The Phosphatidylinositol 3-Kinase/Akt Pathway Enhances Smad3-stimulated Mesangial Cell Collagen I Expression in Response to Transforming Growth Factor-β1*

Transforming growth factor (TGF)-β has been associated with renal glomerular matrix accumulation. We previously showed that Smad3 promotes COL1A2 gene activation by TGF-β1 in human glomerular mesangial cells. Here, we report that the PI3K/Akt pathway also plays a role in TGF-β1-increased collagen I expression. TGF-β1 stimulates the activity of phosphoinositide-dependent kinase (PDK)-1, a downstream target of PI3K, starting at 1 min. Akt, a kinase downstream of PDK-1, is phosphorylated and concentrates in the membrane fraction within 5 min of TGF-β1 treatment. The PI3K inhibitor LY294002 decreases TGF-β1-stimulated α1(I) and α2(I) collagen mRNA expression. Similarly, LY294002 or an Akt dominant negative construct blocks TGF-β1 induction of COL1A2 promoter activity. However, PI3K stimulation alone is not sufficient to increase collagen I expression, since neither a constitutively active p110 PI3K construct nor PDGF, which induces Akt phosphorylation, is able to stimulate COL1A2 promoter activity or mRNA expression, respectively. LY294002 inhibits stimulation of COL1A2 promoter activity by Smad3. In a Gal4-LUC assay system, blockade of the PI3K pathway significantly decreases TGF-β1-induced transcriptional activity of Gal4-Smad3. Activity of SBE-LUC, a Smad3/4-responsive construct, is stimulated by over-expression of Smad3 or Smad3D, in which the three C-terminal serine phospho-acceptor residues are mutated. This induction is blocked by LY294002, suggesting that inhibition of the PI3K pathway decreases Smad3 transcriptional activity independently of C-terminal serine phosphorylation. However, TGF-β1-induced total serine phosphorylation of Smad3 is decreased by LY294002, suggesting that Smad3 is phosphorylated by the PI3K pathway at serine residues other than the direct TGF-β receptor I target site. Thus, although the PI3K-PDK1-Akt pathway alone is insufficient to stimulate COL1A2 gene transcription, its activation by TGF-β1 enhances Smad3 transcriptional activity leading to increased collagen I expression in human mesangial cells. This cross-talk between the Smad and PI3K pathways likely contributes to TGF-β1 induction of glomerular scarring.

Transforming growth factor (TGF)-␤ has been associated with renal glomerular matrix accumulation. We previously showed that Smad3 promotes COL1A2 gene activation by TGF-␤1 in human glomerular mesangial cells. Here, we report that the PI3K/Akt pathway also plays a role in TGF-␤1-increased collagen I expression. TGF-␤1 stimulates the activity of phosphoinositide-dependent kinase (PDK)-1, a downstream target of PI3K, starting at 1 min. Akt, a kinase downstream of PDK-1, is phosphorylated and concentrates in the membrane fraction within 5 min of TGF-␤1 treatment. The PI3K inhibitor LY294002 decreases TGF-␤1-stimulated ␣1(I) and ␣2(I) collagen mRNA expression. Similarly, LY294002 or an Akt dominant negative construct blocks TGF-␤1 induction of COL1A2 promoter activity. However, PI3K stimulation alone is not sufficient to increase collagen I expression, since neither a constitutively active p110 PI3K construct nor PDGF, which induces Akt phosphorylation, is able to stimulate COL1A2 promoter activity or mRNA expression, respectively. LY294002 inhibits stimulation of COL1A2 promoter activity by Smad3. In a Gal4-LUC assay system, blockade of the PI3K pathway significantly decreases TGF-␤1-induced transcriptional activity of Gal4-Smad3. Activity of SBE-LUC, a Smad3/4responsive construct, is stimulated by over-expression of Smad3 or Smad3D, in which the three C-terminal serine phospho-acceptor residues are mutated. This induction is blocked by LY294002, suggesting that inhibition of the PI3K pathway decreases Smad3 transcriptional activity independently of C-terminal serine phosphorylation. However, TGF-␤1-induced total serine phosphorylation of Smad3 is decreased by LY294002, suggesting that Smad3 is phosphorylated by the PI3K pathway at serine residues other than the direct TGF-␤ receptor I target site. Thus, although the PI3K-PDK1-Akt pathway alone is insufficient to stimulate COL1A2 gene transcription, its activation by TGF-␤1 enhances Smad3 transcriptional activity leading to increased collagen I expression in human mesangial cells. This crosstalk between the Smad and PI3K pathways likely contributes to TGF-␤1 induction of glomerular scarring.
