Dual Role of the β2-Adrenergic Receptor C Terminus for the Binding of β-Arrestin and Receptor Internalization*

Homologous desensitization of β2-adrenergic and other G-protein-coupled receptors is a two-step process. After phosphorylation of agonist-occupied receptors by G-protein-coupled receptor kinases, they bind β-arrestins, which triggers desensitization and internalization of the receptors. Because it is not known which regions of the receptor are recognized by β-arrestins, we have investigated β-arrestin interaction and internalization of a set of mutants of the human β2-adrenergic receptor. Mutation of the four serine/threonine residues between residues 355 and 364 led to the loss of agonist-induced receptor-β-arrestin2 interaction as revealed by fluorescence resonance energy transfer (FRET), translocation of β-arrestin2 to the plasma membrane, and receptor internalization. Mutation of all seven serine/threonine residues distal to residue 381 did not affect agonist-induced receptor internalization and β-arrestin2 translocation. A β2-adrenergic receptor truncated distal to residue 381 interacted normally with β-arrestin2, whereas its ability to internalize in an agonist-dependent manner was compromised. A similar impairment of internalization was observed when only the last eight residues of the C terminus were deleted. Our experiments show that the C terminus distal to residue 381 does not affect the initial interaction between receptor and β-arrestin, but its last eight amino acids facilitate receptor internalization in concert with β-arrestin2.

least for some receptors) the continuous presence of agonist (4). GRKs are unique among serine/threonine kinases in that they do not recognize a well-defined consensus sequence but instead show high specificity for agonist-activated receptors. This lack of a consensus sequence has made mapping of phosphorylated residues in G-protein-coupled receptors difficult. For example, the residues phosphorylated by GRK2 in the ␤ 2 -adrenergic receptor have been mapped to the C terminus, either between amino acids 384 and 411 (5) or between amino acids 355 and 364 (6 -10). It is now clear from a variety of studies (see Ref. 11 for a review) that ␤-arrestins do not simply act as "phosphoreceptor-specific antibodies." Rather, the interaction of ␤-arrestins with GRK-phosphorylated receptors is believed to lead to a conformational change in the ␤-arrestin molecule, which enables it to bind to other parts of the receptor with higher affinity. Furthermore, it has also been demonstrated that ␤-arrestins sense the activated conformation of the receptor (4,12). For example, ␤-arrestin mutants have been described that do not require GRK-mediated phosphorylation to bind to a receptor but still only interact with a receptor when it is activated by agonist (12). The molecular determinants on the receptor molecule for this exquisite selectivity are still unknown. Pull-down assays using fusion proteins have suggested a role for the third intracellular loop for arrestin interaction with muscarinic, ␣ 2 -adrenergic, ␦-opioid, serotonin 2A , and vasopressin V1 receptors, respectively (13)(14)(15)(16)(17)(18)(19). Arrestin translocation assays have shown a role for the second intracellular loop of the 5-HT 2C , ␤ 2 -adrenergic, ␣ 2A -adrenergic, and neuropeptide Y2 receptors (20). Moreover, it has been reported that the C terminus of the dopamine D1 receptor imposed an inhibitory effect on arrestin binding to the receptor, which could be relieved by its GRK-mediated phosphorylation (21).
In this study, we set out to investigate the role of the ␤ 2 -adrenergic receptor (␤ 2 AR) C terminus for the interaction with ␤-arrestin2. To this end, we assessed the internalization and ability to interact with ␤-arrestins of a set of ␤ 2 AR variants containing various truncation or substitution mutations in the C-terminal tail.

