Divergent β-Arrestin-dependent Signaling Events Are Dependent upon Sequences within G-protein-coupled Receptor C Termini*

Background: Many GPCRs that utilize β-arrestins differ with respect to downstream signaling and cellular consequences. Results: Exchanging the C termini of two GPCRs switches the β-arrestin responses and relative affinities for the two receptors. Conclusion: Sequences within the C termini of different GPCRs are important for determining the nature of β-arrestin recruitment and signaling. Significance: These studies provide new insight regarding receptor-specific β-arrestin signals. β-Arrestins are multifunctional adaptor proteins that, upon recruitment to an activated G-protein-coupled receptor, can promote desensitization of G-protein signaling and receptor internalization while simultaneously eliciting an independent signal. The result of β-arrestin signaling depends upon the activating receptor. For example, activation of two Gαq-coupled receptors, protease-activated receptor-2 (PAR2) and neurokinin-1 receptor (NK1R), results in drastically different signaling events. PAR2 promotes β-arrestin-dependent membrane-sequestered extracellular signal-regulated kinase (ERK1/2) activation, cofilin activation, and cell migration, whereas NK1R promotes nuclear ERK1/2 activation and proliferation. Using bioluminescence resonance energy transfer to monitor receptor/β-arrestin interactions in real time, we observe that PAR2 has a higher apparent affinity for both β-arrestins than does NK1R, recruits them at a faster rate, and exhibits more rapid desensitization of the G-protein signal. Furthermore, recruitment of β-arrestins to PAR2 does not require prior Gαq signaling events, whereas inhibition of Gαq signaling intermediates inhibits recruitment of β-arrestins to NK1R. Using chimeric receptors in which the C terminus of PAR2 is fused to the N terminus of NK1R and vice versa and a critical Ser/Thr mutant of PAR2, we demonstrate that interactions between β-arrestins and specific phosphoresidues in the C termini of each receptor are crucial for determining the rate and magnitude of β-arrestin recruitment as well as the ultimate signaling outcome.

The classic paradigm for ␤-arrestin recruitment to GPCRs suggests that ␤-arrestins are recruited after phosphorylation of the receptor C termini by second messenger kinases or G-protein-coupled receptor kinases (GRKs) (15)(16)(17). Studies have shown that broad-spectrum PKC inhibitors can block PAR 2 desensitization, and both PKC and GRKs have been reported to play a role in NK1R down-regulation (2,9,18,19). We hypothesized that the differences in ERK1/2 activation by the two receptors and the subsequent downstream events might be dependent upon interactions between ␤-arrestin-1/2 and the receptor C termini, which might in turn be dependent upon the phosphorylation state of the receptors. To address this possibility, we switched the C-tails of the two receptors and examined Ca 2ϩ mobilization, ␤-arrestin recruitment kinetics, ERK1/2 activation and localization, and two downstream signaling events: proliferation and cell migration.

EXPERIMENTAL PROCEDURES
Materials-All chemicals were from Sigma unless otherwise stated. The following primary antibodies were used for Western blotting, immunostaining, or in-cell Western assays: mouse monoclonal antibodies to FLAG M2 (Sigma, 1:100 for in-cell Western, 1:250 for IF); rabbit anti-phosphocofilin (Cell Signaling, 1:1000 for WB); mouse anti-total cofilin (BD Biosciences, 1:1000 for WB); rabbit anti-phospho-ERK1/2 and mouse antitotal ERK1/2 (Cell Signaling, 1:1000 for WB and in-cell Western, 1:250 for IF); mouse monoclonal antibody to EEA-1 (BD Biosciences, 1:250 for IF); rabbit anti-LAMP1 (Santa Cruz Biotechnology, Inc. (Santa Cruz, CA), 1:250 for IF). Alexa546tagged secondary antibodies to mouse and rabbit were obtained from Invitrogen (1:500 for IF). IRDye680-and IRDye800tagged secondary antibodies (1:45,000 for WB) were from Rockland. 