Constitutive Protease-activated Receptor-2-mediated Migration of MDA MB-231 Breast Cancer Cells Requires Both β-Arrestin-1 and -2*

Protease-activated receptor-2 (PAR-2) is activated by trypsin-like serine proteases and can promote cell migration through an ERK1/2-dependent pathway, involving formation of a scaffolding complex at the leading edge of the cell. Previous studies also showed that expression of a dominant negative fragment of β-arrestin-1 reduces PAR-2-stimulated internalization, ERK1/2 activation, and cell migration; however, this reagent may block association of many proteins, including β-arrestin-2 with clathrin-coated pits. Here we investigate the role of PAR-2 in the constitutive migration of a metastatic breast cancer cell line, MDA MB-231, and use small interfering RNA to determine the contribution of each β-arrestin to this process. We demonstrate that a trypsin-like protease secreted from MDA MB-231 cells can promote cell migration through autocrine activation of PAR-2 and this correlates with constitutive localization of PAR-2, β-arrestin-2, and activated ERK1/2 to pseudopodia. Addition of MEK-1 inhibitors, trypsin inhibitors, a scrambled PAR-2 peptide, and silencing of β-arrestins with small interfering RNA also reduce base-line migration of MDA MB-231 cells. In contrast, a less metastatic PAR-2 expressing breast cancer cell line does not exhibit constitutive migration, pseudopodia formation, or trypsin secretion; in these cells PAR-2 is more uniformly distributed around the cell periphery. These data demonstrate a requirement for both β-arrestins in PAR-2-mediated motility and suggest that autocrine activation of PAR-2 by secreted proteases may contribute to the migration of metastatic tumor cells through β-arrestin-dependent ERK1/2 activation.

One of the earliest steps in tumor cell metastasis is the reorganization of the actin cytoskeleton and initiation of cell migration (1). A variety of extracellular signaling molecules including growth factors and chemokines can promote these events and recent evidence suggests trypsin-like proteases se-creted from tumor cells might generate migratory signals through the activation of cell surface protease-activated receptor-2 (PAR-2) 1 (2)(3)(4)(5)(6)(7). PAR-2 is highly expressed in a variety of normal and tumor cells and is activated by nanomolar concentrations of several serine proteases, including trypsin, tryptase, Factors VIIa and Xa, and membrane-type serine protease-1 (MT-SP1I) (8 -10). Proteolytic cleavage of the PAR-2 extracellular N terminus reveals a tethered ligand that binds to and activates the receptor. Activating peptides (APs) corresponding to the tethered ligand sequence SLIGKV/RL (human/mouse) and a chemically modified version of AP (2-furoyl-LIGRL-ornithine-NH 2 (2f-AP)) also activate PAR-2, in the absence of proteolytic cleavage (9,11). A number of PAR-2-mediated events, including activation of ERK1/2 and cell migration are blocked by expression of a dominant negative fragment of ␤-arrestin-1 that encodes the clathrin binding domain (12). However, the specificity of this reagent is unclear as it acts by binding ␤-arrestin contact sites on clathrin and may affect other clathrinregulated processes as well. In recent studies, we demonstrated that PAR-2 promotes chemotaxis by a mechanism involving localization of activated ERK1/2 and its upstream regulatory kinases to the pseudopodia. This pseudopodial localization appears to occur through the formation of a scaffolding complex containing ␤-arrestin(s) that we refer to as an endosomal scaffold (12,13).
While ␤-arrestins were originally thought to mediate signal termination, there has been a growing body of evidence that they can also serve as signaling scaffolds for a number of receptors (14 -16) and that further functional differences exist between the two ␤-arrestin family members. Recent studies using siRNA to specifically knockdown either ␤-arrestin-1 or 2 suggest that ␤ 2 AR primarily utilizes ␤-arrestin-2 for internalization, while the angiotensin II type 1a receptor (AT1aR) can use either for internalization but requires ␤-arrestin-2 for signaling to ERK1/2 (17). Additional differences may exist between other cell types and receptors. The role of ␤-arrestins in cell migration is supported by genetic studies demonstrating impaired CXCR-mediated motility in lymphocytes from ␤-arrestin-2 knock-out mice (18). However, the specific ␤-arrestin family member(s) that mediate(s) PAR-2-promoted cell motility has not been investigated nor has the role of ␤-arrestins in tumor cell migration. The purpose of these studies was 2-fold. First, using siRNA to specifically knockdown each ␤-arrestin, we wished to determine which ␤-arrestins mediate ERK1/2 activation and chemotaxis downstream of PAR-2. Second, using the metastatic breast cancer cells (MDA MB-231) that migrate constitutively and cells with low metastatic potential (MDA MB-468) (19,20) that do not migrate without stimulation, we wished to investigate how PAR-2 and ␤-arrestins might modulate cell migration in a metastatic tumor cell.
