CCN2 Is Necessary for Adhesive Responses to Transforming Growth Factor-β1 in Embryonic Fibroblasts*

CCN2 is induced by transforming growth factor-β (TGFβ) in fibroblasts and is overexpressed in connective tissue disease. CCN2 has been proposed to be a downstream mediator of TGFβ action in fibroblasts; however, the role of CCN2 in regulating this process unclear. By using embryonic fibroblasts isolated from ccn2–/–mice, we showed that CCN2 is required for a subset of responses to TGFβ. Affymetrix genome-wide expression profiling revealed that 942 transcripts were induced by TGFβ greater than 2-fold in ccn2+/+ fibroblasts, of which 345 were not induced in ccn2–/–fibroblasts, including pro-adhesive and matrix remodeling genes. Whereas TGFβ properly induced a generic Smad3-responsive promoter in ccn2–/–fibroblasts, TGFβ-induced activation of focal adhesion kinase (FAK) and Akt was reduced in ccn2–/–fibroblasts. Emphasizing the importance of FAK and Akt activation in CCN2-dependent transcriptional responses to TGFβ in fibroblasts, CCN2-dependent transcripts were not induced by TGFβ in fak–/–fibroblasts and were reduced by wortmannin in wild-type fibroblasts. Akt1 overexpression in ccn2–/–fibroblasts rescued the TGFβ-induced transcription of CCN2-dependent mRNA. Finally, induction of TGFβ-induced fibroblast adhesion to fibronectin and type I collagen was significantly diminished in ccn2–/–fibroblasts. Thus in embryonic fibroblasts, CCN2 is a necessary cofactor required for TGFβ to activate the adhesive FAK/Akt/phosphatidylinositol 3-kinase cascade, FAK/Akt-dependent genes, and adhesion to matrix.

conserved cysteine-rich residues and a heparin-binding domain and is chemotactic and mitogenic for connective tissue cells (1)(2)(3)(4). However, the physiological role of CCN2 is largely unknown.
As an initial approach to elucidate the physiological function of CCN2, mice deleted for the ccn2 gene were recently generated (5). Mice homozygous for a deletion of the ccn2 gene die soon after birth, displaying an inability of the rib cage to ossify properly (5). The phenotype of these mice is consistent with a role for CCN2 in matrix synthesis and remodeling as ccn2Ϫ/Ϫ embryos show reduction in the expression of bone-specific matrix proteins, such as aggrecan (5). We recently found that embryonic fibroblasts isolated from ccn2Ϫ/Ϫ mice showed reduced basal adhesive signaling, including a reduction of FAK and ERK phosphorylation and delays in ␣-smooth muscle actin (␣-SMA) stress fiber formation (6), suggesting that CCN2 plays a key role in mediating the formation of attachments between the cell and matrix at focal adhesions.
Although CCN2 was discovered over a decade ago, the precise biological function of CCN2 has remained elusive. CCN2 is expressed in mesenchymal cells in development, is induced during wound healing (4,8), and is overexpressed in fibrosis (7)(8)(9)(10)(11). Indeed, an enzyme-linked immunosorbent assay detecting the amino-terminal portion of the CCN2 protein has indicated that the appearance of CCN2 in the blister fluid of scleroderma patients can be considered an excellent surrogate marker for the severity of skin fibrosis (12). One of the most potent inducers of CCN2 is TGF␤, which promotes CCN2 expression in dermal fibroblasts, but not in epidermal keratinocytes, through a complex network of transcriptional interactions requiring Smads, protein kinase C, and Ras/mitogen-activated protein kinase/extracellular signal-regulated kinase kinase/ERK and a consensus transcription enhancer factor binding-element in the CCN2 promoter (11,13). CCN2 protein production in response to TGF␤ does not occur in smad3Ϫ/Ϫ mouse embryonic fibroblasts (MEFs) or in the presence of the mitogen-activated protein kinase/extracellular signal-regulated kinase kinase inhibitor U0126 (10,14).
