Transforming Growth Factor β1 Induces αvβ3 Integrin Expression in Human Lung Fibroblasts via a β3 Integrin-, c-Src-, and p38 MAPK-dependent Pathway*

In response to transforming growth factor β1 (TGFβ) stimulation, fibroblasts modify their integrin repertoire and adhesive capabilities to certain extracellular matrix proteins. Although TGFβ has been shown to increase the expression of specific αv integrins, the mechanisms underlying this are unknown. In this study we demonstrate that TGFβ1 increased both β3 integrin subunit mRNA and protein levels as well as surface expression of αvβ3 in human lung fibroblasts. TGFβ1-induced αvβ3 expression was strongly adhesion-dependent and associated with increased focal adhesion kinase and c-Src kinase phosphorylation. Inhibition of β3 integrin activation by the Arg-Gly-Asp tripeptide motif-specific disintegrin echistatin or αvβ3 blocking antibody prevented the increase in β3 but not β5 integrin expression. In addition, echistatin inhibited TGFβ1-induced p38 MAPK but not Smad3 activation. Furthermore, inhibition of the Src family kinases, but not focal adhesion kinase, completely abrogated TGFβ1-induced expression of αvβ3 and p38 MAPK phosphorylation but not β5 integrin expression and Smad3 activation. The TGFβ1-induced αvβ3 expression was blocked by pharmacologic and genetic inhibition of p38 MAPK- but not Smad2/3-, Sp1-, ERK-, phosphatidylinositol 3-kinase, and NF-κB-dependent pathways. Our results demonstrate that TGFβ1 induces αvβ3 integrin expression via a β3 integrin-, c-Src-, and p38 MAPK-dependent pathway. These data identify a novel mechanism for TGFβ1 signaling in human lung fibroblasts by which they may contribute to normal and pathological wound healing.

In response to transforming growth factor ␤1 (TGF␤) stimulation, fibroblasts modify their integrin repertoire and adhesive capabilities to certain extracellular matrix proteins. Although TGF␤ has been shown to increase the expression of specific ␣v integrins, the mechanisms underlying this are unknown. In this study we demonstrate that TGF␤1 increased both ␤3 integrin subunit mRNA and protein levels as well as surface expression of ␣v␤3 in human lung fibroblasts. TGF␤1-induced ␣v␤3 expression was strongly adhesion-dependent and associated with increased focal adhesion kinase and c-Src kinase phosphorylation. Inhibition of ␤3 integrin activation by the Arg-Gly-Asp tripeptide motif-specific disintegrin echistatin or ␣v␤3 blocking antibody prevented the increase in ␤3 but not ␤5 integrin expression. In addition, echistatin inhibited TGF␤1-induced p38 MAPK but not Smad3 activation. Furthermore, inhibition of the Src family kinases, but not focal adhesion kinase, completely abrogated TGF␤1-induced expression of ␣v␤3 and p38 MAPK phosphorylation but not ␤5 integrin expression and Smad3 activation. The TGF␤1-induced ␣v␤3 expression was blocked by pharmacologic and genetic inhibition of p38 MAPKbut not Smad2/3-, Sp1-, ERK-, phosphatidylinositol 3-kinase, and NF-B-dependent pathways. Our results demonstrate that TGF␤1 induces ␣v␤3 integrin expression via a ␤3 integrin-, c-Src-, and p38 MAPK-dependent pathway. These data identify a novel mechanism for TGF␤1 signaling in human lung fibroblasts by which they may contribute to normal and pathological wound healing.
One of the key events in wound repair is the infiltration of fibroblasts from surrounding tissue to the extracellular matrix (ECM) 2 in which they proliferate and differentiate into myofibroblasts. Under normal conditions myofibroblasts play a crucial role in ECM deposition and subsequent wound contraction and then disappear as the fibrotic response diminishes and normal structure and function are achieved (1). However, their retention, uncontrolled proliferation, and excessive synthesis of ECM proteins represents a pathologic process that ultimately results in fibrosis (2). Both fibroblast proliferation and differentiation, as well as ECM protein synthesis, are profoundly influenced by growth factors such as TGF␤ as well as cell adhesion (3)(4)(5)(6). Adhesion of cells to ECM is mediated by a family of transmembrane proteins known as integrins that are expressed on the cell surface as ␣/␤ heterodimers (7,8). Importantly, integrins not only support cell attachment but also act in concert with receptors for several growth factors, including TGF␤, to regulate survival, migration, proliferation, and differentiation of fibroblastic, epithelial, and endothelial cells (reviewed in Refs. 7,8). Over the past few years a close relationship between ␣v integrins (recognizing RGD motif) and TGF␤ signaling pathways has been identified (9). These include activation of latent TGF␤ complexes by ␣v␤6 and ␣v␤8 integrins in airway epithelium (8,10), augmented TGF␤ signaling by ␣v␤3 and ␣v␤5 integrins in scleroderma fibroblasts (11), and TGF␤ receptor type II (TGF␤RII)-␣v␤3 integrin interaction-dependent proliferation and differentiation of human lung fibroblasts (4). Previous studies have also identified several molecules as inducers of ␣v integrin expression in various tissue culture systems (12)(13)(14). It was also reported that growth factors are able to activate integrins and that this activation provided an additional mechanism for a growth factor to induce a broad spectrum of cellular responses (15)(16)(17). Recently we demonstrated that TGF␤1 not only synergistically interacts with ␣v␤3 integrins but also induces their gene transcription in human * This work was supported by the British Columbia Lung Association, the Canadian Institutes of Health Research, The Wolfe and Gita Churg Foundation, and the National Health and Medical Research Council of Australia. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article ( lung fibroblasts (4). However, the mechanisms involved in this process are still unknown. Therefore, this study was undertaken to delineate the signaling mechanisms that mediate ␣v␤3 up-regulation in response to TGF␤1 stimulation. The results of this study demonstrate that TGF␤1-dependent induction of ␤3 integrin expression does not involve Smad2/3 or Sp1 transcription factors, but it is mediated by selective and specific activation of the integrin itself and c-Src and p38 MAPK pathways.
