Caveolin-1 Up-regulation during Epithelial to Mesenchymal Transition Is Mediated by Focal Adhesion Kinase*

Emerging evidence has shown that caveolin-1 is up-regulated in a number of metastatic cancers and can influence various aspects of cell migration. However, in general, the role of caveolin-1 in cancer progression is poorly understood. In the present study, we examined alterations in caveolin-1 expression during epithelial-to-mesenchymal transition (EMT) and the ability of caveolin-1 to alter cancer cell adhesion, an aspect of cell motility. We employed two EMT cell models, the human embryonic carcinoma cell line NT2/D1, and TGF-β1-treated NMuMG cells, which are derived from normal mouse mammary epithelia. Caveolin-1 expression was substantially up-regulated in both cell lines following the induction of EMT and was preceded by increased activation of focal adhesion kinase (FAK) and Src, two known tyrosine kinases involved in EMT. We hypothesized that caveolin-1 expression could be influenced by increased FAK phosphorylation, to which Src is a known contributor. Examination of FAK+/+ and FAK-/- mouse embryonic fibroblasts revealed that in cells devoid of FAK, caveolin-1 expression is strikingly diminished. Using FAK and superFAK constructs and the novel FAK inhibitor PF-228, we were able to demonstrate that indeed, FAK can regulate caveolin-1 expression. We also found that Src can contribute to increases in caveolin-1 expression, however, only in the presence of FAK. From the culmination of this data and our functional analyses, we conclude that caveolin-1 expression can be up-regulated during EMT, and further, once expressed, caveolin-1 can greatly influence cancer cell adhesion.

Caveolin-1 (Cav-1), 2 the principal structural protein of the cholesterol-rich plasma membrane invaginations known as caveolae (1), was first discovered as a Src substrate in Rous sarcoma virus-transformed cells (2,3), and appears to play opposing roles in the context of cancer (4). Studies have labeled it both as a tumor suppressor (5)(6)(7) and an oncogene (8 -12). Support for the tumor suppressor theory includes the fact that caveolin-1 is known to sequester and dampen the activity of a variety of signaling molecules that can cause cell transformation (13) and the localization of caveolin-1 near a tumor suppressor locus on chromosome 7 (14). Clues that caveolin-1 may be serving a tumor promoter effect first arose with the observation that prostate cancer tissue expresses more caveolin-1 than does normal tissue (15). The list has since expanded and includes esophageal squamous cell carcinoma, multiple myeloma, Ewing's sarcoma family tumors, clear cell renal carcinoma, urinary bladder tumors, and non-small cell lung cancer (4,8,16,17). This suppressor/promoter discrepancy could be due to the examination of various cell types, cancer stages, variations in in vitro versus in vivo data, or an indicator that caveolin can serve different roles depending on the context in which its function is being examined. One trend that has more recently arisen, however, is that caveolin-1 is up-regulated in a number of metastatic cancers including prostate and lung, and consistent with this, we and others (18 -21) have shown that caveolin-1 plays an important role in cell migration.
Epithelial-to-mesenchymal transition (EMT) is a process that occurs during embryonic development (as stem cells differentiate into various components of the organism) (22) and during cancer progression (as tumor cells gradually acquire a more motile phenotype). Cancer cells that undergo EMT, which entails an intricate series of changes in cell-cell contacts (23), cell-matrix interactions, and cell signaling (24), can gain the ability to invade, extravasate, and potentially re-establish as a metastatic lesion in a lymph node or distant organ. The integrin and cadherin families of proteins demonstrate changes in expression during the process of EMT, and recently, a crossregulation was described between these families via the small GTPase Rap1 (25). Correspondingly, increased activation of FAK and Src, both players in focal adhesion complexes, has been shown in tumor progression/metastasis, and this activation can greatly enhance cell invasion (26 -29). Recently, FAK has also been implicated in the regulation of cadherins, with reports of FAK decreasing Rac1 activity and affecting N-cadherin adhesions (30) and FAK localizing to/aiding in the disruption of E-cadherin complexes (31). It is well documented that Src can contribute largely to the phosphorylation of FAK (32,33), hence making Src an important consideration on processes affected by FAK activity.
Realizing the complexity of caveolin in cancer and the potential of caveolin-1 expression to change during EMT, we wanted to examine two EMT cell models, NT2/D1 and NMuMG, to examine if and in what cellular context (denoting potential sources of regulation) caveolin-1 expression is altered during EMT. Further, we wanted to determine the impact of caveolin-1 on aspects of cancer cell adhesion, because of the implicated role of caveolin in cell motility (18,19,34). The pluripotent human embryonic carcinoma cell line NT2/D1 has the potential to differentiate into a variety of cell types upon appropriate stimulation (35)(36)(37) and can accordingly undergo EMT. It is well documented that the mouse mammary epithelial cells NMuMG undergo EMT upon treatment with TGF-␤1 (38). Prior to induction of EMT, the NT2/D1 and NMuMG cells display scant caveolin-1 protein expression, thus allowing us to readily examine any alterations in expression during the course of EMT.
