Fascin Protein Is Critical for Transforming Growth Factor β Protein-induced Invasion and Filopodia Formation in Spindle-shaped Tumor Cells*

Background: Fascin is a pro-metastasis actin-bundling protein overexpressed in metastatic tumors. Results: TGFβ induced fascin expression in spindle tumor cells through Smads. Conclusion: Fascin is a TGFβ target gene essential for the pro-invasion activity of TGFβ. Significance: Our data shed new light on how TGFβ dysregulates actin cytoskeleton to promote tumor metastasis. Fascin, an actin-bundling protein overexpressed in all carcinomas, has been associated with poor prognosis, shorter survival, and more metastatic diseases. It is believed that fascin facilitates tumor metastasis by promoting the formation of invasive membrane protrusions. However, the mechanisms by which fascin is overexpressed in tumors are not clear. TGFβ is a cytokine secreted by tumor and mesenchymal cells and promotes metastasis in many late stage tumors. The pro-metastasis mechanisms of TGFβ remain to be fully elucidated. Here we demonstrated that TGFβ induced fascin expression in spindle-shaped tumor cells through the canonical Smad-dependent pathway. Fascin was critical for TGFβ-promoted filopodia formation, migration, and invasion in spindle tumor cells. More importantly, fascin expression significantly correlates with TGFβ1 and TGFβ receptor I levels in a cohort of primary breast tumor samples. Our results indicate that elevated TGFβ level in the tumor microenvironment may be responsible for fascin overexpression in some of the metastatic tumors. Our data also suggest that fascin could play a central role in TGFβ-promoted tumor metastasis.

Fascin, an actin-bundling protein overexpressed in all carcinomas, has been associated with poor prognosis, shorter survival, and more metastatic diseases. It is believed that fascin facilitates tumor metastasis by promoting the formation of invasive membrane protrusions. However, the mechanisms by which fascin is overexpressed in tumors are not clear. TGF␤ is a cytokine secreted by tumor and mesenchymal cells and promotes metastasis in many late stage tumors. The pro-metastasis mechanisms of TGF␤ remain to be fully elucidated. Here we demonstrated that TGF␤ induced fascin expression in spindleshaped tumor cells through the canonical Smad-dependent pathway. Fascin was critical for TGF␤-promoted filopodia formation, migration, and invasion in spindle tumor cells. More importantly, fascin expression significantly correlates with TGF␤1 and TGF␤ receptor I levels in a cohort of primary breast tumor samples. Our results indicate that elevated TGF␤ level in the tumor microenvironment may be responsible for fascin overexpression in some of the metastatic tumors. Our data also suggest that fascin could play a central role in TGF␤-promoted tumor metastasis.
One essential characteristic of metastatic cancer cells is enhanced motility, which facilitates the infiltration of metastatic cells into lymphatic and blood vessels and extravasation out of the circulation (1). The forces that drive tumor cell migration and invasion are provided by the actin cytoskeleton underlying the critical membrane protrusions in migrating tumor cells (2). To efficiently drive the formation of membrane protrusions, it is crucial to cross-link actin filaments into bundles as individual actin filaments are flexible. Cross-linking by bundling protein provides the essential rigidity to counter the compressive forces from the plasma membrane (3).
Fascin is an actin-bundling protein critical for tumor metastasis (4 -7). Expression levels of fascin are very low or not detected in normal epithelia, but are highly elevated in malignant tumors (4). Overexpression of fascin protein is associated with poor prognosis in patients (7)(8)(9)(10)(11)(12). Knockdown of fascin expression inhibited tumor cell migration and invasion in vitro and decreased tumor metastasis in mouse models. Moreover, ectopic expression of fascin promoted tumor cell invasion and metastasis (5,6,13). The causal role of fascin overexpression in tumor metastasis is well established; however, the molecular mechanisms underlying elevated fascin level in metastatic tumors are not clear.
TGF␤ is a cytokine in the tumor microenvironment that regulates various tumor progressions in a context-dependent manner (14). In early stage tumors, TGF␤ is a potent proliferation inhibitor that deters tumor growth; however, late stage tumors are often able to evade the growth inhibition and secrete elevated levels of TGF␤ to promote metastasis (15). The mechanisms by which late stage tumors use TGF␤ signaling to promote tumor spreading are largely unknown.
