β-Arrestin-dependent Regulation of the Cofilin Pathway Downstream of Protease-activated Receptor-2*

β-Arrestins are pleiotropic molecules that mediate signal desensitization, G-protein-independent signaling, scaffolding of signaling molecules, and chemotaxis. Protease-activated receptor-2 (PAR-2), a Gαq/11-coupled receptor, which has been proposed as a therapeutic target for inflammation and cancer, requires the scaffolding function of β-arrestins for chemotaxis. We hypothesized that PAR-2 can trigger specific responses by differential activation of two pathways, one through classic Gαq/Ca2+ signaling and one through β-arrestins, and we proposed that the latter involves scaffolding of proteins involved in cell migration and actin assembly. Here we demonstrate the following. (a) PAR-2 promotes β-arrestin-dependent dephosphorylation and activation of the actin filament-severing protein (cofilin) independently of Gαq/Ca2+ signaling. (b) PAR-2-evoked cofilin dephosphorylation requires both the activity of a recently identified cofilin-specific phosphatase (chronophin) and inhibition of LIM kinase (LIMK) activity. (c) β-Arrestins can interact with cofilin, LIMK, and chronophin and colocalize with them in membrane protrusions, suggesting that β-arrestins may spatially regulate their activities. These findings identify cofilin as a novel target of β-arrestin-dependent scaffolding and suggest that many PAR-2-induced processes may be independent of Gαq/11 protein coupling.

␤-Arrestins were originally identified as terminators of G-protein-coupled receptor signaling and later as signaling scaffolds that regulate activation, subcellular localization, and signal duration of mitogen-activated protein kinases (MAPKs). Over the past 5 years, siRNA 2 studies silencing individual ␤-arrestins have revealed both redundant and selective roles for either ␤-arrestin-1 or -2 in the regulation of additional signaling molecules such as phosphatidylinositol 3-kinase and RhoA (1)(2)(3)(4)(5)(6)(7)(8) and for cellular processes such as chemotaxis, cytoskeletal reorganization, and epithelial permeability (9 -15). Until recently, it was assumed that these ␤-arrestin scaffolds formed only after initial heterotrimeric G-protein activation, but studies on several G-protein-coupled receptors have now demonstrated that ␤-arrestins can mediate signals independently of G-protein coupling (11,16,17), highlighting the importance of identifying the signaling cascades they regulate.
Protease-activated receptor-2 (PAR-2), which requires ␤-arrestins for desensitization, localization of ERK1/2, and cell migration, was one of the first G-protein-coupled receptors shown to utilize ␤-arrestin-dependent scaffolding (9,10). PAR-2 is reported to trigger a wide variety of cellular responses (e.g. proliferation, chemotaxis, ion transport, epithelial barrier function, and tumor cell metastasis) in a cell type-specific manner (10, 18 -25). It is activated by proteolytic cleavage of its N terminus, which unveils a tethered ligand (SLIGRL/SLIGKV, human/mouse) leading to G␣ q coupling, hydrolysis of phosphatidylinositol 4,5-bisphosphate, and mobilization of intracellular Ca 2ϩ (25,26). Most PAR-2 actions are thought to lie downstream of Ca 2ϩ mobilization, although for some events, ␤-arrestin-dependent signaling can oppose the G␣ q /Ca 2ϩ pathway (27). To date, the molecular mechanisms by which PAR-2 mediates processes such as cytoskeletal reorganization and chemotaxis have not been elucidated. Because cell migration requires the cell to extend pseudopodia at the leading edge while inhibiting extensions and dissolving stress fibers at the trailing edge (28 -30) and ␤-arrestins are capable of spatially restricting activity of signaling molecules downstream of PAR-2 activation, we hypothesized that ␤-arrestins might contribute to the spatial and temporal control over actin assembly/ disassembly through scaffolding of signaling molecules traditionally associated with chemotaxis. We focused on the actin filament-severing protein cofilin, the activity of which allows for rapid changes in cytoskeletal architecture, creation of new barbed ends for elongation, and recycling of ATP monomers (28, 30 -33). Cofilin activation is controlled by opposing actions of LIMKs (which inactivate it by phosphorylation on Ser 3 ) and cofilin-specific phosphatases (chronophin (CIN) and slingshot) that activate it (33)(34)(35)(36) as well as by intracellular pH and phosphatidylinositol 4,5-bisphosphate levels (37). In the studies described here, we investigated the following: 1) whether PAR-2 promotes cofilin dephosphorylation and activation through a ␤-arrestin-dependent mechanism, 2) whether ␤-arrestins can scaffold and direct subcellular localization of cofilin, 3) whether ␤-arrestin-dependent cofilin activation is independ-ent of classic G␣ q signaling, and 4) whether upstream cofilin regulators CIN and LIMK are regulated by ␤-arrestins in response to PAR-2 activation.

