Differential Affinities of Visual Arrestin, b Arrestin1, and b Arrestin2 for G Protein-coupled Receptors Delineate Two Major Classes of Receptors*

Visual arrestin, b arrestin1, and b arrestin2 comprise a family of intracellular proteins that desensitize G pro-tein-coupled receptors (GPCRs). In addition, b arrestin1 and b arrestin2 target desensitized receptors to clathrin-coated pits for endocytosis. Whether arrestins differ in their ability to interact with GPCRs in cells is not known. In this study, we visualize the interaction of arrestin family members with GPCRs in real time and in live cells using green fluorescent protein-tagged arrestins. In the absence of agonist, visual arrestin and b arrestin1 were found in both the cytoplasm and nucleus of HEK-293 cells, whereas b arrestin2 was found only in the cytoplasm. Analysis of agonist-mediated arrestin translocation to multiple GPCRs identified two major classes of receptors. Class A receptors ( b 2 adrenergic receptor, mu opioid receptor, endothelin type A receptor, dopamine D1A receptor, and a 1b adrenergic receptor) bound b arrestin2 with higher affinity than b arrestin1 and did not interact with visual arrestin. In contrast, class B receptors (angiotensin II type 1A receptor, neurotensin receptor 1, vasopressin V2 receptor, thyrotropin-releasing hormone receptor, and substance P receptor) bound both b arrestin isoforms with similar high affinities and also interacted with visual arrestin. Switching the carboxyl-terminal 53.5 0.5% less than 63% reduction arr2-GFP, and we 57.0 6 1.3% in cytoplasmic b arr1- b arr2CT-GFP the 34% for b arr1-GFP. These results demonstrate that residues in the carboxyl-terminal domains of b arrestin1 b arrestin2 mediate distinct subcellular distribution patterns of the two isoforms in the absence of agonist. These residues also determine the extent to which each isoform translocates and/or clathrin. In addition, they indicate that initial affinity arrestins is residues in the amino-terminal domains of the arrestin isoforms is consistent with from

G protein-coupled receptors (GPCRs) 1 comprise a large gene family of more than 1000 members that mediate distinct physiological functions as diverse as phototransduction, olfaction, vascular tone, cardiac output, digestion, and pain. GPCR signaling is regulated by a small family of intracellular arrestin proteins that includes visual arrestin (S-antigen), ␤arrestin1 (arrestin2), and ␤arrestin2 (arrestin3) (1,2). Visual arrestin is 60 and 65% identical in amino acid composition to ␤arrestin1 and ␤arrestin2, respectively, and predominantly localized in rod photoreceptor cells of the retina but can also be found in other tissues (3,4). The ␤arrestins are 78% identical in amino acid composition and widely expressed in tissues, but their expression level varies in a cell type-specific fashion (5,6). The variations in arrestin homology, localization, and expression level suggest that arrestin family members may differ in their abilities to regulate GPCR signaling.
Arrestins bind to GPCRs that are phosphorylated by G protein-coupled receptor kinases (GRKs) (7)(8)(9). The binding of a single arrestin to a phosphorylated receptor competitively blocks agonist-mediated signal transduction by uncoupling the receptor from heterotrimeric G proteins (5,7,8,10,11). This process is termed desensitization. In contrast to visual arrestins that primarily desensitize light-activated rhodopsin (10), the nonvisual arrestins ␤arrestin1 and ␤arrestin2 also participate in processes controlling re-establishment of receptor responsiveness (12)(13)(14)(15). ␤Arrestins target desensitized receptors for endocytosis and resensitization by functioning as docking proteins that link receptors to components of the endocytic machinery such as AP-2 and clathrin (16,17). They also regulate the rate at which endosomal receptors are dephosphorylated and recycled back to the plasma membrane through interactions with specific clusters of GRK-phosphorylated residues in the GPCR carboxyl terminus (18).
Arrestins bind numerous GPCRs at the plasma membrane (19), and this finding suggests that receptors contain common arrestin recognition motifs. In vitro studies using purified proteins have shown that specificity exists between arrestin family members and GPCRs. For example, visual arrestin binds rhodopsin in preference to the ␤ 2 -adrenergic receptor (␤ 2 AR) and m2 muscarinic acetylcholine receptor (m2mAChR), whereas ␤arrestin1 and ␤arrestin2 bind the ␤ 2 AR and m2mAChR in preference to rhodopsin (5,8,20,21). Furthermore, ␤arrestin1 binds the ␤ 2 AR with a 2.5-fold greater affinity than ␤arrestin2, and ␤arrestin2 binds the m2mAChR with a 1.5-fold greater affinity than ␤arrestin1 (21). However, whether specificity exists between arrestin family members and GPCRs in cells has not been explored.
In the following study we investigate the dynamic interactions between arrestin family members and GPCRs in live cells by assessing the redistribution of fluorescent arrestins from the cytoplasm to agonist-activated receptors at the plasma membrane. Two classes of GPCRs, designated A and B, are identified that differ in their affinities for the arrestin isoforms. Class A receptors, such as the ␤ 2 AR, do not interact with visual arrestin and bind ␤arrestin1 with less affinity than ␤arrestin2. Class B receptors, such as the angiotensin II type 1A receptor (AT1AR), interact with visual arrestin and bind both ␤arres-tin1 and ␤arrestin2 with similar high affinities. The molecular determinants underlying this classification appear to reside in specific serine residues located in the receptor carboxyl-terminal tail. These findings reveal a potential role for visual arrestin in the regulation of GPCRs outside the visual system. Moreover, they suggest that the particular cellular complement of arrestin isoforms and their distinct interactions with intracellular proteins will play a critical role regulating the pattern of GPCR desensitization, sequestration, and resensitization.

