N-terminal Tyrosine Modulation of the Endocytic Adaptor Function of the β-Arrestins*

The highly homologous β-arrestin1 and -2 adaptor proteins play important roles in the function of G protein-coupled receptors. Either β-arrestin variant can function as a molecular chaperone for clathrin-mediated receptor internalization. This role depends primarily upon two distinct, contiguous C-terminal β-arrestin motifs recognizing clathrin and the β-adaptin subunit of AP2. However, a molecular basis is lacking to explain the different endocytic efficacies of the two β-arrestin isoforms and the observation that β-arrestin N-terminal substitution mutants can act as dominant negative inhibitors of receptor endocytosis. Despite the near identity of the β-arrestins throughout their N termini, sequence variability is present at a small number of residues and includes tyrosine to phenylalanine substitutions. Here we show that corresponding N-terminal (Y/F)VTL sequences in β-arrestin1 and -2 differentially regulate μ-adaptin binding. Our results indicate that the β-arrestin1 Tyr-54 lessens the interaction with μ-adaptin and moreover is a Src phosphorylation site. A gain of endocytic function is obtained with the β-arrestin1 Y54F substitution, which improves both the β-arrestin1 interaction with μ-adaptin and the ability to enhance β2-adrenergic receptor internalization. These data indicate that β-arrestin2 utilizes μ-adaptin as an endocytic partner, and that the inability of β-arrestin1 to sustain a similar degree of interaction with μ-adaptin may result from coordination of Tyr-54 by neighboring residues or its modification by Src kinase. Additionally, these naturally occurring variations in β-arrestins may also differentially regulate the composition of the signaling complexes organized on the receptor.

receptors desensitize in the continued presence of ligand, and one of the contributing mechanisms involves ligand-activated GPCRs binding either of the two ubiquitously expressed ␤-arrestin proteins (1 or 2) (1)(2)(3)(4). Evidence indicates that in addition to competitively preventing G protein activation, the ␤-arrestins can serve both as endocytic adaptors for the GPCRs and simultaneously initiate a signaling phase of their own. The latter phase originates from complexes nucleated by receptor-activated ␤-arrestins and agonist-occupied receptors (5,6). ␤-arrestin-receptor complexes form signaling scaffolds in clathrin rich areas of the plasma membrane (7)(8)(9) by binding partners, such as Src, extracellular signal-regulated kinase (ERK), Akt, and JNK kinases (see Ref. 10 for recent review).
Clathrin-mediated endocytosis is an uptake mechanism used by many membrane proteins (11)(12)(13), and adaptors mediating cargo recognition play a central role in this process. The main clathrin adaptor AP2 contains four distinct subunits, including the homologous 100-kDa ␣ and ␤ chains that interact with clathrin, a 50-kDa chain, which acts as the predominant cargo adaptor, and a smaller 19-kDa chain (14). Membrane proteins containing tyrosine and di-leucine based internalization motifs are recognized and sorted by the -adaptin subunit of AP2 (15)(16)(17). However, ligand-activated GPCRs presumably do not directly interact with -adaptins, but instead use ␤-arrestin1 and -2 as primary adaptors. ␤-arrestins have been shown to bind the ␤-subunit of AP2, clathrin, and also recruit Src kinases that regulate internalization checkpoints (7,18,19). Src kinases phosphorylate tyrosine residues on clathrin pit structural proteins including clathrin itself and dynamin, which enables receptor-laden vesicles to pinch off the plasma membrane (19 -23).
␤-arrestin1 and -2 differ considerably in their ability to internalize ␤2AR, one of many GPCRs for which this disparity occurs (24), with ␤-arrestin2 being considerably better than ␤-arrestin1 in this role. Given the high degree of ␤-arrestin homology, the marked difference in their endocytic abilities must arise from minor variations in their structures. The ␤-ar-restin1/2 N termini contain several positions where tyrosine is not conserved, suggesting possible regulatory sites underlying some of the differences between their endocytic properties. In this study we show that the conservative substitution of ␤-ar-restin1 tyrosine 54 by phenylalanine (corresponding to phenylalanine 55 in ␤-arrestin2) increases ␤-arrestin1 interaction with -adaptin and enhances ␤-arrestin1-mediated ␤2AR endocytosis to the levels mediated by ␤-arrestin2. This phenylalanine substitution also results in the loss of a Src phosphorylation site. These findings provide a molecular basis for the preferential internalization of the ␤2AR via ␤-arrestin2 and uncover new regulatory functions originating from a single substitution of the ␤-arrestin N terminus.

