Phosphoinositide 3-Kinase Induces Scattering and Tubulogenesis in Epithelial Cells through a Novel Pathway*

Hepatocyte growth factor/scatter factor (HGF/SF) treatment of the Madin-Darby canine kidney epithelial cell line causes scattering of cells grown in monolayer culture and the formation of branching tubules by cells grown in collagen gels. HGF/SF causes prolonged activation of both the mitogen-activated protein (MAP) kinase extracellular signal-regulated kinase 2 (ERK2) and the phosphoinositide 3-OH kinase (PI 3-kinase) target protein kinase B (PKB)/Akt; inhibition of either the MAP kinase pathway by the MAP kinase/ERK kinase inhibitor PD98059 or the PI 3-kinase pathway by LY294002 blocks HGF/SF induction of scattering, although in morphologically distinct ways. Expression of constitutively activated PI 3-kinase, Ras, or R-Ras will cause scattering, but activated Raf will not, indicating that activation of the MAP kinase pathway is not sufficient for this response. Downstream of PI 3-kinase, activated PKB/Akt and Rac are both unable to induce scattering, implicating a novel pathway. Scattering induced by Ras or PI 3-kinase is sensitive to PD98059, as well as to LY294002, suggesting that basal MAP kinase activity is required, but not sufficient, for the scattering response. Induction of MDCK cell tubulogenesis in collagen gels by HGF/SF is inhibited by PD98059; expression of activated Ras and Raf causes disorganized growth in this system, but activated PI 3-kinase or R-Ras causes branching tubule formation similar to that seen with HGF/SF treatment. These data indicate that multiple signaling pathways acting downstream of Met and Ras are needed for these morphological effects; scattering is induced primarily by the PI 3-kinase pathway, which acts through effectors other than PKB/Akt or Rac and requires at least basal MAP kinase function. Elevated PI 3-kinase activity induces tubulogenesis, but total inhibition and excess activation of the MAP kinase pathway both oppose this effect.

Hepatocyte growth factor/scatter factor (HGF/SF) 1 is a polypeptide cytokine produced by mesenchymal cells that induces a range of responses in epithelial cells. In addition to causing proliferation, HGF/SF will also cause normal epithelial cells grown on tissue culture plastic to "scatter," showing an increase in motility and reduction in cell-cell interaction that leads to the cells growing as a dispersed culture rather than in confluent islands (1)(2)(3). A related motility response is seen with carcinoma cells, the invasiveness of which is increased by HGF/SF (4). Another morphological effect of HGF/SF is to cause epithelial cells grown in three-dimensional collagen gels to form branching tubules reminiscent of epithelial ducts in the breast or kidney (5,6).
The effects of HGF/SF are mediated through its cell surface receptor, a tyrosine kinase encoded by the c-met proto-oncogene (7,8). In recent years, considerable advances have been made in understanding the signaling pathways triggered by ligand activation of Met. The receptor autophosphorylates at two carboxyl-terminal tyrosines, Tyr 1349 and Tyr 1356 (9). Both sites are capable of binding the Src homology 2 domains of p85 (the regulatory subunit of phosphoinositide 3-OH kinase (PI 3-kinase)), pp60 c-src , phospholipase C-␥, and Shc; in addition, Tyr 1356 can bind to the Src homology 2 domain of Grb2 (10,11). As a result of these interactions, HGF/SF stimulates the activation of PI 3-kinase, pp60 c-src , phospholipase C-␥, Ras, and the mitogen-activated protein kinase ERK2 (9,12,13).
Although the immediate effects of HGF/SF on intracellular signaling pathways are well characterized, the mechanisms involved in longer term responses, such as scattering and tubulogenesis, are less well understood. In the case of scattering in the dog kidney epithelial cell line MDCK, it has been shown that normal function of Ras is absolutely necessary for HGF/ SF-induced motility (14,15). Expression of activated Ras in MDCK cells causes a scattered appearance (14,16), although microinjection of activated Ras protein induces only the early stages of the scattering response, cell spreading and actin rearrangement (15). Inhibition of the function of the Ras-related GTPase Rac inhibits HGF/SF-induced scattering, although microinjection of activated Rac itself does not induce this response (15). Activated Rho protein inhibits HGF/SFinduced scattering (15). Inhibition of PI 3-kinase by wortmannin also inhibits HGF/SF-induced scattering (17). It is therefore likely that Ras, PI 3-kinase, and Rac all play essential roles in scattering; a signaling pathway connecting these three proteins has recently been described (18). However, it is probable that other pathways will be involved in this protracted and complex response. In particular, modulation of cell-cell and cellmatrix interactions is important (19), possibly involving HGF/ SF-induced phosphorylation of ␤-catenin and plakoglobin (20).