Transfections and Cell Lines
MEFs from wild type (MEFwt) and ␤-arrestin-1/2 Ϫ/Ϫ knockout mice (MEF␤arrDKO) were gifts from Dr. Robert Lefkowitz (Duke University Medical Center) and have been described previously (38). MDA-MB-468 cells were purchased from ATCC and grown in Dulbecco's modified Eagle's medium and 10% fetal calf serum. FLAG-and GFP-tagged ␤-arrestin-1 and -2 were a gift from Dr. Robert Lefkowitz (Duke University Medical Center) and have been described previously (39,40). Dominant negative CIN and Myc-tagged CIN constructs have been described previously (36). Transient transfections were performed on 70 -80% confluent cells using Lipofectamine (Invitrogen), and experiments were performed between 24 and 48 h of transfection.

Protein Analysis and Western Blotting
5 ϫ 10 5 cells/60-mm dish (grown for 24 h) were serumstarved for 2 h, treated with 100 nM 2fAP for 0 -90 min at 37°C, lysed in 0.25 ml of lysis buffer (phosphate-buffered saline, pH 7.6, 1% Triton X-100, 1 mM EGTA, 2 mM NaF, 100 M sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, and 1 g/ml each aprotinin, leupeptin, and benzamidine). 25 g of cleared lysate protein was analyzed by SDS-PAGE (15% for cofilin, 10% for LIMK, and 12.5% for ␤-arrestin) followed by Western blotting. Blots were imaged using the LICOR Odyssey imaging system, and LICOR software was used to calculate integrated intensities of bands. Images of Western blots were assembled using Adobe Photoshop 5.0 and imported into Canvas. Some gels were spliced to eliminate blank lanes or lanes containing samples unrelated to the figure.
LIMK/␤-Arrestin Interactions-FLAG-␤-arrestin-1 and -2 were overexpressed in NIH3T3 cells and immunoprecipitated with anti-FLAG (M2)-agarose. Washed beads from transfected or untransfected cells were mixed with recombinant His-tagged LIMK (amino acids 285-639) for 1 h at room temperature, pelleted, washed, and analyzed by SDS-PAGE followed by Western blotting with anti-FLAG and anti-His 6 . Alternatively GST-␤-arrestin-1, GST-␤-arrestin-2, or GST alone were expressed Escherichia coli (strain BL21DE3), protein expression was induced with 0.2 mM isopropyl 1-thio-␤-D-galactopyranoside for 4 h, and GSTtagged proteins were purified with glutathione-Sepharose to Ͼ85% purity. Some breakdown of both GST-␤-arrestin proteins was observed, accounting for the appearance of 29-and 42-45-kDa bands visible by Coomassie. 5 g each of each GST fusion protein was incubated with 1 or 5 g of Histagged LIMK (amino acids 285-639) for 1 h, LIMK-bound proteins were pulled down with cobalt resin, and both bound and unbound proteins were analyzed by SDS-PAGE followed by staining with Coomassie.

Pyrene-Actin Filament Severing Assay
Actin polymerization assays were performed using an Actin Polymerization Biochem kit (Cytoskeleton Inc.). It has been demonstrated by others that addition of depolymerizing factors such as cytochalasin or actin-depolymerizing factor/cofilin to pyrene-actin filaments at steady state will cause a decrease in fluorescence reflective of depolymerization/ severing activity (44 -47). Although cell extracts will likely contain a myriad of other factors that can affect the actin assembly kinetics, when they are added to filaments that have reached steady state, a decrease in fluorescence reflects the presence of a depolymerizing activity. Cell lysates were prepared from MDA-MB-468 cells and MEFs in lysis buffer (20 mM Hepes, pH 7.0, 4 mM MgCl 2 , 1 mM ATP, 100 mM KCl, 9 mM EGTA, and protease inhibitors). Pyrene-G-actin (10% labeled) was incubated in polymerization buffer (containing 10 mM imidazole, pH 7.5, 2 mM MgCl 2 , 50 mM KCl, and 1 mM ATP) for 2 h at room temperature, the resulting pyrene-Factin was diluted to 2 M (final concentration) in Buffer G (5 mM Tris-HCl, pH 8.0, 0.2 mM CaCl 2 , 0.2 mM ATP, and 0.5 mM dithiothreitol), and fluorescence was monitored for 3000 s. Cell extracts (40 g of protein/10 l of lysis buffer) or 10 l of lysis buffer alone were added after 1000 s. Severing was monitored by measuring the decrease in fluorescence at 386 nm on a Spex Fluoromax-2 fluorometer. Cofilin has been shown to generate free barbed ends available for elongation (31); thus if the fluorescence decrease observed upon addition of PAR-2-stimulated extracts was due to filament sev-ering activity, addition of G-actin should result in increased polymerization. To confirm that a decrease in fluorescence was accompanied by an increase in free barbed ends, at t ϭ 2200 s, pyrene-G-actin (2 M) was added back to reactions. This also rules out the possibility that the decrease in pyrene fluorescence was due to quenching. Data were collected using Data Max software 2.2 and analyzed using Kaleidagraph Version 4.0.