2-Furoyl-LIGRLO-NH 2 , Sar-Met-SP, U73122, and BAPTA-AM were purchased from Tocris. FLAG-tagged PAR 2 WT and PAR 2 S363A/T366A in the pBJ1 vector, Renilla luciferase-tagged ␤-arrestin-1 and -2 constructs, and FLAGtagged ␤-arrestin-1 and -2 constructs were obtained as gifts from Dr. JoAnn Trejo (University of California San Diego, La Jolla, CA), Dr. Michel Bouvier (University of Montreal), and Dr. Robert Lefkowitz (Duke University Medical Center), respectively. FLAG-tagged PAR 2 WT and PAR 2 S363A/T366A were subcloned from the pBJ1 vector into the p-eYFP-N1 vector using HindIII and BamHI. Human PAR 2 and NK1R, cloned into eYFP-N1 or eGFP-N1, at BamHI and HindIII sites have been described previously (3,10,20). The C termini of PAR 2 (nucleotide 1081 to the end) and NK1R (nucleotide 1227 to the end) were amplified by PCR with primers containing N-terminal EcoRI and C-terminal BamHI sites, digested, and ligated into eGFP-N1 to generate NK1RC-GFP and PAR 2 C-GFP. The N termini of PAR 2 (nucleotides 1-1080) and NK1R (nucleotides  were amplified by PCR with N-terminal HindIII and C-terminal EcoRI sites, digested, and ligated into EcoRI/ HindIII-digested NK1RC-GFP and PAR 2 C-GFP, respectively. Sequences and reading frames across the fusion junction were confirmed by ABI sequencing. Cell Culture and Transfection-Human embryonic kidney 293 (HEK293), mouse embryonic fibroblasts (MEFs) from wild type and ␤-arrestin-1/2 double knock-out mice (from Robert Lefkowitz), and Chinese hamster ovary (CHO) cell lines were maintained as described previously (3) . All plasmids were stably transfected in CHO cells using Lipofectamine and Plus reagent (Invitrogen) following the manufacturer's protocols. Stable CHO cell lines were selected by flow cytometry and were not single cell clonal cell lines. Transient transfections were carried out using FuGene (Roche Applied Science) following the manufacturer's protocols.
Single Cell Calcium Response Measurements-Cells were preloaded with Fura-2, washed, and mounted in a perfusion chamber on the stage of a Nikon TE300 microscope, and Ca 2ϩ mobilization was determined as described previously (3). Briefly, either 2fAP or Sar-Met-SP was added to the bath, and fluorescence was detected in individual cells using a Nikon video camera and a video microscopy program (Metafluor). Fluorescence was quantified at 340-and 380-nm excitation and 510-nm emission. The ratio of the fluorescence at the two excitation wavelengths, which is proportional to the [Ca 2ϩ ], was determined. To determine maximum Ca 2ϩ levels, readings were followed by the addition of the calcium ionophore ionomycin. For determination of the amount of calcium response (or the duration of signaling), areas under the calcium response curves were determined using Kaleidagraph (version 4.0), and total [Ca 2ϩ ] released was calculated using the Grynkiewitz equation, where K d (for Ca 2ϩ binding to Fura-2/AM) ϭ 220 nm, R ϭ A 340 /A 380 at each time point, R min ϭ A 340 /A 380 under Ca 2ϩ -free conditions, R max ϭ A 340 /A 380 at saturating conditions (with the addition of ionomycin), Sf2 represents the base-line fluorescence at 340 nm without Fura, and Sb2 represents the base-line fluorescence at 340 nm with Fura.
Bioluminescence Resonance Energy Transfer (BRET)-YFPtagged receptor constructs were transiently co-expressed with either ␤-arrestin-1-luciferase or ␤-arrestin-2-luciferase in HEK293 cells. 48 h post-transfection, the cells were treated with appropriate concentrations of 2fAP or Sar-Met-SP and 5 M coelenterazine. For dose curves, coelenterazine was added, and readings were taken 15 min after agonist stimulation. For kinetic measurements, agonist and coelenterazine were added simultaneously, and readings were initiated at that time. Pharmacological inhibitors were added at appropriate concentrations, and the cells were incubated at 37°C 10 min prior to the addition of 2fAP. Light emission was detected (460 -500 nm for RLuc and 510 -550 nm for YFP) using a TRISTAR LB941 multilabel plate reader from Berthold Technologies. BRET signal was calculated as the ratio of the light emitted by eYFP and the light emitted by luciferase. For a negative control, cells trans-fected with the ␤-arrestin-luciferase construct alone were used to determine the background. The ratio observed in ␤-arrestin luciferase-only-transfected cells was subtracted from that observed in the presence of YFP-tagged receptors to give the net BRET values. Half-lives (t1 ⁄ 2 ) of the kinetics reactions were determined from five separate experiments. For titration curves, ␤-arrestin was held constant and co-transfected with increasing amounts of YFP-tagged receptor; BRET readings were taken for each acceptor/donor ratio as described for dose curves but using a constant concentration of 1 M 2fAP. Protein expression levels were confirmed by luciferase emission and YFP fluorescence, respectively.
