Phosphatidylinositol 3-Kinase Function Is Required for Transforming Growth Factor β-mediated Epithelial to Mesenchymal Transition and Cell Migration*

We have studied the role of phosphatidylinositol 3-OH kinase (PI3K)-Akt signaling in transforming growth factor β (TGFβ)-mediated epithelial to mesenchymal transition (EMT). In NMuMG mammary epithelial cells, exogenous TGFβ1 induced phosphorylation of Akt at Ser-473 and Akt in vitro kinase activity against GSK-3β within 30 min. These responses were temporally correlated with delocalization of E-cadherin, ZO-1, and integrin β1 from cell junctions and the acquisition of spindle cell morphology. LY294002, an inhibitor of the p110 catalytic subunit of PI3K, and a dominant-negative mutant of Akt blocked the delocalization of ZO-1 induced by TGFβ1, whereas transfection of constitutively active p110 induced loss of ZO-1 from tight junctions. In addition, LY294002 blocked TGFβ-mediated C-terminal phosphorylation of Smad2. Consistent with these data, TGFβ-induced p3TP-Lux and p(CAGA)12-Lux reporter activities were inhibited by LY294002 and transiently expressed dominant-negative p85 and Akt mutants in NMuMG and 4T1 cells. Dominant-negative RhoA inhibited TGFβ-induced phosphorylation of Akt at Ser-473, whereas constitutively active RhoA increased the basal phosphorylation of Akt, suggesting that RhoA in involved in TGFβ-induced EMT. Finally, LY294002 and neutralizing TGFβ1 antibodies inhibited ligand-independent constitutively active Akt as well as basal and TGFβ-stimulated migration in 4T1 and EMT6 breast tumor cells. Taken together, these data suggest that PI3K-Akt signaling is required for TGFβ-induced transcriptional responses, EMT, and cell migration.

The transforming growth factor ␤ (TGF␤) 1 family of secreted factors is involved in the control of different biological processes including cell proliferation, differentiation, and apoptosis (1). TGF␤ signals through the activation of heteromeric complexes of TGF␤ type I (T␤RI) and type II (T␤RII) receptors (1,2). Activated T␤RI phosphorylates receptor-associated Smads (Smad2 and Smad3), which then bind Smad4 and translocate to the nucleus where they regulate transcription of target genes (3,4). TGF␤ exhibits a tumor suppressor activity, and components of its signaling pathway are frequently mutated or silenced in colon and pancreatic cancers (1,5). However, accumulating data indicate that TGF␤ can positively affect tumorigenesis and contribute to the progression and invasiveness of tumors (5)(6)(7)(8). Moreover, it was recently reported that inhibition of autocrine TGF␤ signaling in carcinoma cells reduces cell invasiveness and tumor metastases (9,10). These effects of TGF␤ are associated with its ability to induce an epithelial to mesenchymal transition (EMT) and stimulate cell migration.
The EMT induced by TGF␤ results in the disruption of the polarized morphology of epithelial cells, formation of actin stress fibers, and enhancement of cell migration (8,9). Two species of T␤RI, Alk2 and Alk5, have been implicated in the induction of EMT by TGF␤ in mammary epithelial cells (11,12). It has also been reported that high levels of ectopic Smad2 and Smad3 can induce some features of EMT in mammary epithelial cells in the context of expression of an activated type I receptor (12). However, considering the complexity of TGF␤ signaling (3,(13)(14)(15)(16), it is conceivable that other molecules can also contribute to EMT. For example, members of the AP-1 family of transcription factors have been shown to induce EMT and promote tumor invasiveness (17,18). AP-1 complexes can be activated in response to TGF␤ (19 -21), physically interact with Smads (13,14), and cooperate with Smads in the control of gene expression (19 -21). In addition, several other downstream signaling pathways can also be activated by TGF␤, including p38Mapk (21), c-jun N-terminal kinase (22,23), and phosphatidylinositol 3-OH kinase (PI3K) (24,25). These signaling pathways can potentially contribute to TGF␤1-mediated EMT, but their significance for EMT and cell migration mediated by TGF␤ remains unclear.