In our present study, both cell models employed demonstrated an up-regulation of caveolin-1 expression during EMT. Importantly, we found that both the presence and phosphorylation of FAK have a significant impact on caveolin-1 expression. The combination of our expression and functional data provide insight into the regulation of caveolin-1 expression during EMT and clues to the impact of this expression on aspects of cancer cell metastasis.

EXPERIMENTAL PROCEDURES
Materials-Antibodies were purchased from the following sources: caveolin-1, FAK, and E-cadherin from BD Bioscience (San Jose, CA); paxillin, pY31-paxillin, pY397-FAK, pY861-FAK, and pY418-Src from BIOSOURCE (Camarillo, CA); caveolin-1 and fibronectin from Santa Cruz Biotechnology (Santa Cruz, CA); actin (coumarin phallacidin) from Molecular Probes (Carlsbad, CA); Src from Upstate Biotechnology (Lake Placid, NY); ZO-1 and N-cadherin from Zymed Laboratories Inc. (South San Francisco, CA); and GAPDH from Research Diagnostics, Inc. (Concord, MA). Secondary antibodies and exposure reagents for Western blot analysis were purchased from Pierce. The transfection reagents Lipofectin and Oligofectamine were purchased from Invitrogen. All media and cell culture reagents were purchased from Invitrogen.
Cell Lines and Culture Conditions-The human embryonic carcinoma cell line NTERA2 cl. D1 (NT2/D1) and the control cell line 2102Ep were cultured in high glucose Dulbecco's modified Eagle's medium without sodium pyruvate and supplemented with 5 mM L-glutamine, 1% penicillin/streptomycin, and 10% fetal bovine serum. The mouse mammary epithelial cell line NMuMG and human lung cancer cell line H1703 were purchased from ATCC (Manassas, VA) and cultured as recommended. FAK Ϫ/Ϫ , SYF Ϫ/Ϫ , and wild-type (WT) mouse embryonic fibroblasts (MEFs) were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 1% L-glutamine, and 1% penicillin/streptomycin. Caveolin-1 WT and KO MEFs were isolated and cultured by our laboratory as previously described (21). All cell cultures were kept in 37°C incubators maintained under 5% CO 2 .
Induction of Differentiation-For NT2/D1 and control 2102Ep cells, EMT was induced by switching the culture media to a stimulatory media consisting of MCDB 131 supplemented with 20% newborn calf serum, 5% normal human serum, 0.05% bovine brain extract, 1% penicillin/streptomycin, and 1% L-glu-tamine. Induction of EMT in the NMuMG cells was accomplished by adding 2 ng/ml TGF-␤1 to the normal growth media.
Inhibitor Treatment-PF-573,228 (referred to as PF-228) is an inhibitor of FAK catalytic activity, as previously described (39). PF-228 was prepared as previously described (39), supplied by Pfizer Inc. (Central Research Division, Groton, CT), and used with the concentrations and for the times indicated in the individual experiments. Two previously characterized Src inhibitors, PP2 (40) and SU6656 (41) (Calbiochem, La Jolla, CA), were used, again as described in the text. Control cells were treated with DMSO only.
Cell Transfection and Lysate Preparation-The FAK-GFP and superFAK constructs were gifts from Dr. Michael Schaller (University of North Carolina, Chapel Hill). The active Src construct c-Src Y527F was a gift from Dr. Daniel Flynn (Mary Babb Randolph Cancer Center, Morgantown, WV). The WT caveolin-1 (21) and caveolin-1 Y14F constructs were generated in our laboratory by utilizing the EcoRI and BamHI sites of the pEGFP-N1 vector (BD Biosciences Clontech, Palo Alto, CA) and the QuikChange Site-directed Mutagenesis Kit (Stratagene, Cedar Creek, TX). Cells were transfected with 3 g of the desired plasmid DNA using either Lipofectin reagent (Invitrogen) or the Amaxa biosystems (Gaithersburg, MD) nucleofection system, both according to the manufacturer's detailed instructions. Complete medium was placed on the cells following transfection or nucleofection recovery. Cav-1 or control siRNA (as used previously in our laboratory, Ref. 18) was transfected into cells using oligofectamine. Cells were lysed by incubating cells with cold radioimmune precipitation assay buffer (50 mM Tris-HCl, pH 7.4, 5 mM EDTA, 150 mM NaCl, 0.5% deoxycholate, 0.1% Triton X-100, 2 mM Na 3 VO 4 , 1ϫ phenylmethylsulfonyl fluoride, and protease inhibitor) for 25 min or used in assays as described.
