Disabled-2 (Dab2) Is Required for Transforming Growth Factor β-induced Epithelial to Mesenchymal Transition (EMT)*

Transforming growth factor β (TGFβ) induces an epithelial to mesenchymal transition (EMT) during both physiological and pathological processes; however, the mechanism underlying this transition is not fully elucidated. Here, we have demonstrated that TGFβ induces the expression of the adaptor molecule disabled-2 (Dab2) concomitant with the promotion of EMT. We show that TGFβ induces a transient accumulation of Dab2 to the membrane and increases Dab2 binding to β1 integrin. Furthermore, small interfering RNA (siRNA)-mediated silencing of Dab2 expression in mouse mammary gland epithelial cells results in inhibition of integrin activation, shown by a decrease of both TGFβ-induced focal adhesion kinase phosphorylation and cellular adherence, leading to apoptosis and inhibition of EMT. Forced re-expression of human Dab2, not targeted by the mouse siRNA sequence, rescues cells from apoptosis and restores TGFβ-mediated integrin activation and EMT. These results are confirmed in the F9 teratocarcinoma cell line, a model for retinoic acid-induced visceral endoderm differentiation in which we demonstrate that ablation of retinoic acid-induced Dab2 expression levels, by stable siRNA silencing of Dab2, blocks visceral endoderm differentiation. Our findings indicate that Dab2 plays an important regulatory role during cellular differentiation and that induction of differentiation in the absence of Dab2 expression commits the cell to apoptosis.

Transforming growth factor ␤ (TGF␤) induces an epithelial to mesenchymal transition (EMT) during both physiological and pathological processes; however, the mechanism underlying this transition is not fully elucidated. Here, we have demonstrated that TGF␤ induces the expression of the adaptor molecule disabled-2 (Dab2) concomitant with the promotion of EMT. We show that TGF␤ induces a transient accumulation of Dab2 to the membrane and increases Dab2 binding to ␤1 integrin. Furthermore, small interfering RNA (siRNA)mediated silencing of Dab2 expression in mouse mammary gland epithelial cells results in inhibition of integrin activation, shown by a decrease of both TGF␤induced focal adhesion kinase phosphorylation and cellular adherence, leading to apoptosis and inhibition of EMT. Forced re-expression of human Dab2, not targeted by the mouse siRNA sequence, rescues cells from apoptosis and restores TGF␤-mediated integrin activation and EMT. These results are confirmed in the F9 teratocarcinoma cell line, a model for retinoic acid-induced visceral endoderm differentiation in which we demonstrate that ablation of retinoic acid-induced Dab2 expression levels, by stable siRNA silencing of Dab2, blocks visceral endoderm differentiation. Our findings indicate that Dab2 plays an important regulatory role during cellular differentiation and that induction of differentiation in the absence of Dab2 expression commits the cell to apoptosis.
Epithelial to mesenchymal transition (EMT) 1 is a fundamental multistep process that occurs during both physiological and pathological states (1,2). By allowing cells to detach from the epithelial tissue where they originate and to migrate, such transitions are necessary for proper embryonic development. However, they also provide a way for epithelium-derived tumors from a benign stage to become invasive and metastasize throughout the body. It has been demonstrated that there is commonality in the signaling pathways regulating EMT during both embryonic development and tumor progression, suggesting that similar molecular mechanisms underlie these pro-cesses and raising the possibility that tumor metastasis might simply be considered as a reactivation of some aspects of the embryonic program of EMT.
The process of EMT requires loss of cell-cell interaction and acquisition of fibroblastic morphology with increased expression of mesenchymal markers, such as N-cadherin (3). Among growth factors that can promote this structural conversion, transforming growth factor ␤ (TGF␤) has been well characterized as an important inducer of EMT during development as well as during cancerogenesis (4). Indeed, although TGF␤ is considered a tumor suppressor during the first stage of tumorigenesis, principally through its ability to cause growth arrest and apoptosis in many non-transformed epithelial cell types, numerous reports show that TGF␤ can also promote tumor progression, particularly through its ability to promote EMT (4,5). Moreover, it has been shown that blocking TGF␤ signaling in transgenic mice that develop multifocal metastatic mammary tumors reduces mammary tumor intravasation and lung metastasis and increases mammary tumor cell apoptosis (6).
TGF␤ exerts its biological effects by inducing the formation of a heteromeric complex composed of type I and type II serine/ threonine kinases receptors. The activated receptor complex, in turn, phosphorylates and activates the receptor-regulated Smads, Smad2 and Smad3. Once activated these Smads complex with the common mediator, Smad4, to transmit the signal to the nucleus, where the Smads regulate transcription of target genes through their interaction with a wide variety of transcriptional regulators (4). In addition, TGF␤ has been shown to activate diverse parallel downstream pathways, for example, extracellular signal-regulated kinase, c-Jun NH 2 -terminal kinase, or p38 kinase (4). Intensive investigation into the molecular mechanism of TGF␤-induced EMT in mammary epithelium has led to the identification of different mediators of this process including the Smads, phosphatidylinositol 3-kinase, RhoA, p38, and ␤1 integrin, among others (3,(7)(8)(9)(10)(11). However, it is clear from the results that the process of EMT is quite complex, involving not only structural and cytoskeletal reorganization but also extensive genetic changes. Thus, much remains to be determined regarding the molecular events mediating a TGF␤-induced EMT.