In this study, we used the NMuMG mammary epithelial cell line as a model for TGF␤1-induced EMT (11). Two metastatic breast tumor cell lines, 4T1 and EMT6, that express high levels of TGF␤ ligands and TGF␤ receptors were used in transcription and migration studies. We report that TGF␤-induced EMT and cell migration depend on the PI3K-Akt pathway. We also show that the phosphorylation of Smad2 and transcriptional responses induced by TGF␤ are inhibited by pharmacological and molecular antagonists of the PI3K-Akt pathway. TGF␤1 can induce phosphorylation and activation of Akt/PKB in a PI3K-dependent manner, and this activation requires the Rho GTPase function. Taken together, our data suggest that PI3K-Akt signaling is required for the morphogenic, transcriptional, and migratory activities of TGF␤.

EXPERIMENTAL PROCEDURES
Antibodies and Other Reagents-TGF␤1 was from R & D Systems (Minneapolis, MN) and EGF from CLONTECH (Palo Alto, CA). Antibodies to E-cadherin and integrin ␤ 1 were from Transduction Laboratories (Lexington, KY), to p85 from Upstate Biotechnology (Lake Placid, NY), and to ZO-1 from Chemicon (Temecula, CA). Phalloidin-fluorescein isothiocyanate (actin) was from Molecular Probes (Eugene, OR). The TGF␤1-neutralizing 2G7 monoclonal IgG 2 was a gift from B. Fendly (Genentech, Inc.) and has been described previously (26). Antibodies to phospho-Ser-473 Akt and total Akt were from New England BioLabs (Beverly, MA), to Smad2 (N19) from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA), and to C-terminal phospho-Smad2 from Upstate Biotechnology. Antibodies to phospho-ERK1/2 and total ERK1/2 were from Promega (Madison, WI) and New England BioLabs, respectively. Mouse monoclonal antibodies 12CA5 and M2 to HA and Flag epitopes were from Roche Molecular Biochemicals and Sigma, respectively. Anti-Myc mouse monoclonal 9E10 antibody was a gift from J. F. Primus (Vanderbilt University). LY294002, ML7, okadaic acid, PD098059, rapamycin, U0126, and U73122 were purchased from Calbiochem (San Diego, CA). Curcumin was from Sigma. The Rac1 inhibitor SCH51344 was a kind gift from C. Kumar (Schering Research Institute, Kenilworth, NJ) (27). Adenovirus vectors encoding a dominant-negative mutant of Akt (AxAktK179D), a mutant regulatory subunit of p85 (Ax⌬p85), and a constitutively active myristoylated mutant of p110 (AxMyr-p110) were kindly provided by W. Ogawa (Kobe University School of Medicine, Kobe, Japan) (28). The pCMV6-AktK179M mutant was a gift from P. N. Tsichlis (Thomas Jefferson University, Philadelphia, PA). Plasmid vectors encoding Q61LRhoA and N19RhoA mutants were obtained from Dr. Lynn Cross (National Institutes of Health, Bethesda, MD). A plasmid vector encoding a GST-GSK3␤ peptide fusion protein was a gift from C. L. Van Den Berg (University of Colorado, Denver).
Cell Culture and Adenoviral Infection-NMuMG cells were purchased from American Type Culture Collection (Manassas, VA) and maintained in DMEM supplemented with 10% FBS and 10 g/ml insulin. 4T1 tumor cells were provided by F. Miller (Karmanos Cancer Center, Detroit, MI) and EMT6 tumor cells by B. Teicher (Lilly Research Laboratories, Indianapolis, IN); both were cultured in DMEM plus 10% FBS. For adenoviral infection of NMuMG and 4T1 cells, 10 5 cells/well in 6-well plates were transduced with adenovirus vectors at 10 -100 plaque-forming units/cell as described by Sakaue et al. (29). More than 90% of the NMuMG cells infected at a similar multiplicity of infection with an adenovirus expressing ␤-galactosidase (Ad␤-Gal) exhibited blue staining. Infected cells were subjected to further treatment 24 -48 h later.