Wound Healing Scratch Assay-5 ϫ 10 5 cells (see individual experiments for cell types and treatment conditions/controls) were seeded, allowed to form a confluent monolayer, and then wounded by scraping the monolayer with a pipette tip. The wound areas and in-migrated cells in each group were monitored for 10 h and analyzed using NIH ImageJ software.
Boyden Chamber Assay-H1703 cells were treated with control siRNA, caveolin-1 siRNA (18), or mock treatment (oligofectamine only, Invitrogen). 48-h post-transfection, 5 ϫ 10 3 cells of each group were resuspended in 0.5% fetal bovine serum/RPMI and subjected in quadruplicate to the upper chambers of the Boyden apparatus. The lower well of the Boyden chamber contained RPMI 1640/10% fetal bovine serum. The polycarbonate membrane separating the upper and lower wells was coated in 10 g/ml fibronectin overnight at room temperature prior to the experiment. Cells were allowed to migrate for 3.5 h. The membrane was then fixed, stained using hematoxylin, and mounted. Migrated cells were manually quantified.
Western Blot and Densitometry Analysis-Boiled cell lysates were subjected to SDS/PAGE, electrotransferred to a nitrocellulose membrane, and immunoblotted with antibodies as shown in the results. Bands were visualized using Super Signal West Pico Stable Peroxide and Luminol/Enhancer solutions (Pierce). Densitometry analysis of blots and corresponding GAPDH controls was performed using the Eagle Eye II apparatus and Eagle Sight software (Stratagene, La Jolla, CA). All blots shown are representatives of experiments performed at least in duplicate.
Real-time PCR Analysis-Primers compatible with the Roche Universal Probe system were designed for mouse cav-1 (forward: aacgacgacgtggtcaaga and reverse: cacagtgaaggtggtgaagc) and human cav-1 (forward: acagcccagggaaacctc and reverse: gatgggaacggtgtagagatg). Total RNA was isolated using TRIzol reagent (Invitrogen) and converted into cDNA using the Promega Reverse Transcription System (Promega). RNA isolated from cav-1 Ϫ/Ϫ MEFs served as an experimental negative control for mouse-derived samples. All samples were also run in duplicate with primer/probe pairs for either 18 S (mouse samples) or GAPDH (human samples), which served as a total loading control. Relative mRNA levels were determined by using the comparative Ct method, where the cav-1 target is normalized to the control and then compared with a reference sample (assigned a relative value of 1) by the equation: 2 Ϫ⌬⌬Ct .
Electric Cell Substrate Impedance Sensing (ECIS)-1 ϫ 10 5 cells (see individual experiments for cell types and any additional treatment conditions) were seeded into fibronectincoated wells of an 8 W10E ECIS chamber/electrode. Cell attachment and spreading were immediately analyzed using the ECIS1600R (Applied BioPhysics, Troy, NJ) attachment program with data collection every 30 s at 4000 Hz for 3 h.
Immunofluorescence, Confocal Microscopy, and Manual Cell Analysis-Cells were seeded onto fibronectin-coated glass coverslips (10 g/ml in 1ϫ phosphate-buffered saline) and fixed using 2% paraformaldehyde. Cells were then permeabilized using 0.5% Triton X-100, washed, and nonspecific binding was blocked by application of 5% normal goat serum for 30 min. Following primary antibody incubation, cells were washed and incubated with the appropriate secondary antibody (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) for 30 min. Finally, slips were mounted using the anti-fade reagent Fluoromount-G (SouthernBiotech, Birmingham, AL). Cells were imaged using a 40ϫ water objective on a Zeiss Axioplan 2 upright microscope and images were captured via the Zeiss LSM510 laser scanning confocal system (NIOSH, Morgantown, WV). Cell spreading area and number of focal adhesions were manually quantified using NIH Image J software. The Student's t test was used to analyze statistical significance, with a p value of less than 0.05 considered significant.

