Induction by transforming growth factor-β1 of epithelial to mesenchymal transition is a rare event in vitro

Introduction Transforming growth factor (TGF)-β1 is proposed to inhibit the growth of epithelial cells in early tumorigenesis, and to promote tumor cell motility and invasion in the later stages of carcinogenesis through the induction of an epithelial to mesenchymal transition (EMT). EMT is a multistep process that is characterized by changes in cell morphology and dissociation of cell–cell contacts. Although there is growing interest in TGF-β1-mediated EMT, the phenotype is limited to only a few murine cell lines and mouse models. Methods To identify alternative cell systems in which to study TGF-β1-induced EMT, 18 human and mouse established cell lines and cultures of two human primary epithelial cell types were screened for TGF-β1-induced EMT by analysis of cell morphology, and localization of zonula occludens-1, E-cadherin, and F-actin. Sensitivity to TGF-β1 was also determined by [3H]thymidine incorporation, flow cytometry, phosphorylation of Smad2, and total levels of Smad2 and Smad3 in these cell lines and in six additional cancer cell lines. Results TGF-β1 inhibited the growth of most nontransformed cells screened, but many of the cancer cell lines were insensitive to the growth inhibitory effects of TGF-β1. In contrast, TGF-β1 induced Smad2 phosphorylation in the majority of cell lines, including cell lines resistant to TGF-β1-mediated cell cycle arrest. Of the cell lines screened only two underwent TGF-β1-induced EMT. Conclusion The results presented herein show that, although many cancer cell lines have lost sensitivity to the growth inhibitory effect of TGF-β1, most show evidence of TGF-β1 signal transduction, but only a few cell lines undergo TGF-β1-mediated EMT.

In contrast to the growth inhibitory effects of TGF-β1 in the early stages of carcinogenesis, TGF-β1 can also act as a promoter of tumor cell invasion and metastasis in the later stages of tumorigenesis [5,6]. Increased production of TGF-β1 is observed in epidermal [35], gastric [36], renal [37], breast [38][39][40][41], and prostate carcinomas [42] when compared with normal tissues. In mice with polyomavirus middle T antigen expression targeted to the mammary gland, blockade of TGF-β1 by administration of Fc:TβRII results in increased apoptosis in primary tumors and reduced tumor cell motility, intravasation, and metastasis [43]. Chronic exposure of mouse epidermal cells to TGF-β1 results in loss of TGF-β1-mediated growth inhibition and marked changes in cell morphology, downregulation of E-cadherin and cytokeratins, upregulation of vimentin, and formation of spindle cell carcinomas in mice [44,45]. Further studies show that carcinomas with excess TGF-β1 production are more motile and invasive, and exhibit increased tumor cell metastasis in athymic mice [36,40,[45][46][47][48][49][50][51].
One mechanism by which TGF-β1 can promote tumor cell motility and invasion is through the induction of EMT [52]. EMT is a complex process that involves changes in cell morphology and dissociation of cell-cell contacts [53,54]. Cells undergoing EMT change from a cobblestone-like appearance to an elongated, mesenchymal phenotype. Accompanying this morphologic change is the delocalization of adherens and tight junctional proteins from the cell-cell junctions, and remodeling of the actin cytoskeleton [53,54]. Characteristics associated with EMT, such as the dissociation of cell-cell and cell-extracellular matrix contacts, acquisition of an elongated cell morpho-logy, and rearrangement of the cytoskeleton, can facilitate cell migration and invasion [55][56][57].
Although TGF-β1 is thought to play a key role in EMT in vivo, the frequency of TGF-β1-induced EMT in vitro is not known. To identify alternative cell systems in which to study TGF-β1-mediated EMT, we screened primary cultures of two human epithelial cell types and 18 established mouse and human cell lines for TGF-β1 responsiveness. We also included six additional cancer cell lines as a comparison for TGF-β1 responsiveness. We found that many of the cell strains displayed morphological changes and exhibited actin stress fiber responses to TGF-β1. However, only in the NMuMG and MCT cells were those changes accompanied by a loss of E-cadherin and zonula occludens (ZO)-1 at cell-cell junctions after 48 hours of TGF-β1 treatment. Whereas all of the nontransformed cells were growth inhibited by TGF-β1, many of the cancer cell lines were insensitive to the growth inhibitory effects of TGF-β1. The TGF-β1-mediated growth inhibition was accompanied by an increase in phosphorylated Smad2 protein levels, but this was not unique to growth inhibited cells because changes in Smad2 phosphorylation occurred in a majority of cells after TGF-β1 treatment. In addition, prolonged TGF-β1 treatment induced a decrease in total Smad2 and/or total Smad3 in some cell lines. Our findings show that, although many cancer cells lost sensitivity to the growth inhibitory effect of TGF-β1, only two murine cell lines underwent TGF-β1-mediated EMT and these cells retain growth inhibitory response to TGF-β1. Cell lines and culture conditions   A549, BT549, DU145, H1299, HBL100, MCF10A,  MCF7, MDA-MB-231, MDA-MB-361, MDA-MB-435S,  MDA-MB-436, MDA-MB- A549, Colo357, DU145, EpH4, H1299, HaCaT, HBL100, MCT, Panc-1, and 4T1 cells were maintained in Dulbecco's modified Eagle's medium (DMEM)/high-glucose medium (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (FBS). BT549 cells were maintained in RPMI-1640 medium (HyClone) supplemented with 10% FBS and 10 µg/ml insulin. NMuMG cells were maintained in DMEM/high-glucose medium supplemented with 10% FBS and 10 µg/ml insulin. HMEC-1016 and HMEC-1012 were grown in DMEM:F12 medium (1:1) (GibCoBRL, Grand Island, NY, USA) supplemented with 1% FBS, 10 µg/ml ascorbic acid, 2 nmol/l β-estradiol, 35 µg/ml bovine pituitary extract, 1 ng/ml cholera toxin, 12.5 ng/ml epidermal growth factor (EGF), 0.1 mmol/l ethanolamine, 0.1 mmol/l phospho-ethanolamine, 1 µg/ml hydrocortisone, 1 µg/ml insulin, 0.2 mmol/l L-glutamine, 10 nmol/l T3, 10 µg/ml transferrin, and 15 nmol/l sodium selenite.
All cell lines were maintained at 37°C in 5% CO 2 , except for the NMuMG, MK, and KC cells, which were grown in 7% CO 2 .

