Smad4-dependent Regulation of Urokinase Plasminogen Activator Secretion and RNA Stability Associated with Invasiveness by Autocrine and Paracrine Transforming Growth Factor-β*

Metastasis is a primary cause of mortality due to cancer. Early metastatic growth involves both a remodeling of the extracellular matrix surrounding tumors and invasion of tumors across the basement membrane. Up-regulation of extracellular matrix degrading proteases such as urokinase plasminogen activator (uPA) and matrix metalloproteinases has been reported to facilitate tumor cell invasion. Autocrine transforming growth factor-β (TGF-β) signaling may play an important role in cancer cell invasion and metastasis; however, the underlying mechanisms remain unclear. In the present study, we report that autocrine TGF-β supports cancer cell invasion by maintaining uPA levels through protein secretion. Interestingly, treatment of paracrine/exogenous TGF-β at higher concentrations than autocrine TGF-β further enhanced uPA expression and cell invasion. The enhanced uPA expression by exogenous TGF-β is a result of increased uPA mRNA expression due to RNA stabilization. We observed that both autocrine and paracrine TGF-β-mediated regulation of uPA levels was lost upon depletion of Smad4 protein by RNA interference. Thus, through the Smad pathway, autocrine TGF-β maintains uPA expression through facilitated protein secretion, thereby supporting tumor cell invasiveness, whereas exogenous TGF-β further enhances uPA expression through mRNA stabilization leading to even greater invasiveness of the cancer cells.

Metastasis is a primary cause of mortality due to cancer. Early metastatic growth involves both a remodeling of the extracellular matrix surrounding tumors and invasion of tumors across the basement membrane. Up-regulation of extracellular matrix degrading proteases such as urokinase plasminogen activator (uPA) and matrix metalloproteinases has been reported to facilitate tumor cell invasion. Autocrine transforming growth factor-␤ (TGF-␤) signaling may play an important role in cancer cell invasion and metastasis; however, the underlying mechanisms remain unclear. In the present study, we report that autocrine TGF-␤ supports cancer cell invasion by maintaining uPA levels through protein secretion. Interestingly, treatment of paracrine/exogenous TGF-␤ at higher concentrations than autocrine TGF-␤ further enhanced uPA expression and cell invasion. The enhanced uPA expression by exogenous TGF-␤ is a result of increased uPA mRNA expression due to RNA stabilization. We observed that both autocrine and paracrine TGF-␤mediated regulation of uPA levels was lost upon depletion of Smad4 protein by RNA interference. Thus, through the Smad pathway, autocrine TGF-␤ maintains uPA expression through facilitated protein secretion, thereby supporting tumor cell invasiveness, whereas exogenous TGF-␤ further enhances uPA expression through mRNA stabilization leading to even greater invasiveness of the cancer cells.
Malignant tumors are characterized by their ability to metastasize to distant organs. The initial steps of metastasis involve invasive growth of tumors across the basement membrane and migration through the extracellular matrix (ECM). 2 Because the enzymatic degradation of both the basement membrane and ECM barriers requires a number of ECM-degrading proteases (1,2) and is a critical early event in metastasis, invasiveness may be modulated by the expression of ECM-degrading proteases in tumor cells in response to autonomous and microenvironmental signals. Among the increasing number of ECMdegrading proteases implicated in metastasis, considerable attention has been focused on the family of matrix metalloproteinases (MMPs) and the plasminogen activator system. One of the regulators of these ECM-degrading proteases is transforming growth factor-␤ (TGF-␤).
TGF-␤ is a multifunctional cytokine that regulates cell proliferation, differentiation, plasticity, and migration in a contextdependent manner (reviewed in Refs. 3 and 4). TGF-␤ transduces signaling through a transmembrane type II receptor (T␤RII), a constitutively active serine/threonine kinase receptor (5). Upon ligand binding, the T␤RII recruits and transphosphorylates intracellular TGF-␤ type I receptor (T␤RI), thereby stimulating T␤RI serine/threonine kinase activity (6). The T␤RI then activates the downstream effectors, Smad2 and Smad3, by phosphorylation. The activated Smad proteins form complexes with the common Smad mediator, Smad4, and then translocate to the nucleus, where the Smad complexes regulate transcription of TGF-␤ target genes in conjunction with various transcriptional or co-transcriptional regulators. In addition to the Smad pathway, other signaling pathways, including the extracellular signal-regulated kinases (ERK1/2) (7,8), the mitogen-activated protein kinase (p38) (9,10), the Src (11), and the phosphatidylinositol 3-kinase (PI3K) (12) pathways can be activated by TGF-␤ in a context-dependent manner. The precise molecular mechanisms of regulation of these pathways for TGF-␤ signaling and the physiological and pathological roles of TGF-␤ in normal tissues and cancer have not been completely defined.
The importance of autocrine TGF-␤ in tumor progression and metastatic behavior has been documented previously. For instance, disruption of autocrine TGF-␤ signaling by a dominant-negative type II receptor (DNIIR) inhibited the invasive and metastatic potential of mammary and colon carcinoma cells (13). This was attributed to prevention of autocrine TGF-␤-induced epithelial-to-mesenchymal transition, a process believed to promote tumor cell migration and invasion (12). In a different study, overexpression of a soluble TGF-␤ type III receptor antagonized autocrine TGF-␤ activity and resulted in inhibition of tumor cell proliferation and induction of apoptosis (14).