DNA Constructs and Transient Expression in HEK293 Cells-
The amino acid sequences of the various ␤ 2 AR mutants are shown in Fig. 1. The phosphorylation-deficient ␤ 2 AR (PD), lacking all putative phosphorylation sites for GRKs, but not * This work was supported in part by SFB487 "Regulatory Membrane Proteins," by a Leibniz grant from the DFG, and by the Fonds der Chemischen Industrie. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1  protein kinase A, has been described previously (22). Chimeras between the wild-type and the PD receptor (PD-N and PD-C) were constructed by taking advantage of a unique EcoRV restriction site occurring in the open reading frame of the ␤ 2 AR, which enabled us to construct chimeras between the wild-type and the PD receptor. The truncated receptors ⌬C2, ⌬C2-PD, and ⌬Asn405 were constructed by polymerase chain reaction using either the wild-type or PD ␤ 2 AR as a template. ␤ 2 AR mutants were either N-terminally tagged with a FLAG tag, or C-terminally tagged with YFP as described for the wildtype receptor (4). All DNAs were cloned in the pcDNA3 expression vector (Invitrogen, Karlsruhe, Germany) and verified by sequencing. pcDNA3 expression vectors for CFP-and YFPtagged bovine ␤-arrestin2 (␤Arr2-CFP, ␤Arr2-YFP) and human GRK2 have been described previously (4). Human embryonic kidney (HEK293) cells were transfected with Effectene (Qiagen, Hilden, Germany) if not otherwise indicated, and were allowed to express the recombinant proteins for 48 h. Receptor expression levels under these conditions were between 1 and 5 pmol/mg membrane protein if not stated otherwise. Western blots of transfected cells using a polyclonal antibody against GRK2 (sc-562, Santa Cruz Biotechnology, Heidelberg, Germany) showed that GRK2 was overexpressed by at least a factor of 10. The amount of overexpression of arrestin could not be measured because commercially available antibodies do not react well with ␤-arrestin2. Blotting for GFP (Clontech, Saint-Germain-en-Laye, France) showed that GRK2 and arrestin overexpression were independent of the identity of the receptor mutant cotransfected and correlated well with each other. Measurement of Receptor-Arrestin Interaction-FRET was measured in single HEK293 cells as described previously (4). Briefly, cells were transiently cotransfected with YFP-tagged wild-type or mutant ␤ 2 AR, ␤Arr2-CFP, and GRK2. Cells were observed on an Axiovert200 inverted microscope and fluorescence intensities at 480 Ϯ 20 and 535 Ϯ 15 nm were measured with a dual emission photometric system (Till Photonics, Gräfelfing, Germany). Kinetics was fitted using the monoexponential function built into Origin (OriginLab, Northampton, MA).
␤-Arrestin translocation was analyzed in HEK293 cells, which were cotransfected with FLAG-tagged wild-type or mutant ␤ 2 AR, ␤Arr2-YFP, and GRK2. The subcellular distribution of ␤Arr2-YFP was monitored by live cell confocal microscopy performed on a Leica TCS SP2 system as described (4).
Radioligand Binding Assays-Binding of 125 I-cyanopindolol ( 125 I-CYP; GE Healthcare, Munich, Germany) to FLAG-tagged wild-type or mutant ␤ 2 AR was determined in membranes prepared from transiently transfected HEK293 cells as described (23). K D values for 125 I-CYP binding in saturation experiments (5-400 pM) were calculated by nonlinear curve fitting. K i values for isoproterenol were determined in competition binding experiments using 10 Ϫ11 -10 Ϫ4 M isoproterenol. 10 M GTP was included to convert all receptors into a low affinity state.
Receptor Internalization Assays-HEK293 cells were transiently transfected with FLAG-tagged wild-type or mutant ␤ 2 AR (and ␤Arr2-YFP and GRK2 where indicated), and seeded into 6-well plates coated with poly-D-lysine. After stimulation with 1 M isoproterenol for 0, 5, 10, 20, and 30 min at 37°C, cells were kept on ice, and cell surface receptors were quantified by [ 3 H]CGP12,177 (GE Healthcare) binding as described (4). Care was taken to ensure approximately equivalent expression of all receptor mutants compared within one experiment, as receptor expression levels could affect the extent of internalization.
