Transforming Growth Factor-β1-dependent Urokinase Up-regulation and Promotion of Invasion Are Involved in Src-MAPK-dependent Signaling in Human Ovarian Cancer Cells*

Urokinase-type plasminogen activator (uPA) has been implicated in tumor cell invasion and metastasis. We reported previously that transforming growth factor (TGF)-β1 induces a dose- and time-dependent up-regulation of uPA mRNA and protein in highly invasive human ovarian cancer cell line HRA, leading to invasion. To further elucidate the mechanism of the invasive effect of TGF-β1, we investigated which signaling pathway transduced by TGF-β1 is responsible for this effect. Here, we show that 1) nontoxic concentrations of TGF-β1 activated Src kinase; 2) TGF-β1 rapidly phosphorylates ERK1/2 and Akt, but not p38; 3) pharmacological Src inhibitor PP2 or antisense (AS) c-Src oligodeoxynucleotide (ODN) treatment reduced TGF-β1-induced phosphorylation of ERK1/2 and Akt by 85–90% compared with controls; 4) pharmacological inhibition of MAPK by PD98059 abrogated TGF-β1-mediated Akt stimulation, whereas TGF-β1-induced ERK1/2 stimulation was not inhibited by PI3K inhibitor LY294002 or AS-PI3K ODN transfection; 5) up-regulation of uPA mRNA in response to TGF-β1 was almost totally blocked by PP2 and PD98059 and partially (∼55%) by LY294002; 6) TGF-β1-induced uPA mRNA up-regulation was inhibited by treatment with AS ODNs to c-Src or PI3K by 90 or 60%, respectively, compared with control ODN treatment; and 7) blockade of the release of the transcription factor NF-κB by pyrrolidinedithiocarbamate reduced the TGF-β1-induced activation of the uPA gene by ∼65%. In addition, curcumin, a blocker of the transcriptional factor AP-1, partially (35%) canceled this effect. Taken together, these data support a role for TGF-β1 activation of two distinct pathways (Src-MAPK-PI3K-NF-κB-dependent and Src-MAPK-AP-1-dependent) for TGF-β1-dependent uPA up-regulation and promotion of invasion.

The processes of ovarian cancer dissemination are characterized by altered local proteolysis, cellular proliferation, cell attachment, and invasion, suggesting that the urokinase-type plasminogen activator (uPA) 1 could be involved in the patho-genesis of peritoneal dissemination (1). uPA is a serine protease associated with various pathological conditions including tumor invasion and metastasis (2). One of the factors regulating the metastatic process is considered to be transforming growth factor-␤ (TGF-␤), which is a multifunctional cytokine that elicits numerous cellular effects pertinent to the metastatic process (1). TGF-␤ regulates a wide range of physiological and pathological cellular processes, including cell growth, differentiation, invasion, migration, mesenchymal transition, extracellular matrix synthesis, and cell death in many cell types including ovarian cancer cells (3). Recent data demonstrated that TGF-␤ specifically stimulates up-regulation of uPA mRNA and protein in certain types of neoplastic cells (1). The cellular mechanism(s) of the TGF-␤-induced uPA-dependent tumor invasion and metastasis has been extensively studied. The identity of the signaling pathway(s) involved in the TGF-␤-induced uPA up-regulation (4) remains less known. In a well characterized human ovarian cancer HRA cell model using antisense strategies and pharmacological inhibitors, we examined the distinct TGF-␤ signaling events.
Previous work from several laboratories has shown that uPA up-regulation is a major point of convergence for the actions of a variety of effectors that affect cell biology, including Src, mitogen-activated protein kinase (MAPK), phosphoinositide-3kinase (PI3K), and Smads (5). Accumulating data indicate that these pathways have been implicated in a number of phenomena associated with tumor promotion (6). The Src kinase pathway is involved in this early TGF-␤ signaling (3,7). The Smad signaling pathway constitutes a main signal transduction route downstream of TGF-␤ receptors (8). There is increasing evidence for interactions between the Smad and the MAPK pathways (9). There is precedence for the involvement of transcription factor activation in TGF-␤-mediated biological functions. For instance, TGF-␤ activates AP-1 and NF-B in the T cell line, leading to interleukin-2 expression (10). Specific inhibition of PI3K activity and down-modulation of the protein could prevent the uPA up-regulation in HRA cells. 2 PI3K is activated by a large spectrum of cytokines, growth factors, and hormones (11). This activation of PI3K is generally regulated by receptor tyrosine kinase and nonreceptor tyrosine kinase. Moreover, activation of a number of downstream signaling proteins is known to be regulated by PI3K and its lipid products including Akt, p70 S6 kinase, and Rac (12)(13)(14)(15). It was recently shown that Akt activates the transcriptional activity of NF-B (16,17). Indeed, Src, MAPK, and PI3K are present in the HRA cells, a well established model for the study of invasive phenotypic change when the cells are treated with TGF-␤1.