Cell Culture-Normal human diploid lung fibroblasts (HFL-1) were obtained from the American Type Culture Collection (Manassas, VA). Cells were cultured at 37°C and 5% CO 2 in F-12K Nutrient Mixture medium (F-12K) supplemented with 10% FBS (Invitrogen) and antibiotics before they reached 90% confluence. Prior to the experiments cells were washed extensively in PBS and quiesced in serum-free F-12K for 24 h.
Flow Cytometry-HFL-1 cells were cultured to 90% confluence and stimulated with recombinant human TGF␤1 (Pepro-Tech, Rocky Hill, NJ) or epidermal growth factor (EGF) (Sigma) in serum-free conditions for the indicated time periods. For experiments examining signaling pathways, cells were pretreated with pharmacological inhibitors of specific signaling molecules for 40 min prior to the addition of TGF␤1. Adherent cells were collected after trypsin/EDTA immersion, and surface expression of integrins was allowed to recover for 30 min in PBS with 10% FBS at room temperature. This step also served to block nonspecific antibody binding. Cells were then fixed with 2% paraformaldehyde for 10 min on ice. Cell surface expression of ␣v␤3 integrin was analyzed using an anti-human ␣v␤3 integrin antibody (mouse IgG 1 , clone LM609, Chemicon, Temecula, CA) and normal mouse IgG 1 (Santa Cruz Biotechnology, Santa Cruz, CA) as an isotype control, followed by incubation with phycoerythrin-conjugated goat anti-mouse F(abЈ) 2 (Cedarlane Laboratories, Ontario, Canada). Cell-associated fluorescence was acquired by a Coulter EPICS XL flow cytometer (Beckman Coulter, Ontario, Canada) and analyzed using Win-MDI 2.8 software.
RNA Extraction and Real Time Reverse Transcription-PCR-See on-line supplemental material for details.
Transfection-HFL-1 were seeded at 1 ϫ 10 5 cells per well in 6-well tissue culture plates for 24 h prior to transfection. The cells were transiently transfected with 2 g of mouse Smad7 DNA using Lipofectamine Plus (Invitrogen) for 24 h according to the manufacturer's instructions. Confirmation of function was determined by Western blot (WB) analysis of ␣-smooth muscle actin expression in TGF␤1-stimulated cells (see supplemental Fig. 1SA). Human Smad3-specific chimera-RNA interference (Smad3 siRNA) (Abnova Corp, Taipei City, Taiwan) and nonsilencing control siRNA (Qiagen, Ontario, Canada) were transfected into HFL-1 cells by using HiPerFect transfection reagent (Qiagen) as instructed by the manufacturer. Forty eight hours after transfection cells were washed with fresh culture medium and further stimulated with TGF␤1 for 24 h. siRNA transfection efficiency was determined by WB (supplemental Fig. 1SB). The mammalian expression plasmids, pcDNA3-p38-KM, pcDNA3 (empty vector control), and pcDNA3-GFP (transfection efficiency control) were transfected into lung fibroblasts using FuGENE 6 transfection reagent (Roche Applied Science) according to the manufacturer's instructions. The optimal ratio of FuGENE 6 (l) to DNA (g) was determined to be 6:1 for HFL-1 cells (supplemental Fig.  2SA). Expression of plasmids was monitored by WB for p38 MAPK expression and by fluorescence microscopy for GFP (data not shown). Following transfection, cells were incubated with 10% FBS in media for 24 h and then with 1% FBS for the next 24 h. Medium was changed, and cells were cultured for an additional 18 h in the presence or absence of TGF␤1. Replication-defective adenoviruses encoding GFP-FRNK fusion protein and GFP alone were amplified and purified using HEK-293 cells as described (18). Preliminary experiments determined that a concentration of 50 -100 particles of Adv-GFP-FRNK and Adv-GFP per cell strongly induced the expression of these proteins (supplemental Fig. 2SB) and infected virtually every fibroblast (ϳ90% of GFP positive cells by flow cytometry) after 48 h of exposure (data not shown). After infection with Adv-GFP-FRNK and Adv-GFP, cells were cultured in the presence of serum for 24 h, then serum-starved for additional 24 h, and stimulated with TGF␤1 for the indicated time periods.