EXPERIMENTAL PROCEDURES
Materials-Isoproterenol, epinephrine, and dopamine were purchased from Research Biochemicals Inc. Arginine vasopressin, angiotensin II, and substance P were obtained from Sigma. Endothelin and neurotensin were from Peninsula Laboratories. 125 I-Cyanopindolol was purchased from NEN Life Science Products. Thyrotropin-releasing hormone and the thyrotropin-releasing hormone receptor (TRHR) cDNA were kindly provided by Dr. Patricia M. Hinkle (University of Rochester, Rochester, NY).
Plasmids-Constructs were generated by polymerase chain reaction following standard protocols and are depicted schematically in Fig. 1. ␤Arrestin1 with GFP conjugated to its carboxyl terminus (␤arr1-GFP) has been described previously (22). It was constructed by replacing the terminal stop codon of ␤arrestin1 with a SalI restriction site and inserting the modified cDNA in frame into p(S65T)GFP (CLONTECH). In order to match perfectly the 12-amino acid linker in ␤arr1-GFP, we modified our previously described ␤arrestin2-GFP conjugate (19). This new ␤arr2-GFP was constructed by replacing the terminal stop codon of ␤arrestin2 with a SalI restriction site and inserting the modified cDNA in frame into p(S65T)GFP. The ␤arr1-␤arr2CT-GFP chimera contains the first 333 amino acids of ␤arrestin1 (Met-1 to Gly-333) fused to the last 75 amino acids of ␤arrestin2 (Asp-336 to Cys-410). The ␤arr2-␤arr1CT-GFP chimera contains the first 335 amino acids of ␤arrestin2 (Met-1 to Gly-335) fused to the last 85 amino acids of ␤arrestin1 (Leu-334 to Arg-418). For both chimeras, the cDNA was inserted in frame into the SalI restriction site in p(S65T)GFP. ␤Arrestin1 with YFP conjugated to its amino terminus (YFP-␤arr1) was constructed by inserting the ␤arrestin1 cDNA in frame into the SalI restriction site in pEYFPC1 (CLONTECH). ␤Arrestin2 with YFP conjugated to its amino terminus (YFP-␤arr2) was constructed by inserting the ␤arrestin2 cDNA in frame into the SalI restriction site in pEYFPC1. The YFP-␤arr1S412A and YFP-␤arr1S412D mutants were generated by replacing nucleotides TCT, encoding serine at amino acid 412 in YFP-␤arr1, with GCT or GAT that mutates the serine to an alanine or aspartic acid, respectively. Visual arrestin with GFP conjugated to its amino terminus (GFP-arrestin) was constructed by inserting the visual arrestin cDNA in frame into the HindIII site in pEGFPC3 (CLONTECH). Construction of the ␤ 2 AR-V2R chimera and V2R-␤ 2 AR chimera have been described previously (18). The ␤ 2 AR-V2R chimera contains the first 341 amino acids of the ␤ 2 AR (Met-1 to Cys-341) fused to the last 29 amino acids of the V2R (Ala-343 to Ser-371). The V2R-␤ 2 AR chimera contains the first 342 amino acids of the V2R (Met-1 to Cys-342) fused to the last 72 amino acids of the ␤ 2 AR (Leu-342 to Leu-413).
Cell Culture and Transfection-Human embryonic kidney (HEK-293) cells were provided by the American Type Culture Collection (ATCC). HEK-293 cells were grown in Eagle's minimal essential medium with Earle's salt (MEM) supplemented with 10% (v/v) heat-inactivated fetal bovine serum and gentamicin (100 g/ml). HEK-293 cells stably expressing the ␤ 2 AR, ␤ 2 AR-GFP, and AT1AR were generated by standard procedures using G418 (400 g/ml) selection. Transient transfections were performed using a modified calcium phosphate coprecipitation method as described previously (23).
Receptor Expression-␤ 2 AR expression was measured by Scatchard analysis using 125 I-cyanopindolol (23). The expression of the AT1AR was measured by flow cytometry and normalized according to the ␤ 2 AR expression measured in parallel by both flow cytometry and saturation binding as described previously (24). Receptor expression in HEK-293 cells stably overexpressing the ␤ 2 AR was approximately 5.0 pmol/mg whole cell protein. Receptor expression in HEK-293 cells stably overexpressing the AT1AR was approximately 3.0 pmol/mg whole cell protein. Receptor expression in HEK-293 cells stably overexpressing the ␤ 2 AR-GFP was approximately 1.5 pmol/mg whole cell protein.
␤Arrestin Expression-Each ␤arrestin molecule is coupled to a single GFP moiety; therefore, the measured fluorescent intensity in the cytoplasm of cells is proportional to the number of GFP fluorophores and hence proportional to the number of GFP-labeled ␤arrestin molecules. Cells selected for analysis of ␤arrestin translocation possessed low and equivalent levels of GFP fluorescence. The relative level of ␤arrestin-GFP fluorescence (in intensity per pixel) was measured using the "range of interest" analysis provided with the Zeiss LSM-510 confocal microscope software. Settings on the microscope (laser power, pinhole size, detector gain, amplifier offset, amplifier gain, etc.) were held constant within and between experiments to ensure that cells expressing similar amounts of the different arrestin isoforms were compared. Expression of GFP-labeled ␤arrestins in the population of cells was assessed by Western blotting using mouse and rabbit anti-GFP antibodies (CLONTECH).
Confocal Microscopy-Transfected HEK-293 cells were plated on 35-mm glass-bottomed culture dishes. Two hours before analysis of translocation, the medium was replaced with serum-free MEM supplemented with 10 mM HEPES. Confocal microscopy was performed on a Zeiss laser scanning confocal microscope (LSM-510) using a heated (37°C) microscope stage as described previously (18,19). Images were collected sequentially using single line excitation (488 nm). Saturating concentrations of agonist were applied directly over the selected cells 10 or 15 s before the third image within a time series. The relative level of ␤arrestin-GFP fluorescence (in intensity per pixel) was measured in a fixed area of the cytoplasm over the duration of the experiment using the range of interest analysis provided with the microscope software. Data were analyzed using a "plateau with exponential decay" nonlinear regression function in GraphPad Prism.