Transforming growth factor (TGF) 1 -␤ is a pleiotropic cytokine involved in activities such as differentiation, growth, apoptosis, inflammation, tissue remodeling, and wound healing. A number of studies indicate that TGF-␤ plays a critical role in renal matrix accumulation. TGF-␤ has been linked to fibrogenesis in experimental models of glomerulonephritis and diabetic nephropathy (1,2). Its expression is elevated in human glomeruli under conditions associated with increased extracellular matrix (ECM) deposition such as diabetic nephropathy and focal segmental glomerulosclerosis (2).
Members of the TGF-␤ superfamily signal via heteromeric complexes of transmembrane serine/threonine kinases, the type I and type II receptors. The Smad proteins function downstream from the TGF-␤ family receptors to transduce signal to the nucleus (3)(4)(5). Upon ligand binding, the type II receptor recruits and phosphorylates type I receptor (T␤RI). The receptor-regulated or pathway-restricted Smads (R-Smads) contain an SSXS phosphorylation site at their C-terminal end that is a direct target of T␤RI. Once phosphorylated, the R-Smads associate with the common-partner Smad, Smad4. The resulting heteromultimer translocates to the nucleus where it regulates expression of TGF-␤ target genes by direct binding to DNA and/or interaction with other transcription factors (3)(4)(5).
A few studies have suggested that the PI3K signaling path- way could be modulated by members of the TGF-␤ family (10 -13). In human airway smooth muscle cells, the p85 regulatory subunit of PI3K associates with type I and type II TGF-␤ receptors and TGF-␤1 enhances activation of PI3K by epidermal growth factor (14). More recently, Bakin et al. (11) showed that LY294002, an inhibitor of PI3K, blocks TGF-␤1-induced Smad2 phosphorylation in breast cancer cells, suggesting that Smad proteins are potential targets of the PI3K pathway. However, the role of PI3K in fibrosis in response to TGF-␤1 has not been investigated.
We previously showed that TGF-␤1-induced collagen I gene expression is Smad3-dependent in human mesangial cells (15,16). Here, we demonstrate that TGF-␤1 activates PDK-1 and Akt, two downstream targets of PI3K, in these cells. Inhibition of the PI3K pathway with LY294002 or an Akt dominantnegative construct abrogates TGF-␤1-stimulated COL1A2 gene transcription. Furthermore, blockade of the PI3K pathway decreases Smad3 activity in transcriptional assays. These data suggest that cross-talk between Smads and the PI3K pathway regulates collagen I expression in response to TGF-␤1.
Cell Culture-Human mesangial cells were isolated from glomeruli by differential sieving of minced normal human renal cortex obtained from anonymous surgery or autopsy specimens. The cells were grown in Dulbecco's modified Eagle's medium/Ham's F12 medium supplemented with 20% heat-inactivated fetal bovine serum (FBS), glutamine, penicillin/streptomycin, sodium pyruvate, Hepes buffer, and 8 g/ml insulin (Invitrogen Life Technologies, Carlsbad, CA) as previously described (17) and were used between passages 5 and 8.