Microscopy-Cells were seeded onto collagen-coated coverslips for 6 h and serum-starved overnight. Protease inhibitor mixture (PIC: 100 M leupeptin, 2 g/ml aprotinin, 6.25 g/ml ␣ 1 -antitrypsin, and 0.5 mM 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride (AEBSF)) or hAP were added for 5-90 min as described, and cells were then fixed in normal buffered formalin followed by pressure cooker antigen retrieval in 1% sodium citrate and stained with anti-B5 (1:500) overnight and Alexa-595 conjugated secondary. Serial sections (1 m, 40ϫ and 100ϫ objectives) were taken on a Zeiss LSM510 microscope. For images in Fig. 2C, images were taken with a 2.5ϫ zoom. For video microscopy, MDA MB231 cells were observed by phase contrast microscopy in bicarbonate-free medium on a Nikkon Inverted microscope (40ϫ objective), and time-lapse images were taken every 2 min using Metamorph software. A quicktime video with 1.0-s delays (2 s delays indicate addition of PIC and AP) is included in the supplemental material.
Chemotaxis Assays and Pseudopodia Purification-For all migration and pseudopodia purification assays, cells were serum-starved overnight in 0.1% bovine serum albumin then seeded onto collagen (10 g/ml)-coated 12-mm Transwell filters with 8-m pores for migration (10 4 /filter) and 3-m pores for pseudopodia purification (10 5 /filter). Agonists (trypsin, AP, or 2f-AP) were added to the bottom chamber, and migrated cells or pseudopodia and cell bodies were analyzed as described previously (13). For inhibitor experiments, a mixture of serine protease inhibitors (PIC), 1 g/ml soybean trypsin inhibitor (SBTI), 2 g/ml leupeptin or 2 g/ml aprotinin, 0 -100 M 2-furoyl-ILRGL-O, or 10 M PD98059 was added to the top chamber after cells attached for 2 h and migration monitored as described previously.
Subcellular Fractionation-5 ϫ 10 5 cells/60-mm dish (grown 24 h) were serum-starved for 16 h, treated with 50 M AP for 0 -60 min at 37°C, washed twice, lysed by Dounce homogenization (10 strokes) in 0.25 ml of hypotonic lysis buffer, and fractionated as described previously (13). Quantification of band densities was performed using Adobe Photoshop, and pERK band densities were normalized to total ERK band densities for each lane.
Trypsin Secretion-At the start of the experiment 3 ml of fresh medium was added to cells (80% confluent) and samples were collected for 5-90 min. Protein was concentrated by chloroform/MeOH extraction and analyzed by SDS-PAGE followed by immunoblotting with antitrypsin/trypsinogen. To estimate quantities of trypsin(ogen), 1 g of purified trypsin was included, and band densities were compared with purified trypsin using Adobe Photoshop histogram analysis.
Statistics-All graphs and statistical analyses were performed using Kaleidagraph Version 3.5.1 or Graphpad Prism, Version 3.5. Migration assays and phospho-ERK analyses were performed a minimum of three times. Analysis of variance and Tukey t tests were used to determine statistical significance and significant differences of values. Previous studies demonstrated that PAR-2 agonists are chemotactic, as a uniform concentration is less effective than a gradient (added to the bottom chamber only) at eliciting a migratory response (13). To investigate whether the high basal level of MDA MB-231 cell migration was mediated by PAR-2, we first determined invasion in the presence of a scrambled PAR-2 peptide; peptides containing a scrambled tethered ligand sequence have been demonstrated to partially inhibit PAR-2 activation (11,21). scr-AP inhibited basal MDA MB-231 cell migration by 55 Ϯ 10% reduction (Fig. 1B), suggesting that constitutive PAR-2 activation contributes to their migratory behavior.

