Integrin (cid:1) 1 Signaling Is Necessary for Transforming Growth Factor- (cid:1) Activation of p38MAPK and Epithelial Plasticity*

Transforming growth factor- (cid:1) (TGF- (cid:1) ) can induce epithelial to mesenchymal transdifferentiation (EMT) in mammary epithelial cells. TGF- (cid:1) -meditated EMT in-volves the stimulation of a number of signaling pathways by the sequential binding of the type II and type I serine/threonine kinase receptors, respectively. Integrins comprise a family of heterodimeric extracellular matrix receptors that mediate cell adhesion and intra-cellular signaling, hence making them crucial for EMT progression. In light of substantial evidence indicating TGF- (cid:1) regulation of various (cid:1) 1 integrins and their extracellular matrix ligands, we examined the cross-talk between the TGF- (cid:1) and integrin signal transduction pathways. Using an inducible system for the expression of a cytoplasmically truncated dominant negative TGF- (cid:1) type II receptor, we blocked TGF- (cid:1) -mediated growth inhibition, transcriptional activation, and EMT progression. Dominant negative TGF- (cid:1) type II receptor expression inhibited TGF- (cid:1) signaling to the SMAD and AKT pathways, but did not block TGF- (cid:1) -mediated p38MAPK activation. Interestingly, blocking integrin (cid:1) 1 function inhibited TGF- (cid:1) -mediated p38MAPK However, blocking integrin (cid:1) 1 adhesion inhibited TGF- (cid:1) -mediated p38MAPK stimulation and EMT progression. The findings suggest integrin (cid:1) 1 adhesion and p38MAPK signaling are necessary but not sufficient for the complex process of TGF- (cid:1) -induced EMT.

TGF-␤ 1 has emerged as a multifunctional cytokine involved in autocrine and paracrine mediation of development, proliferation, wound healing, and pathologic processes (1). TGF-␤ was initially described as a secreted factor that induced an anchorage-independent transformed phenotype in fibroblasts in vitro (2). TGF-␤ also induces phenotypic changes in epithelial cells and activates transcription of integrin genes as well as their extracellular matrix (ECM) ligands. The spectrum of TGF-␤ responses are initiated by its interaction with cell surface receptors that form a heterotetrameric complex of two type I and two type II serine/threonine kinase receptors (3,4). The type II receptor binds to the TGF-␤ ligand to recruit and phosphorylate the type I receptor (4). The activated ligand-receptor complex in turn conveys the signal to the nucleus via the phosphorylation and activation of the SMAD signaling pathway (5). TGF-␤ also stimulates parallel downstream signaling pathways involving RhoA (6), stress kinases (i.e. JNK, p38MAPK) (7-9), phosphatase2A (10), and PI3-kinase/AKT (11). Despite the cross-talk among the various signaling pathways, there is compelling data to suggest a separation of function among the individual signaling molecules (for review, see Ref. 12). Current evidence suggests that the SMAD pathway is involved in growth inhibition, while the Rho proteins, and PI3-kinase pathways are likely operative in phenotypic changes (6,13,14).
Epithelial cell adhesion and communication with the ECM and neighboring cells play fundamental roles in epithelial transdifferentiation into a mesenchymal phenotype (EMT). EMT has been defined by the loss of cell polarity, break down of cell-cell junctions, and the acquisition of an elongated cell shape. While tight junctions and adherence junctions mediate epithelial cell-cell interactions, its interaction with the substratum is through the heterodimeric integrin receptors (15). A large body of literature shows that the expression of integrins is regulated by TGF-␤, in particular the upregulation of ␤ 1 integrins (16 -18). In addition, integrinmediated adhesion potentiates many of the same signaling pathways regulated by TGF-␤ such as the PI3-kinase-AKT, stress kinases, and ERK1/2-stimulated pathways, (19). However, little is known about the underlying cross talk of TGF-␤ and integrin signaling.
Given the convergence of TGF-␤ and integrin signaling on common downstream factors as well as their central effects on cytoskeletal organization, we hypothesized that there may be cooperation between TGF-␤ and integrins in both signal transduction and the modulation of EMT progression. To isolate the role of TGF-␤ we disrupted TGF-␤ signaling by inducibly expressing a cytoplasmically truncated T␤RII (DNIIR, (20)) in non-transformed mouse mammary epithelial cells, NMuMG. Elevated expression of DNIIR caused the loss of TGF-␤ responsiveness to growth, SMAD-mediated transcriptional regulation, and EMT progression without the apparent loss of p38MAPK activation. However, blocking integrin ␤ 1 adhesion inhibited TGF-␤-mediated p38MAPK stimulation and EMT progression. The findings suggest integrin ␤ 1 adhesion and p38MAPK signaling are necessary but not sufficient for the complex process of TGF-␤-induced EMT.