HGF/SF promotion of tubulogenesis is a yet longer term response, about which less is known. Tyrosine 1356 of Met has been shown to be required for tubulogenesis, whereas it is not essential for scattering (21). Recently, Gab1, a novel adaptor protein related to IRS-1, has been described (22); it has been reported that this interacts specifically with Met and that this interaction correlates with induction of branching tubulogen-esis by HGF/SF (23). Tyrosine 1356 is required for Gab1 association with Met (24). Tpr-Met-induced transformation also correlates with Gab1 binding to this site (25). At present, the precise early signaling pathways involved in the induction of tubulogenesis in response to HGF/SF are not clear; however, it is probable that the time course of these signals will be very important in this long term change in cell behavior. As is the case with scattering, modulation of cell-cell and cell-matrix interaction is also likely to be critical in this response (26).
In this paper, we have set out to study the role of signaling pathways controlled by Ras in the scattering and tubulogenesis responses of MDCK cells to HGF/SF. Ras is known to control several signaling pathways, including the Raf/MAP kinase pathway, PI 3-kinase, and Ral GDP dissociation stimulator, a guanine nucleotide exchange factor for the Ras-related protein Ral (27). Partial loss of function mutants of Ras have been described that activate a known subset of these pathways (18,28,29). Using these mutants and also activated versions of Ras effectors such as Raf and PI 3-kinase, we show that expression of activated Ras or PI 3-kinase, but not of Raf, is sufficient to induce scattering. Scattering in response to Ras, PI 3-kinase, or HGF/SF is sensitive to pharmacological inhibition of either the MAP kinase kinase MEK or of PI 3-kinase. This suggests that activation of PI 3-kinase can induce scattering dependent on basal MAP kinase function. In the case of tubulogenesis, cells expressing activated Ras or Raf fail to form tubules but grow in a disorganized manner, whereas cells expressing activated PI 3-kinase form tubules constitutively. Tubulogenesis in response to both HGF/SF and expression of activated PI 3-kinase is inhibited by MEK inhibitor, again suggesting that basal function of this pathway is required in addition to PI 3-kinase activity. In addition to Ras, we also studied the involvement of R-Ras in HGF/SF-induced events. R-Ras is a ubiquitously expressed protein with transforming potential (30); it is 55% identical to Ras at the amino acid level and interacts with many of the same downstream target proteins, activating PI 3-kinase but not Raf (31). Activated R-Ras constitutively induces both scattering and tubulogenesis, suggesting the possibility that this protein may be an important component of the normal cellular response to HGF/SF.

Expression Vectors
The Ras and R-Ras expression vectors were generated as described in Ref. 31. Activated PI 3-kinase constructs were made by the addition of a carboxyl-terminal farnesylation signal from H-Ras to p110 to cause its membrane localization (Myc-tagged p110-CAAX in pcDNA3). 2 Raf-CAAX cDNA were provided by R. Marais and C. J. Marshall (Institute of Cancer Research, London). Gag-PKB and hemagglutinin-tagged PKB cDNAs were provided by B. Burgering and P. Coffer (University of Utrecht).

Cell Lines
Early passage MDCK cells were provided by J. Taylor-Papadimitriou (Imperial Cancer Research Fund) and were maintained in DMEM supplemented with 10% FBS. To generate stably transfected lines, wild type cells were transfected with the various constructs by lipofection (LipofectAMINE, Life Technologies, Inc.). Cells were plated at 10 5 cells/ well of a six-well plate and transfected the following day with 4 g of DNA (plus 0.4 g of a hygromycin resistance plasmid for those vectors not encoding a neomycin resistance gene) and 10 l of LipofectAMINE/ well. After 48 h, cells were replated into a 10-cm dish and selected with 250 g/ml hygromycin or 500 g/ml G418 as appropriate. Control cells were transfected with empty vector alone or were left untransfected. After selection in antibiotic for 10 -14 days, individual resistant colonies were picked and expanded. At least 24 clones were picked for each transfection, and expression of protein was assessed by Western blotting of cell lysates using appropriate antibodies. Expression of V12 H-Ras and the Ras mutant proteins (C40 and G37) varied between 5-fold and 10-fold excess over endogenous Ras levels. Gag-PKB protein levels were 3-5-fold greater than endogenous Akt/PKB levels. Clones expressing Raf-CAAX, p110*, V38R-Ras, and V14Rho were identified by Western blotting using an epitope tag antibody (anti-Myc, clone 9E10). Akt/PKB kinase activity in the clones expressed as fold increase over wild type cells growing under basal conditions were as follows: V12Ras, 10 -16-fold; Raf-CAAX, 0.6 -1.5-fold; p110*, 5-7-fold; V12C40Ras, 4.4 -9.6-fold; gag-PKB, 9 -12-fold; V38R-Ras, 3.4 -7-fold; V12G37Ras, 0.7-1.3-fold. ERK activity was increased 4 -7-fold in V12Ras clones and 7-20-fold in Raf-CAAX clones and was basal in all of the others. Over this range of overexpression of signaling proteins, there was no clear dose-response relationship with respect to the scattering and cytoskeletal rearrangement phenotype.