As CCN2 is induced by TGF␤, it has been hypothesized that CCN2 is a downstream mediator of TGF␤ action (15). Indeed, this notion is the chief operating paradigm in the field. In fact, it is unclear whether phys-iologically CCN2 is a downstream mediator or co-activator of TGF␤ action. Furthermore, the contribution of CCN2 to particular TGF␤ signaling pathways is unknown. Finally, although CCN2 has been proposed to be required for TGF␤ responses in fibroblasts, but not in epithelial cells (15), whether CCN2 is required for all, or a subset, of TGF␤ responses in fibroblasts is completely unknown.
In this study, we take advantage of the fact that CCN2 is expressed by MEFs and is induced by TGF␤ in this cell type (10). Consequently, examining the impact of loss of CCN2 expression by MEFs is likely to be a useful predictive tool to assess the impact of CCN2 in development, but should significantly affect our understanding of CCN2 action in pathological conditions where CCN2 is constitutively expressed, such as cancer and fibrosis (16). Such an approach should also allow an appreciation of the potential effect of anti-CCN2 therapies on these pathologies. Thus, in this study, we perform Affymetrix gene profiling, Western blot, and real time (RT)-PCR analyses to evaluate the contribution of CCN2 to the activity of TGF␤ in MEFs. Our results provide new insights into the role of CCN2 in mesenchymal biology.

MATERIALS AND METHODS
Cell Culture and Harvesting-ccn2ϩ/ϩ and ccn2Ϫ/Ϫ MEFs (embryonic day 14.5) were isolated as described previously (5) and grown in DMEM containing 10% fetal calf serum, 2 mM L-glutamine, antibiotics (100 units/ml penicillin and 100 g/ml streptomycin), and 1 mM sodium pyruvate (Invitrogen). Cells were grown at 37°C, 5% CO 2 , harvested at 90 -95% confluence, and washed twice in PBS. The cells were then scraped in PBS, and pellets were collected after centrifugation at 2000 rpm for 5 min at 4°C and resuspended in 2% SDS, quantified (BCA kit; Pierce), and placed in Laemmli sample buffer containing complete protease and phosphates inhibitors mixture (Roche Applied Science). fakϩ/ϩ and fakϪ/Ϫ MEFs (American Type Culture Collection) were similarly cultured.
Western Blot Analysis-Equal amounts of protein (20 g) were subjected to SDS-PAGE. Gels were electrophoretically transferred to nitrocellulose (Invitrogen). Membrane was blocked with 5% nonfat dry milk in Tris-buffered saline, 0.1% Tween 20 (Sigma), and immunoblotting was performed using anti-phospho-ERK, anti-ERK, anti-FAK, or antiphospho-FAK antibodies (Cell Signaling Technology), anti-CCN2 (Abcam), anti-type I collagen (BIODESIGN International), anti-␣-SMA, and anti-vinculin (Sigma) antibodies as described by the manufacturer. Anti-GAPDH and anti-vimentin (Sigma) antibodies were used as loading controls. Blots were then developed by incubation with biotinylated anti-rabbit or anti-mouse antibodies (1:1000; Vector Laboratories) as secondary antibodies, followed by incubation with ABC reagent (Vector Laboratories). Signal was detected using a luminescence kit (ECL kit; Amersham Biosciences) and x-ray film. Densitometry was performed using Gel Base/Gel-Blot Pro (Synoptics).
