Trafficking of Proteinase-activated Receptor-2 and β-Arrestin-1 Tagged with Green Fluorescent Protein

Proteases cleave proteinase-activated receptors (PARs) to expose N-terminal tethered ligands that bind and activate the cleaved receptors. The tethered ligand, once exposed, is always available to interact with its binding site. Thus, efficient mechanisms must prevent continuous activation, including receptor phosphorylation and uncoupling from G-proteins, receptor endocytosis, and lysosomal degradation. β-Arrestins mediate uncoupling and endocytosis of certain neurotransmitter receptors, which are activated in a reversible manner. However, the role of β-arrestins in trafficking of PARs, which are irreversibly activated, and the effects of proteases on the subcellular distribution of β-arrestins have not been examined. We studied trafficking of PAR2 and β-arrestin1 coupled to green fluorescent protein. Trypsin induced the following: (a) redistribution of β-arrestin1 from the cytosol to the plasma membrane, where it co-localized with PAR2; (b) internalization of β-arrestin1 and PAR2 into the same early endosomes; (c) redistribution of β-arrestin1 to the cytosol concurrent with PAR2 translocation to lysosomes; and (d) mobilization of PAR2 from the Golgi apparatus to the plasma membrane. Overexpression of a C-terminal fragment of β-arrestin-319–418, which interacts constitutively with clathrin but does not bind receptors, inhibited agonist-induced endocytosis of PAR2. Our results show that β-arrestins mediate endocytosis of PAR2 and support a role for β-arrestins in uncoupling of PARs.

cellular N termini to expose tethered ligand domains that bind to and activate the cleaved receptors.
Despite the irreversible nature of proteolytic activation, signaling by PARs, in common with other GPCRs, is rapidly terminated. For neurotransmitter receptors, which are activated in a reversible manner, the mechanisms of signal termination have been extensively investigated (9). These mechanisms include (a) receptor phosphorylation and uncoupling from heterotrimeric G-proteins and (b) receptor endocytosis. G-protein-coupled receptor kinases (GRKs) and ␤-arrestins are cytosolic proteins that translocate to the plasma membrane to mediate uncoupling and endocytosis of GPCRs for certain neurotransmitters. For example, GRK2 and -3 phosphorylate ␤ 2adrenergic receptors (␤ 2 ARs) (10 -12). ␤-Arrestins interact with GRK-phosphorylated ␤ 2 ARs to disrupt their association with G-proteins and terminate signaling (12)(13)(14) and also to serve as adaptor proteins for clathrin-mediated endocytosis of ␤ 2 ARs (15,16). Receptor endocytosis and trafficking also contribute to resensitization and signal transduction. Resensitization requires endocytosis, dephosphorylation, and recycling of ␤ 2 ARs to the plasma membrane (17)(18)(19)(20), and endocytosis of ␤ 2 ARs is required for stimulation of mitogen-activated protein kinases (21).
PAR2 is highly expressed in the gastrointestinal tract and pancreas (6,8,28,29). Pancreatic trypsin activates PAR2 in enterocytes and in epithelial cells of the pancreatic duct to regulate prostanoid secretion and the activity of ion channels (29,30). Tryptase, a major protease of human mast cells (31), activates PAR2 in myocytes and enteric neurons (8,32), and PAR2 may mediate some of the mitogenic and pro-inflammatory effects of tryptase (33). In view of the potential importance of PAR2 in normal regulation, inflammation, and mitogenesis, it is important to understand the mechanisms that terminate PAR2 signaling. Although we have shown that trypsin induces endocytosis and lysosomal degradation of PAR2 (23), agonistinduced trafficking of PAR2 and ␤-arrestin1 has not been ex-amined in real time, and the role of ␤-arrestins in PAR2 endocytosis is unknown.
The purpose of the present study was to investigate agonistinduced trafficking of ␤-arrestin1 and PAR2 and to determine the role of ␤-arrestins in trafficking of PAR2. To examine trafficking of PAR2 and ␤-arrestin1 in real time, we expressed chimeras of these proteins and green fluorescent protein (GFP).
Antibodies-A rabbit polyclonal antibody (HA.11) to the hemagglutinin 12CA5 epitope (YPYDVPDYA) and a monoclonal antibody against the resident Golgi protein mannosidase II were from Berkeley Antibody Co. (Richmond, CA). Monoclonal M1 antibody to the Flag epitope (DYK-DDDDK) was from International Biotechnologies, Inc. (New Haven, CT). A transferrin receptor monoclonal antibody was from Dr. Ian Trowbridge (Salk Institute, San Diego, CA). A monoclonal antibody (GM10) recognizing lysosomal acidic membrane protein-1 (LAMP-1) was from Dr. John Hutton (Cambridge, United Kingdom). A rabbit polyclonal antibody raised to a fusion protein of GFP and glutathione S-transferase (number 9708) was provided by Dr. John H. Walsh (CURE/UCLA Antibody Core). Affinity purified goat anti-mouse and anti-rabbit IgG conjugated to Texas Red were from Jackson ImmunoResearch (West Grove, PA) and Cappel Research Products (Durham, NC). Goat anti-mouse IgG conjugated to (R)-phycoerythrin were from Caltag Laboratories (Burlingame, CA). Goat anti-rabbit IgG conjugated to horseradish peroxidase was from Pierce.
Generation of PAR2-GFP and ␤-Arrestin1-GFP Constructs-Constructs with EGFP at the C terminus of human PAR2, rat ␤-arrestin1, and dominant negative rat ␤-arrestin1 319 -418 were generated by polymerase chain reaction ( Fig. 1) using the following primers and Pfu DNA polymerase. For PAR2, the forward primer was 5Ј-CTTCGAATTCGC-CACCATGCGGAGCCCCAGCGCGGCG-3Ј (EcoRI site underlined, Kozak translation initiation site, and N terminus of PAR2 cDNA in bold), and the reverse primer was 5Ј-CGGTGGATCCCGATAGGAGG-TCTTAACTGTGGTTGAAC-3Ј (BamHI site underlined, C terminus of PAR2 cDNA in bold). A 1319-base pair fragment was amplified and separated on agarose gel and purified using QiaEx extraction kit. For ␤-arrestin1, the forward primer was 5Ј-GGCCGGAAGCTTGCCACCA-TGGGCGACAAAGGGACACGA-3Ј (HindIII site underlined, N terminus of ␤-arrestin1 cDNA in bold), and the reverse primer was 5Ј-GGCCGGCCGCGGTCTGTTGTTGAGGTGTGGAGA-3Ј (SacII site underlined, C terminus of ␤-arrestin1 cDNA in bold). A 1284-base pair fragment was amplified and purified. A fragment of bovine ␤-arrestin1 (residues 319 -418) acts as a dominant negative mutant (34). A construct of rat ␤-arrestin1 319 -418 and EGFP was generated using the forward primer 5Ј-GGCCGGAAGCTTGCCACCATGGTTTCCTACAA-AGTCAAAGTG-3Ј (HindIII site underlined, methionine, and sequence corresponding to ␤-arrestin1 319 -325 in bold) and the same reverse primer as for wild type ARR-GFP. All constructs were ligated into pEGFP-N1 vector and used to transform JM109 Escherichia coli in Luria Broth containing 30 g/ml kanamycin. The sequences of the chimeras were verified using a reverse primer (5Ј-CGTCGCCGTCCA-GCTCGACCAG-3Ј) corresponding to the 46 -67-base pair region of the GFP.