Receptor Phosphorylation Assays-HEK293 cells transiently cotransfected with FLAG-tagged wild-type and mutant ␤ 2 AR and human GRK2 were labeled with [ 32 P]orthophosphate, treated (or not) with 1 M isoproterenol for 5 min, and receptors were immunoprecipitated with M2 anti-FLAG-Sepharose (Sigma-Aldrich) and subjected to SDS-polyacrylamide gel electrophoresis, as described previously (24,25). Incorporated radioactivity was quantified by phosphorimager analysis. Equal expression of the mutant receptors was verified in parallel by Western blots with M2 anti-FLAG antibody (Sigma-Aldrich).
Adenylyl Cyclase Assays-To determine receptor-stimulated adenylyl cyclase activity, cAMP production by freshly prepared membranes (100 g of protein) from transiently transfected HEK293 cells was measured for 30 min as described (23). Care was taken that similar expression levels were achieved for wildtype and mutant receptors. Concentration-response curves (3 nM to 50 M isoproterenol) were analyzed by nonlinear curve fitting.

RESULTS
Ligand Binding of the Receptor Mutants-It has previously been shown that a mutant ␤ 2 AR, in which all eleven serine and threonine residues distal to amino acid 354 are mutated, is not phosphorylated by GRK in response to agonist treatment (22). Starting from this phosphorylation-deficient (PD) mutant, we constructed further mutants (sequences are shown in Fig. 1) to analyze the role of the C terminus for ␤-arrestin2 binding and receptor internalization in more detail. All receptor mutants ( Fig. 1) bound 125 I-cyanopindolol with fairly similar affinities (although the affinity of the ⌬C2-YFP mutant was significantly lower than some of the PD mutants), and addition of a C-terminal YFP tag did not influence ligand binding ( Table 1). The ability of the mutants to bind isoproterenol in competition experiments with 125 I-cyanopindolol was also fairly similar although addition of a C-terminal YFP tag lowered the affinity FIGURE 1. Sequence alignment of the human ␤ 2 AR mutants used in this study. Shown are all residues starting at phenylalanine 336, which is located at the beginning of the intracellular C-terminal tail. Serine and threonine residues are marked with arrowheads above the wild-type sequence, and amino acids used for replacement are shaded in the corresponding mutant sequences.
somewhat (Table 1). Similar findings have been reported previously for the PD (22) and the PD-N (6) mutant.
␤-Arrestin Interaction and Internalization of Full-length Receptors-We and others have previously shown that a mutant ␤ 2 -adrenergic receptor, which lacks all serine and threonine residues in the C terminus (22), does not interact with ␤-arrestin2 (4). This interaction is frequently assessed by confocal microscopy in which agonist-induced translocation of ␤-arrestin2-GFP fusion proteins to the plasma membrane is visualized. In HEK293 cells transiently transfected with ␤ 2 AR and ␤Arr2-YFP, stimulation of the cells with 10 M isoproterenol led to a pronounced translocation of fluorescence after 3 min of isoproterenol treatment (Fig. 2). This translocation was not apparent when the PD mutant was used instead of the wildtype receptor (Fig. 2). Similarly, the PD-N mutant, in which only the proximal group of four serine/threonine residues (between residues 355 and 364) was mutated, failed to recruit ␤Arr2-YFP to the plasma membrane (Fig. 2). In contrast, the PD-C mutant, in which all serines/threonines of the distal cluster (residues 384 and following) were mutated, behaved very much like wild-type receptors with respect to ␤Arr2-YFP translocation (Fig. 2).