Accompanying the loss of capacity to differentiate during tumor progression, tumor cells undergo a transition in gene expression, reorganization of cytoskeleton, and acquisition of spindle cell morphology (16). Tumors with spindle cell morphology were characterized as highly malignant and invasive (17). In breast cancer, fascin is overexpressed in the estrogenϪ receptor-negative, basal-like subgroup (11), a highly metastatic group of breast cancers typically with spindle cell morphology (18). In melanoma, elongated, spindle-like tumor cells showed intense fascin staining, whereas rounded, amoeboid-like melanoma cells were generally fascin-negative (19).
Here, we demonstrate that TGF␤ elevates fascin protein expression and promotes invasion and filopodia formation in tumor cells with spindle morphology, but not in tumor cells with epithelial-like, polygonal morphology. We also show that transcription of fascin mRNA induced by TGF␤ is independent of de novo protein synthesis but relies on the canonical Smaddependent pathway. Furthermore, fascin is essential for TGF␤ to promote invasion and filopodia formation in spindle tumor cells. Therefore, our data suggest that fascin is an immediate TGF␤ target gene essential for its pro-invasion activity. Our data also suggest that TGF␤ might be responsible for the fascin overexpression in some metastatic tumors.
TGF␤ and Inhibitor Treatment-Unless stated otherwise, all cells were treated with 10 ng/ml TGF␤1 (Peprotech, Rocky Hill, NJ) in growth medium for 2 days before being used for assays. Inhibitors, when used, were added together with TGF␤ to growth medium.
Transwell Cell Migration Assay-Cells (1 ϫ 10 5 ) suspended in starvation medium were added to the upper chamber of an insert (6.5-mm diameter, 8-m pore size, BD Biosciences), and the insert was placed in a 24-well dish containing starvation medium with or without 10% FBS (21,22). Migration assays were carried out for 4 -6 h for spindle tumor cells and 12-24 h for polygonal tumor cells. Cells were fixed with 3.7% formaldehyde and stained with crystal violet staining solution, and cells on the upper side of the insert were removed with a cotton swab. Three randomly selected fields (10ϫ objectives) on the lower side of the insert were photographed, and the migrated cells were counted. The migration was expressed as the average number of migrated cells in a field.
Cell Invasion Assay-Cells (1 ϫ 10 5 ) suspended in starvation medium were added to the upper chamber of a Matrigel-coated insert (6.5-mm diameter, 8-m pore size, BD Biosciences), and the insert was placed in a 24-well dish containing medium with or without serum. Invasion assays were carried out for 16 h, and cells were fixed with 3.7% formaldehyde. Cells were stained with crystal violet staining solution, and cells on the upper side of the insert were removed with a cotton swab. Three randomly selected fields (10ϫ objectives) on the lower side of the insert were photographed, and the cells on the lower surface of the insert were counted.
Immunofluorescence Microscopy-A549 cells cultured on collagen-coated glass coverslips were fixed with 3.7% paraformaldehyde in PBS for 10 min at room temperature, permeabilized with 0.1% Triton X-100 for 5 min, and then washed with PBS three times. To block nonspecific binding, the cells were incubated with a solution of PBS containing 1% bovine serum albumin for 30 min and then incubated with Alexa Fluor 594-labeled phalloidin (Invitrogen). The coverslips were then fixed onto slides and imaged using a Zeiss fluorescence microscope.
Live Cell Imaging-A549 cells with or without TGF␤ pretreatment were plated on collagen-coated glass-bottomed 35-mm tissue culture dishes (MatTek) overnight. The membrane protrusion dynamics and cell movement were recorded with a differential interference contrast microscopy under a 40ϫ objective using a Zeiss inverted microscope equipped with a live imaging chamber. The temperature and CO 2 concentration in the chamber were maintained at 37°C and 5%, respectively. Time-lapse images were recorded at 10-s intervals.
qRT-PCR-Total RNA was extracted from cultured cells using TRIzol reagent (Invitrogen), and the reverse transcription was performed using the iScript cDNA synthesis kit (Bio-Rad). The quantitative real-time PCR (qRT-PCR) 3 assay was carried out with the Applied Biosystems 7900HT fast real-time PCR system using Applied Biosystems SYBR Green PCR master mix. Primers for qRT-PCR are shown in supplemental Table S1. All reactions were performed in triplicate, and the experiment was repeated three times.