MATERIALS AND METHODS
Cell Culture-The NMuMG and T47D cells were purchased from the American Type Culture Collection (Manassas, VA) and propagated in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 10 g/ml insulin. The ecdysone-inducible mammalian expression system was used to regulate the expression of the DNIIR gene (a kind gift from R. Derynck, (20)) in NMuMG cells as described by the manufacturer (Invitrogen Co., Carlsbad, CA). Briefly, clones of stable lines expressing the pVgRXR plasmid (containing the ecdysone receptor) were generated and selected for maximal expression. Next, the DNIIR cDNA in the inducible expression plasmid (pIND) was introduced into the selected pVgRXR-expressing clone. Finally, NMuMG clones expressing both plasmids were identified by sensitivity to TGF-␤ (by growth inhibition and transcriptional reporter assays). Ponasterone A (Invitrogen Co., 0 -5 M) was added to the cells 24 h prior to experiments to achieve the desired DNIIR expression level. Ponasterone (5 M) was added to wild type NMuMG cells (not transfected with the pIND vector) to examine its effects on TGF-␤-mediated epithelial differentiation and p38MAPK activation. No ponasterone-associated results were observed (data not shown).
Virus Generation-The DNIIR cDNA was subcloned into pBabe retroviral vector. Then both DNIIR and T␤RII (kindly provided by W. Grady, Vanderbilt University) retrovirus constructs were transfected into amphetrophic retrovirus producing Phoenix cells (a gift from Gary Nolan, Stanford University). Conditioned media was allowed to incubate with target cells for 24 h, after which point the media was replaced for subsequent experiments. The DNp38MAPK adenovirus (21) was amplified in human embryonic kidney cells, HEK293, cesium chloridepurified and titrated by serial dilution (Takara Shuzo Co. Ltd.). Experiments were performed 48 h after retrovirus or adenovirus infection.
Thymidine Incorporation and Luciferase Reporter Assay-Cell growth was assayed by [ 3 H]thymidine (PerkinElmer Life Sciences) incorporation of cells treated for 24 h with TGF-␤ and loaded with [ 3 H]thymidine 2 h prior to harvesting (22). Cells were washed and measured by scintillation counting. Transcriptional activation was tested in cells transfected with the luciferase (firefly) reporter construct cDNAs indicated in conjunction with a cytomegalovirus-driven renal luciferase plasmid (Promega Inc.). The 3TP-Lux reporter from J. Massagué (Memorial Sloan-Kettering Cancer Center, NY) and CAGA reporter from J.-M. Gauthier (Laboratoire Glaxo Wellcome, Les Ulis Cedex, FranceTranscriptional) were used to asses TGF-␤-signaling to the nucleus in the presence and absence of DNIIR expression. The dual-luciferase assays were performed on lysed cells as indicated by the manufacturer, Promega Inc. The firefly luciferase activity determined by luminometer measurements were normalized to renela luciferase activity, and data is reported in relative luminescent units.
Kinase Assays-Cell extracts (400 g of protein) in HEPES buffer were subjected to immunoprecipitation by an antibody to p38MAPK (C-20, Santa Cruz Biotechnology) for 2 h at 4°C, followed by incubation with protein G-Sepharose (Sigma) for 1 h. The beads were washed and resuspended in 3 M [␥-32 P]ATP, 100 M ATP, 10 mM MgCl 2 , 1 mM dithiothreitol, 10 mM ␤-glycerol phosphate, and ATF2 as substrate. The assay was allowed to incubate at 30°C for 20 min prior to denaturation in Laemmli buffer and separation on a 5-20% gradient polyacrylamide gel. The gel was then fixed and visualized by phosphorimaging.