In a second approach, we analyzed the receptor:␤-arrestin interaction in transiently transfected HEK293 cells by fluorescence resonance energy transfer between YFP-tagged receptor and CFP-tagged ␤-arrestin2 (4). In this assay, the receptor:␤arrestin interaction is again strictly dependent on GRK phosphorylation of the receptor (4). Fig. 3A shows a representative time course of interaction between wild-type ␤ 2 AR-YFP and ␤Arr2-CFP in the presence of GRK2, which occurred with a t1 ⁄ 2 of 8.3 s ( Table 2). No agonist-induced FRET between the YFPtagged ␤ 2 AR-PD and ␤Arr2-CFP was observed (Fig. 3B). Similarly, the YFP-tagged PD-N receptor showed only very little interaction with ␤Arr2-CFP, which was barely detectable in the FRET assay (Fig. 3C). In contrast, the YFP-tagged PD-C mutant produced a clear FRET signal with ␤Arr2-CFP (Fig. 3D), with a t1 ⁄ 2 of 4.3 s, which was not significantly different from the WT and ⌬C2 mutant ( Table 2; one-way analysis of variance with Bonferroni's Multiple Comparison test; p Ͼ 0.05). We have shown earlier (4) that this initial binding kinetics mirrors the kinetics of GRK phosphorylation of the receptor. Therefore, our data suggest that the PD-C mutant is phosphorylated slightly faster by GRK2 than the WT receptor.
When the receptor was repeatedly stimulated, ␤-arrestin2 bound fast and with similar kinetics to both WT and PD-C receptors ( Table 2, t1 ⁄ 2 of 1.1 s for the WT receptor and 1.5 s for the PD-C receptor). We have shown earlier (4) that such rapid ␤-arrestin2 binding is most likely due to binding to already phosphorylated receptors. Notably, the FRET amplitude obtained with the PD-C construct was reproducibly smaller than with the wild-type receptor, similar to the differences that we observed previously between the two agonists isoproterenol and norepinephrine. This suggests that the C terminus of the PD-C receptor assumes a different conformation compared with the WT when binding ␤-arrestin2.
Next, we analyzed agonist-stimulated internalization of the ␤ 2 -adrenergic receptor, which is a ␤-arrestin-dependent process and should, therefore, be affected in the PD and PD-N receptor mutants that did not interact with ␤-arrestin2 in our previous experiments. In fact, the internalization behavior of these mutants corresponded well with their ability to interact with ␤-arrestin: while the PD-C mutant internalized with a time course similar to the wild-type receptor, internalization of the PD and PD-N mutants was virtually abolished (Fig. 4A). A similar internalization pattern for the PD mutant has been reported previously (26), in contrast to the original investigation with this mutant (22). Taken together, our data support the proximal serine/threonine cluster (between residues 355 and 364) being the critical site for GRK phosphorylation and subsequent efficient ␤-arrestin2 binding.
Interestingly, internalization of the PD and PD-N mutant receptors could be partially rescued by overexpression of  ␤Arr2-YFP (Fig. 4B), indicating that mutation of the proximal serine/threonine residues did not completely prevent ␤-arres-tin2 from interacting with the receptors. However, no direct interaction between YFP-tagged PD or PD-N mutant receptors and ␤Arr2-CFP could be detected in FRET measurements under comparable conditions (Fig. 3, B and C), suggesting that either the amount of receptor:␤-arrestin complex is too low to be detected by FRET or the observed ␤-arrestin2-mediated res-cue may not directly correspond to primary binding of ␤-arrestin2 to the proximal serine/threonine cluster.