TGF␤ Promotes the Migration of Spindle-shaped Tumor Cells-
It was recently reported that exposure to TGF␤ within the tumor microenvironment might predispose tumor cells for metastasis to distant organs (14). To evaluate the effects of TGF␤ exposure on the migration of tumor cells, we treated a panel of tumor cell lines with TGF␤. We used two breast cancer lines (MDA-MB-231, and MCF-7), two non-small cell lung cancer lines (A549 and H1299), and two melanoma lines (CHL-1 and WM115). Three of the six cell lines (MDA-MB-231, A549, and CHL-1) have spindle cell morphology, whereas the other three (MCF-7, H1299, and WM115) have polygonal, epithelial-like morphology (Fig. 1A). After 48 h of treatment, we found that TGF␤ induced morphology change in the three spindle-shaped cell lines, with cells becoming more elongated and disperse. Extremely long and finger-like protrusions were also observed in some TGF␤-pretreated cells. However, the effects of TGF␤ on the morphology of the three polygonal, epithelial-like tumor cells were much less noticeable (Fig. 1A).
Next, we examined the effects of TGF␤ pretreatment on cell migration using the Transwell migration assay. Interestingly, TGF␤ priming significantly increased the migration of the three spindle-shaped tumor cells by about 3-10-fold (Fig. 1B). In sharp contrast, TGF␤ had little effect on (MCF-7 and H1299) or slightly inhibited (WM115) the migration of the three polygon-shaped tumor cell lines (Fig. 1B).
To characterize the functional characteristics of spindleshaped and polygon-shaped tumor cells in the absence of TGF␤, we used the Transwell assay to evaluate their mobility and invasiveness. Tumor cells were allowed to migration/invade for 6 and 12 h, respectively. As shown in supplemental Fig.  S1, the three spindle-shaped tumor cell lines migrate and invade much faster than the three polygonal tumor cells, consistent with the notion that spindle-shaped tumor cells are more invasive and metastatic.
TGF␤ Induces Hyperactive Membrane Protrusions-Cell migration is a multistep process involving the formation of lamellipodia, turnover of focal adhesions, contraction of cell body, and retraction of trailing tail (23). To investigate the mechanisms underlying the enhanced cell migration in TGF␤pretreated spindle tumor cells, we used live cell imaging to study the membrane protrusion dynamics at the leading edge of migrating cells. As shown in supplemental Movies 1 and 2, TGF␤ pretreatment induced hyperactive membrane protrusion in A549 cells. The membrane protrusion dynamics in the TGF␤-pretreated cells followed a protrusion-ruffle-protrusion pattern (supplemental Fig. S2). New lamellipodia extended from the edge of membrane ruffles, sometimes expanding in the space between adjacent filopodia. At the end of the lamellipodium expansion, the edge of lamellipodia pulled back and formed membrane ruffles. New filopodia or lamellipodia would extend from the edge of , and CHL-1) became more elongated and scattered after TGF␤ treatment. Tumor cells were cultured in growth medium with or without 10 ng/ml TGF␤. The cell morphologies were recorded with phase contrast microscopy using a 10ϫ objective. Scale bar, 50 m. B, the effects of TGF␤ pretreatment on tumor cell migration. TGF␤ pretreatment promoted cell migration in spindle-shaped tumor cells, but not in polygon-shaped tumor cells. The data presented are mean Ϯ S.D. of migrated cells per field from three randomly selected 10ϫ fields. C-F, kymograph analysis of membrane protrusion dynamics in control A549 cell (C and D) and TGF␤-pretreated A549 cells (E and F); C and E, individual frames from time-lapse movies (supplemental Movies 1 and 2) used to generate the kymographs (D and F). Descending and ascending contours in the kymographs indicate membrane protrusion and withdrawal events, respectively. Arrowheads in D and F indicate new membrane protrusions. G, the extension of membrane protrusions in control and TGF␤-treated cells. NOVEMBER 11, 2011 • VOLUME 286 • NUMBER 45 membrane ruffles and begin the next protrusion cycle. In contrast, lamellipodia usually pulled back without membrane ruffle formation in control cells (supplemental Fig. S2 and supplemental Movies 1 and 2).