PDK-1 Activity Assay-Cells in medium containing 1% FBS were treated with 1 ng/ml TGF-␤1 for various time periods leading up to simultaneous harvest in radioimmune precipitation assay buffer (50 mM Tris-HCl, pH 7.5; 150 mM NaCl; 1% Nonidet P-40; 0.1% SDS) containing protease inhibitors (Sigma). After clarification by centrifugation, the protein content was determined by Bradford protein assay (BioRad). Cell lysates were incubated with a SGK1 peptide substrate (Upstate, Waltham, MA) in the presence of [␥-32 P]ATP for 10 min at 30°C. The reactions were spotted on phosphocellulose and washed four times with 1% phosphoric acid followed by an acetone rinse. The amount of radioactivity incorporated into the substrate was determined by scintillation counting. Active PDK-1 (Upstate Biotechnologies) was used as a positive control.
Preparation of Cell Lysates, Immunoprecipitation, and Western Blot Analysis-Cells were switched to medium containing 1% FBS and treated with 1 ng/ml TGF-␤1 or 5 ng/ml PDGF for various time periods leading to simultaneous harvest. In some experiments, the cells were preincubated with 10 M LY294002 or Me 2 SO as vehicle control for 1 h prior to TGF-␤1. For the immunoprecipitation experiments, 500 g of lysates were immunoprecipitated with 1 g of anti-Smad2/3 antibody (N-19) followed by precipitation with 40 l of protein G-Sepharose as previously described (16). The resulting immunoprecipitates or whole cell lysates were analyzed by immunoblotting with anti-phosphoserine antibody (1 g/ml); anti-Smad2/3, anti-Smad1/2/3 or anti-Smad4 antibody (0.2 g/ml); anti-phospho-Smad2 antibody (1 g/ml); anti-phospho-Smad3 antibody (2 g/ml); anti-phospho-Akt or anti-Akt antibody (1:1,000). The blots were developed with chemiluminescence reagents according to the manufacturer's protocol (Santa Cruz Biotechnology). Autoradiograms were scanned with an Arcus II Scanner (AGFA) in transparency mode, and densitometric analysis was performed using the NIH Image 1.61 program for Macintosh.
Cell Fractionation-Cells were scraped into a detergent-free buffer (20 mM Tris-HCl, pH 7.5; 0.5 mM EDTA; 0.5 mM EGTA; 10 mM ␤-mercaptoethanol) containing protease and phosphatase inhibitors. The cells were then disrupted by 15 strokes of a Dounce homogenizer. After 45 min centrifugation at 100,000 ϫ g, the supernatant was removed and saved as the S-100 fraction, representing the cytosolic fraction. The pellet, representing the particulate fraction, was resuspended in buffer containing 0.5% Triton X-100 and syringe sheared. After 30 min of incubation on ice, the insoluble material was removed by 10-min centrifugation at 18,000 ϫ g, and the resulting supernatant was saved as the Triton X-100 soluble fraction, representing the membrane-enriched fraction. After determination of the protein content, each fraction was analyzed by immunoblotting with anti-Akt antibody as described above.
Transient Transfection and Luciferase Assay-The day before the transfection, cells were seeded on 6-well plates at 6.5 to 8 ϫ 10 4 cells per well. Eighteen hours later, cells were switched to 1% FBS medium and transfected with the indicated constructs along with CMV-SPORT-␤galactosidase as a control of transfection efficiency. Transfection was performed with the FuGENE 6 transfection reagent (Roche Applied Science, Indianapolis, IN) as previously described (15). After 3 h, 1 ng/ml TGF-␤1 or control vehicle was added to the cells. In some experiments, the transfected cells were pretreated for 1 h with LY294002 before adding TGF-␤1. 18 -24 h later, the cells were harvested in 300 l of reporter lysis buffer (Promega). Luciferase and ␤-galactosidase activities were measured as previously described (15). Luciferase assay results were normalized for ␤-galactosidase activity. Experimental points were performed in triplicate in several independent experiments.
Plasmid Constructs-The 376COL1A2-LUC construct containing the sequence 376 bp of the ␣2(I) collagen (COL1A2) promoter and 58 bp of the transcribed sequence fused to the luciferase (LUC) reporter gene was previously described (15). The SBE-LUC (18) reporter construct was kindly provided by Dr. B. Vogelstein. The vectors expressing the indicated Smad3 variants (19) were kindly provided by Drs. H. F. Lodish and X. Liu. The Gal4-Smad constructs (20) were kindly provided by Dr. M. P. de Caestecker. The constitutively active p110 PI3K construct (p110␣K227E) (21) was kindly provided by Dr. J. Downward. The CMV-SPORT-␤-galactosidase was purchased from Invitrogen, the pFR-LUC from Stratagene (La Jolla, CA) and the Myc-Akt and Akt dominant-negative constructs from Upstate.
