Autocrine Transforming Growth Factor-β Signaling Mediates Smad-independent Motility in Human Cancer Cells*

Transforming growth factor-β (TGF-β) is a pleiotropic growth factor that plays a critical role in modulating cell growth, differentiation, and plasticity. There is increasing evidence that after cells lose their sensitivity to TGF-β-mediated growth inhibition, autocrine TGF-β signaling may potentially promote tumor cell motility and invasiveness. To understand the molecular mechanisms by which autocrine TGF-β may selectively contribute to tumor cell motility, we have generated MDA-MB-231 breast cancer cells stably expressing a kinase-inactive type II TGF-β receptor (TβRII-K277R). Our data indicate that TβRII-K277R is expressed, can associate with the type I TGF-β receptor, and block both Smad-dependent and -independent signaling pathways activated by TGF-β. In addition, wound closure and transwell migration assays indicated that the basal migratory potential of TβRII-K277R expressing cells was impaired. The impaired motility of TβRII-K277R cells could be restored by reconstituting TGF-β signaling with a constitutively active TGF-β type I receptor (ALK5TD) but not by reconstituting Smad signaling with Smad2/4 or Smad3/4 expression. In addition, the levels of ALK5TD expression sufficient to restore motility in the cells expressing TβRII-K277R were associated with an increase in phosphorylation of Akt and extracellular signal-regulated kinase 1/2 but not Smad2. These data indicate that different signaling pathways require different thresholds of TGF-β activation and suggest that TGF-β promotes motility through mechanisms independent of Smad signaling, possibly involving activation of the phosphatidylinositol 3-kinase/Akt and/or mitogen-activated protein kinase pathways.

dysregulated expression or mutational inactivation of various components of the TGF-␤ signaling pathway can contribute to tumorigenesis (21)(22)(23). In addition, there is increasing evidence that after cells lose their sensitivity to TGF-␤-mediated growth inhibition, autocrine TGF-␤ signaling may promote tumorigenesis. The importance of autocrine TGF-␤ signaling in tumor progression has been highlighted by several studies that have shown that expression of a dominant-negative type II TGF-␤ receptor (dnT␤RII) in various tumor cells can prevent the conversion of cells from an epithelial to an invasive mesenchymal phenotype, delay tumor growth, and reduce metastases (24 -27). These data suggest that TGF-␤ can act directly on tumor cells to promote tumor maintenance and progression. In addition to promoting epithelial to mesenchymal transformation of tumor cells, TGF-␤ can stimulate the motility of many cell types in vitro, suggesting that TGF-␤ production in vivo may enhance migration of tumor cells and thus contribute to tumor invasiveness and metastases. There is also evidence that TGF-␤ can increase cellular motility without affecting proliferation, suggesting that the effects on motility and proliferation may occur via different biochemical pathways (28).
To understand the molecular mechanisms by which autocrine TGF-␤ may selectively contribute to tumor cell motility, we have generated MDA-MB-231 breast cancer cells stably expressing dnT␤RII. MDA-MB-231 cells express TGF-␤ receptors (29), secrete TGF-␤ (30), and, although they are resistant to the growth inhibitory effects of TGF-␤ (29), can respond to TGF-␤ with an increase in spreading (31) and invasiveness (32). In addition, there is evidence that blocking TGF-␤ signaling by administration of a neutralizing TGF-␤ antibody can inhibit MDA-MB-231 cell tumorigenicity and metastases in nude mice (33). In this paper we show that expression of dnT␤RII in MDA-MB-231 cells impairs their basal migratory potential. This impairment in motility can be restored by expression of a constitutively active type I TGF-␤ receptor (ALK5 TD ) but not by overexpression of Smad2/4 or Smad3/4. In addition, the levels of ALK5 TD expression sufficient to restore motility in the cells expressing dnT␤RII are associated with an increase in phosphorylation of Akt and ERK1/2, but not Smad2, suggesting that Smad signaling is dispensable for autocrine TGF-␤-mediated motility and that this response depends on alternative signaling pathways activated by TGF-␤.
Generation of Stable Cell Lines-To generate MDA-MB-231 cells stably expressing dnT␤RII, we obtained a construct encoding a kinaseinactive T␤RII mutant in which the lysine at position 277 has been mutated to arginine (pGABE-T␤RII-K277R) (25) from Martin Oft (UCSF, San Francisco, CA). Lysine 277 corresponds to an invariant lysine found in the ATP-binding site of subdomain II in all protein kinases, and even its substitution with arginine results in loss of kinase activity (36). The pGABE vector is a modified version of the commonly used retroviral vector pBABE in which the puromycin cassette has been replaced by enhanced green fluorescent protein (GFP). In this construct, T␤RII-K277R is HA-tagged, and its expression is driven by the viral long terminal repeat, whereas expression of GFP is driven by the SV40 promoter. MDA-MB-231 cells were transfected with the control pGABE vector or the pGABE-T␤RII-K277R vector utilizing Lipo-fectAMINE reagent (Invitrogen) according to the manufacturer's instructions. Following transfection, cells expressing GFP were sorted by flow cytometry. Clones were then isolated by sorting individual cells from the Ͼ95% positive GFP pool.
