Rho GTPases Control Polarity, Protrusion, and Adhesion during Cell Movement

LY294002 (Calbiochem) was prepared as a stock solution (1,000×) of 20 mM in DMSO and frozen at −20°C. PD98059, 2-(2′-amino-3′methoxyflavone), was purchased from Calbiochem and a stock solution (30 mM; 500×) was prepared in DMSO and frozen at −20°C. The effectiveness of PD98059 was confirmed by its ability to inhibit wound-induced phosphoERK activation. Cycloheximide was prepared as a stock solution (1,000×) of 100 mg/ml in DMSO. The myosin inhibitor 2,3-butanedione monoxime (BDM; Sigma Chemical Co.) was prepared as a stock solution of 0.5 M in H₂O and was added to a final concentration of 15 mM and was readded after 3 h. The effectiveness of BDM was assessed by its ability to disrupt the organization of myosin II in stress fibers. The p160ROCK inhibitor, Y-27632 (Uehata et al., 1997; Hirose et al., 1998), was prepared as a stock solution of 3 mM in DMSO. The effectiveness of Y-27632 was assessed by its ability to disrupt actin stress fibers and focal adhesions.

Primary REF cells were provided by Dr. Durward Lawson (University College London, London, UK). Cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, streptomycin, and penicillin at 37°C in 10% CO₂. Wound-induced cell migration assays were performed on secondary cells between passages 3 and 7.

Recombinant V14Rho (RhoA isotype), V12Rac (Rac1 isotype), V12Cdc42 (G25K isotype), C3 transferase, N17Rac, N17Cdc42, and WASp (Cdc42 binding fragment comprising amino acids 201–321) were expressed as glutathione S-transferase fusion proteins in Escherichia coli and purified on glutathione-agarose beads essentially as described in Self and Hall (1995). The proteins were released from the beads by thrombin cleavage and dialyzed against microinjection buffer (50 mM Tris, pH 7.5, 50 mM NaCl, 5 mM MgCl₂, 0.1 mM DTT) and concentrated as required. For the GTP-binding proteins, active protein concentrations were determined by filter binding assay using [³H]GDP of [³H]GTP as described previously (Self and Hall, 1995). N17Cdc42 has a low affinity for [³H]GDP (Self and Hall, 1995) and therefore the protein concentration for N17Cdc42 was estimated using a protein assay kit (Bio-Rad Laboratories). The protein concentrations of C3 transferase and the WASp fragment were also determined by this method. Purified neutralizing antibody to Ras, Y13-259, was a kind gift of Dr. Hugh Paterson (Chester Beatty Laboratories, London, UK).

REFs for wound assays were seeded at a high density, 12 × 10⁴ cells, on 13-mm glass coverslips, and wounded 1 d later when the cells formed a confluent monolayer. The wound was made by scraping a microinjection needle (broken to its shaft and flame polished) through/across the cell monolayer. The wound width was consistently between 100 and 130 μm and wounds reproducibly took between 5 and 6 h to close. Cells were pretreated with inhibitors for 20 min or, in the case of Y-27632, 1 h before wounding. Since most wound edge cells round up immediately upon wounding and thus are difficult to inject, wounds were left for 1 h to allow cell respreading and to facilitate microinjection. Proteins were injected into the cell cytoplasm along with a marker protein (either FITC- or Texas red–conjugated, lysinated dextrans at 2 mg/ml). Recombinant proteins were microinjected at concentrations as indicated in the text. The neutralizing anti-Ras antibody, Y13-259, was microinjected at a concentration of 8–9 mg/ml. Expression vectors (pRK5-myc) encoding N17Rac1, N17Cdc42 (G25K isoform), WASp fragment, and V12HaRas were injected into the cell nucleus at a concentration of 200 μg/ml in PBSA and expressed protein was visualized using anti-myc antibodies (9E10) or in the case of Ras with the rat monoclonal antibody, Y13-238. Previously we have shown that at least 90% of DNA-injected cells express the pRK5 construct (Lamarche et al., 1996) and myc-tagged protein could be detected by 30 min after cell injection.