Cell Scattering
MDCK were seeded at 10 4 cells/well of a 24-well plate on glass coverslips in DMEM, 10% FBS 48 h before the assay. Cells were pretreated with the PI 3-kinase inhibitor LY294002 (20 M) (Biomol) or the MEK inhibitor PD98059 (30 M) (Calbiochem) as indicated for 30 min in DMEM ϩ 2 mg/ml BSA (fatty acid-free, Sigma) prior to the addition of EGF (20 ng/ml, Sigma) or recombinant human HGF/SF (10 ng/ml, Sigma). For phase-contrast, cells were fixed in 0.5% glutaraldehyde in PBS (30 min at room temperature) and stored at 4°C under PBS.
Staining for Actin-Cells were fixed in 3.7% formaldehyde in PBS for 30 min at room temperature. After three washes in PBS, coverslips were incubated in PBS ϩ 0.2% Triton X-100 for 10 min. After three washes in 1.5% BSA/PBS, cells were incubated in tetramethyl rhodamine isothiocyanate-phalloidin (Sigma) for 60 min at 37°C, washed six times in PBS, and mounted. Coverslips were examined by confocal microscopy.
Staining for ␤-Catenin-Staining was performed essentially as for actin staining, except that ␤-catenin monoclonal antibody (Transduction Laboratories, C19220) was used instead of phalloidin at a dilution of 1:200 in 1.5% BSA/PBS for 20 min at 20°C, followed by washing once with 1.5% BSA/PBS for 30 min at 4°C and twice with 1.5% BSA/PBS for 5 min at room temperature. Secondary antibody was fluorescein isothiocyanate anti-mouse IgG (The Jackson Laboratory, 115-095-100) at a dilution of 1:200 in 1.5% BSA/PBS for 20 min at room temperature in the dark, followed by three washes in 1.5% BSA/PBS for 5 min each. Coverslips were mounted and examined by confocal microscopy.

Kinase Assays
PKB/Akt-Cells were plated at 10 5 cells/well of a six-well plate and transfected the following day with 4 g of hemagglutinin-tagged PKB/ Akt DNA and 10 l of LipofectAMINE/well in 1 ml of Opti-MEM (Life Technologies, Inc.) for 6 h. 1 ml of DMEM, 10% FBS was added, and cells were incubated overnight. After removal of the transfection medium, cells were starved in DMEM, 0.2% FBS for 18 -24 h prior to assay. After stimulation with growth factors, cells were lysed in lysis buffer containing 150 mM NaCl, 50 mM Hepes (pH 7.5), 1% Triton X-100, 5 mM EDTA, 10 g/ml aprotinin/pepstatin/leupeptin, 1 mM sodium orthovanadate, and 1 mM phenylmethylsulfonyl fluoride. Lysates were centrifuged at 13000 ϫ g for 10 min at 4°C, and the supernatant was incubated with 2 g of anti-hemagglutinin antibody (12CA5) for 2 h and then with protein G-agarose for 1 h. Pellets were washed twice in wash buffer (PBS, 1% Triton X-100, 1 mM EDTA), once in high salt buffer (0.5 M LiCl, 0.1 M Tris, pH 8, 1 mM EDTA), and once in kinase buffer wash (50 mM Tris, pH 7.5, 10 mM MgCl 2 ). The pellet was resuspended in a kinase mixture consisting of kinase buffer (50 mM Tris, pH 7.5, 10 mM MgCl 2 , 1 mM dithiothreitol), 2.5 g histone H2B (Boehringer Mannheim), 1 M protein kinase inhibitor (Sigma), 50 M ATP, and 3 Ci of [␥ 32 P]ATP (Amersham Pharmacia Biotech); after 30 min of incubation at room temperature, the reaction was stopped by adding sample buffer and boiling for 5 min. Following SDS-polyacrylamide gel electrophoresis and transfer to nitrocellulose, the lower part of the membrane was analyzed for H2B phosphorylation and quantified by PhosphorImager Analysis (Molecular Dynamics). The upper part of the membrane was probed with polyclonal anti-PKB/Akt antibody (kindly provided by Dr. B. Burgering, Utrecht, Netherlands) and quantified using ECF/Storm imaging (Amersham Pharmacia Biotech and Molecular Dynamics). The H2B phosphorylation was then corrected for hemagglutinin-PKB/Akt expression.