Cell Lysis and Immunoblot Analysis-Cells were lysed in EBC buffer (20 mM Tris, pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 10% glycerol, 20 mM NaF, 1 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, and 2 g/ml leupeptin), and protein concentrations in cell lysates were determined by the Bradford method. Protein extracts (50 g/lane) were separated by 12.5% SDS-PAGE and transferred to nitrocellulose membranes (100 mA, 2.5 h). Membranes were blocked with 5% milk in TBST buffer (containing 20 mM Tris-HCl, pH 7.6, 137 mM NaCl, 0.1% Tween 20 (v/v)) for 1 h at room temperature and then incubated with primary antibodies in TBST plus 1% milk for 16 h at 4°C followed by incubation with secondary antibodies for 1 h at room temperature. Membranes were washed three times in TBST and immunoreactive bands visualized by ECL (Pierce).
Akt/PKB in Vitro Kinase Assay-Akt/PKB was precipitated from protein extracts (150 g) with GST-GSK-3␤ fusion protein immobilized on agarose beads (Sigma) or GST-agarose beads for 2 h at 4°C. An in vitro kinase reaction was performed by adding 10 Ci of [ ␥-32 P]ATP (specific activity, 3000 Ci/mmol; PerkinElmer Life Sciences) for 20 min at 30°C in the presence of 10 M PKA peptide inhibitor (Calbiochem). Reaction was terminated by the addition of 5ϫ Laemmli buffer and heating followed by 15% SDS-PAGE. Quantitative analysis of 32 P-labeled bands was performed using a PhosphorImager (Molecular Dynamics, Sunnyvale, CA).
Transcriptional Assays-NMuMG, 4T1, and EMT6 cells (0.5 ϫ 10 6 ) were seeded in 60-mm dish and transfected the following day with 0.5 g/ml p3TP-Lux (provided by J. Massague, Memorial Sloan-Kettering Cancer Center, New York, NY) or p(CAGA) 12 -Lux (provided by J.-M. Gauthier, Laboratoire Glaxo Wellcome, Les Ulis Cedex, France), each with 0.002 g/ml pCMV-Rl (Promega) using 4 l of FuGENE6 reagent (Roche Molecular Biochemicals)/g of DNA according to the manufacturer's protocol. The next day, cells were seeded in equal amounts in 24-well dishes and incubated for 16 h in low serum (0.5-2%) followed by treatment with 1 ng/ml TGF␤1 for 4 or 16 h. Firefly luciferase (Luc) and Renilla reniformis luciferase (RlLuc) activities in cell lysates were determined using the Dual Luciferase Reporter Assay System (Promega) according to the manufacturer's protocol in a Monolight 2010 luminometer (Analytical Luminescence Laboratory, San Diego, CA). Luc activity was normalized to RlLuc activity and presented as relative luciferase units. All assays were done in triplicate wells, and each experiment was repeated at least twice.
Immunofluorescent Microscopy-NMuMG cells (10 5 cells/well) were grown in DMEM, 5% FBS on glass coverslips (22 ϫ 22 mm) for 24 h before treatment with 2 ng/ml TGF␤1. Cells were fixed with methanol for 10 min at Ϫ20°C or with 2% paraformaldehyde in phosphatebuffered saline (PBS) at room temperature. For permeabilization, cells were incubated with 0.1% Triton X-100 for 5 min at room temperature. Cells were washed three times in PBS after each treatment. Cells were blocked with 3% milk in PBS for 30 min at room temperature, incubated with primary antibodies diluted in 1% milk/PBS (1/300 for ZO-1, 1/500 for integrin ␤ 1 , and 1/2000 for E-cadherin), and then incubated with fluorescent secondary antibodies (1/500) for 1 h at room temperature. Coverslips were mounted onto 25 ϫ 75-mm microslides (VWR Scientific, West Chester, PA) using AquaPolyMount (Polysciences, Warrington, PA). Fluorescent images were captured using a Princeton Instruments cooled CCD digital camera from a Zeiss Axiophot upright microscope.