In-cell Western Assays-For pERK assays, cells were seeded into 96-well plates, serum-starved overnight, and then stimulated with 2fAP or Sar-Met-SP for 0 -60 min, after which they were fixed, blocked, and incubated with anti-rabbit pERK and anti-mouse total ERK1/2 overnight at 4°C. For cell surface receptor detection, HEK293 cells were transiently transfected with empty vector, FLAG-tagged PAR 2 WT, or PAR 2 S363A/ T366A, transferred to 24-well plates, fixed, blocked, and incubated with mouse anti-FLAG monoclonal antibody. Following incubation with IRDye680 and IRDye800-conjugated secondary antibodies, the plates were scanned using the LI-COR Odyssey imaging system, and the integrated intensities of the wells were quantified using the LI-COR software.
Immunofluorescence and Confocal Microscopy-For the receptor internalization assays, 3 ϫ 10 4 CHO cells stably expressing GFP-tagged receptors were cultured overnight on collagen-coated coverslips. When required, the cells were transiently transfected with DsRed-ERK1/2 or FLAG-tagged ␤-arrestin-1 or -2. 36 h post-transfection, medium was changed to serum-free DMEM, and the cells were treated with appropriate agonists (i.e. 1 M 2fAP or 100 nM Sar-Met-SP) for 0 -120 min, fixed, and blocked, and immunostaining was carried out as described previously (3).
Cell Migration Assay-CHO cells stably expressing GFPtagged PAR 2 , NK1R, PAR 2 -NK1R, NK1R-PAR 2 , and HEK293 cells transiently transfected with FLAG-tagged PAR 2 WT or PAR 2 S363A/T366A were used in these assays. 3 ϫ 10 4 cells were seeded onto collagen-coated transwell supports (8-m pore size) and allowed to attach for 2 h at 37°C and then treated with 10 M 2fAP or 100 nM Sar-Met-SP, added to the lower chambers, for 4 h. Non-migratory cells were removed with a cotton swab, filters were stained with crystal violet, and the total number of cells that migrated to the bottom was quantified counting in four fields of vision under the ϫ20 objective of a Nikon phase-contrast microscope. Alternatively, cells were grown to confluence in 35-mm dishes and serum-starved overnight, and a wound was generated by scratching across a monolayer with a 10-l pipette tip. After treatment with or without 1 M 2fAP for 6 -24 h, cell migration into the wound area was monitored using the ϫ4 objective of a Nikon Eclipse TE2000U microscope. The wound area was quantified using ImageJ software (wound area at time 0 minus the wound area at each time point). The migration index was calculated as a -fold change in area covered by cells at the 0 min time point over that at the 24 h time point.
Proliferation Assay-10 4 CHO cells expressing GFP-tagged PAR 2 , NK1R, PAR 2 -NK1R, and NK1R-PAR 2 were seeded onto 35-mm dishes, attached for 2 h, and serum-starved overnight and then treated with 2fAP, Sar-Met-SP, or serum (positive control) for 48 h. Cells were detached, stained with propidium iodide, and resuspended in 1 ml of flow cytometry buffer. Cells per ml were determined using a Beckman flow cytometer. Cell number was determined by multiplying the number of propidium iodide-excluding GFP-positive cells times the total volume. Untransfected cells were used to determine the gating strategy for identifying GFP-positive cells.