Assay of LIMK Activity
Cleared cell lysates (LIMK lysis buffer: Hepes, pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 1 mM MgCl 2 , 1 mM MnCl 2 , and protease inhibitors) were incubated with anti-LIMK for 8 h at 4°C. Phosphatase inhibitors were added (1 mM NaF and NaVO 4 ) to cleared lysates just before addition of antibody. Washed beads were incubated in LIMK lysis buffer supplemented with 5 M cold ATP, 5 Ci of [␥-32 P]ATP, and 2 g of recombinant GST-cofilin at 30°C for 45 min. Reactions were terminated by transfer of the reaction mixture to 2ϫ Laemmli sample buffer followed by SDS-PAGE, Coomassie staining, and autoradiography. GST-cofilin bands were excised, and incorporated radiolabel was determined by scintillation counting.

Ca 2؉ Mobilization Assays
Cells were incubated in physiological salt solution (PSS: 137 mM NaCl, 4.7 mM KCl, 0.56 mM MgCl 2 , 2 mM CaCl 2 , 1.0 mM Na 2 HPO 4 , 10 mM Hepes, 2.0 mM L-glutamine, and 5.5 mM D-glucose, pH 7.4) containing 0.1% bovine serum albumin and 5 M Fura-2/AM for 20 min at 37°C. Cells were washed and mounted in a microincubator containing 1 ml of PSS-bovine serum albumin at 37°C on the stage of a Nikon inverted microscope (40ϫ objective). Agonists were directly added to the bath. The Ca 2ϩ ionophore, ionomycin, was added at the end of all experiments to determine maximum [Ca 2ϩ ]. Fluorescence was detected in individual cells using a Nikon charge-coupled device camera and a video microscopy acquisition program (Metafluor). Fluorescence was measured 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 ] i , was determined, and intracellular [Ca 2ϩ ] was calculated using the Grynkiewicz equation (48). Average [Ca 2ϩ ] i values for a range of 10 -12 cells were calculated and presented as a single trace.

␤-Arrestins and Cofilin Regulation
Microscopy Cells were seeded onto collagen-coated coverslips and allowed to attach overnight. After agonist treatment, cells were fixed in normal buffered formalin and prepared as described previously (9). Serial sections (1 m; 100ϫ and 63ϫ objectives) were taken on a Zeiss LSM510 at 1ϫ and 4ϫ zoom. Overlays of four individual Z-sections and enlarged images were obtained using Adobe Photoshop 5.0.

Data and Statistical Analysis
All graphs and statistical analyses were performed using Kaleidagraph Version 4.0 or Microsoft Excel 2003. All experiments were performed a minimum of three times. Phosphoprotein levels were normalized to total protein levels before calculating -fold change with respect to untreated controls. For LIMK assays, cpm values were normalized to total LIMK immunoprecipitated and total GST-cofilin levels before calculating -fold change with respect to controls. Analysis of variance and Tukey t tests were used to determine statistical significance of and significant differences between values under different conditions.