EXPERIMENTAL PROCEDURES
Materials-Monoclonal antibody against AP-50/2 (-adaptin) was from BD Biosciences. Anti-hemagglutinin antibody was obtained from Roche Pharmaceuticals, X22 anticlathrin antibody was purified from a hybridoma cell line (9,25). Texas Red transferrin was purchased from Molecular Probes. Cell culture reagents were obtained from Life Sciences Technologies and the yeast two-hybrid vectors and reagents from Clontech.
Plasmid Constructs-cDNA vectors for the ␤2AR/␤-arres-tin2 full-length and N-terminal fusion proteins used for binding, adenylyl cyclase, and sequestration measurements were constructed in the following manner using the vector containing ␤2AR/GFP (26). A 5Ј primer containing a 9-base overhang, a SalI restriction site, and 18 -21 bases of ␤-arrestin2, and a 3Ј primer with a 9-base overhang, a BglII restriction site followed by a stop codon, and 18 -21 bases of ␤-arrestin2 were used to generate cDNA inserts by polymerase chain reaction (PCR) employing rat ␤-arrestin2 as a template. The fragments were digested using SalI and BglII and ligated into the ␤2AR/GFP vector cut with SalI and BamH1. GFP variants of the corresponding fusion proteins were similarly constructed by eliminating the stop codon from the 3Ј primer. Fusion proteins of ␤-arrestin and GFP containing point mutations in the N terminus were constructed using a similar strategy and cloned into the ␤-arrestin/GFP vector (8). The cDNA for glutathione S-transferase (GST) fusion protein was also constructed by PCR. A fragment derived from the N-terminal domain (1-80) of the human ␤-arrestin2 was cloned into EcoRI and XhoI of pGEX-4T-1 (Amersham Biosciences AB). Sequences were confirmed using an automated ABI DNA sequencer. Point mutations were generated in ␤-arrestin1(Y54F) and ␤-arrestin2(F55Y) using the QuikChange site-directed mutagenesis system (Stratagene).
Receptor Binding Assays-Agonist and antagonist binding assays using 125 I-CYP were performed as described (27).
Adenylyl Cyclase Assays-Measurement of whole cell cAMP was performed using Dowex and Alumina column chromatography as described (28).
Sequestration Assays-Measurement of receptor sequestration by flow cytometry was performed as described in (27). Measurement of sequestration by [ 3 H]CGP-12177 was performed as follows. HEK-293 cells plated at a density of 3 ϫ 10 6 /100-mm dish were transfected with receptor cDNA augmented with empty vector cDNA to a total of 5 g using a calcium phosphate protocol (28). The cells were trypsinized 5-6 h later and plated at a density of 750,000 cells/well in a 6-well plate and incubated in minimal essential medium (MEM) containing 10% fetal bovine serum. After 48 h, the cells were washed in phosphate-buffered saline and the media replaced with warm MEM or MEM containing 500 nM isopro-terenol. After 30 min at 37°C in a 5% CO 2 incubator the cells were washed three times in ice-cold phosphate-buffered saline, covered with 1 ml of ice-cold phosphate-buffered saline, and suspended by trituration at a density of 1.25 ϫ 10 6 /ml. 100 l volumes of cells were incubated in 15 ml of polypropylene tubes (Fisher) with 50 l of 30 nM [ 3 H]CGP-12177 (Amersham Biosciences) with or without 1 M propranolol for 90 min at 4°C. Cells were washed and collected on a Brandel harvester with Whatman GF/C filters for ␤ analysis.
Yeast Two-hybrid Screening-Fusion genes expressing either ␤-arrestin2 or the N-terminal 80 amino acids of ␤-arrestin2 were transformed into Tyr-187 yeast strains or AH-109 as described previously (9). The ability of the expressed proteins to interact in yeast was assessed by a growth assay using histidine-or adenine-deficient media as opposed to a primary ␤-galactosidase activity screen as described previously (9). Sequences inserted into the yeast vectors were verified by dideoxy sequencing.