Generation of Cell Lines-KNRK cells were stably transfected with the PAR2-GFP chimera vector containing a neomycin resistance gene using Lipofectin (KNRK-PAR2-GFP cells) (23). Clones were screened by fluorescence microscopy and enriched for expression of PAR2-GFP by fluorescence-activated cell sorting. The generation of a cDNA construct encoding N-terminal proopiomelanocortin signal peptide, the Flag epitope, human PAR2, and a C-terminal HA.11 epitope has been described ( Fig. 1) (23). This construct was subcloned into pcDNA3.1 hygro (ϩ). KNRK cells were simultaneously transfected with cDNA encoding this PAR2 construct, in a hygromycin-resistant vector, and ARR-GFP or ARR 319 -418 -GFP, in neomycin-resistant vectors using Lipofectin. Cells were selected in medium containing 0.8 mg/ml G418 and 0.3 mg/ml hygromycin B. After 1 week, cells were analyzed by fluorescence-activated cell sorting to detect PAR2 and ARR-GFP or ARR 319 -418 -GFP. Single cells expressing both constructs were sorted into 96-well plates, expanded, and screened for by immunofluorescence to detect PAR2 and ARR-GFP or ARR 319 -418 -GFP. Cells were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, 0.4 mg/ml G418, and 0.15 mg/ml hygromycin B (as appropriate). Cells were plated 24 -48 h before experiments on glass coverslips coated with poly-L-lysine for measurement of [Ca 2ϩ ] i and for microscopy or on plastic for other studies.
Flow Cytometry-Flow cytometry was used to monitor expression of constructs, enrich populations of cells, and assess expression of PAR2 at the plasma membrane after agonist treatment. To detect PAR2-GFP, ARR-GFP, and ARR 319 -418 -GFP, cells were dissociated in enzyme-free buffer, aliquoted to 1.5 ϫ 10 6 cells per tube, and resuspended in 1 ml of DMEM containing 5% enzyme-free cell dissociation buffer, 0.3% fetal bovine serum, and 2 g/ml propidium iodide. In cells co-expressing PAR2 and ARR-GFP or ARR 319 -418 -GFP, PAR2 was detected using the M1 antibody to the extracellular Flag epitope and a phycoerythrinconjugated secondary antibody. Cells were aliquoted to 1.5 ϫ 10 6 cells per tube and resuspended in 1 ml Iscove's medium containing 1 mg/ml bovine serum albumin (BSA) and 1 mM CaCl 2 (required for the binding of the M1 antibody). Cells were incubated with 10 g/ml M1 antibody for 1 h at 4°C, washed, and incubated with 2 g/ml phycoerythrinconjugated goat anti-mouse IgG for 1 h at 4°C. Cells were washed and suspended in 1 ml of DMEM containing 1 mM CaCl 2 , 5% enzyme-free cell dissociation buffer, 0.3% fetal bovine serum, and 2 g/ml propidium iodide. To examine the effects of agonists on expression of PAR2 at the cell surface, cells were incubated with 500 M AP for 0 -30 min at 37°C, and surface Flag immunoreactivity was quantified as described. Like trypsin, AP induces PAR2 endocytosis (23) but without removing the Flag epitope. Cells were analyzed with a Facscan Flow Cytometer (Becton Dickinson, Franklin Lakes, NJ). Fluorophores were excited at 488 nm, and emission was collected at 530/30 nm for GFP and 575/25 nm for phycoerythrin. Viability was assessed by exclusion of propidium iodide.
Western Blotting-Western blotting was used to confirm expression of GFP chimeras (35). Cells (10 7 ) were pelleted and lysed in 1 ml of Laemmli buffer, lysates were fractionated by SDS-polyacrylamide gel electrophoresis (4 -15%), and proteins were transferred to nitrocellulose. Membranes were incubated with 3% non-fat milk in 100 mM phosphate-buffered saline (PBS, pH 7.4) for 1 h and with GFP antibody 9708 (1:1,000 -20,000) overnight at room temperature. Membranes were washed and incubated with goat anti-rabbit IgG conjugated to horseradish peroxidase (1:8,000) for 1 h at room temperature. Blots were washed, and bands were detected on film using the ECL detection kit. Controls included preabsorption of the diluted primary antibody with the GFP fusion protein (1-2 g/ml) for 1 h at 37°C and use of non-transfected KNRK cells.
Measurement of [Ca 2ϩ ] i -[Ca 2ϩ ] i was measured in cells expressing GFP chimeras using Fura-2/AM (36). Fluorescence was measured at 340 and 380 nm excitation and 510 nm emission, and the results were expressed as the ratio of the fluorescence at the two excitation wavelengths, which is proportional to the [Ca 2ϩ ] i . To generate concentrationresponse curves, cells were exposed once to trypsin or AP. To examine desensitization, cells were challenged repetitively with agonist. All observations were in n Ͼ3 experiments.
Microscopy and Immunofluorescence-To examine trafficking of GFP-labeled proteins in real time, cells were maintained at 37°C in DMEM containing 0.1% BSA (DMEM/BSA). The same cells were observed before and after addition of 10 nM trypsin. To localize proteins by immunofluorescence, cells were incubated in DMEM/BSA containing 10 nM trypsin for 0 -60 min at 37°C, washed, and incubated for up to 4 h in trypsin-free medium. Cells were fixed with 4% paraformaldehyde in 100 mM PBS, pH 7.4, for 20 min at 4°C. PAR2 was localized using GFP, the HA.11 antibody to the C-terminal 12CA5 epitope, or the M1 antibody to the N-terminal Flag epitope (Fig. 1). We have described use of HA.11 (23). M1 was used at 5 g/ml with 1 mM CaCl 2 , overnight at 4°C. ␤-Arrestin1 was detected using GFP. The Golgi apparatus, early endosomes, and lysosomes were localized using antibodies to mannosidase II, the transferrin receptor, and LAMP-1, respectively (23). All observations were made in duplicate in Ͼ3 separate experiments. In some experiments cells were preincubated with 10 mM NH 4 Cl for 30 min or with 70 M cycloheximide, 10 g/ml brefeldin A, or 50 M BAPTA-AM for 60 min prior to stimulation with trypsin; these drugs were included in all solutions throughout the experiment (23,37). Cells were observed with a Zeiss Axiovert microscope, an MRC 1000 confocal microscope (Bio-Rad) with a krypton/argon laser, and a Zeiss plan-Apochromat ϫ 100 oil-immersion objective (numerical aperture of 1.4, ϱ0.7). For observation of live cells, care was taken to minimize laser exposure by limiting the number of optical sections (2-3 per time point) and the laser intensity (Ͻ3%, aperture of 3-4 mm), to maintain fluorophore stability.