PAR-2 Promotes Cofilin Dephosphorylation and Filament Severing
Activity-Cofilin can be activated by dephosphorylation on Ser 3 ; thus, we investigated whether cofilin phosphorylation was altered in response to the PAR-2 activation using the specific PAR-2 activating peptide 2fAP (49,50), inactive reverse 2fAP, or trypsin and probing cell lysates with a phospho (Ser 3 )specific cofilin antibody. In the human breast cancer cell line (MDA-MB-468), we observed a robust decrease in phosphorylation as early as 30 s after PAR-2 activation ( Fig. 1, A-C). This effect was dose-dependent ( Fig. 1B) and was observed in another breast cancer cell line, MDA-MB-231 (not shown). Maximum cofilin dephosphorylation (66% for 2fAP and 75% for trypsin) was observed at 5 min; by 90 min, some of the cofilin remained dephosphorylated (53 Ϯ 5% for 2fAP and 48 Ϯ 16% for trypsin) (Fig.  1, A and C). No dephosphorylation was observed in response to reverse 2fAP (Fig. 1A). Dephosphorylation was observed even when agonist was removed at each time point (supplemental Fig. 1), suggesting that the prolonged response was due to maintenance of dephosphorylated cofilin by downstream mechanisms rather than to prolonged agonist exposure. Because peak dephosphorylation was observed at 5 D, pyrene-labeled F-actin was prepared as described under "Materials and Methods," and fluorescence was monitored for 3000 s. At t ϭ 1200 s, either lysis buffer alone (ϫ) or MDA-MB-468 cell extracts prepared under the following conditions were added: IgG (Ⅺ and f) or cofilin-immunodepleted extracts (E and F) from untreated (Ⅺ) or 2fAP-treated (5 min) cells (f and F), cofilin-depleted extracts supplemented with 10 nM recombinant cofilin (E), or 10 nM recombinant cofilin alone (ƒ). A decrease in fluorescence intensity is representative of filament severing activity. Inset, Western blot demonstrating cofilin and actin (loading control) levels after depletion with IgG or anti-cofilin. E, average maximal filament severing levels (ϮS.E.) observed in the presence of untreated and 2fAP-treated cells compared with buffer controls are shown (n ϭ 3). F, pyrene-G-actin was added to severing reactions described in C at t ϭ 2200 s. Inset, initial slopes after monomer addition for boxed regions. unt, untreated; CNT, control. JULY 13, 2007 • VOLUME 282 • NUMBER 28 min, many of the remaining studies examined events primarily after 5 min of PAR-2 activation. These data demonstrate that an early event in PAR-2 signaling is cofilin dephosphorylation and that after the initial cofilin activation some, but not all, is rephosphorylated over time. Although cofilin dephosphorylation results in its activation, there are still factors in the cell that can prevent its severing activity even when it is dephosphorylated (33,51). We used a modified pyrene-actin assembly assay to confirm that PAR-2 actually promoted cofilin-dependent actin filament severing (see "Materials and Methods" for a detailed explanation of the assay). Briefly addition of severing factors to filaments that are already at steady state results in a decrease in pyrene fluorescence, which provides a readout of filament severing activity (44,47). Extracts were immunode-pleted with either IgG, as a negative control (Fig. 1B, f and Ⅺ), or cofilin antibody (Fig. 1B, E and F) to determine whether cofilin was mediating the severing activity. Addition of IgG-depleted extract from 2fAP-treated MDA-MB-468 cells decreased pyrene fluorescence by 63%, whereas extract from untreated cells did not (Fig. 1,  B and C). Addition of fresh pyrene-Gactin ϳ1200 s after 2fAP-treated lysate addition resulted in a sharp increase in fluorescence (Fig. 1D) that was not observed with untreated lysates, indicating that the observed fluorescence decrease was due to filament severing and the subsequent generation of free barbed ends for elongation and not to fluorescence quenching. Importantly the PAR-2-induced severing activity was lost by immunodepletion of cofilin (Fig. 1B, F). Furthermore severing activity in immunodepleted extracts could be restored by addition of recombinant cofilin (Fig. 1B, E). These results clearly show that PAR-2 promotes cofilin dephosphorylation and activation, leading to increased actin filament severing activity.

PAR-2-stimulated Cofilin Dephosphorylation and Filament Severing
Require ␤-Arrestins-To determine whether ␤-arrestins were required for cofilin activation, we assessed cofilin dephosphorylation in MDA-MB-468 cells after siRNA knockdown of either ␤-arrestin-1, ␤-arrestin-2, or both ( Fig. 2 and supplemental Fig. 2) or in embryonic fibroblasts from wild type (MEFwt) and ␤-arrestin-1/2 knockout mice (MEF␤arrDKO) (Fig. 3). PAR-2 signaling in MDA-MB-468 cells has been well characterized in our laboratory; thus, they serve as an excellent model system for studying PAR-2-induced cofilin activation (9,10,27). MEFs from ␤-arrestin knock-out mice provide the best model to address the role of ␤-arrestins in signaling events, but they express less PAR-2; thus PAR-2 responses are less robust (40). Two sets of siRNA oligos that have been described previously (10, 16, 17, 40 -43) were used to knock down each human ␤-arrestin separately or both simultaneously; data from one set is shown. Specificity of siRNA oligos in these studies was confirmed by quantitative Western analysis, which showed a 70 -80% knockdown of each ␤-arrestin with no discernible effect on expression of an unrelated protein, tubulin ( Fig. 2A), and by rescue of knockdown effects with expression of FLAG-tagged rat ␤-arrestin-1 and -2 (Fig. 2, D and E). A full time course revealed that in MDA-MB-468 cells knockdown of either ␤-arrestin inhibited PAR-2evoked cofilin dephosphorylation between 30 and 90 min (supplemental Fig. 2). The maximum difference between control siRNA-transfected and either ␤-arrestin-1 or -2 siRNA-transfected cells was observed at 30 min; thus, subsequent knockdown experiments were performed at 0, 5, and 30 min. Expression of ␤-arrestin-1 siRNA alone had no significant effect on cofilin dephosphorylation at 5 min but inhibited it by 75% at 30 min (Fig. 2, B and C). Expression of ␤-arrestin-2 siRNA alone inhibited cofilin dephosphorylation by 43% at 5 min and by 90% at 30 min (Fig. 2, B and C). siRNA oligos that target both ␤-arrestin-1 and -2 inhibited cofilin dephosphorylation by greater than 85% at both 5 and 30 min (Fig. 2, B and C). Expression of FLAG-tagged rat ␤-arrestin-1 and -2 rescued the effect of individual siRNA knockdown (Fig. 2, D and E). The inhibition of cofilin dephosphorylation observed with double knockdown at 5 min could be rescued by expression of either ␤-arrestin-1 or -2 (Fig. 2, F and G), suggesting that the two ␤-arrestins may be redundant with respect to initial cofilin dephosphorylation. In MEFs, cofilin dephosphorylation was abolished by genetic deletion of both ␤-arrestins (Fig. 3). These data demonstrate that PAR-2 promotes cofilin dephosphorylation by a ␤-arrestin-dependent mechanism and that initial cofilin dephosphorylation can be mediated by either ␤-arrestin. To confirm that the requirement for ␤-arrestins in cofilin dephosphorylation was reflected in filament severing activity, we performed pyreneactin severing assays in the presence of extracts from cells transfected with ␤-arrestin-1 or -2 siRNA. ␤-Arrestin-2 siRNA eliminated and ␤-arrestin-1 siRNA reduced PAR-2-induced severing (by 60%) compared with control siRNA-transfected cells (supplemental Fig. 3). These data clearly demonstrate that PAR-2-induced cofilin dephosphorylation and filament severing activity are mediated by ␤-arrestins.