Western Blotting-After TGF␤1 stimulation, control or cell signaling inhibitor-pretreated cell monolayers were lysed in protein extraction buffer with protease and phosphatase inhibitor cocktails (Sigma). Equal concentrations of protein lysates were resolved by SDS-PAGE, transferred to nitrocellulose membranes, and probed with a mixture of primary antibodies. p38 MAPK phosphorylation was determined using mouse mAb against human phosphorylated p38 MAPK (Thr 180 /Tyr 182 ) and rabbit polyclonal antibodies against human p38 MAPK (both from Cell Signaling Technology, Danvers, MA). FAK phosphorylation was determined using mouse mAb against human phosphorylated FAK (pY397) (BD Biosciences) and rabbit polyclonal Ab against C-terminal region of human FAK (pp125 FAK , Sigma). c-Src and Smad3 phosphorylation was determined using rabbit mAb against phospho-Src (Tyr-416) (Cell Signaling Technology), rabbit polyclonal Ab against c-Src (Santa Cruz Biotechnology), and rabbit mAb against phospho-Smad3 (Ser4 23/425 ) (Epitomics, Burlingame, CA), respectively, and anti-␤-tubulin mAb (Upstate Biotechnology Inc., Lake Placid, NY) to control equal protein loading. Expression of ␤3 integrin chain, ␤5 integrin chain, and fibronectin was detected with mouse mAb (BD Biosciences) or rabbit polyclonal Ab (Cell Signaling Technology) against human ␤3 integrin, polyclonal rabbit Ab against human ␤5 integrin (Abcam, Cambridge, MA), and cellular fibronectin (Chemicon), respectively. Detection was performed with IR700 and IR800 anti-mouse and anti-rabbit antibodies (Cell Signaling Technology) and the Odyssey Infrared Imaging System (LI-COR Biotechnology, Lincoln, NE) using the manufacturer's protocol. Density of the bands was analyzed with Odyssey software 1.1 (LI-COR Biotechnology) using two infrared channels independently. The results are expressed as a phosphorylated protein/nonphosphorylated protein density ratio or protein/␤-tubulin density ratio.
Statistical Analysis-Data are expressed as mean Ϯ S.E. of at least three independent experiments. Statistical comparisons were performed using ANOVA with post hoc Fisher's protected least significant difference. Probability values were considered significant if they were less than 0.05. All tests were done using StatView 5.0 software (SAS Institute Inc., Cary, NC).

TGF␤1 Increases ␤3 Protein Expression and Enhances Cell Surface Expression of ␣v␤3 Integrin on Human Lung Fibroblasts-
Recently we showed that TGF␤1 increased ␤3 steady-state mRNA expression (4). Consistent with the increased ␤3 transcription, fibroblasts also increased ␤3 subunit and ␣v␤3 integrin expression on the cell surface after exposure to TGF␤1 at a concentration of 10 ng/ml and higher. As can be seen in Fig. 1A, ␤3 protein levels significantly increased after 24 h of exposure to TGF␤1 at a concentration of 10 ng/ml. As shown in Fig. 1B, surface expression of ␣v␤3 heterodimer was also significantly elevated after cell stimulation with TGF␤1 (60% increase in MFI and 2-fold increase in the percentage of ␣v␤3-positive cells). Increased integrin expression on the cell surface was observed for up to 48 h, although the magnitude of expression was not significantly different from that seen at 24 h (data not shown). In contrast, EGF treatment did not significantly modify ␤3 protein production or surface expression of ␣v␤3 after 24 h of exposure (Fig. 1, A and B). Moreover, when EGF was simultaneously added with TGF␤1, it abrogated both TGF␤1-induced ␤3 subunit and ␣v␤3 cell surface expression (Fig. 1, A  and B). In parallel we examined the effect of TGF␤1 on another ␣v partner, the ␤5 integrin. We found that incubation of HFL-1 with TGF␤1 for 18 h induced robust expression of ␣v␤5 (supplemental Fig. 3S) by increasing ␤5 protein production ( Fig. 1C) similar to the effects on ␣v␤3 expression. In addition, both removal of exogenously added TGF␤1 (washing) and neutralizing of endogenously produced TGF␤ with a pan-TGF␤ blocking antibody (clone 1D11), dramatically attenuated the effect of TGF␤1 on ␤3 and ␤5 integrin expression (Fig. 1C). These results demonstrate that the TGF␤1 effects on ␣v␤3 and ␣v␤5 integrin expression are: 1) associated with up-regulation of corresponding ␤ integrin chain expression, and 2) highly specific and not mediated by a secondary mediator.
TGF␤1-induced Expression of ␣v␤3 Integrin by Human Lung Fibroblasts Does Not Require Smad and Sp1-In our previous studies we have shown that both Smad2 and p38 MAPK were activated in human lung fibroblasts within 10 min of TGF␤1 stimulation, and peak activation was reached at 1 h (19). In pilot experiments we found that both TGF␤1-induced Smad3 and p38 MAPK phosphorylation was still detectable after 18 h and correlated with the increased levels of ␤3 and ␤5 integrins (supplemental Fig. 4S and Fig. 1C).