[ 3 H]Thymidine incorporation assays
Subconfluent cells were treated with TGF-β1 (5 ng/ml) for 46 hours in 12-well plates and pulsed for 2 hours with 4 µCi/well [ 3 H]thymidine (Perkin Elmer Life Sciences, Boston, MA, USA). Cells were fixed with 1 ml 10% trichloroacetic acid for 30 min at 25°C, followed by two washes with 10% trichloroacetic acid. DNA was solubilized by incubation in 600 µl 0.2 N NaOH for 30 min, and radioactivity was counted using 200 µl solubilized DNA in 4 ml scintillation fluid.

Flow cytometry analyses
Subconfluent cells were treated with TGF-β1 (5 ng/ml) for 48 hours and approximately 10 6 cells were incubated in propidium iodide solution (50 µg/ml propidium iodide (Sigma), 5 µg/ml RNase A, 0.1% Triton X-100, and 1 µg/ml sodium citrate) for 5-10 min. Stained cells were analyzed using a FACS Caliber (Becton-Dickinson, San Jose, CA, USA) and the data stored as listmode files. DNA cell cycle histograms were analyzed and modeled using ModFit and WinList software (Verity Software House, Topsham, ME, USA). Fifteen thousand events were analyzed for each sample.

Immunofluorescence microscopy
Subconfluent cells, grown on 22 mm 2 glass cover slips (VWR Scientific, Atlanta, GA, USA), were treated for 48 hours or 6 days with 5 ng/ml TGF-β1. After treatment, cells were washed once with phosphate buffered saline (PBS) and fixed with 100% cold methanol (for rat ZO-1 and E-cadherin antibody use), 4% paraformaldehyde/PBS (for F-actin staining), or 1% paraformaldehyde/PBS (for rabbit ZO-1 antibody use) for 20 min at 25°C. Cells were washed four times with PBS, permeabilized by incubation with 0.2% Triton X-100 (in PBS) for 10 min at 25°C, and then washed two more times with PBS. For ZO-1 and E-cadherin staining, nonspecific binding sites were blocked by cellular incubation for 2 hours with 5% serum (goat serum for rabbit ZO-1 and E-cadherin antibody use and rabbit serum for rat ZO-1 antibody use) in PBS, incubated in primary antibodies diluted in 5% serum/PBS (rat ZO-1 1:1000, rabbit ZO-1 1:100, and E-cadherin 1:2500) for 1 hour at 25°C, followed by four washes with PBS. Cells were incubated in appropriate biotinylated secondary antibodies diluted in 5% serum/PBS (1:250) for 1 hour at 25°C, washed four times with PBS, and incubated in streptavidin-conjugated Cy3 diluted in 5% serum/PBS (1:1000) for 1 hour at 25°C. Cells were washed four times with PBS and nuclei were counterstained by incubation with 0.1 µg/ml Hoechst (33258) for 15 min at 25°C. For F-actin staining, cells were incubated in Texas Red-X phalloidin (Molecular Probes Inc., Eugene, OR, USA) diluted in PBS (1:200) for 30 min, washed with PBS three times, and were counterstained with Hoechst as described above.