Migration Assays-4T1 and EMT6 tumor cells (4 ϫ 10 4 /well) were plated in DMEM, 10%FBS in the upper chamber of 8-m pore (24-well) transwells (Corning Costar, Cambridge, MA) and incubated alone or with variable concentrations of TGF␤1 in the absence or presence of LY294002 or the TGF␤1-neutralizing 2G7 IgG 2 . Three days later, the cells that had migrated through pores and reattached to the lower chamber were trypsinized and cell numbers measured in a Coulter counter.

RESULTS
The PI3K-Akt Pathway Is Involved in EMT Induced by TGF␤1-TGF␤1 induced a mesenchymal transition in NMuMG cells within 24 h. Cells treated with 2 ng/ml TGF␤1 changed their shape from a cuboidal to a more elongated form (Fig. 1A, DIC). Concomitantly, TGF␤1 induced the delocalization of E-cadherin from adherens junctions, ZO-1 from tight junctions, and the delocalization of integrin ␤ 1 from the cell surface ( Fig. 1). There were no detectable differences in the intracellular staining of E-cadherin, ZO-1, and integrin ␤ 1 between treated and untreated cells. In addition, no detectable changes in E-cadherin were found by immunoblot analysis of whole cell extract (Fig. 1B).
To determine the signaling pathways that contribute to TGF␤-induced EMT, we examined the ability of different pharmacological agents to block the changes in cell morphology and in localization of epithelial markers at cell junctions. We found that LY294002, a synthetic inhibitor of the p110 catalytic subunit of PI3K (30), blocked the morphological transition, the delocalization of ZO-1 from cell junctions, and the reorganization of actin fibers ( Fig. 2A). Inhibitors of MEK1/2 (PD098059 ( Fig. 2A) and U0126), c-jun N-terminal kinase (curcumin), mTOR (mammalian target of rapamycin), phospholipase C (U73122), Rac1 (SCH51344), MLCK (myosin light chain kinase; ML7), and PP2A (okadaic acid) did not affect TGF␤mediated transition (data not shown), suggesting that signaling pathways associated with these molecules may not contribute to EMT mediated by TGF␤1. Inhibition of EMT by LY294002 suggested that PI3K is involved in EMT induced by TGF␤1. To further test this hypothesis, NMuMG cells were infected with adenovirus encoding a constitutively active mutant of p110 (ca-p110), the catalytic subunit of PI3K. Cells expressing Myc-tagged ca-p110 showed a higher level phosphorylation of Akt at Ser-473, confirming its functional activity (Fig. 2B). Similar to exogenous TGF␤1, infection with the ca-p110 virus resulted in the delocalization of ZO-1 from tight junctions. However, the cells retained their epithelial morphology, whereas infection with a ␤-galactosidase adenovirus (Ax␤-Gal) did not alter cell morphology nor ZO-1 staining at adherens junctions (Fig. 2B). Finally, we examined whether Akt/ PKB, a downstream effector of PI3K, would affect EMT. Transduction of NMuMG cells using a dominant-negative mutant Akt (AktK179D) adenovirus inhibited TGF␤-induced delocalization of ZO-1 from tight junctions as well as changes in cell morphology (Fig. 2C). These data suggest that the PI3K-Akt pathway is required for some of the phenotypic hallmarks associated with TGF␤-mediated EMT.
Activation of the PI3K-Akt Pathway in Response to TGF␤1-To further test that the PI3K pathway is activated by TGF␤1, we examined the phosphorylation status and kinase activity of Akt. Immunoblot analyses with antibodies specific to the phosphorylated form of Akt showed that TGF␤ induced phosphorylation of Akt at Ser-473 within 30 min, achieving a detectable maximum at 2 h (Fig. 3A). Phosphorylation of Ser-473 Akt was inhibited by 20 M LY294002 (Fig. 3A, last lane), indicating that Akt activation requires PI3K function. The activity of Akt/PKB was measured using an in vitro kinase assay with GST-GSK3␤ fusion protein containing GSK-3␤ peptide in frame with GST and immobilized on agarose beads as a substrate. Treatment of NMuMG cells with TGF␤1 for 2 h stimulated a 4-fold induction in the incorporation of 32 P into GST-GSK3␤ (Fig. 3B). Next, we tested the TGF␤1 dose dependence of phosphorylation of Akt and Smad2. Treatment with 0.5 ng/ml (20 pM) TGF␤1 was sufficient to induce a maximal phosphorylation for both Ser-473 Akt and Smad2 (Fig. 3C). TGF␤1 and EGF, a known agonist of PI3K, induced similar levels of Ser-473 Akt phosphorylation. EGF induced activating phosphorylation of ERK1/2, whereas TGF␤1 did not stimulate ERK activation at any concentration tested (Fig. 3C).