Disabled-2 (Dab2) was first identified as DOC-2, for differentially expressed in ovarian carcinoma (12), and subsequently as a protein whose phosphorylation is stimulated by CSF-1 (13). Dab2 has been shown to be involved in regulation of cytoskeleton-based functions by regulating cell positioning (14), inducing macrophage adhesion (15), or regulating protein trafficking by interacting with different proteins implicated in endocytosis (16 -22). An important role for Dab2 in regulating the homeostasis of epithelial differentiation has been suggested (14), a role recently confirmed by several genetic studies demonstrating a function for Dab2 in endodermal cell formation, differentiation, and organization (16,23). Dab2 also par-ticipates in diverse signal transduction pathways. Indeed, it has been reported that Dab2 interferes with mitogenic growth factor signal transduction (24 -27) as well as with Wnt signaling (28). Dab2 can also be a positive mediator of TGF␤ signaling by bridging the heteromeric TGF␤ receptor complex to the Smad proteins (29). However, the biological role of Dab2 in TGF␤ signaling has not yet been fully elucidated.
Here, we report for the first time a role for Dab2 during TGF␤-induced EMT in normal murine mammary gland (NMuMG) cells. NMuMG epithelial cells were chosen because they are a well established model system for studying TGF␤induced EMT. We observed that during TGF␤-induced EMT in these cells, the expression of Dab2 is increased and Dab2 transiently accumulates in the membrane fraction and binds to ␤1 integrin. Furthermore, stable siRNA-mediated silencing of Dab2 in NMuMG cells results in inhibition of integrin activation, shown by a decrease of both TGF␤-induced FAK phosphorylation and cellular adherence, leading to apoptosis and inhibition of EMT. Thus, we propose a model in which Dab2 acts as a critical switch during TGF␤-mediated EMT, determining whether the cell proceeds toward an EMT rather than apoptosis.

EXPERIMENTAL PROCEDURES
Cell Culture and Reagents-Non-transformed NMuMG epithelial cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 10 g/ml insulin, and antibiotics (penicillin and streptomycin). F9 mouse embryonic carcinoma cells were from American Type Culture Collection (ATCC) and cultured on gelatin-coated tissue culture plates in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. In some experiments, NMuMG cells were plated onto culture dishes precoated overnight at 4°C with 10 g/ml fibronectin (Sigma). TGF␤2 was a generous gift from Genzyme Inc. (Cambridge, MA) and was used at a final concentration of 5 ng/ml. All-trans-retinoic acid (RA) was purchased from Sigma and was used at a final concentration of 100 nM. Antibodies to Dab2, ␤1 integrin, and FAK were purchased from BD Transduction Laboratories. The antibody to phospho-397 FAK was purchased from BIOSOURCE. The antibody to phosphotyrosine 4G10 (PY) was purchased from Upstate Biotechnology. The polyclonal antibody to Bim was purchased from BD Biosciences. Antibodies to Hsp90 and to N-cadherin were purchased from Santa Cruz Biotechnology.
Cell Transfection-Stable transfected cell lines were generated by co-transfection of NMuMG cells with pSUPER-siDab2 or pSUPER (control vector), respectively (see above), and a plasmid expressing puromycin resistance, pPuro (Clontech). Cells were selected in medium containing 2 g/ml puromycin and maintained as a population. NMuMG-siDab2humanDab2 cells were obtained by co-transfection of the human form of Dab2, WTDab2, with a plasmid expressing neomycin resistance, pSV2Neo. Cells were selected in medium containing G418 (1 mg/ml) and maintained as a population. A control population of cells was transfected with pSV2Neo alone. F9-siDab2 cells were generated by co-transfection of the pSUPER-siDab2 and pPuro plasmids. Cells were selected in medium containing 2 g/ml puromycin and maintained as a population. Transfections were performed using Lipofectamine reagent (Invitrogen) according to the manufacturer's protocols.