The urokinase plasminogen activator (uPA) is a serine protease capable of initiating cascades of activation of ECM-degrading enzymes (15) and eliciting intracellular signaling through receptor binding. Clinically, elevated uPA expression in tumors is associated with tumor aggressiveness and poor outcome in patients (16,17) and numerous studies have linked uPA to invasive and metastatic phenotype of tumors in vitro and in animal models (18 -21). The metastatic MDA-MB-231 breast cancer cells secrete active TGF-␤ (22,23) and are TGF-␤-responsive (24). These cells also express both the matrix metalloproteinases-9 (MMP-9) and uPA (20,25,26). We hypothesized that autocrine TGF-␤ may function as a tumor promoter by regulating MMP-9 or uPA activity in MDA-MB-231 cells. The present study provides evidence that autocrine TGF-␤ regulates both cell invasiveness and uPA secretion. Inhibition of uPA activity is sufficient to suppress tumor cell invasion to the same extent as inhibition of autocrine TGF-␤ signaling, suggesting that autocrine TGF-␤ stimulation of invasiveness occurs via its regulation of uPA release. The Smad pathway appears to be required for the regulation of uPA release as silencing of Smad4 protein expression suppressed uPA secretion. Interestingly, although autocrine TGF-␤ regulates uPA production through protein secretion, exogenous TGF-␤ further increases uPA expression through RNA stabilization also through a Smad4-dependent fashion. Finally, this work demonstrates that a pharmacological kinase inhibitor of TGF-␤ receptors inhibits both uPA secretion and tumor cell invasiveness, thereby providing evidence for the potential efficacy of targeting TGF-␤ signaling for therapeutic intervention in cancer and suggests that uPA expression or secretion may be an important mediator of such effects.

MATERIALS AND METHODS
Cell Cultures and Reagents-MDA-MB-231 cells from the American Type Culture Collection (Manassas, VA) were maintained in the Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum in a 37°C incubator with 5% CO 2 . TGF-␤ was purchased from R & D Systems (Minneapolis, MN). The recombinant active human PAI-1 (cat. no. 1092) and human urokinase (cat. no. 124) were from American Diagnostica, Inc. (Greenwich, CT). The pharmacological inhibitor of the TGF-␤ type I receptor (LY364947) (27,28) was provided by Eli Lilly (Indianapolis, IN).
Immunoblot Analysis-To harvest protein lysates, cells were washed with cold phosphate-buffered saline (PBS) and lysed in radioimmune precipitation assay buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 1 g/ml aprotinin, 1 g/ml leupeptin, 1 g/ml pepstatin, 1 mM phenylmethylsulfonyl fluoride, 1 mM sodium orthovanadate, and 1 mM sodium fluoride) for 20 min on ice. Lysates were sonicated and then clarified by centrifugation at 15,000 ϫ g for 15 min at 4°C. Protein contents of lysates were determined by the Bradford Assay (Bio-Rad). Proteins in the lysates were resolved by SDS-PAGE and transferred to polyvinylidene difluoride membranes. Membranes were blocked in 5% milk PBS-T (0.1% Tween 20 (v/v) in PBS) for 1 h at room temperature and then probed with primary antibodies in 5% milk PBS-T overnight at 4°C. After several washes with PBS-T, membranes were incubated in PBS-T containing horseradish peroxidase-conjugated secondary antibodies for 1 h at room temperature and washed again with PBS-T. Immunoreactive bands were visualized by chemiluminescence reaction using ECL reagents (Amersham Biosciences) followed by exposure of the membranes to XAR5 films (Kodak, Rochester, NY). To detect secreted uPA, conditioned media were collected, centrifuged at 15,000 ϫ g for 5 min to remove cell debris, and then subjected to immunoblotting under non-reducing conditions. 2-to 8-h conditioned media were concentrated using Microcon YM-10 centrifugal filter devices from Millipore (Billerica, MA). The volumes of conditioned media loaded on gels were normalized to the protein concentrations of cell lysates. The fibronectin antibody was purchased from BD Transduction Laboratories, Inc. The Smad2 and phospho-Smad2 antibodies were obtained from Upstate Biotechnology, Inc. (Lake Placid, NY). The uPA antibody (cat. no. 394) was obtained from American Diagnostica, Inc. The polyadenosine diphosphate ribose polymerase and Rho GDI (guanine nucleotide dissociation inhibitor) antibodies were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). The actin and FLAG antibodies were purchased from Sigma-Aldrich, Inc.
Transient Transfection and Luciferase Reporter Assay-Cells at 50 -70% confluence on 12-well plates were co-transfected with 0.5 g of a firefly luciferase promoter-reporter construct and 0.01 g of the Renilla reniformis luciferase reference reporter construct, phRL-TK (Promega, Madison, WI) using Lipofectamine Plus reagents (Invitrogen). Four hours after transfection, cells were cultured back in regular media. Fortyeight hours after transfection, firefly and R. reniformis luciferase activities were measured using the Dual Luciferase Reporter Assay System kit (Promega) in an Optocomp II Luminometer (MGM Instruments, Inc., Hamden, CT). Normalized firefly luciferase activity was plotted as mean Ϯ S.D. from three independent experiments. The phuPA-Luc reporter containing the nucleotide sequence Ϫ2345 to ϩ30 of the human uPA promoter was kindly provided by Dr. Shuji Kojima (29). The p3TP-Lux reporter was a generous gift from Dr. Joan Massague (30).