Rho-like GTPases Mediate Activation of the PI3K-Akt Pathway in Response to TGF␤1-Recent studies have suggested that RhoA is involved in TGF␤1-mediated transcription (22,23) and that TGF␤1 can activate RhoA in NMuMG cells. 2 Therefore, we tested whether RhoA GTPase affected the activation of PI3K-Akt mediated by TGF␤1. NMuMG cells transiently transfected with a dominant-negative RhoA mutant (N19RhoA) showed a significantly reduced level of Akt phosphorylation compared with a control (Fig. 3D). Transfection of the constitutively active form of RhoA (Q61LRhoA) resulted in an increase of basal phosphorylation of Akt (Fig. 3E). These results suggest that RhoA may be involved in TGF␤1-mediated activation of the PI3K-Akt pathway.
Transcriptional Responses to TGF␤1 Involve the PI3K-Akt Pathway-TGF␤ transcriptional responses can be controlled through the subcellular localization of Smads. It has been shown that SARA, a recently identified mediator of TGF␤ signaling, controls recruitment of Smad2 to TGF␤ receptors (31). The function of SARA depends on its FYVE homology domain, which binds phosphatidylinositols phosphorylated by PI3K (31). In addition, recent data have suggested that microtubules (MTs) may control Smad-dependent TGF␤1 transcriptional responses (32). It has been shown that PI3K associates tightly with ␣and ␤-tubulins (33), and it is involved in the function of MTs (34). Therefore, we next examined whether PI3K is involved in the regulation of TGF␤-mediated transcription. Two TGF␤-responsive reporter constructs were used in transcriptional assays: p3TP-Lux, containing the firefly luciferase reporter gene under the control of three 12-O-tetradecanoylphorbol-13-acetate (TPA) response elements and a fragment of the PAI-1 promoter (1), and p(CAGA) 12 -Lux, a reporter gene containing 12 repeats of Smad binding sequences from the PAI-1 promoter (35). In NMuMG cells transiently transfected with p3TP-Lux, TGF␤-mediated induction of luciferase was inhibited by LY294002 in a dose-dependent manner at 4 and 16 h (Fig. 4A). Similar results were obtained with 4T1 and EMT6 mammary tumor cell lines (data not shown). LY294002 also inhibited TGF␤-stimulated reporter activity in both NMuMG and 4T1 cells transfected with p(CAGA) 12 -Lux (Fig.  4B). We next examined whether an adenovirus vector encoding a dominant-negative mutant of p85 (dn-p85), the regulatory subunit of PI3K, would emulate the effects of LY294002. Expression of dn-p85 significantly reduced a basal phosphoryla- tion of Ser-473 Akt (Fig. 2B), confirming its functional activity. TGF␤-induced p3TP-Lux reporter activity was reduced by 75% in both NMuMG and 4T1 cells infected with the dn-p85 adenovirus vector but not with a control adenovirus encoding ␤-galactosidase (Fig. 4C). Finally, transient transfection of a dominant-negative mutant of Akt (AktK179M) markedly inhibited TGF␤-induced p3TP-Lux transcription (Fig. 4D). These data suggest that the PI3K-Akt pathway is involved in TGF␤ transcriptional responses.