For each experimental run, control wounds were fixed soon after the wound was made (for inhibitor experiments), 1 h after wounding (for microinjection experiments) and immediately after the wound edges have met as monitored by frequent observation using phase-contrast optics. Experimental wounds were fixed at the time that control wounds had closed. For wound closure measurements, cells were stained for filamentous actin as previously described (Nobes and Hall, 1995). In brief, cells were fixed in 4% paraformaldehyde/1% glutaraldehyde/PBS (in order to preserve fine actin structures such as filopodia), permeabilized in 0.2% Triton X-100/PBS, blocked with sodium borohydride (0.5 mg/ml) in PBS, and stained with rhodamine-conjugated phalloidin (0.1 μg/ml) in PBS. Cells for immunostaining were fixed with 4% paraformaldehyde/PBS, permeabilized, blocked with sodium borohydride, and incubated with primary antibodies diluted in PBS for 1 h at room temperature. Cells were immunostained to reveal vinculin, phosphotyrosine, and myc-tag as described previously (Nobes and Hall, 1995; Lamarche et al., 1996). To reveal c-fos, we used a rabbit polyclonal fos antibody (Oncogene Science Inc.) diluted 1:100 followed by FITC-conjugated goat anti–rabbit (1:200; Pierce and Warriner) and a tertiary layer of FITC-conjugated donkey anti–goat (1:200; Jackson ImmunoResearch Labs, Inc.). Dually phosphorylated ERK-1 and -2 were detected using monoclonal anti–MAP kinase, activated (Sigma Chemical Co.; M8159) and diluted 1:50, followed by FITC-conjugated goat anti–mouse (1:200; Pierce and Warriner) and a tertiary layer of FITC-conjugated donkey anti–goat. Cells were stained for myosin II using affinity-purified rabbit anti–human platelet nonmuscle myosin II antibody (Biomedical Technologies, Inc.) diluted 1:100. To detect expressed Ras protein, cells were permeabilized using 0.1% saponin/80 mM potassium Pipes (pH 6.8)/1 mM MgCl₂/5 mM EGTA for 2 min before fixation with 4% paraformaldehyde and incubation with rat monoclonal antibodies, Y13-238 (a kind gift of Mike Olson, Chester Beatty Laboratories), followed by Cy3-conjugated donkey anti–rat antibodies (Jackson ImmunoResearch Labs, Inc.). Injected cells were detected with Cascade blue avidin to biotin dextran which was coinjected with the Ras expression vector. Coverslips were mounted by inverting them onto 5 μl movial mountant containing p-phenylenediamine (a few grains/5 ml) as an antifade agent. The coverslips were examined on a Zeiss axiophot microscope using Zeiss 10× and 20× air lenses, and 40× and 63 × 1.4 oil immersion lenses. Fluorescence images were recorded on Kodak T-MAX 400 ASA film.

Semiquantitative measurements were made of control wounds (t = 0) and of control and inhibitor-treated wounds (after the control wound had closed). Three randomly chosen regions of a single wound (each 700 μm long), stained with rhodamine-conjugated phalloidin, were photographed at a magnification of 20. A mean wound width in micrometers was determined (by averaging the width every 30 μm) and an average percent wound closure was calculated.

For microinjection experiments, a region of wound was randomly chosen and a continuous stretch of 10–15 leading edge cells on opposing sides of the wound was microinjected. Since not only front row cells participate in the wound response, we also injected cells in rows 2, 3, and 4 on both sides of the wound. For microinjected wounds, wound width was measured as the distance between microinjected cells (on either side of the wound) visualized by coinjected fluorescent dextran or by anti-myc staining, since some noninjected cells often crawl between injected cells into the wound. In this case, measurements were made every 15 μm over a length of wound of ∼300–400 μm. From the average values for each injection, percent wound closure was calculated.