Modeling of ␤Arrestin Translocation-In rod photoreceptor cells, rhodopsin is expressed at higher levels than visual arrestin (25). However, in the heart, ␤arrestin appears to be expressed at higher levels than the ␤ 2 AR so that the intracellular arrestin concentration is not a limiting factor in receptor desensitization (26). In the latter case, most of the intracellular arrestins will remain in the cytosol with only a small fraction translocating to agonist-activated receptors. Experimental conditions, however, can be modified in order to increase the translocating fraction of arrestin proteins. The expression of endogenous receptor protein can be augmented in a permanent or transient manner utilizing receptor cDNA transfection methods. Receptor overexpression creates an excess of potential binding sites for arrestin. These surplus sites create an environment permitting much of the arrestin to bind at the membrane to activated receptors, resulting in an increased cytosolic depletion of the arrestin protein.
Our experiments were performed in HEK-293 cells stably overexpressing the ␤ 2 AR (or AT1AR) at approximately 5.0 pmol/mg whole cell protein and transiently overexpressing GFP-labeled ␤arrestins. ␤Arrestin translocation was analyzed only in cells expressing a defined amount of the GFP-labeled ␤arrestin (measured as described above). By using as a standard a permanent line of HEK-293 cells expressing a known amount of ␤ 2 AR-GFP (1.5 pmol/mg whole cell protein), we were therefore able to estimate the receptor/␤arrestin ratio in the selected cells as shown below in Equations 1 and 2. Visualization of Arrestin Translocation to GPCRs total cell ␤ 2 AR total cell ␤arrestin ϭ pmol/mg of wild-type ␤ 2 AR pmol/mg of cell ␤arrestin-GFP (Eq. 2) Our measurements indicate receptors were expressed at levels 2.5-5fold greater than the level of ␤arrestin-GFP. To ensure the simultaneous activation of these receptors, we used a large excess of agonist. Therefore, our ␤arrestin translocation experiments utilized both the overexpression and simultaneous activation of receptors and can be modeled as follows.
The homologous desensitization of GPCRs first requires receptor phosphorylation by GRKs (7)(8)(9). The phosphorylation of agonist-activated membrane receptors, which is essentially irreversible at the plasma membrane, can be described by first order rate Equation 3, where R(t) is the number of non-GRK-phosphorylated receptors in the presence of a saturating concentration of agonist; R T is the time-independent total number of membrane receptors; k 1 is a measure of the rate of phosphorylation, and exp is the abbreviation for the exponential function. If R phos (t) is the amount of free phosphorylated membrane receptor not bound to arrestin, and R B (t) is the number of membrane receptors bound to arrestin, conservation of mass requires the following (see Equation 4).
Likewise if B T is the total amount of cellular ␤arrestin, B M (t) is the time-dependent amount of receptor-associated arrestin on the membrane, and B(t) is the amount of cytosolic arrestin, then we determine Equation 5.
Experiments indicate that a single phosphorylated receptor binds a single arrestin (i.e. B M ϭ R B ) (5,7,8). This observation and the exponential decay of non-GRK phosphorylated membrane receptors described by rate Equation 3 provide normalized rate equations (Equations 6 or 7) for the disappearance of cytosolic arrestins Note that the kinetic parameters k 2 and k 3 indicating the removal and reappearance of free cytosolic arrestins may depend on the type of arrestin present. At the beginning of an experiment essentially all of the arrestin is cytosolic so that if an experiment is designed so that R T , the total number of membrane receptors, is significantly greater than B T , the total amount of arrestins in the cytosol that can possibly bind and desensitize these receptors. Moreover, the observation that for t Ͼ 0 the rate of cytosolic arrestin disappearance must always be slower than the rate of GRK-mediated receptor phosphorylation (i.e. 1 Ϫ X(t) Ͻ 1 Ϫ exp(Ϫk 1 t)) further justifies neglecting this term. The rate of disappearance of cytosolic arrestins then simplifies to the relationship shown in Equation 8.
With the definition that ␣ ϭ k 3 ϩ k 2 R T the solution to Equation 8 is given in Equation 9.
Our experiments were designed to measure the rate of ␤arrestin disappearance in a region of cytosol in single cell that is defined by the laser-illuminated volume imaged in the confocal microscope. Since each ␤arrestin molecule is coupled to a single GFP moiety, the measured fluorescence intensity, I(t), in the cytosol will be proportional to the number of imaged GFP fluorophores (with proportionality constant I 0 ) and hence proportional to the number of ␤arrestin-GFP molecules X(t) in the illuminated volume. Therefore, the time-dependent change in the measured fluorescence intensity, I(t) (Equation 10), or the normalized fluorescent intensity, I(t)/I 0 , can be calculated from Equation 9 as follows: Data were analyzed by Equation 10 using the nonlinear regression function Minerr of Mathcad version 6.0 (Mathsoft, Cambridge, MA).

RESULTS
Translocation of ␤Arrestin1 and ␤Arrestin2 to the ␤ 2 AR-GFP-labeled proteins not only provide a means for analyzing intracellular protein-protein interactions without having to disrupt the plasma membrane, but they also allow for the assessment of the kinetics of interactions that occur in seconds (27,28). Therefore, to compare the ability of ␤arrestin1 and ␤arrestin2 to interact with agonist-activated GPCRs in real time and in live cells, we fused the green fluorescent protein (GFP) to the carboxyl terminus of ␤arrestin1 (␤arr1-GFP) and ␤arrestin2 (␤arr2-GFP) (see Fig. 1). The GFP-tagged ␤arrestins were then transiently transfected into HEK-293 cells stably overexpressing the ␤ 2 AR. These cells provide a cellular platform in which the affinities of the different arrestin isoforms for the ␤ 2 AR can be reliably compared since each cell will express the same number of receptors. The transfected ␤arrestins were expressed at comparable levels in the population of cells (Fig. 2C, inset). However, to ensure further an accurate comparison of the two ␤arrestin isoforms, single cells expressing equivalent amounts of ␤arr1-GFP or ␤arr2-GFP were selected for analysis of translocation by matching the relative intensity of GFP fluorescence as described under "Experimental Procedures." In the absence of agonist, we observed a difference in the subcellular distribution of the two ␤arrestin isoforms. ␤arr1-GFP was distributed in both the cytoplasm and nucleus of cells, whereas ␤arr2-GFP was distributed in the cytoplasm of cells but excluded from the nucleus (Fig. 2, A and B, compare 0-s images). Upon agonist addition, both ␤arr1-GFP and ␤arr2-GFP redistributed from the cytoplasm to the receptor at the plasma membrane (Fig. 2, A and B). However, as indicated by the increase in fluorescence at the plasma membrane, ␤arr2-GFP translocated faster and to a greater extent than did ␤arr1-GFP.