ERK2-Cells were plated at 2 ϫ 10 5 cells/well of a six-well plate in DMEM, 10% FBS. Medium was replaced the following day with DMEM, 2 mg/ml BSA, and cells were starved for 18 -24 h. Following growth factor stimulation, cells were lysed as described above, and ERK 2 was 2 S. Wennström and J. Downward, manuscript in preparation. immunoprecipitated using a polyclonal antibody (provided by C. Marshall) and protein A-agarose. Lysates were washed twice with wash buffer and once with kinase buffer (25 mM Tris, pH 8, 10 mM MgCl 2 , 2 mM MnCl 2 ). Kinase assay was carried out in kinase buffer plus 2.5 g MBP, 10 M ATP, and 2 Ci [␥ 32 P]ATP for 30 min at room temperature and quantified by PhosphorImager analysis after SDS-polyacrylamide gel electrophoresis.

Collagen Gel Cultures
Collagen (Collagen Corp., Palo Alto, CA) solution was mixed with 10ϫ DMEM and sterile water to a final collagen concentration of 2 mg/ml and neutralized with NaOH on ice. 0.7 ml/35-mm dish was allowed to gel at 37°C. MDCK cells were harvested using trypsin/EDTA and plated at 5 ϫ 10 3 /well in a further 1 ml of collagen solution on top of the previously gelled collagen layer. After 15 min at 37°C to allow the solution to gel, 1 ml of complete medium (DMEM, 10% FBS) was added. When indicated, this was supplemented with HGF/SF (20 ng/ml), and the medium was changed every 2-3 days. Cultures were routinely photographed at 7-10 days or as indicated.

Stimulation of MDCK Cells with HGF/SF Leads to a Prolonged Activation of PKB/Akt and ERK2 Protein Kinases-
MDCK cells undergo a scattering response when stimulated with HGF/SF (1) but show weak scattering in response to EGF. Scattering has been shown to involve signaling pathways, including Ras, Shc, and PI 3-kinase (9,14,17,32), all of which are activated in response to both HGF/SF and EGF. It is possible, however, that the kinetics of activation of Ras and PI 3-kinase pathways may modify the motogenic response. There is precedence for such a situation in PC12 cells, which respond to EGF by proliferation and to NGF by differentiation, depending on the kinetics of activation of the Ras/Raf/ERK pathway. Fig. 1A confirms the effects of HGF/SF and EGF on MDCK motility. In order to evaluate activation of the PI 3-kinase pathway, we measured PKB/Akt kinase activation in response to EGF and HGF/SF. Fig. 1B shows that the activation of PKB/Akt by EGF was transient: it peaked at 5 min (3.9-fold increase over basal activity) and returned to baseline by 30 min. In contrast, the activation by HGF/SF was protracted, with a slower onset (1.85-fold increase at 5 min; peak, 3.83-fold at 120 min) and kinase activity that was still elevated (2.28fold) 6 h after the addition of HGF/SF. The activation of ERK2 by EGF showed a peak at 5-15 min (3.1-fold over basal) and a steady decline in activity to reach a basal level by 2-3 h. In response to HGF/SF, there was again a slower onset of action (peak 4-fold at 15 min) and more prolonged activation with significantly elevated activity at 4 h (2.86-fold). Therefore, the kinetics of activation of both the PI 3-kinase and ERK pathways are markedly different in response to HGF/SF compared with EGF, with HGF/SF stimulation resulting in prolonged activation of both signaling pathways.
Inhibition of Either PI 3-Kinase or MEK Blocks the Motogenic Response to HGF/SF-It has been shown previously that inhibition of PI 3-kinase reduces HGF/SF-induced scattering (17). We confirmed that the PI 3-kinase inhibitor LY294002 will prevent scattering in response to HGF/SF (Fig. 2). We also utilized the MEK inhibitor PD98059 to examine the role of the MAP kinase pathway in scattering. When used at a concentration that blocks HGF/SF-induced ERK activation completely, PD98059 prevented cell scattering. A comparison of the effects of LY294002 and PD98059 on HGF/SF-induced changes in the actin cytoskeleton was also carried out. HGF/SF treatment led to a reduction in the number of actin stress fibers, a breakdown in cell-cell contacts, and dispersion of colonies. Cells pretreated with LY294002 still showed the HGF/SF-induced reduction in stress fibers and some loss of cell-cell contacts within the colony, but the cells did not disperse. In contrast, pretreatment with PD98059 maintained, or even increased, stress fibers, and there was no loss of cell-cell contact.
Therefore, although the net result of the inhibitors on scattering is similar, it appears likely that this is brought about in different ways. The breakdown of cell-cell contacts and the increase in cell motility may be regulated separately by the Raf/MEK/MAP kinase and PI 3-kinase dependent pathways. Activated Ras can influence both pathways, although the relative contribution of endogenous Ras to PI 3-kinase activation varies between cell type and stimulus (33)(34)(35).