Affinity Labeling of Cells with 125 I-TGF-␤1 and Immunoprecipitation of HA-tagged T␤RII-K277R-125 I-TGF-␤1 was obtained from PerkinElmer Life Sciences. Near confluent MDA-MB-231 cells, as well as clones and pools stably expressing GFP alone (GABE) or GFP and T␤RII-K277R (dnT␤RII) in 12-well plates, were washed three times over 30 min with 500 l of ice-cold 0.1% bovine serum albumin dissolved in Dulbecco's phosphate-buffered saline (D-PBS) containing Ca 2ϩ and Mg 2ϩ . The cells were then affinity labeled with 100 pM 125 I-TGF-␤1 as described previously (37), with slight modifications. Briefly, after a 3-h incubation with 100 pM 125 I-TGF-␤1 at 4°C, the cells were washed with 500 l of ice-cold D-PBS, and the ligand-receptor complexes were crosslinked with 400 l of 1 mM bis(sulfosuccinimidyl)suberate (BS 3 ; Pierce) for 10 min on ice. The cross-linking reaction was stopped with the addition of 100 l of 500 mM glycine. Cells were washed twice with 500 l of D-PBS and solubilized with 125 l of 20 mM Tris buffer, pH 7.4, containing 1% Triton X-100, 10% glycerol, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, 2 g/ml leupeptin, 2 g/ml pepstatin, and 2 g/ml soybean trypsin inhibitor. Solubilized material was centrifuged for 10 min at 4°C to pellet cell debris. The supernatants were transferred to one-fifth volume of 5ϫ electrophoresis sample buffer, boiled, and vortexed. All samples were analyzed using 3-12% SDS-PAGE and visualized by autoradiography. For immunoprecipitation experiments, the radiolabeled cell lysate from a T75 flask was centrifuged at 5000 ϫ g, and the supernatant was split into eight equal aliquots and incubated with antibodies directed against ALKs 1, 2, and 5, the type II TGF-␤ receptor, or HA overnight at 4°C. Aliquots of radiolabeled cell lysates incubated with normal rabbit serum or no antibody were used as controls.
Immunoblot Analysis-Cells were washed twice with ice-cold D-PBS and lysed with 50 mM Tris, 150 mM NaCl buffer containing 1% Nonidet P-40, 0.25% deoxycholate, 1 mM EDTA, 20 mM sodium fluoride, 1 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, 2 g/ml leupeptin, 2 g/ml pepstatin, and 2 g/ml soybean trypsin inhibitor. Protein content was quantitated utilizing the BCA protein assay reagent (Pierce). Protein extracts were separated by 7.5% or 10% SDS-PAGE and transferred to nitrocellulose membranes at 100 volts for 2 h. Membranes were blocked with 5% nonfat dry milk in TBS-T (20 mM Tris-HCl, pH 7.6, 150 mM NaCl, 0.1% Tween 20 (v/v)) for 1 h at room temperature and incubated with primary antibodies diluted in TBS-T plus 2.5% nonfat dry milk over night at 4°C. The membranes were then washed four times for 10 min with TBS-T, incubated with horseradish peroxidase-conjugated secondary antibodies for 1 h at room temperature, and rewashed four times for 10 min with TBS-T. Immunoreactive bands were visualized by chemiluminescence (Pierce).
Immunofluorescence-Cells grown on glass coverslips (22 ϫ 22 mm) in 35-mm wells were washed twice with D-PBS, fixed with 4% paraformaldehyde in D-PBS for 15 min, permeabilized with 0.1% Triton X-100 for 10 min, and blocked with 3% nonfat dry milk in D-PBS for 60 min, all at room temperature. Cells were then incubated with primary mouse monoclonal antibodies diluted in 1% nonfat dry milk/D-PBS for 1 h at room temperature and washed three times with D-PBS, followed by incubation with a CY3-conjugated anti-mouse antibody in 1% nonfat dry milk/PBS for an additional hour at room temperature. In some experiments, cell nuclei were stained with 1 g/ml Hoechst for 10 min at room temperature. After three 10-min washes with D-PBS, coverslips were mounted onto 25 ϫ 75-mm microslides using AquaPolyMount (Polysciences Inc., Warrington, PA). Fluorescent images were captured using a Princeton Instruments cooled CCD digital camera from a Zeiss Axiophot upright microscope.
Transcription Reporter Assays-Cells were transiently transfected with 1 g per 35-mm dish of the Smad-dependent heterologous promoter reporter construct p(CAGA) 12 -Luciferase (38) provided by Dr. Jean-Michel Gauthier (Laboratoire Glaxo Wellcome, Les Ulis Cedex, France) along with 0.01 g per 35-mm dish of pCMV-Renilla using FuGENE 6 reagent (Roche Molecular Biochemicals) according to the manufacturer's protocol. The following day, cells were split into 24-well plates, and ϳ45 h post-transfection, cells were either left unstimulated or were stimulated with 40 pM TGF-␤1 for 16 -20 h. All cells were then washed with D-PBS and lysed. Firefly and Renilla reniformis luciferase activities were measured using Promega's dual luciferase reporter assay system according to the manufacturer's protocol. Luciferase activity was normalized utilizing the ratio of Firefly to R. reniformis luciferase activity and presented as -fold induction. All assays were done in triplicate wells, and each experiment was repeated at least twice.