To quantitate the differences in the translocation profiles of ␤arr1-GFP and ␤arr2-GFP to the ␤ 2 AR, we measured the timedependent loss of ␤arrestin fluorescence from the cytoplasm after treatment with agonist. These data were then analyzed using a plateau with exponential decay nonlinear regression function (GraphPad Prism) that revealed three major differences in the translocation profiles of the two ␤arrestin isoforms (Fig. 2C). First, translocation of ␤arr2-GFP began sooner than ␤arr1-GFP. The delay in time between agonist addition (arrow) and the initial loss of ␤arrestin from the cytoplasm was 1.5 Ϯ 1.0 s for ␤arr2-GFP and 8.2 Ϯ 1.4 s for ␤arr1-GFP. Second, translocation of ␤arr2-GFP occurred faster than ␤arr1-GFP. The half-life of ␤arrestin depletion from the cytoplasm was 19.2 Ϯ 1.1 s for ␤arr2-GFP and 35.8 Ϯ 2.0 s for ␤arr1-GFP. Third, translocation of ␤arr2-GFP proceeded to a greater extent than ␤arr1-GFP. The fraction of cytoplasmic ␤arr2-GFP that translocated to the plasma membrane was 63.2 Ϯ 0.4%, whereas only 33.8 Ϯ 0.4% of the cytoplasmic ␤arr1-GFP translocated to the plasma membrane.
Fusion of the GFP moiety to the carboxyl terminus of ␤ar-restin1 and ␤arrestin2 may differentially alter their affinities for the ␤ 2 AR. Therefore, we fused the yellow fluorescent protein (YFP) variant of GFP to the amino terminus of ␤arrestin1 (YFP-␤arr1) and ␤arrestin2 (YFP-␤arr2), and we compared the ability of these ␤arrestin conjugates to translocate to the ␤ 2 AR stably overexpressed in HEK-293 cells. In the absence of agonist, YFP-␤arr1 was distributed in both the cytoplasm and nucleus of cells, whereas YFP-␤arr2 was distributed in the cytoplasm but excluded from the nucleus (Fig. 3, A and B, compare 0-s images). Upon agonist addition, both ␤arrestin isoforms redistributed from the cytoplasm to the receptor at the plasma membrane with profiles very similar to their GFP counterparts (Fig. 3, A and B). Translocation of YFP-␤arr2 began sooner, occurred faster, and proceeded to a greater extent than YFP-␤arr1 translocation (Fig. 3C). The delay in translocation was 1.95 Ϯ 0.8 s for YFP-␤arr2 and 8.34 Ϯ 3.2 s for YFP-␤arr1. The half-life of translocation was 28.0 Ϯ 0.8 s for YFP-␤arr2 and 44.3 Ϯ 4.9 s for YFP-␤arr1. The extent of ␤arrestin translocation corresponded to a 66.0 Ϯ 0.3% reduction in cytoplasmic YFP-␤arr2 and a 39.7 Ϯ 1.0% reduction in cytoplasmic YFP-␤arr1. These results demonstrate that, independent of the location of the fluorescent moiety, ␤arrestin2 has a greater affinity for the ␤ 2 AR than does ␤arrestin1.
Translocation of ␤Arrestin1 Phosphorylation Mutants to the ␤ 2 AR-It has been recently demonstrated that ␤arrestin1 is constitutively phosphorylated on a carboxyl-terminal serine (Ser-412) when it is in the cytoplasm (29). However, ␤arrestin1 is dephosphorylated when it is recruited to the agonist-activated ␤ 2 AR at the plasma membrane (29). Since ␤arrestin2 does not possess the Ser-412 residue, it may not be subject to the same type of regulation. To test whether the phosphorylation status of ␤arrestin1 regulates its translocation to the ␤ 2 AR, we mutated Ser-412 to an alanine (S412A) or to an aspartic acid (S412D) to mimic the dephosphorylated and constitutively phosphorylated forms of ␤arrestin1, respectively (29). HEK-293 cells stably overexpressing the ␤ 2 AR were transiently transfected with YFP-␤arr1, YFP-␤arr1S412A, or YFP-␤arr1S412D. In the absence of agonist, no differences were observed in the subcellular distribution of the wild-type and mutant ␤arrestins as each protein was distributed in both the cytoplasm and nucleus of cells (Fig. 4, A-C, compare 0-s images). Upon agonist addition, the wild-type and mutant ␤arrestins translocated to the ␤ 2 AR with very similar profiles (Fig.  4, A-C). As shown in Fig. 4D, the delay in translocation was 5.7 Ϯ 2.6 s for YFP-␤arr1, 4.9 Ϯ 1.8 s for YFP-␤arr1S412A, and 5.2 Ϯ 2.7 s for YFP-␤arr1S412D. The half-life of translocation was 44.8 Ϯ 4.2 s for YFP-␤arr1, 47.4 Ϯ 2.7 s for YFP-␤arr1S412A, and 46.9 Ϯ 4.2 s for YFP-␤arr1S412D. The extent of ␤arrestin translocation corresponded to a 42.0 Ϯ 0.9, 40.6 Ϯ 0.6, and 44.6 Ϯ 1.0% reduction in cytoplasmic YFP-␤arr1, YFP-␤arr1S412A, and YFP-␤arr1S412D, respectively. Therefore, the translocation profile of ␤arrestin1 to the ␤ 2 AR (which is delayed in onset, slower in rate, and reduced in magnitude compared with ␤arrestin2 translocation) is not influenced by the phosphorylation status of the Ser-412 residue.