PAR-2 Promotes Association of ␤-Arrestins with Cofilin and Redistribution of Both to Membrane Protrusions-To address
whether PAR-2 promotes interaction of cofilin with ␤-arrestins, cofilin was immunoprecipitated from MDA-MB-468 cells after PAR-2 activation with rabbit anti-cofilin, and immune complexes were analyzed by SDS-PAGE followed by Western blotting with mouse anti-␤-arrestin. Both ␤-arrestin-1 and -2 (Fig. 4A) associated with cofilin at 5 and 90 min of PAR-2 activation. Furthermore when ␤-arrestin-2 was overexpressed, association with cofilin was observed even in the absence of receptor activation (Fig. 4B). PAR-2 promoted redistribution of both ␤-arrestins (Fig. 5A) and cofilin (Fig. 5B) to membrane protrusions where they were colocalized (Fig. 5C). Interestingly although ␤-arrestin-2 was found predominantly   JULY 13, 2007 • VOLUME 282 • NUMBER 28 in the back of the F-actin-rich protrusions, ␤-arrestin-1 was found predominantly at the tips (Fig. 5C). These data are consistent with a role for ␤-arrestins in the spatial restriction of cofilin activity.

␤-Arrestins and Cofilin Regulation
PAR-2 Promotes Cofilin Dephosphorylation Independently of G-protein Coupling-To determine whether ␤-arrestindependent cofilin activation represented a G-protein-independent pathway, we analyzed cofilin dephosphorylation after 1) treatment with BAPTA-AM to block intracellular Ca 2ϩ mobilization (Fig. 6, A and C), 2) treatment with U73122 to block PLC activity (Fig. 6, B and C), 3) treatment with a G␣ q antagonist (GP2A peptide (52, 53)) (Fig. 6, B and C), or 4) siRNA knockdown of G␣ q/11 (Fig. 6, E and F). Cofilin dephosphorylation was potentiated by BAPTA-AM pretreatment and G␣ q/11 knockdown, suggesting that it does not require G␣ q/11 activation and subsequent Ca 2ϩ mobilization and may be negatively regulated by the G␣ q /Ca 2ϩ signaling axis (Fig. 6, A-E). Further confirming independence of G␣ q signaling, neither the GP2A peptide nor U73122 significantly inhibited PAR-2-stimulated cofilin dephosphorylation (Fig. 6, B and C). The ability of BAPTA-AM, U73122, and GP2A to block PAR-2/G␣ q signaling was confirmed by their ability to inhibit PAR-2-stimulated Ca 2ϩ mobilization (Fig. 6D). G␣ q/11 knockdown was confirmed by quantitative Western analysis; note that levels of another heterotrimeric G-protein (G␣ i ) were unaltered, whereas G␣ q/11 levels were reduced by ϳ70% (Fig. 6E, lower panel). Although the pharmacological agents used in these studies may affect pathways other than G␣ q , the fact that they do not decrease PAR-2-evoked cofilin dephosphorylation strongly supports a model for G␣ q /PLC␤/Ca 2ϩ -independent activation of cofilin. Involvement of G␣ i and G␣ o/12 is unlikely as PAR-2-evoked cofilin dephosphorylation was also insensitive to pretreatment with pertussis toxin or transfection of dominant negative RhoA (not shown). The data thus far demonstrate that PAR-2 can promote cofilin activation via a G␣ q protein/ Ca 2ϩ -independent, ␤-arrestin-dependent pathway (Fig. 6G).
PAR-2-stimulated Cofilin Dephosphorylation Requires CIN-The next two sets of studies were directed at elucidating the molecular mechanisms by which PAR-2 promotes cofilin dephosphorylation by examining the role of the positive (phosphatase) and negative (LIMK) regulators of cofilin activity. We first addressed the possibility that PAR-2 promotes activation of a phosphatase leading to cofilin dephosphorylation. GST-cofilin was phosphorylated with recombinant LIMK in vitro and then incubated with lysis buffer alone or with extracts from cells treated with 2fAP for 0 -45 min. Addition of PAR-2-activated extracts decreased phosphocofilin levels by 50%, suggesting that PAR-2 activates a cofilin phosphatase (Fig. 7A). This activity was insensitive to nonspecific phosphatase inhibitors NaVO 4 and NaF. 3 We next examined whether the recently identified cofilin phosphatase, CIN, was required for PAR-2stimulated cofilin dephosphorylation. Expression of dominant negative CIN (DN-CIN-GFP) inhibited cofilin dephosphorylation by 50% at 5 min and abolished cofilin dephosphorylation at 90 min, demonstrating that PAR-2-stimulated cofilin dephosphorylation requires the activity of CIN (Fig. 7B). CIN also was redistributed to the membrane protrusions upon PAR-2 activation similar to what was observed with cofilin and ␤-arrestins (Fig. 8A). Co-immunoprecipitations revealed that ␤-arrestin-1 associated with CIN upon 5 and 90 min of PAR-2 activation (Fig. 8, B and C), whereas ␤-arrestin-2 could not be co-precipitated with CIN (not shown). Thus, ␤-arrestin-1 may spatially restrict cofilin dephosphorylation through scaffolding of its upstream phosphatase in membrane protrusions.