Northern Blot Analysis-RNA was prepared using the RNeasy kit from Qiagen. 20 g of total RNA were denatured, size fractionated on 1.2% formaldehyde-agarose gel, transferred to Nytran membrane (Schleicher & Schuell), and immobilized by UV cross-linking. Equivalent loading of intact RNA was assured by visualization of ethidium bromide-stained 28 and 18 S ribosomal RNA bands. pcDNA3-Dab2 vector (29) was linearized with KpnI, purified using a Wizard Prep column (Promega), and used as a template for riboprobe synthesis. The Dab2 riboprobe was then generated by in vitro transcription by using SP6 polymerase and the RiboMAX kit (Promega). Blot was then hybridized to the [ 32 P]-labeled-riboprobe overnight at 65°C in northernMax Hyb buffer (Ambion), washed 3ϫ 20 min with 0.5 SSC/0.1 SDS at 65°C, 1ϫ 20 min with 0.2 SSC/0.1 SDS at 65°C, and then autoradiographed. Subcellular Fractionation, Western Blot, and Immunoprecipitation Analysis-Cytosolic and membrane fractionation was performed as described (28). Immunoprecipitation was performed on 500 g of the membrane lysates with ␤1 integrin antibody or normal mouse IgG overnight, followed by adsorption to Sepharose-protein G (Amersham Biosciences) for 2 h. The beads were washed five times in lysis buffer, and samples were analyzed by SDS-polyacrylamide gel electrophoresis and immunoblotting. For Western blot analysis, cells were lysed at 4°C in lysis buffer TNMG (20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 5 mM MgCl 2, 10% glycerol, 0.5% Nonidet P-40, 1 mM Na3VO4, 25 mM ␤-glycerophosphate, and EDTA-free protease inhibitor mixture Complete from Roche Diagnostic) followed by sonication and then centrifugation at 10,000 rpm for 10 min at 4°C. The lysates were fractionated by SDS-polyacrylamide gel electrophoresis, transferred to Immobilon-P membrane (Millipore), and probed with the indicated primary antibody. Specific binding of antibodies was then detected with peroxidase-conjugated secondary antibodies and visualized by an enhanced chemiluminescence detection system (Amersham Biosciences).
Adhesion Assay-Cells were trypsinized, resuspended in 1 ml of medium at 5 ϫ 10 6 cells/ml. 100 l of this cell suspension was added per well to a 96-well tissue culture plate and allowed to adhere at 37°C for 1 h. After incubation, non-adherent cells were removed by washing. The color reaction substrate (MTT; 100 l) was added to each well, and the plate was incubated for 3 h at 37°C. The supernatant was discarded, and the formazan salt was dissolved in 100 l of isopropanol. The optical density was measured at 570 nm using a microplate reader.
Apoptosis Assays-For DNA fragmentation assay, cells were collected by centrifugation, washed, and lysed in HL buffer (10 mM Tris, pH 8.0, 1 mM EDTA, and 0.2% Triton X-100). After centrifugation for 15 min at room temperature at 14,000 rpm, the supernatants were collected. The supernatants were extracted with an equal volume of phenol and then phenol:chloroform:isoamyl alcohol (25:24:1) and precipitated at Ϫ20°C with an equal volume of isopropanol and 0.1 volume of 5 M NaCl. After centrifugation at 14,000 rpm for 20 min, the precipitated DNA was washed with 70% ethanol, resuspended in Tris/EDTA, pH 8.0, containing DNase-free RNase A, and incubated at 37°C for 30 min. The electrophoresis was performed in 1ϫ TBE (Tris borate-EDTA) buffer on 2% agarose gel containing ethidium bromide.
Apoptosis was also measured using the Cell Death Detection ELISA plus kit (Roche Diagnostics) following the kit protocol. Briefly, cells were collected and resuspended in 200 l of kit lysis buffer. The lysate was centrifuged, and 20 l of the supernatant, containing the cytoplasmic fraction, was applied to streptavidin-coated microplates. A mix of biotinylated anti-DNA antibody and peroxidase-conjugated antihistone antibody was added to the wells, and binding was allowed for 3 h. The plates were washed thoroughly followed by the addition of 2,2Ј-azino-di(3-ethylbenzthiazolin-sulfonate) (ABTS) substrate. Color development was monitored spectrophotometrically at 405 nm. Fig. 1A by microscopic examination, treatment of NMuMG cells with TGF␤ induces a drastic alteration in cell shape. Whereas untreated cells display the typical cuboidal cobblestone-like characteristic of epithelial cells, a 24-h treatment with TGF␤ induces an elongated and spindle-shaped phenotype characteristic of fibroblasts ( Fig. 1A) (31). Confirming this morphologic analysis of a TGF␤-induced EMT, the biochemical results presented in Fig. 1B demonstrate that TGF␤ treatment induces a progressive increase of Ncadherin, a fibroblastic marker (Fig. 1B) (3). Because we have previously identified Dab2 as a TGF␤ signaling molecule (29), we wished to determine whether its level of expression varies during TGF␤-induced EMT. As can be seen in Fig. 1C by immunoblot analysis of NMuMG cellular lysate, Dab2 levels are significantly up-regulated in response to TGF␤. Analysis of the mRNA levels of Dab2 by Northern blot shows a TGF␤induced increase in Dab2 mRNA in accordance with the corresponding increase in Dab2 protein levels (Fig. 1D). This result was also confirmed by reverse transcription PCR (data not shown). Taken together, these data show a significant upregulation of Dab2 during TGF␤-induced EMT, suggesting a potential role of Dab2 during this process.