Immunofluorescence-Cells grown on glass coverslips with or without collagen I coating (10 g/ml, Collaborative Labs) were fixed with 100% cold methanol for immunolocalization of ZO-1 (Chemicon) and E-cadherin (Santa Cruz Biotechnology) or by 4% paraformaldehyde for Smad1/2/3 (Santa Cruz Biotechnology). Subsequently PBS containing 0.1% Triton X-100 and 1% horse serum was used for permeabilization and blocking. The primary antibody (1:1000) was incubated for 1 h, and Cy3 (Sigma) secondary antibody was used for fluorescence visualization. F-actin and nuclear staining was done by fixing the cells in 4% paraformaldehyde followed by incubation with Texas-Red-conjugated phalloidin (Molecular Probes, Eugene, OR) or Hoechst 33253, respectively.

RESULTS
DNIIR Expression-When the cytoplasmic region of T␤RII is truncated and lacks the serine/threonine kinase domain it is able to bind TGF-␤ and recruit the type I receptor, but is unable to signal via the SMAD signaling pathway, and is thus termed a dominant negative T␤RII, DNIIR (20). We sought to generate stable clones in non-transformed mouse mammary ductal epithelial cells, NMuMG, that have an inducible DNIIR gene to identify down stream signaling pathways that may be differentially regulated by TGF-␤. Fig. 1A shows an increase in FLAG-tagged DNIIR protein expression that corresponds to increasing concentrations of ponasterone (an ecdysone analogue). To verify that the expression of DNIIR resulted in a functional loss of growth inhibitory and SMAD-mediated transcriptional activation by TGF-␤, as reported earlier (20), thymidine incorporation assays and transcriptional reporter as-

FIG. 1. Inducible expression of DNIIR in NMuMG cells indicated concentrations of ponasterone (Pon) were added to cells 24 h prior to assaying for protein expression, TGF-␤mediated growth inhibition, or TGF-␤-induced transcriptional activation.
A, ponasterone-induced cells were harvested for immunoblotting with anti-FLAG antibody. This is representative of three blots for the FLAG-tagged DNIIR. B, increasing concentrations of TGF-␤ were incubated (24 h) with ponasteroneinduced cells and assayed for [ 3 H]thymidine incorporation (in triplicate, n ϭ 2). The graph is represented as a percent of growth, having the absence of TGF-␤ expressed as 100%. C, the activation of the 3TP-Lux and CAGA luciferase-conjugated reporters were assayed after TGF-␤ treatment for 24 h (in triplicate, n ϭ 2). says were performed, respectively. As the DNIIR expression was induced with 0 -5 g/ml ponasterone, TGF-␤-mediated growth inhibition was essentially abolished by 2 and 5 M ponasterone (Fig. 1B). Next, the role of DNIIR expression in TGF-␤-mediated transcriptional activation was examined by 3TP-lux (the heterologous PAI-1 promoter with 3ϫ 12-O-tetradecanoylphorbol-13-acetate-responsive elements (23)) and CAGA (the SMAD DNA binding domain (24)) luciferase reporter assays. The addition of TGF-␤ had a reduced induction on both the reporters when ponasterone was present (Fig. 1C). At 1 g/ml ponasterone TGF-␤ transcriptional activation of 3TP-Lux was reduced 1.5-fold, and the CAGA activity by 2.7fold. Treatment with 2 and 5 M ponasterone resulted in TGF-␤-mediated transcriptional activity that was reduced to basal levels. These results suggest that responses to TGF-␤ depending either solely on SMAD signaling or more complex signaling activating 3TP-Lux is blocked by the expression of DNIIR. Thus the expression of DNIIR in stable inducible clones cause cells to become insensitive to TGF-␤-regulation of growth and SMAD-mediated transcriptional activation.
To further understand the repercussions of DNIIR expression in NMuMG cells we studied its role in TGF-␤-mediated EMT. The expression of an apical tight junction-associated protein, ZO-1, was used as a marker to indicate cell polarity and cell-cell junction stability for the epithelial phenotype of NMuMG cells. We have previously shown that TGF-␤-mediated fibroblastic conversion causes a distinct loss of ZO-1 expression at tight junctions (14). NMuMG cells in the presence and absence of ponasterone and TGF-␤ (200 pM for 16 h) were examined for immunolocalization of ZO-1 and Smad2/3, along with corresponding phase contrast images for phenotypic changes. Fig. 2 illustrates TGF-␤-induced Smad2/3 nuclear localization accompanied by the loss of ZO-1 expression and the acquisition of a mesenchymal phenotype. Increasing expression levels of DNIIR reduced TGF-␤-mediated NMuMG fibroblastic conversion and Smad2/3 nuclear localization. We were able to follow the sequential loss of junctional ZO-1 staining and subsequent acquisition of the elongated cellular phenotype as ponasterone concentrations were reduced in the presence of TGF-␤. These results highlight the importance of the functionality of T␤RII in TGF-␤-mediated SMAD nuclear localization and the complexity of EMT progression.