Generation of Cell Lines Expressing PAR2-GFP and PAR2
Plus ARR-GFP or ARR 319 -418 -GFP KNRK cells were used for transfection since we have extensively studied signaling and trafficking of PAR2 and other GPCRs in this cell line (23,37). The mechanism of PAR2 desensitization in KNRK cells resembles that observed in enterocytes (23). Thus, KNRK cells regulate PAR2 similarly to cells that naturally express this receptor.
We transfected KNRK cells with cDNA encoding PAR2-GFP in a neomycin-resistant vector or with PAR2 in a hygromycinresistant vector plus ARR-GFP or ARR 319 -418 -GFP in neomycin-resistant vectors. We screened cells by flow cytometry and fluorescence microscopy to obtain clonal cell lines. KNRK-PAR2-GFP clone 3C was used in all experiments since analysis by flow cytometry revealed a single population of cells that expressed PAR2-GFP at a high level ( Fig. 2A). Fluorescence microscopy confirmed the high, uniform expression of PAR2-GFP, which was detected at the plasma membrane and in a prominent perinuclear pool (Fig. 2B). KNRK-PAR2 ϩ ARR-GFP clone 16 was selected since analysis by flow cytometry identified a single population of cells that expressed PAR2 and ␤-arrestin1 at a high and uniform level ( Fig. 2A). ARR-GFP was cytosolic with no detectable localization at the plasma membrane (Fig. 2B). Fluorescence microscopy confirmed the co-expression of PAR2 and ARR-GFP (e.g. Fig. 6). Thus, there is no co-localization of PAR2 and ␤-arrestin1 in the unstimulated state. KNRK-PAR2 ϩ ARR 319 -418 -GFP clone 17 was selected since flow cytometry indicated that a large proportion of these cells co-expressed PAR2 and ARR 319 -418 -GFP ( Fig. 2A). ARR 319 -418 -GFP had a punctate distribution throughout the cell which resembles the distribution of clathrin in KNRK cells (37) (Fig. 2B). Since ARR 319 -418 -GFP constitutively interacts with clathrin (34), it is probable that this protein is co-localized with clathrin in KNRK-PAR2 ϩ ARR 319 -418 -GFP cells.
We analyzed cells by Western blotting with a GFP antibody to determine whether chimeric proteins of the predicted size were expressed. The antibody detected recombinant GFP with an apparent mass of ϳ27 kDa (Fig. 2C). A broad protein band of ϳ60 -90 kDa was detected in extracts of KNRK-PAR2-GFP cells. The predicted mass of human PAR2 is ϳ44 kDa, but there is one consensus site for N-linked glycosylation. Therefore, this protein probably represents glycosylated PAR2-GFP. A prominent protein of ϳ75 kDa was detected in extracts of KNRK-PAR2ϩARR-GFP cells, which corresponds to the known mass of ␤-arrestin1 (ϳ48 kDa) plus GFP (ϳ27 kDa). A protein band of ϳ40 kDa was detected in extracts of KNRK-PAR2ϩARR 319 -418-GFP cells, which corresponds to the known mass of ␤-arrestin-319 -418 (ϳ12 kDa) plus GFP (ϳ27 kDa). The level of expression of ARR-GFP was higher that of ARR 319 -418 -GFP, as judged by the intensity of the immunoreactive proteins. Thus, the GFP antibody recognizes proteins of the predicted sizes in transfected cell lines. The GFP antibody is specific because it did not interact with proteins in extracts of untransfected KNRK cells, and preabsorption of the antibody with GFP-GST fusion protein markedly diminished the signal in transfected cells.

Functional Characterization of Cells Expressing PAR2-GFP and PAR2 Plus ARR-GFP or ARR 319 -418 -GFP
Although GFP is a compact protein that has been attached to other GPCRs and ␤-arrestin2 without affecting signaling or trafficking (38 -41), it is important to verify that GFP does not affect the function of each protein to which it is coupled. To determine if GFP affected PAR2 signaling, we measured trypsin-and AP-induced Ca 2ϩ mobilization in KNRK-PAR2-GFP cells, KNRK-PAR2 ϩ ARR-GFP cells, and KNRK-PAR2 ϩ ARR 319 -418 -GFP cells. In all cell lines, trypsin and AP strongly stimulated Ca 2ϩ mobilization, although the concentration-response curve for cells expressing PAR2-GFP was right-shifted for both trypsin and AP (Fig. 3A). We did not complete full concentration-response curves for all cell lines, and thus we cannot accurately determine the efficacy and EC 50 values of the responses. However, approximate EC 50 values were as follows: KNRK-PAR2-GFP cells, trypsin 10 nM, AP 20 M; KNRK-PAR2 ϩ ARR-GFP cells, trypsin 2 nM, AP 2 M; KNRK-PAR2-ARR 319 -418 -GFP cells, trypsin 2 nM, AP 7 M (mean, n ϭ 3 experiments). In KNRK cells expressing wild type human PAR-2, we have previously reported that trypsin and AP stimulate Ca 2ϩ mobilization with potencies of 2 nM and 18 M, respectively (28). Thus, trypsin and AP are slightly less potent agonists in cells expressing PAR2-GFP. In all cell lines, 10 nM trypsin induced a prompt increase in [Ca 2ϩ ] i which was maximal within a few seconds and which declined to 50% of the maximal levels within 19 Ϯ 2 s for KNRK-PAR2-GFP cells, 21 Ϯ 1 s for KNRK-PAR2 ϩ ARR-GFP cells, and 99 Ϯ 13 s for KNRK-PAR2 ϩ ARR 319 -418 -GFP cells (mean Ϯ S.E., n ϭ 3 experiments) (Fig. 3, B-D). Therefore, the increase in [Ca 2ϩ ] i is more prolonged in cells expressing ARR 319 -418 -GFP.
To examine desensitization, we exposed cells to 10 nM tryp-sin, 100 M AP, or carrier (control) for 2 min, washed cells, and challenged them again with 10 nM trypsin or 100 M AP 5 min after the first challenge. The magnitude of responses to the second challenge was compared with the response of carriertreated cells to determine the extent of desensitization (Fig. 3, B-D). In KNRK-PAR2-GFP cells, exposure to trypsin caused 88 Ϯ 5% desensitization to a second trypsin exposure, and exposure to AP caused 45 Ϯ 4% desensitization to a second AP exposure (mean Ϯ S.E., n ϭ 3 experiments). In KNRK-PAR2 ϩ ARR-GFP cells, desensitization to trypsin was 52 Ϯ 10%, and desensitization to AP was 21 Ϯ 18%. In KNRK-PAR2-ARR 319 -418 -GFP cells, desensitization to trypsin was 91 Ϯ 2%, and desensitization to AP was 48 Ϯ 5%. Thus, in all cell lines, trypsin induces stronger desensitization than AP. We have previously reported that trypsin more strongly desensitizes Ca 2ϩ mobilization than does AP in KNRK cells expressing wild type PAR2 (23). Desensitization to trypsin and AP was less in cells expressing ARR-GFP. Together, these results indicate that GFP has only minor effects on PAR2-mediated signaling, on desensitization of signal transduction, or on the ability of ␤-arrestin1 to participate in desensitization. A, cells were exposed to graded concentrations of trypsin or AP (SLIGKV-NH 2 ), and the peak response was subtracted from the basal value to determine the change in the fluorescence ratio (n ϭ 3-6 observations). B and C, cells were exposed to carrier (control) or 10 nM trypsin for 2 min, washed, and then exposed to 10 nM trypsin 5 min after the first challenge. Representative results from 3 observations are shown. In the left panel, the blot was probed with the GFP antibody (1:20,000 for all lanes except KNRK-PAR2-GFP, where 1:1,000 was used). In the right panel (control) the blot was probed with the GFP antibody preabsorbed with 1-2 g/ml GFP-GST that was used for immunization. Note the specific detection of proteins of the predicted masses in transfected cell lines.