Stimulated NT2/D1 and NMuMG Cells Both Undergo EMT-
To investigate potential alterations in caveolin-1 expression during EMT, we employed two cell models, the human embryonic carcinoma cell line NT2/D1 and the mouse mammary epithelial cell line NMuMG. The NT2/D1 cells have the potential to undergo EMT and be differentiated into a number of cell types, including neuronal-like cells in response to retinoic acid (42). To characterize the changes occurring in the NT2/D1 cells using our stimulation method (see "Experimental Procedures"), we first immunofluorescently labeled unstimulated cells and cells that had been stimulated for 1 week for SSEA-4 and SSEA-1, markers for undifferentiated and differentiating human cells, respectively. The stimulated cells demonstrated a sizable loss of SSEA-4 and a gain in cytoplasmic SSEA-1 staining (data not shown), indicating that these cells are indeed differentiating following stimulation. The differentiating cells were next examined for evidence of EMT. The morphological differences seen between the undifferentiating and differentiating NT2/D1 cells included cell elongation and a decrease in cell-cell contacts (Fig. 1, A and B). A classically described signature of EMT is the down-regulation of epithelial markers, such as E-cadherin, and up-regulation of mesenchymal markers, such as N-cadherin and fibronectin (43). Western blot analysis of the stimulated NT2/D1 cells revealed that E-cadherin expression was significantly suppressed 24 h post-stimulation, and scant E-cadherin expression remained at day 4 ( Fig. 1C, top panel). In contrast, fibronectin, an EMT marker in some cell lines (43), was barely expressed in the unstimulated NT2/D1 cells, but was substantially up-regulated as the cells differentiated (Fig. 1C). Further, N-cadherin was slightly up-regulated by day 4 (Fig. 1C). As E-cadherin, N-cadherin, and the junctional protein ZO-1 have been shown to not only alter their expression but also cellular localization during EMT (23, 44 -46), we decided to immunofluorescently label the day 0 and day 4 NT2/D1 cells for these proteins to examine their staining patterns. Fig. 1G (top panels) shows that stimulation of the NT2/D1 cells causes a weakened fluorescence intensity and decrease in cell-cell junction localization of both E-cadherin and ZO-1. In contrast to this, N-cadherin staining intensity is enhanced in the stimulated NT2/D1 cells. Classically, cells that undergo EMT demonstrate enhanced motility and invasive capabilities (47). By using a scratch wound healing assay, we show that the stimulated NT2/D1 cells migrate more effectively and fill the wound-gaps up to 6-fold faster than that of the unstimulated cells (Fig. 1H). Day 0 and day 4 NT2/D1 cells were also subjected to collagen-based Transwell invasion assays, and the day 4 cells showed an acquired ability to invade compared with the day 0 cells, which were actually unable to invade (data not shown). Finally, we used real-time PCR analysis to examine Snail expression (a potent contributor to E-cadherin repression during EMT, Ref. 48) in day 0 to day 4 NT2/D1 cells and found that Snail mRNA was significantly up-regulated following stimulation of these cells (data not shown).
TGF-␤1 treatment of NMuMG cells is a well documented cell model to examine epithelial to mesenchymal transition (38,49,50). Treatment of cells with TGF-␤1 has been shown to trigger cell motility (51,52), with TGF-␤1 prompting NMuMG cells to switch from migrating as a slow moving sheet of cells to migrating more effectively as individual, mesenchymal-like cells (50). Therefore, to confirm the morphological and molecular changes in the NT2/D1 cells, we chose to examine EMT in the NMuMG cells. Fig. 1, D and E show the morphology differences seen in NMuMG cells that have been cultured in either normal growth media or growth media supplemented with 2 ng/ml TGF-␤1 for 4 days, respectively. Western blot analysis revealed that TGF-␤1 treatment increased N-cadherin expression and slightly decreased E-cadherin (Fig. 1F), as previously reported (53). Similar to the results seen in the NT2/D1 cells, past reports have also shown increases in fibronectin but no change in vimentin expression in TGF-␤1-treated NMuMG cells (38,53). Immunofluorescence labeling revealed that TGF-␤1 stimulation of the NMuMG cells caused similar changes in the staining intensity and cellular localization of E-cadherin, ZO-1, and N-cadherin as that seen in the NT2/D1 cells (Fig. 1G), again suggesting a switch to a mesenchymal cell phenotype. Finally, we wanted to verify that the TGF-␤1 treatment enhanced the ability of NMuMG cells to migrate. Day 0 or day 4 NMuMG cells were seeded and subjected to a scratch wound healing assay. As shown in Fig. 1H, TGF-␤1 treated NMuMG cells do indeed increase their ability to migrate.
The collective changes in each of these cell lines following stimulation are consistent with characteristic changes in cells undergoing an epithelial to mesenchymal transition (54 -56), which includes a decrease in cell-cell contacts and increased migratory capabilities. Therefore, we decided to utilize these two EMT cell models to explore alterations in caveolin-1 expression during EMT and the cellular context in which any changes occur.
Caveolin-1 Expression Is Up-regulated during EMT-Prior to induction of EMT, the NT2/D1 and NMuMG cell models showed little caveolin-1 protein expression. NT2/ D1 cells were therefore stimulated for a period of 4 days, as shown in Fig. 2A. Starting at day 2, caveolin-1 expression is increased in these cells. Real-time PCR analysis revealed a significant increase in caveolin-1 mRNA between the day 0 and day 4 cells, suggesting the noted up-regulation of caveolin-1 seen in the NT2/D1 cells may result from an increase in mRNA (data not shown). Even after 3 weeks, caveolin-1 protein expression was still maintained (data not shown) indicating that this increased caveolin-1 expression was not simply a transient occurrence in these cells. Fielding et al. (57) reported that cholesterol and other sterols have the ability to alter caveolin-1 mRNA levels. Therefore, to ensure that the contents of our high serum media did not cause caveolin-1 expression to be altered independent of the differentiation process, we employed a control cell line, 2102Ep, which is a human embryonic carcinoma cell line that is unable to differentiate (35,58) and thereby also does not undergo EMT. After receiving identical stimulation as the NT2/D1 cells, indeed, no morphological changes occurred in these cells. Further, even after 3 weeks of culturing  MAY 16, 2008 • VOLUME 283 • NUMBER 20 the 2102Ep cells under the stimulatory conditions, no caveolin-1 expression could be detected (data not shown).