Matrigel Invasion Assay-A modified Boyden chamber assay was performed using Transwells (12-m pore size, 12 mm in diameter) from Costar (Cambridge, MA) and Matrigel (BD Biosciences). Each Transwell insert was first coated with 100 l of 2.5 mg/ml Matrigel diluted in serum-free media for 1 h at 37°C, uPA Protein Secretion and RNA Stability through Smad4 and then 10 l of Matrigel was added in the center of the Transwell 2 h before use. Cells were trypsinized, washed with serumfree media twice, re-suspended in 0.2% bovine serum albumin serum-free medium, seeded in Transwell inserts (150,000 cells/ insert), and grown in the presence of 10% fetal bovine serum media in the lower chamber. After 16 h of incubation, Matrigel and cells remaining inside the inserts were removed with Q-tips, and the cells that had traversed to the reverse side of the inserts were rinsed with PBS, fixed in 4% formaldehyde for 30 min at room temperature, and stained with 1% crystal violet for 1 h to overnight at room temperature. Cells were counted under a light microscope (at 200ϫ power), and invasive cell numbers were the averages of those from five areas on each insert. Each invasion assay was performed in triplicate and repeated three times.
Adenovirus-mediated Overexpression of a Dominant-negative TGF-␤ Type II Receptor (DNIIR)-To amplify adenoviruses, 293T cells cultured in 5% serum/Dulbecco's modified Eagle's medium at 80% confluence on P100 plates, were infected with adenoviruses in 1 ml of fresh 5% serum medium with rocking. After 3 h, 9 ml of 5% serum medium was added in each plate without removing infection medium. Three days after infection, cells were trypsinized, collected in 1 ml of 5% fetal bovine serum medium, and subjected to three freeze/thaw cycles at Ϫ20°C/37°C. Adenovirus-containing supernatant was obtained from the cell suspension by centrifugation at 15,000 ϫ g for 20 min at 4°C and stored at Cell Proliferation Assay-Cells were seeded in 96-well plates (50,000/well), and the relative viable cell numbers were determined by MTT assay using the CellTiter 96 Non-radioactive Cell Proliferation Assay kit (Promega), following the manufacturer's protocol. MTT hydrolysis was determined by measuring the absorbance at 570 nm using a plate reader.
Preparation of Plasma Membrane Fractions-Cells were collected into buffer containing 0.15 M NaCl, 20 mM HEPES, 2 mM CaCl 2 , 100 g/ml leupeptin, 2.5 mg/ml pepstatin A, and 1 mM phenylmethylsulfonyl fluoride (pH 8.0) by scraping and then lysed by freeze/ thaw (liquid N 2 /42°C) cycles. Nuclei were isolated from the suspension of lysed cells by centrifugation at 500 ϫ g for 20 min at 4°C, washed three times, and re-suspended in radioimmune precipitation assay buffer. The nucleus-free supernatant was spun at 100,000 ϫ g for 1 h at 4°C. The resulting supernatant was compose of cytoplasmic fractions, and the pellets were subsequently washed three times with 3 ml of the cell resuspension buffer and dissolved in radioimmune precipitation assay buffer as membrane fractions.
MMP Zymography-Serum-free conditioned medium was mixed with 2ϫ sample buffer (0.5 M Tris-HCl, pH 6.8, 5% SDS, 20% glycerol, and 1% bromphenol blue) and subject to SDS-PAGE using 10% SDS-gelatin (1 mg/ml final concentration) gels under a non-reducing condition. After electrophoresis, gels were soaked in washing buffer (50 mM Tris-HCl, pH 7.5, 0.1 M NaCl, and 2.5% Triton X-100) for 1 h at room temperature to remove SDS and then in reaction buffer (50 mM Tris-HCl, pH 7.5, and 5 mM CaCl 2 , pH 8.0) overnight at 37°C. Subsequently, gels were stained in staining buffer (0.15% Coomassie Blue R250 in 10% acetic acid and 30% methanol) and de-stained in the same staining solution without Coomassie Blue R250. Clear bands of pro-and active MMP-9 (92 and 84 kDa, respectively) were observed against the blue background of stained gels. Northern Blot Analysis-Total RNA isolated with the TRIzol reagent was resolved on formaldehyde-agarose gels, transferred, and immobilized onto Hybond-N nylon membranes (Amersham Biosciences). Blots were blocked in the ULTRAHyb buffer (Ambion, Austin, TX) for 3 h at 65°C and then probed with 32 P-labeled antisense Riboprobes (5 ϫ 10 5 cpm/ml) in the ULTRAHyb buffer overnight at 68°C. After several washes with low stringency (2ϫ SSC/0.1% SDS) and high stringency (0.1ϫ SSC/0.1% SDS) buffers at 68°C, images were acquired by autoradiography using a PhosphorImager. To prepare Riboprobes, cDNA plasmids were linearized, purified, and then subject to in vitro transcription using the MAXIscript kit (Ambion) in the presence of 50 Ci of [␣-32 P]UTP (800 Ci/mmol) for 1 h at 37°C. Unincorporated nucleotides were removed using Nuc-Away columns (Ambion).