Fascin in TGF␤-induced Tumor Cell Invasion
Next, we employed kymography to analyze the membrane protrusion kinetics in control cells and TGF␤-pretreated cells. A line was drawn at the leading edge of a migrating cell; intensity values along the defined line region in each image of a time-lapse series were extracted and assembled together sideby-side to generate a kymograph montage (Fig. 1, C-F). As shown in the kymograph in Fig. 1, C and D, transient protrusions frequently extended out from the lamellipodium region of the control A549 cells and quickly withdrew to the starting point, generating spike-like structures on a flat basal line on the kymograph. The leading edge of the control A549 cells barely moved forward during the recording period (Fig. 1, D and G). In contrast, the kymographs of TGF␤-pretreated cells appeared as stair-like structures, indicating that the lamellipodia in these cells extended in "bursts"; in addition, the newly formed protrusions␤ in TGF␤-treated cells were able to hold their positions for a period of time until the next extension, instead of withdrawing to the starting point in the control cell. Consequently, the leading edges moved steadily forward following each extension (Fig. 1, F and G). The hyperactive membrane protrusions may contribute to increased mobility in TGF␤treated cells.
TGF␤ Induces Actin Cytoskeleton Remodeling-The morphology changes and hyperactive membrane protrusions suggested remodeling of the actin cytoskeleton in TGF␤-treated spindle tumor cells. Therefore, we used fluorescence microscopy to visualize actin filaments in the cells. As shown in Fig.  2A, the phalloidin staining in control A549 cells showed typical cortical actin staining, with the majority of F-actin in the cortical region. On the outer periphery of the strong actin staining, numerous weak and wavy filopodia-like protrusions were also observed in most of the control cells ( Fig. 2A, gray arrowheads). In contrast, stress fibers were noted in the majority of the TGF␤-pretreated A549 cells, consistent with previous observations that TGF␤ promoted stress fiber formation (24,25). Most strikingly, TGF␤ induced very long and straight filopodia (filopodia protruded more than 20 m out of the cell boundary) ( Fig. 2A). Typically, several such long filopodia extended around the cells, rendering a distinct "spiky" morphology in TGF␤-treated A549 cells. The extremely long filopodia were observed in 70% of TGF␤-treated cells (110 out of a total of 158 cells); in contrast, only less than 5% (8 out of a total of 228 cells) of control cells had such filopodia (Fig. 2B).
Elevated Fascin Expression in Spindle Tumor Cells after TGF␤ Treatment-Filopodia are finger-like protrusions critical for tumor cell invasion and metastasis (26). Cross-linking of parallel actin filaments is considered to be crucial for filopodia formation as individual filaments lack the rigidity required to overcome the compressive forces from plasma membrane (3). Fascin is a key actin-bundling protein in filopodia (27), and fascin overexpression has been reported in all carcinomas examined to date (4). The long filopodia in TGF␤-pretreated cells prompted us to examine the possibility that TGF␤ might regulate fascin expression in tumor cells. The effects of TGF␤ treatment on fascin protein levels in these tumor cells were evaluated with Western blotting using a fascin-specific antibody. TGF␤ treatment elevated fascin protein levels by 2-6fold in the three spindle tumor cells. In sharp contrast, no TGF␤-induced fascin expression was observed in any of the polygonal tumor cells (Fig. 2, C and D).
To investigate whether TGF␤ regulates the degradation of fascin protein, we used cycloheximide (CHX), a protein synthesis inhibitor, to inhibit the new synthesis of fascin in A549 cells and used Western blotting to monitor fascin degradation at different time points after CHX treatment. No degradation of fascin protein in control cells or in TGF␤-treated cells was detected even 24 h after CHX treatment, indicating that fascin is a very stable protein (data not shown). Next, we investigated whether TGF␤ regulated fascin expression at the mRNA level in two spindle tumor cell lines (A549 and MDA-MB-231 cells) and one polygonal cell line (MCF-7 cells) using qRT-PCR. Consistent with our Western blot results, we found that TGF␤ treatment elevated fascin mRNA levels in A549 and MDA-MB-231 cells, but not in MCF-7 cells, suggesting that TGF␤ promoted the transcription of fascin only in spindle tumor cells, but not in polygonal tumor cells (Fig. 2E).