To record the position of the Golgi apparatus in migrating wound edge cells, wider wounds (∼200 μm wide) were made to monolayers of REFs and cells were fixed 2, 4, and 6 h after wounding. The Golgi apparatus was localized by immunolabeling, as described above, using a rat monoclonal antibody (23C) which recognizes β-COP. The position of the immunolabeled Golgi apparatus in each cell was recorded as described in Kupfer et al. (1982). Wound edge cells were divided up into three 120° sectors centering on the nucleus, one of which faced the edge of the wound (see Fig. 5 A). Cells in which the Golgi apparatus (i.e., 50% or more of the fluorescent image) was within the sector facing the wound were scored positive and for each time point at least 100 cells were examined. For the zero time point the random value of 33.3% of cells showing Golgi apparatus in the forward facing sector was given. To examine the effect of mutant GTPases on Golgi apparatus reorientation, wounds were injected with expression vectors encoding myc-tagged N17Rac, N17Cdc42, WASp, and V12Cdc42 1 h after the wound was made. Expressed myc-tagged proteins were detected 1 h later when 50% of cells repositioned their Golgi apparatus to the forward facing sector. The cells were fixed 5 h later and Golgi apparatus position was recorded in wound edge cells expressing myc-tag.

A CCD camera and time-lapse controller (EOS Electronics) were attached to a Zeiss inverted microscope. Injected or inhibitor-treated cells were placed in a slide flask (Cat. no. 170920; Nunc). Microscope images were collected at a rate of 10 frames every 60 s on a Sony betacam video recorder. Individual frames were transferred from videotape to a Macintosh computer with a frame grabber and processed using Adobe Photoshop. After time-lapse, cells were fixed and stained to identify microinjected cells.

To investigate the role of small GTPases in fibroblast motility, small scrape wounds were made across a confluent monolayer of primary REFs (Fig. 1 A). The wound width is approximately two to three cells across (mean 120 μm) and during a period of 5–6 h, cells move bodily forwards to fill the gap without the necessity for cell division (Fig. 1, B and D). The response to wounding is rapid; within 30 min, rounded and retracted cells at the wound edge respread and extend filopodia and lamellipodia into the open space. 2 h after wounding, front row cells have clearly developed a polarized morphology, with lamellipodia and membrane ruffling localized only at the leading edge and not at the sides or rear edge of cells, where they are in contact with neighboring cells (Fig. 1 C). The wound edge cells remain fairly tightly packed (Fig. 1 C) and adherens junctions, visualized with cadherin antibodies, remain intact during the course of the assay (data not shown). Rhodamine-conjugated phalloidin reveals that all cells in the monolayer contain abundant stress fibers, but in addition, cells at the leading edge also have actin-rich lamellipodial and filopodial extensions (Fig. 1 E). Antiphosphotyrosine antibodies reveal elongated focal adhesions in all cells and cells at the leading edge also have smaller focal complexes intimately associated with lamellipodia and filopodia (see arrow in Fig. 1 F). After ∼5 h, as opposing cells approach each other, time-lapse cinematography reveals that cell ruffling ceases within minutes of cell–cell contact (data not shown).

Wounding results in the rapid induction of various transcription factors such as c-fos and Egr-1 in cells both at the wound edge and several rows back into the monolayer (Verrier et al., 1986; Martin and Nobes, 1992; Pawar et al., 1995). We observe c-fos upregulation in four or five rows of cells back from the leading edge (data not shown). However, inhibition of protein synthesis, by pretreatment of the cells with cycloheximide (100 μg/ml), has no effect on wound closure (percent wound closure is 90 ± 3.6; control = 96 ± 2.9%) and we conclude that changes in gene transcription and protein synthesis are not required for forward movement in this short-term assay.

