NHERF-1 and the Cytoskeleton Regulate the Traffic and Membrane Dynamics of G Protein-coupled Receptors*

The sodium-hydrogen exchange regulatory factor 1 (NHERF-1/EBP50) interacts with the C terminus of several G protein-coupled receptors (GPCRs). We examined the role of NHERF-1 and the cytoskeleton on the distribution, dynamics, and trafficking of the β2-adrenergic receptor (β2AR; a type A receptor), the parathyroid hormone receptor (PTH1R; type B), and the calcium-sensing receptor (CaSR; type C) using fluorescence recovery after photobleaching, total internal reflection fluorescence, and image correlation spectroscopy. β2AR bundles were observed only in cells that expressed NHERF-1, whereas the PTH1R was localized to bundles that parallel stress fibers independently of NHERF-1. The CaSR was never observed in bundles. NHERF-1 reduced the diffusion of the β2AR and the PTH1R. The addition of ligand increased the diffusion coefficient and the mobile fraction of the PTH1R. Isoproterenol decreased the immobile fraction but did not affect the diffusion coefficient of the β2AR. The diffusion of the CaSR was unaffected by NHERF-1 or the addition of calcium. NHERF-1 reduced the rate of ligand-induced internalization of the PTH1R. This phenomenon was accompanied by a reduction of the rate of arrestin binding to PTH1R in ligand-exposed cells. We conclude that some GPCRs, such as the β2AR, are attached to the cytoskeleton primarily via the binding of NHERF-1. Others, such as the PTH1R, bind the cytoskeleton via several interacting proteins, one of which is NHERF-1. Finally, receptors such as the CaSR do not interact with the cytoskeleton in any significant manner. These interactions, or the lack thereof, govern the dynamics and trafficking of the receptor.

The sodium-hydrogen exchange regulatory factor 1 (NHERF-1/EBP50) interacts with the C terminus of several G proteincoupled receptors (GPCRs). We examined the role of NHERF-1 and the cytoskeleton on the distribution, dynamics, and trafficking of the ␤ 2 -adrenergic receptor (␤ 2 AR; a type A receptor), the parathyroid hormone receptor (PTH1R; type B), and the calcium-sensing receptor (CaSR; type C) using fluorescence recovery after photobleaching, total internal reflection fluorescence, and image correlation spectroscopy. ␤ 2 AR bundles were observed only in cells that expressed NHERF-1, whereas the PTH1R was localized to bundles that parallel stress fibers independently of NHERF-1. The CaSR was never observed in bundles. NHERF-1 reduced the diffusion of the ␤ 2 AR and the PTH1R. The addition of ligand increased the diffusion coefficient and the mobile fraction of the PTH1R. Isoproterenol decreased the immobile fraction but did not affect the diffusion coefficient of the ␤ 2 AR. The diffusion of the CaSR was unaffected by NHERF-1 or the addition of calcium. NHERF-1 reduced the rate of ligand-induced internalization of the PTH1R. This phenomenon was accompanied by a reduction of the rate of arrestin binding to PTH1R in ligand-exposed cells. We conclude that some GPCRs, such as the ␤ 2 AR, are attached to the cytoskeleton primarily via the binding of NHERF-1. Others, such as the PTH1R, bind the cytoskeleton via several interacting proteins, one of which is NHERF-1. Finally, receptors such as the CaSR do not interact with the cytoskeleton in any significant manner. These interactions, or the lack thereof, govern the dynamics and trafficking of the receptor.
Cell membranes are highly heterogeneous structures consisting of an ensemble of fluctuating microdomains with distinct lipid and protein compositions. These microdomains play important functions in signal transduction processes by increasing the rate and efficiency of coupling of key intermolecular interactions involved in specific signaling processes (1)(2)(3). The cytoskeleton has also been implicated in the regulation of signal transduction processes by serving as a substrate for the anchoring of specific proteins (4), regulating traffic (5), and partitioning the cell membrane into microdomains through the formation of effective barriers to the diffusion of lipids and proteins present in the bulk of the plasma membrane (6,7). However, this general model of the cytoskeleton's role in the regulation of signaling processes depicts a somewhat passive picture. Most studies of the relationships among cytoskeletal structures and signaling processes focused on the effects of extracellular signals on cytoskeletal reorganization rather than on the effects of the cytoskeleton upon signaling pathways. This view has been recently challenged by the discovery of a family of proteins containing N-terminal postsynaptic density protein (PSD95)/Drosophila disc large tumor suppressor (DlgA)/Zo-1 protein (PDZ) 2 domains (which interact with a variety of signaling molecules) and a C-terminal Ezrin-Radixin-Moesin (ERM)-binding domain (that enables these proteins to interact with cytoskeletal structures) (8 -10). Two of the members of this family, the Na ϩ /H ϩ exchange regulatory factors 1 and 2 (NHERF-1 and -2), interact with and modulate the function of several G protein-coupled receptors (10 -13). However, although the data show unequivocally that NHERF-1 and NHERF-2 modulate the function of GPCR, a unified hypothesis to explain the multiple roles of these scaffolding molecules is lacking.