Induction of Fascin Expression by TGF␤ Requires Smad3 and
Smad4-To explore the mechanisms through which TGF␤ elevates fascin expression, we first used SB-431542, a TGF␤ receptor I (T␤RI) kinase inhibitor, to inhibit the activity of the TGF␤ receptor complex. Treatment with SB-431542 completely blocked the overexpression of fascin in TGF␤-treated MDA-MB-231 cells and A549 cells, suggesting that T␤RI was essential to mediating the TGF␤ effects (Fig. 3A).  NOVEMBER 11, 2011 • VOLUME 286 • NUMBER 45 TGF␤ could regulate gene transcription either through the Smad-dependent pathways or through the Smad-independent pathways by activation of Erk, JNK, and p38 MAPK (28). We first examined the role of Erk and JNK because it was recently reported that the core promoter region of fascin contains a binding site for AP-1/CREB, which could be activated by Erk and JNK through phosphorylation (29,30). The addition of the MEK inhibitor UO126 and JNK inhibitor SP600125 significantly reduced the amount of phospho-Erk 1/2 and phosphoc-Jun, respectively, in both control cells and TGF␤-treated cells, indicating the successful inhibition of MAPK and JNK activity (Fig. 3, B and C). However, neither UO126 nor SP600125 had any effect on TGF␤-induced fascin expression (Fig. 3, B-D). We further examined the role of p38 MAPK using p38 kinase inhibitor SB203580 (supplemental Fig. S3). Treatment with p38 inhibitor failed to inhibit TGF␤-induced fascin expression at the mRNA level or the protein level, suggesting that ERK, JNK, and p38 MAPK were not involved (supplemental Fig. S3 and Fig. 3, B-D).

Fascin in TGF␤-induced Tumor Cell Invasion
To explore whether TGF␤-induced fascin expression is mediated through the TGF␤-Smad pathway, we employed small hairpin RNA (shRNA) to knock down the expression of Smad4. The knockdown of Smad4 was confirmed with qRT-PCR (supplemental Fig. S4C). MDA-MB-231 and A549 cells stably expressing control shRNA or Smad4 shRNAs were cultured in TGF␤ or in control medium, and the expression of fascin was evaluated with qRT-PCR and Western blotting. As shown in Fig. 3, E and G, Smad4 shRNA almost completely inhibited TGF␤-induced fascin expression at both the mRNA and the protein level, arguing that fascin is regulated by TGF␤ through the canonical T␤RI-Smad pathway. To investigate whether either or both of the two receptor-regulated Smads (R-Smads) are involved, we further used shRNA to decrease the expression of Smad2 and Smad3 (supplemental Fig. S4, A and  B). Smad3 shRNA, but not Smad2 shRNA, reduced the TGF␤induced expression in both MDA-MB-231 cells and A549 cells, suggesting that fascin expression is regulated by TGF␤ through the T␤RI-Smad3-Smad4 pathway, whereas Smad2 is not required (Fig. 3, E and F).
Characterization of TGF␤ Signaling in Spindle-shaped and Polygon-shaped Tumor Cells-To gain insight into the differential regulation of fascin by TGF␤ in spindle-and polygonshaped tumor cells, we used qRT-PCR to examine the expression levels T␤RI and T␤RII in the six cell lines (Fig. 3H). As shown in Fig. 3H, TGF␤ receptor I and receptor II mRNA levels were higher in the spindle-shaped tumor cells (MDA-MB-231, A549, and CHL-1)than in the polygonal cells (MCF7, H1299 and WM115). We further examined the total Smad3 and phoso-Smad3 level in the two breast cancer cell lines (MDA-MB-231 and MCF7) and the two non-small cell lung cancer cell lines (A549 and H1299). As shown in Fig. 3I, both the total Smad3 and the TGF␤-induced phosho-Smad3 levels are higher in the MDA-MB-231 and A549 cells when compared with MCF7 and H1299 cells, respectively. The lower T␤RI, T␤RII, and phospho-Smad3 levels in the polygonal tumor cells may lead to global inhibition of TGF␤ responses. To examine this possibility, we examined the effect of TGF␤ treatment on the transcription of several other genes regulated by TGF␤, includ-ing p21 Cip , p27 Kip , N-cadherin, and vimentin, and we noted that TGF␤ failed to induce the transcription of any of these genes in MCF-7 cells (supplemental Fig. S5). Therefore, the global inhibition of TGF␤ response in some cells might at least partially explain the lack of TGF␤-induced fascin expression.