Cell motility is generally associated with the protrusion of two types of actin-rich structure, lamellipodia and filopodia, at the leading edge of cells and in Swiss 3T3 fibroblasts, these are controlled by Rac and Cdc42, respectively (Ridley et al., 1992; Kozma et al., 1995; Nobes and Hall, 1995). To determine whether Rac plays a role in wound-induced cell motility, patches of cells on opposite sides of a wound were microinjected with either dominant negative Rac (N17Rac) protein, or with an expression vector encoding N17Rac. Rhodamine-conjugated phalloidin staining revealed that inhibition of Rac blocked all lamellipodial activity (Fig. 2, compare B with A). Cells were left with a “scalloped” shape, although filopodia could still be seen (Fig. 2 B, arrow). Inhibition of Rac had severe effects on cell movement (Fig. 2, C and D) and quantitative analysis (Fig. 3) revealed that in cells injected with N17Rac protein wound closure was blocked by ∼80%, whereas in cells injected with the N17Rac expression construct cell movement was inhibited by 98%. The difference in these values likely reflects the instability of N17Rac protein (half-life ∼3 h; Ridley et al., 1992). Microinjection of constitutively active Rac protein (V12Rac) had no inhibitory effect on cell movement (Fig. 3). We conclude that Rac is activated upon wounding and that Rac activity is essential for cell movement.

To determine whether Cdc42 also plays a role in wound closure, patches of cells on opposite sides of a wound were injected with either dominant negative Cdc42 (N17Cdc42) protein, or with an expression vector encoding N17Cdc42. It can be seen from Fig. 3 that inhibition of Cdc42 results in a 50% inhibition of cell movement. Microinjection of cells with the Cdc42-binding domain of the Cdc42 target protein, WASp, resulted in a similar level of inhibition (Fig. 3), confirming that Cdc42 is required for efficient movement but is not absolutely essential.

To examine the role played by Cdc42 during wound closure, we looked more carefully at migrating cells lacking Cdc42 activity. Fig. 4 A shows the typical polarized morphology of a wound edge cell (visualized with injected fluorescent dextran) with ruffling activity restricted to the front edge. In contrast, Fig. 4 B shows the morphology of a wound edge cell in which Cdc42 has been inhibited. Cell polarity is completely lost and protrusive, lamellipodial activity can be seen all around the cell's periphery. Expression of constitutively active Cdc42 (V12Cdc42) protein had no effect on cell polarity (Fig. 4 C) or forward movement (Fig. 3).

The polarized morphology of migrating cells also involves the reorganization of the microtubular network and the alignment of the Golgi apparatus in the direction of movement (Gotlieb et al., 1981, 1983; Kupfer et al., 1982; Euteneuer and Schliwa, 1992; Bershadsky and Futerman, 1994). Although no major reorganization of the microtubular network could be seen during the REF wound healing assay (data not shown), disruption of microtubules with nocodazole did inhibit wound closure by 60 ± 6% (n = 3). Realignment of the Golgi apparatus, on the other hand, could be clearly observed during the assay. As seen in Fig. 5, A and B, during the course of the experiment the percentage of cells with their Golgi apparatus orientated in the direction of movement rose from 33% (essentially random) to >80%. Although the relatively long time course suggests that this realignment is not essential for movement, it may facilitate movement and is a clear read-out of a developing polarized morphology. To determine whether Cdc42 activity is required for this aspect of cell polarity, Golgi apparatus reorientation was measured in wound edge cells after injection with dominant negative Cdc42 or with the WASp fragment. Fig. 5 (C–E) shows that inhibition of Cdc42 completely prevents Golgi apparatus realignment. Wound-induced Golgi apparatus reorientation was not inhibited by constitutively active Cdc42 (V12Cdc42) or by dominant negative Rac (N17Rac) (Fig. 5 E).