FAK-mediated Caveolin-1 Up-regulation during EMT
To confirm EMT-associated increases in caveolin-1 expression, NMuMG cells were treated with 2 ng/ml TGF-␤1 for the times indicated in Fig. 2, and caveolin-1 expression was determined by Western blot analysis. As seen in Fig. 2B, caveolin-1 expression is up-regulated ϳ2-fold by day 3 post-initiation of stimulation. As with the NT2/D1 cells, this increase was not transient, as caveolin-1 expression remained elevated in the NMuMG cells with longer TGF-␤1 treatment times (data not shown). Thus, these results demonstrate that, following changes in cell morphology (see Fig. 1), caveolin-1 can be upregulated and is seemingly a product of EMT.
Phosphorylation of FAK and Src Is Increased during EMT-As discussed, caveolin-1 has been implicated in affecting the phosphorylation of FAK and Src, both of which can also influence EMT (26). We therefore decided to examine the phosphorylation of FAK and Src in relation to caveolin-1 expression in our two cell models during EMT. We subjected day 0 to day 4 NT2/D1 cell lysates to Western blot analysis for pFAK (Y397), pFAK (Y861), FAK, pSrc (Y418), and Src (Fig. 3A). Total FAK and Src showed no change in expression throughout; however, FAK and Src phosphorylation did change over time. FAK-Y397 becomes phosphorylated (and peaks at day 2), and Src shows a large spike in Tyr-418 phosphorylation at day 1. Correspondingly, the phosphorylation of FAK at Tyr-861, which is Src-dependent (32), also increased (and peaked) at day 1 (Fig. 3A).
Similar changes in the day 0 to day 4 NMuMG cells can be seen in Fig. 3B. Notably, pSrc (Y418) shows a transient increase at day 1 post-stimulation, and is correlated the phosphorylation of FAK at Tyr-861. Increases in phosphorylation of FAK at Tyr-861 has been previously described in NMuMG cells during EMT (53,59).
It has been reported that the knockdown of caveolin-1 expression causes a loss in FAK phosphorylation upon integrin clustering (60). In fact, in a study examining caveolin-1 expression in small and non-small cell lung cancers, cell lines lacking caveolin-1 expression also showed no phosphorylation of FAK, although total FAK was unaffected (8). Based on these reports, we found the increases in pFAK (at tyrosines 397 and 861) prior to increases in caveolin-1 expression in the NT2/D1 and NMuMG cells during EMT very intriguing. We therefore hypothesized that FAK may be a mediator of caveolin-1 expression.
FAK Is a Critical Mediator of Caveolin-1 Expression-To explore this hypothesis, we first examined FAK ϩ/ϩ and FAK Ϫ/Ϫ mouse embryonic fibroblasts (MEFs) for caveolin-1 expression. As shown in Fig. 4A, caveolin-1 expression is barely detectable in the FAK Ϫ/Ϫ MEFs, a notable contrast to the abundance of caveolin-1 in the FAK ϩ/ϩ MEFs. To further confirm the positive effect of FAK on caveolin-1 expression, FAK Ϫ/Ϫ MEFs were transfected with a plasmid encoding either FAK or superFAK (a catalytically active FAK mutant, Ref. 61), or mocktransfected. Fig. 4B shows that caveolin-1 expression is increased almost 4-fold in the superFAK-transfected cells compared with the mock-transfected control. Expression of FAK also increased caveolin-1 expression by 1.4-fold over mocktreated; however, caveolin-1 expression was not completely restored to levels seen in FAK ϩ/ϩ MEFs. This ineffective restoration of caveolin-1 expression was likely caused by transfection efficiency (in our hands, the nucleofection efficiency for MEFs was ϳ65% based on examination of cells nucleofected with constructs encoding GFP). Furthermore, this result suggests that the phosphorylation of FAK could specifically be important in the regulation of caveolin-1 expression. The recently characterized focal adhesion kinase inhibitor PF-573,228 (PF-228) has been shown to inhibit FAK catalytic activity via ATP-binding pocket interactions (39). For this reason, we decided to employ PF-228 as a means to examine the influence of FAK activity on  caveolin-1 expression. FAK ϩ/ϩ MEFs were treated with either DMSO only or increasing concentrations of PF-228 for 24 h, followed by Western blot analysis for caveolin-1, pFAK (Y397), and FAK. As shown in Fig. 4C, treatment with PF-228 decreases FAK phosphorylation in a dose-dependent manner, and when using 5 M PF-228, both phosphorylation of FAK and expression of caveolin-1 are reduced by almost 50% compared with levels seen in controls. Additionally, we wanted to determine the impact of FAK activity on caveolin-1 expression during EMT. Intriguingly, stimulated NT2/D1 cells treated with PF-228 also demonstrated decreased caveolin-1 expression compared with control cells during EMT (Fig. 4D). NT2/D1 cells required slightly higher doses of PF-228 to effectively decrease FAK phosphorylation (below 50% of control) compared with that of MEFs. However, this varying effective dose of PF-228 seen between different cell lines was also reported in the initial characterization of the inhibitor (39). These results suggest that FAK phosphorylation can greatly influence caveolin-1 expression and correspondingly has the potential to alter caveolin-1 expression during EMT.