Here, we explored the signal transduction mechanisms downstream of the TGF-␤ receptors that result in uPA upregulation and cell invasion. However, mechanisms by which TGF-␤ up-regulates uPA expression are heterogeneous and depend upon the particular complement of signaling molecules expressed within a given cell type. We therefore investigated the effects of TGF-␤1 on the Src-dependent uPA up-regulation and invasiveness of parental and transfected HRA cells in relation to the status of MAPK, PI3K, and Smad in HRA cells. We then examined involvement of the Src, extracellular signalregulated kinase 1/2 (ERK1/2), and PI3K/Akt signaling pathways in NF-B and AP-1 activation. The key experiments were repeated with a few additional ovarian cancer cell lines to assess the potential generality of the results. These findings have important implications for our understanding of the role of TGF-␤ in up-regulating uPA expression.

Materials-LipofectAMINE
Plus reagent was purchased from Invitrogen. High molecular weight recombinant uPA and Glu-type plasminogen were obtained from American Diagnostics (Greenwich, CT). Boyden-type cell invasion chambers (BioCoat Matrigel TM invasion chambers) were obtained from Collaborative Biomedical (Franklin Lakes, NJ). Ultrapure natural human TGF-␤1 was from Genzyme (Cambridge, MA) and R&D Systems (Minneapolis, MN). The nude mice (Balb-c, nu/nu) were obtained from SLC (Hamamatsu, Japan). Culture media, penicillin, streptomycin, and fetal bovine serum were purchased from Invitrogen. Tissue culture plastics were purchased from Costar/ Corning (Cambridge, MA) and Falcon (Becton Dickinson and Co., Bedford, MA). Bovine serum albumin, Tris-base, dithiothreitol, phenylmethylsulfonyl fluoride, and ammonium persulfate were commercially obtained from Sigma. Acrylamide, bisacrylamide, and polyvinylidene difluoride membrane were from Bio-Rad. X-ray film was purchased from Eastman Kodak Co. ECL was purchased from Amersham Biosciences. Protein estimation reagents (BCA kit) were from Pierce. All other chemicals were analytical grade.
Pharmacological Inhibitors-The inhibitors were dissolved in Me 2 SO (cell culture grade; Sigma) and used in the following concentrations: wortmannin (100 nM; specific inhibitor of PI3K), LY294002 (10 M; specific inhibitor of the p110 catalytic subunit of PI3K), herbimycin A (35 nM; tyrosine kinase inhibitor), SB202190 (30 M; P38 kinase inhibitor), and PD98059 (50 M; specific inhibitor of mitogen-activated protein kinase/extracellular signal-regulated kinase kinase). All of the inhibitors except wortmannin (Sigma) were obtained form Calbiochem. The inhibitors diluted in normal growth medium were added to wells containing confluent cells and incubated for 30 min to 1 h. TGF-␤1 (10 ng/ml) was added to serum-free medium containing the respective inhibitors and incubated for the indicated periods of time, after which time the conditioned medium and cells were separately collected, and the cells were counted. The samples were stored at Ϫ80°C until measured. Me 2 SO (0.05%, v/v) diluted in medium was used as a negative control.
Cell Culture-The ovarian cancer cell line, HRA, was grown in Dulbecco's modified Eagle's medium (Sigma) supplemented with 10% (v/v) fetal bovine serum, penicillin (100 units/ml), and streptomycin (100 g/ml) in a 5% CO 2 atmosphere with constant humidity (18). SKOv-3 and OVCAR-3 ovarian cancer cells were obtained from the American Type Culture Collection (Manassas, VA). These cells were maintained in RPMI 1640 supplemented with 10% (v/v) fetal bovine serum and antibiotics. For all experiments in which TGF-␤1 was added, cells were incubated in the serum-free medium. TGF-␤1 (10 ng/ml) was added either alone or in combination in cancer cells preincubated for 30 min to 1 h with pharmacological inhibitors, curcumin (20 M, AP-1 blocker; Sigma), or pyrrolidinedithiocarbamate (PDTC; 100 M; NF-B inhibitor; Sigma). The protein concentrations in the supernatants of cell extracts were measured by the Bio-Rad protein assay. Total RNA isolations were done using the Trizol reagent (Invitrogen).