Morphology of MDCK Clones Grown on a Solid Substratum-It has been shown previously that MDCK cells expressing constitutively activated Ras have an altered morphological appearance with increased membrane ruffling and extended cell processes (16). In contrast to parental cells, they do not form discrete colonies at low cell densities but have a more dispersed appearance, with reduced cell-cell contacts. In order to investigate which of the signaling pathways controlled by Ras may be responsible for this phenotype, we utilized activated versions of various proteins known to lie downstream of Ras. Wild type MDCK cells were transfected with constructs expressing activated forms of H-Ras (V12 Ras), Raf (Raf-CAAX), and PI 3-kinase (p110*) and with partial loss of function mutant Ras proteins that distinguish between the various effector pathways downstream of Ras: V12S35 Ras selectively activates the Raf pathway, V12C40 Ras selectively activates the PI 3-kinase pathway, and V12G37 Ras selectively activates the Ral GDP dissociation stimulator pathway (18). In addition, cells were transfected with the activated Ras-related protein R-Ras (V38 R-Ras) and with activated RhoA (V14 Rho). Fig. 3A shows phase-contrast morphology of wild type MDCK and of clones stably expressing the various constructs. The morphological appearances shown are representative of at least four separate clones expressing the same constructs.
MDCK Clones with an Activated PI 3-Kinase Pathway Have a Scattered Appearance-Wild type MDCK cells grow as discrete colonies at subconfluent cell densities as do cells expressing activated Rho. As previously reported, cells expressing activated Ras do not form compact colonies but show extended processes and lamellipodia. A similar appearance is seen with cells expressing activated PI 3-kinase, the V12C40 Ras mutant that activates PI 3-kinase, and with cells expressing activated R-Ras. V38 R-Ras is able to interact with and activate PI 3-kinase in several cell types (31), and MDCK clones expressing activated R-Ras show elevated activity of the PKB/Akt kinase, which lies in a pathway downstream of PI 3-kinase (data not shown). In contrast, clones expressing the V12G37 Ras mutant, which does not interact with PI 3-kinase or Raf, form compact colonies. MDCK cells expressing activated Raf grow in loosely associated colonies: although the cells are not scattered, the regions of cell-cell contact are not as well formed as those of wild type or V12G37 Ras cells. Cells expressing the V12S35 Ras mutant had a similar appearance to those expressing Raf-CAAX (data not shown). Although activated PI 3-kinase induces scattering, both activated PKB/Akt and Rac, the two well studied targets of this enzyme, fail to do so ( Fig. 3A and data not shown). PI 3-kinase, therefore, presumably uses another, probably as yet unknown, pathway to mediate these changes in cell behavior. (Fig. 3B) of wild type cells showed the presence of stress fibers and peripheral bundles of actin fibers at the edges of colonies (15). A similar appearance was seen in cells expressing V12G37 Ras. Cells expressing activated p110*, V12 Ras, V12C40 Ras, or activated V38 R-Ras had few stress fibers or peripheral bundles. The expression of activated V14 Rho led to an increase in stress fibers and peripheral bundles; these cells failed to scat-ter in response to HGF/SF (data not shown). Activated Raf-CAAX clones showed a reduction in stress fibers, but some cells had peripheral bundles, even though cells were not arranged in compact colonies. All of the different cell types were able to form continuous epithelioid sheets when grown to confluence (not shown).

MDCK Clones with Either Activated Raf or PI 3-Kinase Show a Reduction in Actin Stress Fibers-Actin staining
In summary, the constitutive activation of PI 3-kinase, either by the expression of activated p110 or by expression of Ras and R-Ras proteins that activate PI 3-kinase, leads to cell scattering, a more fibroblastic appearance, and some reduction of actin stress fibers under basal conditions. Although the scattering response to HGF/SF requires both PI 3-kinase and MEK activation, it appears that stable expression of activated PI 3-kinase alone is sufficient to lead to cell scattering. This may be a result of the continuous high level activation of PI 3-kinase in these clones. Cells expressing activated Raf do not show widespread dispersion of colonies but do have fewer stress fibers and form less effective cell-cell contacts and less compact colonies than control cells. These data are consistent with the findings, detailed above, that activation of the Raf/MEK pathway is permissive, although not fully sufficient, for HGF/SFinduced cell scattering.
