaptin, and might involve residues Glu-849, Glu-902, and Tyr-888. However, the failure of ␤ 2 -adaptin mutants to bind ␤-arrestin might simply reflect a change in the ␤ 2 -adaptin subdomain structure rather than the removal of critical determinants involved in the association of the two proteins. To rule out this possibility, we examined the ability of ␤ 2 -adaptin-(664 -937)-Y888A, -E849A, and -E902A mutants to bind other endocytic proteins. We hypothesized that if such a structural alteration occurred in the ear domain of ␤ 2 -adaptin, then the mutants may be affected in their ability to bind other high affinity and/or low affinity ␤ 2 -adaptin-interacting-proteins, such as epsin and clathrin (28). GST fusion proteins of wild type and the mutant ␤ 2 -adaptin (Y888A, E849A, or E902A) were incubated with cytosolic extracts from mouse brain. Associated epsin and clathrin in the affinity-purified complex were detected by immunoblotting (Fig. 6). The results indicate that the ␤ 2 -adaptin-(664 -937) wild type, -E849A, and -E902A proteins were equally effective at binding epsin and clathrin. However, we did not detect any association of epsin or clathrin with GST-␤ 2 -adaptin-(664 -937)-Y888A or GST alone. These results indicate that the replacement of Glu-849 and Glu-902 by alanine residues does not significantly alter the tertiary structure of ␤ 2 -adaptin C-terminal subdomain, and suggest that ␤-arrestin directly interacts with these residues. Replacement of the tyrosine residue may result in a more drastic effect on the folding of the molecule because it prevents the interaction of ␤ 2 -adaptin with other endocytic proteins, or Tyr-888 may play a direct role in epsin and clathrin binding. Taken together, our results strongly suggest that Glu-849 and Glu-902 in the C terminus of ␤ 2 -adaptin participate in ␤-arrestin binding.

␤-Arrestin/AP-2 Interaction in GPCR Internalization
focal microscopy (Fig. 7). In the absence of agonist, ␤ 2 AR-GFP was uniformly distributed at the plasma membrane, whereas ␤ 2 -adaptin mutants were found both in the cytosol and at the plasma membrane (Fig. 7, left panels). Upon isoproterenol stimulation of cells expressing ␤ 2 AR-GFP alone, the membrane-delimited fluorescence almost totally disappeared in favor of the emergence of numerous puncta inside the cell (Fig.  7). However, in cells expressing both ␤ 2 AR-GFP and the ␤ 2adaptin-(664 -937) mutant and stimulated with agonist, the receptors did not appear to internalize. This is shown by the lack of significant decrease in receptor fluorescence from the plasma membrane (compare the lower cell expressing both ␤ 2 AR-GFP and ␤ 2 -adaptin-(664 -937) with the upper cell ex-pressing the receptor alone). Similar results were also obtained when the ␤ 2 -adaptin-(592-937) mutant was expressed, although this construct appeared to affect significantly the expression level of ␤ 2 AR (data not shown). The ability of ␤ 2 -adaptin C-terminal constructs to interact with proteins involved in the endocytic process such as clathrin suggests that the expression of ␤ 2 -adaptin-(592-937) or ␤ 2adaptin-(664 -937) might affect the internalization of several different classes of receptors. For instance, the transferrin receptor (TfR), which internalizes via clathrin-coated vesicles, is believed to be linked to the clathrin cages through the interac- were incubated with a GST-␤-arrestin-1 C terminus fusion protein (GST-␤arr-CT), or with GST alone. Affinity-purified proteins (AP) were resolved by SDS-PAGE, transferred onto a nitrocellulose membrane, and Ponceau S-stained for detection of fusion proteins (bottom panel). Membranes were immunoblotted with anti-HA and clathrin antibodies (top and middle panels) for the detection of ␤ 2 -adaptin minigene constructs and endogenous clathrin, respectively. Results show that replacement of tyrosine 888, glutamate 849, and 902 residues with alanine (Y888A, E849A and E902A, respectively) in the HA-␤ 2 -adpaptin 664 -937 construct impaired their association with ␤-arrestin C terminus. Total represents 5% of the total amount of expressed HA-␤ 2adaptin constructs or endogenous clathrin used in each pull-down assay. Results are representative of at least three independent experiments.