In Swiss 3T3 fibroblasts, Rho is required for both the formation and maintenance of actin stress fibers and focal adhesions (Ridley and Hall, 1992; Hotchin and Hall, 1995). REF monolayers have abundant stress fibers, and after wounding motile edge cells continue to display many actin stress fibers oriented both towards the wound and radially across the cell (Fig. 6 A). To determine whether these structures are required for cell movement, cells were microinjected with the Rho inhibitor, C3 transferase, at a concentration of 65–75 μg/ml. As shown in Fig. 6, B and C, at this concentration of C3 transferase, cells are devoid of all visible stress fibers and focal adhesions, yet they continue to extend lamellipodia and filopodia normally and wounds close (Fig. 6 D).

Although Rho activity, as visualized by stress fiber formation, is not required for wound closure, microinjection of C3 transferase at a fourfold higher concentration (300 μg/ml) results in loss of substrate adhesion and severe cell retraction (data not shown) and under these conditions wound closure is significantly inhibited (80% inhibition, see Fig. 6 D). Using this experimental approach, we were unable, therefore, to assess whether the absence of actin stress fibers/focal adhesions could enhance cell movement, since inhibition of Rho with C3 transferase has at least two distinct effects. However, stress fibers and focal adhesions can be inhibited more directly by blocking the activity of one of the downstream targets of Rho, p160ROCK (Uehata et al., 1997), with compound Y-27632. In the presence of 20 μM Y-27632, wound edge cells contained few actin stress fibers (data not shown), but remained attached to the coverslip (Fig. 6 F). Under these assay conditions, there is a significant increase in the speed of wound closure by ∼30% as indicated by the position of the wound edges in the presence (Fig. 6 F) and absence (Fig. 6 G) of Y-27632.

Finally, microinjection of a constitutively active Rho protein (V14RhoA), which is unable to cycle through an inactive GDP-bound state, severely inhibits wound closure (95% inhibition; see Fig. 6 D). Injection of V14Rho does not obviously increase the number of actin stress fibers in wound edge cells, which are already abundant, although in some cases stress fibers and focal adhesions appear to be thicker than in control cells (data not shown).

We conclude that Rho-dependent stress fibers and focal adhesions are not required for cell movement, but that a basal Rho activity is required to maintain cell substrate adhesion as previously described (Takaishi et al., 1993; Ridley et al., 1995; Santos et al., 1997). Furthermore, focal adhesions and stress fibers must undergo dynamic turnover during movement and this requirement results in an inhibitory effect on the speed of movement.

Ras has been proposed to play a role in regulating the turnover of focal adhesions (Schlaepfer and Hunter, 1998). To determine whether Ras is activated during the assay, wounded cells were visualized with an antibody that recognizes dually phosphorylated ERK, a downstream target of the Ras pathway. Fig. 7 shows that ERK is activated rapidly after wounding (within 5 min), but that the levels decrease to background by 1 h. Interestingly, ERK activation is seen to extend further back from the wound edge than c-fos induction (eight rows as opposed to only four). Wound-induced ERK activation can be blocked by microinjection of the neutralizing Ras antibody, Y13-259 (data not shown), or by preincubation with the MEK inhibitor (PD98059) (Fig. 7), indicating a requirement of the Ras/ MEK pathway for wound-induced ERK activation.

To determine whether Ras is required for cell movement, patches of cells around a wound were microinjected with the neutralizing Ras antibody Y13-259 (Furth et al., 1982). Inhibition of Ras has no effect on lamellipodia protrusions (Fig. 8, A and B), nor does it affect cell polarity (data not shown). Nevertheless, cell movement is inhibited by 80% (Fig. 8 G). Further analysis by time-lapse videomicroscopy of cells in which Ras has been inhibited revealed normal protrusive activity, but no net cell movement. However, we did observe that actin stress fibers and focal adhesions in the Ras-inhibited cells (Fig. 8, A–D) appeared to be somewhat thicker and larger than those in control cells, similar to what we observed in cells injected with constitutively active Rho. These observations suggested a possible link between Ras and focal adhesion and stress fiber turnover. To test this, cells were injected with the Ras inhibitor Y13-259 and were either coinjected with the Rho inhibitor C3 transferase, or were treated with the p160ROCK inhibitor Y-27632. Under both conditions, stress fibers and focal adhesions were lost as expected (data not shown), but now cells migrated normally resulting in wound closure (Fig. 8 G).