RESULTS
The C Termini of PAR 2 and NK1R Determine ␤-Arrestin-dependent Desensitization and Signaling Patterns-To determine the importance of the receptor C termini in directing specific ␤-arrestin-dependent signaling events, we constructed chimeric receptors, in which either the cytoplasmic C terminus of PAR 2 was fused to the N terminus of NK1R at the seventh transmembrane domain (NK1R-PAR 2 ) or the C terminus of NK1R was fused to the N terminus of PAR 2 (PAR 2 -NK1R) ( Fig.  1A). They were expressed in CHO cells, and agonist-induced Ca 2ϩ mobilization was determined using Fura-2/AM (PAR 2 and PAR 2 -NK1R received 2fAP, and NK1R and NK1R-PAR 2 received Sar-Met-SP). Mobilization of Ca 2ϩ from intracellular stores is an early signaling event downstream of G␣ q -coupled receptors, such as PAR 2 and NK1R, and is rapidly terminated upon ␤-arrestin recruitment. Thus, Ca 2ϩ mobilization is often used as a read-out for G␣ q -coupled receptor activation and for examining defects in desensitization (1,3,15,21). The chimeric receptors, like their wild type C-terminal parent receptors, are expressed at the cell surface ( Fig. 1B), colocalize with ␤-arrestins upon activation (supplemental Fig. S1), and are able to mobilize calcium in response to agonist in a dose-dependent fashion, demonstrating that all four receptors are functional ( Fig. 1, C and D). No significant differences in the dose responses of chimeric receptors and the corresponding parent receptors were observed ( Fig. 1, C and D). As observed previously, the duration of the NK1R-induced Ca 2ϩ signal was longer than that following PAR 2 activation, resulting in a 2.2-fold increase in the total concentration of intracellular Ca 2ϩ (Fig.  1E). The chimeric receptors display Ca 2ϩ mobilization patterns similar to their C-terminal parent (i.e. PAR 2 -NK1R (Fig. 1I) displays a prolonged 2fAP-induced Ca 2ϩ response, and NK1-PAR 2 (Fig. 1H) displays a transient Sar-Met-SP-induced response). All four receptors also demonstrated agonist-induced internalization (Fig. 2). As reported previously, PAR 2 internalized rapidly, as did NK1R-PAR 2 , colocalizing with EEA1 after 5 min (supplemental Fig. S2, A and C) and with lysosomal marker LAMP1 after 1 h of 2fAP treatment (supplemental Fig. S3). NK1R and PAR 2 -NK1R colocalized with EEA1 at 15 min (supplemental Fig. S2, B and D). Unlike PAR 2 and NK1R-PAR 2 , NK1R and PAR 2 -NK1R colocalized with EEA1 at 15 and 30 min of agonist treatment, showing very little colocalization with LAMP1 (supplemental Figs. S2 (B and D) and S3). Thus, the C terminus of the two receptors is also important for determining the pattern of internalization.

Determinants of ␤-Arrestin-dependent Signaling by PAR 2 and NK1R
FEBRUARY 1, 2013 • VOLUME 288 • NUMBER 5 Both PAR 2 and NK1R activate ERK1/2 through G-proteinand ␤-arrestin-dependent pathways, but although activated ERK1/2 downstream of PAR 2 is primarily found at the membrane and in the cytoplasm, downstream of NK1R, activated ERK1/2 is primarily nuclear (1-3, 5). Western blot analysis with anti-pERK shows that the chimeric receptors were also able activate ERK1/2 in response to the appropriate agonist (Fig. 3, A-E). To examine the subcellular localization of ERK1/2, cells expressing each of the four GFP-tagged receptors were treated with or without 2fAP or Sar-Met-SP, stained for phospho-ERK ( Fig. 3F) or total ERK1/2 (supplemental Fig. S4A), and examined by confocal microscopy. Additionally, cells expressing each receptor were co-transfected with DsRed-ERK2, treated with 2fAP or Sar-Met-SP, and examined by confocal microscopy (supplemental Fig. S4B). As previously reported, in cells expressing PAR 2 , phospho-ERK1/2 was primarily observed in the cytoplasm and near the plasma membrane after activation, whereas in cells expressing NK1R, most of the phospho-ERK1/2 translocated into the nucleus (Fig. 3F). After treatment with Sar-Met-SP, activated ERK1/2 distribution in cells expressing NK1R-PAR 2 was predominantly cytoplasmic and near the plasma membrane (Fig. 3F), whereas 2fAP treatment of cells expressing PAR 2 -NK1R resulted in nuclear translocation of ERK1/2 (Fig. 3F). Total ERK1/2 localization showed a similar pattern of localization (supplemental Fig. S4).