␤-Arrestins and Cofilin Regulation
LIMK was determined. Consistent with its ability to trigger rapid cofilin dephosphorylation, PAR-2 promoted a rapid decrease in LIMK activity (Fig. 9, A and C). Interestingly knockdown of ␤-arrestin-1 abolished and knockdown of ␤-arrestin-2 attenuated LIMK inhibition; in fact inhibition of ␤-arrestins unmasked a PAR-2-stimulated increase in LIMK activity (Fig.  9, B and C). These findings demonstrate that the PAR-2-evoked inhibition of LIMK activity requires ␤-arrestins and once again suggest that, in their absence, PAR-2 has the opposite effect, i.e. promoting events leading to cofilin phosphorylation. Confocal microscopy revealed a redistribution of LIMK to membrane protrusions and colocalization with ␤-arrestins upon PAR-2 activation similar to what was observed with cofilin and CIN (Fig. 10A). Furthermore we were able to detect interaction between ␤-arrestins and the C terminus of LIMK in vitro. Recombinant GSTtagged ␤-arrestin-1 expressed in E. coli was pulled down by cobalt-Sepharose-bound His-tagged LIMK (amino acids 285-639), whereas GST alone and GST-␤-arrestin-2 were not. No ␤-arrestin-1 was bound to cobalt in the absence of LIMK (Fig. 10B). When the concentration of the LIMK was increased by 5-fold, we observed a corresponding increase in ␤-arrestin-1 binding (Fig. 10B, compare upper and lower panels). Surprisingly when FLAG-tagged ␤-arrestins were expressed in and immunoprecipitated from mammalian cells, both proteins pulled down the recombinant LIMK (Fig. 10C), suggesting that ␤-arrestin-2 might be modified or might interact with another protein, either of which confers the ability to interact with LIMK. Although we could not detect PAR-2-evoked interaction of ␤-arrestins with LIMK in intact cells by co-immunoprecipitation (perhaps due to a transient interaction), these data suggest that ␤-arrestins may also contribute to the spatial regulation of cofilin phosphorylation through localized interaction with LIMK.

