Integrin Engagement Differentially Modulates Epithelial Cell Motility by RhoA/ROCK and PAK1^*

Cell migration is essential for a number of biological and pathological processes, including normal development, angiogenesis, wound repair, and tumor invasion and metastasis. The process of cell spreading and migration on extracellular matrix (ECM)¹ involves integrin receptors and dynamic changes in the cytoskeleton. Migration represents a multi-step process including formation of adhesive protrusions at the leading edge, release of adhesions, and cell rear retraction (1). The complex interplay between integrins and the cytoskeleton is regulated by specific signaling pathways that are not completely understood.

The Rho family GTPases, particularly Rho, Rac, and Cdc42, modulate many aspects of cytoskeletal function that occur during migration (2–6). Rac1 seems to be essential in most cells for the protrusion of lamellipodia at the leading edge and for forward cell movement. In contrast, RhoA is required to maintain substrate adhesion during cell movement and to produce contractile force in the forwarding migrating cell. The main function of Cdc42 is to maintain cell polarity and initiate the formation of filopodium.

The role of individual Rho GTPases in migration may depend on cell type. The contribution of RhoA in motility has been established in specific cell types such as colon carcinoma cells, hepatoma cells, and lymphoma cells (7–9). However, many studies on the role of Rho proteins in migration have used fibroblasts (10, 11). Other studies show that increased Rac activity promotes migration and invasion of lymphocytes (7, 12). For lung adenocarcinoma cells (13), Rac1 can inhibit the motility of epithelial cells because of an increase in the formation of cadherin junctional adhesions (14). High RhoA activity inhibits movement in fibroblasts and lung adenocarcinoma cells (13, 15). Thus, the balance among RhoA, Rac, and Cdc42 activities will determine whether a given cell remains stationary or is migratory.

It is well known that ECM proteins can trigger cell spreading and motility through integrin-dependent regulation of Rho family members. The type of ECM also appears to have dramatic effects on the migratory response of the cell. For example, in Madin-Darby canine kidney epithelial cells, Rac activation enhanced migration on collagen but suppressed migration on laminin-1 or fibronectin substrates (16). During the initial phase of spreading on fibronectin, RhoA activity is reduced through activation of p190 RhoGAP as a result of Src and FAK signaling in fibroblasts (15, 17, 18); subsequently, RhoA activity increases markedly (17). Rac1 and Cdc42 activities are known to be high during spreading and membrane protrusion (19, 20). Other ECM proteins, such as laminin-10/11 (13) and laminin-8 (21), activate Rac1 to promote cell migration. However, α4β1 integrin interaction with fibronectin down-regulated RhoA activity and induced melanoma cell migration (22). Another important issue concerns differences between the behavior of cells studied on two-dimensional surfaces versus three-dimensional matrices (23).

Previously, we showed that for squamous cell carcinoma (SCC) cells, laminin-5 ligand promotes rapid cell scattering, whereas fibronectin and collagen I do not (24). In the present study, we analyzed the integrin-mediated regulation of Rho GTPases and their downstream effectors resulting from adhesion to laminin-5 and type I collagen substrates in SCC cells. On laminin-5 substrate, α3β1 integrin preferentially inactivated RhoA and induced activation of Cdc42 and PAK1, thereby promoting migration of oral SCC cells. In contrast, on type I collagen, α2β1 integrin strongly activated RhoA, leading to enhanced focal contact formation, thereby hindering cell migration. These results suggest that Rho signaling in SCC may be important in defining cell phenotype.

EXPERIMENTAL PROCEDURES

Reagents and Antibodies—Y-27632 and C3 transferase were purchased from Calbiochem; myelin basic protein (MBP) and poly-l-lysine (PLL) were from Sigma; type I collagen was from Cohesion Technologies (Palo Alto, CA). MAbs against Rac1, Cdc42, and paxillin (clone 165) were purchased from BD Transduction Laboratories (Lexington, KY). Anti-RhoA mAb and anti-PAK (N-20) polyclonal antibody were obtained from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Anti-hemagglutinin mAb was purchased from Covance (Richmond, CA). Anti-laminin-5 blocking antibody (P3H9–2) and anti-vinculin mAb were obtained from Chemicon International, Inc. (Temecula, CA); anti-phospho-PAK1 (Thr⁴²³) polyclonal antibody and anti-Myc mAb were purchased from Cell Signaling Technology Inc. (Beverly, MA); VC5 (mouse anti-α5 integrin) and GoH3 (rat anti-α6 integrin) were purchased from BD Pharmingen (San Diego, CA). J143 (anti-human α3 integrin) was obtained from ATCC. VM1 and VM2 mAb (anti-α2 and α3 subunits, respectively) were described previously (25). AIIB2 (rat anti-human β1 integrin) was provided by Caroline Damsky (University of California, San Francisco). Anti-mouse and anti-rabbit secondary antibodies, conjugated to horseradish peroxidase for immunoblotting, or conjugated to fluorescein isothiocyanate for confocal and immunofluorescence microscopy, were obtained from Jackson ImmunoResearch Laboratories (West Grove, PA). Rhodamine-conjugated phalloidin was obtained from Molecular Probes (Eugene, OR).