LY294002 and PD98059 Reverse the Dispersed Appearance of Cells Expressing V12 Ras or Activated PI 3-Kinase-To investigate whether PI 3-kinase or MEK inhibition could reverse the morphological appearance of cells expressing activated Ras or PI 3-kinase, cultures were treated with inhibitors for 8 h. Incubation with LY294002 led to the appearance of more closely associated colonies in clones expressing V12 Ras, p110*, and V12C40 Ras (Fig. 4A). In addition, PD98059 also caused a reversion to a wild type morphology in all three cell types. Although clones expressing V12 Ras have elevated ERK activity, this is not the case for cells expressing p110* or V12C40 Ras (36) However, this level of PD98059 reduces ERK activity to less than 30% of basal in wild type and in p110* and V12C40 Ras clones under these growth conditions (not shown). This suggests that basal MEK activity is essential for the scattered morphology of cells expressing p110* and V12C40 Ras.
Actin staining of V12 Ras, Raf-CAAX, and p110* cells treated with LY294002 and PD98059 shows that PD98059 treatment leads to increases in the levels of actin stress fibers in each cell type (data not shown). In addition, cells expressing activated Raf appear to have closer cell-cell contacts following treatment with PD98059. In order to investigate the effects of these signaling pathways on cell-cell junctions, cell lines expressing V12 Ras, Raf-CAAX, and p110* and wild type MDCK cells were stained for ␤-catenin in either the absence or the presence of LY294002 and PD98059. As shown in Fig. 4B PD98059, the cells showed much more ␤-catenin staining at sites of contact, which were far more extensive. Similar effects were seen for V12 Ras-expressing cells. In the case of Raf-CAAX-expressing cells, the cells normally grow in a nonscattered manner but show reduced tightness of the colonies; PD98059, but not LY294002, resulted in a tightening of the colonies and an increase in the amount of ␤-catenin staining at cell-cell junctions.
Branching Tubulogenesis in Collagen Induced by HGF/SF Is Inhibited by PD98059 and the Expression of V14 Rho and V12G37 Ras-Wild type MDCK cells grown in three-dimensional collagen gels formed simple cystic structures. In the presence of HGF/SF, there was marked branching tubulogenesis (Fig. 5A). Co-incubation with the MEK inhibitor PD98059 prevented effective tubule formation. Prolonged incubation with the PI 3-kinase inhibitor LY294002 resulted in extensive cell death. Cells expressing activated V14 Rho and V12G37 Ras, which have reduced motogenic responses to HGF/SF, also failed to form tubules in response to HGF/SF (Fig. 5B).
Effect of Expression of V12 Ras, Raf-CAAX, Activated PI 3-kinase, and R-Ras on Tubulogenesis-MDCK clones stably expressing V12 Ras, Raf-CAAX, p110*, V12C40 Ras, and V38 R-Ras were grown in collagen gels in the absence of HGF/SF (Fig. 6). Both activated Ras and Raf did not produce wellformed cysts but led to the development of diffusely spread clusters lacking organized structure. Cells expressing activated PI 3-kinase or V12C40 Ras (the mutant that activates PI 3-kinase) formed relatively simple tubular structures with less branching than was seen with HGF/SF-treated wild type cells. Clones expressing activated V38 R-Ras formed tubular structures with extensive branching. These were also seen to a lesser extent in cells expressing high levels of wild type R-Ras. The branching structures were seen earlier after cell seeding in collagen in clones expressing V38 R-Ras when compared with wild type cells treated with HGF/SF. Typically, in V38 R-Ras cells extensive branching tubules were seen 4 -6 days after the start of culture, compared with 8 -10 days with wild type cells treated with HGF/SF. The effects of expressing activated PKB/ Akt, a downstream target of R-Ras and PI 3-kinase, on tubulogenesis was also assessed; gag-PKB-expressing cells behaved like wild type cells. Electron microscopic analysis of the branching tubules formed in response to activated PI 3-kinase, V12C40 Ras, and V38 R-Ras revealed that they contained lumens and were similar to those formed in wild type MDCK cells in response to HGF/SF (data not shown).