Nuclear Run-on Assay-Cell were collected, washed twice with PBS, and then re-suspended in lysis buffer (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl 2 , and 1 mM dithiothreitol). Nonidet P-40 was then added to a final concentration of 0.2-0.5%, depending on cell types. After a 5-min incubation on ice, nuclei were pelleted at 500 ϫ g for 5 min, washed once with nuclear freezing buffer (50 mM Tris-HCl, pH 8.3, 40% glycerol, 5 mM MgCl 2 , 1 mM dithiothreitol), and re-suspended in fresh nuclear freezing buffer. In vitro run-on transcription was performed using 2 ϫ 10 7 nuclei in 150 l of reaction buffer (5 mM Tris-HCl, pH 8.0, 2.5 mM MgCl 2 , 150 mM KCl, 1 mM of ATP, CTP, or GTP, 150 Ci of [␣-32 P]UTP (800 Ci/mmol), 80 units of RNasin, and 2.5 mM dithiothreitol) for 30 min at 30°C. Transcription was terminated by adding 350 l of deoxyribonuclease I solution (20 mM Tris-HCl, pH 7.4, 10 mM CaCl 2 , and 300 units of RNase-free DNase I) and a 30-min incubation at 37°C. Next, proteins were digested by adding 50 l of proteinase K solution (1% SDS, 5 mM EDTA, 1 mM Tris-HCl, pH 7.4, and 300 g/ml proteinase K) and a 30-min incubation at 50°C. The 32 P-labeled RNAs were phenol/ chloroform purified and precipitated in 10% trichloroacetic acid plus 20 g of yeast tRNA. After centrifugation at 15,000 ϫ g for 1 h, the RNAs were re-suspended in RNasefree H 2 O, denatured for 10 min at 65°C, and then chilled on ice. The radiolabeled RNAs were hybridized to cDNAs pre-immobilized on membranes in hybridization buffer (50% formamide, 5ϫ SSC, 5 mM EDTA, 5ϫ Denhardt's solution, 0.1% SDS, and 100 g/ml denatured salmon sperm DNA) for 48 h at 42°C. Next, the membranes were washed several times in 2ϫ SSC (1ϫ SSC ϭ 0.15 M NaCl and 12.5 mM sodium citrate, pH 7) and then in 2ϫ SSC plus 10 g/ml RNase A for another 30 min or not depending on intensity of background. Signals were acquired and quantified by a PhosphorImager. To immobilize cDNAs to nitrocellulose membranes, 1 g of linearized plasmid was denatured in 0.2 M NaOH for 30 min at room temperature and then neutralized with 10 volumes of 6ϫ SSC. The DNAs were applied onto nitrocellulose membranes using a slot blot apparatus and immobilized by UV cross-linking.
RNA Interference-To perform Smad4 silencing, 50% confluent cells were transfected with 50 -200 pM of a pool of four Smad4 or scrambled siRNAs (Dharmacon, Lafayette, CO) using Oligofectamine (Invitrogen) according to manufacturer's guideline. Conditioned media (48 -72 h post transfection) and protein lysates (72 h post transfection) were harvested and subjected to immunoblotting for Smad4 and uPA. To determine uPA mRNA stability under a Smad4-silencing condition, cells were first transfected with 100 nM Smad4 siRNA for 6 h and grew in serum media overnight, and then uPA mRNA stability was determined following treatment of TGF-␤ overnight.
Statistical Analysis-p values for multiple comparison tests were derived by an analysis of variance (ANOVA) with a Bonferroni correction.
The inhibitory effect of LY364947 on TGF-␤ signaling was first validated by examining Smad2 phosphorylation. Basal levels of Smad2 phosphorylation were undetectable by immunoblotting. However, LY364947 abolished Smad2 phosphorylation induced by exogenous TGF-␤ without altering total Smad2 protein levels (Fig. 1A). Expression of fibronectin is induced by TGF-␤ through a Smad-independent pathway (31). LY364947 decreased both basal and exogenous TGF-␤-induced fibronec-tin expression (Fig. 1B). In addition, we evaluated the effect of LY364947 on TGF-␤-induced promoter activation by reporter assays using p3TP-Lux, a luciferase reporter construct highly responsive to TGF-␤ (30) and observed that LY364947 significantly inhibited both basal and exogenous TGF-␤induced promoter activation (Fig.  1C). To investigate whether autocrine TGF-␤ has a role in regulation of MDA-MB-231 invasiveness, Matrigel invasion assays were performed with or without LY364947 treatment. LY364947 inhibited cell invasion in a dose-dependent manner (Fig. 1D). These data suggest that autocrine TGF-␤ plays a role in basal invasive growth of MDA-MB-231 cells.