Finally, we looked at the ability of ␤ 2 -adaptin-(592-937) and -(664 -937) mutants to block the endocytosis of other GPCRs. We chose the V2R, another receptor that internalizes in a ␤-arrestinand clathrin-dependent manner (8), and compared its internalization pattern to that of the ␤ 2 AR. Cells expressing either receptor with or without the ␤ 2 -adaptin mutants were ␤-Arrestin/AP-2 Interaction in GPCR Internalization stimulated with their respective agonist for 20 min, and the level of receptor internalization was assessed (Fig. 9). The inhibitory effect of the ␤ 2 -adaptin mutants was also compared with another dominant negative construct of the clathrin-mediated pathway, K44A dynamin. When expressed with the ␤ 2 AR, K44A dynamin inhibited 60% of receptor internalization compared with cells expressing ␤ 2 AR alone (Fig. 9A). The ␤ 2adaptin-(592-937) and -(664 -937) mutants, although not as potent inhibitors as the K44A dynamin, were equally effective at blocking ␤ 2 AR (ϳ30%). A similar pattern of inhibition was observed when these dominant negative constructs were used to block the internalization of the V2R. Cells expressing either ␤ 2 -adaptin-(592-937) or -(664 -937) mutants showed a decrease in the V2R internalization of more than 40%. Expression of K44A dynamin also resulted in the inhibition of receptor internalization (80% of inhibition). These ␤ 2 -adaptin mutants also acted as dominant negative for the agonist-mediated internalization of the AT1R (data not shown). When the internalization of AT1R was assessed in the presence of the ␤ 2adaptin-(592-937) mutant, a significant reduction in the endocytosis of the receptors was observed at early time points of agonist stimulation (i.e. a reduction of 50% in receptor internalization was already detected after 2 min of agonist stimu-lation compared with cells expressing wild type receptors alone; data not shown). These data indicate that the association of AP-2, via its ear domain, to receptor/␤-arrestin complexes is important for clathrin-mediated endocytosis of GPCRs. DISCUSSION In the present study, we have identified critical residues in the ␤ 2 -adaptin C-terminal domain responsible for the binding to ␤-arrestins. Expressing an AP-2 ␤-subunit mutant containing the ␤-arrestin-binding domain acts as a dominant negative of GPCR endocytosis without affecting the internalization of other classes of receptors such as the transferrin receptor. These results extend our previous findings that ␤-arrestins are endocytic scaffold proteins and indicate that the association of ␤-arrestin with AP-2 plays an important role in the clathrinmediated internalization of other members of the GPCR family.
We find that residues Glu-849 and Glu-902 in the platform C-terminal subdomain of ␤ 2 -adaptin are involved in the interaction of AP-2 with ␤-arrestin. Mutational analysis of both proteins indicates that residues Glu-849 and Glu-902 in ␤ 2adaptin are potentially linked through ionic interactions with residues Arg-394 and Arg-396 of ␤-arrestin (23). These findings, however, do not exclude the possibility that other residues and/or contact points may be required to stabilize and regulate the association between the two proteins. The platform subdomain of ␤ 2 -adaptin has been shown to contain binding sites for other endocytic adaptor/accessory proteins such as AP180, epsin, and eps15 (28). However, residues in ␤ 2 -adaptin implicated in these interactions seem to differ from those involved in the association of ␤-arrestin. For example, Owen et al. (28) have reported that substituting residue Glu-902 in ␤ 2 -adaptin for an arginine (E902R) had no effect on the binding of AP180, epsin, and eps15. Consistent with these data, we find that a similar mutation in ␤ 2 -adaptin (E902A) retains its binding capability for epsin, but does not associate with ␤-arrestin. Other residues such as Tyr-888 located in the platform subdomain of ␤ 2 -adaptin seem to be involved in many, if not most of these interactions. We showed, however, that replacement of residue Tyr-888 with alanine (Y888A) not only impaired ␤ 2 -adaptin binding to ␤-arrestin but also abrogated its association with clathrin and epsin. Owen et al. (28) also reported that the replacement of residue Tyr-888 by a valine (Y888V) compromised the binding of ␤ 2 -adaptin to clathrin, AP180, epsin, and, to a lesser extent, eps15. This would suggest that Tyr-888 in ␤ 2 -adaptin is a critically conserved residue that acts as a regulatory point of contact for different endocytic proteins. Alternatively, the loss of multiple interactions between different endocytic proteins and the ␤ 2 -adaptin tyrosine mutant may reflect a structural change in the platform subdomain.