RNA Isolation and Northern Blot-Cells were switched to medium containing 1% FBS. They were preincubated with 10 M LY294002 for 1 h before addition of 1 ng/ml TGF-␤1, 5 ng/ml PDGF or control vehicle for 24 h. Total RNA was harvested using TRIzol (Invitrogen Life Technologies) and analyzed by Northern blot as previously described (17). The same blots were successively rehybridized with additional probes after complete stripping. cDNAs for human ␣1(I) (clone Hf677, Ref. were considered significant. The difference between two comparative groups was further analyzed by unpaired Student's t test.

RESULTS
Activation of the PI3K Pathway by TGF-␤1-Only a limited number of studies have examined the role of PI3K in TGF-␤ signaling. None have investigated the potential role of the PI3K pathway in fibrosis induced by TGF-␤1. We used human glomerular mesangial cells treated with TGF-␤1 as our model to study the mechanism of abnormal matrix accumulation by kidney cells. First, we investigated whether the PI3K pathway is activated in response to TGF-␤1 in these cells. We examined whether the activity of PDK-1 (a downstream target of PI3K) is increased by TGF-␤1 treatment. Human mesangial cells were treated with 1 ng/ml TGF-␤1 for different time periods leading up to simultaneous harvest. Lysates were used for an in vitro kinase assay with a SGK1 peptide substrate. Fig. 1 shows that TGF-␤1 stimulates PDK-1 activity beginning at 1 min. The increased activity is sustained for up to 24 h. These data suggest the PI3K pathway is activated by TGF-␤1 in human mesangial cells.
Because one major target of PDK-1 is Akt, we sought to determine whether PDK-1 activation by TGF-␤1 leads to increased Akt activity as measured by membrane translocation and phosphorylation. Cells, treated with TGF-␤1 for various durations, were fractionated into cytosol and membrane fractions for analysis by immunoblot with an anti-Akt antibody. Akt translocates to the membrane-enriched fraction in a biphasic manner following TGF-␤1 treatment. An initial peak of membrane-associated Akt is observed 5 min after adding TGF-␤1 (Fig. 2, A and B). A second peak is detected at 30 min, and membrane association remains elevated at up to 24 h of treatment. Increased membrane association correlates with increased phosphorylation (Fig. 2, C  and D). LY294002, a PI3K inhibitor, blocks Akt membrane translocation (Fig. 2E) and phosphorylation (Fig. 2F). These results indicate that activation of Akt follows TGF-␤1-stimulated PDK-1 activity.

TGF-␤1-induced PI3K Modulates Collagen I Gene Activity-
We previously have shown that TGF-␤1 stimulates ␣2(I) collagen (COL1A2) gene expression in mesangial cells (15). Since the PI3K pathway is activated by TGF-␤1, we investigated whether this pathway modulates COL1A2 promoter activity in response to TGF-␤1. We performed transient transfection experiments with the 376COL1A2-LUC construct (15) and determined the effect of inhibiting the PI3K pathway on its activity. The transfected cells were pretreated for 1 h with LY294002, a specific inhibitor of PI3K. TGF-␤1 was then added for 24 h and luciferase activity was determined. Inhibition of PI3K almost completely blocks TGF-␤1-induced COL1A2 promoter activity (Fig. 3). Similarly, co-transfection of a vector expressing a kinase-deficient (K179M) Akt (shown to act as a dominantnegative Ref. 12) prevents the response to TGF-␤1. Of note, basal promoter activity is also decreased by the dominantnegative construct. These results suggest that the PI3K pathway is necessary for the transcriptional activation of COL1A2 by TGF-␤1.