PD98059 Treatment of V12 Ras-expressing Cells Induces Tubulogenesis-Because Raf-CAAX expression caused dispersed growth of MDCK cells in collagen gels, seemingly with little cell-cell interaction, whereas activated PI 3-kinase caused tubulogenesis, it seemed likely that activated Ras was promoting both signaling pathways, but the Raf effect was dominant over the PI 3-kinase effect. In order to investigate this further, V12 Ras-expressing cells were treated with PD98059 in collagen gels. As shown in Fig. 7, moderate levels of the MEK inhibitor induced formation of branching tubules in these cells; tubules formed rapidly, in 4 -6 days, compared with 8 -10 days for wild type cells treated with HGF/SF. High levels of PD98059 caused the V12 Ras MDCK cells to form small cysts only, perhaps due to inhibition of cell proliferation, which is likely to be a prerequisite for branching tubule formation. It therefore appears that strong activation of the Raf/MAP kinase pathway downstream of Ras prevents tubulogenesis, possibly by disrupting cell-cell interactions, whereas activation of PI 3-kinase promotes tubulogenesis: the Raf effect is dominant over the PI 3-kinase effect. DISCUSSION PI 3-kinase activity has previously been shown to be required for HGF/SF-induced scattering in MDCK cells (17). In this study, we show that constitutive activation of PI 3-kinase is sufficient to cause scattering provided that basal MAP kinase function is not inhibited. Expression of activated PI 3-kinase, Ras, or R-Ras, all of which lead to long term elevation of phosphatidylinositol 3,4,5-trisphosphate and phosphatidylinositol 3,4-bisphosphate levels (31), causes scattering, even though PI 3-kinase and R-Ras do not induce MAP kinase activation (36). Expression of activated Raf, which does activate FIG. 6. Effect of expression of activated signaling proteins on spontaneous tubule formation in collagen gels. MDCK clones stably expressing the proteins as indicated were grown in collagen gels in the absence of HGF/SF for eight days and examined by phase-contrast. Similar results were seen in three separate experiments utilizing two or three independent clones for each construct. WT, wild type. MAP kinase (36), fails to cause scattering. From these data, it would appear that activation of the PI 3-kinase pathway alone is sufficient to induce the scattering response. However, although LY294002, a PI 3-kinase inhibitor, blocks scattering as expected, PD98059, a MEK inhibitor, also blocks scattering, even though in the case of cells expressing activated PI 3-kinase or R-Ras the MAP kinase pathway is not obviously activated. A likely explanation of this is that basal signaling through the MAP kinase pathways is essential, but not sufficient, for scattering. All of the assays described in this paper were carried out in the presence of 10% serum (which is not sufficient to induce scattering), so a constitutive basal level of activity of both the MAP kinase and PI 3-kinase pathways would be expected. PD98059 inhibits the basal activity of MAP kinase in cells grown under these conditions by more than 70% (data not shown).
Inhibition of PI 3-kinase and of MEK has distinct morphological effects on the response of MDCK cells to scattering stimuli. In the case of LY294002, HGF/SF-treated cells failed to disperse, but they did show reduced cell-cell contacts and stress fibers. In contrast, PD98059 caused HGF/SF-treated cells to remain largely in contact and made stress fiber loss less apparent. Cells expressing Raf-CAAX did not scatter constitutively, yet they did show reduced cell-cell contact and stress fibers. It therefore appears likely that activation of the Raf/ MAP kinase pathway, at least to basal levels, is required for the dissolution of cell-cell contacts, and perhaps stress fibers, whereas activation of the PI 3-kinase pathway is needed for the cells to disperse. It may be that the primary role of the Raf/ MAP kinase pathway is to effect cell-cell contact, whereas the primary role of the PI 3-kinase pathway is to induce cell motility. Both signaling components are needed for scattering, and both are induced by treating cells with HGF/SF. The failure of EGF to induce efficient scattering is likely to be due to the fact that this growth factor only induces transient activation of MAP kinase and PI 3-kinase, whereas HGF/SF causes prolonged activation of these pathways.
One feasible explanation of these results is as follows. Expression of constitutively activated PI 3-kinase in cells results in elevated cell motility; in cells experiencing basal stimulation of the MAP kinase pathway by serum, sufficient weakening of cell-cell contacts and stress fibers may occur to allow the cells to break away from each other and move apart. Inhibition of MEK by PD98059 prevents this and thus blocks scattering. Conversely, in cells expressing constitutively activated Raf, the strength of cell interactions is reduced, but the cells fail to move apart because there is insufficient stimulation of PI 3-kinase to induce cell movement. Recently it has been reported that MAP kinase phosphorylates and activates myosin light chain kinase, providing another possible mechanism for the effects of PD98059 on MDCK cell scattering (37). Other indications that the Raf/ MAP kinase pathway might lead to reduction in cell-cell contacts come from the observations that overexpression of c-Jun or c-Fos results in rapid destabilization of adherens junctions (38,39).
The mechanism of the tubulogenesis response of MDCK cells to HGF/SF is similar to that of the scattering response in certain aspects, but not others: it is likely that tubulogenesis involves the use of a larger number of signaling pathways than scattering, including the STAT pathway (40). PD98059 blocks both responses, suggesting that basal MAP kinase signaling is required in tubulogenesis as well as scattering. LY294002 induces apoptosis in the time course of the tubulogenesis assay, although from the data with scattering it would seem likely that PI 3-kinase is required for tubulogenesis in addition to its role in providing a survival signal for epithelial cells (36). However, constitutive activation of MAP kinase by expression of activated Raf leads to disorganized growth in collagen gels and an inability to form tubules in response to HGF/SF. It appears likely that the reduction in cell-cell interaction in cells with strongly activated Raf makes it impossible to form multicellular complexes. In contrast, expression of constitutively activated PI 3-kinase, or a Ras mutant that activates only PI 3-kinase and not other effectors, is sufficient to cause tubule formation by itself. Again, this is inhibited by PD98059 and is thus dependent on basal function of the MAP kinase pathway (data not shown).