After Hoechst staining, cells were washed three times with PBS and cover slips were mounted onto 25 × 75 mm microslides (Fisher Scientific, Pittsburgh, PA, USA) using AquaPolyMount (Polysciences, Warrington, PA, USA). Phase contrast images were captured using the Zeiss Epifluorescence inverted microscope and Zeiss AxioCam digital camera (Carl Zeiss, Jena, Germany). Fluorescent images were captured using the Zeiss Axiophot upright microscope (Carl Zeiss) and the Princeton Instruments cooled CCD digital camera (Princeton Scientific Instruments, Inc., Monmouth Junction, NJ, USA).

Transforming growth factor-β β1 induced epithelial to mesenchymal transition in NMuMG and MCT cells
To determine the frequency of TGF-β1-mediated EMT in vitro, we performed a screen of 18 established cell lines from human and mouse epidermis, mammary gland, lung, pancreas, kidney, and prostate (Table 1). In addition to the established cell lines, primary cultures of human epidermal keratinocytes (HEKs) and human mammary gland epithelial cells (HMEC-1012 and HMEC-1016) were also screened ( Table 1). In the present study, EMT was defined as acquisition of a spindle-shaped morphology, loss of ZO-1 and E-cadherin from cell-cell junctions, and formation of actin stress fibers after TGF-β1 treatment. Using phase contrast and immunofluorescence microscopy, we determined that only two cell lines underwent EMT. After 48 hours of TGF-β1 treatment, the NMuMG and MCT cells became elongated (Fig. 1a,e,i,m), lost E-cadherin (Fig. 1b,f,j,n) and ZO-1 ( Fig. 1c,g,k,o) staining at cell-cell junctions, and developed actin stress fibers (Fig. 1d,h,l,p).
The primary cultures of breast epithelial cells (HMEC-1012 and HMEC-1016) acquired a spindle-shaped morphology, whereas the primary cultures of HEKs did not change cell morphology after 48 hours of TGF-β1 treatment (Fig. 2a,e,i,m; and data not shown). In addition, the HMECs and HEKs developed actin stress fibers after treatment (Fig. 2d,h,l,p; and data not shown). However, the HMECs and HEKs did not lose junctional E-cadherin staining (Fig. 2b,f,j,n; and data not shown) after TGF-β1 treatment for 48 hours. These cells also had little or no ZO-1 localized to their cell-cell junctions ( Fig. 2c,g,k,o; and data not shown).
The human breast epithelial cell line MCF10A did not become elongated (Fig. 3a,e) or lose junctional E-cadherin or ZO-1 staining (Fig. 3b,f,c,g) in response to 48 hours of TGF-β1 treatment. After treatment, these cells exhibited some actin stress fiber formation (Fig. 3d,h). However, the MCF10A cells acquired a spindle-shaped morphology ( Fig. 3i,m) and actin stress fibers (Fig. 3l,p), and lost much of the junctional E-cadherin and ZO-1 ( Fig. 3j,n,k,o) after TGF-β1 treatment for 6 days. The HaCaT, UNC10, A549, Colo357, DU145, MK, KC, and 4T1 cells also exhibited an actin stress fiber response to TGF-β1, without delocalization of E-cadherin or ZO-1 from the cell-cell junctions, similar to the MCF10A cells after 48 hours of TGF-β1 treatment (Fig. 4b,f,r,v,c,g,s,w,d,h,t,x; and data not shown).
Other cell lines screened did not exhibit any changes in cell morphology or relocalization of E-cadherin, ZO-1, or actin after treatment with TGF-β1. These cell lines included  Table 2.