Cell Culture—Cells from the human SCC cell lines HSC-3 and UM-SCC-10A were maintained as described previously (24) in DMEM (Mediatech, Inc., Herndon, VA) supplemented with 10% fetal bovine serum (Gemini Bio-Products, Woodland, CA) and cultured at 37 °C in a humidified atmosphere of 5% CO₂ and 95% air. Cells used for adhesion and motility assays were serum-starved overnight.

Purified laminin-5 was used as described previously for adhesion and motility assays (24, 25). Laminin-5 matrix substrate was derived from human SCC cells as described previously (24). In brief, cells were grown to confluence on tissue culture plates for 48 h. After washing with phosphate-buffered saline, cells were removed by treatment with 20 mm NH₄OH for 5 min according to the method described previously (26, 27). The matrix was then extensively washed with phosphate-buffered saline before use.

Adenoviral Infection—Adenovirus encoding constitutively active RhoA (V14RhoA) or dominant-negative RhoA (N19RhoA) was kindly provided by A. Hassid (University of Tennessee, Memphis, TN). Adenovirus encoding GFP or dominant-negative Cdc42 (N17Cdc42) was a gift of G. E. Davis (Texas A&M University System, College Station, TX). Adenovirus encoding wild-type (WT) PAK1 or constitutively active PAK1 (E423 PAK1) was from W. T. Gerthoffer (University of Nevada, Reno, NV). HSC-3 cells were infected with adenovirus at a multiplicity of infection of 500 in 1 ml of culture medium in 6-well plates or in 4 ml of culture medium in 10-cm plates. After 2 h incubation, 2 or 6 ml of DMEM was added to the 6-well or 10-cm plates, respectively. After 24 h in culture, cells were processed for experiments as described below.

Inhibitor Treatments—For treatment with the Rho inhibitor C3 transferase, cells were preincubated with 5 μg/ml C3 transferase in serum-free medium overnight. For treatment with the Rho-associated coiled-coil kinase (ROCK) inhibitor Y-27632, cells were preincubated with 25 μm Y-27632 in DMEM for 30 min before and included during the experiments.

Cell Spreading Measurements—Suspended cells were seeded onto plates previously coated with collagen I or laminin-5 substrate for the indicated times at 37 °C in serum-free medium. After fixation with 4% paraformaldehyde, cells were stained with 2% Coomassie Brilliant Blue (Sigma) in 45% methanol and 10% acetic acid for 10 min (15). The relative areas in pixels of more than 20 individual cells were generated using Metamorph and NIH Image software. The average of the relative areas of cells plated on laminin-5 substrate for 30 min was chosen as the maximal cell area. The ratio of the cell area at the indicated time point to the maximal cell area was then determined.

Immunofluorescent Staining—Cells were seeded onto chamber slides (Nalge Nunc International, Naperville, IL) coated with collagen I or laminin-5 and incubated at 37 °C for 1 h. After fixation with 4% paraformaldehyde for 10 min and permeabilization with 0.5% Nonidet P-40 in phosphate-buffered saline for 5 min, cells were incubated with primary antibodies (anti-paxillin mAb or anti-vinculin mAb) for 1 h, followed by incubation with goat anti-mouse fluorescein isothiocyanate-conjugated secondary antibodies. Rhodamine-conjugated phalloidin was used to co-stain polymerized actin filaments. Slides were mounted with Vectashield (Vector, Burlingame, CA) and viewed using a Nikon fluorescence microscope or a Bio-Rad Laboratories laser scanning confocal microscope (model MRC-1024).

Immunoblotting—Cells were extracted with lysis buffer (50 mm Tris, pH 7.2, 500 mm NaCl, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 10 mm MgCl₂, and complete protease inhibitor mixture (Roche Molecular Biochemicals) and processed for SDS-PAGE after adjusting for equal protein loading, estimated by using the BCA protein assay kit (Pierce). After transfer to nitrocellulose membranes (Millipore Corp., Bedford, MA), proteins were probed with primary antibodies and secondary horseradish peroxidase-coupled antibody. Blots were developed by chemiluminescence using the ECL system (Amersham Biosciences). The band intensities were measured by densitometry using NIH Image software (Scion Corp., Frederick, MD).

Adhesion Assay—Cell adhesion was measured as described previously (28). Briefly, 96-well plates were coated with type I collagen (10 μg/ml) and laminin-5 (2.5 μg/ml) at 37 °C for 1 h, followed by blocking with 0.1% bovine serum albumin. HSC-3 cells were incubated with or without blocking mAbs to integrin subunits for 30 min at 4 °C, and 2 × 10⁴ cells were added to each well and allowed to attach for 20 min at 37 °C. Adherent cells were then quantified by a microcolorimetric assay (28).

Migration Assay—Time-lapse video microscopy was performed as described with a modification (29). Briefly, cells were seeded onto 6-well plates (Falcon, Becton Dickinson Labware) coated with different substrates (collagen I (10 μg/ml), laminin-5 (5 μg/ml), or PLL (10 μg/ml)) for 30 min. Plates were then examined in a Zeiss Axiovert inverted microscope with an X-Y scanning motorized stage (Carl Zeiss MicroImaging, Inc., Thornwood, NY) and maintained at 37 °C and 5% CO₂. Images were collected at the indicated time intervals using a SPOT-RT CCD camera (Molecular Dynamics) and analyzed with the Openlab software system (Improvision Inc., Lexington, MA). The positions of individual nuclei were tracked to determine the relative migration rates.