Disabled-2 (Dab2) Expression Is Increased during TGF␤induced EMT-As shown in
Dab2 Is Necessary during TGF␤-induced EMT-To deter-mine the significance of the increase in Dab2 levels in cells undergoing an EMT, we decided to suppress Dab2 activity. Because no dominant negative form of Dab2 exists, we chose to stably down-regulate Dab2 expression by using a mammalian expression vector that directs intracellular synthesis of small interfering RNA (30). NMuMG cells that stably express Dab2 siRNA (NMuMG-siDab2), but not the control cells expressing the empty vector (NMuMG-siControl), exhibited a reduction in the basal expression levels of Dab2 as well as in the induction of Dab2 during the differentiation process induced by TGF␤ ( Fig. 2A). We then monitored TGF␤-induced EMT in these cells. As shown in Fig. 2B, we observed that 24 h of TGF␤ treatment in NMuMG-siDab2 cells does not result in EMT, whereas under the same conditions TGF␤ can induce a mesenchymal transition in control cells. We confirmed that the silencing of Dab2 expression blocks TGF␤-mediated fibroblastic conversion by demonstrating that there was a concomitant loss in the up-regulation of N-cadherin levels in response to TGF␤ in NMuMG-siDab2 cells (Fig. 2C).
Interestingly, in addition to retaining their epithelial morphology in response to TGF␤, we observed that some NMuMG-siDab2 cells displayed apoptosis-like morphological changes.
These cells rounded-up and became detached following TGF␤ treatment (Fig. 2B). The presence of DNA fragmentation visualized by oligonucleosomal DNA ladder formation on agarose gel (Fig. 2D) or quantified by ELISA (Fig. 2E) confirmed that, indeed, cell death through apoptosis was occurring upon treatment of the NMuMG-siDab2 cells with TGF␤. We (52) and others (32) have recently reported that the BH3 domain-only pro-apoptotic protein Bim is induced during apoptosis. We, therefore, examined whether TGF␤-mediated apoptosis in NMuMG-siDab2 cells is associated with changes in Bim expression. Indeed, we found that TGF␤ treatment provokes a marked increase in Bim expression levels in NMuMG-siDab2 cells, in contrast to NMuMG-siControl cells (Fig. 2F). Taken together, these results suggest that Dab2 up-regulation is necessary for a proper TGF␤-induced EMT, apparently required to prevent apoptosis.
To determine whether the role of Dab2 in EMT and differentiation is specific to TGF␤ and/or the NMuMG cells, we investigated Dab2 levels in response to other differentiation factors and in another differentiation model. Previous reports have shown that Dab2 is induced in response to RA-induced differentiation in the F9 teratocarcinoma cell line in which RA induces embryonic stem cell-like epithelial cells to differentiate into visceral endoderm (24). First, we investigated whether RA could induce differentiation in NMuMG cells and whether there was a concomitant increase in the level of Dab2 as demonstrated above with TGF␤ stimulation. The results demonstrate that RA (100 nM) significantly increases Dab2 protein levels in a time-dependent fashion (Fig. 3A) that is accompanied with an increase in the differentiation marker N-cadherin (Fig. 3A) and morphological alteration characteristic of EMT (data not shown). We next chose to analyze the role of Dab2 in the F9 teratocarcinoma cell line. As shown in Fig. 3B, and similar to that previously reported (24), RA (100 nM) dramatically induces Dab2 expression in F9 cells in a time-dependent fashion with maximal effects observed following 72 h. RA treatment of F9 cells has also been shown to be accompanied by growth suppression (24) and morphological alteration phenotypic of cells differentiating into visceral endoderm (Fig. 3D) and characterized by the endoderm differentiation marker GATA-4 ( Fig. 3C). In F9 cells, in which we stably ablated Dab2 expression levels with the Dab2 siRNA vector (F9-siDab2), we did not observe RA induction of Dab2 protein levels (Fig. 3B), and the cells failed to differentiate into visceral endoderm (Fig.  3D) and express the endoderm marker GATA-4 (Fig. 3C). As described above for NMuMG and TGF␤, F9 cells stably expressing the siRNA to Dab2 not only failed to differentiate in response to the differentiating agent RA but instead proceeded to undergo apoptosis as analyzed by induction of the pro-apoptotic protein Bim (Fig. 3C). These results demonstrate that Dab2 protein levels mediate an important regulatory step in the differentiation process and that inhibition in the induction of Dab2 levels in the presence of a differentiating agent (i.e. TGF␤ or RA) induces apoptosis instead of differentiation.
Re-expression of Dab2 Rescues TGF␤-induced Apoptosis and Restores TGF␤-induced EMT-To confirm that the effects observed in NMuMG-siDab2 cells are specific to a down-regulation of Dab2 expression, we attempted to restore Dab2 expression in these cells. Because the siRNA sequence that we used targets only mouse Dab2, we stably transfected a cDNA encoding the human form of Dab2 to ectopically re-express Dab2 protein. As shown in Fig. 4A, the human form of Dab2 is not targeted by stable expression of mouse Dab2 siRNA and is stably expressed in the cells. Restoration of Dab2 expression in NMuMG-siDab2 cells reverses the apoptotic effects of TGF␤ and, most importantly, restores TGF␤-induced EMT as de- were treated with TGF␤ (5 ng/ml) for the times indicated, and total cellular lysates were analyzed by Western blot using ␣-N-cadherin antibody. Equivalent protein loading was verified by ␣-Hsp90 immunoblotting. C, Dab2 expression during TGF␤-induced EMT was analyzed by Western blot analysis using ␣-Dab2 antibody. D, Dab2 expression was also evaluated by Northern blot using 20 g of total RNA isolated from cells treated with TGF␤ for the times indicated. Visualization of 18 and 28 S ribosomal RNA by ethidium bromide staining of the gel was used as a loading control.