Interestingly, when V12Ras is expressed at relatively low levels and at early times (2 h) after microinjection of an expression construct, we clearly observe Ras concentrated at focal adhesion sites using the anti-Ras antibody, Y13-238 (see Fig. 8 E and vinculin costain in Fig. 8 F). After longer times (6 h), V12Ras has a significant affect on cell morphology, inducing loss of adhesions and changes in cell shape, typical of a transformed morphology.

Ras regulates a number of signaling cascades including the ERK-MAP kinase pathway and in some, but not all, cells PI3-kinase (Rodriguez-Viciana et al., 1997). To examine whether either of these two activities might be responsible for the Ras requirement seen here, we have used the MEK inhibitor PD98059 and the PI3-kinase inhibitor LY294002. Either of these inhibitors alone results in a relatively weak inhibition of wound closure (25 and 30%, respectively; see Fig. 8 H) and even in combination, wound closure is only inhibited by ∼50%. We conclude that neither of these downstream pathways is likely to account for the specific Ras requirement during wound closure.

The analysis of a model cell line, Swiss 3T3 fibroblasts, has led to the identification of three signal transduction pathways responsible for controlling the organization of the actin cytoskeleton and each is regulated by a member of the Rho family of small GTPases (reviewed in Ridley, 1996; Van Aelst and D'Souza-Schorey, 1997; Hall, 1998). Rho promotes the assembly of actin–myosin contractile filaments (stress fibers) and associated focal adhesion complexes, Rac induces the formation of protrusive lamellipodial structures (Ridley and Hall, 1992; Ridley et al., 1992), while Cdc42 triggers filopodia formation (Kozma et al., 1995; Nobes and Hall, 1995). Since the assembly and disassembly of filamentous actin structures drive cell movement, axon guidance, phagocytosis, and cytokinesis, it is likely that Rho GTPases will play key roles in controlling these fundamental biological processes. Here we examine the individual contributions of Rho, Rac, and Cdc42 to cell movement.

Animal cells move by crawling along a surface and this involves a combination of protrusive activity at a leading edge combined with forward movement of the nucleus and cell body (Lauffenburger and Horwitz, 1996; Mitchison and Cramer, 1996). Cells may move either as individuals or as a group, often in sheets, and examples of both types of movement can be found in the embryo and in the adult. To analyze how cells move cooperatively, we have used an in vitro wound healing assay. Individual cells in a confluent fibroblast monolayer are nonmotile and appear to have low levels of Rac and Cdc42 activity since only a few, sporadic membrane protrusions can be seen. Rho, on the other hand, appears to be highly active, since cells display many stress fibers and focal adhesions. Wounding of the monolayer induces cells at or close to the leading edge to crawl forward to close the gap and during the 5-h course of our experiments, the cells move cooperatively as a sheet, retaining close contacts with their neighbors.

The most obvious and immediate effect of wounding is to induce (within minutes) dynamic activity, lamellipodial and filopodial protrusions, and membrane ruffling, at a leading edge. This is consistent with a wound-induced activation of Rac and Cdc42, although it is formally possible that these GTPases are already active in the monolayer, but functionally silent due to cell–cell contacts. In any event, microinjection of wound edge cells with a Rac inhibitor prevents lamellipodia formation and membrane ruffling (filopodia are still seen) and there is no forward movement over the time course of the experiment. Others have reported that Rac is required for chemotaxis of fibroblasts towards PDGF (Anand-Apte et al., 1997) and macrophages towards CSF-1 (Allen et al., 1998), for hepatocyte growth factor–induced scattering of MDCK epithelial cells (Takaishi et al., 1994; Ridley et al., 1995), and for the invasive movement of some metastatic cells (Habets et al., 1994; Michiels et al., 1995). Furthermore, genetic disruption of Rac in the Drosophila embryo prevents the actin cytoskeleton changes required to drive dorsal closure, a process involving the movement of sheets of epithelial cells around the early embryo (Riesgo-Escovar et al., 1996; Glise and Noselli, 1997). It appears, therefore, that Rac may be universally required to drive cell crawling, consistent with its unique role in controlling the formation of lamellipodia.