In the transwell migration assay, the undersides of the transwell (8-μm pore size; Corning Costar Corp., Cambridge, MA) were precoated with collagen (10 μg/ml) and laminin-5 (0.5 or 1.25 μg/ml). Next, 2 × 10⁵ cells were loaded onto the upper chamber of the transwell, and the lower chamber was filled with serum-free medium. Cells were incubated for 3 h at 37 °C, fixed with 4% paraformaldehyde, and stained with crystal violet. Non-migrating cells retained on the upper side were removed by wiping with a cotton swab. Cells that had migrated through the filter were counted and averaged from 10 randomly chosen microscopic fields using a ×20 objective. Migration was taken as 100% for cells infected with control virus.

Invasion Assay—Polymerized type I collagen gels were prepared by overlaying 30 μl of DMEM containing 2.4 mg/ml type I collagen to the upper chamber of each transwell and allowing gelation at 37 °C for 1 h. Next, HSC-3 cells (2 × 10⁵) in 200 μl of serum-free DMEM were added on top of the collagen gel. Serum-free medium was then added to the lower chamber and incubated for 24 h at 37 °C. Cells were fixed and stained with crystal violet. Collagen gel and associated cells were removed with a cotton swab. Cells that had penetrated the collagen I gel and reached the underside of the filter membrane were then counted in 10 randomly chosen microscopic fields using a ×20 objective. For each experimental condition, three invasion chambers were used. The mean ± S.E. were determined. Data were expressed as the percentage of treated migratory cells compared with that of control cells.

RhoA, Rac1, and Cdc42 Activity Assays—The construct expressing the Rho binding domain of ROCK fused to glutathione S-transferase (GST) was provided by M. A. Woodrow (University of California, San Francisco). The construct expressing PAK 75–132 fused to GST was a gift from P. N. Lowe (Medicines Research Center, GlaxoSmithKline, United Kingdom). Proteins were expressed in Escherichia coli BL21 and purified as described for pull-down assays (17).

Prior to the assay, cells were cultured under serum-free conditions overnight. Cells were plated on different substrates for the indicated times and were lysed in 500 μl of 50 mm Tris, pH 7.5, 150 mm NaCl, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 10 mm MgCl₂,1mm sodium vanadate, 1 mm NaF, 1 mm phenylmethylsulfonyl fluoride, 10 μg/ml aprotinin, and 2.5 μg/ml leupeptin. The lysates were then clarified by centrifugation. For estimating RhoA activity, 40 μg of cell lysates was used to measure total RhoA, and 1.5 mg of cell lysates was mixed with 30 μg of GST-ROCK and 90 μl of glutathione-agarose beads (Sigma) for 60 min at 4 °C. For estimating activity of Rac1 or Cdc42, 40 μg of cell lysates was used to measure total Rac1 or Cdc42, and 1.5 mg of cell lysates was mixed with 30 μg of GST-PAK and 90 μl of glutathione-agarose beads for 60 min at 4 °C. Beads were washed, and bound protein was eluted by boiling in Laemmli buffer. Samples were separated on 12% SDS-polyacrylamide gels, transferred to nitrocellulose, and then immunoblotted with anti-RhoA mAb, anti-Rac1 mAb, or anti-Cdc42 mAb.

Protein Kinase Assay—Serum-starved HSC-3 cells or WT PAK1-transduced cells were plated on collagen I or laminin-5 substrate for 30 min. Cell lysates were collected using lysis buffer (10 mm HEPES, pH 7.5, 150 mm NaCl, 1 mm EDTA, 0.6% Triton X-100, 20 mm glycerophosphate, 10% glycerol, 5 mm sodium fluoride, 1 mm sodium vanadate, 1 mm phenylmethylsulfonyl fluoride, 10 μg/ml leupeptin and aprotinin) as described by Royal et al. (30). Cell lysate (800 μg) was immunoprecipitated with anti-PAK1 (N-20) antibody for 1 h, followed by incubation with 40 μl of protein G-Sepharose (40% suspension) for a second 1-h incubation at 4 °C. The immune complex was washed twice with the lysis buffer and once with the kinase buffer (20 mm HEPES, pH 7.5, 20 mm MgCl₂, 1 mm EDTA, 20 mm glycerophosphate, 1 mm dithiothreitol, 5 mm sodium fluoride, 0.1 mm sodium vanadate, 1 mm phenylmethylsulfonyl fluoride, 10 μg/ml leupeptin and aprotinin). The immunoprecipitated PAK1 activity was assayed using MBP as a substrate. The kinase reaction was performed for 30 min at 30 °C in 50 μl of kinase buffer containing 5 μg of MBP, 100 μm cold ATP, and 6 μCi of [γ-³²P]ATP. The reaction was terminated by adding Laemmli sample buffer, followed by boiling. Proteins were separated by electrophoresis on a 12% SDS-polyacrylamide gel. Following autoradiography, band intensities were measured by densitometry. PAK1 expression levels were evaluated by Western blotting with the anti-PAK1 (N-20) antibody.

RESULTS

Enhanced Cell Spreading and Motility on Laminin-5 Substrates—Previous studies using cell aggregates had revealed that, compared with collagen, fibronectin, and laminin-1, laminin-5 had the most potential to disrupt cell-cell adhesions and promote cell scattering (24). We analyzed the effects of type I collagen and laminin-5 on single cell spreading and migration. Cells were able to attach but spread slowly on type I collagen over the 30-min time period (Fig. 1A). On laminin-5 substrates, cells spread rapidly and were nearly fully spread by 20 min after seeding, whereas cells on collagen I showed only a partial spreading during this time period. In addition, cells displayed an array of lamellopodia and microspikes on the laminin-5 substrate but on type I collagen cells showed only small lamellopodia (Fig. 1B). The time course of adhesion of cells to type I collagen and laminin-5 was indistinguishable (data not shown).