Induction of Fascin Expression by TGF␤ Requires No de Novo
Protein Synthesis-To explore the kinetics of TGF␤-induced fascin mRNA expression, we used qRT-PCR to determine fascin mRNA levels in MDA-MB-231 cells after treatment with TGF␤ for different periods of time. As shown in Fig. 4A (upper  panel), there was a 3-h lag in TGF␤-induced fascin transcription. Fascin mRNA levels increased steadily after the lag, reaching the plateau level (ϳ6.4-fold) after 24 h and remaining at this plateau for at least another 24 h. Next, we further evaluated the effects of TGF␤ on the protein expression kinetics of fascin. As shown in Fig. 4A (lower panel), TGF␤ significantly elevated the fascin protein level after 24 h, and the fascin protein level remained high for up to at least 72 h. A similar time course was also observed in A549 cells (data not shown), suggesting that TGF␤ was able to induce and maintain elevated fascin expression in spindle tumor cells.
We further examined the kinetics of Smad3 phosphorylation in MDA-MB-231 cells. Smad3 phosphorylation was increased by more than 10-fold 3 h after TGF␤ treatment. Phospho-Smad3 level decreased gradually after 3 h but remained at a fairly high level for at least 26 h (Fig. 4B).
Next, we investigated whether TGF␤-induced fascin transcription required de novo protein synthesis. TGF␤ regulates the expression of a plethora of genes. Some are immediately downstream of the TGF␤ signaling cascade, whereas others are indirectly modulated through TGF␤-regulated transcription factors (31,32). To determine whether the induction of fascin gene transcription by TGF␤ was immediately downstream of the TGF␤ signaling cascade, we used CHX to block de novo protein synthesis. CHX treatment did not change the TGF␤induced expression of fascin mRNA in MDA-MB-231 cells and A549 cells, indicating that elevated fascin expression does not require new protein synthesis (Fig. 4C). As controls, CHX treatment successfully blocked the inhibition of E-cadherin transcription and the induction of N-cadherin mRNA expression by TGF␤ (Fig. 4D), confirming that gene transcriptions indirectly regulated by TGF␤ through other transcription factors required new protein synthesis.
It has been previously reported that TGF␤ signaling accelerates the degradation of Smad co-repressors such as Ski and SnoN through the ubiquitin-proteasome pathway (33). To determine whether the ubiquitin-proteasome pathway might be required of TGF␤-induced fascin transcription, we used proteasome inhibitor MG132 to treat MDA-MB-231 cells. As shown in Fig. 4E, treatment with MG132 significantly decreased TGF␤-induced fascin transcription, indicating that degradation of co-suppressors might be involved in the regulation of fascin expression.
Fascin Is Required for TGF␤-induced Invasion and Filopodia Formation-To explore the role of fascin in TGF␤-induced actin cytoskeleton remodeling and cell migration, we employed shRNA to knock down fascin expression in spindle tumor cells. The successful knockdown of fascin protein was confirmed by Western blotting (Fig. 5A). To assess the role of fascin in TGF␤-induced filopodia formation, A549 cells stably expressing control shRNA or fascin shRNA were treated with TGF␤ and stained for F-actin. TGF␤ induced long filopodia and spiky morphology in control shRNA cells (Fig. 5B). TGF␤-induced long filopodia were noted in about 67% of A549 cells expressing control shRNA (183 out of a total of 274 cells examined) (Fig. 5C). In contrast, TGF␤-induced long filopodia were only observed in roughly 20% of A549 fascin shRNA cells (60 out of a total of 294 cells examined), suggesting that fascin was critical for TGF␤-induced filopodia formation (Fig. 5, B and C).
Next, we investigated the role of fascin in TGF␤-promoted spindle tumor cell migration. Using the Transwell migration assay, we found that TGF␤-promoted cell migration was significantly lower in fascin shRNA-treated cells when compared with the control shRNA-expressing cells, suggesting that fascin was important for TGF␤-induced cell migration. (Fig. 5D).