The other protrusive structures seen after wounding are filopodia and, as expected from work with Swiss 3T3 cells, these can be completely blocked by inhibition of Cdc42 in wound edge cells. However, this results in only a partial block (∼50%) of wound closure and it appears that although Cdc42 activity is required for efficient cell movement, in this assay it is not absolutely essential. Nevertheless, closer inspection of cells in which Cdc42 has been inhibited reveals an extremely interesting role for this protein. Normally, cells at the wound margin develop a morphological polarity showing a clear leading edge (with membrane ruffles and filopodia), but no protrusional activity at their sides or rear, where they are in contact with neighboring cells. In addition, migrating cells polarize their Golgi apparatus to a position forward of the cell nucleus and in the direction of migration (see also Kupfer et al., 1982) and disruption of the Golgi apparatus directly with brefeldin A, or indirectly with nocodazole (to disrupt microtubules) inhibits directed cell migration (Bershadsky and Futerman, 1994). We observed that inhibition of Cdc42 in wound edge cells caused a complete loss of polarity; cells protrude lamellipodia all around their periphery, irrespective of whether there are cell–cell contacts, and Golgi apparatus reorientation does not occur. We conclude that Cdc42 is not required to activate Rac, but that it is required to restrict Rac activity to the leading edge through the generation of a polarizing signal.

These observations point to some interesting analogies with the yeast Saccharomyces cerevisiae, where Cdc42 is required for the formation of the bud during cell growth (Drubin, 1991; Drubin and Nelson, 1996). Inactivation of Cdc42 causes a disruption of the actin cytoskeleton and a loss of cell polarity and directed secretion; as a consequence, cell growth becomes isotropic (Adams et al., 1990). These results bare striking similarities with what we observe in the wound assay. However, one significant difference is the effect of a constitutively active Cdc42; in yeast this also causes loss of cell polarity (Ziman et al., 1991) but in wound edge cells it has no obvious effect on their polarity. Perhaps the presence of cell–cell contacts in fibroblast monolayers provides an additional constraint, and in this respect, it is interesting to note that both dominant negative and constitutively activated Cdc42 prevent directed cell migration, though not motility per se, of isolated macrophages in a chemotaxis assay and reorientation of the MTOC in isolated T cells after interaction with antigen-presenting cells (Stowers et al., 1995; Allen et al., 1998). Our observations point to some interesting interplay between Cdc42 activity and cell–cell contacts.

The mechanism by which Cdc42 generates polarity in multicellular organisms is unknown, although in yeast, Cdc42 is directed to the bud site by Bud1, another small GTPase in the Ras family. It is not even clear whether filopodia are required or whether they are formed after polarity has been established. Our experiments suggest two extreme roles for Cdc42: (a) it directs protrusion (i.e., Rac activity) and secretion (microtubule-dependent Golgi apparatus reorientation) to the leading edge; or (b) it is required for inhibition of protrusive activity from regions involved in cell–cell contact. By analogy with yeast, we favor the first model but we cannot exclude the second.