Cell Transfections-Transfections of ccn2ϩ/ϩ and ccn2Ϫ/Ϫ MEFs were performed essentially as described previously (11,13). Briefly, 2 ϫ 10 5 cells were seeded into each well of a 6-well plate. The next day, cells were transfected using FuGENE (Roche Applied Science) in a ratio of 3 l of FuGENE per 2 g of DNA (1.5 mg of reporter, 0.5 mg of CMV-␤galactosidase (Clontech)). Luciferase expression was determined and adjusted for ␤-galactosidase expression, which was used to control for differences in transfection efficiencies among wells (Applied Biosystems). Experiments were performed three times in triplicate. A representative experiment is shown. In rescue experiments, cells were similarly transfected with either a vector encoding constitutively active Akt1 or an empty expression vector (Upstate Biotechnology). Cells were incubated after transfection for 24 h in serum-free media, followed by further incubation in the presence or absence of 4 ng/ml TGF␤1 for 6 h. RNA was harvested and subjected to real time-PCR analysis. To verify the functionality of the Akt1 construct, a reporter DNA construct containing multimers of an NFB-responsive element upstream of the secreted enhanced alkaline phosphatase gene (NFB-SEAP; Clontech) was transfected into NIH 3T3 fibroblasts (American Type Culture Collection) using PolyFect (Qiagen) at a ratio of 10 l of PolyFect per 2 g of DNA (0.25 mg of CMV-␤-galactosidase; 0.25 mg of NFB-SEAP; 1 g of empty expression vector or expression vector containing constitutively active Akt1).
RNA Quality Assessment, Probe Preparation, and GeneChip Hybridization and Analysis-Microarrays and analysis were performed essentially as described previously (17,18). All GeneChips were processed at the London Regional Genomics Centre (Robarts Research Institute, London, Ontario, Canada). RNA was harvested (Trizol, Invitrogen) and quantified, and quality was assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies Inc.) and the RNA 6000 Nano kit (Caliper Life Sciences). Quality data were then analyzed using the degradometer (mean degradation factor 1.99, S.D. 0.0678). Biotinylated cRNA was prepared from 10 g of total RNA as per the Affymetrix GeneChip Technical Analysis Manual (Affymetrix). Double-stranded cDNA was synthesized using SuperScript II (Invitrogen) and oligo(dT) 24 primers. Biotin-labeled cRNA was prepared by cDNA in vitro transcription using the Bizarre High Yield RNA Transcript Labeling kit (Enzo Brioche) incorporating biotinylated UTP and CTP. Fifteen g of labeled cRNA was hybridized to Mouse Genome 430 2.0 GeneChips for 16 h at 45°C as described in the Affymetrix Technical Analysis Manual (Affymetrix). GeneChips were stained with streptavidin/phycoerythrin, followed by an antibody solution, and a second streptavidin/phycoerythrin solution, with all liquid handling performed by a GeneChip Fluidics Station 450. GeneChips were scanned with the Affymetrix GeneChip Scanner 3000 (Affymetrix, Santa Clara, CA). Signal intensities for genes were generated using GCOS1.2 (Affymetrix) using default values for the statistical expression algorithm parameters and a target signal of 150 for all probe sets and a normalization value of 1. Normalization was performed in GeneSpring 7.2 (Agilent Technologies Inc.). The Robust Multichip Average preprocessor was used to import data from the .cel files. Data were first transformed (measurements less than 0.01 set to 0.01) and then normalized per chip to the 50th percentile and per gene to wildtype control samples. Experiments were performed twice, and fold changes were identified using the GeneSpring filter. Data presented in Table 1 are an average of these independent studies. The fold change between treated and untreated samples had to be at least 2-fold to identify a transcript as being altered. These criteria had to be met in both sets of experiments. Clustering using gene ontologies was filtered using a 2-fold cutoff using the level two filtering parameter by the Affymetrix website. Probes annotated by the Gene Ontology Consortium were used. Categories that were considered over-represented in the set of transcripts that were CCN2-dependent had to have at least four members.
Real Time-PCR-Cells were serum-starved for 24 h and treated with 4 ng of TGF␤ for different lengths of time, as indicated. Total RNA was isolated using Trizol (Invitrogen), and the integrity of the RNA was verified by gel electrophoresis or Agilent Bioanalyzer. For initial time course analysis, total RNA (25 ng) was reverse-transcribed and amplified using TaqMan Assays on Demand (Applied Biosystems) in a 15-l reaction volume containing two unlabeled primers and 6-carboxyfluorescein-labeled TaqMan MGB probe.
Samples were combined with TaqMan one-step mastermix (Applied Biosystems). Amplified sequences were detected using the ABI Prism 7900 HT sequence detector (PerkinElmer Life Sciences) according to the manufacturer's instructions. Triplicate samples were run; transcripts were measured in picograms, and expression values were standardized to values obtained with control 28 S RNA primers as described previously (19). Statistical analysis was performed by the Student's paired t test.