Peptide Phosphorylation-Peptides were synthesized (Sigma Genosys) and diluted to a working concentration of ϳ0.5-1 mM in sterile water. Reactions were performed according to the manufacturer's protocol provided with the purified, active c-Src (Upstate Biotechnology). Briefly, 20 units of active c-Src was added to a reaction mixture containing substrate peptide (100 M/assay), [␣-32 P]ATP (1 mCi/100 l; 3000 Ci/mmol; PerkinElmer Life Sciences), 1ϫ Src reaction buffer (100 mM Tris-HCl, pH 7.2, 125 mM MgCl 2 , 2 mM EGTA, 0.25 mM sodium orthovanadate, 2 mM dithiothreitol). After 10 min at 30°C, 40% trichloroacetic acid was added to the mixture and incubated at room temperature for 5 min. The reaction was transferred to the center of a 2 ϫ 2 cm 2 P81 Whatman paper square. The squares were washed five times for 5 min each wash in 0.75% phosphoric acid. The squares were then washed once in acetone, transferred to a scintillation vial with scintillation mixture, and incorporated counts were measured in a scintillation counter. Background reactions (no enzyme) for each peptide were performed. This recorded value (no enzyme added) was then subtracted from the respective reactions. The normalized (substrate concentration), background-subtracted values were plotted using GraphPad Prism. The positive control was a peptide optimized for Src substrate specificity (Upstate Biotechnology).
␤-Arrestin1 Purification and Phosphorylation-␤-Arres-tin1-HT (histidine tagged) and ␤-arrestin1(Y54F)-HT were subcloned into the pET-21b vector (Novagen) and expressed in DH5␣ cells (37°C) with a 16 h isopropyl 1-thio-␤-D-galactopyranoside induction. ␤-arrestins were partially purified from bacterial lysates using Probond nickel-charged resin (Invitrogen) according to the manufacturer's protocol. Phosphorylation of ␤-arrestin1 proteins (done with increasing amounts of purified protein) was performed with the same reaction conditions as with the peptide reactions (described above) for 10 min at 30°C. SDS sample buffer (Invitrogen) was added to the reac-tions, and they were immediately boiled for 3 min and loaded to a 10% polyacrylamide gel. Separated proteins were transferred to nitrocellulose and exposed directly to film or blotted for antityrosine (PY99, SantaCruz Biotechnology) or for ␤-arrestin1 (BD Biosciences). Western blots were visualized using ECL (Pierce).
␤-Arrestin Phosphorylation in HEK-293 Cells-␤-Arrestin-FLAG cDNAs were transfected into HEK-293-␤2AR stable cell lines grown in 12-well plates using Lipofectamine (Invitrogen). Forty-eight hours post-transfection, cells were stimulated with 100 nM isoproterenol for 1 min, lysed in mammalian protein extraction reagent (Pierce) with protease inhibitors added (Roche). Lysates were diluted 10-fold with lysis buffer (1% Triton X-100 in Tris-buffered saline: 25 mM Tris-HCl, pH 7.5, 150 mM NaCl). M2 anti-FLAG conjugated agarose (Sigma) beads were added to the diluted lysates and rocked overnight at 4°C. Beads were washed five times in Tris-buffered saline, and 2ϫ SDS sample buffer (Invitrogen) was added directly to the beads and gently mixed. The supernatants were resolved by 10% SDS-PAGE, transferred to nitrocellulose, and probed with antibodies against phosphotyrosines (Santa Cruz Biotechnology) and the FLAG tag (Sigma). Blots were visualized with ECL (Pierce), and densitometry was performed with the FluoroS Multi-Imager (Bio-Rad). The phosphotyrosine signal was confirmed by directly treating the immunoprecipitates with 10 -20 units of YOP (Yersinia outermembrane protein) tyrosine phospha-tase (Upstate) at 30°C for 3 min followed by Western analysis as described above.