Integrins Prevent Apoptosis during TGF␤-induced EMT-We next explored the possible mechanism whereby the absence or down-regulation of Dab2 mediates apoptosis in response to TGF␤. Rosenbauer et al. (15) have previously reported that Dab2 can augment bone marrow-derived macrophage adhesion. We, therefore, compared the adhesive capability of NMuMG-siDab2 cells to NMuMG-siControl cells. As seen in Fig. 5A, 2-4 h after seeding the cells in tissue culture plates most of the NMuMG-siDab2 cells were still "rounded up" compared with the control cells, which had begun to exhibit an adhered and "spread-out" shape. Quantitatively, NMuMG cells that lack Dab2 were ϳ50% less adherent than control NMuMG cells (Fig. 5B). We, therefore, hypothesized that the decreased adherence and spreading observed in NMuMG-siDab2 cells could play a role in the ability of TGF␤ to induce apoptosis in these cells.
Previous results in the NMuMG/TGF␤ EMT model have demonstrated that integrin ␤1 is required for TGF␤-mediated EMT when cells were grown on collagen-coated dishes (8). We next examined whether the loss of adhesion in cells with the downregulated Dab2 expression reported above could be due to lack of integrin engagement. Confirming the work of Bhowmick et al. (8), the results demonstrate that the RGD peptide, which blocks integrin ␤1, when added to NMuMG cells is able to block TGF␤-induced EMT (Fig. 5C) and TGF␤ induction of N-cadherin (Fig.  5D). The results demonstrate not only that RGD blocks TGF␤induced EMT but that in the presence of RGD and lack of integrin ␤1 engagement TGF␤ addition induces increased apoptosis, as evidenced morphologically in Fig. 5C.
TGF␤ Transiently Induces Accumulation of Dab2 to the Membrane Fraction-To gain further insight into the link between Dab2 and integrins, we first investigated the cellular distribution of Dab2. As shown in Fig. 6A, TGF␤ induces the progressive accumulation of Dab2 to the membrane fraction during the first 12 h of treatment, which is followed by its re-distribution into the cytosolic fraction at 24 h. Because Dab2 and integrins both share, transiently, the same membrane subcellular compartment, we wished to determine whether in response to TGF␤ Dab2 associates with ␤1 integrin. Indeed, ␤1 integrin has already been shown to be expressed in NMuMG cells and its expression level increased by TGF␤ (8, 33). The results of Fig. 6B (lower panel) confirm that TGF␤ induces an increased expression of ␤1 integrin but, most importantly,

FIG. 4. Re-expression of Dab2 rescues TGF␤-induced EMT in
NMuMG-siDab2 cells. A, NMuMG-siDab2 cells were stably transfected with the human form of Dab2 (NMuMG-siDab2-humanDab2) cells. Re-expression of Dab2 in these cells was determined by Western blot analysis using ␣-Dab2 antibody. B-E, NMuMG-siDab2 and NMuMG-siDab2-humanDab2 cells were treated in the absence (Control) or presence (ϩTGF␤) of TGF␤ for 24 h. B, cells were analyzed for morphological changes by microscopy. C, N-cadherin expression was analyzed by Western blot analysis using ␣-N-cadherin antibody. Equivalent protein loading was verified by ␣-Hsp90 immunoblotting. D-E, re-expression of Dab2 rescues TGF␤-induced apoptosis. D, agarose gel electrophoresis reveals no DNA ladder in NMuMG-siDab2-humanDab2 cells treated for 24 h with TGF␤. E, Bim expression was analyzed by Western blot analysis using ␣-Bim antibody.
demonstrate that under endogenous conditions Dab2 can interact transiently with ␤1 integrin and that this association is TGF␤-dependent (Fig. 6B, upper panel). These results suggest that TGF␤-mediated EMT requires the interaction of Dab2 and ␤1 integrin and that Dab2 functions through ␤1 integrin.