Transforming growth factor-β β1-induced inhibition of S-phase
Previous studies analyzing TGF-β1 induction of EMT have emphasized the stages in which TGF-β1 acts as a promoter of carcinogenesis. It is proposed that cells first lose sensitivity to the growth inhibitory effects of TGF-β1, before TGF-β1 can promote EMT and cancer progression [59]. Therefore, we examined whether there was a correlation between TGF-β1-mediated EMT and loss of sensitivity to the growth inhibitory effects of TGF-β1 by examining [ 3 H]thymidine incorporation and flow cytometry.  Table 1 Tissue origin, description, and morphology of cell strains used    MDA-MB-231, A549, DU145, and EpH4 ( Fig. 5 and Table 3).
Transforming growth factor-β β1-mediated Smad2 phosphorylation did not correlate with S-phase reduction To determine the relationship of Smad2 phosphorylation with reduction in S-phase and EMT, after TGF-β1 treatment, we analyzed phosphorylated Smad2 protein levels.
Western analyses of protein extracts showed that, after 2 hours of treatment with TGF-β1, phosphorylated Smad2 protein levels increased in all of the cell strains screened except for the MCF7, MDA-MB-361, and MDA-MB-435S cells when normalized to actin ( Fig. 7 and Table 3). After treatment with TGF-β1 for 48 hours, levels of phosphorylated Smad2 remained elevated in the cells with increased levels of phosphorylated Smad2 after 2 hours of treatment, except HMEC (1016), HBL100, MCF10A, MDA-MB-436, H1299, MK, and MCT when normalized to actin (Figs 7 and 8, and Table 3). Immunoblot analysis of the MCF7, MDA-MB-361, and MDA-MB-435S cells with the phosphorylated Smad2 antibody (lot #21034) resulted in a band (indicated by an asterisk) that migrated higher than the phosphorylated Smad2 protein (Fig. 7).
Cell strains that exhibited TGF-β1-induced phosphorylation of Smad2 did not necessarily exhibit a reduction in   Table 3).

Prolonged transforming growth factor-β β1 treatment led to decreased total Smad2 and Smad3 protein levels in many cell strains
Total Smad2 and Smad3 protein levels decreased with TGF-β1 treatment in 51% and 48% of the cell lines, respectively (Fig. 8).    (Fig. 8).
Decreased total Smad2 and total Smad3 protein levels did not correlate with an increased reduction in S-phase after TGF-β1 treatment for 48 hours. For instance, the HMEC (1012), MK, and KC cells had a 87%, 81%, and Available online http://breast-cancer-research.com/content/6/3/R215 R223 Table 3 Summary of transforming growth factor-β β1-mediated Smad signaling and effect on S-phase in vitro  Table 3).
Also, decreased total Smad2 and Smad3 protein levels did not correlate with an increase in the levels of phosphorylated Smad2 protein after TGF-β1 treatment (Figs 7 and 8, and Table 3). An increase in phosphorylated Smad2 protein levels occurred in all cell strains screened after 2 hours of TGF-β1 treatment except for the MCF7, MDA-MB-361, and MDA-MB-435S cells (Fig. 7), whereas only a subset of those had decreased total Smad2 and/or Smad3 protein levels after TGF-β1 treatment for 48 hours (Fig. 8 and Table 3). Additionally, the MDA-MB-361 and the MDA-MB-435S cells did not have increased phosphorylated Smad2 levels or decreased total Smad3 Breast Cancer Research Vol 6 No 3 Brown et al.