Translocation of ␤Arrestin1 and ␤Arrestin2 Chimeras to the ␤ 2 AR-Once bound to the GRK-phosphorylated ␤ 2 AR, ␤arres-tin1 and ␤arrestin2 function as docking proteins that link the receptor to clathrin-coated pits by interacting with components of the endocytic machinery such as AP-2 and clathrin (16,17). ␤Arrestin2 binds clathrin with a 6-fold greater affinity than ␤arrestin1, and the two ␤arrestin isoforms may differ in their capacity to bind AP-2 (16,17). To test whether differences in the ability of ␤arrestin1 and ␤arrestin2 to interact with these proteins contribute to their different translocation profiles, we switched the carboxyl-terminal domains of ␤arrestin1 and ␤ar-restin2 because this region contains both the clathrin-and AP-2-binding sites (see Fig. 1) (16,30). The ability of these chimeric ␤arrestins to translocate to the ␤ 2 AR stably overexpressed in HEK-293 cells was then assessed. In the absence of agonist, the subcellular distribution of the chimeric ␤arrestins was completely reversed from their wild-type counterparts (Fig. 5, A and B, compare 0-s images). The ␤arr1-␤arr2CT-GFP chimera was distributed in the cytoplasm of cells but excluded from the nucleus, whereas the ␤arr2-␤arr1CT-GFP chimera was distributed in both the cytoplasm and nucleus of cells. Both chimeras translocated to the ␤ 2 AR upon agonist addition (Fig.  5, A and B). Similar to our findings for ␤arr1-GFP and ␤arr2-GFP, translocation of ␤arr2-␤arr1CT-GFP began sooner and occurred faster than ␤arr1-␤arr2CT-GFP translocation (Fig.  5C). The delay in translocation was 1.8 Ϯ 1.5 s for ␤arr2-␤arr1CT-GFP and 8.6 Ϯ 2.3 s for ␤arr1-␤arr2CT-GFP. The half-life of translocation was 30.6 Ϯ 1.8 s for ␤arr2-␤arr1CT-GFP and 50.7 Ϯ 4.1 s for ␤arr1-␤arr2CT-GFP. However, in contrast to our findings for ␤arr1-GFP and ␤arr2-GFP, the two chimeric ␤arrestins translocated to similar extents (Fig. 5C). We measured a 53.5 Ϯ 0.5% reduction in cytoplasmic ␤arr2-␤arr1CT-GFP which is significantly less than the 63% reduction observed for ␤arr2-GFP, and we measured a 57.0 Ϯ 1.3% reduction in cytoplasmic ␤arr1-␤arr2CT-GFP which is significantly more than the 34% reduction observed for ␤arr1-GFP.
These results demonstrate that residues in the carboxylterminal domains of ␤arrestin1 and ␤arrestin2 mediate the distinct subcellular distribution patterns of the two ␤arrestin isoforms in the absence of agonist. These residues also determine the extent to which each isoform translocates to the agonist-activated ␤ 2 AR presumably through interactions with AP-2 and/or clathrin. In addition, they indicate that the initial affinity of ␤arrestins for the ␤ 2 AR is mediated primarily by residues in the amino-terminal domains of the two ␤arrestin isoforms which is consistent with findings from in vitro studies (21).
Translocation of ␤Arrestin1 and ␤Arrestin2 to Other GPCRs-Once ␤arrestin binds to agonist-activated GPCRs at the plasma membrane, the fate of the receptor-␤arrestin com-plex differs among receptors. For some GPCRs, such as the ␤ 2 AR, the receptor-␤arrestin complex dissociates at or near the plasma membrane shortly after the formation of clathrincoated pits, and ␤arrestin is excluded from receptor-containing endocytic vesicles (18,22). However, for other GPCRs, the receptor-␤arrestin complex remains intact, and ␤arrestin is internalized with the receptor into endosomes (18,22). The angiotensin II type 1A receptor (AT1AR) belongs to the latter group of receptors; therefore, we compared the ability of ␤ar-restin1 and ␤arrestin2 to translocate to the AT1AR stably overexpressed in HEK-293 cells. In marked contrast to our findings for the ␤ 2 AR (Figs. 2 and 3), both ␤arrestins translocated to the AT1AR with nearly identical profiles (Fig. 6, A and  B). As shown in Fig. 6C, the delay in translocation was 5.4 Ϯ 1.5 s for ␤arr2-GFP and 10.2 Ϯ 0.9 s for ␤arr1-GFP. The half-life of translocation was 26.6 Ϯ 1.9 s for ␤arr2-GFP and 25.1 Ϯ 1.2 s for ␤arr1-GFP. The extent of ␤arrestin translocation corresponded to a 61.3 Ϯ 0.8% reduction in cytoplasmic ␤arr2-GFP and a 61.7 Ϯ 0.6% reduction in cytoplasmic ␤arr1-GFP. Moreover, with a longer agonist treatment, both ␤arres- A and B, the distribution of ␤arr1-␤arr2CT-GFP and ␤arr2-␤arr1CT-GFP fluorescence was visualized with the confocal microscope before (0 s) and after (35 s and 180 s) treatment with 10 M isoproterenol. C, quantitation of ␤arr1-␤arr2CT-GFP and ␤arr2-␤arr1CT-GFP translocation to the ␤ 2 AR. ␤arr1-␤arr2CT-GFP and ␤arr2-␤arr1CT-GFP fluorescence were measured in the cytoplasm of cells before and after treatment with 10 M isoproterenol. Confocal images were collected every 24.8 s (scan time ϭ 3.9 s), and agonist was added (arrow) 10 s before the third scan. Data represent the mean Ϯ S.E. of three independent experiments (n ϭ 11-13 cells) and were analyzed using a plateau with exponential decay nonlinear regression function in Graph-Pad Prism. For comparison, translocation profiles determined for ␤arr1-GFP and ␤arr2-GFP in Fig. 2C are included (broken gray lines). 