Differential Signaling Activation in the Presence of DNIIR Expression-As EMT is a well described phenotypic consequence of TGF-␤ signaling, we investigated the activation states of various signaling proteins potentially involved in the inhibition of EMT by the expression of DNIIR in NMuMG cells. Fig. 3 shows that Smad2 phosphorylation occurs within 30 min of incubation with 200 pM TGF-␤ and maintains its phosphorylated status at 3 h. The expression of DNIIR, using 5 M ponasterone, diminished TGF-␤-mediated Smad2 phosphorylation. Total Smad2 levels did not change as a consequence of DNIIR expression over the time course. Next, we examined other signaling pathways activated by TGF-␤. In the absence of DNIIR expression, cells showed modest TGF-␤-induced p38MAPK and AKT (an indicator of PI3-kinase activity) phosphorylation with negligible ERK1/2 activation. TGF-␤-mediated AKT phosphorylation was blocked by DNIIR expression. However, DNIIR expression did not block TGF-␤-stimulated p38MAPK phosphorylation. DNIIR expression did not seem to alter ERK1/2 phosphorylation levels with respect to total ERK1/2 levels in the NMuMG cells. Cells were treated with EGF as a positive control for the activation of ERK1/2 and AKT. The mitogen-activated protein kinase inhibitor, PD98059 (Cell Signaling Technology), was used as a negative control. These results indicate that the receptor-ligand complex consisting of the type I TGF-␤ receptor and DNIIR results in the loss of downstream regulation of the AKT and SMAD signaling pathways but were ineffective in preventing TGF-␤-stimulated phosphorylation of p38MAPK.
DNIIR Effects on Integrin Expression-A large body of evidence indicates that TGF-␤ signaling is involved in integrin expression. Thus it was important to determine whether the down-regulation of TGF-␤ autocrine signaling through the induced expression of DNIIR affects integrin expression. As NMuMG cells express the collagen receptor ␣ 2 ␤ 1 that is known to be important in integrin-mediated activation of p38MAPK (25), we assessed whether overexpression of DNIIR affected expression of this integrin. To do this, NMuMG cells were incubated for 24 h in serum-free medium in the presence or absence of DNIIR expression (by 5 M ponasterone) after which the cells were subjected to flow cytometry. As seen in Fig. 4 no changes in the cell surface expression of this integrin was observed. Furthermore, there was also no alteration in expression of the other major collagen-integrin, ␣ 1 ␤ 1 . These results suggest that overexpression of DNIIR for 24 h does not alter the expression levels of the major collagen binding integrins in NMuMG cells under serum-free conditions.
p38MAPK Activation by TGF-␤ Is Dependent on Integrin Ligation-Integrin ␤ 1 ligation is known to induce p38MAPK phosphorylation and activation (19,25). The inducible DNIIR expression system was ideal for examining if cooperative interactions exists between TGF-␤ and integrin signaling at the level of p38MAPK activity. Since NMuMG cells express the collagen I binding integrins, and the ligation of integrin ␣ 2 ␤ 1 by collagen I is known to activate p38MAPK, these cells were grown on collagen I to examine the specific role of ␤ 1 integrins in TGF-␤ signaling. Under these growth conditions TGF-␤ induced Smad2 and p38MAPK phosphorylation as before (Fig.  5A). However, incubation of the cells with a neutralizing integrin ␤ 1 antibody caused a decrease in p38MAPK phosphorylation but had no effect on TGF-␤ induction of Smad2 phosphorylation. Further, p38MAPK phosphorylation was observed when DNIIR-induced NMuMG cells were treated with TGF-␤ on either the tissue culture plates or collagen I; however, this was diminished by the co-incubation with an integrin ␤ 1 -neu-tralizing antibody (Fig. 5B). The neutralizing antibody did not alter the ERK1/2 or AKT responses to TGF-␤ (data not shown).