PAR 2 and NK1R Recruit ␤-Arrestins with Different Apparent Affinity and Kinetics-The data described thus far suggest that the C termini of the two receptors determine the mechanism of ERK1/2 activation and subsequent subcellular localization. Although both receptors have been shown to interact with ␤-arrestins by multiple groups (1,2,4,10,22,23), the differences in downstream signaling events suggest that they might direct distinct patterns of ␤-arrestin recruitment. To examine the more subtle differences in recruitment of ␤-arrestin-1 and -2 to PAR 2 and NK1R, BRET assays were employed. First, agonist dose responses of ␤-arrestin recruitment to each receptor at 15 min were determined, which revealed no significant differences in the agonist dose required to recruit ␤-arrestin to either PAR 2 or NK1R (Fig. 5, A and B). We then examined the kinetics of recruitment by examining BRET in response to a single dose of agonist over 20 min. Agonist stimulation of PAR 2 led to rapid recruitment of both ␤-arrestin-1 and -2 at equivalent rates (Fig. 5, C-E). In contrast, recruitment of both ␤-arrestins to NK1R was 2-fold slower than for PAR 2 (Fig. 5, C-E), and recruitment of ␤-arrestin-1 to NK1R was significantly slower than recruitment of ␤-arrestin-2. Another observation apparent in both the dose response and kinetic assays was that net BRET values for ␤-arrestin-1 and -2 recruitment to PAR 2 were higher than those for NK1R, which could reflect a higher affinity of both ␤-arrestins for PAR 2 or a conformational difference in the receptor/␤-arrestin conformation that results in a greater distance between the luciferase and YFP tags. To distinguish between these possibilities, we monitored net BRET as a function of the acceptor/donor ratio (receptor-YFP/␤-arrestinluciferase) and determined the acceptor-donor ratio at which half-maximal BRET (BRET 50 ) is observed. These studies revealed several key differences between the two receptors. First, BRET 50 values for NK1R/␤-arrestin interactions were higher than those for PAR 2 /␤-arrestin interactions (nearly 2-fold for ␤-arrestin-2 and 2.7-fold for ␤-arrestin-1), suggesting that PAR 2 has a higher relative affinity for both ␤-arrestins than NK1R. Second, although no significant difference in BRET 50 was observed for interactions between either ␤-arrestin-1 or -2 and PAR 2 , the BRET 50 value was significantly higher for ␤-arrestin-1/NK1R than for ␤-arrestin-2/NK1R, suggesting that NK1R preferentially binds ␤-arrestin-2 over ␤-arrestin-1 (Fig.  5F). Consistent with our data demonstrating a requirement for the C terminus in a number of ␤-arrestin-dependent events, the chimeric receptors displayed ␤-arrestin recruitment rates similar to their C-terminal parent (i.e. PAR 2 -NK1R preferentially recruited ␤-arrestin-2, whereas NK1R-PAR 2 recruited both equally) (Fig. 6A).
pathways. To determine whether swapping the C termini of the two receptors also affected the cellular events resultant from ␤-arrestin engagement, we looked at proliferation and cell migration downstream of all four receptors. As previously reported, activation of PAR 2 promoted a 4-fold increase in cell migration; however, NK1R did not significantly affect cell migration (Fig. 6B). Conversely, activation of NK1R-PAR 2 increased cell migration by 3-fold, whereas activation of PAR 2 -NK1R did not (Fig. 6B). Similarly, NK1R and PAR 2 -NK1R, but not PAR 2 and NK1R-PAR 2 , promoted a 2-fold increase in proliferation (Fig. 6C). Thus, the ultimate cellular consequence of ␤-arrestin recruitment to PAR 2 and NK1R is determined by the C terminus of each receptor.
Recruitment of ␤-Arrestins to PAR 2 but Not to NK1R Is Independent of G␣ q Signaling-Although PAR 2 can promote ␤-arrestin-dependent signaling independent of G␣ q , NK1R appears to require integration of both G␣ q and ␤-arrestin signaling to activate ERK1/2. To determine whether the two receptors differ in their requirement for initial G␣ q coupling, we examined ␤-arrestin-1/2 recruitment in the presence of pharmacological inhibitors of phospholipase C␤ (U73122) and PKC (GFX) and a chelator of intracellular Ca 2ϩ (BAPTA-AM) using BRET. Both ␤-arrestin-1 and -2 were recruited to PAR 2 , following agonist stimulation in the presence of all three inhibitors (Fig. 7A). On the contrary, all three inhibitors led to a significant reduction in recruitment of ␤-arrestins to NK1R (Fig. 7B).