A Dominant-negative TGF-␤ Type II Receptor Suppresses Autocrine TGF-␤ Signaling and Cell Invasion-To substantiate the results obtained using the TGF-␤ receptor kinase inhibitor, we assessed invasiveness of MDA-MB231 cells following suppression of autocrine TGF-␤ signaling with a dominantnegative TGF-␤ type II receptor (T␤RII). T␤RII is the receptor responsible for ligand binding and for activation of TGF-␤ type I receptor through its kinase activity. T␤RII devoid of the kinase domain (DNIIR) acts as a dominant-negative mutant by competing with wild-type receptors for TGF-␤ ligands (32). Expression of FLAGtagged dominant-negative DNIIR was achieved using an adenoviral vector and was confirmed by immunoblotting for FLAG ( Fig. 2A). DNIIR expression was increased with increasing amounts of adenovirus, whereas no DNIIR was detected in parental or the ␤-galactosidase adenovirus-infected cells. Expression of DNIIR inhibited TGF-␤-stimulated Smad2 activation (Fig. 2B) and decreased both basal and exogenous TGF-␤-induced fibronectin expression (Fig. 2C). DNIIR also suppressed basal and exogenous TGF-␤-stimulated p3TP-Lux promoter activation (Fig. 2D), demonstrating the inhibitory effects of DNIIR on TGF-␤ signaling. As expected, DNIIR expression significantly decreased MDA-MB-231 cell invasion (Fig. 2E). This effect is not a result of inhibition of cell proliferation as determined by MTT assay (data not shown). Thus, consistent with the results using the pharmacological inhibitor, inhibition of autocrine TGF-␤ signaling by the dominant-negative TGF-␤ type II receptor (DNIIR) down-regulated MDA-MB-231 invasiveness. , and subjected to immunoblotting for uPA. C, the effect of LY364947 on uPA protein levels on cellular membrane. Cells were treated with 5 M LY364947 or 5 ng/ml TGF-␤ for 24 h. Membrane, cytoplasmic, and nuclear fractions were isolated as described under "Materials and Methods" and subjected to immunoblotting for uPA, polyadenosine diphosphate ribose polymerase (a nuclear marker), Rho guanine nucleotide dissociation inhibitor (GDI) (a cytoplasmic marker), and ␤-actin. D, down-regulation of uPA secretion by the dominant-negative TGF-␤ type II receptor (DNIIR). Cells were infected with ␤-galactosidase (Ad-␤-Gal) or DNIIR (Ad-DNIIR) adenoviruses at the indicated m.o.i. values. 48 h later, cells were cultured in fresh medium and allowed to grow for 16 h. Conditioned media, protein lysates, and total RNAs were harvested and subjected to immunoblotting for uPA and ␤-actin and Northern blotting for uPA and cyclophilin. E, the effect of LY364947 on pro-MMP-9 and active MMP-9 protein levels. Cells were treated with LY364947 at the indicated concentrations for 24 h. Condition media were collected and subjected to gelatin zymography.

Disruption of Autocrine TGF-␤ Signaling Suppresses uPA
Secretion-The MDA-MB-231 cells secrete uPA (20). To address whether autocrine TGF-␤ can modulate uPA expression, uPA levels were determined after blockade of autocrine TGF-␤ signaling using LY364947. Immunoblotting for uPA was conducted under non-reducing conditions, which detect both uPA and complexes of uPA and its inhibitor, plasminogen activator inhibitor-1 (PAI-1) (33). The low motility bands with molecular masses around 100 kDa represent uPA⅐PAI-1 complexes (uPA ϳ55 kDa and PAI-1 ϳ52 kDa). Our results show that LY364947 decreased uPA secretion in a dose-dependent manner (Fig. 3A) and that the decreased level of free uPA in the conditioned medium did not result from increased uPA/PAI-1 association and depletion of uPA. The kinetics of the LY364947 effect was determined by a time-course experiment from 2 to 24 h after treatment with 5 M LY364947. The result (Fig. 3B) shows that uPA secretion was inhibited starting 8 h after treatment. MDA-MB-231 cells constitutively express uPA receptor mRNA and exhibit detectable membrane-associated uPA (20). To determine whether the decreased level of released uPA after LY364947 treatment was the result of increased uPA receptor expression and increased uPA binding to uPA receptor, levels of membrane-bound uPA were determined after LY364947 treatment. The purity of the membrane fractions was confirmed by the absence of nuclear and cytoplasmic proteins, polyadenosine diphosphate ribose polymerase, and Rho GDI (guanine nucleotide dissociation inhibitor), respectively (Fig.  3C). Our results show that neither exogenous TGF-␤ treatment nor LY364947 treatment altered membrane uPA levels (Fig.  3C). Inhibition of autocrine TGF-␤ signaling by expression of DNIIR also suppressed uPA secretion in a dose-dependent fashion (Fig. 3D). In contrast, ␤-galactosidase adenovirus infection did not alter uPA secretion, further supporting that inhibition of autocrine TGF-␤ signaling suppresses uPA secretion. Of note, expression of DNIIR was not detectable at an m.o.i. of 10 by immunoblotting ( Fig. 2A); however, this low level of DNIIR expression was sufficient to decrease uPA secretion (Fig.   3D). These data demonstrate that disruption of autocrine TGF-␤ signaling inhibited uPA secretion.
Steady-state levels of uPA mRNA did not change with either LY364947 treatment or DNIIR expression (Fig. 3, A and D). Further, whereas uPA release into the medium was decreased, the intracellular uPA were increased after blockade of autocrine TGF-␤ signaling (Fig. 3, A and D), suggesting that autocrine TGF-␤ stimulates uPA secretion and that inhibition of autocrine TGF-␤ signaling reduces the level of secretion without impairing uPA production, thus leading to intracellular accumulation of uPA. We also examined uPA production in response to TGF-␤ in the Panc-1 human pancreatic cancer cell line and found that TGF-␤ treatment stimulated overall uPA production (free uPA plus uPA in complex with PAI-1), whereas LY364947 decreased overall uPA production with no change in uPA mRNA levels (Fig. 4A). In this cell line, total uPA levels also correlate with invasive activity (Fig. 4B). These data demonstrate that the TGF-␤-regulated uPA secretion is not limited to a single cell line, and the Panc-1 produced TGF-␤ also regulated uPA production and invasion in an autocrine manner.