Src-induced Increases in Caveolin-1 Expression Are FAK-dependent-Once FAK is autophosphorylated on tyrosine 397, other proteins, such as Src, can bind and phosphorylate other FAK sites, including tyrosines 576, 577, 861, and 925 (33,62), hence enabling FAK to impact cell spreading and migration (63). Therefore, we wanted to explore the possibility that Src could contribute to the regulation of caveolin-1 expression by increasing FAK phosphorylation. SYF Ϫ/Ϫ MEFs transfected with the Src construct cSrc-Y527F (a constitutively active Src) showed increases in both FAK phosphorylation (at both Tyr-397 and Tyr-861) and caveolin-1 expression compared with mocktransfected control cells (Fig. 5A).
To determine if Src-induced increases in caveolin expression could be seen in the absence of FAK, FAK Ϫ/Ϫ MEFs were transfected with cSrc-Y527F and caveolin-1 expression was compared with that of FAK ϩ/ϩ MEFs and mock control (Fig. 5B). No increase in caveolin-1 expression could be detected in the absence of FAK, suggesting that FAK is required for Src-induced increases in caveolin-1 expression. Alternatively, expression of super-FAK in SYF Ϫ/Ϫ MEFs still leads to increased expression of caveolin-1 (Fig. 5C), suggesting that Src is not critical for FAK-induced caveolin-1 expression. We also tested the ability of the Src inhibitors PP2 and SU6656 (40,41) to impact caveolin-1 expression during EMT. PP2, given during day 2 post-stimulation, almost completely blocked day 3 NT2/D1 caveolin-1 expression (Fig. 5D). Although not as dramatic, treatment with SU6656 could also decrease day 3 caveolin-1 expression in the NT2/D1 cells (Fig. 5E).
Finally, to probe for alterations in caveolin-1 mRNA in response to FAK, we isolated RNA from FAK ϩ/ϩ cells, FAK Ϫ/Ϫ cells, and FAK Ϫ/Ϫ cells nucleofected with superFAK or Src-Y527F constructs. Genetic depletion of FAK led to a more than 2-fold decrease in caveolin-1 mRNA compared with wild-type MEFs (Fig. 5F). Expression of superFAK significantly increased caveolin-1 mRNA levels by 4-fold over mock-transfected FAK Ϫ/Ϫ cells, while Src-Y527F did not show this effect (Fig.  5G). These results show that the presence and phosphorylation of FAK are critical to both caveolin-1 mRNA and protein expression (see Fig. 4).
Down-regulation of Caveolin-1 Affects Tumor Cell Adhesion and Migration-The data obtained from these EMT models made us speculate that caveolin-1 may be a component of a more motile cell phenotype. In the past, we have speculated that caveolin-1 may play a role in focal adhesion dynamics (18), and it has been reported that caveolin contributes to the membrane order of focal adhesions (64). Given the fact that mesenchymal cells are highly motile in nature (26), we hypothesized that  caveolin-1 expression may impact total cell spreading area and adhesion in cancer cells. To test this hypothesis, we treated the human lung cancer cell line H1703 with caveolin-1 or control siRNA and examined the effect on cell attachment, spreading, and focal adhesion numbers. The H1703 cell line was chosen for this analysis because these cells express caveolin-1, are independent of one another (mesenchymal morphology), adhere/ spread nicely, and can be readily transfected. Also, as in the other cell lines we examined, caveolin-1 expression in the H1703 cell line can be regulated by FAK and Src (Fig. 6, A and B); and treatment of the NT2/D1, NMuMG, or H1703 cells with the PF-228 FAK inhibitor also led to a significant reduction in motility compared with controls (data not shown), an effect that has been previously reported with this inhibitor in other cell lines (39). Treatment with caveolin-1 siRNA effectively suppressed caveolin-1 expression by more than 80% in the H1703 cells (Fig. 6D). Control or caveolin-1 siRNA-treated H1703 cells were then subjected to an ECIS attachment assay. With this technique, as cells contact the gold electrodes, the reported resistance to the flow of electrons is increased. Fig. 6C shows that H1703 cells treated with caveolin-1 siRNA have a much greater reported resistance over time compared with the control siRNA-treated cells. The initial dip in the resistance in both groups is reflective of the changes in temperature and electron flow upon addition of the cells, as reported previously (65). These results suggest that decreases in caveolin-1 expression can increase resistance upon attachment, reflecting some alteration in H1703 cell attachment (focal adhesions) or cell spreading. Similar results were also obtained when subjecting low passage caveolin-1 WT and KO MEFs to ECIS attachment analysis, with the caveolin-1 KO MEFs displaying a large increase in resistance over that of WT (data not shown).