To further assess the role of fascin in TGF␤-promoted metastasis in spindle cell tumors, we evaluated the effects of TGF␤ on the invasion ability of spindle tumor cells stably expressing fascin or control shRNA. TGF␤ pretreatment promoted the invasion of control shRNA-expressing spindle tumor cells by about 3-5-fold, consistent with the pro-metastasis function of TGF␤. Expression of fascin shRNA dramatically inhibited TGF␤-promoted invasion in all three of the spindle tumor cells. In contrast to the 3-5-fold invasion increase in control shRNA cells, TGF␤ was only able to promote invasion by about 1.2-1.5-fold in fascin shRNA cells (Fig. 5E).
To further evaluate the hypothesis that fascin is critical for TGF␤-induced tumor cell migration and invasion, we overexpressed fascin in tumor cells expressing Smad4 shRNA. As shown in Fig. 5, F and G, Smad4 knockdown nearly completely abolished TGF␤-induced migration and invasion in MDA-MB-231 cells, consistent with the observation that Smad4 was␤ essential for TGF␤-induced fascin expression; ectopic fascin expression restored shSmad4 cell migration and invasion to levels similar to TGF␤-primed control cells, consistent with the notion that fascin is critical for TGF␤-induced migration and invasion in metastatic tumor cells. Correlation between TGF␤1, T␤RI, and Fascin Expression Levels in Primary Breast Tumor Samples-To validate our cell culture-based discoveries in breast cancer patients, we examined the correlation between fascin levels and TGF␤ as well as TGF␤ receptor levels in a cohort of 99 primary breast tumor samples collected at the Memorial Sloan-Kettering Cancer Center (34). Pearson correlation coefficient (r) and probability (p) values between the two fascin probe sets and each of the probe sets for TGF␤1, TGF␤2, and TGF␤3 and TGF␤ receptor I and II were calculated. Fascin expression significantly correlates with TGF␤1 and TGF␤ receptor I, but not with TGF␤2 and TGF␤3 or TGF␤ receptor II (Fig. 6A and data not shown). Next, tumor samples were sorted according to TGF␤1 expression level (Fig. 6B). The 50 tumor samples with TGF␤1 level at or below median level were assigned to the "TGF␤ low" group, and the other 49 samples with TGF␤1 level above the median level were assigned to the "TGF␤1 high" group. Fascin (FSCN1) levels in the TGF␤1 high group were about 2-fold as high as the TGF␤1 low group (p Ͻ 0.002) (Fig. 6C), suggesting that elevated TGF␤1 levels in primary breast tumors contributed to fascin overexpression.

DISCUSSION
Fascin Expression Is Regulated by TGF␤-Although there are a plethora of studies about fascin overexpression in various carcinomas (4,5,7,8,11,12), the factors contributing to fascin overexpression in metastatic tumors remain largely unknown. Binding sites for several transcription factors, including ␤-catenin-TCF (T-cell factor), CREB, and AP-1, in the FSCN1 promoter region have been previously reported (5,29,30,35). However, it is not clear whether and how factors in the tumor microenvironment contribute to fascin overexpression in metastatic tumor cells. Our results here demonstrated that TGF␤ promoted fascin expression in spindle tumor cells, implicating that cytokines in the tumor microenvironment could affect fascin expression.  The up-regulation of fascin by TGF␤ was surprising in a sense because TGF␤ target genes have been studied previously in various epithelial and tumor cell lines (including MDA-MB-231), and fascin was not among the TGF␤ response genes identified through microarray screening (14,36). We noticed that previous screenings focused on early response genes and that tumor cells were treated with TGF␤ for no more than 3 h (14,36). Therefore, we explored the kinetics of TGF␤-induced transcription of fascin. Our data revealed a 3-h lag in TGF␤-induced fascin expression (Fig. 3D), which may explain the discrepancy between our data and previous studies. The lag might be due to the requirement for TGF␤-induced degradation of Smad co-suppressors as proteasome inhibitor MG132 inhibited TGF␤-induced fascin transcription. Despite the lag, fascin expression is likely directly downstream of TGF␤ signaling because TGF␤ was able to promote the expression of fascin mRNA in the presence of CHX, a protein synthesis inhibitor.