Fluorescence and Confocal Microscopy-Fluorescence microscopy and immunostaining of hemagglutinin-tagged receptor or clathrin was performed as described (9,26). Images were obtained on a Zeiss LSM-510 confocal microscope.
We employed a GST fusion approach to test further whether ␤-arrestin2 could bind to an AP2 complex in vitro. This requires modifying the C terminus of ␤-arrestin2 that contains clathrin and ␤-adaptin binding motifs, because isolation of -adaptin might occur from its association with a clathrin-␤adaptin complex. In the C-terminal ␤-arrestin2 truncation mutant, ␤-arrestin2(N1-372) the clathrin and ␤-adaptin bind-ing motifs are absent. Columns containing C-terminal GST fusion proteins of either full-length ␤-arrestin2 or ␤-arrestin2(N1-372) were exposed to a mouse brain lysate. Anti -adaptin antibody immunoblotting of the retained protein eluate from either the ␤-arrestin2 or ␤-arrestin2(N1-372) column demonstrated a 50-kDa band characteristic of -adaptin. The relative amount of each GST-fused ␤-arrestin protein present on a column is shown by the Coomassie staining (Fig.  2B, lower panel). This result suggests that a site responsible for -adaptin binding exists in ␤-arrestin2 that is independent of the C-terminal ␤-adaptin and clathrin binding motifs. Surprisingly, ␤-arrestin2(N1-372) bound 3-fold more -adaptin than full-length ␤-arrestin2. Han et al. (31) have hypothesized that C-terminal ␤-arrestin truncation mutants assume pre-activated conformations. This suggests that a -adaptin binding motif of ␤-arrestin2 may become fully exposed only upon ␤-arrestin-receptor binding, in a manner analogous to that postulated for the ␤-arrestin-␤-adaptin binding motif (31).
We next passed a mouse brain lysate through a column containing a C-terminal GST fusion protein of ␤-arrestin2(N1-80) to determine whether the ␤-arrestin2 N terminus could be responsible for the previous observations. The immunoblot corresponding to GST-␤-arrestin2(N1-80), but not GST alone, also yielded a 50-kDa band when probed with antiadaptin antibody (Fig. 2C). The immunostaining of the total brain lysate, which contains -adaptin as part of the AP2 complex, is shown as positive control.
If residues 55-58 of ␤-arrestin2 are implicated in -adaptin binding and act as a receptor sorting motif, then fusion of the N-terminal part of ␤-arrestin2 to the ␤2AR could affect the cellular localization of the receptor. To test this hypothesis we constructed a ␤2AR fused to the first 78 residues of ␤-ar-restin2, ␤2AR/␤-arrestin2(N1-78) and characterized its response to agonist.
We first assessed some pharmacological properties of ␤2AR/ ␤-arrestin2(N1-78) compared with the wild type ␤2AR. The two receptors have equal affinities for the antagonist iodocyanopindolol ([ 125 I]CYP, Table 1) and the agonist isoproterenol (as measured with membrane preparations of receptors by displacement of 125 I-CYP). They also express similarly in terms of the proportion of receptors in the G protein-bound high affinity state (Table 1), and despite a shift in the EC 50 for cAMP production (Table 1), the ␤2AR/␤-arrestin2(N1-78) chimera-ac- . Images are representative of the yeast growth in each dish in two independent experiments. B, a mouse brain homogenate was purified over a column containing either GST alone, GST-␤-arrestin2 fulllength, or rat ␤-arrestin2 truncated after residue 372 (proximal to the clathrin and ␤-adaptin binding motifs). The GST-conjugated proteins were removed from the columns using reduced glutathione, and equal total protein aliquots of eluate were assessed for GST-␤-arrestin expression using SDS-PAGE and Coomassie protein staining. The immunoblot revealed a characteristic 50-kDa band that corresponded to the -adaptin subunit of AP2 (see the lysate fraction in the right-most column). Data are representative of two independent experiments. C, a mouse brain homogenate was purified over a column either containing GST beads alone or containing a GST fusion protein of the N-terminal 80 amino acids of rat ␤-arrestin2. No -adaptin was isolated using GST alone (left-most column), but a 48 -50-kDa protein corresponding to the size of -adaptin was observed with the antibody in the brain lysate (right-most column) and the proteins retained using the GST-␤-arrestin2(N1-80) (center column). Data are representative of three independent experiments.