Previous studies have suggested that mutation of two putative PKC phosphorylation sites in PAR 2 (Ser-363 and Thr-366; PAR 2 S363A/T366A) inhibits stable colocalization of ␤-arrestin with PAR 2 , receptor desensitization, and internalization and membrane activation of ERK1/2. This mutant receptor robustly promoted nuclear translocation of activated ERK1/2 and proliferation, whereas the wild type receptor did so only weakly (2). Because inhibition of PKC did not abolish recruitment of ␤-arrestin to PAR 2 , we examined whether more subtle features of ␤-arrestin/receptor interactions were affected by mutation of these residues. Both receptors are expressed on the cell surface, and both are capable of recruiting ␤-arrestin-1 and -2 at the same agonist dose (Fig. 8, A and B). Kinetic BRET assays revealed that the rate of recruitment to PAR 2 S363A/ T366A was reduced 1.5-and 2-fold for ␤-arrestin-1 and -2, respectively, compared with the wild type PAR 2 (Fig. 8, C and  D). Furthermore, BRET 50 values were increased for the phosphomutant, suggesting that it had a lower relative affinity for both ␤-arrestins than the wild type receptor. The BRET max value was also decreased, indicating that the nature of the receptor/␤-arrestin interactions was different such that the acceptor and donor tags were in closer proximity with the wild type receptor. To examine whether PAR 2 S363A/T366A was deficient in other aspects of ␤-arrestin-dependent PAR 2 signal-FIGURE 6. ␤-Arrestin-dependent cellular effects of the chimeric receptors are similar to their respective C-terminal parent. A, maximal net BRET signal was determined for PAR 2 , NK1R, PAR 2 -NK1R, or NK1R-PAR 2 with either ␤-arrestin-1 or -2 as described in the legend to Fig. 5. *, significant difference from N-terminal parent; #, significant difference from ␤-arrestin-1 (p Ͻ 0.05, n ϭ 3). B, cells transfected with each of the four receptors were seeded onto transwell filters. Cell migration is expressed as a mean Ϯ S.E. (error bars) increase in number of cells that migrated after 2fAP or Sar-Met-SP treatment compared with untreated cells. *, significant increase in cell migration; #, significant difference between bracketed groups (p Ͻ 0.01, n ϭ 4). C, cells transfected with each receptor serum-starved and then treated with or without 2fAP, Sar-Met-SP, or serum (positive control) for 24 h. Mean Ϯ S.E. increase in total cell number is shown. *, significant increase in proliferation (p Ͻ 0.01, n ϭ 4). ing, we examined cofilin dephosphorylation and cell migration. Consistent with its decreased ability to stably recruit ␤-arrestins, PAR 2 S363A/T366A promoted a transient increase in cofilin phosphorylation, which returned to base line within 10 min, and no cofilin dephosphorylation (activation) was observed (Fig. 9, A and B). In contrast, activation of PAR 2 WT resulted in a 50% dephosphorylation of cofilin (Fig. 9, A and B). Similarly, whereas wild type PAR 2 promoted a 2-fold increase in cell migration, PAR 2 S363A/T366A did not promote cell migration (Fig. 9C).

DISCUSSION
Since the discovery of ␤-arrestin-dependent signaling, it has become apparent that not all GPCRs elicit the same signals upon recruitment of ␤-arrestins. The studies presented here demonstrate that differences in ␤-arrestin-dependent signaling are determined in part by specific interactions between ␤-arrestins and the C terminus of each receptor. Here we used two G␣ q -coupled receptors, PAR 2 and NK1R, both of which utilize ␤-arrestins for activation of ERK1/2 but do so by distinct mechanisms with different outcomes, to examine the receptor features underlying these signaling differences. We provide evidence that by swapping the C termini of the two receptors, we can force PAR 2 agonists to induce a ␤-arrestin-dependent signaling pattern similar to that of NK1R and NK1R agonists to induce a signaling pattern similar to that of PAR 2 . This result suggests that each receptor may have a unique "␤-arrestin fingerprint" that ultimately determines the specific scaffolds formed on the receptor-bound ␤-arrestin as well as the downstream signaling events. This model would predict that interactions with different GPCR C-terminal residues may induce subtle differences in ␤-arrestin conformation that then expose distinct sets of binding partner sites.
In these studies, we compared and contrasted different ␤-arrestin functions: 1) recruitment to each receptor, 2) duration of G␣ q response, 3) signaling events (e.g. ERK1/2 activation mechanism and cofilin activation), and 4) functional responses (e.g. cell migration and proliferation). The results of these comparisons suggest that PAR 2 , which promotes ␤-arrestin-dependent membrane-associated ERK1/2 activity, cofilin activation, and chemotaxis, rapidly recruits ␤-arrestins and demonstrates equal apparent affinities for both ␤-arrestin-1 and -2. In contrast, NK1R, which promotes ␤-arrestin-dependent nuclear ERK1/2 activity and proliferation, does not activate cofilin and preferentially recruits ␤-arrestin-2 over ␤-arrestin-1. Furthermore, recruitment of both ␤-arrestins to NK1R occurs significantly more slowly, and with a lower apparent affinity, than to PAR 2 . These results allow us to make certain predictions regarding the relationship between ␤-arrestin recruitment patterns and downstream effects.