Inhibition of Autocrine TGF-␤ Signaling Does Not Affect MMP-9 Protein Levels and Activity-MDA-MB-231 cells express the MMP-9 protein (20) that is particularly important among at least 19 MMP proteins identified to date in tumor invasion and metastasis due to its ability to degrade the basement membrane component, type IV collagen. We determined MMP-9 protein levels and activity following abrogation of autocrine TGF-␤ signaling with LY364947. Gelatin zymography shows no change in both pro-(92 kDa) and active MMP-9 (84 kDa) protein levels after LY364947 treatment (Fig. 3E). These results indicate that autocrine TGF-␤ does not regulate MMP-9 activity or protein expression in the MDA-MB-231 cells. Consistent with previous reports (20), we did not detect MMP-2 (72 kDa) expression in MDA-MB-231 cells.
Inhibition of Basal uPA Activity Impairs MDA-MB-231 Cell Invasion-Inhibition of autocrine TGF-␤ signaling resulted in decreased invasiveness and uPA secretion as described above. We next investigated a relationship between uPA activity and MDA-MB-231 cell invasiveness. Inhibition of uPA activity using an anti-catalytic uPA blocking antibody attenuated cell invasion by ϳ70% as compared with no treatment or IgG treatment in Matrigel invasion assays (Fig. 5A). These data suggest a correlation between uPA activity and MDA-MB-231 cell invasiveness. Parallel MTT assays suggest that the decreased cell invasion was not due to inhibition of cell proliferation (Fig. 5B). To test the hypothesis that uPA is the molecule that mediates the autocrine TGF-␤ pro-invasive effect, we examined whether addition of uPA in the medium can reverse the inhibitory effect

uPA Protein Secretion and RNA Stability through Smad4
of LY364947 on cell invasion. Our result shows LY364947 failed to decrease cell invasion in the presence of recombinant uPA (Fig. 5C), suggesting that uPA is the cellular molecule that mediates TGF-␤-regulated cell invasion. Interestingly, addition of recombinant human PAI-1, an inhibitor of uPA, did not decrease basal cell invasion. As PAI-1 has been previously shown to have a positive role in cell attachment and invasion (34), and concomitant elevation of uPA and PAI-1 has been observed in cancer and linked to poor outcome (35), our data support the hypothesis that invasive potential can be up-regulated by increased uPA expression in conditions with elevated PAI-1 expression.
Smad4 RNA Interference Decreases uPA Secretion-TGF-␤ receptors transduce signals through both Smad and non-Smad pathways, including the extracellular signal-regulated kinases (ERK1/2) (7,8), the p38 mitogen-activated protein kinase (9,10), the phosphatidylinositol 3-kinase (PI3K) (12), and the Src (11) pathways. Whether autocrine TGF-␤ regulates these Smadindependent pathways was evaluated by immunoblotting. LY364947 treatment (up to 10 M) did not significantly alter activation status of ERK1/2, p38, Akt, and Src proteins (Fig. 6A). In contrast, as little as 0.1 M LY364947 suppressed uPA secretion (Fig. 3A), implicating that these pathways are probably not involved in autocrine TGF-␤ regulation of uPA secretion. We hypothesized that the Smad pathway mediates the autocrine TGF-␤ effect on uPA secretion. In this case, down-regulation of the common Smad mediator, Smad4, should suppress uPA release in response to autocrine TGF-␤ signaling. Smad4 silencing was achieved using a mixture of four Smad4 siRNAs. Immunoblotting results confirmed down-regulation of Smad4 protein expression in Smad4 siRNAtransfected cells but not in parental or scrambled control siRNA-transfected cells (Fig. 6B). Smad2 protein expression was unaffected by the Smad4 siRNA transfection, further confirming specific silencing of Smad4 protein expression by the Smad4 siRNAs. Interestingly, uPA secretion was also decreased by inhibition of Smad4 protein expression (Fig. 6B), suggesting that the Smad pathway is required for the regulation of uPA secretion by autocrine TGF-␤. To further confirm the Smad4 role in uPA secretion, we tested whether forced expression of Smad4 can rescue the inhibitory effect of Smad4 silencing on uPA secretion. Fig. 6C shows that adenovirus-mediated overexpression of Smad4 prevented the inhibited uPA secretion induced by Smad4 silencing.
Paracrine TGF-␤ Increases uPA mRNA Levels through RNA Stabilization-Paracrine/exogenous TGF-␤ was previously shown to stimulate uPA expression and increased MDA-MB-231 cell invasiveness (20). We further examined uPA regulation by exogenous TGF-␤. Unlike autocrine TGF-␤, exogenous TGF-␤ increased both uPA mRNA and protein levels in a dosedependent manner (Fig. 7A). A time-course study shows that uPA mRNA was induced by TGF-␤ in a time-dependent fashion to near maximal levels by 16 h after treatment, and the level was sustained for at least 48 h (Fig. 7B). Increased mRNA . The Smad pathway is required for uPA secretion. A, LY364947 did not alter the activation status of ERK1/2, Akt, p38, and Src proteins. Cells were treated with 0, 1, and 10 M LY364947 for 24 h and then lysed. Immunoblotting of the lysates was performed using the antibodies against total and phospho-ERK1/2, Akt, p38, and Src proteins. B, Smad4 silencing decreased uPA secretion. Cells were transiently transfected with Smad4 or scrambled control siRNAs at the indicated concentrations for 24 h. Cells were then allowed to grow in fresh medium for another 24 h. Conditioned media and lysates were harvested and subjected to immunoblotting for uPA, Smad2, Smad4, and ␤-actin. C, forced expression of Smad4 blocked the inhibitory effect of Smad4 silencing on uPA secretion. Cells were transfected with Smad4 siRNAs. On the next day, cells were infected with ␤-galactosidase or Smad4 adenoviruses at the indicated m.o.i. 48 h after viral infection, conditioned media and lysates were harvested and subjected to immunoblotting for uPA, Smad4, and ␤-actin.