Next, the role of specific integrin receptors involved in mediating adhesion was determined using anti-integrin blocking antibodies. HSC-3 cells have significant levels of α2, α3, and α6 integrins and low levels of α5 and αv integrins (24, 31). Treatment of cells with anti-α2 or anti-β1 integrin antibody effectively blocked cell adhesion to type I collagen, whereas antibodies to α3 or α5 integrin was without effect. On laminin-5, treatment of cells with anti-α3 or anti-β1 integrin antibody inhibited cell adhesion (Fig. 1C). Interestingly, anti-α6 antibody had little inhibitory activity, suggesting that α6β1 and α6β4 integrins are not the dominant adhesion receptors for laminin-5 in these cells.

Formation of Focal Adhesions on Type I Collagen and Laminin-5—To establish the role of substrate adhesions in migration, we seeded cells onto the different ECM ligands and then evaluated focal adhesion formation by immunofluorescent staining. On type I collagen, we found that cells formed a high density of large, mature focal adhesions strongly stained by paxillin or vinculin antibodies (Fig. 1D). However, on laminin-5 substrate, paxillin- or vinculin-positive focal adhesions were limited in number and poorly stained. When cells were costained with rhodamine-phalloidin, an extensive array of polymerized actin and stress fibers was visible on collagen I but was minimal on laminin-5 (data not shown).

The ability of type I collagen and laminin-5 to promote cell migration was assessed using time-lapse video microscopy. Analysis of cell tracts revealed that although HSC-3 cells attached to type I collagen and the poly-l-lysine control, they migrated poorly on these substrates (Fig. 2A). In contrast, cells plated on laminin-5 showed a mostly random course of migration but moved at a rate severalfold greater than cells plated on type I collagen (Fig. 2, A and B). Another SCC cell line, UM-SCC-10A, responded similarly as HSC-3 cells to the two substrates (Fig. 2C). To evaluate the role of integrin receptors in HSC-3 cell migration, we performed motility assays in the presence of inhibitory antibodies. Blocking antibodies against integrin α3 or β1 most effectively inhibited cell movement on laminin-5, whereas blocking antibodies against integrin α2 and α6 had negligible effect on motility (Fig. 2D). Although α6β1 and α6β4 can act as receptors for laminin-5 (32) and are both expressed on HSC-3 cells (31), these integrins do not appear to be crucial for cell adhesion or migration. This indicates that α3β1 integrin is the primary receptor for adhesion to, and migration on, laminin-5. Importantly, these cells were mostly immobile on type I collagen but showed high rates of locomotion on laminin-5.

Time-lapse microscopy showed that cells seeded on collagen type I initially showed a partially spread morphology, and then slowly emitted lamellopodia around their peripheral edges. Over time the cells formed minimal membrane protrusions consisting of slowly forming and retracting lamellopodia and few filopodia (Fig. 2E, a). However, cells on laminin-5 substrate rapidly emitted a complex set of protrusions consisting of frontal filopodia followed closely behind by lamellopodia that correlated with cell forward movement (Fig. 2E, b). Typically, cells almost immediately began to locomote over the substrate and often generated a fan-like morphology or exhibited multiple large pseudopodia and shuffled forward. In both types of locomotion, filopodia and lamellopodia were frequently generated in the direction of movement (Fig. 2E, b). Immunofluorescent staining for α3 integrin showed the integrin concentrated at the lateral basal surface of the forward cell face and associated lamellopodia that colocalized with polymerized actin stained with rhodamine-phalloidin (Fig. 2E, d–f).

Collagen and Laminin-5 Differentially Regulate Rho GTPases—To gain insight into how the ECM may regulate cell motility, we investigated the potential role of Rho GTPases in this process. It is well established that the reorganization of actin cytoskeleton is associated with alterations in cell morphology and is regulated by Rho GTPases (2, 33). Therefore, we examined the activation of Rac1, RhoA, and Cdc42 in cells seeded for 30 min on type I collagen, laminin-5, or PLL substrates (Fig. 3A). Compared with cells in suspension, Rac1 was activated on type I collagen but less so on laminin-5 or the nonspecific poly-l-lysine control. RhoA activity in suspended control cells was found to be variable but tended to display significant activity. Others have reported that RhoA activity is transiently elevated following detachment of cells (34). Adhesion to collagen induced significant RhoA activation, whereas cells on laminin-5 generated only a low level of activation compared with control. In suspended cells, Cdc42 showed a relatively low level of activation, but adhesion to laminin-5 induced high activity with lower levels on collagen type I and poly-l-lysine.

The enhanced motility on laminin-5 appeared to correlate with low RhoA activation and high Cdc42 activation. Analysis of RhoA GTPases activity on laminin-5 over the long term revealed a consistent pattern. In the case of Rac1, a significant level of activation was detected that remained fairly constant over the time period (Fig. 3B). For RhoA activity, again suspended cells had a high basal level but, following seeding on laminin-5, RhoA activity sharply decreased and remained at this basal level for 3 h. Cdc42 activity progressively intensified with increasing seeding time on the laminin-5 substrate. This stimulation was evidenced as early as 15 min after plating and remained high even after 180 min.