Fascin Is Regulated by TGF␤ through the Canonical T␤RI-Smad Pathway-Our data further demonstrated that fascin expression was regulated through the canonical T␤RI-Smad pathway. Smads in the TGF␤ pathway recognized the consensus DNA sequence CAGAC (37). To determine whether the human fascin 1 gene contains Smad-binding sites, we examined the promoter region of FSCN1 and discovered that there were two CAGAC sequences at Ϫ381 and Ϫ1225 positions, respectively. It is possible that activated Smad3 and Smad4 induced fascin expression by binding to one or both of the Smad-binding sites.
It is intriguing to note that TGF␤ only induced fascin expression in spindle tumor cells, but not in polygonal cells. One way for tumor cells to overcome growth inhibition effects exerted by TGF␤ is through loss of expression or functional inactivation of T␤RI and T␤RII (38). We noted that that T␤RI and T␤RII levels in the three spindle tumor cells were higher than the three polygonal cells. The reduced expression of TGF␤ receptors might partially explain the lack of TGF␤-induced fascin expression in polygonal cells.
The Smad co-activator context in cells may also contribute to the regulation of fascin by TGF␤. Although Smad3 and Smad4 can directly bind to DNA to mediate gene transcription, interaction between Smads and their target sequence is of low affinity, and DNA-binding co-factors are required to provide specific regulation of gene transcriptions (37). Consequently, the transcriptomic output of activated Smads is highly dependent on the cellular context of Smad co-factors (15). The conversion from polygonal morphology to spindle morphology during tumor progression is a recessive event resulting from the change of gene expressions that control epithelial differentiation (16). It is possible that the transition from epithelial-like to mesenchymal-like tumor cells provides the cellular co-factor context required for the induction of fascin expression by activated Smads. We noted that some polygonal tumor cells (e.g. H1299 and WM115, data not shown) have fairly high fascin levels when compared with epithelial cells, suggesting that factors other than TGF␤ also contribute to fascin overexpression in tumors.
Role of Fascin in TGF␤-promoted Tumor Metastasis-The presence of TGF␤ in the tumor microenvironment has pro-found impacts on tumor progression and metastasis (14,15,17). Activated TGF␤ signaling in primary tumor was linked to lung metastasis in estrogen receptor-negative breast tumors (14). Our data demonstrated that TGF␤ induced fascin expression and promoted migration/invasion in MDA-MB-231 breast tumor cells. More importantly, TGF␤1 and TGF␤ receptor I significantly correlate with fascin (FSCN1) level levels in a cohort of primary breast tumors, suggesting that activation of TGF␤ signaling may contribute to fascin overexpression in primary tumors. Interestingly, fascin overexpression in breast tumor has been associated with lung metastasis, and fascin is one of the top performing lung metastasis signature genes (34). It is possible that elevated fascin expression is critical for TGF␤ to promote tumor cell migration, invasion, and lung metastasis in estrogen receptor-negative breast tumors. Indeed, TGF␤ promoted migration and invasion in all three spindle tumor cell lines with elevated fascin expression, but not in the three polygonal tumor cells, in which TGF␤-induced fascin expression was absent.
More importantly, knockdown of fascin expression with shRNA significantly inhibited migration and invasion in TGF␤treated spindle tumor cells. The near complete inhibition of TGF␤-promoted invasion by fascin shRNA was quite remarkable considering that TGF␤ also up-regulates the expression of other pro-invasion genes such as matrix metalloproteases 2 and 9 (17,39). During invasion, tumor cells use protrusion structures termed invadopodia to secrete matrix metalloproteases and to coordinate the degradation of extracellular matrix (40). The formation and elongation of invadopodia requires filopodia-associated protein components (41). Fascin is the major actin-bundling protein in filopodia protrusions (27). It was recently reported that fascin was critical to stabilize actin in invadopodia (26,41,42). Depletion of fascin with shRNA destabilized invadopodia and impaired extracellular matrix degradation (42). Given the similar components and regulatory mechanism between invadopodia and filopodia, it has been suggested that invadopodia are "invasive filopodia" (26,42). Our data showed that fascin was required for long filopodia induced by TGF␤. It is possible that fascin played a central role in the invasive machinery mobilized by TGF␤ in spindle tumor cells. By stabilizing actin cytoskeleton within invadopodia and filopodia, fascin coordinates the invasion and metastasis promoted by TGF␤.