PAR-2 Mediates Base-line Migration in
To address the hypothesis that constitutive PAR-2 activation might be due to the presence of serine protease activity in the medium, we investigated the effects of a protease inhibitor mixture (PIC, see "Experimental Procedures" for description) or individual protease inhibitors on MDA MB-231 cell migration. To rule out the possible role of thrombin receptors, we included the thrombin inhibitor, hirudin. PIC and SBTI inhibited basal MDA MB-231 cell migration by 46 Ϯ 5% and 50 Ϯ 22%, respectively, while leupeptin and aprotinin alone had no effect (Fig. 1C). Hirudin addition caused a mild increase in MDA MB-231 migration, consistent with reported observations in this cell line showing that PAR-1 inhibition of PAR-2 induced cell migration (3). To demonstrate that PAR-2 was responsible for the protease-mediated motility and that the protease inhibitors were not cytotoxic, we added hAP (SLIGKV) or 2f-AP in the presence of PIC. 2f-AP is a recently described synthetic PAR-2 agonist that achieves the same response as APs at nanomolar rather than micromolar concentrations (11). hAP and 2f-AP restored cell migration to 101 Ϯ 5% and 180 Ϯ 10% of basal in PIC-treated cells (Fig. 1C), suggesting that constitutive proteolytic cleavage of PAR-2 strongly contributes to the migratory behavior of MDA MB-231 cells. PIC did not block PAR-2-stimulated migration in MDA MB-468 cells (data not shown), further confirming that the inhibitors were not cytotoxic. Consistent with the inhibitory effects of PIC on MDA MB-231 cell migration, addition of PIC to MDA MB-231 cells results in retraction of cell processes and subsequent addition of hAP results in their membrane ruffling and filipodia exten-sion ( Fig. 1D and supplemental video 1).
MDA MB-231 Cells Secrete Biologically Active Trypsin-The ability of trypsin inhibitors to block MDA MB-231 cell migration suggests they might secrete trypsin, leading to autocrine activation of PAR-2. We compared the trypsin secreted from MDA MB-231 and MDA MB-468 cells by Western blot of conditioned medium (CM), sampled over 3 h. Trypsin appeared in the medium of MDA MB-231 cells almost immediately and continued to accumulate, while no detectable trypsin appeared in the medium of MDA MB-468 cells (Fig. 3A). The approximate concentration of trypsin of conditioned medium was calculated by comparing the density of antitrypsin immunoreactive bands to a purified trypsin control and was estimated to be in the low nanomolar range (ϳ1-5 nM). We then tested the ability of CM from MDA MB-231 cells to or MDA MB-468 cells to promote chemotaxis of MDA MB-468 cells. While MDA MB-231 CM resulted in a 12.5 Ϯ 1.4-fold increase in cell migration, MDA MB-468 CM had no effect. Incubation of CM with PIC abolished cell migration demonstrating that the major chemotactic factor present in the medium is a serine protease. Purified trypsin induced a similar (14 Ϯ 2-fold) increase in cell migration (Fig. 3B).

PAR-2 Is Constitutively Localized to Pseudopodia in MDA
MB-231 Cells-Western blot analysis of total cell extracts suggests that both MDA MB-231 and MDA MB-468 cells express similar amounts of PAR-2 (Fig. 4A), but the patterns of posttranslational processing differ slightly. We observe that PAR-2 is heavily glycosylated (indicated by the arrowheads in Fig.  4A), and these bands collapse into the lower band (indicated by an arrow) upon tunicamycin treatment of cells. An intermediate band (indicated by an asterisk) is both glycosylated and phosphorylated as it shifts down upon Alkaline phosphatase treatment of cell lysates. 2 The uppermost glycosylated band may contain additional modifications. These different species of PAR-2 are more evident in some cell lines than others, and we are currently investigating the role of differentially processed PAR-2 in its signaling properties.