Here, we examined the effects of the expression of NHERF-1/EBP50 in the dynamics of three GPCRs: the ␤ 2 -adrenergic receptor (␤ 2 AR), the parathyroid hormone receptor (PTH1R), and the calcium-sensing receptor (CaSR). The ␤ 2 AR and the PTH1R are type A and B receptors, respectively, and were chosen because of their well documented interactions with NHERF-1 (10, 12, 14 -18). The CaSR is a type C receptor that does not interact with NHERF-1, thus serving as an important control. The data show that the ␤ 2 AR and the PTH1R are closely associated to actin stress fibers by a mechanism that is modulated by their interactions with NHERF-1. NHERF-1 expression was not required for the cytoskeletal association of the PTH1R; however, the ␤ 2 AR was found in bundles only in cells expressing NHERF-1. The CaSR was not associated with the cytoskeleton independently of NHERF-1 expression. The diffusion of the PTH1R and the ␤ 2 AR was strongly influenced by the expression of NHERF-1. Finally, we show that the effects of NHERF-1 on PTH1R traffic are consistent with a novel model in which NHERF-1 interferes with the binding of arrestin to the activated PTH1R.

EXPERIMENTAL PROCEDURES
Materials and Constructs-Anti-NHERF-1 antibody was purchased from Upstate Biotechnology, Inc. Anti-FLAG (M1) was from Sigma, and anti-hemagglutinin (HA11) was purchased from Covance. Anti-rabbit IgG antibody conjugated to TRITC was obtained from Jackson ImmunoResearch. All other materials were purchased from Sigma unless otherwise noted. A pEGFP-N2 plasmid encoding a full-length human PTH1R carboxyl-terminal eGFP fusion protein (PTH1R-eGFP) was kindly provided by C. Silve (INSERM, Paris, France). Similar plasmids coding for the ␤ 2 -adrenergic receptor (␤ 2 -AR-eGFP) and the calcium-sensing receptor (CaSR-eGFP) were kindly supplied by Dr. Jeffrey Benovic (Thomas Jefferson University) and Dr. Gerda Breitwieser (Geisinger Medical Center), respectively.
Confocal Microscopy-An Olympus Fluoview 1000 confocal microscope equipped with a SIM scanner was used for all experiments unless otherwise indicated. For live cell imaging experiments, the cells were kept at 37°C using an open perfusion microincubator (Harvard Apparatus Inc.).
Immunocytochemistry-Cells were cultured on glass coverslips, transfected with the desired plasmids, and allowed to grow until 80% confluent. The coverslips were washed in phosphate-buffered saline and fixed for 20 min in 4% paraformaldehyde in phosphate-buffered saline at 4°C. Cells were permeabilized with 5% nonfat milk, 0.1% Triton X-100 for 1 h at 4°C and then stained with either phalloidin-TRITC (3 nM) or anti-NHERF-1 antibody (1 g/ml) overnight at 4°C. Phalloidinstained cells were washed four times with phosphate-buffered saline and mounted using gelvatol. Anti-NHERF-1-stained cells were further treated with anti-rabbit IgG conjugated with TRITC (1:1000) in 5% nonfat milk for an additional 2 h at room temperature. The cells were then washed four times with phosphate-buffered saline, mounted with gelvatol, and examined by confocal microscopy.  Fluorescence Recovery after Photobleaching (FRAP)-All FRAP measurements were done focusing the microscope onto the plasma membrane adjacent to the coverslip. This region of the cell was chosen for two main reasons: 1) it provides a large surface, and 2) many supplemental studies were done using total internal reflection fluorescence (TIRF), a technique that is limited to the observation of the cell membranes adjacent to the coverslip. The plasma membrane was located using a Z-stack scanning procedure (20). Circular regions of interest (between 90 and 200 m 2 ) were selected and bleached with a 1-s pulse from a 405-nm laser line using the Fluoview 1000 SIM scanner, whereas recovery data were acquired using the instrument's main scanner and the 488-nm line of an argon gas laser. This short pulse was selected to ensure a Gaussian bleaching spot. To maximize reproducibility of the experimental conditions, all data were acquired in the photon-counting mode of the instrument. Sixty images were then collected at intervals of 1-1.6 s. The images were exported to Meta-Morph (Universal Imaging, Inc.), and the average fluorescence intensities of the bleached regions and control regions in other cells or far removed regions of the same cell were obtained. The data were fitted to a single exponential decay and plotted using GraphPad Prism. The diffusion coefficient was calculated using the Stokes-Einstein equation FIGURE 3. PTH1R distribution on the plasma membrane of CHO-N10 cells. A, TIRF images of the plasma membrane of CHO-N10 cells transfected with PTH1R-eGFP. The cells shown in this image were not induced with tetracycline and therefore did not express NHERF-1. B, the addition of latrunculin A (200 ng/ml; 15 min prior to observation) induced the redistribution of the PTH1R from bundles to the bulk membrane. C, colocalization of PTH1R (green) with F-actin (phalloidin stain; red, left) and co-localization of PTH1R (green) with NHERF-1 (red, right) in CHO-N10 cells induced with tetracycline. D, a mutation of the PDZ domain binding motif of the PTH1R ( 590 ETVM 593 3 590 ETVA 593 ) reduces the interactions of the receptor with cytoskeletal fibers. E, overexpression of a truncated NHERF-1 mutant that does not contain the ERM-binding domain significantly reduces the binding of PTH1R to cytoskeleton fibers. Scale bars, 10 m.

TABLE 1 Effects of NHERF-1 expression on receptor distribution
CHO-N10 cells expressing the indicated constructs were imaged by TIRF. The distribution of fluorescent receptors was determined using a threshholding procedure to determine "bundle" and "bulk" fluorescence in regions of interest that comprised most of the surface of each examined cell. The percentage of receptors bound to cytoskeletal bundles was measured from the ratio of the integrated fluorescence of identified bundles to the total integrated fluorescence of the regions of interest. At least six independent Mattek plates were examined for each condition. The number n denotes the total number of cells examined for each condition. The statistical analyses were done by ANOVA followed by Tukey's multiple comparison tests.