We used two approaches to evaluate the role of Smad signaling in regulating the effect of TGF␤1 on integrin expression. In the first set of experiments, a murine Smad7-expressing con- The whole cell lysates were resolved by SDS-PAGE, transferred to a nitrocellulose membrane, and probed with anti-␤3 and anti-␤-tubulin antibodies (as shown in representative immunoblot). ␤-Tubulin was used as a loading control and for densitometry normalization in all experiments performed. Protein concentrations were quantitated by scanning densitometry as described under "Experimental Procedures." The bar graph demonstrates an increase of ␤3 protein in cytokine-treated cells over controls (bars represent means Ϯ S.E., n ϭ 3; *, p Ͻ 0.05 compared with control and EGF-treated cells). B, ␣v␤3 cell surface expression. HFL-1 cell were treated with TGF␤1 or EGF (10 ng/ml) or in combination for 24 h following staining with anti-␣v␤3 antibody (LM609, solid line histograms) or normal mouse IgG (isotype control, dashed line histogram) and analyzed by flow cytometry. Histogram plots are a representative experiment of data generated from at least three individual experiments (in parenthesis MFI ratios between the anti-␣v␤3 and isotype control antibodies). Graph bars represent the means Ϯ S.E. percentage of ␣v␤3 positive cells (n ϭ 3; *, p Ͻ 0.05 compared with control and EGF-treated cells). C, effect of TGFneutralizing antibody on ␤3 and ␤5 integrin expression in HFL-1 cells. Fibroblasts were serum-starved overnight, pulse-stimulated with TGF␤1 for 0.5, 1.5, and 3 h, washed with fresh medium, and incubated for a total of 18 h in the presence of either 10 g of pan-TGF␤ neutralizing antibody or normal mouse IgG as a negative control. ␤3 and ␤5 integrins were determined by immunoblotting (one representative immunoblot of two independent experiments is shown). struct was used. Although mouse Smad7 is 95% homologous with the human cDNA, experiments investigating the effect of this construct on TGF␤1-induced ␣-smooth muscle actin expression were performed. TGF␤1-induced increase in ␣-smooth muscle actin expression in fibroblasts was completely abolished by Smad7 overexpression, confirming that the mouse protein is functional in human fibroblasts (supplemental Fig. 1SA). However, transfection of the Smad7 construct did not significantly influence the effect of TGF␤1 on ␣v␤3 expression ( Fig. 2A). To further prove that TGF␤1 effect on ␤3 integrin expression is Smad-independent, fibroblasts were transfected with control or Smad3 siRNA, and the level of ␤3 and ␤5 integrin expression was determined by immunoblotting. The control experiments showed that total and Ser 423/425 -phosphorylated Smad3 protein was knocked down by 60% compared with control (supplemental Fig.  1SB). In keeping with the results from Smad7 overexpression, inhibition of Smad3 did not influence the expression of ␤3 integrin induced by TGF␤1. In contrast, Smad3 siRNA significantly attenuated ␤5 integrin expression induced by TGF␤1 (Fig. 2B).
Cooperation between the Smad proteins and the transcription factor Sp1 may represent a general mechanism for conferring TGF␤1 inducibility of several genes, including integrins (13,20). We sought to determine whether Sp1 was involved in the up-regulation of ␣v␤3 integrin induced by TGF␤1. As shown in Fig. 2C, the Sp1 inhibitor mithramycin had no effect on ␤3 subunit, but it efficiently prevented the increase in fibronectin expression induced by TGF␤1. Together these results indicate that the Smad/Sp1 pathway does not play a major role in TGF␤1-induced up-regulation of ␣v␤3 integrins on human lung fibroblasts.

TGF␤1-induced ␤3 Expression Requires Cell Adhesion and Is
Enhanced by Integrin Activation with ECM Proteins-Our previous studies have demonstrated that several TGF␤-mediated effects in human lung fibroblasts are adhesion-and integrin-dependent (3,4,6). In preliminary experiments we found that TGF␤1 was unable to increase ␣v␤3 integrin expression in cells in suspension but did so in cells adherent on a plastic surface (data not shown). Therefore, we determined whether integrin activation following adhesion to different ECM proteins influences the ability of TGF␤1 to induce ␣v␤3 expression. Fig. 3 demonstrates that cell adhesion to either fibronectin (FN) or collagen did not significantly modify basal ␤3 expression; however, FN strongly potentiated the effect of TGF␤1. These findings suggest that signals mediated by integrin activation following cell adhesion are necessary for TGF␤1 to elevate ␤3 integrin expression.
TGF␤1-induced ␤3 Integrin Expression Is Dependent on ␣v␤3 Activation-To evaluate the role of integrin activation on enhanced ␣v␤3 expression, we treated adherent fibroblasts with the disintegrin echistatin, which has been shown to inhibit the activation of RGD-binding integrins in several cell types (17,21,22). Echistatin had no effect on cell attachment but dose-dependently impaired the ability of fibroblasts to spread  and support monolayer integrity (supplemental Fig. 5S). As shown in Fig. 4A, exposure of cells to echistatin abrogated the effect of TGF␤1 on ␤3, but not on ␤5 integrin expression. To show that echistatin prevents activation of ␣v␤3 and downstream integrin-mediated signaling, we determined levels of Tyr 397 -FAK and Tyr 416 -Src phosphorylation following exposure to the disintegrin. Echistatin dramatically inhibited TGF␤1-induced c-Src kinase (Fig. 4B) and FAK phosphorylation (data not shown) in a concentration-dependent manner.