Identification of a Major G␣ q / Ca 2ϩ -independent PAR-2 Signaling
Arm-We and others have shown previously that ␤-arrestins are required for activation of a number of signaling proteins and that they can scaffold these molecules to regulate their localization and activity (4,9,27). Furthermore a recent paradigm shift suggesting that some receptors can signal through ␤-arrestins (16,38,42), independently of G-protein engagement, has led to the speculation that ␤-arrestin-dependent events may represent an independent signaling arm for a subset of receptors. The studies described here are the first to demonstrate regulation of the cofilin pathway by either PAR-2 or ␤-arrestins and point to the importance of evaluating both ␤-arrestinand Ca 2ϩ -dependent signaling events when assessing PAR-2 action. We show that PAR-2, a G␣ q -coupled receptor, pro-  ). B, pretreatment with vehicle control, U73122, and GP2A. C, bar graph depicting PAR-2-evoked changes at 5 and 90 min in normalized phosphocofilin levels (ϮS.E.) for the above treatment groups (n ϭ 5). Significant differences between BAPTA-AM and vehicle pretreatment are indicated by * (p Ͻ 0.05), and significant cofilin dephosphorylation is indicated by # (p Ͻ 0.05). D, functional blockade of intracellular Ca 2ϩ mobilization by each pharmacological agent demonstrates their viability. Shown are traces of average [Ca 2ϩ ] i in response to 2fAP after pretreatment with vehicle control, BAPTA-AM, GP2A, and U731222. E, transfection with control (CNT) or G␣ q/11 siRNA. Upper panel, representative Western blot of phospho-(pC) and total cofilin (tC) levels after PAR-2 activation for 0, 5, and 90 min. Lower panels, Western blots of G␣ q/11 (upper) and G␣ i (lower) to show knockdown efficiency and specificity of G␣ q/11 siRNA. F, bar graph depicting changes in normalized phosphocofilin levels (ϮS.E.) in siRNA-transfected cells after 5 and 9 min (m) of exposure to 2fAP (n ϭ 5). Significant differences between G␣ q and control siRNA values are indicated by * (p Ͻ 0.05), and significant changes in cofilin phosphorylation at each time point with respect to untreated cells are indicated by # (p Ͻ 0.05). G, schematic summarizing the effects of blocking ␤-arrestin-dependent and G␣ q -dependent pathways on cofilin activation. IONO, ionomycin; IC, intracellular. JULY 13, 2007 • VOLUME 282 • NUMBER 28 motes ␤-arrestin-dependent cofilin dephosphorylation and activation by a mechanism that is independent of the classic G␣ q /Ca 2ϩ signaling axis. ␤-Arrestin dependence is demonstrated by the fact that siRNA-mediated silencing of either ␤-arrestin-1 or -2 partially inhibited cofilin dephosphorylation and knockdown or genetic ablation of both ␤-arrestins abolished it. Independence from the classic G-protein pathway is demonstrated on multiple levels: siRNA-mediated silencing of G␣ q/11 itself, inhibition of G␣ q /receptor coupling by the GP2A, and inhibition of signaling intermediates PLC␤ (by U73122) and Ca 2ϩ (by BAPTA-AM). The effects of U73122 were less pronounced than those of the other inhibitors possibly because it can also block PLC␥ activation, which is not activated by G␣ q ; a possible ␤-arrestin-dependent role for PLC␥ in PAR-2 signaling has never been investigated. In the case of G␣ q/11 siRNA and BAPTA-AM treatment, cofilin dephosphorylation was actually potentiated. Conversely in the absence of both ␤-arrestins, PAR-2 was able to promote a mild increase in cofilin phosphorylation. The prevalence of one pathway over the other in a given cell type may depend not only on the relative abundance of ␤-arrestins and G␣ q/11 but on the expression of other components of the cofilin pathway. Together these results suggest that ␤-arrestin-dependent signaling can operate not just independently of, but in opposition to, G-proteindependent signaling. At first glance, this model may appear counterintuitive; i.e. why would a cell want a single receptor to send opposing signals simultaneously? However, it is consistent with the concept that during a process in which a cell must polarize the same molecular events occurring in one cellular microdomain need to be suppressed in another. The physiological relevance of this ␤-arrestin-dependent pathway is highlighted by the fact cofilin activation can lead to increased cell motility and tumor metastasis (31, 54 -57), and we have demonstrated previously that constitutive activation of ␤-arrestin-dependent PAR-2 signaling is associated with the basal migration of metastatic tumor cells (10).