Role of the Distal C Terminus for ␤-Arrestin2 Binding-The above data suggested that residues in the distal part of the ␤ 2 AR C terminus are not required for ␤-arrestin binding. To investigate whether this part of the receptor contributes to ␤-arrestin binding and receptor internalization at all, we constructed a truncated receptor mutant (⌬C2), in which the C terminus distal to leucine 381, containing the serine and threonine residues which were mutated in PD-C, was deleted (Fig.  1). As shown in Fig. 5, the ⌬C2 receptor was fully capable to interact with ␤-arrestin2, as revealed by an agonist-dependent, quantitative translocation of ␤Arr2-YFP to the plasma membrane (Fig. 5A) and a strong signal detected in the FRET assay (Fig. 5B). Evaluation of the binding curves suggested that the kinetics of ␤-arrestin2 binding to ⌬C2 was similar to the wild-type receptor after the first stimulation ( Table 2). If the receptor had already been prestimulated, ␤-arrestin2 bound significantly faster to the ⌬C2 mutant than to the wild-type receptor (t1 ⁄ 2 of 1.1 s for wild-type receptor versus 0.8 s for ⌬C2). These data suggest that the distal C terminus is dispensable for efficient ␤ 2 AR phosphorylation and slightly impedes ␤-arrestin2 binding to the phosphorylated receptor.
Phosphorylation of PD-C and ⌬C2 Mutants-To explore whether the PD-C and ⌬C2 mutants showed altered phosphorylation by GRK2, we compared their phosphorylation state to that of the wild-type receptor in unstimulated and agoniststimulated cells. Fig. 6A shows that, in unstimulated cells, all three receptors showed little phosphorylation, which could be clearly increased by treating the cells for 5 min with 1 M isoproterenol. Quantitative analysis of repeated experiments revealed that there was no significant difference between the three receptors when the amount of agonist-stimulated phosphorylation was normalized to the basal phosphorylation (Fig. 6B).
Activity of the PD-C and ⌬C2 Mutants-To investigate a possible role for the distal receptor C terminus in receptor activation, we determined the cAMP production by WT, PD-C, and ⌬C2 receptors in membranes prepared from transiently transfected HEK293 cells and compared them. Because constitutive activity also depends on the amount of receptor expression, we took care to compare only membrane preparations that contained similar amounts of receptor, as determined by 125 I-CYP binding. Table 3 shows that both the PD-C and ⌬C2 mutants

TABLE 2 Kinetic constants for ␤ Arr2-YFP binding to various receptor mutants
There was no binding of ␤ Arr2-YFP to the PD and PD-N mutants (Fig. 3); there was some binding to the ⌬C2-PD mutant (Fig. 7), but rates could not be determined (ND) because of insufficient signal-to-noise ratio. The rates for the first stimulus are not significantly different from each other but for the second stimulus, binding of arrestin to ⌬C2 is significantly faster than to the WT (p Ͻ 0.05; one-way ANOVA with Bonferroni's multiple comparison test).  NOVEMBER 14, 2008 • VOLUME 283 • NUMBER 46 coupled to adenylyl cyclase very much like wild-type receptor in response to increasing amounts of isoproterenol. While WT and PD-C mutant showed the same basal activity, basal activity of the ⌬C2 mutant was increased but the difference did not reach statistical significance. ␤-Arrestin Interaction of the ⌬C2 Mutant Requires Receptor Phosphorylation-As shown above, binding of ␤-arrestin2 to the ⌬C2 receptor is accelerated after repeated stimulation, when compared with binding after the first stimulus. This suggested that ␤-arrestin2 binding to ⌬C2 still required receptor phosphorylation. To explore this, we constructed the ⌬C2-PD mutant in which the proximal serine and threonine residues between residues 355 and 364 were mutated to alanine and glycine (Fig. 1), leading to a mutant receptor corresponding to a C-terminally truncated PD-N. Again, mutation of the proximal serine/threonine cluster resulted in a receptor in which ␤-ar-restin2 binding was severely impaired. The ⌬C2-PD receptor was unable to induce quantitative translocation of ␤Arr2-YFP to the plasma membrane (Fig. 7A), and the FRET signal between YFP-tagged ⌬C2-PD and ␤Arr2-CFP was dramatically reduced (Fig. 7B). These findings indicate that efficient ␤-arres-tin2 binding still requires receptor phosphorylation at the proximal serine/threonine residues, even when the distal C terminus is removed. However, in contrast to the full-length PD-N receptor, the ⌬C2-PD mutant was able to recruit a small amount of ␤Arr2-YFP, which appeared as a distinct vesicular staining close to the plasma membrane after agonist treatment (Fig. 7A). Similarly, the FRET signal obtained with ⌬C2-PD was strongly reduced, but not completely abolished (Fig. 7B); unfortunately, the data were too noisy to be evaluated by curve fit-  ting. These data suggest that there is some residual binding of ␤-arrestin2 to the activated ⌬C2-PD receptor, even though the main GRK2 phosphorylation site between residues 355 and 364 is missing. In the full-length receptor mutants (PD and PD-N), no such binding could be detected (Fig. 3, B and C). Thus, removal of the distal C terminus slightly facilitates the access of ␤-arrestin2 to the ⌬C2-PD receptor, so that the effect becomes detectable. Therefore, mutation of the GRK2 phosphorylation sites seems to only decrease (but not completely abolish) the affinity of the activated receptor for ␤-arrestin2.