Trypsin-induced Trafficking of PAR2-GFP and ARR-GFP in Real Time
PAR2-GFP-We have previously examined trypsin-and APinduced trafficking of PAR2 in populations of cells fixed at various times after stimulation (23), but agonist-induced trafficking of PAR2 has not been examined in real time. We examined the effects of trypsin on the subcellular distribution of PAR2-GFP in individual KNRK-PAR2-GFP cells in real time. Before addition of trypsin, PAR2-GFP was detected at the plasma membrane and in a perinuclear compartment (Fig. 4). After 2-5 min incubation with 10 nM trypsin at 37°C, PAR2-GFP was observed at the plasma membrane and in a few superficial vesicles. After 10 min, PAR2-GFP was prominently detected in vesicles in a peripheral and a perinuclear location, and there was diminished localization at the plasma membrane. After 30 -60 min, PAR2-GFP was present in vesicles in a perinuclear location.
ARR-GFP-In a similar manner we examined the effects of trypsin on the subcellular distribution of ARR-GFP in individual KNRK-PAR2 ϩ ARR-GFP cells in real time. Before addition of trypsin, ARR-GFP was uniformly distributed in the cytosol and was not detected at the plasma membrane or in vesicles (Fig. 5). After 2-5 min incubation with 10 nM trypsin at 37°C, ARR-GFP was prominently detected at the plasma membrane and in a few superficial vesicles. After 10 min, ARR-GFP was present in vesicles in a peripheral and a perinuclear location and was no longer present at the plasma membrane. After 30 -60 min, ARR-GFP was detected in vesicles in a perinuclear location, and the intensity of the signal in the cytoplasm was diminished.
Thus, trypsin induces rapid redistribution of PAR2-GFP from the plasma membrane and into vesicles in a superficial and then a perinuclear location and stimulates translocation of ARR-GFP from the cytosol to the plasma membrane, followed by redistribution to endosomes. At 2-5 min, PAR2-GFP was detected in endosomes and at the plasma membrane, and ARR-GFP was usually present at the plasma membrane. Both proteins were clearly detected in endosomes at 10 min. If ␤-arres-tin1 mediates endocytosis of PAR2, we would expect membrane translocation of ␤-arrestin1 to precede PAR2 endocytosis and that ␤-arrestin1 would co-localize with PAR2 in vesicles. However, this possibility could not be investigated by observation of different cell populations in real time, when only a few cells were examined in any one experiment.

Simultaneous Localization of PAR2 and ARR-GFP
To examine the effect of trypsin on the subcellular distribution of PAR2 and ␤-arrestin1 in the same cells, we simultaneously localized these proteins in KNRK-PAR2 ϩ ARR-GFP cells by using an antibody to the C-terminal 12CA5 epitope and a Texas Red-labeled secondary antibody to detect PAR2, and the GFP signal to detect ␤-arrestin1. This analysis permitted evaluation of the relative localization of both proteins in multiple cells at precisely defined times. In unstimulated cells, PAR2 was present at the plasma membrane, and ARR-GFP was detected in the cytosol with minimal co-localization (Fig.  6). After 2 min incubation with 10 nM trypsin at 37°C, PAR2 remained at the plasma membrane, and ARR-GFP was detected at or close to the plasma membrane. After 5 min, PAR2 and ARR-GFP co-localized at the plasma membrane. After 10 -60 min, vesicles in a superficial and perinuclear location that contained PAR2 had the same size, shape, and location as those containing ARR-GFP, as indicated by superimposition of confocal images, where a yellow color denotes co-localization. Thus, trypsin induces translocation of ARR-GFP to the plasma membrane, where it co-localizes with PAR2. Notably, this translocation precedes endocytosis of PAR2. ARR-GFP and PAR2 internalize into the same vesicles in a superficial and then perinuclear location.
To define the duration of co-localization of PAR2 and ␤-ar-restin1 and to determine when they resumed their steady state distribution, we incubated KNRK-PAR2 ϩ ARR-GFP cells with 10 nM trypsin for 15 min at 37°C, washed cells, and incubated them in trypsin-free medium at 37°C. Cells were fixed 60 -240 min after initial exposure to trypsin, and PAR2 and ARR-GFP were localized as described. Within 60 -120 min of initial exposure to trypsin, PAR2 and ARR-GFP were mostly co-localized in vesicles clustered in a perinuclear region (not shown). Within 240 min, PAR2 was detected at the plasma membrane, and ARR-GFP was prominently detected in the cytosol (Fig. 7).
PAR2 and ARR-GFP were also detected in perinuclear region in some cells. Thus, PAR2 and ARR-GFP mainly resume their steady state distribution at the plasma membrane and in the cytosol, respectively, by 240 min after exposure to trypsin.
There are large pools of PAR2 in the Golgi apparatus (23,28). Disruption of these pools with brefeldin A inhibits resensitization of cellular responses to PAR2 agonists, suggesting that PAR2 is mobilized from the Golgi apparatus to the plasma membrane. To determine whether the recovery of the steady state distribution of PAR2 and ARR-GFP requires mobilization of Golgi pools of these proteins, we treated cells with brefeldin A. Within 240 min of initial exposure to trypsin in cells treated with brefeldin-A, PAR2 was detected in numerous vesicles throughout the cell with no detectable surface immunoreactivity (Fig. 7). ARR-GFP was detected in the cytosol and in vesi-FIG. 6. Localization of PAR2 and ARR-GFP in KNRK-PAR2 ؉ ARR-GFP cells. Cells were exposed to carrier (0 min, control) or 10 nM trypsin for 2-60 min at 37°C. PAR2 was localized with the HA.11 antibody and a Texas Red-conjugated secondary antibody (left panels). ␤-Arrestin1 was localized with GFP (center panels). Images in the right panels are formed by superimposition of images from the two other panels in the same row. In carrier-treated cells at 0 min, PAR2 was at the cell surface (arrow heads) and ARR-GFP was cytosolic (arrows), with no co-localization. After 2 min incubation with trypsin, PAR2 was still at the cell surface (white arrowheads) and ARR-GFP was also at the plasma membrane (yellow arrowheads) in close proximity to PAR2. After 5 min with trypsin, both PAR2 and ARR-GFP were co-localized at the plasma membrane (white arrowheads). After 10 -60 min with trypsin, PAR2 and ARR-GFP were co-localized in vesicles in the superficial, then in the perinuclear location (white arrows). Scale bar ϭ 10 m.
cles that were distinct from those containing PAR2. Thus, mobilization of Golgi stores of PAR2, but not ␤-arrestin1, is required for resumption of the steady state distribution.