TABLE 1 Pharmacological profile and basal sequestration of ␤2AR chimeras in HEK-293 cells
Binding of ͓ 125 I͔CYP or CGP-12177 was measured on membrane preparations of HEK-293 cells expressing the ␤2AR or ␤2AR/␤-arrestin2(N1-78). Results are representative of three independent experiments. Competitive displacement of ͓ 125 I͔CYP by isoproterenol in membrane preparations derived from HEK-293 cells that were transfected with bovine Gs-␣ subunits and either ␤2AR or ␤2AR/␤-arrestin2(N1-78). Normalized adenylyl cyclase activity was measured in HEK-293 cells in response to increasing concentrations of isoproterenol for those cells transfected with either of the receptor chimeras. Basal sequestration of the two receptor isoforms in HEK-293 cells as measured by ligand binding was not significantly different. Results are presented as mean Ϯ S.D.

An Arrestin Endocytic Signal for -Adaptin
tivated adenylyl cyclase with similar efficacy to the ␤2AR (not shown).
Upon agonist treatment ␤2AR/␤-arrestin2(N1-78) sequesters to a greater extent than ␤2AR as measured by flow cytometry (data not shown). This suggests that the ␤-arrestin2(N1-78) peptide may contain a sorting motif that underlies this enhanced ability to sequester. To identify the amino acids responsible, sequestration of ␤2AR/␤-arrestin2 fusion proteins constructed using different peptide segments of the ␤-arrestin2 N terminus was assessed by the binding of the hydrophilic ligand [ 3 H]CGP-12177 (Fig. 3A). The ability of the peptide fusions to augment sequestration disappeared with truncation at Phe-55, suggesting that amino acids 50 -60 may contain a canonical sorting signal.
If ␤-arrestin2(N1-78) does contain an internalization signal normally employed by ␤-arrestin2 for endocytosis of GPCRs, then ␤2AR/␤-arrestin2(N1-78) should traffic to an endosomal vesicular compartment containing wild type ␤2AR. In Fig. 3B isoproterenol-activated wild type ␤2AR-GFP is observed to redistribute from the plasma membrane to endosomes that contain a fluorescent transferrin marker (32). In the absence of agonist a GFP fusion protein of ␤2AR/␤-arrestin2(N1-78) is also predominantly plasma membrane bound (Fig. 3C) like the ␤2AR (26). When expressed in the same cell, ␤2AR-RFP (Fig.  3C, lower left panel) and ␤2AR/␤-arrestin2N(1-78/GFP) (Fig.  3C, upper right panel) redistribute to a common intracellular vesicular compartment in response to agonist (Fig. 3C, lower  right panel). The type of fluorescent protein tag (RFP versus GFP) attached to the ␤2AR does not influence its endosomal localization (data not shown). These data establish that the ␤2AR and ␤2AR/␤-arrestin2(N1-78) share a similar pattern of cellular distribution before and after agonist stimulation. Alto-gether, our data are consistent with a hypothesis that the N terminus of ␤-arrestin2 functions as an FVTL -adaptin binding motif (YVTL in ␤-arrestin1) that could participate in the regulation of ␤2AR internalization.
An interesting difference between the two putative -adaptin binding motifs is the potential for regulation of YVTL in ␤-arrestin1 by tyrosine phosphorylation. However, there are four tyrosines present in the N-terminal part of ␤-arrestin1 that are not conserved in ␤-arrestin2 (Fig. 4A). Using an in vitro assay, we investigated whether ␤-arrestin1 is a Src substrate because ␤2AR stimulation leads to a rapid association of the activated tyrosine kinase Src to the receptor in a ␤-arrestin1-dependent fashion (19).