Clathrin-mediated endocytosis of membrane proteins requires the coordinate regulation, both spatially and temporally, of multiple protein-protein interactions. For example, the AP-2 complex recruits different accessory proteins and links receptors to CCVs through direct interactions or via accessory/adaptor proteins associated with the AP-2 complex (18,22). Some of these interactions may affect the stability of the receptor in the CCV. The ␤ 2 AR and V2R have been shown to internalize in a ␤-arrestin-dependent fashion (5,8,33), and ␤-arrestin may serve this function by linking the receptors to CCVs by simultaneously binding the ␤-subunit of AP-2 as well as clathrin. Therefore, by preventing the association of both clathrin and AP-2 with ␤-arrestin, this may have additive inhibitory effects on the endocytosis of the receptors compared with conditions where only each individual interaction is blocked. Unexpectedly, the ␤ 2 -adaptin-(664 -937) mutant, which binds poorly to clathrin but retain its ␤-arrestin-binding capability, was able to Agonist-mediated internalization of receptors was assessed as described under "Experimental Procedures" and represents the lost of receptor fluorescence for the ␤ 2 AR after 20 min of isoproterenol stimulation or the fraction of [ 3 H]AVP bound to V2R after the same incubation period that was resistant to acid washes. Results are the mean Ϯ S.E. of three to four or three to five experiments for ␤ 2 AR and V2R, respectively. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.

␤-Arrestin/AP-2 Interaction in GPCR Internalization
block the endocytosis of both receptors as efficiently as the ␤ 2 -adaptin-(592-937) mutant, which contains a high affinity binding site for ␤-arrestin and clathrin. An explanation for this effect could come from the fact that both ␤ 2 -adaptin constructs are blocking the binding of ␤-arrestin to the receptors, thus preventing the targeting of the receptors to CCVs. However, when ␤ 2 -adaptin mutants were expressed with a green fluorescent protein-tagged ␤-arrestin (GFP-␤-arrestin-2) and the ␤ 2 AR, GFP-␤-arrestin-2 was recruited to the receptor at the plasma membrane following agonist stimulation (data not shown). Alternatively, the association of ␤-arrestin with AP-2 and clathrin may be a sequential event. The binding of ␤-arrestin to AP-2 may precede the association of clathrin with ␤-arrestin. Thus, by preventing the latter interaction such as by using ␤ 2 -adaptin dominant negatives containing the ␤-arrestin-binding domain, we might affect the subsequent formation of higher order endocytic complexes.
In conclusion, we have mapped a ␤-arrestin binding domain in ␤ 2 -adaptin. This interaction involves two glutamic acid residues (Glu-849 and Glu-902) in the platform subdomain of the ear of ␤ 2 -adaptin, where other endocytic accessory/adaptor proteins have been shown to interact. Our data provide additional evidence for the role of ␤-arrestins as endocytic scaffold proteins and indicate that the interaction of ␤ 2 -adaptin with ␤-arrestin is important in recruiting GPCRs for clathrin-coated vesicle-mediated endocytosis.