To demonstrate that echistatin inhibits ␤3 integrin expression by blocking integrin, but not TGF␤1 signaling, FN expression in response to TGF␤1 exposure was determined. In contrast to ␤3, FN expression induced by TGF␤1 was enhanced in the presence of echistatin indicating that TGF␤1 up-regulates ␤3 integrin expression directly by activation of the integrin on the cell surface rather than indirectly through enhanced production of ECM proteins. Echistatin also had no effect on Smad3 activation per se (Fig. 4B), but it inhibited TGF␤1induced p38 MAPK in parallel with ␤3 integrin expression (Fig. 4, A and B).
Because echistatin may also influence other RGD integrins apart from ␣v␤3, we next aimed to confirm the identity of the integrin involved in TGF␤1-induced ␤3 expression. To do this, we incubated fibroblasts with 10 g/ml monoclonal blocking antibodies to human ␣v␤3 integrin (clone LM609) and ␣v␤5 integrin (clone P5H9) 1 h before addition of TGF␤1. Similar to the effect seen with echistatin, LM609, but not P5H9 or control IgG, completely abrogated the effect of TGF␤1 on ␤3 integrin expression (Fig. 4C). These results provide further support for the concept that the effect of TGF␤1 on ␤3 subunit expression is dependent on the activation of ␣v␤3, but not ␣v␤5 integrin, on the cell surface and recruitment of nonreceptor protein tyrosine kinases, including FAK and c-Src kinase.

Adenovirally Mediated Overexpression of FRNK Inhibits TGF␤1-induced FAK Activation but Not ␣v␤3
Expression-Based on the observation that integrin activation and clustering results in activation of FAK, we determined the effect of TGF␤1 on FAK phosphorylation. Fig. 5A demonstrates that TGF␤1 induced time-dependent phosphorylation of FAK on Tyr 397 , its major autophosphorylation site. The enhanced Tyr 397 phosphorylation of FAK was somewhat delayed, being first observed 3 h following TGF␤1 exposure and further increased at later time points (Fig. 5A, p Ͻ 0.05) coincidentally with the induction of ␤3 expression. To confirm whether FAK activation was involved in TGF␤1-induced ␣v␤3 expression, fibroblasts were infected by a replication-defective adenovirus encoding a GFP-FRNK fusion protein (Adv-GFP-FRNK) and exposed to  . Echistatin (2 g/ml) did not change Smad3 phosphorylation but inhibited TGF␤1-induced p38 MAPK activation (right panel, representative immunoblot for two independent experiments). C, ␣v␤3 blocking antibody (clone LM609, PBS solution without preservatives and carrier proteins, 10 g/ml), but not ␣v␤5 blocking antibody (clone P5H9, carrier proteins-free solution) and mouse IgG isotype control applied at the same concentration 1 h before stimulation with TGF␤1, prevents the increase in ␤3 protein induced by cytokine in human lung fibroblasts. Data are means Ϯ S.E. of three independent experiments (*, p Ͻ 0.05 compared with controls and ␣v␤5 blocking antibody). Ctrl, control.
Inhibition of c-Src Kinase Activity Blocks TGF␤1-induced ␣v␤3 Integrin Expression-To confirm the involvement of c-Src kinase in TGF␤1-induced ␣v␤3 integrin expression, we exposed cells to the Src kinase inhibitor, PP2, and measured ␣v␤3 expression following TGF␤1 stimulation. As shown in Fig. 6A, exposure of cells to PP2 completely abrogated the stimulatory effect of TGF␤1 on ␣v␤3 surface expression. Furthermore, incubation with PP2 suppressed the effects of TGF␤1 on ␤3 gene expression over a 6-h time period (Fig. 6B). It has been demonstrated that Src kinases regulate signaling-dependent adhesion as well as FAK-mediated signaling events by regulating FAK kinase activity. As expected, PP2, but not the inactive PP3, almost completely inhibited both Tyr 416 -Src and Tyr 397 -FAK phosphorylation induced by TGF␤1 (data not shown). TGF␤1-induced ␤3 protein increase was also strongly inhibited by PP2 but not PP3 (Fig. 6C).
Recently, it has been demonstrated that PP2 can also inhibit TGF␤ receptor I and II kinase activity (25). Therefore, we also tested the more specific SFK inhibitor SU6656, which does not inhibit TGF␤ receptor function. As shown in Fig. 6C, SU6656 inhibited TGF␤1-induced ␤3 protein in a dose-dependent manner. Moreover, SU6656 inhibited TGF␤1-induced p38 MAPK activation but not Smad3 phosphorylation (Fig. 6D). These results further support a role for autocrine integrin signaling in TGF␤1-induced ␤3 expression and suggest that c-Src kinase and p38 MAPK activation, but not Smad3, is a key element in the process.