Arrestin-independent Role of ␤ 2 AR C-tail in Internalization
The Distal C Terminus Is Required for ␤-Arrestin2-dependent Receptor Internalization-Next we analyzed whether removal of the distal C terminus influenced receptor internalization. As shown in Fig. 8A, internalization of the ⌬C2 receptor was impaired (24% internalized receptors after 30 min), when compared with the wild-type receptor (32% internalized receptors). This result was rather unexpected, since the ⌬C2 receptor bound ␤-arrestin2 very efficiently (Fig. 5). Therefore, additional elements seem to exist in the distal C terminus which are required for efficient internalization of the ␤ 2 AR. Interestingly, internalization of the ⌬C2 receptor could be fully rescued by overexpression of ␤-arrestin2 (Fig. 8B), which argues for a combined action of ␤-arrestin2 and the ␤ 2 AR distal C terminus. This finding was corroborated by the internalization pattern of the ⌬C2-PD receptor, in which both the proximal and distal parts of the C terminus are altered. Such a "double mutant" receptor was completely unable to internalize in an agonist-dependent manner. Importantly, its internalization was only poorly rescued by overexpression of ␤-arrestin2 (Fig. 8B) when compared with that of PD mutants with a full-length C terminus (Fig. 4B). To further map the site on the ␤ 2 AR C terminus that aids in receptor internalization, the ⌬Asn405 mutant was constructed. It lacks the last eight amino acids. This mutant displayed wild-type-like ␤-arrestin2 binding as determined by FRET ( Table 2, t1 ⁄ 2 of 10.5 s after the first and t1 ⁄ 2 of 1.1 s after the second stimulus), but impaired receptor internalization (Fig.   FIGURE 6. Agonist-induced phosphorylation of the PD-C and ⌬C2 ␤ 2 AR mutants. A, HEK293 cells transiently transfected with FLAG-tagged wild-type or mutant ␤ 2 AR and GRK2 were labeled with [ 32 P]orthophosphate. Labeled cells were treated (or not) with 1 M isoproterenol for 5 min. Afterward, receptors were immunoprecipitated with M2 anti-FLAG-Sepharose, and incorporated radioactivity was quantified by phosphorimager analysis. Shown is a representative result from six experiments. B, averaged data from the six experiments, with the agonist-stimulated phosphorylation normalized to basal phosphorylation.

TABLE 3
Receptor-stimulated adenylyl cyclase activity in membranes from HEK293 cells transiently transfected with FLAG-tagged wild-type or mutant human ␤ 2 AR cAMP production was determined in the absence (E 0 ) or presence of 3 nM to 50 M isoproterenol for 30 min. E max and EC 50 values were calculated by nonlinear curve fitting. Each value represents the mean Ϯ S.E. of at least three experiments performed in triplicate.  8A, 18% internalized receptors after 30 min). The ⌬Asn405 mutant internalized even slightly less efficiently than the ⌬C2 mutant. These data suggest that the last eight amino acids of the ␤ 2 AR contain a domain that is involved in receptor internalization, but not arrestin binding.