Preparation of Conditioned Medium and Cell Lysate-The cell monolayers treated with or without various agents for the indicated times were washed with phosphate-buffered saline. 1 ϫ 10 6 cells were lysed in 750 l of lysis buffer containing 20 mM Tris-HCl (pH 7.5), 12.5 mM 2-glycerophosphate, 150 mM NaCl, 1.5 mM MgCl 2 , 2 mM EGTA, 10 mM NaF, 0.5% Triton X-100, 2 mM dithiothreitol, 1 mM sodium vanadate, and 1 mM phenylmethylsulfonyl fluoride at 4°C for 15 min and scraped with a rubber policeman. Cell extracts were then centrifuged at 3,000 ϫ g to remove cell debris. All samples were stored at Ϫ80°C until use. In parallel, cells treated in the same condition in different dishes were harvested and counted using a hemocytometer.
Northern Blot Hybridization with cDNA Probes-Northern blot hybridization was carried out as described previously (19,20). Samples of total RNA (10 g) were separated by electrophoresis through denaturing 1.2% agarose gels containing 1% formaldehyde and transferred onto nylon or nitrocellulose membranes using standard molecular biological techniques. Hybridization was carried out with [␣-32 P]dCTP by random oligonucleotide priming to specific activities of 0.4 -0.9 ϫ 10 9 cpm/g. uPA cDNA was prepared as described (19,20). Filters were reprobed with the cDNA for glyceraldehyde-3-phosphate dehydrogenase to correct for the amount of RNA loaded onto the filters (19,20). After hybridization, the membranes were washed and exposed on Kodak BioMax MS-1 film at Ϫ70°C. Filters were quantitated by scanning densitometry using a Bio-Rad model 620 video densitometer with a one-dimensional Analyst software package for Macintosh.
Preparation of Oligodeoxynucleotides and Lipofection of HRA Cells-Antisense ODNs were selected for sequence target to c-Src (antisense c-Src, 5Ј-GGG CTT GCT CTT GCT GCT CCC CAT-3Ј; sense c-Src, 5Ј-ATG GGG AGC AGC AAG AGC AAG CCC-3Ј) and PI3K p85 (antisense PI3K, 5Ј-GTA CTG GTA CCC CTC AGC ACT CAT-3Ј; sense PI3K, 5Ј-ATG AGT GCT GAG GGG TAC CAG TAC-3Ј) (21). The corresponding sense ODN was used as control for each antisense ODN. Furthermore, each inverted antisense (iAS) oligonucleotide (iAS c-Src ODN or iAS PI3K ODN) provided additional controls for the vehicle and transfection. The ODNs were synthesized, purified, and modified with phosphorothioate. Oligonucleotides mixed with Lipofectin reagent were incubated for 15 min at room temperature. Thereafter, the oligonucleotide-liposome complexes were then added to cells and washed twice with medium (22). After 8 h of incubation at 37°C, the cells were collected by centrifugation, washed three times in phosphatebuffered saline, resuspended in medium plus 10% fetal bovine serum, and grown for 48 h.
Western Blot Analysis-Each cell was harvested, and cell pellets were lysed as described above. Centrifuged lysates (50 g) from each cell were analyzed by SDS-polyacrylamide gel electrophoresis and transferred to a polyvinylidene difluoride membrane by semidry transfer. Membranes were blocked for 1 h at room temperature in Trisbuffered saline containing 0.1% Tween 20 (TBST) and 2% bovine serum albumin. Blots were probed with the following primary antibodies overnight at 4°C: monoclonal anti-uPA (antibody 3471 plus 3689), anti-␤actin, anti-phospho-Src, anti-Src, anti-phospho-Akt, ant-Akt, anti-phospho-ERK1/2, or anti-ERK1/2 antibodies. This was followed by incubation with the appropriate horseradish peroxidase-conjugated secondary antibody at a dilution of 1:5,000 for 1 h. Detection was achieved by enhanced chemiluminescence (Amersham Biosciences) and exposed to film. Densitometric analysis of Western blots was carried out using the Macintosh Image System.
Cell Growth Assay-To examine the proliferation of each cell line, 5 ϫ 10 3 cells were seeded, and the number of cells in each cell line was counted in triplicate after 24 h to assess plating efficiency. Each experiment was done in triplicate.