TGF␤1-mediated Phosphorylation of Smad2 Requires PI3K-The transcriptional data using the p(CAGA) 12 Lux reporter (Fig. 4B) suggested that PI3K is involved in the control of Smad-dependent transcription. Therefore, we examined the effect of PI3K blockade on TGF␤-induced phosphorylation of Smad2. Immunoblot analysis with antibodies specific to Smad2 phosphorylated at the C terminus showed that C terminus phosphorylation of Smad2 was induced by TGF␤1 within 15 min, reaching a maximum by 1 h. However, co-incubation with 20 M LY294002 markedly reduced ligand-mediated Smad2 phosphorylation without detectable changes in total Smad2 protein levels (Fig. 5A). At the same time, phosphorylation of Ser-473 Akt was completely blocked by LY294002 (Fig. 5B). The induction of the C-terminal phosphorylation of Smad2 and phosphorylation of Ser-473 Akt in response to TGF␤1 appears to occur with similar kinetics and TGF␤1 dose dependence (Figs. 3 and 5). To test whether the PI3K-Akt pathway is directly involved in the C-terminal phosphorylation of Smad2, NMuMG cells were transfected with dn-Akt followed by TGF␤1 treatment and immunoblot analysis of C terminus phosphorylation of Smad2. The level of Smad2 phosphorylation was similar in control cells and cells transfected with dn-Akt, suggesting that Akt is not involved in C-terminal phosphorylation of Smad2 (Fig. 5C). Infection of cells with ca-p110 also did not induce ligand-independent phosphorylation of Smad2 (Fig. 5D).

FIG. 3. TGF␤1 activates the Akt/PKB kinase in NMuMG cells.
A, NMuMG cells were stimulated with 2 ng/ml TGF␤1 for the indicated times. After lysis in EBC buffer, protein extracts (50 g/lane) were subjected to SDS-PAGE followed by immunoblot analysis with antibodies for phospho-Ser-473 Akt (1:1000) or total Akt (1:1000). B, Akt/PKB kinase activity in protein extracts (150 g) from NMuMG cells treated with 2 ng/ml TGF␤1 was determined by an in vitro kinase assay as described under "Experimental Procedures." Phosphorylated GST-GSK-3␤ was resolved by SDS-PAGE, and 32 P incorporation was analyzed by PhosphorImager. C, immunoblot analysis of protein extracts from NMuMG cells treated with different concentration of TGF␤1 or 2 ng/ml EGF for 60 min using phospho-specific antibodies to phospho-Smad2, Akt, and ERK1/2. GST, glutathione S-transferase. D, immunoblot analysis of phospho-Ser-473 Akt and total Akt in cells transiently transfected with a dominant-negative RhoA mutant (dn-RhoA). E, immunoblot analysis of phospho-Ser-473 Akt and total Akt in cells transiently transfected with a constitutively active RhoA mutant (ca-RhoA). Therefore, we examined whether PI3K is involved in TGF␤induced cell migration. We used 4T1 and EMT6 mouse tumor cells, which exhibit high levels of TGF␤ receptors that mediate transcriptional responses (Fig. 4) but are not growth inhibited by exogenous TGF␤1. 3 TGF␤1 enhanced migration of both cell lines in a dose-dependent manner with an EC 50 of approximately 0.1 ng/ml (4 pM). LY294002 blocked both basal and TGF␤-stimulated cell migration (Fig. 6A) without an effect on tumor cell proliferation (data not shown). The TGF␤1-neutralizing 2G7 monoclonal antibody also reduced basal cell migration, suggesting that this phenotypic response was partially dependent on autocrine TGF␤ signaling (Fig. 6B). Furthermore, both LY294002 and 2G7 reduced the basal level of phosphorylation at Ser-473 Akt in 4T1 and EMT6 cells (Fig. 6C), suggesting a causal association between autocrine TGF␤ sig-naling with basal PI3K-Akt signaling and the subsequent migration of tumor cells.