Wound Closure and Transwell Motility Assays-For wound closure assays, confluent cell monolayers were wounded by manually scraping the cells with a pipette tip. Following wounding, wound size was verified with an ocular ruler to ensure that all wounds were the same width. The cell culture medium was then replaced with fresh medium, and wound closure was monitored by microscopy at various times.
Transwell motility assays were performed utilizing 5-m pore, 6.5-mm polycarbonate transwell filters (Corning Costar Corp., Cambridge, MA). For these assays, single cell suspensions were seeded in serum-free medium containing 0.1% bovine serum albumin onto the upper surface of the filters and allowed to migrate toward various concentrations of FCS. After a 16 -20-h incubation period, cells on the upper surface of the filter were wiped off with a cotton swab, and the cells that had migrated to the underside of the filter were fixed, stained with 0.5% crystal violet, and counted by brightfield microscopy at ϫ200 in five random fields.
Adenoviral Expression of ALKs and Smads-The adenoviral con-struct encoding FLAG-tagged Smad4 (39) was obtained from Dr. Harold Moses (Vanderbilt University, Nashville, TN). All other adenoviral constructs encoding FLAG-tagged Smads or HA-tagged constitutively active mutants of the TGF-␤ (ALK5 T204D ), activin (ALK2 Q207D ), and BMP (ALK3 Q233D ) type I receptors (40) were generously provided by Dr. Kohei Miyazono (Japanese Foundation for Cancer Research, Tokyo, Japan). Stocks of recombinant viruses for each of these constructs were generated in 293 cells and titered utilizing the Takara assay (Takara Biomedicals, Tokyo, Japan). Cells were then infected with these or a control ␤-galactosidase adenovirus at a multiplicity of infection (m.o.i.) that resulted in Ͼ90% cell infection (ϳ15 plaque-forming units/cell or less). The efficiency of infection was evaluated by in situ staining of cells for ␤-galactosidase activity 48 h following infection.

T␤RII-K277R Is Expressed in MDA-MB-231
Cells-To abrogate TGF-␤ signaling in MDA-MB-231 cells, an expression vector encoding GFP and kinase-inactive T␤RII was transfected into cells. Expression of the kinase-inactive T␤RII-K277R mutant was verified by affinity labeling cell surface receptors with 125 I-TGF-␤1 (Fig. 1A). Because T␤RII-K277R has an intact extracellular domain, it can still bind TGF-␤ and should therefore co-migrate with endogenous T␤RII. Cell surface labeling of parental cells resulted in the labeling of three proteins corresponding to the endogenous type I, II, and III TGF-␤ receptors. There was little or no change in the amount of receptor labeling observed in the control cells expressing GFP alone (GABE 15 Pool) compared with parental cells. However, in the pool expressing T␤RII-K277R (dnT␤RII 15 Pool), Pool; right panel) were affinity labeled with 100 pM 125 I-TGF-␤1, cross-linked with BS 3 , lysed, and incubated with normal rabbit serum (NRS), polyclonal rabbit antisera directed against various type I TGF-␤ superfamily receptors (ALK1, ALK2, and ALK5), the type II TGF-␤ receptor (T␤RII), or HA as indicated. Immunoprecipitates were resolved by SDS-PAGE using a 3-12% polyacrylamide gel and visualized by autoradiography. Affinity labeled but non-immunoprecipitated (NIP) cells were loaded as a reference (lane 1).
there was a significant increase in the amount of labeled type II receptor, suggesting that the exogenous receptor was expressed. Individual clones obtained from each of these pools expressing either GFP alone (G15-5, -6) or GFP and T␤RII-K277R (dn15-2, -3, -5, -11) displayed a similar pattern of labeling.
To confirm that this increase in the labeling of T␤RII was indeed because of expression of the HA-tagged T␤RII-K277R, extracts from affinity labeled cells were immunoprecipitated with an HA antibody. As shown in Fig. 1B, the HA antibody immunoprecipitated a labeled type II receptor in the pool and clones expressing T␤RII-K277R but not in the control pool or clones expressing GFP alone, confirming transgene expression. The type I TGF-␤ receptor appeared to co-immunoprecipitate with T␤RII-K277R in these experiments. This was confirmed in subsequent co-immunoprecipitation experiments (see Fig. 2).
T␤RII-K277R Is Functional in MDA-MB-231 Cells-Having ascertained that T␤RII-K277R was expressed, we then examined whether it was functional. Immunoprecipitation experiments revealed that when affinity labeled cells expressing T␤RII-K277R were precipitated with an HA antibody, a labeled protein the size of a type I receptor co-precipitated with T␤RII-K277R (see Fig. 1B and Fig. 2, right panel, lane 8). We confirmed that this was T␤RI by precipitating similarly labeled cells with various TGF-␤ superfamily type I receptor antibodies, including ALK1, -2, and -5. Only the ALK5 (T␤RI), but not the ALK1 or ALK2 antibodies, precipitated the cross-linked type I receptor (Fig. 2). Although the T␤RII antibody co-precipitated ALK5 efficiently, the ALK5 antibody co-precipitated T␤RII only weakly (see Fig. 2, lane 5, both panels). In the control GABE 15 Pool, the HA antibody failed to precipitate any proteins, as expected. Immunoprecipitations with a T␤RII antibody were carried out in both pools as a positive control, and both resulted in the co-precipitation of T␤RI (Fig. 2, lane 7, both panels). These data indicate that T␤RII-K277R associates with T␤RI.