Adhesion Assay-Adhesion assays were performed essentially as described (6). 96-Well plates were coated overnight at 4°C with 4 g/ml type I collagen, 4 g/ml fibronectin, or 4 g/ml bovine serum albumin (Sigma) in 1ϫ PBS. Nonspecific binding sites on substrates were blocked for 1 h in 1% BSA in PBS at room temperature prior to addition of cells. Fibroblasts were cultured in DMEM, 0.5% fetal calf serum for 24 h and then were harvested with 2 mM EDTA in PBS (20 min, room temperature), washed twice with DMEM containing 0.5% BSA (Sigma), and resuspended in the same medium at 2.5 ϫ 10 5 cells/ml. Cells (100 l/well) were added to wells of the 96-well plate and incubated at 37°C  for 60 min. An acid phosphatase assay was used, in which adherent cells were quantified by incubation with 100 l of substrate solution (0.1 M sodium acetate, pH 5.5, 10 mM p-nitrophenyl phosphate, and 0.1% Triton X-100) for 2 h at 37°C. The reaction was stopped by the addition of 15 l of 1 N NaOH/well, and A 450 was measured. Statistical analysis was performed by the Student's unpaired t test.

RESULTS
CCN2 Is Required for the Ability of TGF␤ to Induce ␣-SMA and Type I Collagen but Is Not Required for Induction of MMP-3 or a Smad3-respon-sive Promoter-CCN2 protein is induced by TGF␤ yet is expressed at a basal level in MEFs (9) (Fig. 1). To assess whether CCN2 was required for transcriptional responses to TGF␤, real time-PCR was performed on mRNAs prepared at various time points after TGF␤ treatment of ccn2ϩ/ϩ and ccn2Ϫ/Ϫ MEFs. Primers detecting matrix metalloproteinase-3 (MMP-3), COL1A1, COL1A2, and ␣-SMA were selected for these initial analyses, given that these are known markers of an activated fibroblast (20). To our surprise, the ability of TGF␤ to induce COL1A1, COL1A2, and ␣-SMA was impaired even at the earliest time points, paralleling the induction of CCN2 mRNA in ccn2ϩ/ϩ fibroblasts (Fig. 2). Results were con- firmed using Western blot analyses employing anti-␣-SMA and anti-type I collagen antibodies (Fig. 1). Conversely, TGF␤ was able to induce MMP-3 in the absence of CCN2 (Fig. 2). These results suggest that the ability of CCN2 to mediate TGF␤-induced gene expression is dependent on the basal, constitutive expression of CCN2 by MEFs rather than on the TGF␤induced increment of CCN2 expression. Given that MMP-3 was induced by TGF␤ in ccn2Ϫ/Ϫ MEFs, these results suggest that CCN2 is not generally required for TGF␤-dependent transcriptional responses but, in the context of MEFs, that CCN2 may be a cofactor for TGF␤ action.
To confirm the notion that CCN2 was not generally required for the ability of TGF␤ to induce gene expression, ccn2ϩ/ϩ and ccn2Ϫ/Ϫ fibroblasts were transfected with a DNA construct bearing multiple copies of a Smad-responsive element subcloned upstream of the luciferase reporter gene (SBE-lux; Fig. 3). This element is responsive to activation of the Smad3 pathway, which is required for TGF␤ responses in fibroblasts (21). The generic TGF␤-responsive Smad signaling pathway was fully active in ccn2Ϫ/Ϫ MEFs as the ability of the Smad reporter to respond to TGF␤ was not impaired in ccn2Ϫ/Ϫ MEFs (Fig. 3). Collectively, these results suggest that CCN2 was not generally required for TGF␤ signaling in fibroblasts but instead affects a specific subset of TGF␤-dependent responses.