DISCUSSION
The tumor-promoting activity of TGF␤1 associated with the induction of EMT has been documented for different tumor types (5)(6)(7)(8)(9). Several reports have shown that TGF␤ can induce a reversible mesenchymal transition in mammary epithelial NMuMG cells (11,12). In this study, we present data to support the role of the PI3K-Akt pathway in TGF␤-mediated EMT. We found that either the blockade of PI3K activity by a synthetic inhibitor, LY294002, or by expression of dn-Akt significantly inhibited EMT (Fig. 2). These observations led us to hypothesize that the PI3K-Akt pathway is directly involved in this transition. Similar to TGF␤1, forced expression of constitutively active PI3K (ca-p110) was sufficient to promote the disruption of cellular junctions but did not induce per se the 3 C. L. Arteaga, unpublished data. changes in cell morphology associated with EMT (Fig. 2B). The dissolution of tight junctions and the disruption of adherent junctions induced by TGF␤1 are relatively early processes, occurring within 4 -8 h after the addition of TGF␤1, whereas changes in the cell shape occur later. This result suggests that PI3K function is required for the early changes during TGF␤mediated EMT but that other events associated with the reorganization of cytoskeleton leading to changes in cell morphology may not depend on the PI3K-Akt pathway. The observed delocalization of E-cadherin, integrin ␤ 1 , and ZO-1 from cellular junctions occurred without detectable changes in their cellular content, suggesting that these TGF␤-mediated effects may involve PI3K-dependent endocytosis. These observations are consistent with the studies implicating PI3K in endocytosis and vesicular trafficking (41)(42)(43). Similar to TGF␤, hepatocyte growth factor can also disrupt epithelial cell-cell junctions and induce the delocalization of E-cadherin from cell junctions (44). In this process, hepatocyte growth factor induces the delocalization of both E-cadherin and the hepatocyte growth factor receptor, c-Met, via PI3K-mediated co-endocytosis (44). This co-endocytosis can be blocked by dominant-negative mutants of RhoA and Rab5, a component of early endosomes (44). In addition, Rab5-mediated endocytosis is also regulated by Akt/ PKB (45). Thus, TGF␤-mediated delocalization of epithelial markers from cell junctions may involve the function of PI3K-Akt and Rho-like GTPases.
The activation of PI3K in response to TGF␤ has been reported in two other cell systems (24,25). In NMuMG cells, TGF␤1 induced phosphorylation and activation of Akt/PKB with kinetics similar to the C-terminal phosphorylation of Smad2 (Figs. 3 and 5). Activation of Akt depends on PI3K, since it can be blocked by a synthetic inhibitor of PI3K (Figs. 3 and 5) and by expression of dn-p85 (Fig. 2B, inset). These results suggest that the PI3K-Akt pathway is activated directly by TGF␤1. This conclusion is further supported by recent reports showing co-precipitation of p85, the regulatory subunit of Each bar represents the mean Ϯ S.D. of 3 wells. B, 4T1 and EMT6 cells were seeded under identical conditions as described in A in the absence or presence of the TGF␤-neutralizing monoclonal antibody 2G7. Cells migrating through the 8-m pores were counted 3 days later. Data represent the mean Ϯ S.D. of 3 wells. C, exponentially growing 4T1 and EMT6 cells in DMEM, 5% FBS were incubated with 1 ng/ml TGF␤1 with or without 20 M LY294002 for 4 h. Where indicated, the 2G7 monoclonal antibody (10 g/ml) was added for 24 h. Whole cell lysates were prepared, and 50 g of total protein/lane were subjected to SDS-PAGE followed by immunoblot analyses for phospho-Ser-473 Akt and total Akt as indicated under "Experimental Procedures." PI3K, with TGF␤ receptors and stimulation of PI3K activity by TGF␤1 in other cell types (24,25). We also confirmed a direct association p85 with both type I and type II TGF␤ receptors in NMuMG cells. 4 Because of the reported role of Rho family GTPases in TGF␤1 signaling and their interaction with the PI3K pathway (46), we tested the role of the RhoA GTPase in TGF␤-mediated activation of Akt. Expression of dominant-negative N19RhoA mutant disrupted ligand-induced phosphorylation of Akt at Ser-473. On the other hand, expression of a constitutively active mutant, Q63LRhoA, resulted in an increase of the basal phosphorylation of Akt. These findings suggest that RhoA GTPase is involved in TGF␤1-mediated activation of Akt, which is consistent with recent reports that Rho-like GTPases can synergize with TGF␤ signaling (22,23). Therefore, RhoA may function as an upstream effector of Akt activation in response to TGF␤1.