p38 MAPK Inhibition Blocks TGF␤1-induced ␣v␤3 Integrin Expression-Besides Smad-mediated transcription, TGF␤1 activates other signaling cascades, including MAPK and NF-B   26). Importantly, Src kinases can interfere with these pathways directly or indirectly, and our results demonstrate that p38 MAPK activation by TGF␤1 is dependent on integrin activation and SFK recruitment. There-fore, we assessed whether inhibition of MAPK, including p38 and p42/ p44, NF-B, and PI3K affected TGF␤1-induced ␣v␤3 integrin expression. As seen in Fig. 7A, no changes in the cell surface expression of ␣v␤3 integrin were observed after treatment with the MEK/ERK inhibitor U0126, the NF-B inhibitor PDTC, and the PI3K inhibitor wortmannin. Similar effects were seen for TGF␤1-induced ␤3 protein (data not shown). However, the p38 MAPK inhibitor SB203580 (10 M), and the more potent inhibitor SB202129 at a concentration of 0.2 M and higher, decreased ␤3 subunit and ␣v␤3 integrin expression induced by TGF␤1 (Fig. 7, A and B). To further validate the role of p38 MAPK, we transiently overexpressed a kinase mutant p38 MAPK (p38-KM) construct in fibroblasts (20). As expected, an increased level of p38 MAPK protein was seen in p38-KMtransfected cells but not in cells transfected with empty pcDNA3 vector (supplemental Fig. 2SA). As seen in Fig. 7C, TGF␤1 failed to induce ␤3 integrin expression in p38-KM-transfected cells. Importantly, the Smad signaling pathway was completely preserved and TGF␤1 efficiently induced ␤5 integrin expression in p38-KM-transfected cells (supplemental Fig. 2SA and Fig. 7C). Induction of p38 MAPK phosphorylation by treatment with recombinant tumor necrosis factor-␣ did not upregulate ␤3 integrin in HFL-1 cells (data not shown). Thus, limiting TGF␤1-induced p38 MAPK activity did not alter Smad-dependent induction of ␤5 integrin, but it substantially inhibited ␤3 protein and ␣v␤3 cell surface expression.

DISCUSSION
Adhesion of fibroblasts to the ECM via specific integrins can alter cell responses to growth factors, including TGF␤ (3-6), which are often produced in excess and may promote specific cell-ECM interactions giving a rise to several pathological cycles and abnormal behavior of the cells in the settings of fibrotic diseases (2,11).
In this study, we demonstrate that TGF␤1 increases ␤3 integrin expression and subsequent ␣v␤3 surface expression on Graph bars are the mean Ϯ S.E. of three independent experiments (*, p Ͻ 0.05 compared with TGF␤1-stimulated cells without inhibitor). C, PP2 and the more specific Src inhibitor SU6656 almost completely abrogated TGF␤1-induced ␤3 integrin expression in human lung fibroblasts. PP3, pharmacologically inactive form of PP2, used at the same concentrations and experimental conditions as PP2 did not influence TGF␤1-induced ␤3 protein (*, p Ͻ 0.05 compared with PP2 and SU6656 inhibitors, n ϭ 3). D, HFL-1 cells treated with the Src inhibitor SU6656 for 45 min and then stimulated with TGF␤1 for additional 45 min. Phosphorylation of p38 MAPK and Smad3 was determined by immunoblotting and quantified by densitometry. The Src inhibitor SU6656 abrogated TGF␤1-induced p38 MAPK phosphorylation but increased phosphorylation of Smad3 in HFL-1 cells (*, p Ͻ 0.05 compared with controls (Ctrl); #, p Ͻ 0.05 compared with TGF␤1 alone, n ϭ 3).
human lung fibroblasts via an autocrine pathway involving c-Src-and p38 MAPK-dependent mechanisms. Significantly, this effect is independent of the conventional Smad/Sp1 signaling pathway. Similarly, although the effect was dependent on cell adhesion and FAK was phosphorylated by TGF␤1, inhibiting FAK activation did not influence the effect of TGF␤1 on ␣v␤3 expression. We show that the effect of TGF␤1 is dependent on ␣v␤3-mediated cell adhesion and activation of c-Src kinase and p38 MAPK pathway. Blockade of c-Src activity also completely abrogated the stimulatory effect of TGF␤1 on activation of p38 MAPK, suggesting that c-Src is upstream of this signaling kinase.
To date there have been several reports investigating the effects of TGF␤1 in modulating expression and function of ␣v␤3 integrins on different cell types. For instance, TGF␤1 has been shown to have no effect on ␣v␤3 expression in human airway epithelial cells (27). In contrast, in human vascular smooth muscle cells (28), human WI-38 fibroblasts (12), and a variety of malignant cell lines, TGF␤1 increases ␣v␤3 expression. Although these studies focused on how the temporal expression of ␣v␤3 integrins is influenced by exposure to TGF␤1, they highlight that the effects of TGF␤1 on integrin expression may be cell type-specific, and little is known about the underlying signal transduction pathways involved.