NHERF-1(؊)
NHERF-1(؉) for two-dimensional diffusion (D ϭ r 2 /(4 d ), where r is the radius of the bleached spot, and d is the half-life of fluorescence recovery). Image Correlation Spectroscopy (ICS) and Image Cross-correlation Spectroscopy (ICCS)-These studies were based on the experimental techniques described by Wiseman et al. (21)(22)(23) and recently reviewed by Bacia and Schwille (24) and Kim et al. (25). The technique measures the correlation of an image with itself after a defined latency time . This correlation is a function of the mobility of the fluorescent molecules. The experimental design is, in principle, identical to that used for fluorescence correlation spectroscopy, except that in ICS a larger region of the cell is imaged. This allows the acquisition of spatial information that is not available in typical fluorescence correlation spectroscopy experiments. Two different experimental designs were used for these studies. In some cases, the data were collected with a Fluoview 1000 confocal microscope focused onto the plasma membrane of the cell adjacent to the coverslip. In others, the data were obtained using an Olympus-based TIRF system (see below). For confocal microscopy experiments, a small section of the plasma membrane (Ͻ50 m 2 ) was rapidly scanned (50 -60 ms/frame) under low laser power, and up to 300 images were collected. For the TIRF experiments, up to 200 images were obtained consecutively using a 500-ms exposure time. With either experimental set-up, fluorescence loss due to photobleaching of the sample was almost negligible. The image data were exported to ImageJ and analyzed using a plug-in specifically written to calculate the autocorrelation function of the data. This plug-in was based on code developed by A. Tully and E. Levitan (University of Pittsburgh). The resulting autocorrelation data were exported to GraphPad Prism and fit to a single species two-dimensional diffusion model (G() A, fluorescence recovery after photobleaching. CHO-N10 cells not induced with tetracycline were imaged at 1-s intervals with a Fluoview 1000 confocal microscope. At time 0, several circular spots were bleached (using a 405-nm laser line at 99% power applied for 1 s) where indicated using the SIM scanner of the confocal microscope without interrupting image acquisition. B, summary of the diffusion data. NHERF-1(Ϫ), CHO-N10 cells not induced with tetracycline to express NHERF-1; cells in which the expression of NHERF-1 had been induced are denoted with NHERF-1(ϩ). The data represent averages of 5-16 plates of cells. Statistical analysis was done by ANOVA followed by Tukey's post hoc tests to determine the statistical significance of the differences between tetracycline-treated and untreated cells. a, a statistically significant difference with regard to the NHERF-1(ϩ) state (p Ͻ 0.025 in all cases). C, effects of various mutations on the diffusion of PTH1R. The mutant M593A-PTH1R has a defective PDZ domain binding motif that does not bind PDZ domains efficiently. S1S2, cells that were cotransfected with a NHERF-1 mutant in which the sequences of the core binding motifs of both PDZ domains had been mutated to eliminate PTH1R binding. ⌬ERM, cells co-transfected with PTH1R-eGFP and a C-terminal truncation mutant of NHERF-1 that does not possess an ERM binding motif. Diffusion coefficients were determined from ICS data. The data represent a summary of 5-8 cells of at least five independent plates for each condition. Only data from cells treated with tetracycline to induce NHERF-1 expression are presented. Statistical analysis was done by ANOVA followed by Tukey's post hoc tests. b, statistically significant differences with both basal (NHERF-1(Ϫ)) and tetracycline-treated (NHERF-1(ϩ)) cells (p Ͻ 0.01 for all cases). D, the effects of latrunculin A on the dynamics of PTH1R. CHO-N10 cells were treated with 200 ng/ml latrunculin A 1 h before imaging. Diffusion coefficients were determined from ICS data. The graph summarizes results obtained with 5-10 cells from at least five different plates per condition. c, statistically significant differences with cells that had not been treated with latrunculin (p Ͻ 0.002); d, significant differences with cells that do not express NHERF-1 (p Ͻ 0.01). teristic time constant, K is a proportionality factor, and G 0 is a term that accounts for spatial autocorrelation) as described by Hebert et al. (21). The data were also fit to a two-state diffusion model without any significant improvement in the quality of the fit. Therefore, all analyses reported in this paper were done using the single-species diffusion model. For calibration purposes, the diffusion of free intracellular GFP was determined using ICS and FRAP. The values for the diffusion coefficient of GFP obtained by both methods were comparable. ICCS experiments were conducted essentially using the same design, except that images were collected from both green (eGFP) and red (monomeric dsRed) channels. The cross-correlation calculation was performed using an ICCS plug-in for ImageJ developed for this purpose. Fractional binding was calculated from the ratio of the amplitude of the cross-correlation function to the amplitude of the autocorrelation function for PTH1R. These calculations are based on the method described by Kim et al. (25).
Total Internal Reflection Microscopy-TIRF microscopy studies were conducted using an Olympus 1X71 equipped with a 150-milliwatt argon laser, an Olympus ϫ60 TIRF objective, a Princeton CCD camera, and Sutter filter wheels. The imaging work station was controlled with SimplePCI software (Compix, Inc.). Typical exposure times of 500 ms were used for most experiments. For ICS measurements, up to 200 images were collected continuously. For endocytosis experiments, images were taken once every 30 s for up to 20 min. The data collected were exported to ImageJ and analyzed using the appropriate plug-ins.