Extracellular Matrix Invasion Assay-Chemoinvasion assays were carried out in a Boyden chamber as described (23). The upper surface of chamber was precoated with a layer of artificial basement membrane, Matrigel. The cell suspension (1 ϫ 10 5 cells/well) was added to the upper chamber. The lower chamber was filled with fibroblast-conditioned medium, which acted as a chemoattractant. To measure invasion, incubation was at 37°C for 24 h. The invaded cells in the lower side of the filter were stained with hematoxylin. Triplicate filters were used for each cell type and assay condition, and 10 random fields were counted per filter under a microscope (ϫ 400). The experiments of inhibition of cell invasiveness were performed as follows. Because wortmannin is unstable at 37°C in culture medium, it was added every 6 h to the upper chamber of the Matrigel invasion assay.
Statistics-Data are expressed as mean Ϯ S.D. of at least three independent triplicate experiments. Statistical analysis was performed by one-way analysis of variance followed by Student's t test. p Ͻ 0.05 was considered statistically significant. (4). In the present study, we examined whether TGF-␤1 induces phosphorylation of Src in HRA (lanes 1-7) and SKOv-3 (lanes 8 and 9) cells by Western blot. We observed that TGF-␤1 induced a marked rise in the level of phosphorylated Src protein in a time-dependent manner ( Fig. 1). As compared with nonstimulated HRA cells, a 6-fold increase in phosphorylated Src was observed at 20 min in response to 10 ng/ml TGF-␤1. The anti-Src antibodies immunoblotted similar amounts of Src protein in HRA cells, irrespective of whether cells were stimulated with TGF-␤1. Furthermore, TGF-␤1 (10 ng/ml, 20 min) also induces phosphorylation of Src in SKOv-3 cells (lanes 8 and 9).

TGF-␤1 Stimulates Src Phosphorylation-We have been investigating signaling pathways involved in TGF-␤1 activation
TGF-␤1 Activates Phosphorylation of Akt through PI3K-In the previous study (4), we observed concentration-and timedependent ERK1/2 phosphorylation in response to TGF-␤1 beginning with only 1 ng/ml TGF-␤1. Here, we show the ability of TGF-␤1 to induce Akt phosphorylation by probing Western blots of proteins from TGF-␤1-stimulated HRA cells using antibodies to phosphorylated Akt or total Akt, respectively. TGF-␤1 induced Akt phosphorylation in a time-dependent manner. TGF-␤1 produced an early increase (230% increase at 10 min; lane 3) of PI3K/Akt activity that plateaued after 20 min (600% of base line; lane 4) and returned to base line after 60 min (lane 6) (Fig. 2, A and B). We also observed concentrationdependent phosphorylation of Akt beginning with only 0.4 ng/ml TGF-␤1 (lane 3) (Fig. 2, C and D). This increase in Akt phosphorylation could be blocked when HRA cells were preincubated with wortmannin (100 nM, 1 h; lane 7) or LY294002 (10 M, 30 min; lane 8), respectively. Pretreatment of HRA cells with 50 M PD98059 (lane 9) also abrogated the TGF-␤1-stimulated Akt phosphorylation. The role of PI3K in Akt activation was more specifically demonstrated through inhibition with PI3K antisense ODN. We used antisense ODN targeting of the gene for PI3K and corresponding control ODNs (S PI3K ODN and iAS PI3K ODN). We evaluate by Western blot analysis, on total cell lysate, the PI3K p85 expression after transfection experiments. As expected, Fig. 2E provides data on the extent to which PI3K p85 and Src protein expression is reduced by the antisense strategy. Furthermore, antisense PI3K ODN transfection abrogated Akt phosphorylation in HRA cells (Fig. 2C, lane 11), irrespective of whether cells were treated with TGF-␤1, whereas sense PI3K ODN (Fig. 2C, lane 12) and an inverted antisense ODN (lane 13) did not. Note that TGF-␤-stimulated Akt phosphorylation was not impaired by sense PI3K ODN or inverted antisense ODN. In contrast, TGF-␤1-dependent Akt phosphorylation was not suppressed by p38 MAPK inhibitor, SB202190 (lane 10).