To determine whether the p38MAPK phosphorylation was integrin ␤ 1 -dependent and not simply adhesion-dependent we assessed the effect of TGF-␤ on NMuMG cells cultured on poly-L-lysine-coated plates. This allowed cells to attach based on electrostatic interactions, whereas cells plated onto collagen-coated dishes spread due to integrin engagement (26). Similar to the results seen with the integrin ␤ 1 -neutralizing antibody, TGF-␤-mediated Smad2 phosphorylation was not effected by poly-L-lysine-mediated cell adhesion; however, there was no induction of TGF-␤-mediated p38MAPK phosphorylation (Fig. 5C). These results suggest that integrin-mediated cell adhesion is required for TGF-␤ to stimulate p38MAPK phosphorylation, and TGF-␤-induced Smad2 phosphorylation is independent of integrin-mediated adhesion.
The regulation of p38MAPK phosphorylation was then directly determined by examining kinase activity in the context of T␤RII and integrin ␤ 1 disruption. The in vitro kinase assay using immunoprecipitated p38MAPK from treated cells and ATF2 as substrate verified a TGF-␤-dependent increase in p38MAPK activity in the presence and absence of ponasterone (Fig. 6). The preincubation of cells with the integrin ␤ 1 -neutralizing antibody prior to TGF-␤ stimulation diminished the levels of ATF2 phosphorylation. However, p38MAPK activation by ultraviolet light exposure was not effected by the integrin ␤ 1 antibody, demonstrating that the requirement of integrin ␤ 1 ligation was specific for TGF-␤-induced p38MAPK activation. These results were consistent with p38MAPK phosphorylation previously observed (see Figs. 3 and 5) in confirming that DNIIR expression does not block TGF-␤-mediated p38MAPK activation, and integrin ␤ 1 -mediated adhesion is necessary for TGF-␤ signaling to the stress kinase.
T␤RII Expression Required for Differential Signaling Activation-Next we sought to determine whether TGF-␤-mediated p38MAPK activation in our inducible system was mediated by the formation of a complex consisting of the type I TGF-␤ receptor and DNIIR or the type I TGF-␤ receptor and endogenous T␤RII. To do this we expressed either DNIIR or T␤RII by retroviral infection in a T␤RII null mammary epithelial cell line, T47D. Not surprisingly, T47D cells were insensitive to TGF-␤ for SMAD and p38MAPK activity (Fig. 7). The transduction of DNIIR expression was not able to restore either SMAD or p38MAPK phosphorylation by TGF-␤. However, T␤RII transduction restored both TGF-␤-induced Smad2 and p38MAPK phosphorylation. Further preincubation with integrin ␤ 1 antibody blocked p38MAPK phosphorylation, but did not effect Smad2 activity, as observed with NMuMG cells. These results suggest that p38MAPK activation in the NMuMG cells is transduced by the endogenous type I and type II TGF-␤ receptor complex and that the level of DNIIR expres- sion achieved by the inducible system was insufficient to completely suppress endogenous T␤RII activity. Thus the threshold of T␤RII activity for p38MAPK activation may be lower than that for SMAD activation.
Integrin ␤ 1 and p38MAPK Signaling Involved in TGF-␤mediated Phenotypic Change-Since integrin function appears to be involved in certain aspects of TGF-␤ signaling and is important for cell-ECM interactions that effect cell morphology (27), we examined whether integrin ␤ 1 signaling affected TGF-␤-induced EMT progression. NMuMG cells, grown on collagen I-coated coverslips, were incubated with integrin ␤ 1 -neutralizing antibodies or rabbit-IgG in the presence or absence of 200 pM TGF-␤ for 16 h under serum-free conditions. Cell phenotype was subsequently examined by phase contrast microscopy concurrent with actin, ZO-1, and E-cadherin subcellular localization by immunofluorescent detection (Fig. 8). When NMuMG cells were grown on collagen I, stress fiber formation was observed (Fig. 8, inset) accompanied by junctional localization of ZO-1 and E-cadherin. TGF-␤ treatment caused increased stress fiber formation and the acquisition of a fibroblastic morphology along with the delocalization of ZO-1 and E-cadherin. When cells were grown in the presence of anti-integrin ␤ 1 antibody prominent cortical actin staining accompanied with reduced ZO-1 tight junctional localization was observed in untreated cells. The adhesion junctions, however, seemed to be intact with E-cadherin localization at the cell-cell contacts. Subsequent TGF-␤ treatment did not seem to alter the epithelial morphology of the cells in the presence of the integrin ␤ 1 antibody. The incubation with rabbit IgG had little effect on the integrity of the cell junctions, and TGF-␤ caused a phenotypic differentiation to take place exhibiting the loss of ZO-1 and E-cadherin junctional localization (data not shown). Thus specifically disrupting integrin ␤ 1 function inhibited TGF-␤-induced EMT in NMuMG cells.