Figure 6
Transforming growth factor (TGF)-β1 induced a robust decrease in S-phase in non-cancer cell strains. Human primary cells and human and mouse established cell lines were treated with (TGF-β1) or without (control) 5 ng/ml TGF-β1 for 48 hours. The cells were then harvested and incubated with propidium iodide, and DNA content was determined using a FACS Caliber (Becton-Dickinson). Data were modeled and analyzed using ModFit and WinList software (Verity Software House). Fifteen thousand events were analyzed for each sample. Representative histograms are presented illustrating cells that had a robust decrease in S-phase (NMuMG), modest decrease in S-phase (EpH4), and no decrease in S-phase (MDA-MB-436) as a result of TGF-β1 treatment. G 0 /G1 and G2 phases are indicated by black shading and S phase is indicated by gray shading on the histograms. The percentage of cells in G 0 /G1 and in S phases of the cell cycle are indicated. Three independent experiments were performed with three replicates per experiment.
levels, but they did exhibit decreased total Smad2 levels after 48 hours of TGF-β1 treatment (Figs 7 and 8, and Table 3).
The decrease in total Smad2 and Smad3 protein levels, observed after TGF-β1 treatment, did not associate with the ability of the cell line to undergo TGF-β1-mediated EMT. NMuMG and MCT cells underwent EMT (Fig. 1) and they both exhibited decreased total Smad2 and no decrease in total Smad3 protein levels after TGF-β1 treatment for 48 hours (Fig. 8). The other cell lines that had decreased total Smad2 and no decrease in total Smad3 levels after TGF-β1 treatment did not undergo TGF-β1-induced EMT (Figs 2-4 and 8, and Table 3; and data not shown).

Discussion
TGF-β1 is an inhibitor of epithelial cell growth in the early stages of carcinogenesis. However, it also promotes tumor cell invasion and metastasis through the induction of EMT [52]. Many studies on the mechanism of TGF-β1induced EMT are limited to a few murine cell lines and mouse models. Therefore, to identify alternative cell systems in which to study TGF-β1-induced EMT, we performed a TGF-β1 sensitivity screen using a panel of human primary cells and established human and mouse cell lines.

Figure 7
Transforming growth factor (TGF)-β1 induced Smad2 phosphorylation in many cancer and noncancer cell strains. Twenty-four human and mouse established cell lines and three primary human cell cultures were treated with TGF-β1 (5 ng/ml) for 2 hours. Extracts from control (-) and treated (+) cells were prepared and subjected to immunoblot analyses using antibodies for phosphorylated Smad2 (Ser 465/467), total Smad2, total Smad3, or actin. A signal of higher molecular weight, that is not phosphorylated Smad2, is indicated by an asterisk.
Changes in cell morphology, from a cobblestone-like appearance to more elongated shape, and actin stress fiber formation, were observed in the NMuMG and MCT cells, and in primary cultures of HMECs (Fig. 1a,d,e,h,i,l,m,p; Fig. 2i,l,m,p; and data not shown). However, only the NMuMG and MCT cells lost E-cadherin and ZO-1 staining at the cell-cell junctions after 48 hours of TGF-β1 treatment (Fig. 1b,f,j,n,c,g,k,o). E-cadherin is a protein found at adherens junctions that allows for cell-cell adherence, and ZO-1 is a tight junctional protein that forms a selective barrier between cells. Loss of these proteins at cell-cell junctions is used as a marker of EMT [54,[60][61][62][63]. Cortical actin is a measure of cell integrity and loss thereof, or formation of actin stress fibers, is also used to define EMT [63,64]. Therefore, we concluded that only the NMuMG and MCT cells underwent TGF-β1-mediated EMT, after 48 hours of treatment, in this screen. Results from previous studies also show that NMuMG and MCT cells undergo TGF-β1-mediated EMT [60,61,[63][64][65]. Other studies reported that TGF-β1-mediated EMT occurs in HaCaT, Colo357, and Panc-1 cells 48 hours after TGF-β1 treatment [66,67]; however, they did not undergo TGF-β1induced EMT according to the conditions used in the present study (Fig. 4). Other cell lines screened in the Breast Cancer Research Vol 6 No 3 Brown et al.