TGF-␤1 Stimulates Phosphorylation of ERK1/2 or Akt via an Src-dependent Mechanism-Src activation has been linked to the activation of MAPK and/or PI3K in some systems (7). We therefore examined the involvement of Src in TGF-␤1-induced activation of the MAPK or PI3K pathway. We found that pharmacological Src inhibitor PP2 inhibited the entire time course of ERK1/2 (Fig. 3A, lane 3) and Akt (Fig. 3C, lane 3) activation by 90% (up to 2 h, data not shown), indicating that Src activation is an early event in this process. To further examine the involvement of the MAPK and PI3K pathways in TGF-␤1 signaling via a Src cascade, AS ODN targeting of the gene for c-Src and corresponding control ODNs was used to treat cells with subsequent TGF-␤1 (10 ng/ml) stimulation. Indeed, Fig. 2E, lower panel, provides data on the extent to which Src protein expression is reduced by the antisense strategy. Antisense c-Src ODN (lane 4) caused a reduction in the level of phospho-ERK1/2 expression by 90%, apparently through specific reduction in c-Src (Fig. 3, A and B). Also, AS c-Src ODN (lane 4) inhibited phospho-Akt expression by 85% (Fig. 3, C and D). In contrast, S c-Src ODN (lane 5) or iAS ODN (lane 6) caused no significant effect on phosphorylation of ERK1/2 and Akt. Further, general tyrosine kinase inhibition with genistein (lane 7) or herbimycin A (lane 8) significantly attenuated TGF-␤1-induced ERK1/2 or Akt phosphorylation as did the specific Src family kinase inhibitor PP2 (Fig. 3). These results show that ERK1/2 and Akt down-regulation via pharmacological inhibition of Src and c-Src ODN treatment supports the critical position of Src upstream of MAPK and PI3K pathways.
PI3K Is a Downstream Target of MAPK-We examined the outcome of MAPK inhibitor on TGF-␤1-stimulated PI3K/Akt activation or whether PI3K inhibitors can inhibit TGF-␤1-stimulated ERK1/2 phosphorylation. Pretreatment of HRA cells for 30 min with PD98059 (Fig. 4A, lane 3) almost completely abrogated the TGF-␤1-stimulated Akt phosphorylation. In contrast, pharmacological PI3K inhibitors wortmannin (Fig. 4B, lane 4) or LY294002 (lane 5) did not change TGF-␤1-mediated ERK1/2 phosphorylation. These results suggest that ERK1/2 is involved in the signaling route leading to Akt activation and PI3K is a downstream target of ERK1/2.  1-7) and SKOv-3 (lanes 8 and 9) cells were stimulated with TGF-␤1 (10 ng/ml) for various time points as indicated. Cells were lysed with lysis buffer, and the protein content of each sample was quantitated. Each sample (50 g) was resolved by 10% SDS-PAGE and probed with anti-phospho-c-Src (p-Src; 0.5 g/ml) to detect the phosphorylated Src protein. Blots were stripped and reprobed with anti-c-Src antibody (0.5 g/ml). Experiments were repeated three times with essentially identical results. B, blots were scanned and analyzed for quantification with the Macintosh software. Band intensities for phospho-c-Src were normalized to the corresponding band intensities for total Src. Data from three experiments were averaged and are represented as the mean Ϯ S.D., expressed as -fold increase with respect to nonstimulated cells (time 0). Unlike letters (a-c) represent statistical differences (p Ͻ 0.05).

Effect of Pharmacological Inhibitors on TGF-␤1-induced
Expression of uPA mRNA and Protein-Our previous data showed that ERK1/2 phosphorylation correlated closely with TGF-␤1induced uPA expression by enzyme-linked immunosorbent assay, Western blot, and Northern blot analyses in HRA cells (4). The involvement of Src, MAPK, and PI3K in this TGF-␤1stimulated functional response was assessed by the use of the pharmacological inhibitors PP2, PD98059, wortmannin, or LY294002, respectively. As shown in Fig. 5, A and B

c-Src ODN (lane 4) or iAS ODN (lane 5)-transfected cells.
Note that TGF-␤-stimulated uPA mRNA expression was not impaired by S c-Src ODN or iAS ODN. To further assess the role of PI3K in uPA up-regulation, we transiently transfected HRA cells with PI3K ODN, in sense and antisense orientation (Fig. 6, A and B). The response of the transfected cells was consistent with a PI3K modulation. In fact, when probed by Western blot, the amount of PI3K in cells transfected with the AS construct was markedly (ϳ90%) reduced (data not shown). uPA mRNA up-regulation after TGF-␤1 treatment overlapped that of S PI3K-transfected cells (lane 7) or iAS ODN-transfected cells (lane 8). TGF-␤1-stimulated uPA mRNA expression was not impaired by S PI3K ODN or iAS ODN. On the contrary, TGF-␤1 treatment of cells transfected with the antisense construct (lane 6) resulted in abrogation of uPA expression by ϳ60% (Fig. 6, A and B).