To determine whether T␤RII-K277R prevented TGF-␤ signaling, we examined its effect on the ability of T␤RI to phosphorylate Smad2. Immunoblot analysis of TGF-␤1-treated cell lysates using a phospho-specific Smad2 antibody revealed that although TGF-␤1 could induce phosphorylation of Smad2 in both GABE clones (G15-5 and G15-6), its ability to do so in the T␤RII-K277R clones (dn15-2, -3, -5, -11) was impaired (Fig.  3A). This impairment was not because of a decrease in total Smad2 protein, as reprobing with an antibody directed against total Smad2/3 did not reveal any significant change in protein levels. We next examined the effect of T␤RII-K277R expression on TGF-␤-induced Smad translocation to the nucleus by immunofluorescence (Fig. 3B). In the GABE clone Smad2 staining was relatively diffuse, but upon TGF-␤1 treatment for 60 min, Smad2 staining became concentrated in the nucleus. In contrast, in the T␤RII-K277R clones, there was little or no change in Smad2 staining following TGF-␤1 treatment, suggesting impaired TGF-␤-mediated translocation of Smad2 to the nu-

FIG. 3. Expression of T␤RII-K277R impairs TGF-␤ signaling.
A, near confluent clones expressing GFP alone (G15-5 and G15-6) or GFP and T␤RII-K277R (dn15-2, dn15-3, and dn15-11) were incubated overnight under serumfree conditions, stimulated with 80 pM TGF-␤1 for the times indicated, washed, and lysed. Protein extracts (50 g/lane) were separated by 10% SDS-PAGE followed by immunoblot analysis for phospho-Smad2 (p-Smad2) and total Smad2/3. B, cells were grown on glass coverslips for 48 h and serum-starved for 16 h followed by treatment with 40 pM TGF-␤1 for 1 h. Cells were then prepared for indirect immunofluorescence staining of Smad2. Nuclear localization of Smad2 was confirmed by staining the same cells with Hoechst. C, cells were transiently transfected with p(CAGA) 12 -Luciferase along with pCMV-Renilla. The following day, cells were split to six wells of a 24-well plate, treated with 40 pM TGF-␤1 for 16 h, washed, and lysed. Firefly and Renilla luciferase activities were measured using Promega's dual luciferase reporter assay system. Fold induction of luciferase activity (y axis) is based on the ratio of Firefly to Renilla luciferase activities. Each data point represents the mean Ϯ S.D. of three wells. D, near confluent pools expressing GFP alone (G15) or GFP and T␤RII-K277R (dn15) were incubated overnight under serum-free conditions, stimulated with 80 pM TGF-␤1 for the times indicated, and prepared for immunoblot analysis as in A. The fibronectin blot (Fn) was probed with an antibody directed against actin as a loading control, whereas the phospho-Smad2 (p-Smad2) and phospho-p38 MAPK (p-p38 MAPK) blots were probed with antibodies directed against the unphosphorylated forms of the respective proteins to verify equal loading. cleus. We then examined the effect of T␤RII-K277R on TGF-␤1-induced transcription. A reporter construct containing twelve Smad binding elements repeated in tandem, p(CAGA) 12 -Luciferase, was transiently transfected into the GABE and T␤RII-K277R clones, along with pCMV-Renilla. Normalized luciferase activity indicated that TGF-␤1 could induce transcription of both reporter constructs in the GABE clones, but its ability to do so in the T␤RII-K277R clones was impaired (Fig. 3C).
To determine whether signaling pathways other than the Smad pathway were also affected by expression of T␤RII-K277R, we examined fibronectin expression, which has been reported to be induced by TGF-␤ in a JNK-dependent but Smad4-independent manner (6). We chose to perform these experiments in our pools as the results obtained in these cells are representative of those obtained in the clones (compare phospho-Smad2 blots in Fig. 3, A and D). Following TGF-␤ stimulation for 24 h, we observed an increase in fibronectin expression in the GABE pool, but this induction was decreased significantly in the pool expressing T␤RII-K277R, as was the basal level of fibronectin expression (Fig. 3D). We were unable to detect any induction of phosphorylation of JNK in response to TGF-␤ in our GABE pools (data not shown). However, we did observe an increase in phosphorylation of p38 MAPK following TGF-␤ stimulation for 60 min, and this induction of phosphorylation was slightly attenuated in the pool expressing T␤RII-K277R (Fig. 3D).