To assess the phosphorylation of the N-terminal tyrosines in ␤-arres-tin1 that are not conserved in ␤-ar-restin2, we synthesized the corresponding peptides (Fig. 4, A and B). The site prediction program NetPhos (Expasy Tools) had scored two of the ␤-ar-restin1 N-terminal tyrosines as phosphorylation sites (Fig. 4B). Even though it did not score highly as a phosphorylation candidate, ␤-arrestin1 peptide 2 was the only substrate that underwent significant in vitro Src phosphorylation (Fig. 4B). Interestingly, peptide 2 contains the canonical Y54XX sorting signal and corresponds to the F55XX sequence that we described above as a potential -adaptin binding motif in ␤-arrestin2.
Examination of the crystal structure of inactive ␤-arrestin1 indicates that Tyr-54 is only partially exposed. Therefore we sought to establish whether this residue could be phosphorylated by Src in full-length ␤-arrestin1. Co-incubation in vitro of purified Src and ␤-arrestin1 revealed that Src was able to phosphorylate full-length ␤-arrestin1 (Fig. 4C), and the single point mutation Y54F was sufficient to prevent it (Fig. 4D).
To confirm that similar Src phosphorylation occur in cells, we measured the tyrosine phosphorylation of over-expressed ␤-arrestin1 before and after exposing the ␤2AR to agonist. The specificity of the phosphotyrosine antibody used in this assay was assessed by immunoprecipitating FLAG-tagged ␤-arrestin1 and treating the immunoprecipitates with YOP tyrosine phosphatase (Fig. 5A). Tyrosine phosphatase treatment almost completely eliminated the anti-phosphotyrosine PY99 signal confirming the specificity. We next observed in cells that ␤-arrestin1 as well as ␤-arrestin1(Y54F) are both constitutively tyrosine phosphorylated (Fig. 5B). However, isoproterenol exposure leads to a significant increase in ␤-arrestin1 tyrosine phosphorylation that is not observed in ␤-arrestin1(Y54F) (Fig. 5, B and C). These data suggest that ␤-arrestin1 is tyrosine phosphorylated on residue 54 in response to ␤2AR stimulation.
Using GST pull-down, we evaluated the interaction of ␤-arrestin1, ␤-arrestin1(Y54F), and ␤-arrestin2 with -adaptin to determine whether the gain of function of ␤-arrestin1(Y54F) in promoting sequestration correlated with an increase in its affinity toward -adaptin. The three GST ␤-arrestin fusion proteins interact equally well with clathrin, and the ␤-arrestin1 GST fusion interacts much less well with -adaptin than the ␤-arrestin2 fusion protein (Fig.  5E). Remarkably, ␤-arrestin1(Y54F) is able to pull-down -adaptin like ␤-arrestin2 (Fig. 5E). No clathrin or -adaptin was retained by the GST alone (Fig. 5E, left lane control). These data strongly support the hypothesis that the FVTL sequence in ␤-arrestin2 mediates an interaction with the -adaptin subunit of the AP2 complex during receptor endocytosis. Moreover it suggests that the tyrosine at position 54 of ␤-arrestin1 reduces the extent of this interaction.

DISCUSSION
Sequence differences between the two ␤-arrestin isoforms are highly conserved across species, suggesting that differences in their activities may arise even from seemingly conserved amino acid substitutions. To date, few differences between the regulatory behaviors of the ␤-arrestins have been demonstrated, but cell-based studies have indicated a preference of many GPCRs for ␤-arrestin2 during clathrin-mediated internalization (24).
Our results indicate that the natural Tyr to Phe variation between position 54 in ␤-arrestin1 and position 55 in ␤-arres- tin2 generates differences in ␤-arrestin binding to -adaptin and ␤-arrestin-promoted ␤2AR internalization. Interestingly, despite ␤-arrestin1 Y54LTV defining a canonical YXX -adaptin binding motif, ␤-arrestin1 has less affinity for -adaptin than ␤-arrestin2. However, the ␤-arrestin1(Y54F) mutant binds -adaptin as well as ␤-arrestin2 and is a gain of function mutation for ␤2AR sequestration. According to available structural information, in the absence of tyrosine 54 phosphorylation, this same tyrosine is presumably engaged via a hydrogen bond with Arg-52 (31), constraining its availability for interaction with -adaptin. Crystal structure of the ␤-arrestin indicates that the Tyr/Phe of this putative motif is not fully exposed, however the in vitro and in cellulo phosphorylation of Tyr-54 by Src demonstrate the accessibility of this residue.