We further compared RhoA activation by collagen and laminin-5 substrates. In suspension cells plated on type I collagen RhoA activity was prominently increased compared with control cells after 10 min, and then decreased somewhat by 30 min (Fig. 4A). This increase in RhoA activity correlates with stress fiber formation on collagen I substrate (Fig. 1D). However, as before, RhoA activity remained at only modest levels when cells were plated on laminin-5 (Fig. 4B). To confirm that laminin-5 was responsible for the suppression of RhoA activity, we included a blocking laminin-5 antibody (P3H9-2) during the assay to interfere with adhesion to laminin-5. RhoA was inactivated in cells with or without the control antibody treatment but, in the presence of blocking laminin-5 antibody, the suppression of RhoA activity was released (Fig. 4C). This is consistent with the finding that suggests that adhesion to laminin-5 suppresses RhoA activation.

In the next set of experiments, we identified which integrins may be modulating RhoA activity by examining the effects of artificial integrin engagement. To test the effects of integrin engagement and cross-linking on RhoA activity, we used specific anti-integrin antibodies immobilized on the plate surface or added to cells in suspension. In the first approach, cells were seeded on plates coated with immobilized anti-α2 or anti-α3 mAb to cross-link specific integrins, and were then assessed for RhoA activity. Seeding cells onto immobilized anti-α2 integrin antibodies led to enhanced levels of RhoA activation that gradually increased up to 90 min (Fig. 4D). This increase in RhoA activity generally mimicked the activation seen in cells adherent to collagen but the induction was slower compared with cells adhering to collagen substrate (Fig. 4A). In contrast, when similar experiments were performed to engage α3 integrin with anti-integrin mAb, a progressive decrease in RhoA activity over this time period was observed (Fig. 4E), similar to the response of cells seeded on laminin-5. In the second approach, we directly tested the role of specific integrins in RhoA activation by inducing integrin cross-linking in suspended cells using an antibody-mediated clustering assay. HSC-3 cells in suspension culture were first treated with the anti-integrin mAb, followed by incubation with a secondary cross-linking antibody. Treatment of suspended cells with anti-α2 mAb lead to a strong activation of RhoA (Fig. 4F). However, treatment with anti-α3 mAb or control antibodies failed to lead to an increase (Fig. 4F). These findings show that α2 integrin, but not α3 integrin, when artificially ligated and cross-linked with specific antibody, will induce the activation of RhoA. Overall, these data are consistent with the idea that adhesion to collagen induces α2 integrin signaling and RhoA activation, whereas laminin-5/α3 integrin signaling induces suppression of RhoA activity.

Suppression of RhoA Facilitates Cell Migration—To determine whether elevated RhoA activity is associated with the poor spreading and migration observed on type I collagen, we infected cells with adenovirus encoding hemagglutinin-tagged constitutively active V14RhoA or dominant-negative N19RhoA. The expression of exogenous RhoA was recognized by anti-hemagglutinin mAb (Fig. 5A). Inhibiting RhoA by expression of N19RhoA caused a nearly 200% increase in haptotaxic migration toward type I collagen compared with a partial reduction caused by expression of V14RhoA (Fig. 5B). Because activity assays show that laminin-5 suppresses RhoA activity, we examined the effects of RhoA on cell migration toward laminin-5. Cells were infected with adenovirus encoding V14RhoA or N19RhoA and subjected to a transwell migration assay. Neither expression of dominant-negative N19RhoA nor expression of constitutively active V14RhoA enhanced cell migration toward immobilized laminin-5, but both led to a slight suppression of migration (Fig. 5C). Moreover, time-lapse video microscopy revealed that the Rho inhibitor exoenzyme C3, dominant-negative N19RhoA, and ROCK inhibitor Y-27632 all resulted in cell migration rates 2–3-fold greater on type I collagen compared with controls (Fig. 5D). Thus, treatments that inhibit the RhoA pathway uniformly stimulated motility on collagen substrates.

To investigate the involvement of Rho in tumor cell invasion, we used the collagen gel assay in which cells were seeded onto a polymerized three-dimensional type I collagen matrix and then evaluated for penetration of the ECM barrier. Analysis revealed that in cells expressing the dominant-negative N19RhoA, invasion was substantially increased. In contrast, cells that expressed the constitutively active V14RhoA displayed a 75% inhibition of invasion (Fig. 5E). The results were consistent with those observed for migration on collagen substrates and confirmed that RhoA suppressed cell locomotion.

To evaluate the importance of downstream effectors of Rho, we tested the effect of the ROCK inhibitor, Y-27632, on focal adhesion formation using immunofluorescent staining of cells plated on collagen type I substrates (Fig. 5F). As before, the HSC-3 cells typically form an array of mature focal adhesions on collagen (Fig. 1D). We found that in the presence of Y-27632, the intensity of paxillin and vinculin staining at focal adhesions decreased dramatically and was replaced with small punctate adhesions. Thus, inhibiting RhoA and ROCK activity appeared to decrease focal adhesion assembly and stabilization, which may contribute to enhanced motility.