T␤RII-K277R Impairs the Motility of MDA-MB-231 Cells-Next we examined the effect of T␤RII-K277R expression on the motility of MDA-MB-231 cells in a wound closure assay. In the GABE clones, cells migrated into the wounded area and closed the wound within 24 h, whereas in the T␤RII-K277R clones the wound remained open at 24 h (Fig. 4A). This difference in motility did not appear to be because of an effect on proliferation, because when the experiment was performed in the presence of mitomycin C, a compound that inhibits cell division, the same results were obtained (data not shown). Thus, expression FIG. 4. Expression of T␤RII-K277R impairs motility. A, confluent cell monolayers of clones expressing GFP alone (G15-5 and G15-6) or GFP and T␤RII-K277R (dn15-2, dn15-3, and dn15-11) were wounded with a pipette tip. Following wounding, cell culture medium was replaced with fresh medium, and wound closure was monitored by microscopy at the times indicated. B, single cell suspensions of pools expressing GFP alone (G15 Pool) or GFP and T␤RII-K277R (dn15 Pool) in serum-free medium containing 0.1% bovine serum albumin were seeded onto 5 M polycarbonate transwell filters and allowed to migrate toward increasing concentrations of FCS, as indicated. After 20 h, cells on the underside of the filters were fixed, stained, and counted. The results are represented quantitatively in the bar graph below the representative filter micrographs. Each data point represents the mean Ϯ S.D. of two wells. of T␤RII-K277R in MDA-MB-231 cells appears to impair their motility, independent of changes in proliferation. As an alternative measure of cell motility, we also examined the effect of T␤RII-K277R on the ability of cells to migrate in a transwell assay system. We observed a 3-to 4-fold reduction in the ability of cells expressing T␤RII-K277R to migrate toward FCS, as compared with control cells expressing GFP alone (Fig. 4B).
The Impaired Motility of T␤RII-K277R Cells Is TGF-␤ Type I Receptor-specific-Because the impaired motility of T␤RII-K277R cells was observed in the absence of exogenous TGF-␤ stimulation, we wished to determine whether this impairment was TGF-␤-specific. To do so, we chose to restore TGF-␤ signaling in T␤RII-K277R cells by expressing a constitutively active mutant of T␤RI. Mutation of threonine 204 in ALK5 (T␤RI) to aspartic acid leads to constitutive activation of the type I receptor kinase, allowing it to induce signals in the absence of ligands or type II receptors (41). Likewise, mutation of corresponding threonine and glutamine residues in the activin (42) and BMP (43) type I receptors to aspartic acid also leads to constitutive activation of these kinases. To test for TGF-␤ specificity, cells expressing T␤RII-K277R were infected with adenoviruses encoding HA-tagged constitutively active mutants of the TGF-␤ (ALK5 TD ), activin (ALK2 QD ), and BMP (ALK3 QD ) type I receptors (40). Uninfected cells or cells infected with a ␤-galactosidase adenovirus at the same m.o.i. were used as controls. The efficiency of infection was Ͼ90% as evaluated by in situ staining of cells for ␤-galactosidase activity 48 h following infection (data not shown). At this time, expression of the mutant type I receptors was confirmed by immunoblot analysis utilizing an HA antibody (Fig. 5A), and their effect on motility was assessed in wound closure assays. Motility was only restored in cells expressing ALK5 TD (Fig. 5C). Although ALK2 QD and ALK3 QD were expressed and functional, as evidenced by their ability to induce Smad1 phosphorylation (Fig.  5B), they failed to restore motility in cells expressing T␤RII-K277R (Fig. 5C). These results suggest that the impaired motility of T␤RII-K277R cells is T␤RI (ALK5)-specific.
Restoration of Smad Signaling Does Not Rescue the Impaired Motility of T␤RII-K277R Cells-Although there is evidence that Smad signaling is critical for the anti-proliferative effects mediated by TGF-␤ (44,45), it is unclear whether TGF-␤mediated motility requires Smad signaling. We reasoned that if Smads are required for TGF-␤-mediated motility, blockade of Smad signaling with dominant-negative Smad mutants or with the inhibitory Smad, Smad7, should impair motility. However, expression of either dominant-negative Smad4 or Smad7 in the MDA-MB-231 parental cells resulted in cell death (data not shown). Therefore, it was not possible to address whether TGF-␤-mediated motility requires Smad signaling utilizing this approach. Instead, we chose to overexpress the TGF-␤ R-Smads, Smad2 or Smad3, along with Smad4, in cells expressing T␤RII-K277R to determine whether reconstitution of Smad signaling could restore motility. Cells were infected with adenoviruses encoding FLAG-tagged Smad2 and Smad4 or FLAG-tagged Smad3 and Smad4, and exogenous Smad expression was confirmed by immunoblot analysis utilizing an anti-FLAG antibody (Fig. 6A). The ability of Smad2/4 and Smad3/4 to activate Smad-dependent signaling was examined utilizing the Smaddependent transcription reporter construct, p(CAGA) 12 -Luciferase. Expression of Smad3/4 resulted in a marked increase in basal transcription (Fig. 6B). Stimulation with TGF-␤1 did not cause any further increase in transcription, suggesting that Smad signaling was activated maximally. Despite this, Smad3/4 failed to restore motility in the cells expressing T␤RII-K277R (Fig. 6C, bottom panel) and had no effect on the motility of control cells expressing GFP alone (Fig. 6C, top panel). Activation of basal transcription was not as marked with Smad2/4, as expected, because Smad2 itself cannot bind DNA (46). However, despite nearly 100-fold induction of transcription following infection at the maximally tolerated m.o.i., this combination also failed to restore motility in the cells expressing T␤RII-K277R (Fig. 6C, bottom panel) and had no effect on the motility of control cells expressing GFP alone (Fig. 6C, top  panel). These results indicate that reconstitution of Smad signaling alone is not sufficient to restore autocrine TGF-␤-mediated motility in cells expressing T␤RII-K277R, nor is it sufficient to enhance the motility of control MDA-MB-231 cells.