Binding of TGF␤1 to TGF␤IIR causes the recruitment and activation of TGF␤IR and subsequently Smad2/3 phosphorylation, which then associates with Smad4. This complex then translocates to the nucleus where it modulates transcription of a large number of genes. In contrast, Smad7 inhibits the TGF␤1-mediated phosphorylation of Smad2/3 through competition for binding to the TGF␤1 receptor (26,29). Using several independent approaches, we determined that the Smad2/3 pathway is not required for the TGF␤ stimulation of ␤3 integrin. Specifically, we have shown that in contrast to ␤5 expression, which was strongly dependent on Smad3 activation, TGF␤1-induced ␤3 integrin was maintained in the cells transfected with either Smad3 siRNA or a dominant Smad7 expression construct. Interestingly, in silico analysis of the mouse and human ␤3 integrin gene promoters (30,31) (GenBank TM accession numbers AF026510 and AF02055, respectively) revealed considerable sequence homology across a 1.3-kb region upstream of the transcription start site and several conserved binding elements for Sp1 but not for Smad proteins. Recently, it has been demonstrated that cooperation between Sp1 and Smad proteins may represent a general regulatory mechanism for conferring the TGF␤1 inducibility of several genes, including ␤5, ␣5, and ␣11 integrins (13,20,32) and different types of collagen (33,34). In these studies the antibiotic mithramycin A selectively and efficiently reduced TGF␤-induced collagen and ␣11 integrin expression in mesenchymal cells via inhibition of Sp1 binding to the gene promoter. In this study, we also demonstrated that mithramycin A significantly inhibited TGF␤1induced expression of cellular FN, but it had no effect on ␤3 integrin expression. These paradoxical observations indicate that the effects of TGF␤1 on integrin expression are dependent on cell type and specific regulatory elements of the individual integrin gene promoters. Significantly, our data also suggest that TGF␤1 is able to activate signaling pathways in human lung fibroblasts to promote the expression of integrin complexes independently of Smad2/3 and Sp1.
In addition to Smad-mediated transcription, it has been shown that TGF␤ also activates several specific signaling cascades, including MAPK, NF-B, and PI3K (19,35,36). However, in this study pharmacological inhibition of NF-B, MEK-ERK, and PI3K pathways did not significantly influence the induction of ␣v␤3 by TGF␤1. In contrast, blocking of p38 MAPK activation by either selective inhibitors or overexpres- sion of a dominant p38-kinase dead mutant significantly attenuated TGF␤1-induced ␤3 transcription and ␣v␤3 surface expression. Furthermore, we found that p38 MAPK phosphorylation induced by TGF␤1 also could be inhibited by blocking of ␣v␤3-mediated c-Src activation. During the submission process of this manuscript, Galliher and Schiemann (37) showed that c-Src phosphorylates Tyr 284 on the TGF␤RII receptor and by that it regulates TGF␤ induction of p38 MAPK signaling pathway in malignant mammary epithelial cells. It would be of interest to determine whether the same mechanisms operate in primary human fibroblasts or, alternatively, if there are any defects of TGF␤-induced p38 MAPK phosphorylation in ␤3 Ϫ/Ϫ mouse fibroblasts. Nevertheless, our study provides evidence for the existence of an alternative c-Src-and p38 MAPK-dependent, Smad2/3-independent signaling pathway induced by TGF␤ in human lung fibroblasts that may operate during a fibrotic process in the lung (Fig. 8).
Several studies have demonstrated physical or functional relationships between integrins and growth factor receptors (reviewed in Refs. 7,38). Previously we demonstrated that TGF␤IIR functionally interacts with ␣v␤3, but not with ␣v␤5 integrins in human lung fibroblasts, and this interaction results in substantial enhancement in TGF␤1-induced proliferation (4). However, whether TGF␤1 alters the affinity or activity of ␣v␤3 integrins and their signal transduction pathways is unknown. It is likely that the activation of growth factor receptors induced by their specific ligands has an important function in organizing the signaling pathways associated with the attachment of integrins to ECM. For example, vascular endothelial growth factor is able to enhance the affinity of ␣v␤1 and ␣v␤3 integrins in human endothelial cells (16). In addition, platelet-derived growth factor or EGF-induced activation of FAK through integrins was proposed as an important proximal link between growth factor receptor and integrin signaling pathways (24). In agreement with these findings, we showed that TGF␤1 significantly increases affinity of ␣v integrins on lung fibroblasts to their respective ligands and also activates FAK.
Traditionally FAK activation has been associated with integrin signaling, because it is recruited and phosphorylated at sites of focal adhesions upon integrin aggregation, and it serves as a bridge between growth factor receptors and ␤ integrin cytoplasmic tails (24). In this study we found that TGF␤1 induced FAK autophosphorylation on Tyr 397 (Tyr(P) 397 -FAK), and FAK protein paralleled the time course of ␤3 integrin mRNA expression (4). Critically, TGF␤1 did not influence ␣v␤3 expression on fibroblasts in suspension, but adhesion of cells on ECM proteins significantly enhanced the effect of TGF␤1 on ␤3 integrin expression. However, although overexpression of the FAK inhibitor, FRNK, completely inhibited basal and TGF␤1-induced Tyr(P) 397 -FAK, it did not significantly influence TGF␤1induced ␣v␤3 expression. In contrast, inhibition of c-Src kinase, which also has been shown to play an important role in downstream signaling of integrin receptors, dramatically blunted TGF␤1-induced ␣v␤3 expression and Tyr(P) 397 -FAK.