Immunofluorescence reveals that PAR-2 localization in MDA MB-231 cells appears to be enriched at the leading edge, while in the non-migratory MDA MB-468 cells it is distributed more uniformly around the cell surface (Fig. 4B). Furthermore, treatment of MDA MB-231 cells with trypsin inhibitors results in redistribution of PAR-2 around the cell periphery, while treatment of MDA MB-468 cells with hAP results in accumulation of surface PAR-2 at membrane ruffles (Fig. 4C). In MDA MB231 cells, a considerable amount of PAR-2 is localized to intracellular compartments, probably both the Golgi apparatus, as has been reported by several other investigators (22), and the lysosomes, due to constitutive receptor activation. Of most interest to us from the perspective of cell migration is that fraction localized to the leading edge.
To determine whether PAR-2 is constitutively sequestered in 2 M. Mathur and K. DeFea, unpublished observations. extending pseudopodia along with other components of the previously described endosomal scaffold (12, 13), we purified pseudopodia and cell bodies from unstimulated MDA MB-231 cells as described previously (13,23). Only MDA MB-231 cells were included because pseudopodia formed in response to agonist in MDA MB-468 cells are smaller and result in too little protein for analysis. In untreated MDA MB-231 cells, 95 Ϯ 5% ␤-arrestin-2, 65 Ϯ 20% ␤-arrestin-1, 75 Ϯ 5% pERK, and 65 Ϯ 5% pRaf are localized to the pseudopodia (Fig. 5). Total ERK1/2 and Raf-1 are distributed approximately equally between pseudopodia and cell bodies, suggesting the phosphorylated forms are specifically restrained at the leading edge. A study by Brahmbhatt and Klemke (24) comparing found that phospho-ERK was specifically associated with extending but not retracting pseudopodia, consistent with the idea that sequestration of signaling molecules to the leading edge might promote persis- and pseudopodia (from bottom) were analyzed by Western blot with antibodies to pRaf-1, pERK1/2, PAR-2, ␤-arrestin-1/2, and histone. A, representative Western blot from three separate experiments is shown on the left, and quantification of pseudopodia to cell body ratios is shown on the right. Differentially modified forms of PAR-2 are indicated: arrowhead ϭ glycosylated, * ϭ phosphorylated, † ϭ high molecular weight form. pRaf and pERK1/2 levels (P) were compared with total ERK1/2 and Raf levels (T). For PAR-2, pseudopodial/cell body ratio was calculated for the sum of all four bands (T) and for each individual band. tent migration. Interestingly, the modified forms of PAR-2 are differentially distributed. While ϳ80% of the glycosylated PAR-2 is found in the pseudopodia, 97% of the phosphorylated protein and 70% of the unglycosylated protein are restricted to the cell body (Fig. 5). Further studies are required to determine whether glycosylation and phosphorylation play a role in localization of PAR-2 to the pseudopodia and subsequent cell migration. These data are consistent with the hypothesis that trypsin secreted from MDA MB-231 cells activates PAR-2 resulting in its incorporation into a signaling complex at the leading edge of the cell.

MDA MB-231 Cell Migration Is Dependent on Both ␤-Arrestin-1 and -2-
Our previous data showed that blocking receptor internalization with a dominant negative fragment of ␤-arrestin-1 inhibits PAR-2-stimulated ERK1/2 activation and motility (12,13). Recent evidence suggests that ␤-arrestins are not merely redundant family members, and while some receptors appear to depend predominantly on one ␤-arrestin for both signaling and signal termination functions, others use one for internalization and another for signaling. It is unclear which one ␤-arrestin PAR-2 "prefers," and the dominant negative ␤-arrestin-1 used previously can inhibit both and may even have a more global effect on membrane trafficking. We used siRNA to knockdown ␤-arrestin-1 and -2 expression either individually or together and found that inhibiting expression of either of the ␤-arrestin-1 or -2 reduced base-line MDA MB-231 cell migration (by 55 Ϯ 10% and 50 Ϯ 12%, respectively), while simultaneous knockdown of both ␤-arrestins resulted in a 71 Ϯ 4% reduction (Fig. 6B). To investigate whether the effects of PAR-2 and ␤-arrestins on cell migration was through the same pathway, we examined the effect of siRNA knockdown on PAR-2-stimulated cell motility, in the presence of PIC. As described in the legend to Fig. 1C, addition of PIC depresses base-line cell migration, and addition of hAP or 2f-AP restores migration. In the presence of PIC, siRNA knockdown of ␤-arrestin-1 expression abolishes 2f-AP-stimulated cell migration, and knockdown of ␤-arrestin-2 reduces 2f-AP-stimulated migration by 92 Ϯ 6%, suggesting that PAR-2-stimulated migration requires both ␤-arrestin-1 and -2 (Fig. 6C) and introducing the question of whether they each have distinct roles in PAR-2 signaling.