Cdc42 Contributes to PAK1 Activation and Cell Migration— Laminin-5 induced a progressive increase in Cdc42 activation (Fig. 3B). When cells were plated onto laminin-5 in the presence of blocking laminin-5 antibody, the active level of Cdc42 was significantly reduced, whereas, control antibody had no effect (Fig. 6A). We next evaluated the potential of Cdc42 to mediate motility on laminin-5 substrates. Cells were infected with N17Cdc42 adenovirus and tested for cell migration using time-lapse microscopy. Cells expressing N17Cdc42 exhibited a significant reduction in cell migration rate, whereas treatment with the ROCK inhibitor Y-27632 had little effect on cell movement (Fig. 6B). However, migration was only partially suppressed by N17Cdc42, suggesting that other pathways that include Rac1 are most likely involved. Among the PAK family of serine/threonine kinases, PAK1 is a major effector protein of Cdc42; therefore, we measured PAK1 activation on different substrates using an in vitro kinase assay (Fig. 6C). Compared with cells attached to collagen type I or suspension control cells, laminin-5-adherent cells expressed more than double the level of PAK1 activity using MBP as substrate. Also, PAK1 was heavily autophosphorylated in cells seeded onto laminin-5 substrate but the level of phosphorylation was low for control cells in suspension or for cells seeded on collagen. This was confirmed using a specific phosphoantibody: PAK1 was observed to be heavily phosphorylated at threonine 423 in cells seeded on laminin-5 substrate compared with cells in suspension or attached to collagen type I (Fig. 6D).

Phosphorylation at threonine 423 has been shown to be strongly correlated with PAK1 activity (35). To determine whether Cdc42 was required for PAK1 activation in cells seeded on laminin-5, cells were infected with adenovirus encoding Myc-tagged dominant-negative N17Cdc42 and used for the in vitro kinase assay. Compared with control cells, expression of N17Cdc42 suppressed both kinase activity for MBP substrate and autophosphorylation of PAK1 (Fig. 6E). These data indicate that a significant portion of PAK1 activity in cells adhering to laminin-5 is induced by activation of Cdc42.

To determine whether kinase activity of PAK1 was required for motility, we infected cells with adenovirus encoding GFP, Myc-tagged wild type PAK1 (wt PAK1), or Myc-tagged constitutively active (E423, Glu⁴²³) PAK1. The cells were then assessed for their cell migration on collagen type I or laminin-5. Levels of the expression for the Myc-tagged transgenes are shown in Fig. 7A. In motility assays, the constitutively active mutant, Glu⁴²³ PAK1, increased migration toward immobilized type I collagen by more than 2-fold compared with the control cells. On laminin-5 substrates, Glu⁴²³ PAK1 increased migration by nearly 3-fold compared with the control (Fig. 7, B and C). In contrast, expression of wt PAK1 failed to enhance motility toward type I collagen compared with the control (Fig. 7B). But on laminin-5, expression of wt PAK1 increased migration by more than 2-fold compared with the control (Fig. 7C). Finally, we tested autophosphorylation and kinase activity of PAK1 in cells overexpressing wt PAK1. As expected, adhesion to laminin-5 resulted in strong autophosphorylation of PAK1 and elevated kinase activity of PAK1 (Fig. 7D). On collagen substrate, minimal activation of PAK1 was observed. This suggests that even in the presence of excess PAK1, adhesion to collagen is not sufficient to generate high PAK1 activity. Only on laminin-5 was overexpressed PAK1 activated and this correlated with enhanced motility. Overall, these data suggest that autophosphorylation and kinase activity of PAK1 by Cdc42 and Rac are required for sustained and prolonged migration.

DISCUSSION

ECM proteins are able to stimulate or constrain cell movement depending on a number of factors, including substratum adhesiveness, integrin receptor activation, and cytoskeletal coordination of locomotion (4, 36–38). Studies over the last several years have demonstrated that ECM regulates the activities of Rho GTPases which, in turn, modulates cell migration (2, 17, 19, 20, 39). Motility involves a set of complex sequential events that include disruption of focal adhesions, lessening of cell substrate adhesion, formation of nascent adhesions with new protrusions and, finally, polymerization of actin cytoskeleton. The data we present provide additional evidence that Rho GTPases serve as a controlling platform for migration regulated by ECM-integrin signaling. Importantly, our results demonstrate that specific ligand-integrin binding results in unique patterns of active Rho GTPases. It is particularly interesting that certain integrins can promote migration while others hinder locomotion. RhoA/ROCK and Cdc42/PAK1 pathways appear to be key mediators in the signaling pathways that control cell motility in squamous cell carcinoma cells.

We show that in squamous epithelial cells, collagen I/α2 integrin adhesive signaling restrains migration via activation of the RhoA/ROCK pathway. RhoA can influence cell migration by modulating various activities that include inducing polymerization of actin filaments and detachment at the cell rear. We found a strong correlation between RhoA activity and the formation of mature focal adhesions on type I collagen. Moreover, the number and size of focal adhesions were decreased by inhibiting ROCK activity. Previous studies have suggested that high levels of adhesiveness inhibit speed of cell migration (36–38) and de-adhesion is the rate-limiting step in regulating cell migration under a condition of high cell-substratum adhesiveness (1, 40, 41). Similarly, cells lacking the tyrosine kinases FAK or Src form more extensive adhesions, and these cells migrate poorly (42–45). It is reasonable to conclude that the high density of mature focal adhesions formed on type I collagen substrate play a role in negatively modulating epithelial cell motility by strengthening immobilization.