Endocytosis Studies-These experiments were done using TIRF microscopy. Cells with or without NHERF-1 were imaged at 37°C at 30-s intervals for 1-2 min before the addition of the ligand. The microscope was refocused after stimulation of the cells, and image acquisition was continued for an additional 20 min at 30-s intervals. Fluorescence intensity data of the whole cell or of specific regions of interest were exported to GraphPad Prism and analyzed by fitting to a single exponential decay.
Statistical Analysis-All multiple comparisons were done using ANOVA followed by Tukey's post-test pairwise comparisons using the analysis routines built in GraphPad Prism. Plates from a minimum of five independent experiments were used for every procedure. Most of the experiments described here were repeated 12 times or more. ICS and ICCS studies were done using ImageJ plug-ins developed in house. These are available upon request.

RESULTS
Expression of NHERF-1 in CHO-N10 Cells-These studies were done using a new cell line, derived from Chinese hamster ovary cells. These cells, termed CHO-N10, express NHERF-1 in a tetracycline-inducible manner. We chose CHO cells as a model system for these studies, because they do not express detectable levels of NHERF-1 (Fig. 1). Fig. 1 also shows the induction of NHERF-1 expression 24 h after the addition of various doses of tetracycline. As shown, the expression of NHERF-1 by CHO-N10 cells is exquisitely sensitive to tetracycline. Therefore, CHO-N10 cells appear to be an ideal system to examine the effects of NHERF-1 expression on the dynamics and biochemistry of its interacting proteins. All experiments reported below were done inducing NHERF-1 expression with 50 ng/ml tetracycline for 24 h.
GPCRs That Contain C-terminal PDZ Binding Motifs Are Tethered to Actin Stress Fibers-Since NHERF-1 binds GPCR and the cytoskeleton, we first examined the effects of NHERF-1/EBP50 on the distribution of GPCR in live cells transfected with ␤ 2 AR-eGFP, PTH1R-eGFP, or CaSR-eGFP. TIRF microscopy revealed that the ␤ 2 AR-eGFP construct aggregated on puncta and bundle-like structures that closely resembled cytoskeletal stress fibers only in cells expressing NHERF-1/ EBP50 (Fig. 2, A and B). In contrast, PTH1R-eGFP accumulated in bundles even in the absence of NHERF-1 expression (Figs. 2, C and D, and 3A). The CaSR chimeras were never observed in bundles (Fig. 2, D and E). Several lines of evidence suggested that these bundle structures were linked to the cytoskeleton. For instance, prolonged serum starvation of the cells reduced significantly the prevalence of receptor bundles. Furthermore, the addition of latrunculin A to disrupt the cytoskeleton rapidly dissipated most of the PTH1R bundles (see, for instance, Fig.  3B), such that 15 min after the addition of latrunculin, no bundles were visible. The relative distribution of the receptors in bundles was determined by morphometric analyses. The results of these measurements are summarized in Table 1. These results illustrate the significant effects of the expression of NHERF-1 on the subcellular distribution of the ␤ 2 AR. The effects of NHERF-1 on the distribution of the PTH1R were much less dramatic although significant. In contrast, NHERF-1 expression had no detectable effects on the distribution of the CaSR. Finally, to demonstrate that these distributions were not peculiar to CHO cells, the distribution of the PTH1R was examined in HEK293, rat osteosarcoma (ROS), and human osteosarcoma (SaOS2) cells. In all cases, a significant fraction of the PTH1R was found in bundle-like structures (data not shown).
To determine the relationship between the receptor bundles and the cytoskeleton, CHO-N10 cells expressing PTH1R-eGFP were fixed and stained with TRITC-conjugated phalloidin. Fig.   3C shows clearly that the bundles parallel actin fibers. As expected, NHERF-1 co-localized with these fibers.
The data shown in Fig. 1 clearly suggested a major difference in the subcellular distributions of the CaSR on one side and the ␤ 2 AR and the PTH1R on the other. We hypothesized that the differences in the distribution of these receptors were a consequence of the fact that the ␤ 2 AR and the PTH1R contain a PDZ binding motif at their C terminus, whereas the CaSR does not. Since the PDZ binding motifs of the ␤ 2 AR and the PTH1R interact with NHERF-1, we proceeded to examine the relevance of these interactions for the stability of the receptor bundles using the PTH1R as a model. CHO-N10 cells were transfected with 1) M593A-PTH1R-eGFP (a PTH1R-eGFP in which the C-terminal methionine has been mutated to alanine, which results in a dramatic reduction of the affinity of the receptor for NHERF-1 (26)) or 2) wt-PTH1R-eGFP co-transfected with a deletion mutant of NHERF-1 that does not contain the ERM-binding domain (⌬ERM-NHERF-1, which has intact PDZ domains and acts as a dominant negative regarding most of the effects of NHERF-1, primarily because it does not interact with the cytoskeleton (18)). In both cases, the PTH1R bundles were still present, albeit much less so than in the cells that expressed wt-PTH1R-eGFP alone (Table 1 and Fig. 3, D and E). Thus, the binding of the PTH1R to cytoskeletal fibers is mediated to a significant extent by the interaction of the C terminus of the receptor with specific PDZ domaincontaining proteins.