Contractile actin–myosin filaments are found in all animal cells; in cultured fibroblasts these are readily seen as well organized bundles (stress fibers), tethered to the plasma membrane at specialized sites (focal adhesions) (Jockusch et al., 1995; Burridge and Chrzanowska-Wodnicka, 1996; Burridge et al., 1997). The REFs used in the wound healing assay have abundant stress fibers and focal adhesions and these are maintained even in leading edge cells after wounding. It is often said that these structures likely inhibit locomotion (Couchman and Rees, 1979), and we have confirmed this using the Y-27632 compound, an inhibitor of the Rho-effector p160ROCK which mediates actin stress fiber assembly (Leung et al., 1996; Amano et al., 1997; Ishizaki et al., 1997; Uehata et al., 1997). Y-27632 induces a loss of actin stress fibers and focal adhesions (without obviously affecting cell shape or attachment) and causes a significant (30%) increase in the rate of cell migration. Interestingly, the myosin ATPase inhibitor BDM, which also stimulates loss of actin stress fibers, inhibited cell motility by 30% (data not shown), suggesting that other contractile, actin–myosin filaments may play some role during cell movement in this assay.

Several groups have identified an essential role for Ras in cell movement, both in chemotaxis and in wound closure assays (Sosnowski et al., 1993; Fox et al., 1994; Kundra et al., 1995). We also find that inhibition of Ras blocks cell movement, but we also show that movement can be fully restored by simultaneous inhibition of focal adhesion/stress fiber assembly by inhibiting either Rho or its effector protein, p160ROCK. This experiment strongly suggests that Ras regulates the turnover of focal adhesions and stress fibers, clearly an important feature of migrating cells (Izzard and Lochner, 1976; Dunlevy and Couchman, 1993). This could occur in at least two ways. First, Ras might regulate the Rho GTPase cycle. Since Rho is required for the formation of focal adhesions, their turnover may depend on Rho inactivation (i.e., formation of Rho.GDP from Rho.GTP), which is likely to be controlled by a GTPase-activating protein (GAP). One obvious candidate molecule that could provide a link between the two GTPases is p120RasGAP, which is known to exist in a complex with p190RhoGAP, a GAP active on Rho. In fact, overexpression of the NH₂-terminal, non-GAP containing region of p120RasGAP leads to loss of stress fibers (Ellis et al., 1990; Settleman et al., 1992; McGlade et al., 1993; Ridley et al., 1993).

The second possibility is that Ras regulates focal adhesion turnover independently of Rho. In support of this model is the observation that although the p160ROCK inhibitor Y-27632 restores cell movement after Ras inhibition, it does not restore cell movement in cells expressing constitutively activated Rho protein (V14Rho) (data not shown). However, the mechanism by which Ras might regulate focal adhesion turnover is not clear. Inhibition of MEK or PI3-kinase (two downstream pathways controlled by Ras) appears to have only minor effects on movement. This is in contrast to previous reports suggesting that PI3-kinase and the MAP kinase pathway are important for motility (Keely et al., 1997; Klemke et al., 1997). Furthermore, the components of the focal adhesion complex that are the target of Ras-mediated signals are also not clear, although focal adhesion kinase is one molecule that has been suggested to play an important role in their turnover (Ilic et al., 1995; Gilmore and Romer, 1996).

In conclusion, we have shown that Rac is essential for producing the leading edge protrusions required for forward movement in a wound healing assay. Cdc42 activity is essential to maintain the polarized phenotype of migrating cells and Rho is required for cell adhesion. Ras is also essential for promoting focal adhesion turnover during migration. Major outstanding questions concern the nature of the biochemical pathways that mediate these GTPase functions and how they are coordinated in time and space.

We thank Toshio Hamasaki (Yoshitomi Pharmaceutical Industries, Ltd., Japan) for providing Y-27632 compound, Tina Bridges and Anne Ridley for providing recombinant proteins, Hugh Paterson for Ras antibodies, Colin Hopkins for Golgi antibodies, and Durward Lawson for primary REF cells. We are grateful to Paul Martin and Louise Cramer for critical reading of the manuscript and to Colin Hopkins and Richard Lamb for helpful discussions.

GAP

GTPase-activating protein

REF

rat embryo fibroblast