Another function of RhoA/ROCK signaling is to negatively regulate membrane protrusive activities (46). It is believed that Rho controls both contraction and retraction events and works in concert with the generation of protrusive structures during migration. In some cases RhoA activity was found to be necessary for migration (47), whereas other studies show that high levels of Rho slow forward locomotion (6). These differences in response may be related to the complexity and biphasic process of initial adhesion to the substrate followed by commencement of cell migration (23). The squamous epithelial cells used in the present study showed a slow rate of spreading on type I collagen, and, once spread, most cells remained immobile but did slowly emit small lamellipodial projections. ROCK inhibitor Y-27632 promoted formation of membrane protrusions and resulted in increased cell motility on collagen substrates. Furthermore, constitutively active V14RhoA-infected cells displayed reduced protrusive structures (data not shown). In the case of laminin-5 substrates, dominant negative RhoA produced a significant inhibition of migration. This effect presumably reflects the need for some level of active RhoA activity that is required for cell body contraction and tail retraction that occurs during the repetitive cycle of cell locomotion (6).

In other studies, RhoA and ROCK activation limited membrane protrusions in monocytic THP-1 cells (46) and in leukocytes (48), apparently by regulating myosin-dependent contractile force. Actin-myosin II-mediated contraction may inhibit membrane protrusion formation. Reducing myosin II activity increases the rate of spreading (49), suggesting that increasing cell rigidity by myosin activation represses cell extension driven by actin polymerization. Enhanced cell spreading has also been observed by inhibiting RhoA via p190 RhoGAP in fibroblasts (15) or truncated RhoA in endothelial cells (50, 51). However, neither infection with N19RhoA nor treatment with Y-27632 was sufficient to promote SCC cell spreading on type I collagen to the same extent as seen on laminin-5 (data not shown).

Several studies have reported that RhoA is inactivated by signaling from certain ECM proteins such as tenascin (52) and ECM receptors such as β1 integrin (8, 53). We provide evidence that laminin-5/α3 integrin signaling also inactivates RhoA. Unlike transient inactivation of RhoA by fibronectin (17), laminin-5/α3 integrin signaling maintained inactivation of RhoA for a sustained period of time (Fig. 3B). However, our results differ from those of an earlier study by Nguyen et al. (54) in which laminin-5 or binding of primary skin keratinocytes through α3 integrin to immobilized mAb activated RhoA. Differences in these two studies may be related to the different cell types used or to contrasting experimental conditions. In other studies, α6β4 integrin and adhesion to laminin 1 in Clone A carcinoma cells induced Rho activation, whereas β1 receptors and adhesion to collagen type caused inhibition of Rho activity (8). It is clear, however, that depending on cell type, different integrins can elicit variable responses in Rho activation.

An important issue is how specific integrins during adhesion lead to differential regulation of individual Rho family proteins. Most likely there are multiple regulatory mechanisms for Rho GTPases that lead to diverse responses. Cellular interaction with the ECM proceeds in a set of sequential events that includes initial adhesion to the substrate followed by spreading and migration. This spatio-temporal relationship is reflected in how Rho proteins are recruited to adhesion contacts such as focal adhesions and influence cell behavior. Additionally, focal adhesions are important signaling devices that act as staging points for signal transduction initiated by FAK and related molecules. Recent studies have shown that one mechanism by which integrins can regulate Rho proteins is by inducing their selective translocation to plasma membrane lipid raft domains that target active Rho and Rac along with their respective effectors, mDia and PAK (55, 56). It is possible that specific integrins are preferentially localized at these lipid rafts and provide unique scaffolding for Rho family member signaling that is dependent on cell type and available ECM ligand.

It has been suggested that variability of integrin response is related to the cell-type specific expression and function of guanine nucleotide exchange factors and GTPase activating proteins (GAPs) (23). For example, the p130^Cas-CrkII-DOCK 180 complex is responsible for Rac activation after α3 integrin engagement (13). Similarly, certain integrins induce p190 RhoGAP leading to RhoA inactivation and increased cell protrusion (15, 53). Other potential signaling pathways may control Rho protein function. It is well known that integrins activate Src tyrosine kinases along with FAK (57) and evidence indicates that integrin signaling can lead to RhoA activation (but not cdc42 or Rac1) in a Src-dependent manner (53). Less is known about how Cdc42 is activated after integrin engagement but its activity is crucial for maintaining cell polarity. Recent studies suggest that PIX acts as a guanine nucleotide exchange factor for both Cdc42 and Rac (58, 59). Interaction of αPIX with β-parvin suggests an involvement of αPIX in integrin-mediated signaling (60). It is likely that the ability of certain integrins to preferentially activate one Rho family member will define a dominant cell phenotype that is distinct from that triggered by other integrin receptors.

Rho family members Cdc42 and Rac1 are known to play a critical role in membrane protrusive activity by means of actin reorganization (61, 62). Rac1 mediates lamellopodia formation, whereas Cdc42 triggers generation of filopodia (11). We found that increased Cdc42 activity was correlated to elevated PAK1 activity in SCC cells, and suggest that Cdc42/PAK1 is the major signaling pathway that may promote filopodial and, indirectly, lamellipodial extension. Cells attached to type I collagen exhibited Rac1 activation but low Cdc42/PAK1 activation, and spread more slowly. By contrast, cells attached to laminin-5 showed moderate Rac1 activity and high Cdc42/PAK1 activity and spread quickly. It has been reported that inhibiting PAK1 and Cdc42 but not Rac1 inhibits spreading of fibroblasts (19). In addition, constitutively active PAK1 was sufficient to enhance cell movement toward immobilized type I collagen.