Ca 2ϩ assays performed in this study specifically measure intracellular Ca 2ϩ , because EGTA is included in the medium bathing the cells, and are thus considered a read-out of G␣ q -dependent signaling. We have shown that activated NK1R promotes a prolonged release of intracellular Ca 2ϩ compared with A, cell surface expression of N-terminally FLAG-tagged PAR 2 WT-YFP and PAR 2 S363A/T366A-YFP was determined by in-cell Western assays using anti-FLAG. *, p Ͻ 0.05, n ϭ 3. B-F, BRET assays were performed with FLAG-tagged PAR 2 -YFP or PAR 2 S363A/T366A-YFP and ␤-arrestin-1/2-Rluc. Net BRET ratio was estimated in response to incremental doses of 2fAP. B, inset, percentage of maximal BRET response to the increasing 2fAP doses. C, net BRET response was quantified over time in response to 1 M 2fAP. D, mean Ϯ S.E. half-lives (t1 ⁄2 ) of BRET kinetics are shown as mean Ϯ S.E. (error bars) (significant differences between bracketed groups: *, p Ͻ 0.03; **, p Ͻ 0.008; n ϭ 3). E, titration curves monitoring net BRET in response to varying acceptor/donor ratios were performed as in Fig. 5F. F, BRET 50 and BRET max for PAR 2 WT or PAR 2 S363A/T366A were computed from E. Both values were significantly different, reduced in PAR 2 S363A/T366A compared with PAR 2 WT (p Ͻ 01, n ϭ 3). PAR 2 , which ultimately results in a higher concentration of total Ca 2ϩ released into the cytosol. We have previously demonstrated that, although both ␤-arrestins are required for full receptor down-regulation, genetic deletion of ␤-arrestin-1 but not ␤-arrestin-2, results in a prolonged Ca 2ϩ signal downstream of PAR 2 (3). Thus, the lower apparent affinity of NK1R for ␤-arrestin-1 and the slower recruitment kinetics of both ␤-arrestins to NK1R are consistent with the longer duration of the Ca 2ϩ signal. Further supporting this model, in which the pattern of ␤-arrestin recruitment determines the signal duration, the chimeric receptor containing the N terminus of PAR 2 and the C terminus of NK1R showed the same pattern of ␤-arrestin recruitment as NK1R, along with a prolonged Ca 2ϩ signal. Conversely, the chimeric receptor containing the N terminus of NK1R and the C terminus of PAR 2 had a short Ca 2ϩ signal and a ␤-arrestin recruitment pattern similar to PAR 2 . Because ␤-arrestins directly uncouple G-protein/GPCR signals as well as facilitate clathrin-mediated endocytosis, these results suggest that residues within the C terminus of each receptor determine the relative affinities for each ␤-arrestin, which in turn affects how efficiently the G-protein signal is terminated.
A more puzzling distinction between PAR 2 and NK1R signaling that is common to many GPCRs is the observation that although both receptors require ␤-arrestins for full activation of ERK1/2, they do so by different mechanisms with drastically different cellular responses. PAR 2 utilizes separate G-protein-/Ca 2ϩdependent and ␤-arrestin-dependent mechanisms, whereas NK1R requires both Ca 2ϩ and ␤-arrestins, and the ␤-arrestin signal induced by NK1R is not independent from the G-protein signal (1,2). Here we demonstrate that recruitment of both ␤-arrestins to PAR 2 occurs even in the presence of inhibitors of the G-protein signaling arm, whereas recruitment to NK1R is impaired by inhibition of the G␣ q effector phospholipase C␤ or by chelation of intracellular Ca 2ϩ . Furthermore, PAR 2 promotes formation of a stable complex containing ␤-arrestins and the entire ERK1/2 module (Raf, MEK1, and ERK1/2), leading to Src-independent, membrane-associated ERK1/2 activity and cell migration (2,5,13). In contrast, NK1R promotes formation of a transient complex containing ␤-arrestins and Src but not Raf, leading to nuclear translocation of ERK1/2 and proliferation (1). Here we show that the subcellular localization of activated ERK1/2 downstream of the chimeric receptors and the ultimate consequence of ERK1/2 activation (e.g. proliferation versus chemotaxis) follow that of the C-terminal parent.