uPA Protein Secretion and RNA Stability through Smad4 NOVEMBER 10, 2006 • VOLUME 281 • NUMBER 45 expression is due to either increased transcription or increased RNA stability. The transcription of uPA in response to exogenous TGF-␤ was determined by a nuclear run-on assay. Exogenous TGF-␤ failed to enhance uPA transcription (Fig. 8A) but strongly stimulated transcription of PAI-1, a TGF-␤ target gene that has been shown to be transcriptionally activated by TGF-␤ (36). Consistent with the nuclear run-on result, transient expression assays using a uPA-promoter-reporter construct containing the nucleotide Ϫ2345 to ϩ30 region of the human uPA promoter (phuPA-Luc) show very little changes in uPA promoter activity after TGF-␤ treatment (Fig. 8B). In contrast, uPA promoter activity was significantly induced by phorbol 12-myristate 13-acetate, a known inducer of activation of the human uPA promoter (37) (Fig. 8B). Next, the stability of uPA mRNA in the presence and absence of exogenous TGF-␤ was determined by examining uPA mRNA levels at various time points after blocking transcription with the RNA polymerase II-specific inhibitor DRB (5,6-dichloro-1-␤-D-ribofuranosylbenzimidazole). Quantitation of the Northern blotting results (Fig. 8C) shows that by 8 h after DRB treatment, uPA mRNA levels were decreased by 50% in untreated cells. In contrast, uPA mRNA levels were only decreased by 25% in TGF-␤treated cells, suggesting that exogenous TGF-␤ enhances the stability of uPA mRNA. Thus, exogenous TGF-␤ increased uPA mRNA levels through mRNA stabilization in MDA-MB-231 cells.
Smad4 Is Required for Paracrine TGF-␤-mediated Stabilization of uPA mRNA-The engagement of the Smad pathway in the paracrine/ exogenous TGF-␤-induced effect on uPA stability was determined. We first examined steady-state levels of uPA mRNA following TGF-␤ treatment in Smad4 knock-down cells. Suppressed Smad4 protein expression was first confirmed following Smad4 siRNA transfection with or without TGF-␤ treatment (Fig. 9A). Interestingly, uPA mRNA induction by TGF-␤ was not observed in the Smad4 siRNAtransfected cells, suggesting that the uPA mRNA-stabilizing effect of exogenous TGF-␤ is lost in the absence of Smad4. We next examined the stability of uPA mRNA following TGF-␤ treatment in control or Smad4 siRNA-transfected cells and observed that TGF-␤ failed to stabilize uPA mRNA in Smad4 siRNA-transfected cells (Fig. 9, A and B). These data suggest that the Smad pathway is required for uPA mRNA stabilization induced by exogenous TGF-␤. There was not complete loss of the TGF-␤ effect on uPA mRNA stability in Smad4 siRNA-transfected cells, presumably attributed to undetected residual expression of Smad4 protein under the Smad4 silencing condition.

DISCUSSION
TGF-␤ is a potent inhibitor of epithelial cell growth through inhibition of proliferation and induction of apoptosis (38, 39)  and is an important tumor suppressor (40). The tumor suppressor role of TGF-␤ is evident in that mice heterozygous for deletion of the TGF-␤ gene with expression of 10 -30% of TGF-␤ levels of wild-type animals, developed an increased number of chemically induced tumors than did wild-type littermates (40). However, escape from the growth inhibitory effects of TGF-␤ occurs frequently in cancer through numerous mechanisms. Moreover, TGF-␤ displays tumor promoting effects in late-staged tumors. TGF-␤ has been shown to facilitate tumor progression by inducing epithelial to mesenchymal transition (41,42), ECM degradation (20), and cyclooxygenase-2 expression (43) or by inhibiting antitumor immune responses (44).
Disruption of tumor autocrine TGF-␤ signaling has been found to delay tumor growth and inhibit metastases (13,45). However, the molecular mechanisms underlying the tumorigenic effects of autocrine TGF-␤ remain unclear. The present study demonstrates that autocrine TGF-␤ regulates both cell invasiveness and uPA secretion. Inhibition of uPA activity decreases tumor cell invasion to the same extent as does the inhibition of autocrine TGF-␤ signaling. Interestingly, although autocrine TGF-␤ regulates uPA availability via effects on protein secretion, paracrine/exogenous TGF-␤ further increases uPA availability by increasing uPA mRNA through RNA stabilization. Our data suggest two distinct levels of regulation of uPA in response to different magnitudes of stimulation by TGF-␤. The amount of active TGF-␤ in MDA-MB-231 cellconditioned media has been quantitated at ϳ0.25 ng/ml (22, 23, 46), whereas at least 10-fold higher concentrations were required to increase uPA mRNA levels (see Fig. 6A).