Affymetrix Gene Profiling of Ccn2ϩ/ϩ and Ccn2Ϫ/Ϫ Fibroblasts Reveal Novel Genes Whose Induction by TGF␤ Is CCN2-dependent-To evaluate to what extent CCN2 was required for the ability of TGF␤ to induce gene expression by MEFs, we cultured ccn2ϩ/ϩ and ccn2Ϫ/Ϫ MEFs until 80% confluence and serum-starved cells for 24 h. Cells were then treated in the presence or absence of TGF␤ (4 ng/ml) for an additional 6 h. This time point was chosen because our real time-PCR analysis indicated that TGF␤ responses in MEFs were maximal at 6 h. Total RNA was prepared from these cells, reverse-transcribed, and applied to Affymetrix MOE430 arrays. Experiments were performed twice, and average induction values were obtained. Analysis of data by GeneSpring revealed TGF␤ induced 942 transcripts greater than 2-fold in ccn2ϩ/ϩ fibroblasts. Of these 345 were not induced in ccn2Ϫ/Ϫ fibroblasts. A representative selection of transcripts is shown in Tables 1 and 2. That the majority of TGF␤-induced genes in ccn2ϩ/ϩ were also induced in ccn2Ϫ/Ϫ MEFs supported our initial impression that CCN2 was not generally required for the ability of TGF␤ to induce gene expression. Pro-fibrotic (adhesion, contraction, and matrix) genes, as revealed by cluster analysis, were revealed to be both independent and dependent on CCN2 (Tables 1 and 2). Similarly, genes involved with cell signaling and metabolism were both independent and dependent on CCN2 (Tables 1 and 2). No significant physiological grouping specific to ccn2dependent or ccn2-independent genes emerged. However, analysis of functional clusters indicated that genes significantly over-represented in the group nonresponsive to TGF␤ in ccn2Ϫ/Ϫ fibroblasts were those involved with RNA processing and transcription. A complete list of these mRNAs is shown in Table 3. These results indicated that loss of CCN2 expression in MEFs is likely to both directly and indirectly affect gene expression in response to TGF␤ (Tables 2 and 3).
To confirm the requirement for target gene induction by TGF␤ for CCN2, we verified our gene array data using real time-PCR analysis of RNA isolated from ccn2ϩ/ϩ and ccn2Ϫ/Ϫ MEFs treated with and without TGF␤ for 6 h (Fig. 4). Whereas TGF␤ induced mRNA encoding the matrix protein COL2A1 and the focal adhesion protein vinculin (22) in ccn2Ϫ/Ϫ MEFs, the ability of TGF␤ to induce mRNAs encoding the matrix remodeling protein MMP14 (matrix metalloproteinase 14) (23), the collagen processing enzyme prolyl 4-hydroxylase (24), the cytoskeletal gene TMP1 (tropomyosin 1) (25), and the matrix and pro-adhesive proteins LIMS1 (26), Akt1 (27), and Emilin1 (28) were significantly impaired in the absence of CCN2 expression.
TGF␤-induced Akt/PI 3-Kinase and FAK Are Impaired by Loss of CCN2-We then wished to gain insights into the signaling pathways responding to TGF␤ that were deficient in ccn2Ϫ/Ϫ MEFs. After our initial experiments ruled out the Smad pathway, we reasoned that loss of ccn2Ϫ/Ϫ must affect the ability of TGF␤ to signal through other routes. Loss of CCN2 expression resulted in a reduced ability for TGF␤ to induce LIMS1, Akt1, and Emilin1 mRNA, proteins involved with adhesive signaling. CCN2 binds adhesive receptors including integrins and heparan sulfate-containing proteoglycans (6, 29 -31). In addition, ccn2Ϫ/Ϫ MEFs show reduced adhesion to and signaling on fibronectin (6). Thus, we reasoned that loss of CCN2 might affect the ability fibroblasts to respond to TGF␤ by inducing adhesive signaling. To confirm this notion, we harvested protein extracts prepared from ccn2ϩ/ϩ and ccn2Ϫ/Ϫ MEFs that had been treated for varying lengths of time in the presence