6. Translocation of ␤arr1-GFP and ␤arr2-GFP to the AT1AR. HEK-293 cells stably overexpressing the AT1AR were transiently transfected with ␤arr1-GFP (A) or ␤arr2-GFP (B). Single cells expressing equivalent amounts of the two ␤arrestin isoforms were selected as described under "Experimental Procedures." A and B, the distribution of ␤arr1-GFP and ␤arr2-GFP fluorescence was visualized with the confocal microscope before (0 s) and after (30 s and 120 s) treatment with 1 M angiotensin II. C, quantitation of ␤arr1-GFP and ␤arr2-GFP translocation to the AT1AR. ␤arr1-GFP and ␤arr2-GFP fluorescence were measured in the cytoplasm of cells before and after treatment with 1 M angiotensin II. Confocal images were collected every 20.2 s (scan time ϭ 3.9 s), and agonist was added (arrow) 10 s before the third scan. Data represent the mean Ϯ S.E. of 5-7 independent experiments (n ϭ 27-36 cells) and were analyzed using a plateau with exponential decay nonlinear regression function in GraphPad Prism. D, the distribution of ␤arr1-GFP and ␤arr2-GFP was visualized with the confocal microscope 20 min after treatment with 1 M angiotensin II. tin isoforms were observed to internalize with the AT1AR into endocytic vesicles (Fig. 6D). These results show that ␤arr1-GFP has the capacity to translocate as quickly and to the same extent as ␤arr2-GFP and that ␤arrestin1 and ␤arrestin2 have similar high affinities for the AT1AR.
We next evaluated the translocation of ␤arrestin1 and ␤ar-restin2 to a variety of other GPCRs including the mu opioid receptor (MOR), endothelin type A receptor (ETAR), dopamine D1A receptor (D1AR), ␣1b-adrenergic receptor (␣1bAR), neurotensin 1 receptor (NTR-1), vasopressin V2 receptor (V2R), thyrotropin-releasing hormone receptor (TRHR), and substance P receptor (SPR). With the exception of the MOR, which was stably overexpressed, all other receptors were transiently overexpressed in HEK-293 cells. A qualitative comparison was then made between the levels of ␤arr1-GFP and ␤arr2-GFP that translocated to the plasma membrane 2 min after agonist treatment (Fig. 7). The amount of ␤arr2-GFP that translocated to the MOR, ETAR, D1AR, and ␣1bAR was greater than the amount of ␤arr1-GFP. In contrast, the amount of ␤arr2-GFP and ␤arr1-GFP that translocated to the NTR, V2R, TRHR, and SPR was very similar as both ␤arrestin isoforms translocated to a large extent. The robust translocation of ␤arr1-GFP to these receptors at the plasma membrane cleared out the cytoplasm and revealed the pool of ␤arr1-GFP distributed in the nucleus. With longer agonist incubations, both ␤arrestin isoforms dissociated from the MOR, ETAR, D1AR, and ␣1bAR at the plasma membrane and did not traffic with these receptors into endocytic vesicles ( Fig. 8 and data not shown). In contrast, both ␤arrestin isoforms remained associated with the NTR, V2R, TRHR, and SPR and internalized with these receptors into endocytic vesicles ( Fig. 8 and data not shown).
These results identify two classes of GPCRs. "Class A" receptors include the ␤ 2 AR, MOR, ETAR, D1AR, and ␣1bAR. These receptors, which recognize three different G proteins (G s , G i , and G q ), bind ␤arrestin2 with higher affinity than ␤arres-tin1. Complexes between class A receptors and ␤arrestin1 or ␤arrestin2 are relatively unstable, however, and dissociate at or near the plasma membrane and are excluded from receptorcontaining vesicles. "Class B" receptors include the AT1AR, NTR-1, V2R, TRHR, and SPR. These receptors, which recognize G s and G q , bind both ␤arrestin1 and ␤arrestin2 with high affinity. Complexes between class B receptors and ␤arrestin1 or ␤arrestin2 are stable and internalize together into endocytic vesicles.
Translocation of Visual Arrestin to GPCRs-Although visual arrestin preferentially binds rhodopsin, in vitro studies have shown that it can also bind the ␤ 2 AR but with an affinity 5-15-fold lower than ␤arrestin1 and ␤arrestin2 (5, 8, 21). To test whether this interaction occurs in cells, we constructed a GFP-arrestin fusion protein (see Fig. 1), and we evaluated its ability to translocate to the ␤ 2 AR stably overexpressed in HEK-293 cells. In the absence of agonist, GFP-arrestin (like ␤arres-tin1) was distributed in both the cytoplasm and nucleus of cells (Fig. 9A, upper panel). Upon agonist addition, however, GFParrestin did not translocate to the ␤ 2 AR (Fig. 9A, upper panel). Even after 60 min of agonist treatment, no translocation was observed (data not shown). Similarly, GFP-arrestin did not translocate to MOR, ETAR, D1AR, and ␣1bAR (data not shown).
We next evaluated the ability of visual arrestin to translocate to the AT1AR stably overexpressed in HEK-293 cells since this receptor is less selective and binds both ␤arrestin isoforms with high affinity. As shown in the lower panel of Fig. 9A, GFP-arrestin translocated to the AT1AR within 2 min of agonist stimulation. Moreover, GFP-arrestin translocated to the other class B receptors (Fig. 9B). The robust translocation of GFP-arrestin to the V2R, TRHR, and SPR at the plasma membrane cleared out the cytoplasm and revealed the pool of GFParrestin distributed in the nucleus. Interestingly, in contrast to our observations for the ␤arrestin isoforms, translocation of visual arrestin resulted in a more diffuse and less punctate pattern of fluorescence at the plasma membrane. This suggests that although visual arrestin binds class B receptors at the plasma membrane, it may not target these receptors to clathrin-coated pits which is consistent with its reported inability to bind clathrin (17). These results not only demonstrate that visual arrestin can interact in cells with GPCRs other than rhodopsin, but they also further establish the existence of two classes of GPCRs that differ in their affinity for arrestin family members.