Activated Ras-expressing cells fail to form tubules, suggesting that the loss of cell organization caused by strong Raf activation overrides the effect of constitutive PI 3-kinase activation. Ras-transformed epithelial cells have been reported to have reduced ␤-catenin/E-cadherin complexes due to constitutive tyrosine phosphorylation of ␤-catenin (41), although it is not clear whether this occurs through Raf/MAP kinase pathway or is sufficient to disrupt cell adhesion. There do not appear to be marked differences in the overall levels of expression of either E-cadherin or ␤-catenin in any of the cell lines studied in this paper (data not shown). In order for tubule formation to occur, it may be necessary for Raf/MAP kinase to be sufficiently active to allow remodeling of cell interactions, but not so active that cell organization is lost. Alternatively, strong inhibition of the MAP kinase pathway may prevent cell proliferation, which is needed to form tubules. Interestingly, inhibition of the Raf/MAP kinase pathway by moderate levels of PD98059 results in tubulogenesis in V12 Ras-expressing MDCK cells (Fig. 7), suggesting that strong activation of the Raf pathway does indeed mask the tubulogenesis signal provided from Ras via PI 3-kinase. Similar dominance of Raf effects over PI 3-kinase effects downstream of Ras has been noted in the influence of Ras on Myc-induced apoptosis in fibroblasts (42).
From the data presented here, PI 3-kinase appears to be a key mediator of scattering and tubulogenesis. The nature of the downstream targets of PI 3-kinase in these pathways is currently being investigated. It is clear, however, that activated PKB/Akt alone is not able to induce tubulogenesis (Fig. 6) or scattering (Fig. 3A). PKB/Akt mediates the survival signal provided by PI 3-kinase (36) but does not influence the cytoskeleton (35). However, it is possible that the ability of PI 3-kinase to provide a survival signal through PKB/Akt could to some extent promote tubulogenesis, and perhaps scattering, by protecting cells from matrix-detachment induced cell death during remodeling of their contacts with the matrix; it has been reported that overexpression of Bcl-2 can promote epithelial cell tubulogenesis (43). Another possible effector downstream of PI 3-kinase is the Rho family GTPase Rac; this could be involved in these morphological responses because it is known to have effects on cell morphology, scattering, and invasiveness (15,44,45). However, because microinjection of activated Rac into MDCK cells is not sufficient to cause scattering (15), and activated Rac expression has recently been shown to increase the strength of cell-cell contacts (46,47), there must be other pathways downstream of PI 3-kinase that are required for the full scattering response. Activated Rac failed to have significant effect on MDCK cell scattering or tubulogenesis in our hands, even when activated PKB/Akt was co-expressed (data not shown).
In contrast to Ras, expression of activated R-Ras is sufficient to cause tubulogenesis of MDCK cells. It has been reported that R-Ras activates PI 3-kinase but not Raf (31); R-Ras therefore is able to promote tubulogenesis through the PI 3-kinase pathway without causing the disorganized proliferation seen with activated Raf or Ras. In addition, like Ras, R-Ras causes good scattering of MDCK cells in monolayer culture. These observations raise the possibility that R-Ras, rather than Ras, may be the key regulator of the tubulogenesis and scattering responses downstream of Met. At present, no proven assay for measuring the activation state of endogenous R-Ras has been developed, so it is not known whether it is activated in response to HGF/SF in the way that Ras is (12). In addition, it is not known whether mutations in R-Ras equivalent to serine to asparagine at residue 17 of Ras have dominant negative activity. R-Ras is not regulated by the ubiquitous Ras exchange factor Sos (30,48), although the brain-specific exchange factor Ras-GRF does act on R-Ras (49); R-Ras is inactivated by the same GTPaseactivating proteins as Ras (50). Activated R-Ras will transform some fibroblast cell lines, such as NIH 3T3, but not Rat 1 (51,52), whereas in MDCK cells it promotes survival in suspension but does not cause growth in soft agar. 3 The data presented here emphasize the importance of the ability of key signaling proteins, such as Met and Ras, to induce activation of multiple pathways to achieve long term physiological alterations in cell behavior. The exact mechanisms involved in HGF/SF-induced scattering and tubulogenesis are likely to be complex and will involve both rapid posttranslational modifications and longer term changes in gene expression. However, it is clear from this work that prolonged activation of PI 3-kinase is sufficient for the induction of both scattering and tubulogenesis provided that basal function of the Raf/MAP kinase pathway is operating.