We also found that AS-Src ODN transfection (lane 3) had almost complete inhibition of expression of the 50-kDa band corresponding to uPA in TGF-␤1-stimulated HRA cells (Fig. 6,  C and D). or S PI3K ODN (lane 7). TGF-␤-stimulated uPA protein expression was not impaired by S ODNs (lanes 4 and 7) or iAS ODNs (lanes 5 and 8). appropriate candidate, because AP-1 is activated by PI3K and MAPK and has also been involved in the induction of uPA (25). Further, Akt has been shown to up-regulate NF-B transcriptional activity in a variety of cells (26); we suspected that the up-regulation of uPA mRNA in HRA cells by Akt might occur via increased NF-B transcriptional activity. Initially, to investigate whether the AP-1 transcriptional complex was implied in the up-regulation of the uPA, experiments were performed in which uPA mRNA and protein expression was measured in the presence and absence of curcumin, an inhibitor of c-Jun/AP-1 (27). Curcumin, a natural product of the Indian spice turmeric, is a potent antioxidant and anti-inflammatory agent. As shown in Fig. 7, curcumin (lane 3) partially (by 35%) canceled the TGF-␤1-induced increase of uPA mRNA and protein expression in HRA cells. Second, to determine whether the modulation of NF-B transcriptional activity by Akt is responsible for its regulation of uPA up-regulation, we assayed expression of uPA mRNA and protein, either with or without the NF-B inhibitor PDTC (lane 4). PDTC is an antioxidant and chelator of heavy metals that blocks NF-B activity by suppressing the release of I-B␣ from NF-B. Inhibitor of the transcription factor NF-B was able to reduce TGF-␤1-induced uPA mRNA and protein expression by 65 and 80%, respectively (Fig. 7). Note that inhibition of expression by curcumin and PDTC do not specifically implicate AP-1 and NF-B transcription factors.
Assessment of Generality of the Results from HRA Cells-Key experiments were repeated with additional ovarian cancer cell lines, SKOv-3 and OVCAR-3. As shown in Fig. 8A, uPA mRNA expression in HRA cells was compared with other ovarian cancer cell lines. uPA mRNA was expressed in all cancer cell lines tested, with relatively higher levels in HRA and SKOv-3 cells and lower levels in OVCAR-3 cells. Consistent with our original findings, a 50-kDa uPA protein was detected in all cancer cells, and uPA protein expression was reproducibly higher in HRA and SKOv-3 cells than in OVCAR-3 cells (Fig.  8B). We asked whether TGF-␤1 treatment of cells results in the induction of uPA mRNA. A significant increase in the steady state level of uPA mRNA in HRA and SKOV-3 ovarian carcinoma cells (TGF-␤ at 10 ng/ml for 3 h: HRA, ϳ4-fold; SKOv-3, ϳ3-fold). There are essentially no differences in the signal transduction mechanisms downstream of the TGF-␤ receptors that result in uPA up-regulation (Fig. 8C) and cell invasion (see Fig. 9) between HRA cells and SKOv-3 cells. TGF-␤-mediated signaling pathways have been associated with the presence of Src, MAKP, PI3K, and transcription factors NF-B and AP-1; this new pathway may not be specific to only the unique cell line HRA. However, it should be noted that OVCAR-3 cells are resistant to TGF-␤ (Fig. 8, A and B). It is likely that OVCAR-3 cell-derived factors may allow cancer cells to escape TGF-␤ invasion promotion. Whether OVCAR-3 cells contain specific TGF-␤-mediated signaling pathways remains to be determined.
Effects of Pharmacological Inhibitors or Antisense Transfection on TGF-␤1-induced Invasive Response-We reported previously increased cell invasion and elevated uPA production as components of TGF-␤1-induced ovarian cancer cell invasion (4). TGF-␤1-stimulated HRA cell invasion though modified basement membrane matrix (Matrigel) requires uPA activity (4). To examine signaling pathways involved in HRA (upper part) and SKOv-3 (lower part) cell invasion, we first tested the effect of pharmacological inhibitors of Src kinase, ERK, PI3K, and p38 MAPK on spontaneous and TGF-␤1-induced invasion by an in vitro HRA cell invasion assay (Fig. 9). Inhibition of Src activation by PP2 was sufficient to reduce invasion to near basal levels. A similar finding was observed with 50 M PD98059. Interestingly, wortmannin and LY294002 have a weak effect (40 -60% inhibition). In contrast, inhibition of p38 MAPK by SB202190 does not reduce invasion.