In the secretory pathway, proteins are transported between intracellular compartments via membranous vesicles that the process involves vesicle formation, motility, and docking, and membrane remodeling and fusion (47,48). Numerous proteins have been implicated in regulation of vesicle trafficking during protein secretion. For example, members of Ypt/Rabs proteins of the Ras GTPase superfamily have been characterized to be key regulators of protein transport (49 -51). A negative role for Rab3, a member of the Ypt/Rab family, in protein secretion has been previously shown by a number of studies (reviewed in Ref. 52). This function of Rab3 suggests that depletion of Rab3 may lead to enhanced protein secretion. Recently, one of the biolog-icalfunctionsofTGF-␤hasbeenshowntobefacilitatingubiquitindependent degradation of protein in a Smad-dependent manner (53). These observations raise the possibility that TGF-␤ signaling may modulate uPA protein secretion by depleting negative regulators of protein secretion through facilitated protein degradation.
Most of TGF-␤ activities involve modulation of gene transcription. However, TGF-␤ has been shown to substantially increase stability of mRNAs such as ␣1(I) collagen mRNA in human hepatic stellate cells (54) and elastin mRNA in human fetal lung fibroblasts (55). In both cases, activation of the p38 MAPK signaling pathway is required for the TGF-␤ effect. Interestingly, expression of inhibitory Smad7, an inhibitor of TGF-␤ signaling but not Smad6, an inhibitor of bone morphogenetic protein action blocked TGF-␤ signaling and dramatically diminished the TGF-␤-stabilizing effect on elastin mRNA (55). Theses results suggests the involvement of both Smad and p38 MAPK pathways in the TGF-␤-mediated stabilization of elastin mRNA. In our study, exogenous TGF-␤ stabilized uPA mRNA in a Smad4-dependent manner. Exogenous TGF-␤ also induced activation/phosphorylation of p38 MAPK (data not shown) in the same cell context. Therefore, it is conceivable that the p38 signaling pathway may participate in the process of TGF-␤ regulation of uPA mRNA stability in conjunction with the Smad pathway.
Stabilization of mRNA involves the binding of RNA proteins to certain cis-elements of mRNAs (56). Adenylate-uridylaterich elements are important regulatory cis-elements present in the untranslated regions of short-lived mRNAs such as protooncogenes, cyclooxygenase-2 (57), and c-fos (58) mRNAs. TGF-␤ and Ras have been shown to synergistically stabilize the COX-2 mRNA through an adenylate-uridylate-rich element in the proximal 3Ј-untranslated regions (57). Given that MDA-MB-231 cells possess an activating K i -Ras mutation (59) and that adenylate-uridylate-rich elements are present in the 3Ј-untranslated region of the uPA mRNA (60), it is possible that TGF-␤ may cooperate with the active Ras to regulate the stability of uPA mRNA through a similar mechanism, but confirmation of this awaits further investigation.
Cells release uPA as a single-chain zymogen. The low level of intrinsic proteolytic activity of pro-uPA (61) can convert the plasminogen in tumor microenvironments or Matrigel (62) to plasmin, which in turn activates pro-uPA. This pro-uPA activation by plasmin and activation of plasminogen by uPA in a cyclic fashion promotes degradation of the ECM or the basement membrane and facilitates cell invasion. In addition, plasmin can potentially activate MMPs (15,63,64) thereby promoting ECM degradation and tumor cell invasion (65). However, we did not find evidence for up-regulation of MMP-9 activity or expression by TGF-␤ in the MDA-MB-231 cells under the conditions studied.
TGF-␤ is released from cells mostly in a latent, inactive form via a constitutive secretion pathway (66). Despite the predominance of latent TGF-␤ in conditioned media in general, MDA-MB-231 cells express detectable active TGF-␤ (46). It is interesting that uPA can proteolytically activate latent TGF-␤ (67). Therefore, TGF-␤ regulation of uPA production may be a positive feedback loop for activation of latent TGF-␤ (Fig. 8), and this relationship can be a cycle in cancer progression because TGF-␤ is overexpressed in both malignant breast tumors and surrounding stroma (68,69), and uPA expression is increased in human breast carcinomas and bone metastases (70). It will be of interest to determine whether the basal level of uPA secretion contributes to the availability of active TGF-␤ and autocrine TGF-␤ signaling.
The plasminogen activator inhibitor (PAI-1) is a TGF-␤ target gene and a strong inhibitor of uPA. The present study demonstrates that TGF-␤ increases levels of both uPA and PAI-1 in the MDA-MB-231 conditioned media. Interestingly, despite the fact that PAI-1 inhibits uPA activity, concomitant elevation of uPA and PAI-1 has been observed in breast cancer and is associated with poor outcome (35), suggesting that tumor progression may occur in the presence of high levels of PAI-1. TGF-␤ has been shown to stimulate attachment and invasion through up-regulation of PAI-1 (34), and these effects may contribute to cell invasion. Our findings suggest that, although PAI-1 is induced by TGF-␤, this induction is not sufficient to prevent the invasion promoting effect of uPA.
In summary, autocrine TGF-␤ regulates cell invasiveness through maintaining uPA levels by facilitated protein secretion, whereas paracrine/exogenous TGF-␤ further increases invasiveness through stimulated uPA expression by RNA stabilization. The Smad pathway appears to be required for the distinct levels of regulation of uPA in response to different magnitudes of TGF-␤ stimulation.