The traditional model for GPCR signaling predicts that receptor phosphorylation, by GRKs or second messenger kinases, creates binding sites for ␤-arrestins. Indeed, studies have suggested that C-terminal phosphorylation of both PAR 2 and NK1R is essential for receptor desensitization, internalization, and ERK1/2 activation (2,19,25). Although regulation of NK1R/␤-arrestin recruitment and internalization by GRKs has been demonstrated, similar GRK-mediated regulation of PAR 2 has only been hypothesized (9,10). Regulation of both receptors by PKC phosphorylation has been suggested by the fact that desensitization and internalization of both are sensitive to broad-spectrum PKC inhibitors, and mutation of putative PKC phosphorylation sites in both receptors hinders receptor desensitization and internalization (2,18,19). However, when BRET was used to more closely monitor real-time ␤-arrestin interactions with a mutant PAR 2 previously shown to be defective in ␤-arrestin colocalization and PKC-induced desensitization, we observed that ␤-arrestin-1 and -2 are both recruited to this mutant receptor but with a significantly slower rate and lower apparent affinity. These data suggest that interactions between phosphorylated Ser-363 and Thr-366 in the C-tail of PAR 2 strengthen its interaction with ␤-arrestins but are not necessary for initial recruitment. Consistent with the decreased apparent affinity for ␤-arrestins, PAR 2 S363A/T366A also fails to promote cofilin dephosphorylation or cell migration, which are hallmarks of ␤-arrestin-dependent signaling downstream of PAR 2 . Importantly, differences in the C-terminal phosphorylation pattern of GPCRs may contribute to the stability of their respective interactions with ␤-arrestins and, thus, to the differences in their downstream signaling patterns. There are a number of GPCRs that appear to elicit similar G-protein-independent, ␤-arrestin-dependent signaling as PAR 2 . Likewise, there are receptors that, like NK1R, promote ␤-arrestin-dependent recruitment of Src and nuclear ERK1/2 activation and proliferation (26). Other studies using chimeric angiotensin and vasopressin receptors have suggested that the stability of the ERK1/2 scaffolding complexes is dependent upon the C termini of the receptors (27). Our previous studies had demonstrated that the native ␤-arrestin⅐ERK1/2 complex formed in response to PAR 2 was sufficiently stable to remain intact on a size exclusion column, whereas the complex formed in response to NK1R required cross-linking in order to survive the purification process (1,2). These studies suggested that the PAR 2 -associated complex is significantly more stable than that formed downstream of NK1R. Thus, the differences in the ␤-arrestin signaling pathways elicited by PAR 2 and NK1R may be dependent upon the stability of the complex formed, which previous studies demonstrated is dependent upon interactions between ␤-arrestins and receptor C termini (27). This hypothesis is further supported by the observation in the BRET studies presented here that ␤-arrestins display a higher affinity for PAR 2 compared with NK1R, and the pattern of ␤-arrestin recruitment is dependent upon the receptor C terminus. Interestingly, the PAR 2 phosphomutant, which was previously shown to be deficient in desensitization and ␤-arrestin ERK1/2 localization (2), displays a lower ␤-arrestin affinity and slower recruitment kinetics than the wild type receptor. Like the NK1R, it does not promote cofilin activation or cell migration, but it promotes proliferation to a greater extent than the wild type receptor. Although there are studies indicating phosphoreceptor-binding sites on ␤-arrestins (28,29), it is likely that there are receptor-specific differences that are yet to be appreciated, which may contribute to more subtle differences in signaling depending upon the phosphorylation state of the receptor. These studies also suggest that differences in the phosphorylation patterns of different GPCRs influence the stability of ␤-arrestin interactions. Ultimately, differences in receptor phosphorylation may in turn affect which binding domains on ␤-arrestin are exposed, resulting in association of certain putative binding partners and exclusion of others. Recent studies using a ␤-arrestin biosensor to detect gross changes in conformation suggested that biased agonists of several GPCRs, capable of eliciting ␤-arrestin-de-pendent signaling in the absence of G-protein coupling, promote a conformation that is distinct from that elicited by standard agonists (30,31). Thus, differences in ␤-arrestin conformation in response to recruitment to PAR 2 versus NK1R might underlie the differences in signaling we observe. Ultimately, these subtle differences in ␤-arrestin/receptor interactions lead to dramatic differences in cellular responses (e.g. proliferation versus cell migration).