DISCUSSION
In this report, we analyzed the role of the ␤ 2 AR C terminus for ␤-arrestin binding and receptor internalization. We restricted our investigation to ␤-arrestin2, which has a higher affinity for the ␤ 2 AR than ␤-arrestin1 (27).
It has been shown previously that simultaneous mutation of all serine/threonine residues in the ␤ 2 AR C terminus prevents agonist-mediated receptor phosphorylation (22) and ␤-arres-tin-mediated internalization (26). Here we show that not all of these residues are required for initial ␤-arrestin binding to the receptor. Previous research has demonstrated that phosphorylation of a cluster of four residues between Ser-355 and Ser-364 is essential for ␤-arrestin binding to the ␤ 2 AR (6 -8, 10). Consistent with these results, we show that mutation of only these proximal residues (the PD-N mutant) yields a receptor that is defective in agonist-induced ␤-arrestin2 binding, as evident from a variety of analyses: (1) loss of receptor-mediated plasma membrane translocation of ␤Arr2-YFP (Fig. 2), (2) the loss of FRET between ␤ 2 AR-YFP and ␤Arr2-CFP (Fig. 3C) and (3) a loss of receptor internalization (Fig. 4A).
Mutation of the proximal serine/threonine cluster did not completely abolish the ability of the ␤ 2 AR receptor to interact with ␤-arrestin2. The affinity of ␤-arrestin2 for an activated, but non-phosphorylated receptor seemed to be dramatically reduced, but not completely abolished. Using FRET, we observed a weak interaction between the activated ⌬C2-PD mutant receptor and ␤Arr2-YFP (Fig. 7). This could be explained in two ways. First, truncation of the C terminus in the ⌬C2-PD mutant could have reduced the distance or improved the orientation of the two fluorophores compared with the situation in the full-length phosphorylation-deficient (PD and PD-N) mutants, resulting in higher sensitivity for detecting the receptor-arrestin complex. Alternatively, ␤-arrestin2 binding to the truncated receptor could be facilitated in comparison to the full-length receptor. Such a mechanism has been proposed for the D1 dopamine receptor (21). A D1 receptor in which all of the GRK phosphorylation sites had been removed by truncation, ␤-arrestin2 could still bind to the activated receptor, and desensitization was still observed. However, internalization of the truncated D1 mutant was dramatically impaired. In the ␤ 2 AR, truncation of the receptor did not abolish the requirement for receptor phosphorylation, as ␤-arrestin2 binding to the ⌬C2 mutant was much stronger than binding to the ⌬C2-PD mutant (which presumably cannot be phosphorylated by GRKs; compare Fig. 7 to Fig. 5). Thus, phosphorylation of the proximal receptor C terminus is required for a high affinity interaction with ␤-arrestin2. If the ␤ 2 AR had been phosphorylated by a first agonist stimulus, application of agonist for a second time caused faster binding of ␤-arrestin2 to the ⌬C2 mutant than to the wild-type receptor ( Table 2). This suggests that the distal C terminus slightly impedes access of ␤-arrestin2 to the rest of the receptor, even in a GRK-phosphorylated state.
The observed residual binding to the ⌬C2-PD mutant was still agonist-dependent, which suggests that other parts of the receptor contain a ␤-arrestin2 binding site(s), which change their conformation upon agonist activation. An additional ␤-arrestin binding site in the ␤ 2 AR has been identified in the second intracellular loop (20), and the third intracellular loop may also contribute to ␤-arrestin binding, as suggested from results with other receptors (13)(14)(15)(16)(17)(18). This is in agreement with the currently accepted model of arrestin binding to G-proteincoupled receptors (11).