Figure 8
Prolonged transforming growth factor (TGF)-β 1 treatment induced a decrease in total Smad2 and/or Smad3 in many cell strains. Twenty-four human and mouse established cell lines and three primary human cell cultures were treated with TGF-β1 (5 ng/ml) for 48 hours. Extracts from control (-) and treated (+) cells were prepared and subjected to immunoblot analyses using antibodies for phosphorylated Smad2 (Ser 465/467), total Smad2, total Smad3, or actin. study formed actin stress fibers with TGF-β1 treatment, but they did not lose ZO-1 from cell-cell junctions (Fig. 3c,d,g,h; and data not shown). We did not examine E-cadherin localization in all of the cell strains because not all of the cell strains exhibited morphologic changes, with TGF-β1 treatment, that were indicative of EMT.
Unlike the mouse NMuMG and MCT cells, none of the 14 human cell lines with an epithelial morphology screened underwent TGF-β1-mediated EMT within 48 hours of TGF-β1 treatment. The MCF10A human breast epithelial cell line took 6 days to undergo a morphologic change, lose junctional E-cadherin and ZO-1, and form actin stress fibers with TGF-β1 treatment (Fig. 3i-p). However, this extended treatment with TGF-β1 to induce EMT is not consistent with previous reports that used times up to and including 48 hours [60,61,66,67]. It may be that the extended treatment with TGF-β1 is necessary to activate secondary and tertiary signaling pathways that are not activated within 48 hours. The Panc-1, Colo357, and HaCaT cells were treated with TGF-β1 for longer than 48 hours because they were shown to undergo TGF-β1induced EMT in previous reports [66,67]. However, these cell lines did not undergo EMT when treated with TGF-β1 for 72 hours (data not shown). These data could suggest that human cells are more resistant to TGF-β1-mediated EMT. Alternatively, the TGF-β1 treatment times were not long enough to induce EMT, the small sample size used may not be representative, or the cells may not have the genetic alterations necessary for TGF-β1 to induce EMT.
It is somewhat unexpected that TGF-β1-induced EMT occurred in nontransformed cell lines, such as the NMuMG and MCT cells, because in vivo mouse studies show that TGF-β1 promotes tumor cell invasion and metastasis in the later stages of carcinogenesis [5,6]. In one of these studies, targeted expression of TGF-β1 to mouse suprabasal keratinocytes results in resistance to the formation of benign skin tumors, after long-term chemical carcinogenesis treatment. However, benign papillomas that develop become malignant at an accelerated rate and metastasis occur more rapidly than spontaneous tumors in control mice. Additionally, there are a high incidence of spindle cell carcinoma development in these mice, pointing to TGF-β1-induction of EMT in vivo [38]. NMuMG and MCT cells may have the genetic alterations that allow for TGF-β1-mediated EMT to occur.
Many studies on TGF-β1-mediated EMT have emphasized that TGF-β1 promotes carcinogenesis in stages. It is proposed that cells first lose sensitivity to the growth inhibitory effects of TGF-β1, and subsequently TGF-β1 can promote tumor progression and EMT [59]. MMH-D3 murine hepatocytes and EpH4 murine mammary gland epithelial cells undergo TGF-β1-mediated EMT only when rendered insensitive to TGF-β1-induced cell cycle arrest or apoptosis, by infection with active H-Ras [59,70,73,81]. However, the present study does not support this model because it shows that resistance to the growth inhibitory effects of TGF-β1 is not a prerequisite for TGF-β1-mediated EMT. The NMuMG and MCT cells exhibited a decrease in S-phase and underwent EMT after 48 hours of TGF-β1 treatment (Figs 1, 5, and 6). Consistent with the present study were experiments performed by Nicolas and coworkers [72] that showed that increased Smad3 in MDCK cells restored growth inhibitory responses to TGF-β1 but did not revert cells from a mesenchymal to an epithelial phenotype. Additionally, Chang and coworkers [49] observed that increased TGF-β1 expression in sarcoma cells increased cell tumorigenicity while inhibiting cell proliferation. All but three of the cell strains screened had an increase in phosphorylated Smad2 protein levels after 2 hours of TGF-β1 treatment, but not all of the cell lines that had increased phosphorylation of Smad2 had reduced S-phase after 48 hours of TGF-β1 treatment. The SCC028, MDA-MB-436, MDA-MB-468, and H1299 cells had an increase in phosphorylated Smad2 levels after 2 hours of TGF-β1 treatment, but they exhibited