Because the ECIS analysis cannot pinpoint the exact cause of the resistance shift, we decided to manually examine H1703 cell spreading and focal adhesion numbers. Again, H1703 cells were treated with either control or caveolin-1 siRNA. Cells were then seeded onto fibronectin-coated coverslips, and after 3 h, they were fixed, immunofluorescently labeled for actin and paxillin, and imaged via confocal microscopy. The resulting cell images were then manually quantified for both total cell area and focal adhesion number using the software described under "Experimental Procedures." As shown in Fig. 6, E and F, downregulation of caveolin-1 increases both total cell area and focal adhesion number compared with the control. To determine if  or control (DMSO only) and blotted for pFAK (Y397) and caveolin-1. C, control or caveolin-1 siRNA-treated H1703 cells were examined for alterations in attachment resistance using an ECIS attachment assay. Targeted knockdown of caveolin-1 using caveolin-1 siRNA (black curve) lead to a significant increase in resistance compared with that of the control (gray curve) (p Ͻ 0.01). D, some of the control or siRNA-treated cells from C were lysed and immunoblotted with antibody for caveolin-1. E, control or caveolin-1 siRNA-treated H1703 cells were immunofluorescently labeled for actin and paxillin, randomly imaged, and relative cell spreading areas were quantified using Image J software. Targeted knockdown of caveolin-1 increases the total cell spreading area compared with cells treated with control siRNA. *, p Ͻ 0.05. F, total number of focal adhesions in either control or caveolin-1-treated cells was quantified using the same images in E. Note that the knockdown of caveolin-1 leads to an increase in the total number of focal adhesions in the cell. *, p Ͻ 0.05. G, to determine whether adhesion numbers were affected by cell spreading area, cells were arbitrarily subdivided into small, medium, and large groups. Bars represent the relative fold increase in focal adhesions in caveolin-1 siRNA-treated cells over that of control siRNA-treated cells (normalized to a value of 1) for each group. *, p Ͻ 0.05. H, control or caveolin-1 siRNA-treated cells were subjected to Boyden chamber migration assays. Knockdown of caveolin-1 significantly decreases cell migration compared with that of controls (*, p ϭ 0.001). GAPDH blots were included as equal protein loading controls in A, B, and D.
the effect of caveolin-1 on focal adhesion number was caused by its effect on cell spreading area (i.e. larger cells could potentially have more focal adhesions than smaller cells), control or caveolin-1 siRNA-treated cells were size-matched into three arbitrary groups (small, medium, and large), and the number of focal adhesions in each group was compared. Fig. 6G shows that cav-1 knockdown increases the number of total cell adhesions, with a maximum effect seen in the cells with the largest spreading area. Finally, caveolin-1 can be phosphorylated on tyrosine 14, and, as mentioned above, was shown to contribute to focal adhesion membrane order (64). By using wild-type and caveolin-1 KO MEFs that were transfected with plasmid encoding either wild-type caveolin-1 or Y14F caveolin-1, we examined whether caveolin-1 phosphorylation could be the factor responsible for these changes in cell adhesion and spreading.
Our results showed that caveolin-1 phosphorylation could not fully explain the noted changes in cell adhesion (data not shown). The results suggest that even in cells of equivalent relative area, there are more focal adhesions in the caveolin-1 siRNA-treated cells, thus making changes in caveolin-1 expression an important factor contributing to the extent to which tumor cells attach and spread.
To test whether down-regulation of caveolin-1 would affect migration of the lung cancer cells, H1703 cells were transfected either with control or caveolin-1 siRNA and subjected to a Boyden chamber motility assay. As shown in Fig. 6H, down-regulation of caveolin-1 causes up to 40% reduction in migration compared with that of control siRNA-treated cells. This observation is consistent with our previous results demonstrating that both knockdown of caveolin-1 using caveolin-1 siRNA and genetic depletion of the caveolin-1 protein results in impeded cell migration (18,21).