or absence of TGF␤. The resultant protein extracts were subjected to Western blot analysis anti-phospho-Akt and anti-phospho-FAK antibodies. As controls, Western blots were performed with antibodies recognizing their unphosphorylated counterparts. We found that TGF␤ induction of Akt and FAK phosphorylation was significantly impaired in ccn2Ϫ/Ϫ MEFs, as the TGF␤ induction of phospho-Akt and phospho-FAK was significantly reduced in the absence of CCN2 (Fig. 5). Please note that basal Akt and FAK activation was lower in ccn2Ϫ/Ϫ fibroblasts and that the response to TGF␤ to induce Akt and FAK was impaired even at even the earliest time points (Fig. 5). That a defect in the early responses to TGF␤ was observed in ccn2Ϫ/Ϫ MEFs is consistent with the notion that the defects displayed in ccn2Ϫ/Ϫ MEFs are likely to be due to loss of constitutive CCN2 expression, rather than due to loss of TGF␤-induced CCN2 expression. Adhesive Signaling Is Required for the TGF␤ Induction of CCN2-dependent Genes-To determine the functional relevance of adhesive signaling to the responses of MEFs to TGF␤, we used fakϩ/ϩ and fakϪ/Ϫ MEFs to investigate whether the genes whose induction by TGF␤ in MEFs depended on CCN2 also depended on FAK expression. Real time-PCR analysis of fakϩ/ϩ and fakϪ/Ϫ MEFs treated with and without TGF␤ for 6 h showed that the ccn2-dependent genes prolyl 4-hydroxylase, COL1A1, and ␣-SMA by TGF␤ also depended on FAK expression (Fig. 6). Conversely, the ccn2-independent gene vinculin was induced by TGF␤ in a fashion independent of FAK (Fig. 6).
In fibroblasts, PI 3-kinase acts within the integrin/FAK/Akt cascade (32). Confirming the notion that impaired FAK/Akt signaling in response to TGF␤ was responsible for the inability of ccn2Ϫ/Ϫ MEFs to respond to TGF␤, we performed real time-PCR analysis of ccn2ϩ/ϩ MEFs treated with TGF␤ in the presence and absence of the PI 3-kinase inhibitor wortmannin. Consistent with the notion that CCN2 mediated the induction of Akt/PI 3-kinase in response to TGF␤, the ability of TGF␤ to induce type I collagen and ␣-SMA, but not vinculin, protein was blocked by wortmannin (Fig. 7). Similarly, the induction of COL1A1, ␣-SMA, and prolyl 4-hydroxylase mRNAs in ccn2ϩ/ϩ MEFs were impaired in the presence of wortmannin (Fig. 8). Collectively, To extend our analysis, we wished to assess whether overexpressing constitutively active Akt1 in ccn2Ϫ/Ϫ fibroblasts could now allow the expression of CCN2-dependent transcripts in response to TGF␤. To verify the functionality of our construct encoding constitutively active Akt1, we showed that co-transfection into fibroblasts of an expression vector constitutively active Akt1, but not empty expression vector, could activate an Akt1-responsive promoter. To perform this analysis, we used an NFB-responsive promoter (NFB-SEAP), as NFB is activated by Akt in fibroblasts (33). As anticipated, co-transfection of an expression vector constitutively active Akt1, but not empty expression vector, resulted in a significant activation of SEAP reporter gene expression, verifying functionality of the construct encoding constitutively active Akt1 (Fig. 9A). Extending these results, and confirming our Western blot data showing reduced basal Akt activity in ccn2Ϫ/Ϫ MEFs (Fig.  5), we showed that, when NFB-SEAP (which is non-TGF␤-responsive; data not shown) was transfected into ccn2ϩ/ϩ and ccn2Ϫ/Ϫ fibroblasts, reporter activity was significantly impaired in ccn2Ϫ/Ϫ fibroblasts, yet was rescued with Akt1 overexpression (Fig. 9A), giving support to the notion that Akt-dependent transcriptional responses are impaired in the absence of CCN2.