NHERF-1 Expression Modulates Receptor Diffusion-We next studied the effects of the cytoskeleton and NHERF-1 expression on PTH1R mobility. Diffusion coefficients were measured using a confocal microscope focused at the plasma membrane closest to the coverslip using FRAP and ICS techniques (Fig. 4). Fig. 4A shows a representative FRAP experiment carried out with cells that do not express NHERF-1. The diffusion coefficients and immobile fractions were calculated for several cells from 5-16 plates for each condition. These results are summarized in Fig. 4B. The diffusion of the PTH1R was significantly slower, and the fraction of immobile receptors was FIGURE 6. Topological analysis of the diffusion of PTH1R. CHO-N10 cells were examined using TIRF microscopy. Images (100 -300) were collected at 500-ms intervals. The cells were then examined to detect bundle-rich and bulk plasma membrane regions. Small circular sections were selected in these regions (Ͻ10-m diameter; as seen in A), and the data were fit to one-species autocorrelation functions. B, the first 30 s of the autocorrelation curves calculated for the regions of interest shown in A. These results were used to calculate diffusion coefficients (shown in C) and immobile fractions (D). Six cell plates for each experimental condition were examined. The distribution among "bundle" and "bulk" regions was done based primarily on the local density of bundles on the cell surface within the selected region of interest. Bundle-rich and bulk regions were selected using thresholding algorithms, and the analysis was limited to well defined bundle and bulk regions. Because of these limitations, roughly 30% of the surface of each cell examined was covered. The data were analyzed by ANOVA followed by post-test Tukey's comparisons between the NHERF-1(ϩ) and NHERF-1(Ϫ) groups. Relevant significant differences are noted in the graph as asterisks (*, p Ͻ 0.001; **, p Ͻ 0.05). These data were collected from six independent coverslips for each condition (1-3 cells were examined in each). significantly greater in cells that expressed NHERF-1. The results obtained with FRAP and ICS were internally consistent; both methods demonstrated that NHERF-1 significantly decreased the diffusion coefficient and mobile fraction of PTH1R (Fig. 4B). It should be noted that the numbers obtained with ICS and FRAP were not identical. These differences are not unusual and have been attributed to the existence of confinement zones that have a much greater relative influence in the data collected by ICS, because the region imaged is significantly smaller (27).
To demonstrate that the effects of NHERF-1 on receptor mobility were primarily due to the interaction of PTH1R with NHERF-1, we examined the diffusion of M593A-PTH1R. M593A-PTH1R diffused very rapidly in comparison with wt-PTH1R in the presence of NHERF-1 (Fig. 4C). Likewise, wt-PTH1R diffused very rapidly when co-expressed with NHERF-1 mutants that do not bind the receptor (such as S1S2, in which the core binding sequences of the PDZ domains have been scrambled) or that fail to interact with the cytoskeleton (such as ⌬ERM-NHERF-1) (Fig. 4C).
Finally, we examined the role of the cytoskeleton in the lateral mobility of PTH1R (Fig. 4D). Blocking actin polymerization with latrunculin A increased dramatically the diffusion coefficient while decreasing the immobile fraction of wt-PTH1R-eGFP. Interestingly, NHERF-1 decreased the diffusion coefficient of PTH1R in latrunculin A-treated cells, suggesting the presence of a macromolecular complex that includes NHERF-1, PTH1R, and other proteins that may interact with the second PDZ domain or with the ERM binding motif of NHERF-1. These observations demonstrate that the restricted mobility of the PTH1R observed in cells that express NHERF-1 is a consequence of two concomitant phenomena: 1) the interactions of the receptor with NHERF-1 and 2) the interactions of NHERF-1 with the cytoskeleton.
Since NHERF-1 also interacts with the ␤ 2 AR with high affinity, we predicted that NHERF-1 expression should have similar effects on ␤ 2 AR diffusion. The results are shown in Fig. 5. NHERF-1 dramatically reduced the mobility of the ␤ 2 AR. Importantly, NHERF-1 had no discernible effects on the diffusion of the CaSR, which was included as a negative control. Therefore, the effects of NHERF-1 are specific and not due to generalized changes in the structure of the cytoskeleton in the vicinity of the plasma membrane. Interestingly, the effects of NHERF-1 on ␤ 2 AR mobility were independent of the presence of ligand, thus suggesting that the interactions between the ␤ 2 AR and NHERF-1 do not require ligand binding.
Local Effects of NHERF-1 on Receptor Lateral Diffusion-Because the distribution of the PTH1R on the cell surface was not homogeneous, we examined the dynamics of the receptor in different subregions of the cell using TIRF microscopy. "Bundle" and "bulk" regions within each cell were identified based on the concentration of observable bundles within the region of interest (ROI) (Fig. 6A). Small (Ͻ2.5 m in diameter) circular ROI were selected within these regions, and the diffusion coefficient of the receptor was measured by ICS. As shown in Fig.  6B, the autocorrelation functions of the PTH1R within "bulk" and "bundle" regions were very different. Within the bundles, in the absence of NHERF-1, about 60% of the receptor molecules diffused with a coefficient of 0.1 m 2 /s (Fig. 6C). In contrast, the receptor diffused very slowly (D ϭ 0.019 m 2 /s) in the bulk region of the cell. This finding strongly suggests that the receptor diffuses along the bundles. However, this diffusion was still very slow when compared with that observed in the presence of latrunculin A; thus, we conclude that the motion of the receptors along the bundles is limited by the binding of the receptor to the cytoskeleton. NHERF-1 expression drastically reduced the diffusion coefficient of the fast moving component within the bundles without exerting a statistically significant effect on the diffusion coefficient of PTH1R molecules within the bulk membrane (Fig. 6C). Conversely, the expression of NHERF-1 increased significantly the immobile fraction of the PTH1R within the bundles. Because the data show that PTH1R bundles coincide with actin stress fibers (Fig.  3C), these results suggest that the effects of NHERF-1 on the lateral diffusion of the PTH1R are effectively limited to the population of receptors located in very close proximity to actin fibers.