decreases in S-phase of only 5%, 1%, 4%, and 4%, respectively (Figs 5-7 and Table 3). The lack of S-phase decrease upon 48 hours of TGF-β1 treatment may point to inactivation of or mutations in downstream proteins that could abolish TGF-β1-mediated growth inhibition in these cells. The MCF7 cells did not exhibit an increase in phosphorylated Smad2 protein levels after 2 hours of TGF-β1 treatment, but there was a 41% decrease in Sphase on 48 hours of TGF-β1 treatment (Figs 5-7 and Table 3). This decrease in S-phase is not as great a decrease as was exhibited by some of the other cell strains when treated with TGF-β1 for 48 hours. The decrease could be explained by very low levels of phosphorylated Smad2 increase on TGF-β1 treatment that was not detected by immunoblot analysis. Alternatively, it is possible that the TGF-β1 signal was propagated by phosphorylation of Smad3.
Smad2 is not solely responsible for the propagation of TGF-β1 signals. Similar to Smad2, on TGF-β1 binding to the TβRII, Smad3 is phosphorylated by the TβRI, forms a complex with Smad4, translocates to nucleus, and regulates the activation of TGF-β1 target genes. However, mutations or deletions in Smad3 are rare in human cancers. Reports by Graham and coworkers [82] and Xu and coworkers [83] suggest that loss of Smad3 may largely be responsible for the nonresponsiveness of some cells to TGF-β1. The proliferation, migration, and invasion of normal extravillous trophoblast cells are under the control of TGF-β1. However, premalignant and malignant trophoblast cells, that have lost the Smad3 protein but retain functional Smad2, are resistant to the antiproliferative and anti-invasive effect of TGF-β1. It is therefore possible that the attenuation of inhibition of S-phase or the induction of EMT by TGF-β1 in the present study may be a result of mutation or loss of Smad3, even though the cells have increased phosphorylation of Smad2 on TGF-β1 treatment.
Interestingly, decreased total Smad2 and Smad3 protein levels were observed after TGF-β1 treatment regardless of changes in the levels of phosphorylated Smad2 protein or of whether cell proliferation was inhibited by TGF-β1 (Figs 5-8). Similar to our findings, decreased total Smad2 protein levels were reported in COS-1 monkey kidney cells after TGF-β1 treatment [84]. In that report, the proteasome inhibitors MG-132 or lactacystin blocked Smad2 from TGF-β1-induced degradation. In addition, Smad3 decreases have been reported during TGF-β1induced EMT in MDCK cells [72,85] and after TGF-β1 treatment of human lung epithelial cells [86]. In these studies, MDCK cells become refractory to the growth inhibitory effects of TGF-β1 [72,85]. TGF-β1 treatment of primary human fibroblasts and HaCaT cells also leads to decreased total Smad3 protein levels [87,88].
A negative feedback loop could explain the decreases in total Smad2 and Smad3 protein levels after 48 hours of TGF-β1 treatment. The E3 ubiquitin ligase Smurf has been shown to interact with Smads and promote their ubiquitination [89,90]. In this model, phosphorylated, nuclear Smad2 is ubiquitinated by Smurf2 and degraded by proteasomes [89,90]. Smad7 also interacts with Smurf2 and induces TGF-β receptor degradation [91,92]. Phosphorylated Smad3 is also ubiquitinated by the ROC1-SCF Fbw1a E3 ligase complex, and subsequently degraded in the proteasome [88]. In order to prevent continuous Smad signaling in the absence of TGF-β1 stimulation, Smad2 and Smad3 are negatively regulated by a number of proteins. Smad6 and Smad7 inhibit Smad2 and Smad3 activation by competing with Smad2 and Smad3 for binding to the TGF-β receptors [93]. Smad6 and Smad7 are induced by activation of TGF-β1 signaling and form a negative feedback loop [94][95][96]. However, the mechanism of decrease in total Smad2 and/or Smad3 in different cell strains on prolonged TGF-β1 treatment remains unclear.

Conclusion
TGF-β1 induction of EMT is a rare occurrence in vitro and is not limited to transformed cell lines. In fact, TGF-β1mediated EMT was only observed in nontumorigenic cells in the present study. Furthermore, there does not appear to be a correlation between loss of TGF-β-mediated growth inhibition and EMT. The inability of TGF-β1 to induce EMT was not due to a total loss of TGF-β1 signaling, as was evidenced by induction of Smad2 phosphorylation. In conclusion, our data indicate that the in vitro models used to study TGF-β1 induction of EMT may not be as relevant to in vivo cancer progression as was previously assumed in the field.