Recruitment of ␤-arrestins to activated G-protein-coupled receptors is known to be a prerequisite for receptor internalization (3). Accordingly we show that for those of our receptor mutants, which did not efficiently bind ␤-arrestin2 (PD, PD-N, and ⌬C2-PD), internalization was consistently severely impaired. However, it appears that effective recruitment of ␤-arrestins is not sufficient for efficient receptor internalization, because deletion of the distal C terminus generated a receptor (the ⌬C2 mutant) that was impaired in internalization despite strong ␤-arrestin2 binding. Deletion of the last eight residues of the ␤ 2 AR (⌬Asn405 mutant) was sufficient to reproduce the phenotype, suggesting that it is these amino acids that aid in internalization. Indeed, three of these amino acids are altered in the PD-C mutant which may explain the slightly reduced internalization of this mutant compared with wildtype (Fig. 4A, 35% internalized receptors after 30 min of stimulation, versus 47% wild-type receptors). Interestingly, overexpression of ␤-arrestin2 rescued the internalization-deficient phenotype of the ⌬C2 mutant, i.e. elevated ␤-arrestin2 levels could compensate for the loss of the receptor distal C terminus.
Based on the evidence from the literature, we believe that the role of the distal ␤ 2 AR C terminus is most likely to coordinate further protein-protein interactions aiding in the internalization process. Proteins that have been shown to bind to the ␤ 2 AR include protein kinase A-anchoring proteins (AKAPs) (28,29), the membrane fusion regulatory protein, N-ethylmaleimidesensitive factor (NSF) (30), or the Na ϩ /H ϩ exchanger regulatory factor (NHERF/EBP50) (31,32). Both NSF and NHERF/ EBP50 have been described to bind the ␤ 2 AR within the part of the C-terminal tail that is deleted in the ⌬Asn405 mutant (30 -32). Of particular interest, NSF can also interact with ␤-arres-tin1, although the interaction between NSF and ␤-arrestin2 was not investigated (33). We suggest that the distal ␤ 2 AR C terminus coordinates the interaction between ␤-arrestins and further proteins, thus aiding in the formation of a protein complex that is required for efficient internalization of the receptor. Our experiments do not address whether these interactions are required for efficient targeting of the receptor-arrestin complex into clathrin-coated pits or for the subsequent transformation of the pit into a clathrin-coated vesicle.
Why did overexpression of ␤-arrestin2 rescue the internalization defect of the PD and PD-N mutant (Fig. 4B) but was not efficient at rescuing the internalization defect of the ⌬C2-PD mutant (Fig. 8B)? In the FRET assay, the receptor with the lowest affinity to ␤-arrestin2 is the PD mutant (compare Figs. 3 and  7), so the trivial explanation that the sensitivity of our FRET microscope is too low to detect low affinity receptor-arrestin interactions (34) seems unlikely. We hypothesize that upon overexpression of ␤-arrestin2, it can bind indirectly to the PD and PD-N mutants via another protein that interacts with the distal ␤ 2 AR C terminus. If the interaction of this putative "bridging" protein with the ␤ 2 AR was dependent on a free receptor C terminus, we would not be able to pick up this complex formation in the FRET assay. Internalization of the ⌬C2-PD mutant cannot be rescued efficiently by overexpression of ␤-arrestin2 because the unknown arrestin binding partner is unable to interact with the receptor. This proposed mechanism might not be restricted to the ␤ 2 AR: it has been reported that phosphorylation-deficient mutants of the ␦-opioid receptor were able to internalize in a ␤-arrestin2-dependent manner (35).
In summary, the experiments reported in this report and our previous publication (4) can be explained by a model (11) in which ␤-arrestins first interact with agonist-activated, GRKphosphorylated ␤ 2 ARs. The phosphorylated residues required for ␤-arrestin binding are located in the proximal C terminus between Ser-355 and Ser-364. Receptor-bound ␤-arrestin2 then acts in concert with the distal receptor C terminus to allow efficient internalization.