Activation of PI3K/Akt Is Not Sufficient for COL1A2 Stimulation-We next sought to determine whether increased PI3K activity is sufficient to stimulate collagen I gene expression. Cells were transfected with the 376COL1A2-LUC construct and a vector expressing a constitutively active p110 (p110*) PI3K subunit (21) or the corresponding empty vector. As shown in Fig. 4A, the constitutively active p110 alone does not increase COL1A2 promoter activity. To confirm that the constitutively active p110 construct was active in mesangial cells, lysates from cells co-transfected with Myc-Akt and either p110* or empty vector were immunoprecipitated with an anti-cMyc antibody, and the immunoprecipitated complexes were analyzed by immunoblotting with anti-phospho-Akt. Akt phosphorylation levels are increased by ϳ2 fold in cells transfected with the constitutively active p110 compared with cells transfected with empty vector (data not shown). PDGF has been shown to activate PI3K in various cell types (24 -26). We examined whether the PI3K/Akt pathway was activated by PDGF in human mesangial cells. As expected, PDGF stimulates phosphorylation of Akt at serine 473 in a time-dependent manner. (Fig. 4B). Thus, PDGF is able to activate the PI3K pathway in mesangial cells, in agreement with Ghosh Choudhury et al. (27) who have demonstrated that PDGF stimulates PI3K activity, as determined by increased PIP production.
Next, we compared the effects of PDGF and TGF-␤1 on collagen I transcription. In mesangial cells that were transfected with the 376COL1A2-LUC reporter construct, the collagen I promoter is not stimulated by PDGF but is responsive to TGF-␤1 (Fig. 4A). We also measured collagen I mRNA expression after treatment with PDGF or TGF-␤1. As expected, PDGF is not able to increase ␣ 1 (I) and ␣ 2 (I) collagen mRNA levels (Fig. 4C). In contrast, TGF-␤1 stimulates collagen I mRNA expression as previously reported (17), and the induction is blocked by LY294002 (Fig. 4C). These data further support a role for PI3K in TGF-␤1-increased collagen I transcription while PDGF, another activator of the PI3K/Akt pathway, is not able to stimulate COL1A2 gene expression.
PI3K Modulation of Collagen I Gene Expression Is through Modulation of Smad3 Activity-We previously have shown that Smad3 stimulates COL1A2 gene transcription (15,16). We investigated whether the reason why PDGF could not activate collagen I expression was due to an inability to stimulate Smad3 activity. Cells were treated with TGF-␤1 or PDGF for different durations and Smad3 phosphorylation levels were examined. TGF-␤1 induces increased Smad3 phosphorylation in a time-dependent manner, as previously shown (15), while PDGF does not activate Smad3 (Fig. 5). Thus, although both PDGF and TGF-␤1 are able to stimulate the PI3K pathway, only TGF-␤1 leads to phosphorylation of Smad3 that correlates with increased COL1A2 promoter activity.
Next, we determined whether inhibition of the PI3K pathway would block Smad3-mediated COL1A2 activation. Mesangial cells were co-transfected with the 376COL1A2-LUC construct and the expression vector for Smad3 or the corresponding empty vector. One hour after adding LY294002 or control vehicle, the cells were treated with TGF-␤1 for 24 h. Inhibition of PI3K reduces both ligand-dependent and ligandindependent Smad3-mediated COL1A2 promoter activity (Fig.  6). Together, these data suggest that TGF-␤1-stimulated PI3K pathway activity contributes to Smad3 induction of COL1A2 gene transcription.