Cycloheximide diminishes resensitization of cellular responses to repetitive stimulation with PAR2 agonists, suggesting that new receptor synthesis also contributes to resensitization when the Golgi stores are depleted (23). To determine whether the recovery of the steady state distribution of PAR2 and ARR-GFP requires new protein synthesis, we treated cells with cycloheximide. When cycloheximide-treated cells were exposed once to trypsin, PAR2 was still detected at the plasma membrane of many, but not all, cells within 240 min of initial exposure to trypsin (Fig. 7). In contrast, in all cycloheximidetreated cells, ARR-GFP resumed its steady state localization in the cytosol. To determine whether protein synthesis is required for recovery of surface PAR2 after depletion of intracellular stores, we challenged cells twice with 10 nM trypsin for two 15-min periods at an interval of 60 min. In control cells not treated with cycloheximide, PAR2 and ARR-GFP were detected at the cell surface and in the cytosol, respectively, within 240 min after the second trypsin challenge (not shown). In contrast, in cycloheximide-treated cells, PAR2 was not detected at the plasma membrane, and there was only weak staining of vesicles (Fig. 7). ARR-GFP was readily detected in the cytosol and in some vesicles, although the intensity of the signal was diminished. Thus, after repetitive stimulation with trypsin, which probably depletes cells of their intracellular stores of PAR2, the recovery of surface receptors mainly depends on new proteins synthesis.
The 12CA5 epitope is C-terminal and thus the HA.11 antibody detects both intact and cleaved PAR2. Therefore, the recovery of cell-surface PAR2 detected using the HA.11 anti-FIG. 7. Recovery of PAR2 and ARR-GFP distribution in KNRK-PAR2 ؉ ARR-GFP cells. Cells were treated with carrier (1st row control), brefeldin A (2nd row ), or cycloheximide (cyclo., 3rd and 4th rows). Cells in the 1st 3 rows were incubated once with 10 nM trypsin for 15 min at 37°C, washed, incubated in trypsin-free medium for 225 min and then fixed. Cells in the 4th row were incubated twice with 10 nM trypsin for 15 min at 37°C at a 60-min interval, washed, incubated in trypsin-free medium for 225 min after the second trypsin exposure, and then fixed. PAR2 was localized with the HA.11 antibody and a Texas Red-conjugated secondary antibody (left panels). ␤-Arrestin1 was localized with GFP (center panels). Images in the right panels are formed by superimposition of images from the two other panels in the same row. In controls (1st row), PAR2 was detected at the plasma membrane (arrowheads), and ARR-GFP was cytosolic (yellow arrows). PAR2 and ARR-GFP were also co-localized in a perinuclear region (white arrows). In brefeldin A-treated cells (row 2), PAR2 was detected in numerous vesicles (white arrows) and was not found at the plasma membrane. ARR-GFP was found in the cytosol and in distinct vesicles (yellow arrows), and there was no co-localization with PAR2. In cycloheximide-treated cells after one trypsin exposure (3rd row), PAR2 was often detected at the plasma membrane (arrowheads), and ARR-GFP was cytosolic (white arrows). In cycloheximide-treated cells after two trypsin exposures (4th row), PAR2 staining was weak, not detected at the cell surface, and confined to vesicles (white arrows), and ARR-GFP was cytosolic and in vesicles (yellow arrows). Scale bar ϭ 10 m. body could represent receptor mobilized from the Golgi apparatus or cleaved and recycled PAR2. To determine if PAR2 detected at the plasma membrane within 240 min of exposure to trypsin is intact, we stained cells with the M1 antibody to the N-terminal Flag epitope. Because the Flag is removed by trypsin, M1 detects intact but not cleaved PAR2. Flag immunoreactivity was readily detected at the plasma membrane within 240 min of exposure to trypsin, indicating the presence of intact PAR2 at the cell surface (Fig. 8). PAR2 was also detected in prominent perinuclear stores, which probably represent the Golgi apparatus. In cells treated with cycloheximide, there was also strong Flag immunoreactivity at the plasma membrane within 240 min of exposure to trypsin (Fig. 8), confirming our results with the HA.11 antibody after a single exposure to trypsin (Fig. 7). Notably, Flag immunoreactivity was not detected in prominent Golgi stores in cycloheximide-treated cells, presumably because they are no longer replenished by synthesis of new receptors. Thus, the recovery intact PAR2 at the plasma membrane after a single exposure to trypsin depends on mobilization of receptor from the Golgi apparatus. These stores are maintained by synthesis of new receptors.

Identification of Organelles Containing PAR2-GFP and ARR-GFP
Immunoreactive PAR2 is present in a prominent Golgi store in unstimulated cells, and trypsin and AP induce translocation of PAR2 from the cell surface to early endosomes and then lysosomes (23,28). To determine if PAR2 agonists cause a similar redistribution of PAR2-GFP, and to identify organelles we stained cells with antibodies to mannosidase II (Golgi apparatus), the transferrin receptor (endosomes), and LAMP-1 (lysosomes).
In unstimulated KNRK-PAR2-GFP cells, PAR2-GFP was detected at the plasma membrane, and in a perinuclear compartment that was also stained with an antibody to mannosidase II and is thus the Golgi apparatus (Fig. 9). In unstimulated KNKR-PAR2 ϩ ARR-GFP cells, ARR-GFP was uniformly distributed throughout the cytosol with no prominent stores that resemble the Golgi apparatus (not shown).
To identify vesicles containing PAR2-GFP and ARR-GFP, KNRK-PAR2-GFP cells and KNRK-PAR2 ϩ ARR-GFP cells were incubated with 10 nM trypsin for 15 min at 37°C, washed, incubated in trypsin-free medium, and fixed 30 -240 min after exposure to trypsin. Within 30 min after initial exposure to trypsin, PAR2-GFP and ARR-GFP were detected in vesicles in a superficial and perinuclear location that also stained with an antibody to the transferrin receptor (Fig. 9). Thus, both PAR2-GFP and ARR-GFP redistribute to early endosomes. In experiments to determine whether PAR2-GFP and ARR-GFP were present in lysosomes, cells were treated with 10 mM NH 4 Cl, to alkalinize lysosomes and thereby diminish protein degradation, which would quench the signal (37). Within 60 -120 min of initial exposure to trypsin, PAR2-GFP was prominently localized to large vesicles some of which stained with an antibody to LAMP-1 (not shown). After 240 min, PAR2 was present at the plasma membrane and also detected within vesicles that stained for LAMP-1 and are thus lysosomes (Fig. 9). ARR-GFP was present in the cytosol throughout the cell and was also concentrated in the vicinity of lysosomes. However, we did not detect ARR-GFP within lysosomes. Thus, whereas PAR2 is sorted to lysosomes, ␤-arrestin1 may be sorted from PAR2 in a pre-lysosomal compartment.