Perturbation of the Effects of NHERF-1 Expression by the Addition of Ligand-The diffusion data shown in Figs. 4 and 5 suggest that NHERF-1 effectively tethers PTH1R and ␤ 2 AR to the cytoskeleton. Inasmuch as the interactions of some GPCR with NHERF-1 appear to be ligand-dependent (28), we hypoth- FIGURE 7. Effects of the addition of ligand on the lateral mobility of cell surface receptors. CHO-N10 cells were transfected with PTH1R-eGFP, ␤ 2 AR-eGFP, or CaSR-eGFP, and NHERF-1 expression was induced with 50 ng/ml tetracycline for 24 h. The diffusion of the fluorescent receptors was examined by TIRF-ICS immediately before and 5 min after the addition of saturating concentrations of ligand (100 nM PTH-(1-34) for the PTH1R, 10 M isoproterenol for the ␤ 2 AR, and 5 mM Ca 2ϩ for the CaSR). One hundred images were collected per set at a rate of 3 images/s. The measured total intensities were corrected for photobleaching and receptor endocytosis using a normalization plug-in for ImageJ. The corrected intensities were fit to a single component two-dimensional diffusion model to determine the diffusion coefficient and the immobile fraction of the receptor. The statistical analysis was done by ANOVA followed by post-test Tukey's comparisons. Statistically significant differences between basal and post-ligand were noted (p Ͻ 0.01, n ϭ 6 independent coverslips (1-6 cells examined/coverslip) for each experiment).
esized that the addition of ligand would result in changes of the diffusion properties of these receptors. Since its diffusion was insensitive to the expression of NHERF-1, once again we used the CaSR for control purposes. TIRF experiments were done using CHO-N10 cells transiently expressing PTH1R-eGFP, ␤ 2 AR-eGFP, or CaSR-eGFP. The results are shown in Fig. 7. The addition of 5 mM calcium did not influence the diffusion of the CaSR. In contrast, the diffusion of the PTH1R was very sensitive to the addition of ligand; the diffusion coefficient of the PTH1R increased almost 4-fold within 5 min of the addition of PTH-(1-34), whereas the immobile fraction was reduced to about half of its original value. The results obtained with the ␤ 2 AR were somewhat different. Prior to the addition of ligand, about 95% of the surface ␤ 2 AR was essentially immobile. The addition of isoproterenol did not alter the diffusion coefficient of the ␤ 2 AR; however, the immobile fraction of the receptor was reduced substantially. These results suggest that ligand induces the dissociation of the ␤ 2 AR⅐NHERF-1 complexes from the cytoskeleton without affecting the binding of NHERF-1 to the receptor.
NHERF-1 Expression Modulates PTH1R Internalization-To investigate the effects of NHERF-1 on receptor traffic, we examined the effects of the addition of ligand on the surface distribution of the PTH1R using TIRF microscopy. In these experiments, images were collected at 30-s intervals after the addition of ligand. PTH-(1-34), a well characterized PTH agonist (18,26), induced the disappearance of these receptor bundles (Fig. 8A). Fig. 8B shows a kymograph of the xy projections of the cells shown in Fig.  8A. Interestingly, the PTH1R internalized significantly more slowly at the edges of the cell. Collected intensity data were fit to a single exponential to estimate an apparent first order rate of internalization. The plateau values from these fits were used to calculate the fraction of internalized receptor. In the presence of PTH-(1-34), 70 -80% of the receptors internalized. Strikingly, NHERF-1 expression reduced the rate of internalization of the receptor by about 50% without affecting significantly the fraction of internalized receptors. In contrast, the antagonist PTH-(7-34) induced slower receptor internalization, which was abrogated by the expression of NHERF-1. We conclude that NHERF-1 expression affected differentially receptor internalization, depending on the nature of the ligand used to induce endocytosis.
Effects of NHERF-1 Expression on the Binding of ␤-Arrestin to the PTH1R-To gain some insight into the mechanisms by which NHERF-1 modulates PTH1R dynamics and internalization, CHO-N10 cells were transfected with PTH1R-eGFP and ␤-arrestin-1 fused to mRFP (a monomeric variant of the Discosoma protein dsRed). These cells were then studied using ICCS analysis, a technique developed to study protein-protein interactions in live cells. This method is based on a time-dependent analysis of the cross-correlation between the fluorescence intensities of two targets labeled with distinct fluorophores (24): proteins that are associated with a common macromolec- The image data were processed as follows. After correction for photobleaching, a y axis maximal projection of the cell was obtained. A pseudocolor look up table was applied to the projection. As shown, the intensity decayed significantly more rapidly in the NHERF Ϫ cell shown in A. Furthermore, receptor internalization was somewhat slower at the edges of both cells. C, summary of the PTH1R endocytosis data. The data show the summary of 12-18 cells (12 separate coverslips) per condition. The internalization rate constant was calculated fitting the intensity data to a single exponential, whereas the fraction of surface receptors internalized was calculated from the span of the decay function obtained from these fits. ANOVA analyses followed by Tukey's post-test comparisons revealed that NHERF-1 expression had a significant effect on the rate of endocytosis of the PTH1R, independently of the ligand used to induce internalization. Moreover, PTH-(7-34)-induced internalization of the receptor was almost completely blocked by NHERF-1 (p Ͻ 0.001, n ϭ 12). ular complex diffuse together, whereas those that are not diffuse in random directions with respect to one another. This approach has been previously used to measure dynamic protein-protein association (24,25).