We further examined the effect of inhibiting these signaling pathways by transfecting the HRA cells with antisense or control ODNs (sense ODN and iAS ODN) to block gene translation/ transcription directly through targeting c-Src and PI3K. The ODN-treated cells stimulated with or without TGF-␤1 (10 ng/ ml) were analyzed for in vitro invasion. Antisense c-Src ODN treatment resulted in a 70 -80% decrease in invasiveness. Antisense PI3K ODN treatment reduced invasion by 50 -60% compared with control ODNs. These data indicate that MAPK and to a lesser degree PI3K pathways contribute to an invasive response in HRA cells, irrespective of whether cells were stimulated with TGF-␤1. Invasion of HRA cells treated with or without TGF-␤1 was not impaired by S ODNs or iAS ODNs.
In a separate experiment, we showed that, similar to HRA cells, TGF-␤1-stimulated SKOv-3 cell invasion was also im- paired by pharmacological inhibitors or antisense ODNs to c-Src and PI3K. DISCUSSION Our previous findings indicated that HRA cells do not produce a detectable level of matrix metalloproteinases as determined by zymography (4) and that invasiveness of HRA cells is sensitive to pretreatment with the neutralizing antibodies directed against uPA but is resistant to anti-matrix metalloproteinase-2/9 (1) or matrix metalloproteinase-2/9 inhibitors. Our findings implicate TGF-␤ induction of uPA expression and invasion as key components of more aggressive ligand-induced invasion, although multiple proteinases seem to be important in ovarian cell invasion. These data support a role for uPA in the TGF-␤1-dependent invasive process. However, little is known about TGF-␤-dependent signaling pathways in ovarian cancer cells. Although our previous results (4) showed that TGF-␤1 stimulation results in phosphorylation of MAPK, leading to uPA up-regulation, these studies did not provide insight into other signaling pathways. The data presented here extend our previous studies (4).
To examine signaling pathways involved in ovarian cancer cell uPA production, we first tested the effect of pharmacological inhibitors of Src kinase, ERK, and PI3K on TGF-␤1-induced uPA expression. We further examined the effect of inhibiting these signaling pathways by transfecting the cells with antisense or control ODNs to block gene translation/transcrip-tion directly through targeting c-Src and PI3K. The ODNtreated cells stimulated with TGF-␤1 were analyzed for uPA expression and invasiveness. The principal finding of this study is that TGF-␤1 up-regulates uPA expression in HRA and SKOv-3 cells and facilitates invasiveness through Src-dependent MAPK and PI3K/Akt activation. Involvement of Src, ERK1/2, or PI3K in invasiveness and metastatic development has been established in many neoplastic cells including HRA cells (28). Cell spreading and invasion are prevented by the pharmacological inhibitors for these kinases (29). Another possibility is that inhibition of invasion by pharmacologic inhibitors of various kinases does not necessarily implicate loss of uPA in the invasive mechanism, and these pathways regulate many cellular behaviors relevant to motility and invasion.
We next investigated downstream events of PI3K/Akt and ERK1/2 activation. Our data using pharmacological inhibitors suggest that ERK1/2 directly regulates AP-1 activation and indirectly regulates NF-B activation via the PI3K/Akt cascade in TGF-␤1-stimulated uPA up-regulation. Therefore, our data support a role for TGF-␤1 activation of at least two distinct pathways (Src-MAPK-PI3K-NF-B-dependent and Src-MAPK-AP-1-dependent) for TGF-␤1-dependent uPA up-regulation and promotion of invasion, demonstrating that the Src-MAPK-PI3K is a main pathway for this response (Fig. 10).
A major downstream signal after MAPK/ERK activation is considered to be the phosphorylation of Elk-1/TCF transcrip- tion factors, leading to the induction of c-fos proto-oncogene expression to form the AP-1 transcriptional complex (30). It is also possible that TGF-␤ activated the Jun kinase pathway, because the AP-1 complex is usually a heterodimer formed by c-Fos and c-Jun. The use of curcumin in the present study does not allow us to exclude the concomitant activation of both pathways, because it prevents the c-Fos-c-Jun complex from binding to the AP-1 motif of DNA (27). Whatever the case is, the promoter of the uPA contains several sites of AP-1 recog-nition, thus supporting our hypothesis of a TGF-␤-MAPK/ ERK-AP-1 interaction.