Role of ␤-Arrestins in PAR2 Endocytosis
For certain GPCRs, such as the ␤ 2 AR, ␤-arrestins serve as adaptor proteins for clathrin-mediated endocytosis (15,16). However, ␤-arrestins are not required for endocytosis of the angiotensin II type 1A receptor (42) or the m1, m3, and m4 muscarinic cholinergic receptors (43). To determine whether ␤-arrestins are required for agonist-induced endocytosis of PAR2, we stably expressed PAR2 and ␤-arrestin 319 -418 (34). This fragment constitutively binds to clathrin but is unable to interact with phosphorylated GPCRs. Thus, ␤-arrestin 319 -418 acts as a dominant negative mutant that inhibits agoniststimulated endocytosis of the ␤ 2 AR (34). We also examined whether overexpression of ␤-arrestin1 accelerated endocytosis of PAR2. KNRK cells expressing PAR2 alone (control), PAR2 ϩ ARR, or PAR2 ϩ ARR 319 -418 were incubated with 500 M AP for 0 -30 min at 37°C. To detect PAR2 at the cell surface, non-permeabilized cells were incubated with Flag M1 antibody (against the extracellular N-terminal Flag epitope) and a phycoerythrin labeled secondary antibody and were analyzed by flow cytometry. In KNRK-PAR2 cells, AP caused a rapid reduction in surface Flag immunoreactivity (Fig. 10). After 5 min incubation with AP, surface Flag immunoreactivity was 54.2 Ϯ 3.2% (mean Ϯ S.E., n ϭ 3 experiments) of that in untreated cells. After 30 min, surface Flag immunoreactivity was 41.0 Ϯ 1.5% of untreated cells. Expression of ARR-GFP did not accelerate endocytosis. In KNRK-PAR2 ϩ ARR-GFP cells, after 5 min surface Flag immunoreactivity was 53.8 Ϯ 7.2% and after 30 min surface Flag immunoreactivity was 42.6 Ϯ 10% of untreated cells. In KNRK-PAR2 ϩ ARR 319 -418 -GFP cells, the effect of AP on surface immunoreactivity was markedly diminished at all time points. After 5 min incubation with AP, surface Flag immunoreactivity was 89.7 Ϯ 2.6% that in untreated cells (p Ͻ 0.05 compared with KNRK-PAR2 and KNRK-PAR2 ϩ ARR-GFP cells, analysis of variance, Student-Newman-Kuels test). After 30 min, surface Flag immunoreactivity was 56.0 Ϯ 3.7% of untreated cells. Thus, although overexpression of ␤-ar-restin1 does not accelerate endocytosis of PAR2, expression of the dominant negative mutant of ␤-arrestin markedly reduces the initial rate of agonist-induced endocytosis of PAR2, which indicates that ␤-arrestins mediate endocytosis of this receptor. Cells were treated with carrier (control), or cycloheximide (cyclo.). Cells were incubated once with 10 nM trypsin for 15 min at 37°C, washed, incubated in trypsin-free medium for 225 min, and then fixed. PAR2 was localized with the Flag M1 antibody, which detects only intact and not trypsin-cleaved receptor, and a Texas Red-conjugated secondary antibody. In controls, PAR2 was detected at the cell surface (arrowheads) and in Golgi stores (arrows). In cycloheximidetreated cells, PAR2 was detected only at the cell surface (arrowheads). Scale bar ϭ 10 m.

The Role of Intracellular Ca 2ϩ in Trypsin-induced
Trafficking of PAR2-GFP and ARR-GFP PAR2 couples to phospholipase C␤ and generation of inositol trisphosphate, which mobilize Ca 2ϩ from intracellular stores (23,29). Ca 2ϩ may be required for the activity of proteins involved in PAR2 and ␤-arrestin trafficking. To determine whether an increase in [Ca 2ϩ ] i is required for trypsin-induced trafficking of PAR2 and ␤-arrestin1, we treated cells with BAPTA/AM, a chelator of intracellular Ca 2ϩ . In untreated cells, exposure to trypsin for 2-30 min caused endocytosis of PAR2 and translocation to perinuclear endosomes and stimulated redistribution of ARR-GFP from the cytosol and to the plasma membrane and endosomes (Fig. 4 -6 and 11). In cells treated with BAPTA/AM, trypsin-mediated trafficking of PAR2-GFP and ARR-GFP was delayed at all times studied (Fig. 11). Even after 30 min, PAR2-GFP was detected at the plasma membrane and in superficial vesicles of BAPTA/AMtreated cells, rather than in perinuclear endosomes. Trypsin stimulated a slow translocation of ARR-GFP from the cytosol to the plasma membrane of BAPTA/AM-treated cells (not shown). After 30 min, ARR-GFP was retained at the plasma membrane and in superficial vesicles of BAPTA/AM-treated cells and was  (1st 3 rows) or KNRK-PAR2 ؉ ARR-GFP cells (4th and 5th rows), respectively. Cells were untreated (0 min, control), or incubated with 10 nM trypsin for 15 min, washed, incubated in trypsin-free medium, and fixed 30 min or 240 min after initial exposure to trypsin. Mannosidase II (MII), transferrin receptor (Tf-R), and LAMP-1 were localized by immunofluorescence using a Texas Red-conjugated secondary antibody (center panels). Images in the right panels are formed by superimposition of images from the two other panels in the same row. In unstimulated cells (1st row), PAR2-GFP was detected at the plasma membrane (arrowheads) and in a perinuclear pool that co-localized with the Golgi marker mannosidase II (arrows). Within 30 min after initial exposure to trypsin, PAR2-GFP (2nd row) and ARR-GFP (4th row) were detected in superficial and perinuclear vesicles that co-localized with the transferrin receptor and are thus early endosomes (arrows). Within 240 min, PAR2-GFP (3rd row) was detected within vesicles stained with LAMP-1 and which are thus lysosomes (arrows). ARR-GFP (5th row) was detected in the cytosol (white arrows) and in the vicinity of lysosomes (yellow arrows). Scale bar ϭ 10 m. not detected in perinuclear endosomes. These results indicate that intracellular Ca 2ϩ is required for trypsin-induced trafficking of PAR2 and ␤-arrestin1.

DISCUSSION
Expression of fully functional PAR2 and ␤-arrestin1 tagged with GFP permitted detailed analysis of the consequences of receptor activation to the subcellular localization of these proteins in real time. Irreversible activation of PAR2 by tryptic cleavage caused a rapid (Ͻ2 min) translocation of ␤-arrestin1 from the cytosol to the plasma membrane, which preceded and was required for PAR2 endocytosis. ␤-Arrestin1 and PAR2 rapidly (Ͻ10 min) translocated into the same endosomes. ␤-Ar-restin1 and PAR2 remained co-localized in vesicles in a perinuclear region for prolonged periods (up to 2 h). Whereas PAR2 was sorted to lysosomes, ␤-arrestin1 returned to the cytosol. Resumption of the steady state distribution of PAR2 at the plasma membrane required mobilization of PAR2 from the Golgi apparatus and synthesis of new receptor. To our knowledge, this is the first demonstration of protease-induced trafficking of PAR2 and ␤-arrestins in real time and the first report that ␤-arrestins mediate endocytosis of PARs.