Re-expression of Smad4 in Smad4-defective Cancer Cells Does Not Enhance Motility-To determine whether Smads are required for cancer cell migration, we examined whether activation of TGF-␤ signaling could promote motility in the absence of Smad signaling utilizing Smad4 null MDA-MB-468 breast cancer cells (47). Smad4 and ALK5 TD were expressed, either alone or in combination, by adenoviral transduction, and their effects on the motility of MDA-MB-468 cells were examined in wound closure assays. Expression of HA-tagged ALK5 TD and FLAG-tagged Smad4 was confirmed by immunoblot analysis (Fig. 7A), and their ability to activate Smad-dependent signaling was examined in transcription reporter assays utilizing the Smad-dependent p(CAGA) 12 -Luciferase reporter construct (Fig. 7B). As expected, in the absence of Smad4 (uninfected, ␤-galactosidase, ALK5 TD alone), TGF-␤1 was unable to stimulate transcription in these cells. However, upon re-expression of Smad4 a marked increase in both TGF-␤1-mediated and ALK5 TD -mediated transcription was observed, indicating that both Smad4 and ALK5 TD were indeed functional in these cells. Despite this, neither Smad4 nor ALK5 TD had any effect on cell motility, whether they were expressed alone or in combination (Fig. 7C). The fact that ALK5 TD could not promote motility, even when Smad4 was co-expressed with it, suggests that MDA-MB-468 cells are not responsive to the pro-migratory effects of TGF-␤.
To determine whether TGF-␤ could induce migration in the absence of Smad signaling in other cells, the Smad4 defective SW480.7 colorectal cells, conditionally expressing Smad4 via an ecdysone-inducible system (34), were utilized. Cells were stimulated with increasing concentrations of TGF-␤1 in the absence or presence of 3 M ponasterone to induce Smad4 expression. Smad4 expression in ponasterone-treated cells was confirmed by immunoblot analysis (Fig. 8A), and its effect on TGF-␤-mediated motility was examined in wound closure assays. Again, as in the MDA-MB-468 cells, these cells failed to FIG. 8. Effect of exogenous TGF-␤1 stimulation on the motility of SW480.7 cells ؊/؉ Smad4. Cells in 6-well plates were either left untreated (left panel) or treated with 3 M ponasterone (right panel) for 40 h to induce Smad4 expression. Cells were then wounded, washed, and incubated with serum-free medium in the presence of 0, 4, 20, or 100 pM TGF-␤1 for 24 h. Wound closure was monitored by microscopy at the times indicated (B). At the conclusion of the wound closure experiment, cells were lysed, and Smad4 expression was examined by immunoblot analysis utilizing a monoclonal antibody directed against Smad4 (A). The blots were also probed with an actin antibody to verify equal loading. Ponasterone was maintained in the culture medium of selected wells throughout the experiment. respond to TGF-␤ both in the absence and presence of Smad4 (Fig. 8B). Taken together, these data indicate that reconstitution of Smad signaling alone is not sufficient to promote migration of cancer cells.
Different Signaling Pathways Require Different Thresholds of TGF-␤ Activation-To determine whether Smad signaling is actually required in addition to other pathways activated by TGF-␤ to promote migration, we examined what signaling pathways were activated under conditions where motility was restored following ALK5 TD expression in cells expressing T␤RII-K277R (see Fig. 5C and Fig. 9B). For these experiments, expression of ALK5 TD was confirmed by HA immunoblot (Fig.  9A), and the activation status of candidate signaling pathways was examined utilizing phospho-specific antibodies. Under these conditions, we observed an increase in the phosphorylation of Akt and ERK1/2, with little or no change in the phosphorylation of JNK or p38 MAPK (Fig. 9A). Interestingly, the levels of ALK5 TD expression that were sufficient to restore motility were not sufficient to induce phosphorylation of Smad2. Additional experiments indicated that ϳ4-fold greater ALK5 TD expression was required for induction of Smad2 phosphorylation (Fig. 9C). At that m.o.i., the viral load per se interfered with the ability of the cells to migrate. These data indi-cate that different signaling pathways require different thresholds of TGF-␤ activation, as do different biological effects mediated by TGF-␤ (27), and suggest that TGF-␤ may promote motility through mechanisms independent of Smad signaling, possibly involving activation of the PI3K-Akt and/or MAPK pathways. Consistent with this, the ability of TGF-␤ to promote migration was impaired in the presence of the PI3K inhibitor, LY294002 (Fig. 10A), as well as in the presence of JNK, MEK, and p38 MAPK pathway inhibitors (Fig. 10B).