Dab2 Expression Is Necessary for Integrin Activation during TGF␤-induced EMT-Many of the signals resulting from integrin activation employ the focal adhesion-associated tyrosine kinase FAK (34). To further ascertain that TGF␤ induces integrin activation through Dab2, we assayed FAK activation during TGF␤-mediated fibroblastic conversion. The data, presented in Fig. 6C, indicate that TGF␤ treatment of NMuMG cells induced a progressive phosphorylation of FAK on Tyr-397, its major autophosphorylation site. TGF␤-induced FAK phosphorylation was observed within 4 h, and maximal effects were observed at 12-24 h. However, we observed a marked reduction in the TGF␤-induced FAK autophosphorylation in NMuMG-siDab2 cells (Fig. 6C), demonstrating that Dab2 is necessary for integrin activation during TGF␤-induced EMT. This finding was further confirmed by immunoprecipitation of NMuMG-siControl and NMuMG-siDab2 cell lysates with anti-FAK antibody followed by immunoblot analysis with anti-phosphotyrosine antibody (PY) (Fig. 6D). Similarly, maximal TGF␤induced FAK phosphorylation was observed at 12-24 h and required the expression of Dab2 (Fig. 6D). Finally, we verified the restoration of integrin signaling after re-expression of Dab2 FIG. 5. Integrin engagement is required for TGF␤-mediated EMT. A, NMuMG-siDab2 cells display a decreased adhesion and spreading. NMuMG-siControl and NMuMG-siDab2 cells were trypsinized, and equal number of cells were allowed to adhere to culture dishes for the times indicated. Representative images from NMuMG-siDab2 and NMuMG-siControl cells are shown. B, NMuMG-siControl or NMuMG-siDab2 cells were equally seeded for an adhesion assay. After a 1-h incubation, non-adherent cells were removed by washing, and the adherent cells were monitored by the MTT adhesion assay as described under "Experimental Procedures." C, NMuMG cells were treated with TGF␤ for 24 h in the presence or absence of RGD peptide followed by examination by phase contrast microscopy. D, NMuMG cells were treated with TGF␤ for 24 h in the presence or absence of RGD peptide. Total cellular lysates were analyzed for N-cadherin expression by Western blot using ␣-N-cadherin antibody.
FIG. 6. Localization and association with ␤1 integrin of Dab2 during TGF␤-induced EMT. A, TGF␤ promotes the accumulation of Dab2 into the membrane. NMuMG cells were treated with TGF␤ (5 ng/ml) for the times indicated. The cytosolic and membrane lysates were subjected to Western blotting with ␣-Dab2 antibody. B, Dab2 transiently binds to ␤1 integrin. NMuMG cells were treated with TGF␤ for the times indicated and membrane cell lysates prepared. ␤1 integrin protein was immunoprecipitated using ␣-␤1 integrin antibody from 500 g of membrane cell lysates, and immunoprecipitates were analyzed for the presence of Dab2 by Western blotting (upper panel). As a control, a TGF␤ lysate treated for 24 h was immunoprecipitated with normal mouse IgG. The blot was reprobed with ␣-␤1 integrin antibody (lower panel). C and D, TGF␤-induced FAK phosphorylation is decreased in NMuMG-siDab2 cells. C, lysates prepared from NMuMG-siControl and NMuMG-siDab2 cells treated with TGF␤ for the times indicated were analyzed for FAK phosphorylation using an antibody specific for FAK autophosphorylation at tyrosine 397 (Y397, upper panel). Equivalent protein loading was verified by ␣-Hsp90 immunoblotting. D, lysates were also immunoprecipitated with ␣-FAK antibody, followed by Western blotting with ␣-phosphotyrosine antibody (PY). The probe was stripped and reprobed with ␣-FAK to assess FAK levels immunoprecipitated (lower panel). E, re-expression of Dab2 restores TGF␤-induced FAK activation. TGF␤-induced activation of FAK was determined by immunoblotting with phosphotyrosine FAK antibody Y397.
Fibronectin Prevents TGF␤-induced Apoptosis and Rescues TGF␤-induced EMT in NMuMG-siDab2 Cells-We next sought to reinforce that apoptosis in NMuMG-siDab2 cells is caused by impairment of ␤1 integrins activation. Because it has been described that the extracellular matrix protein, fibronectin, regulates cell survival by interacting with and activating ␤1 integrin, we tested whether enhanced integrin engagement by cell attachment to fibronectin can counteract TGF␤-induced apoptosis in NMuMG-siDab2 cells. We, indeed, observed that plating NMuMG-siDab2 cells onto dishes coated with fibronectin protected them against TGF␤-induced apoptosis (Fig. 7A). In agreement with this morphological examination, immunoblot analysis with anti-Bim antibody revealed that the adhesion of NMuMG-siDab2 cells to fibronectin prevents TGF␤induced Bim up-regulation (Fig. 7B). We subsequently reasoned that if cell adhesion to fibronectin protects NMuMG-siDab2 cells against TGF␤-induced apoptosis by enforced integrin stimulation, FAK activation should also be restored in these cells. To verify this, an immunoblot was performed using the specific anti-phosphoFAK (Y397) antibody. As shown in Fig. 7C, FAK phosphorylation in response to TGF␤ in NMuMG-siDab2 cells is restored when these cells are seeded on fibronec-tin. Altogether, these results indicate that forced activation of integrin signaling prevents TGF␤-induced apoptosis due to down-regulation of Dab2 expression.