Expression of Functional PAR2-GFP and ARR-GFP-A prerequisite to using GFP is to establish that it does not interfere with normal functioning of the proteins of interest. Trypsin stimulated redistribution of PAR2-GFP from the plasma membrane to endosomes and lysosomes and induced translocation of ARR-GFP from the cytosol to the plasma membrane and endosomes. We have previously shown that trypsin causes a similar redistribution of immunoreactive PAR2 in transfected KNRK cells (23), and agonists of the neurokinin-1 receptor stimulate translocation of immunoreactive ␤-arrestins from the cytosol to the plasma membrane and endosomes in KNRK cells and neurons (35). 2 Therefore, GFP does not affect trafficking of PAR2 and ␤-arrestins. Trypsin and AP stimulated Ca 2ϩ mobilization in all cell lines with approximately similar efficacies and potencies. The concentration response curves for trypsin-and AP-stimulated Ca 2ϩ mobilization were shifted to the right for cells expressing PAR2-GFP when compared with cells expressing ARR-GFP or ARR 319 -418 -GFP (this study) or our previous observations of cells expressing wild type PAR2 (28). This reduced potency may be due to differences in the level of PAR2 expression between cell lines or could indicate that GFP has a small effect on PAR2 signaling. The rate of PAR2 endocytosis was similar in cells expressing ARR-GFP and endogenous ␤-arrestins, suggesting that GFP does not affect the capacity of ␤-arrestins to mediate endocytosis. Repeated exposure to trypsin and AP strongly desensitized Ca 2ϩ mobilization in all cell lines, although desensitization was less in cells expressing ARR-GFP, suggesting that GFP could have a small effect on the ability of ␤-arrestins to mediate uncoupling. The duration of the Ca 2ϩ response to trypsin was considerably longer in cells expressing ARR 319 -418 -GFP, an unexpected observation since ARR 319 -418 should not interact with receptors and would not be expected to affect desensitization. Indeed, Ca 2ϩ mobilization desensitized normally to repeated exposure to trypsin and AP in cells expressing ARR 319 -418 -GFP. Together, our results indicate that GFP has minor effects of the behavior of PAR2 or ␤-arrestins. In support of our results, GFP has been placed at the C terminus of the cholecystokinin A, ␤ 2 -adrenergic, and neurokinin-1 receptors, and of ␤-arrestin2 without affecting trafficking and function (38 -41). 2 Agonist-induced Trafficking of PAR2 and ␤-Arrestin1-The availability of cells expressing functional PAR2-GFP and ARR-GFP permitted examination of trypsin-induced trafficking in real time. By using cells that co-expressed PAR2 and ARR-GFP, we were able to localize simultaneously both proteins at precise times after receptor activation. The earliest detectable alteration was translocation of ␤-arrestin1 from the cytosol to the plasma membrane, which preceded endocytosis of PAR2 ( Fig. 6 and Fig. 12). Agonists of other GPCRs also induce translocation of ␤-arrestins from the cytosol to the plasma membrane in cell lines and in neurons (16,35,41). The mechanism of membrane translocation of ␤-arrestins is unknown. GRK2/3 interact with free ␤␥ subunits of G-proteins and thereby translocate to the plasma membrane (44,45) where they may phosphorylate PAR2. Once phosphorylated, the affinity with which GPCRs interact with ␤-arrestins is markedly enhanced (46), which may promote redistribution of ␤-arrestins to the plasma membrane.
Membrane translocation of ␤-arrestin1 was followed by endocytosis of ␤-arrestin1 and PAR2 into the same endosomes containing the transferrin receptor (Fig. 12). In support of our results, ␤-arrestins co-localize with the ␤ 2 AR and with the neurokinin-1 receptor in endosomes of transfected cells and neurons after exposure to agonists (16,35). 2 Although we do not know if PAR2 and ␤-arrestin1 are physically associated in endosomes, these proteins remained co-localized for up to 2 h. PAR2 and ␤-arrestin1 were sorted in a perinuclear region into distinct pathways. PAR2 was detected within lysosomes, especially at later times after stimulation with trypsin. ␤-Arrestin1 resumed its distribution in the cytosol and was concentrated in the cytosol in close proximity to lysosomes. We did not detect ␤-arrestin1 within lysosomes, but we cannot exclude the possibility that some ␤-arrestin1 translocates to lysosomes with PAR2 where it may be degraded and difficult to detect. The mechanism by which PAR2 and ␤-arrestin1 are sorted into

FIG. 11. Trypsin-induced trafficking of PAR2-GFP (upper panels) and ARR-GFP (lower panels) in control cells (left) and cells treated with BAPTA/AM (right).
Cells were incubated with 10 nM trypsin for 30 min at 37°C, fixed, and observed. In control cells, PAR2-GFP and ARR-GFP were principally detected in perinuclear endosomes (arrows). In BAPTA/AM-treated cells, PAR2-GFP and ARR-GFP were detected at the plasma membrane (arrowheads) and in superficial vesicles (arrows) but were not detected in a perinuclear location. Scale bar ϭ 10 m. different pathways remains to be determined. Dephosphorylation of GPCRs decreases their affinity for ␤-arrestins (46), and endosomes are enriched in GPCR phosphatases (47). Thus, endosomal phosphatases may dephosphorylate PAR2 resulting in dissociation of ␤-arrestin1.
Within 4 h of trypsin treatment, PAR2 recovered its steady state distribution at the plasma membrane. By using an antibody to a Flag epitope placed proximal to the cleavage site, we demonstrated that intact, uncleaved PAR2 is found at the cell surface at this time. Recovery of the steady state distribution of PAR2 was prevented by brefeldin A, suggesting that recovery requires mobilization of presynthesized PAR2 from the Golgi apparatus. There was no visible depletion of the Golgi pools of intact PAR2 4 h after trypsin, unless cells were treated with cycloheximide. Thus, newly synthesized receptors replenish the Golgi stores. We have previously shown that brefeldin A prevents resensitization of trypsin-induced Ca 2ϩ mobilization, further supporting the importance of the Golgi stores of PAR2 (23). Cycloheximide did not prevent recovery of surface PAR2 after a single exposure to trypsin. However, when cells were repetitively challenged with trypsin, cycloheximide prevented recovery of the steady state distribution of PAR2 at the plasma membrane. In support of this observation, cycloheximide diminished resensitization of Ca 2ϩ mobilization after repetitive stimulation by trypsin (23). Together, these results suggest that synthesis of new receptors is required once the Golgi pools are exhausted. The detection of PAR2 in lysosomes, and the finding that cycloheximide inhibits recovery after repeated activation, suggests that most PAR-2 is degraded after endocytosis. However, proof of degradation will require further analysis. It is possible that some activated PAR2 also recycles, as is the case for PAR1 (24,25) and many receptors for neurotransmitters (9,37). However, in contrast to neurotransmitter receptors, which are activated in a reversible manner, cleaved PARs can no longer be activated by a protease. In contrast to PAR2, there were no prominent stores of ␤-arrestin1 in the Golgi apparatus, and neither brefeldin A nor cycloheximide prevented the redistribution of ␤-arrestin1 to the cytosol. Therefore, ␤-arrestin1 redistributes from endosomes to the cytosol to resume its steady state distribution.