Although both type I collagen and laminin-5 activate Rac1, cells move poorly on type I collagen. However, α3β1 integrin binding to laminin-5 directs the stabilization of polarized lamellipodium in keratinocytes through activation of Rac1 (63). On the other hand, the laminin-10/α3β1 integrin-dependent pathway has been shown to preferentially activate Rac through a p130-CrkII-DOCK180 complex, thereby promoting lung adenocarcinoma cell migration (13). Bourne and collaborators (64) recently reported that phosphatidylinositol 3,4,5-trisphosphate-dependent Rac1 signaling is important in mobilizing the leading edge of the cell. Thus, it appears that Rac1 may participate cooperatively with Cdc42 to promote protrusive activity. In the case of squamous cell motility on laminin-5, Cdc42 appears to play a dominant role, but Rac1 must also be important.

During migration on laminin 5 substrates, many cells tend to move in a random walking fashion or in a polarized fan cell morphology where first exploratory filopodia followed by lamellopodia are emitted in the direction of movement. Presumably Rac1 and Cdc42 are needed for this two-step sustained movement. On collagen substrates, lamellopodia are slowly elaborated but these protrusions are transient and form in many directions so the cells tend to remain immobilized with little net movement. Because lamellopodia formation is believed to be primarily controlled by Rac, these results are consistent with the finding that Rac1 activity in cells is significant on both substrates. The sustained activation of Cdc42 on laminin 5 appears to be important for progressive and persistent migration.

Both morphological polarity and directional sensing allow for efficient motility. It is established that Cdc42 is required for polarity. Early work indicated that in some cells (e.g. macrophage (47)), inhibiting Cdc42 does not block migration but in others Cdc42 is needed for migration and invasion (65). Inhibiting Cdc42 does not block actin polymerization and lamellae formation, but impairs polarity and motility (3, 64). In addition, like the phosphatidylinositol 3-kinse pathway, the Cdc42/PAK1 pathway is part of the “molecular compass” required to sense direction (66). Thus, Cdc42/PAK1 appears to be important in sampling external environmental cues, initiating polarized actin polymerization, and controlling direction of movement. In the case of HSC-3 cells on collagen, lamellopodia were observed but filopodia were not common (Fig. 2E). For cells on laminin-5, filopodia were present that seemed to probe the surrounding substrate and this was usually followed by forward movement. On collagen substrates, the normally quiescent cells when treated with ROCK inhibitor (Y-27632) were able to emit multiple lamellipodia in different directions (not shown), indicating that they still expressed this capacity for protrusive activity and movement when RhoA activity is suppressed.

The α3 and α6 integrins are known to recognize laminin-5 as well as other substrates (67, 68). α3β1 Integrin plays a major role in both cell adhesion and migration on laminin-5 (24, 25, 32, 69–71). Our findings are consistent with these previous reports. Unlike α3β1 integrin, α6-containing integrins appear to play a less important role in oral SCC cell migration on laminin-5. However, α6β4 integrin has been shown to stimulate chemotaxis, but has no influence on haptotaxis in breast carcinoma cells (72). Previous studies have suggested that in keratinocytes, α6β4 integrin may restrain motility but is not needed for migration (32). It is possible that in SCC cells, α6β4 integrin does not play a dominant role in ECM-mediated motility, but additional studies are clearly needed.

Some evidence suggests that the RhoA/ROCK pathway plays an important role in invasion and metastasis in various tumors (73–76). For example, ROCK inhibitors Y-27632 (9, 77) and Wf-536 (74) have suppressed invasion of hepatoma and melanoma cells. However, our findings indicate that inactivating ROCK promotes oral SCC cell migration and invasion, suggesting that ROCK inhibitors may facilitate invasion of head and neck SCCs. Similarly, the importance of RhoA in migration has been shown to vary among cell types. An active RhoA mutant was shown to inhibit motility of fibroblasts and lung adenocarcinoma cells (13, 15) but to enhance motility of lymphoma cells, colon carcinoma cells, and hepatoma cells (7–9). Recently it was reported that the morphological phenotype of the cell can influence how RhoA regulates motility. Sahai and Marshall (78) showed that, whereas elongated cells do not require Rho/ROCK signaling and ezrin function for motility, rounded cells using bleb-like extensions for motility do require Rho/ROCK signaling for movement.

In summary, our findings support the idea that different ECM substrates act as environmental cues controlling cell movement through specific integrin-mediated signaling pathways involving Rho GTPases. In the case of squamous epithelial cells, adhesion to type I collagen leads to immobilization of cells. In part, restraint of migration was related to the strong activation of RhoA and induction of ROCK signaling. Consequently, these cells form an extensive actin array and stable focal adhesions that negatively impact cell locomotion. Also important on collagen substrates are the reduced Cdc42 signaling and limited activation of PAK1 kinase levels that seem to compromise protrusive filopodial and lamellipodial activity and lead to reduced polarized migration. In contrast, laminin-5 substrates suppress RhoA and strongly activate Cdc42, leading to a robust motile response. Thus, the functionality of integrin receptors and partner ligands together can selectively modulate Rho family members and control how different tissue cell types respond to their unique ECM microenvironment.