In order to eliminate experimental artifacts due to signal saturation, we selected cells in which the fluorescence intensities of PTH1R-eGFP and mRFP-␤-arrestin-1 were comparable. In all cases, we calculated the cross-correlation function of the same cells, immediately before and at various times after the addition of the ligand. The fractional degree of binding was experimentally determined by measuring the ratio of the amplitudes of the PTH1R autocorrelation curve and the PTH1R-␤Arrestin cross-correlation curve, as described by Bacia et al. (24). The addition of PTH-(1-34) increased substantially the binding of PTH1R to arrestin in a time-dependent manner from a low basal value to almost 100% (Fig. 9A), indicative of the formation of a complex that included arrestin and the receptor. NHERF-1 reduced the rate of formation of this complex 3-fold, suggesting that NHERF-1 binding interferes with arrestin recruitment to the receptor. This observation also indicates that NHERF-1 dissociation probably precedes arrestin binding.

DISCUSSION
Direct, functional interactions of GPCRs with the cytoskeleton were initially suggested several years ago (29). We describe here three different modes of interaction of GPCR with the cytoskeleton: 1) some GPCR, such as the ␤ 2 AR, are linked to the cytoskeleton via their interactions with NHERF-1; 2) some GPCRs, such as the PTH1R, interact with the cytoskeleton through their interactions with other proteins in addition to NHERF-1; and 3) some GPCRs, such as the CaSR, do not appear to interact with the cytoskeleton at all. CHO-N10 cells were transfected with PTH1R-eGFP and selected for 20 days with G418. These cells were further transfected with a ␤-arrestin-1-mRFP construct and treated with vehicle (NHERF Ϫ ) or with tetracycline (NHERF ϩ ) as described under "Experimental Procedures." The cells were then examined with a confocal microscope focused on the plasma membrane. A, kinetics of binding of ␤-arrestin-1 to PTH1R. The fractional binding was calculated from the autocorrelation and cross-correlation curves as described under "Experimental Procedures." The solid lines were obtained by fitting the data to a single exponential. The rate constants obtained were 0.12 and 0.037 min Ϫ1 for the NHERF Ϫ and NHERF ϩ conditions, respectively. The difference between these rate constants is statistically significant (p ϭ 0.002, n ϭ 5). B, the fractional binding of PTH1R to arrestin was determined by image cross-correlation spectroscopy 10 min after the addition of the various ligands. The autocorrelation function for PTH1R-eGFP and the cross-correlation function were determined as described under "Experimental Procedures." The fractional binding of PTH1R to arrestin was calculated from the ratio of the amplitudes of the cross-correlation and autocorrelation functions as described by Kim et al. (25). The differences among groups were examined by ANOVA followed by pairwise Tukey's test comparisons. a, different from preligand basal (p Ͻ 0.05); b, different from NHERF Ϫ (p Ͻ 0.05). Five independent coverslips were examined for each condition. C, CHO-N10 cells were transfected with hemagglutinin-tagged PTH1R and FLAGtagged ␤-arrestin-1. Cells were treated with PTH-(1-34) (100 nM) or PTH-(7-34) (1 M) where stated. Fifteen minutes after the addition of ligand, the cells were lysed, and ␤-arrestin-1 was immunoprecipitated (IP) using agarose-conjugated anti-FLAG antibodies. IB, immunoblot.
The interactions of the ␤ 2 AR with the cytoskeleton appeared to be mediated almost exclusively by NHERF-1, as there was no evidence of cytoskeletal localization of the ␤ 2 AR in cells that did not express NHERF-1. However, surprisingly, the association of PTH1R to cytoskeletal structures was not strictly dependent on the expression of NHERF-1, suggesting that NHERF-1 is only one of several cellular components mediating the interactions of PTH1R with the actin cytoskeleton. Consistent with this, at least two other cytoskeleton-related proteins have been reported to interact with PTH1R: Tctex-1, a dynein light chain (30), reportedly linked to the actin cytoskeleton (31), and 4.1G, a protein directly associated with actin (32). The binding sites for these two proteins are located near the C terminus of PTH1R, upstream of the PDZ domain binding motif (ETVM), which appears to be excluded (30,32). These interactions may be sufficient to promote partial binding of PTH1R to the cytoskeleton. However, interference with the C-terminal PDZ domain binding motif significantly decreased the accumulation of PTH1R in bundles, suggesting that other PDZ domain-containing proteins may participate in the linkage of the PTH1R to the cytoskeleton. Nevertheless, our data strongly suggest strong interactions between the PTH1R and NHERF-1 in live cells; the PTH1R becomes effectively immobilized by its interactions with NHERF-1 and the underlying cytoskeletal fibers. The functional consequences of these interactions are probably multiple. The activation of phospholipase C-dependent pathways by PTH1R appears to be regulated by binding to NHERF and the cytoskeleton (16,33).
Much less is known about the functional consequences of tethering the ␤ 2 AR to NHERF-1 and the cytoskeleton, and most of what we know about these is limited to the effects of NHERF-1 on ␤ 2 AR traffic. It is clear that ␤ 2 AR recycling is tightly regulated by NHERF-1 expression, since disruption of the cytoskeleton impairs ␤ 2 AR recycling and increases receptor degradation (34). More recently, it has been reported that the ability of isoproterenol to activate cAMP production is increased by disruption of the cytoskeleton (35), but this observation has not been linked to NHERF-1 expression or function.
The diffusion of the ␤ 2 AR and the PTH1R but not the CaSR was strongly influenced by NHERF-1. The expression of NHERF-1 reduced the diffusion coefficient of the ␤ 2 AR and the PTH1R while increasing significantly the immobile fraction of these receptors. These results are consistent with the report of Bates et al. (27), who showed that the diffusion of the CFTR was also strongly influenced by NHERF-1 binding. These results prove that NHERF-1 immobilizes its targets by linking them to the cytoskeleton.