To determine whether blocking TGF-␤-mediated p38MAPK activity may be a component in the inhibition of EMT by integrin ␤ 1 neutralizing antibody, we transduced a dominant negative p38MAPK (DNp38MAPK-␤, (21) Y. Wang, Univ. of California, San Diego) by adenovirus infection (20 multiplicity of infection). The expression of DNp38MAPK-␤ in NMuMG cells inhibited the TGF-␤-induced fibroblastic transition. As a negative control, adenoviral expression of ␤-galactosidase did not show any noticeable differences in TGF-␤-induced epithelial differentiation from the un-infected control (data not shown). The DNp38MAPK-␤-mediated down-regulation of p38MAPK phosphorylation in NMuMG cells is shown in the right panel of Fig. 8. Thus limiting p38MAPK activity did not alter TGF-␤-induced actin stress fiber formation, yet it resulted in the consistent localization of ZO-1 and E-cadherin at the cell junctions after 16 h of treatment with TGF-␤. Together, these results suggest that impairing either integrin ␤ 1 or p38MAPK signaling inhibit TGF-␤-induced EMT progression in NMuMG cells. Furthermore, blocking EMT in NMuMG cells through the inhibition of integrin ␤ 1 may in part be because it inhibits p38MAPK signaling.

DISCUSSION
The extracellular microenvironment plays a critical role in the ability of an epithelial cell to transition to a more plastic phenotype. This is exemplified by the influence that TGF-␤ and ECM components have one basic epithelial cell functions such as the morphology, polarity, and adhesion through TGF-␤ receptors and integrin function. Based on the common downstream signaling molecules and overlapping cellular consequences induced by TGF-␤ and integrins, we hypothesized that there may be cooperation between the two pathways in EMT. We utilized an inducible system expressing DNIIR in mammary epithelial cells and employed an integrin ␤ 1 antibody to explore potential cross-talk between the two cell surface receptors. Effectively limiting T␤RII activity by the induction of DNIIR expression was able to block TGF-␤-mediated growth inhibition, SMAD-mediated transcriptional activation, and EMT. TGF-␤-induced Smad2, and AKT phosphorylation was blocked by the expression of DNIIR. Yet, the induced expression of the cytoplasmically truncated T␤RII at the levels that blocked the other measured responses did not block TGF-␤mediated p38MAPK activation. The potential for a lower T␤RII threshold for p38MAPK activation was indicated by the results using cells with a T␤RII null background, T47D. Blocking integrin-␤ 1 -mediated adhesion resulted in the inhibition of TGF-␤-mediated activation of p38MAPK. Further, TGF-␤-mediated EMT was blocked both by the neutralizing integrin ␤ 1 antibody and by limiting p38MAPK activity with a DNp38MAPK-␤-expressing adenovirus. However, as DNIIR expressing cells are insensitive to TGF-␤-mediated EMT progression, our data suggest that p38MAPK activation is necessary but not sufficient for the fibroblastic transition of NMuMG cells.
We investigated TGF-␤ signaling in NMuMG cells expressing an inducible DNIIR gene to assess the effect of a functional reduction of T␤RII expression. The NMuMG cells were used as a model system for the expression of DNIIR because TGF-␤ induces both growth arrest and EMT in these cells (6,28,29). The expression of DNIIR at higher levels limited TGF-␤-induced growth inhibition and diminished transcriptional induction of the luciferase reporters in NMuMG cells (Fig. 1). Similar DNIIR expression levels were also observed to block TGF-␤mediated Smad2/3 nuclear translocation, loss of ZO-1 tight junction expression, and the conversion to a fibroblastic morphology (Fig. 2). To identify signaling pathways that may be regulated in the presence of DNIIR expression subsequent studies were performed in the presence of 5 M ponasterone. The results in Fig. 3 suggested that expression levels of DNIIR sufficient to desensitize the NMuMG cells to TGF-␤-induced Smad2 and AKT activation still permitted TGF-␤ stimulation of p38MAPK phosphorylation. This result was consistent with the results of Chin et al. that showed TGF-␤-mediated p38MAPK phosphorylation in cells expressing another cytoplasmically truncated construct of T␤RII (30). Their expression of the mutant T␤RII construct did not inhibit TGF-␤ induction of p38MAPK-mediated pro-␣ (1) collagen I mRNA expression in mesangial cells (30). However, since DNIIR expression was not able to rescue TGF-␤-mediated p38MAPK phosphorylation in T47D cells, we speculate that a fraction of the receptor tetramer-ligand complex on the cell surface of NMuMG cells may include a combination of DNIIR as well as endogenous type I and type II receptors. Thus if sufficient over-expression of DNIIR was achieved in NMuMG cells one could anticipate the complete competition of T␤RII and the loss of p38MAPK signaling.