Interestingly, we also found that plating NMuMG-siDab2 cells onto dishes coated with fibronectin not only reversed TGF␤-induced apoptosis but allowed them to transdifferentiate in the presence of TGF␤ (Fig. 7A). This observation was confirmed by Western blot analysis, which shows that fibronectin restores TGF␤-induced up-regulation of the fibroblastic marker, N-cadherin (Fig. 7D). Thus, restoration of the integrin signaling pathway can override down-regulation of Dab2 and restore TGF␤-induced EMT. DISCUSSION We have described a role for Dab2 during TGF␤-induced EMT. We found that concomitant with the promotion of EMT, TGF␤ induces an increase of Dab2 expression, the transient translocation of Dab2 to the membrane, and its binding with ␤1 integrin. To determine whether these effects are related to EMT, we decided to stably reduce Dab2 expression. We observed that down-regulation of Dab2 not only prevents TGF␤induced EMT but induces the cells to undergo apoptosis. Thus, up-regulation of Dab2 in response to TGF␤ is crucial to promote integrin-dependent cell adhesion and survival, which is necessary for TGF␤ to induce EMT.
Previous studies have demonstrated an induction of Dab2 expression following treatment of cells with various stimuli. This is the first demonstration that TGF␤ can induce Dab2 mRNA and protein expression levels. Interestingly, in the previous studies the up-regulation of Dab2 was shown to be associated with cellular morphological remodeling. Indeed, Dab2 has been demonstrated to be induced upon both 12-O-tetradecanoylphorbol-13-acetate-mediated megakaryocyte differentiation of human chronic myeloid leukemic K562 cells (35) and RA-mediated endodermal differentiation of F9 embryonic stem cell-like teratocarcinoma cells (24). Herein, we confirm that Dab2 expression in F9 cells is induced upon RA-induced differentiation but, most importantly, demonstrate that ablation of RA-induced Dab2 expression levels blocks differentiation and instead promotes RA-induced apoptosis. Taken together, these results demonstrate that Dab2 expression levels mediate an important regulatory role during TGF␤-induced EMT and during RA-mediated F9 differentiation and provide further support to the previously postulated notion of a role for Dab2 during morphological remodeling or cellular differentiation.
In accord with a role of Dab2 in cellular differentiation, several lines of evidence suggest that Dab2 is involved in regulation of cytoskeleton-based functions. Indeed, Dab2 has been shown to participate in endocytosis by acting as an adaptor for coated vesicles by interacting with AP-2, clathrin, megalin, or myosin VI (17)(18)(19)(20)22). The role of Dab2 in endocytosis has been confirmed in mice conditionally null for Dab2 that exhibit protein transport defect in kidney (16). Moreover, it has also been shown that Dab2 can promote macrophage adhesion and regulate epithelial cell polarity and positioning (14,15,23). However, the molecular mechanism(s) responsible for mediating these events has not been clearly elucidated. In the present study, we show that down-regulation of Dab2 in NMuMG cells prevents FAK phosphorylation in response to TGF␤ and causes a decreased adhesion, two events dependent upon integrin function, suggesting a role for Dab2 in integrin activation. Numerous studies have shown that integrin-mediated cell attachment to extracellular matrix components provides prosurvival effects and that loss of integrin function is associated with apoptosis (34,36). We demonstrate, herein, that modulation of Dab2 levels alters cellular adhesion and that fibronectin, which interacts with and results in forced activation of ␤1 integrin and subsequently FAK, overrides down-regulation of Dab2 effects by rescuing apoptosis and restoring EMT in response to TGF␤. This thus provides evidence that the effects caused by down-regulation of Dab2 in NMuMG cells result from impairment of integrin activation.
The mechanism by which Dab2 mediates the activation of integrin likely involves an association between Dab2 and ␤1 integrin (Fig. 6), presumably through binding the cytoplasmic tail of integrin as has been reported for other intracellular proteins that regulate signal transduction to and from integrin-adhesion receptor (37). Interestingly, it has been recently reported that proteins that contain a protein phosphotyrosine binding domain, like Dab2, can interact with NPXY motifs found within the integrin cytoplasmic tail (38,39). Calderwood et al. (40) have previously demonstrated that Dab2 can directly interact with ␤3 and ␤5 integrin subunits yet failed to observe a direct interaction between Dab2 and ␤1 integrin in vitro. Therefore, it is possible that the association between Dab2 and ␤1 integrin that occurs during TGF␤-induced EMT is likely not a direct one and is mediated by other cytoskeleton proteins. In this regard, integrins can associate with others membrane proteins that can regulate integrin function. Tetraspanin transmembrane proteins, for example, are associated with integrins and link them to the phosphatidylinositol signaling pathway (41). Therefore, we speculate that Dab2, which can directly interact with phosphoinositides, may associate with a member of the tetraspanin superfamily and affect the potency of integrin signal (19). Moreover, integrins can also cooperate with growth factor receptors and together coordinately transduce their signals. It has been proposed that many cellular responses, including cell survival and proliferation in response to soluble growth factors such as epidermal growth factor and platelet-derived growth factor, are dependent on integrinmediated adherence (34,36,42,43). It has previously been shown that impairment of ␤1 integrin function by neutralizing antibody inhibits TGF␤-induced EMT (8). In this study, however, the role of integrin on apoptosis was not determined because cells were grown on collagen during the neutralizing antibody treatment, hence providing a survival signal. Here, we have demonstrated that integrin activation is necessary for TGF␤-induced EMT, because the integrin inhibitory RGD peptide not only blocks EMT but induces TGF␤-mediated apoptosis. Thus, it is possible that similar to integrin binding to the epidermal growth factor receptor and potentiating its signaling, Dab2 may serve to couple or bridge the TGF␤ receptor complex with integrins, thereby modulating TGF␤ and integrin signaling. In a recent study, Huang et al. (44) report on the role of Dab2 during 12-O-tetradecanoylphorbol-13-acetate-induced megakaryocytic differentiation. They also implicate an important integrin modulatory role for Dab2 during differentiation; however, in contrast to our results, they show that Dab2 inhibits adhesion and integrin activation. A priori, these results are contradictory to those presented here; however, it must be stressed that different cell types and differentiation models, as well as different subtypes of ␤ integrin, were under investigation. Additional studies are nevertheless required to further determine with certitude how Dab2 induces integrin modulation in various cell types.