DISCUSSION
In this study, abrogation of autocrine TGF-␤ signaling in MDA-MB-231 breast cancer cells resulted in an impairment in basal cell migration, which could not be restored by reconstituting Smad signaling, suggesting that Smad signaling alone is not sufficient for autocrine TGF-␤-mediated motility. Consistent with this, reconstitution of Smad signaling in the Smad4defective MDA-MB-468 and SW480.7 cells did not promote migration. In addition, restoration of migration following restoration of TGF-␤ signaling in cells expressing T␤RII-K277R was associated with an increase in phosphorylation of Akt and ERK1/2 but not Smad2. These results indicate that Smad signaling is dispensable for TGF-␤-mediated motility and that FIG. 9. Different signaling pathways require different thresholds of TGF-␤ activation. Clone dn15-2 was infected with an adenovirus encoding an HA-tagged constitutively active mutant of the TGF-␤ type I receptor (ALK5 TD ) at m.o.i. values of 0, 5, 10, or 15, as indicated. Approximately 48 h following infection, cells were wounded, and the effect of ALK5 TD expression on wound closure was monitored by microscopy at the times indicated (B). At the conclusion of the migration assay, cells were lysed, ALK5 TD expression was confirmed by HA immunoblot, and the activation status of candidate signaling pathways was examined utilizing phospho-specific antibodies, as indicated (A). Actin was examined as a loading control. In C, clone dn15-2 was re-infected with ALK5 TD at m.o.i. values of 0, 5, 15, 30, or 60, as indicated, and the activation status of candidate signaling pathways was re-examined utilizing phospho-specific antibodies as in A.
this response is instead mediated through alternative pathways activated by TGF-␤. In support of this, the ability of TGF-␤ to promote migration was blocked in the presence of pharmacological inhibitors of the PI3K, p38 MAPK, MEK, and JNK pathways.
Although Smads have been implicated as critical mediators of many TGF-␤ responses (48 -52), the role of Smads in cancer cell migration has, to the best of our knowledge, not been reported. Previous studies in non-transformed cells have generated conflicting data on the requirement of Smad signaling for cell migration (50,(53)(54)(55). In one study, expression of dom-inant-negative Smad3 in non-transformed murine mammary cells had no effect on TGF-␤-mediated motility even though it blocked the anti-mitogenic effect of TGF-␤ (53). This suggests that Smad3 is not required for this response or that residual Smad3 signaling, not blocked by expression of dominant-negative Smad3, is sufficient to mediate motility. This would be consistent with the idea that different biological responses require different thresholds of TGF-␤ signaling (27). Thus, complete abrogation of Smad3 signaling might be required to observe an impairment in TGF-␤-mediated motility whereas partial blockade of Smad function might be sufficient to block the anti-proliferative effects of TGF-␤.
Interestingly, expression of either dnSmad4 or antagonistic Smad7 in MDA-MB-231 cells resulted in cell death. Although overexpression of Smad7 has been reported to sensitize various cell types to cell death (56), expression of dnSmad4 has not been associated with such a response. There is, however, increasing evidence that TGF-␤ can promote the survival of both transformed (7,57,58) and non-transformed (59 -61) cells. Whether Smad signaling is required for these pro-survival effects of TGF-␤ is not known. In addition, re-expression of Smad4 in Smad4-defective SW480 has been reported to induce a more adhesive and flat phenotype (62). These results suggest that blockade of Smad signaling could potentially lead to loss of adhesion and result in anoikis. This could explain our inability to express dnSmad4 and Smad7 in MDA-MB-231 cells.
Because we were unable to assess the requirement for Smads in autocrine TGF-␤-mediated motility by abrogating Smad signaling, we chose to address this question by activating Smad signaling in cells expressing T␤RII-K277R. Having ascertained that the impaired motility of T␤RII-K277R cells was indeed T␤RI-specific, we overexpressed the TGF-␤ R-Smads, Smad2 or Smad3, each with Smad4, to determine whether autocrine TGF-␤-mediated motility was Smad-dependent. Despite their ability to activate Smad-dependent transcription, neither Smad combination restored the impaired motility of the T␤RII-K277R cells. We (data not shown) and others (62) have observed an increase in cell spreading following Smad overexpression. It is tempting to speculate that this increased cell spreading may be associated with increased adhesion, which interferes with cell migration. This could potentially explain why restoration of Smad signaling in T␤RII-K277R cells failed to restore migration. These data suggest that in breast cancer cells, autocrine TGF-␤ signaling mediates motility in a Smadindependent manner or that alternative pathways, in addition to the Smad signaling pathway, are required for these effects.
To address this question, we examined what signaling pathways were activated under conditions where motility was restored in T␤RII-K277R cells following expression of a constitutively active type I TGF-␤ receptor. In these experiments, we observed an increase in the phosphorylation of Akt and ERK1/2 but not Smad2. These data further imply that Smad signaling is not required for TGF-␤-mediated motility. Although expression of ALK5 TD may induce a low level of Smad phosphorylation, which cannot be detected by immunoblot analysis, it is unlikely that Smads are required for autocrine TGF-␤-mediated motility as reconstitution of Smad signaling in both MDA-MB-468 and SW480.7 cancer cell lines failed to promote motility, and expression of both low and high levels of either Smad2/4 or Smad3/4 failed to rescue the impaired motility of MDA-MB-231 cells expressing dnT␤RII. Thus, alternative signaling pathways activated by TGF-␤ are more likely to be important for migration. Indeed, blockade of the PI3K, p38 MAPK, MEK, and JNK pathways with pharmacological inhibitors impaired TGF-␤-stimulated migration. The fact that inhibitors of p38 MAPK and JNK interfered with TGF-␤-induced FIG. 10. TGF-␤-induced motility requires activation of multiple signaling pathways. Confluent MDA-MB-231 cell monolayers were wounded with a pipette tip. Following wounding, cells were washed, the cell culture medium was replaced with serum-free medium, and the cells were stimulated with 80 pM TGF-␤1 in the absence or presence of the PI3K inhibitor, LY294002 (A) or the p38 MAPK (SB202190), MEK (UO126), and JNK (JNKiII) inhibitors, as indicated (B). Wound closure was monitored by microscopy at the times indicated. migration even though ALK5 TD failed to alter their phosphorylation status suggests that these signaling pathways, though not activated further by TGF-␤ in our experimental system, are required for basal cell migration. In agreement with this, we have indeed observed an impairment in the basal migratory potential of these cells in the presence of these inhibitors (data not shown).