Using two reporter constructs, p3TP-Lux and p(CAGA) 12 -Lux, we found that TGF␤1 transcriptional responses in NMuMG and two tumor cell lines are inhibited by both pharmacological and molecular antagonists of the PI3K-Akt pathway, including dominant-negative p85 and Akt mutants (Fig.  4, A-D). The fact that a blockade of the PI3K-Akt pathway affected Smad-dependent transcriptional responses suggested the involvement of PI3K and Akt in TGF␤ intracellular signal transduction. Consistent with this idea, we found that LY294002 significantly reduced TGF␤-mediated C-terminal phosphorylation of Smad2 in NMuMG cells (Fig. 5). However, neither PI3K nor Akt is involved in C-terminal phosphorylation of Smad2, since introduction of ca-p110 or dn-Akt did not affect it. These results, coupled with the inhibitory effect of LY294002 on Smad2 phosphorylation (Fig. 5), suggest that PI3K is involved indirectly in TGF␤-mediated C-terminal phosphorylation of Smad2.
PI3K activity may also be required for the function of intracellular mediators of TGF␤ signaling. Recently, two factors regulating C-terminal phosphorylation of Smad2 were described (31,32). First, the intracellular localization of Smad2 is controlled by SARA, a recently cloned Smad2-binding protein (31). SARA co-localizes with EEA1, an early endosome marker, 5 and this co-localization depends on the FYVE domain of SARA, which binds phosphatidylinositol 3-phosphates (47,48). It has been shown that deletion of the FYVE domain results in the mislocalization of Smad2 and inhibition of TGF␤ transcriptional responses (31). We found that Smad2 co-localizes with EEA1 in the absence of TGF␤ in NMuMG cells. 6 Thus, it is conceivable that the blockade of PI3K activity in NMuMG cells with LY294002, similar to wortmannin (49), will reduce the levels of phosphatidylinositol 3-phosphate, resulting in the mislocalization of Smad2. This is a potential explanation of the inhibitory effect of LY294002 on TGF␤-induced phosphorylation of Smad2 (Fig. 5A), whereas neither ca-p110 nor dn-Akt can directly modulate Smad2 phosphorylation (Fig. 5C, D). In addition, a recent report provides evidence that endogenous Smad2, Smad3, and Smad4 are stored in the MT network (32). It has been suggested that upon TGF␤ treatment, Smad2 and Smad3 dissociate from MT, become phosphorylated by T␤RI, and translocate to the nucleus where they regulate the transcription of TGF␤ target genes. Moreover, destabilization of MTs with nocodazole can facilitate Smad-mediated TGF␤ transcriptional responses per se in the absence of exogenous TGF␤1 (32). On the other hand, TGF␤ has been reported to stabilize MTs (50), potentially limiting Smad signaling. PI3K has also been shown to control the dynamics of the MT network, which is important for intracellular trafficking, cell motility, and other cell functions (51). Therefore, PI3K antagonists may affect the MT network and interfere with TGF␤ signaling. To formally demonstrate that PI3K blockade inhibits TGF␤ signaling via its effects on MTs will require further investigation.
Both TGF␤ and PI3K have been implicated in chemotaxis and cell migration (7, 36 -40). Here, we show that pM concentrations of TGF␤1 enhanced the basal migration of tumor cells, whereas blockade of PI3K with LY294002 reduced both basal and TGF␤-stimulated cell migration (Fig. 6, A and B). These data are in agreement with a critical role of PI3K in cell motility and migration via the modulation of cytoskeletal organization (47,51). These results were generated with tumor cells that exhibit high levels of TGF␤ expression and TGF␤ receptors as well as constitutive activation of Akt in the absence of added TGF␤ ligand. Similar to LY294002, TGF␤1-neutralizing monoclonal antibodies reduced basal cell migration and Ser-473 phosphorylation of Akt, suggesting an association between autocrine TGF␤ signaling with both constitutively activated Akt/PKB and cell invasiveness. Neither exogenous TGF␤, anti-TGF␤ antibodies, nor LY294002 had any effect on 4T1 or EMT6 cell proliferation. These data coupled with the transcription data using TGF␤ reporters in 4T1 and EMT6 cells imply that EMT can be dissociated from the anti-mitogenic effects of TGF␤. In summary, the results presented provide evidence that the PI3K-Akt pathway is causally involved in the morphogenic, transcriptional, and migratory activities of TGF␤.