Src-mediated ␤-arrestin1 phosphorylation appears within a minute of receptor activation, and therefore most likely precedes the recruitment of activated ␤-arrestin-receptor complexes to coated pits, where Src is known to phosphorylate at a minimum both clathrin and dynamin (20 -22). Notably, the Src inhibitor PP2 does not perturb the recruitment rate of ␤-arrestin1 to the ␤2AR (data not shown). Therefore, a Src-dependent Tyr-54 ␤-ar-restin1 phosphorylation could potentially serve either as a negative regulator for stabilizing AP2/␤-ar-restin1/receptor scaffolds or possibly represents a means for recruiting SH2-directed signaling proteins modulating the nature of signaling scaffold organized by ␤-arrestins. Such a prospect is of specific interest for receptors like the mGluR1a, which can recruit, depending on the nature of the ligand, either of the ␤-arrestins or selectively only ␤-ar-restin1 (34,35).
Examination of the crystal structure of the inactive ␤-arrestin1 indicates that a direct -adaptin interaction via the sequence (Y/F)XXL alone would require a major conformational change in ␤-arrestin to unmask the leucine upon receptor binding. Evidence suggests that receptor-activated ␤-arrestin can undergo substantial conformational rearrangement with receptor activation (36). However, we observed an interaction between ␤-arrestin and -adaptin in the absence of receptor activation, suggesting a less substantial molecular rearrangement, but one also consistent with predicted ␤-arrestin behavior (36). Identifying the aggregate of binding determinants on the ␤-arrestin surface that engage -adaptin pre-and post-receptor activation will most probably require co-crystallization of the participating components including adaptins, ␤-arrestins, and receptors.
The ␤-arrestin2 interaction with bothand ␤-adaptin raises the question of the chronology for these interactions. The proposed ␤-arrestin2 C-tail conformational change upon receptor binding likely occurs before the tail is able to interact with ␤-adaptin (31) through arginine residues 394/396 (18). Indeed, arginine 394 forms part of the ␤-arrestin2 polar core and is thus unavailable for binding ␤-adaptin unless a conformational rearrangement occurs in this region secondary to receptor interaction (31). Our data show that inactive ␤-arres-tin2 can interact with the -subunit. Therefore inactive ␤-ar-restin2 might first bind AP2 through -adaptin. Next, a stronger interaction with AP2 would follow because of the additional ␤-adaptin binding that occurs after receptor-mediated conformational changes of ␤-arrestin2.
An increasing number of signaling proteins like mitogenactivated protein kinase (MAPK), JNK, and NF-B, have been Results are representative of three independent experiments and presented as means Ϯ S.E. *, p Յ 0.05. E, mouse brain homogenates were purified over a column containing either GST alone, GST-␤-arrestin1, GST-␤-arrestin1(Y54F), and GST-␤-arrestin2. The GST-conjugated proteins were removed from the columns using reduced glutathione, and equal total protein aliquots of eluate were resolved by SDS-PAGE. The loading of GST fusion proteins was confirmed by Coomassie staining (bottom panel). The three GST-␤-arrestins were equally well able to interact with clathrin (top panel). GST-␤-arrestin1 had a significantly reduced ability to retain -adaptin compared with GST-␤-arrestin1(Y54F) and GST-␤-arrestin2 (middle panel). No immunoreactivity for clathrin or -adaptin was detected in the control line containing GST alone (left-most column). Immunoblots revealed characteristic 50 and 180 kDa bands corresponding to the -adaptin subunit of AP2 and the clathrin heavy chain (see the lysate fraction in the right-most column). Data are representative of two independent experiments. associated with scaffolds formed by receptors and ␤-arrestins. The demonstration that ␤-arrestin-dependent signaling requires clathrin-dependent localization of receptor-␤-arrestin complexes suggests sequence differences between ␤-arrestin1 and 2 that modulate the membrane localization of activated receptors could affect this novel, nonclassical means of GPCR signal transduction. Consequently, elucidating the principles underlying the dynamics of receptor-␤-arrestin scaffold formation will help explain the G-protein-independent cellular response to a large variety of agonists ranging from ions to hormones.