␤-Arrestins and Cofilin Regulation
Molecular Mechanisms of ␤-Arrestin-dependent Cofilin Dephosphorylation-Although ␤-arrestins have been demonstrated to facilitate cell migration in a number of systems (9 -13;15), few studies have revealed a molecular link between ␤-arrestins and proteins that are known to directly affect actin assembly. The cofilin pathway is of particular interest in this regard because it is widely accepted that its activity must be tightly regulated both spatially and temporally, and ␤-arrestins are known to exert spatial control over many signaling molecules (1,2). Here we demonstrate that PAR-2-stimulated cofilin dephosphorylation involves modulation of both its inhibitory kinase, LIMK, and one of its stimulatory phosphatases, CIN (Fig. 11). In addition, ␤-arrestins can interact with cofilin, CIN, and LIMK, and all five proteins were redistributed to membrane protrusions upon PAR-2 activation, consistent with a model where ␤-arrestins scaffold cofilin and its upstream regulators to spatially restrict their activity. Whether CIN, LIMK, and cofilin are present in a single macromolecular complex remains to be determined. For example, if ␤-arrestin-1 can bind to both LIMK and CIN, it may scaffold both in a complex such that in specific cellular microdomains LIMK activity is opposed by CIN-mediated dephosphorylation of cofilin (Fig. 11). There is precedent for this model as another cofilin phosphatase (slingshot) has been identified in a complex with LIMK (34,58). Whether ␤-arrestin-1 binding to CIN facilitates either its redistribution to the leading edge or its activity remains to be determined. Likewise it remains to be determined whether ␤-arrestins directly inhibit LIMK activity, prevent its activation by upstream kinases, or both. Further characterization of the opposing signals from PAR-2 to LIMK is necessary to fully elucidate this mechanism. Similarly char- acterization of CIN activity and its regulation is an uncharted territory, and future studies will be necessary to determine whether ␤-arrestin-1 is required for PAR-2-stimulated CIN activation or whether it merely exerts spatial control over cofilin dephosphorylation by scaffolding the phosphatase. Association of cofilin with ␤-arrestins may affect both its localization and its phosphorylation by LIMK. In support of the former, overexpression of ␤-arrestin-2 appeared to restrict cofilin to the back of membrane protrusions (compare cofilin localization in Fig. 5, B and C). Another possibility is that ␤-arrestin-2 binding might stabilize dephosphorylated cofilin similar to the way in which 14-3-3 was shown to bind and stabilize phosphorylated cofilin (59).
Our previous studies have strongly suggested that both ␤-arrestins are required for ERK1/2 activation, but identifying a selective role for each has remained elusive (10). More recently, we have evidence that ␤-arrestin-1 mediates early ERK activation, receptor internalization, and lysosomal targeting, whereas ␤-arrestin-2 mediates prolonged ERK1/2 activation and internalization (60). When investigating the effects of silencing individual ␤-arrestins on PAR-2-induced cofilin activation here, there were apparent differences in the temporal regulation of cofilin activation by each ␤-arrestin with ␤-arrestin-2 having a greater effect at earlier time points and ␤-arrestin-1 having a greater effect at later time points. Initial dephosphorylation (at 30 s or less) was abolished only when both ␤-arrestins were deleted, suggesting that for the initial increase in cofilin activity the two ␤-arrestins are functionally redundant. However, the two proteins may differ in the mechanisms by which they regulate cofilin. We detected interactions between CIN and ␤-arrestin-1 but not ␤-arrestin-2, suggesting that regulation of cofilin phosphatase activity might be mediated primarily by ␤-arrestin-1. Conversely cofilin constitutively associated with overexpressed ␤-arrestin-2 but not ␤-arrestin-1. Teasing out the subtle differences between ␤-arrestin-1 and -2 in PAR-2 signaling may further clarify the molecular mechanisms demonstrated here.
␤-Arrestin-dependent Actin Filament Severing-Why might ␤-arrestins be necessary for cofilin activation? Other studies have demonstrated that cofilin localizes to the lamellipodium tip immediately upon epidermal growth factor stimulation where it facilitates actin assembly by creating free barbed ends (55,57), and site-specific activation of cofilin in live cells leads to membrane protrusion at that site (56). These observations are consistent with the idea that localized filament severing promotes actin reorganization and assembly through treadmilling of actin filaments at the leading edge and point to the importance of spatially restricting cofilin activity. We observed a PAR-2-induced increase in actin filament severing activity that was abolished by immunodepletion of cofilin and restored by addition of 10 nM recombinant cofilin. Furthermore PAR-2evoked filament severing activity was severely impaired after ␤-arrestin knockdown. Thus, it is likely that ␤-arrestins play a key role in the spatial restriction of cofilin-induced filament severing during PAR-2-evoked cell migration. However, recent studies have raised some controversy as to whether cofilin is merely an actin filament-severing protein because at high concentrations cofilin can promote actin filament assembly instead of severing (61). A model where ␤-arrestins facilitate nucleation through localized cofilin accumulation is also plausible and should be addressed in future studies.
To summarize, PAR-2 activates the actin filament-severing protein cofilin through two molecular mechanisms, both apparently dependent on ␤-arrestins: inhibition of LIMK activity and activation of the cofilin phosphatase. Cofilin activation is independent of PAR-2-mediated G␣ q coupling and mobilization of intracellular Ca 2ϩ . These observations are important from both a biomedical and a basic signaling perspective. First, PAR-2 activation has become an increasingly popular thera-  JULY 13, 2007 • VOLUME 282 • NUMBER 28 peutic target in inflammatory diseases, and its activation is associated with increased migration of tumor cells (10,(62)(63)(64)(65). Because Ca 2ϩ mobilization and G␣ q coupling are the most commonly used readouts of PAR-2 activation, studies using PAR-2 agonists and antagonists in disease models might need to be revisited in the context of ␤-arrestin-dependent signaling. Also tissue-specific differences in molecular components of  . Arrows indicate regions of colocalization after 5 min of PAR-2 activation. Scale bar ϭ 10 m. B, 5 g of recombinant GST-␤arrestin-1 or GST-␤-arrestin-2 was incubated with 10 g (upper panel) or 2 g (lower panel) of His 6 -LIMK (amino acids 285-638), and His 6 -LIMK was precipitated with cobalt resin. Cobalt resin alone was included as a control for nonspecific protein precipitation. Both bound and unbound proteins were analyzed by SDS-PAGE followed by Coomassie staining. C, FLAG-␤-arrestin-1 and FLAG-␤-arrestin-2 were purified with anti-FLAG-agarose from crude cell lysates and incubated with 20 ng of His 6 -tagged recombinant LIMK (amino acids 285-638). Anti-FLAG-agarose pulldowns from untransfected cells were used as a negative control. Protein was eluted from beads (25 l), and Western blots were probed with anti-FLAG (upper panel) or anti-His 6 (lower panel) to demonstrate interaction of both ␤-arrestins. CNT, control; ␤ or ␤-arr, ␤-arrestin. each PAR-2 signaling arm could explain the somewhat disparate set of cellular responses reported for PAR-2. Second, because ␤-arrestins have been implicated in G-protein-independent signaling downstream of other receptors, the molecular mechanisms elucidated here might be utilized by other G-protein-coupled receptors.