Translocation of Visual and Nonvisual Arrestins to Chimeric GPCRs-Binding of arrestin isoforms to many GPCRs appears to be mediated by GRK-phosphorylated serine and threonine residues located in the receptor carboxyl-terminal tail. To test whether the carboxyl-terminal tail of class A and class B receptors contributes to their differential affinities for the visual and nonvisual arrestins, we employed chimeric receptors in which the tails of the ␤ 2 AR (class A) and V2R (class B) were swapped (18). The ␤ 2 AR with the V2R tail (␤ 2 AR-V2R chimera) and the V2R with the ␤ 2 AR tail (V2R-␤ 2 AR chimera) were transiently expressed in HEK-293 cells. We then compared qualitatively the levels of ␤arr1-GFP, ␤arr2-GFP, and GFParrestin at the plasma membrane 2 min after agonist treatment. The amount of ␤arr2-GFP that translocated to the V2R-␤ 2 AR chimera was much greater than the amount of ␤arr1-GFP, and no translocation was observed for GFP-arrestin (Fig.  10, upper panel). Therefore, the V2R-␤ 2 AR chimera displays a "␤ 2 AR-like" affinity for the visual and nonvisual arrestins. In marked contrast, both ␤arrestin isoforms translocated to a similarly large extent to the ␤2AR-V2R chimera (Fig. 10, lower   panel). Moreover, GFP-arrestin translocated to the ␤ 2 AR-V2R chimera at the plasma membrane. Therefore, the ␤ 2 AR-V2R chimera displays a "V2R-like" affinity for the visual and nonvisual arrestins. These results demonstrate that residues in the receptor carboxyl-terminal tail play a central role in determining the affinity of GPCRs for arrestin family members. DISCUSSION In this study we identify two major classes of GPCRs that demonstrate marked differences in their interactions with visual arrestin, ␤arrestin1, and ␤arrestin2. Class A receptors, which bind both biogenic amines and peptide ligands, demonstrate little to no affinity for visual arrestin and have a higher affinity for ␤arrestin2 than ␤arrestin1. Class B receptors, which bind peptide ligands, bind visual arrestin and have a similar high affinity for both ␤arrestin1 and ␤arrestin2. Residues in the receptor carboxyl-terminal tail appear entirely responsible for determining class A or B characteristics rather than the nature of the ligand. This classification paradigm provides a foundation for predicting the manner in which the cellular complement of arrestin isoforms will regulate the desensitization, sequestration, and resensitization of a particular GPCR.
For many GPCRs, homologous desensitization involves both GRK phosphorylation and arrestin binding (1,2). Our findings fit a model in which the kinetics of GRK phosphorylation are rate-limiting and facilitate the interaction of arrestin with receptors (Fig. 11A). In support of this interpretation, GRK phosphorylation of rhodopsin and the ␤ 2 AR is necessary for high affinity binding of arrestins both in vitro and in cells (7-9, 19, 21). Moreover, GFP analogues of GRK2 translocate to the SPR at the plasma membrane prior to the binding of ␤arrestin (28). Some ␤arrestin binding may also occur, however, with agonistactivated receptors in the absence of GRK phosphorylation. For example, in vitro studies have shown that ␤arrestin binds with low affinity to the unphosphorylated ␤ 2 AR and m2mAChR (21). In addition, in experiments with the SPR under conditions that should significantly reduce GRK-mediated phosphorylation in response to agonist (60 min agonist treatment at 4°C), both transfected and endogenous ␤arrestins were found to associate with the receptor at the plasma membrane (31). ␤Arrestin binding to GPCRs is thought to involve the engagement of two regions of the ␤arrestin molecule with two corresponding regions of the receptor (2,21). A large region within the aminoterminal half of ␤arrestin (termed the activation-recognition domain) appears to bind the third loop of agonist-activated GPCRs (32). A positively charged smaller domain (termed the phosphorylation-recognition domain) appears to interact with GRK-phosphorylated residues in the GPCR carboxyl terminus (33). Thus, although ␤arrestins may bind intracellular loops of agonist-activated receptors that are not phosphorylated, the affinity of this interaction is presumably weak. GRK phosphorylation and the engagement of both ␤arrestin recognition domains with the receptor appear to be necessary for a high affinity interaction to take place.
From the data presented in Fig. 2, we observed that translocation of ␤arrestin1 to the ␤ 2 AR was delayed in onset and slower in rate compared with ␤arrestin2 translocation. Reanalysis of this data using the model described under "Experimental Procedures" indicates that these qualitative differences correspond to an approximately 10-fold higher affinity of ␤arrestin2 for the ␤ 2 AR relative to ␤arrestin1 (Fig. 11B). In contrast to these findings in cells, in vitro studies have reported that ␤arrestin1 has a 2.5-fold greater affinity for the ␤ 2 AR than ␤arrestin2 (21). The discrepancy may arise from differences in the manner in which ␤arrestin1 and ␤arrestin2 interact intracellularly with the GRK-phosphorylated ␤ 2 AR carboxyl-terminal tail. The ␤ 2 AR tail contains 13 putative phosphate acceptor sites, and ␤arrestin1 binding in vivo may require that either a greater number of sites or different residues be phosphorylated. Moreover, proteins other than GRKs apparently influence the interaction between ␤arrestins and GPCRs. The ␤arrestin carboxyl-terminal domain contains AP-2 and clathrin-binding sites (16,30). A stronger interaction between ␤arrestin2 and these endocytic proteins may explain why swapping the ␤ar-restin1 and ␤arrestin2 carboxyl-terminal domains increases the fraction of the ␤arrestin1 chimera and decreases the fraction of the ␤arrestin2 chimera that translocate to the ␤ 2 AR at the plasma membrane.