We demonstrated that p38MAPK activation by TGF-␤ is dependent on integrin ␤ 1 adhesion to ECM by examining TGF-␤ signaling in the presence of integrin ␤ 1 -neutralizing antibody. We also found that, integrin ␤ 1 -dependent adhesion was necessary for TGF-␤-mediated p38MAPK activation in both DNIIR-expressing and -non-expressing NMuMG cells. Furthermore, our in vitro kinase results (Fig. 6) lend supportive evidence to the apparent cooperative regulation of p38MAPK by TGF-␤ and integrin ␤ 1 . TGF-␤ signaling to p38MAPK is thought to be mediated through the coupling of TAK1 and TAB for the subsequent recruitment of MKK3 and MKK6 and activation of p38MAPK (31,32). Analogously the transient integrin-mediated activation of p38MAPK upon cell attachment on collagen I is mediated by Rac1-stimulated PAK association with MKK3 and MKK6 (19,25). A potential mechanism for the cooperation of integrins and activated TGF-␤ receptor complexes may be their co-localization in caveolar vesicles (33,34,35). Another possibility is that integrin-dependent adhesion to ECM regulates scaffolding proteins that bring the signaling intermediaries together.
The microscopy results in Fig. 8 illustrate that blocking either integrin ␤ 1 stimulation or inhibiting p38MAPK kinase activity can abrogate TGF-␤-mediated EMT. Blocking integrin ␤ 1 caused a phenotypic change independent of TGF-␤ that is possibly unrelated to p38MAPK activity. This was evident since the integrin ␤-neutralizing antibody alone resulted in the loss of ZO-1 junctional localization. There is recent evidence showing integrin ␤ 1 co-localization together with apical tight junction-associated ZO-1 and occludin proteins (36). The disruption of this association may explain our result. Alternatively, the neutralizing antibody may cause a loss of cell polarity. E-cadherin, an epithelial marker localized at adherence junctions, was not affected by FIG. 8. Adhesion-and p38MAPK-dependent TGF-␤-mediated EMT. NMuMG cells, cultured on collagen-coated coverslips, were incubated in serum-free media (ø) containing anti-integrin ␤ 1 antibody or preinfected with a DN-p38MAPK-␤ adenovirus (DN-p38) prior to treatment with or without TGF-␤ (200 pM for 16 h). The cells were then fixed and stained for ZO-1 and E-cadherin. Corresponding phase contrast images are shown with F-actin staining as insets. The p38MAPK expression and phosphorylation in TGF-␤-treated NMuMG cells control (Ϫ) or DN-p38-infected (ϩ) was determined by immunoblotting using ␤-tubulin as a loading control.
blocking integrin-␤ 1 adhesion. Although the role of integrins in the maintenance of cell-cell junctions is not entirely clear and further complicates the interpretation of these results, it is evident that p38MAPK activity and integrin adhesion are necessary for TGF-␤-induced EMT.
In conclusion we have presented data indicating that TGF-␤ can stimulate p38MAPK activation at a lower functional T␤RII threshold than the activation of Smad2 and AKT. Furthermore, TGF-␤-dependent p38MAPK activation is dependent on integrin ␤ 1 -ECM interactions and TGF-␤-induced epithelial cell EMT is both p38MAPK activation-and integrin-␤ 1 -mediated. Taken together these data suggest that functional T␤RII dosage as well as cooperative signaling between the TGF-␤ receptor and integrins may influence the phenotypic fate of epithelial cells in vivo.