The observation that different levels of ALK5 TD expression resulted in differential activation of downstream targets (Fig.  9C) indicates that different signaling pathways require different thresholds of TGF-␤ activation. In agreement with this, others have reported that expression of dnT␤RII in NMuMG mammary cells impairs TGF-␤-mediated Smad-dependent inhibition of proliferation but not TGF-␤-mediated activation of p38 MAPK (9). In addition, there is evidence that different biological responses mediated by TGF-␤ also require different thresholds of TGF-␤ signaling. For example, expression of dnT␤RII in squamous carcinoma cells has been reported to block the growth inhibitory effects of TGF-␤ but not its ability to induce EMT (26). Likewise, expression of dnT␤RII in 4T1 murine mammary cancer cells impairs TGF-␤-mediated transcription but fails to block motility (27). Because TGF-␤ signaling was not completely abrogated in the squamous and mammary cancer cells (26,27), the molecular mechanisms by which autocrine TGF-␤ may selectively contribute to tumor progression could not be fully addressed in those studies. Because we have expressed T␤RII-K277R at levels high enough to block both Smad and non-Smad pathways in MDA-MB-231 cells, the model we have generated should prove useful in dissecting the signaling pathways required for the diverse effects elicited by TGF-␤ in cancer.
Our data indicate that autocrine TGF-␤-mediated motility of cancer cells is Smad-independent. This implies that non-transformed cells and transformed cells utilize different mechanisms to promote motility as others have reported that Smad3 null monocytes and keratinocytes exhibit significantly reduced migration to TGF-␤1 in transwell motility assays (54). Moreover, Smad3 appears to be required for TGF-␤-mediated monocyte chemotaxis in vivo, as mice lacking the Smad3 gene display a blunted monocyte chemotactic response following cutaneous wounding (54). Studies in Drosophila also suggest that Smads may be required for cell migration as mutations in Mad, the Drosophila receptor-activated Smad, impair migration of the epidermis during dorsal closure (55). Finally, recent studies in endothelial cells have indicated that TGF-␤ acting through ALK1 stimulates migration in a Smad-dependent manner, whereas TGF-␤ acting through ALK5 inhibits cell migration in a Smad-dependent manner (50). Taken together, these studies highlight the importance of Smads in TGF-␤regulated migration of non-transformed cells.
Despite compelling evidence for the role of Smads in nontransformed cell migration, a lack of requirement for Smad signaling in TGF-␤-mediated cancer cell migration is consistent with previous studies that have shown that TGF-␤ can increase cellular motility of prostate cancer cells without affecting proliferation, suggesting that the effects on motility and proliferation may occur via different biochemical pathways (28). Likewise, expression of Smad7 in pancreatic cancer cells has been shown to abrogate the anti-proliferative effects of TGF-␤ but enhance matrix-associated transcriptional responses, highlighting a dissociation between the matrix and anti-proliferative effects induced by TGF-␤ (63). If the biological effects of TGF-␤ that can contribute to tumor progression were Smad-independent, it might be possible to selectively disrupt those pathways, while ensuring that the tumor suppressive, Smad-dependent pathways are maintained. The sig-naling pathways currently implicated in mediating the various pro-and anti-tumorigenic effects of TGF-␤ indicate that this may in fact be possible. For example, recent studies aimed at identifying the mechanisms by which TGF-␤1 elicits EMT in mammary cells have indicated that the PI3K, RhoA, and p38 MAPK pathways are involved in this process (8,9,11,53). However, whether Smad signaling, which has been implicated in both the anti-proliferative (44,45) and pro-apoptotic (64,65) effects of TGF-␤, is also required for TGF-␤-mediated EMT is unclear. In one study, adenoviral expression of low levels of constitutively active ALK5 induced EMT only if Smad2/4 or Smad3/4 were co-expressed (66). In contrast, other investigators have reported that inhibition of Smad signaling either by overexpression of Smad7 or dominant-negative Smad3 did not affect the transdifferentiation, arguing against the involvement of Smads in EMT (53). Because epithelial transdifferentiation to a mesenchymal phenotype is often associated with acquisition of motile properties, the mechanisms through which TGF-␤ mediates EMT may be similar to those required for TGF-␤-mediated motility. Indeed, the PI3K, RhoA, and p38 MAPK signaling pathways, which are required for TGF-␤-mediated EMT, have also been implicated in TGF-␤-mediated motility (8,11,53). Likewise, we have observed that blockade of these and other pathways interfere with TGF-␤-induced motility (Fig. 10), suggesting that multiple pathways cooperate to elicit this effect. It will be of interest to determine whether Smad signaling is required for other effects mediated by TGF-␤, as a dissociation between the pathways required for the tumor suppressive versus the tumor promoting effects of TGF-␤ could lead to opportunities to selectively inhibit the non-desirable effects of TGF-␤ without compromising its tumor suppressive function.