DISCUSSION
Many features of tumor progression, including increased mitogenic signaling, insensitivity to anti-growth signals, evasion of replicative senescence, resistance to apoptosis, angiogenesis, and increases in invasiveness/cell motility are influenced by caveolin-1 (66,67). It is speculated that in the early stages of cancer, caveolin-1 is down-regulated to reduce its inhibitory effects on cell growth, whereas its expression is elevated as the cancer advances to promote tumor progression (68). Recent reports have described the positive correlation of caveolin-1 expression with tumor grade, stage, and metastasis in some cancer types (68). However, the mechanism that regulates caveolin-1 expression during tumor progression remains unclear.
By employing both the NT2/D1 and NMuMG EMT cell models, we were able to examine alterations in caveolin-1 expression during epithelial to mesenchymal transition. These two EMT cell models also allowed us to examine the context in which changes in caveolin-1 expression occurred, and we found that in both models, caveolin-1 is up-regulated following induction of EMT. Notably, cells elongate and start to scatter prior to the up-regulation of caveolin-1, suggesting that changes in caveolin-1 expression may be a product of EMT. Prior to the up-regulation in caveolin-1 expression, we saw substantial increases in the phosphorylation of FAK (at Tyr-397 and Tyr-861) and Src (at Tyr-418) in both cell models. Upon inhibiting FAK (using the FAK inhibitor PF-228), we saw a dramatic suppression of caveolin-1 expression, strongly suggesting a critical role of FAK in mediating caveolin-1 expression during EMT. Src can also contribute to this regulation, but as we report here, only in the presence of FAK. Thus in this study, we have identified a novel signaling event that can regulate alterations in caveolin-1 expression during EMT. Furthermore, FAK has been shown to be up-regulated in a number of invasive and metastatic cancers (69), and many reports implicate the FAK/ Src signaling pathway in the progression through the epithelial to mesenchymal transition and generation of a more highly invasive phenotype in many cancer types (26 -29). Our present results therefore provide a mechanistic explanation for elevated caveolin-1 expression in locally advanced and metastatic cancers. The FAK-dependent up-regulation of caveolin-1 expression may in turn positively affect integrin-mediated FAK signaling as previously shown by the Chapman and Minna laboratories (8,60).
Interestingly, a recent report showed that upon EGF treatment of A431 human carcinoma cells, FAK phosphorylation was greatly suppressed and remained hypophosphorylated for over 24 h (70). Using the same EGF-treated, FAK-hypophosphorylated A431 cells, Hunter and co-workers (71) later showed that EGF treatment caused down-regulation of caveolin-1 expression. These results are consistent with our current study demonstrating the importance of FAK phosphorylation in the regulation of caveolin-1 expression. Thus, the role of FAK phosphorylation and subsequent caveolin-1 expression may depend on the stimuli (EGF versus TGF-␤1, etc.) and/or cell lines used when inducing EMT, as these peptides can regulate EMT (and other cellular functions) through different pathways and behave differently depending on the cell type (72). This view is supported by a recent study showing that the activation of FAK is critical for promoting E-cadherin delocalization and transcriptional up-regulation of mesenchymal markers specifically during TGF␤-mediated EMT (73).
In the past, we reported that caveolin-1 is also an important regulator of cell polarity and directional movement (18,21), and recently, it was shown that caveolin-1 exerts this effect on cell movement by impacting the activity of Rho GTPases (74). Our current data show that increases in caveolin-1 expression follow classically described cellular changes associated with EMT (including changes in cell morphology and expression of the cadherins and fibronectin). Therefore, this up-regulation of caveolin-1 expression is seemingly a product of EMT, and caveolin-1 can be considered a component of the mesenchymal phenotype in the cell models we studied. We report an increase in cell attachment resistance in caveolin-1 siRNA-treated lung cancer cells during ECIS analysis compared with that of control. It is compelling to speculate on the potential of caveolin-1 to aid in mesenchymal cell migration and invasion by regulating aspects of cell adhesion. This effect of caveolin-1 on cell attachment is also thought-provoking when considering FAK Ϫ/Ϫ MEFs. In the past, we and others (18,21,34) have described that the loss of caveolin-1 hampers effective cell motility. Classic studies on FAK Ϫ/Ϫ cells have reported that without FAK, cells display an increased number of focal adhesions and migrate less effectively (75). In our current study, we have shown that FAK Ϫ/Ϫ MEFs barely express any detectable caveolin-1 and that knockdown of caveolin-1 leads to increases in cell adhesion resistance (more focal adhesions and larger total cell spreading area). It is therefore possible that the decreased caveolin-1 expression in FAK Ϫ/Ϫ MEFs may contribute to the characteristic, yet largely unexplained, phenotype of these cells.
These data highlight the fact that caveolin-1 can be up-regulated during EMT, and its expression is tightly regulated by FAK. Furthermore, the FAK-mediated regulation of caveolin-1 expression we have demonstrated provides a novel pathway by which FAK can mediate cell adhesion and motility.