To investigate whether TGF-␤1-stimulated PI3K activation modulates Smad3 transactivation activity. Cells were co-transfected with a reporter construct containing five Gal4 binding sites in front of the luciferase gene and a construct expressing the Gal4 DNA binding domain fused to full-length Smad3 (Gal4-Smad3). The cells were pretreated for 1 h with vehicle or LY294002 and then incubated with TGF-␤1 for 24 h. The PI3K inhibitor significantly decreases TGF-␤1-stimulated Smad3 transcriptional activity (Fig. 7A). Similarly, co-transfection of the dominant-negative Akt construct reduces Smad3 activity (Fig. 7B). These data suggest that the PI3K pathway modulates Smad3 transactivation activity, at least in part, independently of Smad3 DNA binding activity. We also evaluated the effect of PI3K inhibition on the activity of the SBE-LUC reporter construct containing four copies of the GTCTAGAC sequence that has been shown to bind recombinant Smad3 and Smad4 (18). Cells were co-transfected with SBE-LUC and a construct expressing wild-type Smad3, or the empty vector (pEXL). Pretreatment for 1 h with LY294002 decreases TGF-␤1-induced activity of endogenous Smad3 by ϳ60% (Fig. 8, pEXL histograms) as well as transcriptional activity of overexpressed Smad3 (Fig. 8, Smad3 histograms). Smad3D, a construct in which the three C-terminal serine residues of Smad3 are replaced by three aspartic acid residues, can function as a transcriptional activator in the absence of TGF-␤1 as previously demonstrated (19,28). In mesangial cells, LY294002 inhibits Smad3D activity (Fig. 8). These results indicate that modulation of Smad3 transcriptional activity by the PI3K pathway is probably independent of phosphorylation at the C-terminal T␤RI target site.
In addition to the directly receptor-regulated C-terminal phosphorylation motif, several reports have demonstrated that other sites in R-Smads can be regulated by phosphorylation in response to signaling pathways such as extracellular signal regulated kinase (ERK), c-Jun N-terminal kinase (JNK), protein kinase C (PKC), and Ca 2ϩ -calmodulin-dependent kinase II (CaMKII) (29 -34). Thus we investigated whether inhibition of the PI3K pathway affects total serine phosphorylation of Smad3 as determined by Western blot of immunoprecipitated Smad2/3 complexes using an anti-phosphoserine antibody. In the presence of LY294002, total serine phosphorylation of both Smad2 and Smad3 is significantly decreased (Fig. 9, A and B). In contrast, phosphorylation at the C-terminal site is not inhibited by LY294002 pretreament as determined by immunoblotting with a phospho-Smad2 or phospho-Smad2/3 antibody specifically recognizing phosphorylation at Ser465/467 or Ser433/435, respectively (Fig. 9, C and D). Following TGF-␤1 stimulation, activated R-Smads form complexes with Smad4 (35)(36)(37). Paralleling its effect on Smad2/3 phosphorylation, inhibition of the PI3K pathway also decreases association with Smad4 without affecting the expression levels of Smad2, Smad3, or Smad4 (Fig. 9, A and B). These data suggest that TGF-␤1-induced PI3K activation leads to phosphorylation of Smad2/3 outside the C-terminal phosphoacceptor site and facilitates Smad4 association with R-Smads. DISCUSSION In the present study, we demonstrate that TGF-␤1 activates the PI3K pathway in human mesangial cells and that modulation of Smad3 transcriptional activity by PI3K, possibly through an effect on phosphorylation, contributes to increased collagen I expression in response to TGF-␤1. The PI3K downstream targets, PDK-1 and Akt, are rapidly activated by TGF-␤1 in human mesangial cells. Increased PDK-1 activity is detected within 1 min of TGF-␤1 treatment and is followed by increased membrane translocation and phosphorylation of Akt. In the NMuMG mammary epithelial cell line, epithelial-tomesenchymal transition induced by TGF-␤1 is preceded by increased Akt phosphorylation (11). In Swiss 3T3 cells, TGF-␤ stimulates PI3K activity and Akt phosphorylation (10). Ghosh-Choudhury et al. (12) have demonstrated activation of the PI3K pathway by bone morphogenetic protein (BMP)-2, another member of the TGF-␤ family. In contrast, Krymskaya et al. (14) found that, in human airway smooth muscle cells, TGF-␤1 alone does not alter PI3K activation but enhances the stimulatory effect of epidermal growth factor. Other reports describe inhibition by TGF-␤1 of cytokine-induced Akt phosphorylation in lymphocyte cell lines (38,39). Thus, activation or inhibition of PI3K signaling by TGF-␤1 is likely to be cell type-specific. Here, we show stimulation of this pathway by TGF-␤1 in kidney mesangial cells. This activation is likely to be direct considering the timing of increased PDK-1 activity and Akt phosphorylation in response to TGF-␤1. Moreover, coimmunoprecipitation between the p85 subunit of PI3K and both T␤RI and T␤RII have been demonstrated in airway smooth muscle cell (14).