The difference in the ability ␤arrestin1 and ␤arrestin2 to interact in cells with the ␤ 2 AR is typical of class A receptors. However, receptors belonging to class B, such as the AT1AR, appear to bind ␤arrestin1 and ␤arrestin2 equally well (Figs. 6C and 11C). Two lines of evidence suggest that their affinity for ␤arrestins is greater than that possessed by class A receptors. First, AT1ARs and ␤arrestins form stable endocytic complexes, whereas the ␤ 2 AR⅐␤arrestin complexes dissociate at or near the plasma membrane. The ability of ␤arrestins to internalize with the AT1AR into endocytic vesicles may indicate that this receptor internalizes in a ␤arrestin-dependent pathway. Consistent with this possibility, overexpression of ␤arrestin1 potentiates AT1AR sequestration in HEK-293 cells (24). Second, visual arrestin translocates to class B but not class A receptors. The ability of class B receptors, such as the AT1AR and V2R, to bind arrestin family members with such high affinity appears to be mediated entirely by clusters of phosphorylated residues in the receptor carboxyl-terminal tail (18,22). Similar clusters are noticeably absent from the tails of class A receptors. Thus, it is not unexpected that switching tails between class A and class B receptors completely reverses their affinity for both the visual and nonvisual arrestins.
Differences in ␤arrestin affinities for class A receptors suggest that the tissue complement of ␤arrestin1 and ␤arrestin2 may provide an additional degree of receptor regulation. ␤Ar-restin1 is more abundant than ␤arrestin2 in many bovine tissues including the heart, lung, and spleen and is the predominant isoform in rat respiratory epithelial cells (6,34). Conversely, ␤arrestin2 is more abundant than ␤arrestin1 in the rat central nervous system, olfactory epithelial cells, and spermatozoa (5,34). ␤Arrestin2 was also found to be 3-5-fold more abundant than ␤arrestin1 in a variety of transformed cell lines (35). Since ␤arrestin1 and ␤arrestin2 bind class A receptors with different affinities, signaling of class A receptors will be more affected by relative changes in the ratio of ␤arrestin1 and ␤arrestin2 than class B receptors. An important challenge for future research will be to determine the relationship be- FIG. 11. Modeling of arrestin translocation and desensitization of agonist-activated GPCRs. A, schematic depicting the homologous desensitization of GPCRs in the presence of saturating concentrations of agonist (a). The kinetics of arrestin binding to phosphorylated membrane receptors, resulting in their homologous desensitization, is mathematically described under "Experimental Procedures." B and C, time course of ␤arrestin1 and ␤arrestin2 disappearance from the cytoplasm after agonist activation and phosphorylation of the ␤ 2 AR (B) and AT1AR (C). Data from Fig. 2 (␤ 2 AR) and Fig. 6 (AT1AR) were re-analyzed using the relationship in Equation 10 ("Experimental Procedures"). The kinetic parameter k 1 (corresponding to GRK phosphorylation) was set at 0.02 for the ␤ 2 AR and 0.025 for the AT1AR (23). The rate constants k 2 R T and k 3 (corresponding to ␤arres-tin1 and ␤arrestin2 association and dissociation, respectively) were determined by nonlinear regression analysis. For the ␤ 2 AR, k 2 R T and k 3 values were 0.014 and 0.034 for ␤arrestin1 and 0.195 and 0.114 for ␤arrestin2. For the AT1AR, k 2 R T and k 3 values were 0.058 and 0.038 for ␤arrestin1 and 0.083 and 0.056 for ␤arrestin2. tween GPCR behavior and the absolute and relative levels of ␤arrestin1 and ␤arrestin2 in normal cells and in diseased tissues.
The ability of visual arrestin to interact with class B receptors suggests that it may play an unappreciated role regulating the signaling of GPCRs other than rhodopsin. Visual arrestin is expressed primarily in rod photoreceptor cells of the retina, but it is also found in other brain regions such as the pineal gland and in peripheral tissues including the heart, kidney, and lung (3,4). In contrast to ␤arrestin1 and ␤arrestin2, visual arrestin is unlikely to promote receptor sequestration as it does not bind clathrin and interacts very poorly with AP-2 2 (17). Moreover, we observe that visual arrestin does not appear to target class B receptors to clathrin-coated pits. Therefore, visual arrestin by desensitizing class B receptors without promoting internalization may provide a more permanent scaffold for membrane signaling events and an endocytic independent mechanism to resensitize receptors more rapidly than either ␤arrestin1 or ␤arrestin2 (36 -39).
An unanticipated finding of these studies is that the agonistindependent subcellular distributions of the arrestin isoforms differ. Visual arrestin and ␤arrestin1 are distributed in both the cell cytoplasm and nucleus, whereas ␤arrestin2 is found in the cytoplasm but is excluded from the nucleus. The amino acid residues mediating arrestin compartmentalization appear to reside in the arrestin carboxyl terminus. Switching the ␤arres-tin1 and ␤arrestin2 carboxyl-terminal domains completely reverses their subcellular distributions. Notably, reports have shown that functional prostaglandin E2 receptors reside on the nuclear envelope and other GPCRs, including muscarinic and angiotensin receptors, may be there as well (40 -43). Thus, the nuclear pools of visual arrestin and ␤arrestin1 may specifically regulate nuclear GPCR signaling.
In summary, we have shown in real time and in live cells that specificity exists between arrestin family members and GPCRs. Therefore, these findings provide a basis for predicting the arrestin-mediated pattern of GPCR desensitization, sequestration, and resensitization. Moreover, they explain why the regulation of certain GPCRs will be more susceptible to variations in the cellular complement of arrestin isoforms.