The Role of ␤-Arrestins in Endocytosis of PAR2-Translocation of ␤-arrestin1 to the plasma membrane preceded endocytosis of PAR2, and expression of the C-terminal clathrin binding domain of ␤-arrestin (48) inhibited endocytosis of PAR2, as determined by AP-induced loss of Flag immunoreactivity from the plasma membrane. Together, these results indicate that ␤-arrestins participate in agonist-induced endocytosis of PAR2.
It is likely that PAR2 internalizes by a clathrin-dependent mechanism in KNRK cells since ␤-arrestins serve as adaptor proteins for clathrin (16). Although the role of clathrin in trypsin-induced endocytosis of PAR2 in KNRK cells has not been examined, other receptors internalize at sites of clathrin-coated pits in these cells (37). Thrombin also induces endocytosis of PAR1 at sites of coated pits in human megakaryoblastic erythroleukemia cells and CHRF-288 cells (24).
␤-Arrestins participate in endocytosis of other GPCRs, including the ␤ 2 AR, m2 muscarinic acetylcholine receptor, and the neurokinin-1 receptor (15,34,49). 2 Expression of GRK2 promotes endocytosis of ␤ 2 AR-Y326A, which shows defective agonist-induced phosphorylation and endocytosis (15). GRKs probably enhance endocytosis of GPCRs by promoting phosphorylation, which permits ␤-arrestins to couple the receptors to clathrin. The ability of ␤-arrestins to mediate endocytosis also depends upon their phosphorylation state. Cytosolic ␤-arres-tin1 is constitutively phosphorylated at a C-terminal Ser 412 residue, and rapid dephosphorylation at the plasma membrane is required for clathrin binding and receptor endocytosis (50).
Overexpression of ␤-arrestin did not increase the rate of PAR2 endocytosis, presumably because endocytosis is already so rapid in these cells, and the availability of endogenous ␤-arrestins is not rate-limiting. Indeed, endogenous ␤-arres-tin1 is readily detected in KNRK cells by immunofluorescence and Western blotting. 2 Expression of ␤-arrestin 319 -418 significantly reduced the initial rate of endocytosis of PAR2. We do not know the reason for the inability of ␤-arrestin 319 -418 to abolish endocytosis. One possibility is that ␤-arrestin 319 -418 was not expressed at sufficiently high levels to inhibit completely the capacity of wild type ␤-arrestin to couple to clathrin and thereby inhibit PAR-2 endocytosis. Alternatively, PAR2 may undergo endocytosis by a ␤-arrestin-independent pathway when clathrin-mediated endocytosis is inhibited. Agonist-induced endocytosis of the angiotensin II type 1A receptor and the m1, m3, and m4 muscarinic cholinergic receptors do not depend on ␤-arrestins (42,43). Thus, the molecular mechanism of endocytosis of GPCRs may depend on the receptor and also the cellular environment. The nature of the ␤-arrestin-independent pathway for GPCR endocytosis is unknown. Some GPCRs, exemplified by the endothelin 1 and cholecystokinin A receptors, internalize in part by caveolin-dependent pathways (51, 52), but we do not know the role of caveolin in PAR2 endocytosis.
The Role of Intracellular Ca 2ϩ in Agonist-induced Trafficking of PAR2 and ␤-Arrestin1-PAR2 agonists mobilize Ca 2ϩ from intracellular stores (23). We observed that chelation of intracellular Ca 2ϩ using BAPTA/AM delayed, but did not abol- FIG. 12. Hypothesized pathway of agonist-induced trafficking of PAR2 and ␤-arrestin1 in KNRK cells. 1, trypsin cleaves PAR2 and irreversibly activates the receptor. 2, GRKs and ␤-arres-tin1 translocate to the plasma membrane. 3, GRKs phosphorylate PAR2; ␤-arrestin1 interacts with the phosphorylated receptor to uncouple the receptor from G-proteins and terminate signal transduction. 4, ␤-arrestins also serve as adaptor proteins for clathrin-mediated endocytosis of PAR2. 5, PAR2 and ␤-arrestin1 co-localize in endosomes. PAR2 is sorted to lysosomes, and ␤-arrestin1 is sorted to the cytosol. 6 and 7, resensitization requires mobilization of PAR2 from prominent stores in the Golgi apparatus and eventually synthesis of new receptors. Support for this model derives from the present investigation and Ref. 23. ish, trypsin-induced trafficking of PAR2 and ␤-arrestin1. Notably, both PAR2 and ␤-arrestin1 were retained at the plasma membrane and in superficial vesicles of cells treated with BAPTA/AM and were not prominently detected in perinuclear endosomes. Thus, Ca 2ϩ may be required for the normal function of proteins that mediate endocytosis and trafficking of PAR2 and ␤-arrestin1. One candidate is protein kinase C of the C family, which depends on Ca 2ϩ for activity, since protein kinase C appears to be required for endocytosis of other GPCRs (9). Further experimentation will be required to identify such Ca 2ϩ -dependent proteins.
Functional Implications of Agonist-induced Trafficking of PAR2 and ␤-Arrestin1-Agonist-induced redistribution of PAR2 and ␤-arrestin1 alters cellular responsiveness and may contribute to signal transduction. ␤-Arrestins probably interact with GRK-phosphorylated PAR2 at the plasma membrane to uncouple the receptor from G-proteins and terminate signaling. Trypsin causes phosphorylation of PAR2, 2 but the contribution of GRKs and second messenger kinases is unknown. It is known that GRKs participate in desensitization of PAR1 because overexpression of GRK3 inhibits signaling by thrombin (22). Although expression of ␤-arrestin 319 -418 did not prevent desensitization of PAR2, this construct is unable to interact with receptors and would not be expected to affect signaling (34).
Endocytosis of PAR2 depletes the plasma membrane of activated receptors and may thereby separate them from heterotrimeric G-proteins. Endocytosis may also be required for signal transduction. Endocytosis of the ␤ 2 AR is required for stimulation of mitogen-activated protein kinases (21). Since PAR2 also couples to this pathway (53), endocytosis may contribute to signaling by this receptor. Lysosomal degradation of PAR2 prevents recycling and will inevitably terminate signaling. Indeed, lysosomal trafficking of PAR1 prevents persistent signaling (54). Mobilization of PAR2 from the Golgi apparatus and synthesis of new receptors are necessary for recovery of PAR2 at the plasma membrane and for resensitization of responses to trypsin (23). Resensitization of responses to thrombin is also associated with mobilization of intracellular stores of PAR1 and requires synthesis of new receptors (24 -26). Colocalization of PAR2 and ␤-arrestin1 in endosomes persists until PAR2 is directed to lysosomes and ␤-arrestins return to the cytosol, suggesting that ␤-arrestin has other regulatory roles, which remain to be determined.