Our diffusion studies also shed important light on the influence of ligand binding on interactions of GPCR with NHERF-1. We hypothesized that ligands that reduce the affinity for NHERF-1 should increase the lateral mobility of the receptor. That seems to be the case for the PTH1R; ligand binding induced a significant increase of the diffusion coefficient and the mobile fraction of PTH1R in cells that expressed NHERF-1, whereas it was without significant effects in cells that did not. This indicates that, after ligand binding, 1) the interactions of the PTH1R with NHERF-1 and other partners have been altered, and 2) the association of PTH1R with the cytoskeleton has been impaired by either releasing NHERF-1 from the cytoskeleton or by direct disruption of the cytoskeleton. However, the case of the ␤ 2 AR is somewhat different. NHERF-1 expression immobilizes nonstimulated ␤ 2 AR, strongly suggesting that ␤ 2 AR⅐NHERF-1 complexes exist in the absence of ligand. Five minutes after the addition of ligand, the mobile fraction of the ␤ 2 AR was significantly increased without changing its diffusion coefficient. This strongly suggests that the binding of the ␤ 2 AR to NHERF-1 is not significantly altered by the addition of ligand, although the association of ␤ 2 AR⅐NHERF-1 complexes to the cytoskeleton probably is. The permanence of ␤ 2 AR⅐NHERF-1 complexes after the addition of ligand is consistent with previously published data. In fact, it was suggested that the interactions of the ␤ 2 AR and NHERF-1 were induced by ligand (28). However, more recent work by Cao et al. (34) established that GRK5-dependent phosphorylation of Ser 411 impaired NHERF-1 binding to the ␤ 2 AR, suggesting that ligand binding might negatively influence the binding of NHERF-1 to the ␤ 2 AR. Thus, the binding of NHERF-1 to the ␤ 2 AR appears to be temporally and dynamically regulated by signaling cascades downstream the activation of the receptor. Our results suggest that, at least within 5 min after ligand addition, the ␤ 2 AR remains associated to NHERF-1, but the FIGURE 10. Model describing the interactions of PTH1R with ligand, NHERF-1, the actin cytoskeleton, and arrestin. The PTH1R is anchored to the actin cytoskeleton by multiple interacting proteins, one of which is NHERF-1. In the absence of NHERF-1, the PTH1R moves rapidly along actin fibers by a mechanism that may involve dissociation from the cytoskeleton. The interaction of NHERF-1 with the ETVM C-terminal sequence immobilizes the PTH1R and tethers it to actin fibers. The addition of ligand to the PTH1R induces the dissociation of NHERF-1 and the binding of ␤-arrestin. Ligand-induced arrestin binding is significantly slower in the presence of NHERF-1, probably because NHERF-1 dissociation must precede arrestin binding.
interactions of NHERF-1 with the cytoskeleton or the integrity of the cytoskeleton itself are altered as a consequence of the treatment. It should be noted that several effects of cAMP and protein kinase A on the disruption of the cytoskeleton have been documented (36 -39).
The reduced lateral mobility of membrane proteins induced by NHERF-1 is likely to result in the accumulation of receptors in NHERF-1-enriched regions. Furthermore, the reduced mobility of the receptors should have significant modulatory effects on the traffic of these receptors. It was recently shown that, in HEK293 cells (which express NHERF-1 at very high levels), the interactions of PDZ domain-containing proteins with the C terminus of GPCRs regulate the dynamics of the endocytosis of these receptors by delaying the recruitment of dynamin, inducing abortive events that reduce the efficiency of endocytosis, or selecting subpopulations of clathrin-coated pits (40). Here we show that the interactions with NHERF-1 govern the overall dynamics of receptors that contain C-terminal PDZ binding motifs. We propose that, since endocytic events are initiated randomly on the cell surface and have a finite duration (19), a significant reduction of the lateral mobility of the cargo molecules will probably reduce the probability of accumulation of cargo in clathrin-coated pits, thus reducing the net rate of endocytosis. Furthermore, at least in the case of the PTH1R, the interaction of the receptor with NHERF-1 delays the binding of ␤-arrestin. Delayed ␤-arrestin recruitment implies delayed recruitment of AP-2 and clathrin and, as a consequence, delayed recruitment of dynamin to complete the endocytic process. Thus, our findings provide a simple, self-consistent explanation for the reduced rate of endocytosis of GPCR containing C-terminal PDZ binding motifs observed in the presence of NHERF-1. Furthermore, we show that different ligands can induce selective, alternative pathways of GPCR internalization via mechanisms that may or may not involve arrestin and that are tightly regulated by NHERF-1. The general model we propose is shown in Fig. 10. GPCRs that contain a C-terminal PDZ binding motif are tethered to the cytoskeleton via direct interactions with NHERF-1. Some of these, such as the PTH1R, may bind the cytoskeleton via multiple additional interactions that may include other PDZ domain-containing targets. In general, NHERF-1 binding plays a major role in anchoring these receptors to the cytoskeleton, reducing receptor mobility, and dynamically interfering with the binding of arrestin after ligand stimulation. Furthermore, the binding of NHERF-1 to some of these receptors blocks completely arrestin-independent receptor internalization processes, such as PTH-(7-34)-induced PTH1R endocytosis. These data, in conjunction with those of Puthenveedu and von Zastrow (40), suggest that this may be a very general mechanism for the regulation of GPCR trafficking.