DISCUSSION
In the studies described here, we provide evidence supporting the hypothesis that secreted proteases enhance the metastatic potential of tumor cells by inducing migration (2,4,5,10) and introduce a mechanism for these events. We show that a highly metastatic breast cancer cell line, serine protease inhibitors or a scrambled PAR-2 peptide. Subsequent activation of PAR-2 by specific activating peptides restores cell migration in the presence of protease inhibitors to basal levels or higher. These data suggest that the contribution of serine proteases to MDA MB-231 cell migration is mediated by PAR-2. While other serine proteases capable of activating PAR-2 may be present as well, trypsin is a known agonist of the receptor, and its increased secretion in MDA MB-231 cells suggests it may play an important role in constitutive PAR-2 activation and cell migration in these cells. Further strengthening this hypothesis, we observe that medium from MDA MB-231 cells can promote migration of the less metastatic MDA MB-468, and this effect is ameliorated by with serine protease inhibitors. While autocrine activation of PAR-2 is likely to account for only part of the constitutive migration of these and other tumor cells, the data here support a model where trypsin-like proteases contribute to the metastatic potential of a cell by activating migratory pathways through cell-surface receptors and provide a physiologically significant role for ␤-arrestin scaffolding of signaling molecules.
A specific requirement for ␤-arrestin-2 in lymphocyte motility has recently been demonstrated using ␤-arrestin-1 and -2 knock-out mice (18). An interesting outcome of the studies described here is the evidence that both ␤-arrestins are required for PAR-2-mediated motility, suggesting they are not redundant, even for receptors that appear to utilize both. Such a finding is consistent with studies by others demonstrating that the two ␤-arrestin proteins differ in their specific cellular functions (17,18,(25)(26)(27), and their precise roles in the regulation of actin assembly machinery are described in studies from our laboratory currently in preparation. 3 Previous studies in our laboratory demonstrated that migration is dependent upon ␤-arrestins and ERK1/2 and appears to involve trafficking of the activated kinases to the pseudopodia on early endosomal vesicles (12,13). Here we show that the components of this endosomal scaffold are constitutively localized to pseudopodia in MDA MB-231 cells. Interestingly, although both ␤-arrestin-1 and -2 are required for cell migration, ␤-arrestin-2 is enriched to a greater extent in the pseudopodia, consistent with the hypothesis that the two ␤-arrestins may have separate functions in cell motility.
Studies comparing a variety of breast cancer cell lines have demonstrated that MDA MB-468 cells exhibit low levels of endothelial cell invasion, compared with MDA MB-231 cells, and are not metastatic in a nude mouse model (19,20,28). The fact that less metastatic tumor cells both express PAR-2 and migrate in response to its activation is interesting because a number of proteases released during inflammation can activate PAR-2 (8,29), raising the possibility that chronic inflammation in a cancerous tissue might increase the metastatic potential of a tumor.
Most chemotherapeutic agents target cell division; because metastasis involves proliferation of tumor cells after they have invaded the blood vessel wall, the tumor cells are thought to prevent metastasis as well. However, tumor cells may possess multiple mechanisms for cell migration, which they can utilize independent of cell division. Further studies are required to determine whether trypsin secretion and PAR-2 endosomal scaffold formation are altered in a broad spectrum of tumor cells, whether PAR-2 is required for metastasis of certain tumors in vivo, the efficacy of serine protease inhibitors and PAR-2 antagonists in preventing metastasis, and the role of ␤-arrestins in this process. The use of serine protease inhibitors as therapeutics is not a new concept, and trypsin and tryptase inhibitors have been used in the treatment of diseases such as asthma, emphysema, and ␣ 1 -antitrypsin deficiency (25, 30 -32). Over the last few years, however, the utility of targeting serine proteases as a means of preventing tumor metastasis has been an active area of research. While the major focus has been on the role of proteases in degrading the tumor stroma and basement membrane, the studies described here suggest that activation of cell migration pathways through cell surface receptors may also contribute to the prometastatic action of serine proteases.