Previous investigations have indicated that the PI3K/Akt is involved in preventing apoptosis induced by various growth factors in many different cell types (31), thereby promoting cell survival (32). However, our findings suggest that PI3K may be involved in other aspects of tumor progression, including invasion and metastasis. Accumulating data (33)(34)(35)(36) demonstrated that activated Akt promotes the invasion of pancreatic cancer cells. Interestingly, the catalytic subunit of PI3K has been implicated as a putative oncogene in ovarian cancer (37). There are a number of reports demonstrating the role of Src kinase in PI3K activation (38). PI3K has been shown to be directly activated by various tyrosine kinases (39,40). Nonreceptor tyrosine kinases and G protein-coupled receptors are also known to be main activators of PI3K (5,41). It is therefore likely that Src may recruit and activate PI3K via tyrosine phosphorylation of TGF-␤ receptors.
The action of PI3K in TGF-␤1-driven uPA overexpression may be mediated by the downstream targets of PI3K products, including Akt, p70 S6 kinase, cytoskeletal proteins (42), and protein kinase C (41). The link between PI3K and protein kinase C has been recently elucidated (43). Our recent data showed that, using cDNA microarray analysis, PI3K may be a candidate of target genes for bikunin, a Kunitz-type protease inhibitor, also known as an antimetastatic compound (44), and that bikunin significantly inhibits translocation of protein kinase C from plasma membrane to cytoplasm, suggesting a cross-communication between PI3K and protein kinase C (23).
TGF-␤ family members signal through transmembrane Ser-Thr kinase receptors that directly regulate the intracellular Smad pathway. Increased levels of Smad3 or Smad4 can induce apoptosis (45). In the present study, however, TGF-␤1 neither induced apoptosis (data not shown) nor cell growth inhibition in HRA cells. Therefore, the Smad pathway may not be involved in TGF-␤1-dependent signaling cascade in these cell lines. Furthermore, STAT3 might be an interesting candidate, because c-Src and PI3K have all been shown to stimulate STAT3 (24,46). To examine the Smad/STAT-mediated signaling pathways involved in HRA cell uPA production, we have been testing the effect of antisense ODNs designed to specifically block endogenous Smad/STAT gene expression at both transcriptional and translational levels. Our observations support the perspective that specifically targeting ERK1/2, or perhaps PI3K, may represent a valid approach to blocking uPA expression, cell invasion, and metastasis.
There are some conflicting data on biological functions of TGF-␤. TGF-␤ is also a potent inducer of growth inhibition in several cell types, and the TGF-␤ signaling pathway has been implicated in tumor suppression (43). It has been reported that TGF-␤ inhibited MAPK activity in the rat fibroblast cell line 3Y1 and in v-Src-transformed 3Y0 (SR-3Y1), suggesting that TGF-␤1 specifically induces degradation of activated Src kinase (24). Additionally, the previous study provides evidence that TGF-␤ in HRA and SKOv-3 cells as well as HaCaT and Madin-Darby canine kidney cells induces a rapid and transient increase in Src kinase activity. On the other hand, TGF-␤ has been reported to negatively regulate Src kinases in HepG2 and PC3 carcinoma cells (24). Although the reason for the differing results remains unknown, these data are inconsistent with our present results, which may depend on the stage of differentiation, culture conditions, and a cell type. We have no information on whether TGF-␤ can directly induce activation of Src in ovarian cancer cells. We are examining whether Shc will work as an adaptor molecule to mediate phosphotyrosine-dependent signaling events. In fact, several receptor protein-tyrosine kinases have tyrosine autophosphorylation sites, which bind to Shc, and the receptor-bound Shc becomes tyrosine-phosphorylated by the receptor and recruits another Src homology 2-containing adaptor protein to the phosphotyrosine residues. Further studies are therefore warranted to define the role of each signaling pathway in modulating the sensitivity of ovarian cancer cells to TGF-␤.
In summary, we describe two distinct pathways activated by TGF-␤1: the Src-MAPK/ERK-PI3K/Akt-NF-B and the Src-MAPK/ERK-AP-1 pathways. Furthermore, Src kinase appears to be the initial event common to the first of these two pathways (Fig. 10). Through involvement in TGF-␤1 signaling and induction of uPA, Src protein tyrosine kinases, PI3K/ Akt, ERK1/2, and transcription factors such as NF-B and AP-1 may serve as excellent therapeutic targets in the treatment of ovarian cancer and conditions characterized by uPA up-regulation.