The reports suggesting activation of the PI3K pathway by TGF-␤1 have not examined whether TGF-␤1-induced PI3K plays a role in extracellular matrix accumulation. Here we show that inhibition of PI3K with LY294002 decreases TGF-␤1-stimulated ␣1(I) and ␣2(I) collagen mRNA expression. Similarly, LY294002 or a construct expressing an Akt dominant negative mutant blocks COL1A2 promoter activation by TGF-␤1. Although we could not exclude an effect on mRNA stability, our transfection experiments clearly indicate that, in human mesangial cells, the PI3K pathway contributes to TGF-␤1stimulated ␣ 2 (I) collagen expression at the transcriptional level.
Ricupero et al. (40) reported that, in the absence of any stimulus, incubation with LY294002 inhibited basal ␣1(I) collagen mRNA expression in human lung fibroblasts within 2-5 h. In contrast, in our experiments, LY294002 had a minimal effect on basal ␣1(I) and ␣2(I) collagen I mRNA levels. Thus involvement of the PI3K pathway in basal expression of type I collagen could depend on the cell type. Alternatively, PI3K activation could result from autocrine stimulation or continuous activation by the culture conditions. Another possibility is that the timing of inhibition by LY294002 is cell-specific and may reflect a difference in the stability of collagen I mRNA in different cell types. Indeed, Reif et al. (26) have demonstrated that, in hepatic stellate cells, blockade of PI3K with LY294002 leads to a decrease in basal ␣1(I) collagen mRNA levels at a much slower rate (48 -72 h after treatment) than in lung fibroblasts.
In our experiments, LY294002 substantially decreased the response to TGF-␤1-stimulated ␣ 2 (I) collagen mRNA expression and promoter activity, indicating that the PI3K pathway mediates the increase in collagen I expression in response to TGF-␤1, at least in part at the transcriptional level. However, activation of the PI3K pathway alone is not sufficient to increase COL1A2 gene expression since neither a constitutively active p110 PI3K construct nor PDGF is unable to stimulate collagen I promoter activity or mRNA expression, respectively. Smad3 is phosphorylated in response to TGF-␤1 in a time-dependent manner and mediates COL1A2 transcription (this report and Refs. 15 and 16). Although PDGF is able to increase PI3K activity (27) and Akt phosphorylation (this report) in mesangial cells, it does not stimulate Smad3 phosphorylation. Together these results strongly suggest that PI3K can modulate collagen I expression only when the Smad proteins also are activated.