Although c-Src kinase activity is clearly required for rapid and full phosphorylation of Tyr(P) 397 -FAK (39), c-Src kinase activity is not affected in FAK Ϫ/Ϫ cells (40), suggesting that c-Src can be activated independently of FAK. Furthermore, activated Src, unlike FAK, is concentrated in perinuclear and plasma membrane but not in focal adhesions after integrin ligation (39), suggesting some divergence in their signaling. These findings corroborate our results showing that activation of c-Src kinase, but not FAK autophosphorylation, is the critical downstream pathway in TGF␤1-induced ␣v␤3 expression.
Recently, it has been shown that Src associates with ␣v␤3 and is activated following integrin ligation (39,41). Moreover, the cytoplasmic tail of the ␤3 integrin, but not other integrins, directly interacts with Src kinase through the ␤3-specific domain (39,41,42). These findings may explain the specificity of ␣v␤3 integrins for TGF␤1-induced ␤3 subunit expression in human lung fibroblasts. Indeed, our data demonstrate that FIGURE 8. Schematic model of the proposed synergistic interaction between ␣v␤3-mediated c-Src and TGF␤1-induced p38 MAPK activation in the ␤3 subunit expression by human lung fibroblasts. ␣v␤3 is activated upon interaction with the TGF␤ type II receptors and its ligand TGF␤1 (4,37). Following ␣v␤3 activation c-Src, associated with the ␤3 integrin subunit, is phosphorylated at Tyr 416 and then recruited to the kinase domain of TGF␤RII, which results in phosphorylation of the receptor at Tyr 284 (37). In addition, ␣v␤3 binding to ECM proteins also promotes the activation of SFK. Once activated c-Src coordinates the activation of TGF␤1-induced p38 MAPK signaling pathway and thereby the transcriptional regulation of ␤3 integrin subunit expression. We propose the rapid increase in ␤3 integrin leads to high ␣v␤3 cell surface expression giving rise to a positive feedback loop affecting several fibroblast functions such as adhesion, migration, ECM remodeling, and autocrine TGF␤ expression and signaling (zigzag indicates phosphorylation sites). either disintegrin or monoclonal antibody-mediated blockade of ␣v␤3 activation (but not ␣v␤5), prior to TGF␤1 stimulation, efficiently abrogated the increases of both c-Src activation and ␤3 integrin expression.
How then does TGF␤1 activate or promote activation and signaling of ␣v␤3 integrins? Our data suggest through a process of inside-out signaling whereby activated TGF␤RI or TGF␤RII directly interacts with the ␤3 cytoplasmic domain, which in turn induces a conformational change in the integrin resulting in activation of the downstream SFK (4,37). A similar process of inside-out activation of ␣v␤3 integrin by ligation of insulin-like growth factor I receptor in the murine preadipocyte 3T3-L1 cell line has been described recently (17).
TGF␤1 could also induce synthesis of several ␣v␤3 ligands, including FN, vitronectin, and tenascin (3,34,43), and by enhancement of ligand-integrin interactions activate integrinmediated signaling. Indeed, we demonstrated that seeding of fibroblasts on FN significantly enhanced the effect of TGF␤1 on ␤3 subunit expression. However, these exogenously added ECM proteins did not induce or up-regulate ␤3 expression in the absence of TGF␤1. Furthermore, inhibition of FN production by mithramycin did not prevent TGF␤1-induced expression of the ␤3 integrin.
Finally, TGF␤1-induced activation of ␣v␤3 integrin could be modified by the expression level of the ␣v subunit. We showed that TGF␤1-induced expression of both ␤3 and ␤5 subunit is sufficient to up-regulate the cell surface of ␣v␤3 and ␣v␤5, respectively. This is consistent with the notion that the ␣v subunit is constitutively and excessively expressed in the cytoplasm of fibroblasts as a monomer, and the cell surface expression level of ␣v-containing integrins is controlled by the levels of ␤ subunits, which are exclusively involved in the integrin-mediated signaling mechanisms (11,14).
Given that TGF␤1 up-regulates ␤3 expression in a delayed manner, another possibility is the involvement of other factors induced by TGF␤1. EGF is a well known and potent activator of integrins (16,24,36). However, in this study, although EGF induced a robust FAK autophosphorylation it did not influence ␣v␤3 expression, either alone or in combination with TGF␤1. Moreover, TGF␤ neutralization with blocking mAb and removal of recombinant cytokine after a short time exposure almost completely abrogated ␤3 integrin expression suggesting that activation of integrin/c-Src signaling as well as production of autocrine TGF␤ is exclusively dependent on the initial TGF␤R ligation by active TGF␤1. Altogether these data also indicate that the effect on ␣v␤3 expression is highly specific for TGF␤1.
In summary, our results suggest a model where TGF␤1 induces activation of ␣v␤3 integrin, which in turn leads to c-Src phosphorylation and the initiation of p38 MAPK signaling cascade that is essential for ␤3 chain expression (Fig. 8). Our study for the first time describes a mechanism for the regulation of the integrin ␤3 subunit expression in human fibroblasts, which involves TGF␤1 and the ␣v␤3 integrin itself. Significantly, this appears to be a Smad-independent process. Our data support the concept of a functional synergy between TGF␤1-dependent signaling pathways and ␣v␤3-mediated adhesion processes in normal wound healing and pathological phenomena such as fibrosis.