Our data show that the decrease of integrin signaling caused by down-regulation of Dab2 leads not only to a blockade of TGF␤-induced EMT but to an increase in apoptosis. We have previously reported that TGF␤ signaling was abrogated in a fibrosarcoma HT1080-derived parental cell line that expresses a mutated form of Dab2, rendering it less stable (29). Thus our results herein demonstrating that TGF␤ can still induce apoptosis in NMuMG-siDab2 cells are somewhat surprising and prompted us to analyze TGF␤-induced Smad2 and Smad3 phosphorylation in these cells. As in HT1080 fibroblasts, the absence of Dab2 in NMuMG epithelial cells abrogated TGF␤induced Smad2 and Smad3 phosphorylation (data not shown). Thus, we speculate that TGF␤ can still mediate some responses in NMuMG-siDab2 cells, including apoptosis, in that impairment of integrin activation may be mediated through a Smadindependent TGF␤ pathway such as the phosphatidylinositol-3K/Akt, RhoA, or mitogen-activated protein kinase signaling pathways. Further work to determine whether these pathways are still activated in siRNA Dab2 cells is currently under investigation. The results nevertheless highlight an existing balance between apoptosis and EMT. It is consistent with acquisition of apoptosis resistance that has been frequently described to accompany EMT. For example, it has been shown that TGF␤ can induce EMT in a subpopulation of fetal rat hepatocytes, a transition that renders these cells resistant to TGF␤-mediated apoptosis (45). Moreover, Muraoka et al. (6) have shown that inhibiting TGF␤ signaling blocks not only tumor invasiveness but induces apoptosis in vivo. In addition, at a transcriptional level, Grooteclaes et al. (6) have shown that the E1a-interacting corepressor protein (CtBp) represses both epithelial and pro-apoptotic gene expression, potentially contributing both to promote EMT and prevent apoptosis at the same time (6). In the same way, we propose that Dab2 acts as a critical switch during TGF␤-mediated EMT by preventing apoptosis and thus allowing cells to undergo EMT. This hypothesis is further corroborated by the fact that Dab2 nullizygous mice exhibit increased apoptosis (16).
Dab2 is considered a potential tumor suppressor because it is down-regulated in various types of tumors (12,14,(47)(48)(49). Our data documenting a required role for Dab2 during EMT suggest that Dab2 function may be pro-tumorigenic and thus contradictory to its reported tumor suppressor activity. However, it is possible that, similar to TGF␤, Dab2 has a dual role during tumor progression. In becoming invasive and able to metastasize, epithelial cells obviously acquire properties rendering them resistant to apoptosis. By protecting cells against apoptosis, Dab2 would therefore function as a promoter of invasiveness and tumor formation. Alternatively, epithelial cells that first possess some genetic alteration rendering them resistant to apoptosis may become in the absence of Dab2 less adhesive and facilitate their invasiveness and the formation of metastasis during the late phases of carcinogenesis. It is therefore possible that Dab2 has an alternative function during tumor progression.
Considering that EMT and apoptosis are crucial during development, it is highly probable that the modulatory role of Dab2 in regulating the balance between these two cellular events occurs during tissue development or remodeling. In particular, cells from the mammary gland and prostate are not continuously renewed. Their interaction with the matrix appears to promote differentiation, whereas degradation of the extracellular matrix promotes their apoptosis and contributes to their involution (50). Interestingly, Dab2 has also been implicated during prostate remodeling. Indeed, during rat castration, which results in involution of the prostate, Dab2 protein expression is increased and switches from the luminal, where cells undergo apoptosis, to the remaining basal prostatic cells (51). This result and our present finding suggest, therefore, that the apoptotic protective function of Dab2 may occur during physiological processes, in particular during cellular differentiation induced by hormonal level modification. Drs. B. Pratt and S. Ledbetter at Genzyme Inc. for generous provision of TGF␤. We also thank Dr. Hocevar for generating the F9-siDab2 cell line and Marie Zafiropulos for assistance with the F9 results.