Our data indicate that the PI3K pathway positively modulates Smad3 transcriptional activity. Stimulation of COL1A2 by Smad3 is partially blocked by LY294002. Using the Gal4-LUC assay system or the SBE-LUC reporter construct, we further demonstrate inhibition of Smad3 transactivation activity either by LY294002 or an Akt dominant-negative construct. Bakin et al. (11) showed that Smad2 C-terminal phosphorylation was decreased by LY294002 in NMuMG cells, suggesting that PI3K is involved in TGF-␤1-mediated phosphorylation of Smad2. In contrast, we do not detect any change in Smad2 phosphorylation at the C-terminal site in the presence of LY294002 in mesangial cells, while total serine phosphorylation of Smad2 is decreased. Similarly, C-terminal phosphorylation of Smad3 in response to TGF-␤1 is not affected by inhibition of the PI3K pathway, whereas total serine phosphorylation is partially inhibited. Moreover, the transcriptional activity of Smad3 mutated at its C-terminal serines is still inhibited by LY294002. This further supports the hypothesis that activation of Smad3 by the PI3K pathway is independent of Smad3 phosphorylation at its T␤RI phosphorylation target site. Inhibition of PI3K not only decreases total serine phos- FIG. 9. Blockade of PI3K decreases total serine phosphorylation of Smad2/3 and association with Smad4. Cells were pretreated with 10 M LY294002 for 1 h. After 30 min of incubation with 1 ng/ml TGF-␤1, the cell lysates were harvested and immunoprecipitated with an anti-Smad2/3 antibody. Immunoprecipitated complexes (IP) and cells lysates were electrophoresed and immunoblotted with the indicated antibodies. A, representative experiment showing partial inhibition of Smad2/3 total serine phosphorylation and Smad4 association after PI3K blockade. B, graph shows the ratio between total serine phosphorylation and R-Smad for Smad2 and Smad3; and the ratio between associated Smad4 and total Smad2/3. The ratios in the absence of inhibitor are arbitrarily set up as 1. *, p Ͻ 0.05 (n ϭ 3). C, representative experiment showing no change of C-terminal phosphorylation levels. D, graph shows ratio between C-terminal serine phosphorylation and R-Smad, in the absence (open bars) or presence (shaded bars) of LY294002, after TGF-␤1 treatment. phorylation of Smad2/3 but it also decreases association between Smad2/3 and Smad4, suggesting that PI3K activity contributes to Smad2/3 phosphorylation-dependent association with Smad4. Our data are consistent with work from others who have identified additional sites in the R-Smads whose phosphorylation could be mediated by various pathways such as ERK (29 -31), JNK (32), PKC (33), and CaMKII (34). However, Smad2 and Smad3 have not been described as a target for phosphorylation by the PI3K/Akt pathway, and whether Smad2/3 can be directly phosphorylated by Akt or other downstream kinases of PI3K remains to be determined.
Examples of kinases that could act downstream of PI3K to modulate Smad3 activity include the MEK/ERK pathway and PKC. We previously have shown that ERK1/2 is activated by TGF-␤1 in human mesangial cells (41). Similar to PI3K inhibition, blockade of the MEK/ERK pathway with PD98059 or U0126 leads to decreased TGF-␤1-induced total serine phosphorylation of Smad2/3 without affecting phosphorylation of the C-terminal phosphoacceptor site (31). Since PI3K has been shown to activate ERK in some cells lines (42,43), phosphorylation of Smad2/3 by PI3K in response to TGF-␤1 could be mediated by ERK. On the other hand, it has been shown that PI3K associates with PKC␦ in the erythroleukemia cell line TF-1 upon stimulation by cytokines (44) and that phosphorylation of PKC␦ in response to serum in the human embryonic kidney 293 cells is PI3K-and PDK1-dependent (45). Previously, we have demonstrated that TGF-␤1-stimulated PKC␦ activity in human mesangial cells positively regulates Smad3 transactivation activity (28). Moreover, both ERK and PKC have been shown to phosphorylate R-Smads outside the C-terminal phosphorylation site, enhancing or decreasing Smad activity depending upon cell type and/or stimulus (29 -31, 33). Therefore, these kinases could mediate some of the effects of PI3K on Smad activity. However, the role of ERK or PKC␦ in mediating phosphorylation of Smads after direct activation of the PI3K pathway by TGF-␤1 remains to be investigated.
Recently, it was demonstrated that another member of the TGF-␤ family, BMP-2, stimulates the PI3K pathway during osteoblast differentiation and that a dominant negative Akt construct inhibited nuclear translocation of Smad1/5, the specific BMP R-Smads (12). In contrast, inhibition of the PI3K pathway with LY294002 does not seem to affect Smad2/3 translocation in human mesangial cells, 2 suggesting that modulation of Smads by the PI3K pathway might depend on cell type or the initiating stimulus and the R-Smads that are involved.
In summary, we show that, in human mesangial cells, TGF-␤1 stimulates PDK-1 activity as well as Akt phosphorylation and membrane translocation. The PI3K pathway positively regulates Smad3 transcriptional activity and is required for increased collagen I production, suggesting that activation and interaction of multiple signaling pathways contribute to the pathogenesis of glomerular matrix accumulation.