We then investigated whether the construct encoding constitutively active Akt1, as compared with empty expression vector, could rescue the ability of TGF␤ to induce expression of target mRNAs in ccn2Ϫ/Ϫ MEFs. We verified, using real time-PCR analysis of RNAs isolated from transfected MEFs, that we were able to achieve Akt1 overexpression in ccn2Ϫ/Ϫ MEFs transfected with expression vector encoding Akt1 (Fig.  9B). As earlier data examining the effect of loss of CCN2 expression on the induction of a Smad3-responsive promoter suggested that MEFs transfected poorly, we focused our rescue experiments on the mRNA expression of prolyl 4-hydroxylase and ␣-SMA, because the pattern of expression of these mRNAs suggested a high response to TGF␤ in ccn2ϩ/ϩ cells, and thus examination of these mRNAs would enhance our chances of observing a statistically significant alteration in transcriptional responses to Akt1 in transfected ccn2Ϫ/Ϫ cells. Unfortu- FIGURE 5. ccn2؊/؊ mouse embryonic fibroblasts show impaired phosphorylation of FAK and Akt in response to TGF␤1. ccn2ϩ/ϩ and ccn2Ϫ/Ϫ mouse embryonic fibroblasts were serum-starved for 24 h, prior to exposure to TGF␤1 (4 ng/ml) for the durations indicated. Cell layers were harvested, and equal amounts of protein were subjected to Western blot analysis with anti-phospho-FAK and anti-FAK antibodies or anti-phospho-Akt1 and Akt1 antibodies, as indicated. Note that impairment of protein phosphorylation occurred at the earliest time points. FIGURE 6. The induction of CCN2-dependent transcripts by TGF␤ in MEFs requires FAK. MEFs isolated from fakϩ/ϩ and fakϪ/Ϫ mice were cultured in DMEM, 0.5% FBS for 24 h and treated for TGF␤1 (4 ng/ml) for 6 h, as indicated. RT-PCR was performed with the indicated primer sets. Those genes examined whose TGF␤ inductions were CCN2-dependent and also FAK-dependent, whereas ccn2-independent genes were FAK-independent (* indicates significant impairment of induction of mRNA by TGF␤ in the knock-out; p Ͻ 0.05). Values are expressed as fold increase resulting from TGF␤ treatment, relative to basal mRNA expression, which was adjusted to 1, in control fakϩ/ϩ and fakϪ/Ϫ fibroblasts. nately, the relatively low induction in response to TGF␤ in ccn2ϩ/ϩ MEFs of type I collagen mRNA precluded the use of this transcript for this experiment. We found that transfection of constitutively active Akt1 at least partially restored the ability of ccn2Ϫ/Ϫ cells to respond to TGF␤ by inducing prolyl 4-hydroxylase and ␣-SMA (Fig. 9C, P4H and ␣-SMA). Conversely, overexpression of Akt1 seemed to partially attenuate the ability of ccn2Ϫ/Ϫ cells to respond to TGF␤ by inducing vinculin mRNA (Fig. 9C, vinculin). These latter results presumably reflect that overexpression of active Akt1 results in a bias in TGF␤ responsiveness toward Akt1-dependent genes and away from Akt1-independent genes. Collectively, our results point to a critical role of CCN2 in mediating the TGF␤ induction of adhesion and adhesive signaling in MEFs.
CCN2 Is Required for TGF␤-induced Adhesion-Based on our analysis of the effect of loss of CCN2 on gene expression and signaling in response to TGF␤, and our previous data showing that a principal defect of ccn2Ϫ/Ϫ fibroblasts was a decreased ability of cells to adhere to fibronectin (6), we reasoned that loss of CCN2 might affect the ability of MEFs to adhere to extracellular matrix proteins after TGF␤ treatment. To evaluate this hypothesis and provide a functional context for our microarray and signal transduction analyses, we treated ccn2ϩ/ϩ and ccn2Ϫ/Ϫ MEFs with TGF␤ for 24 h, and cell adhesion to fibronectin and type I collagen was assessed. As anticipated, we found that loss of CCN2 resulted in reduced cell adhesion to matrix, consistent with our previous data showing that CCN2 is required for maximal cell adhesion to fibronectin (6) (Fig. 10). However, although TGF␤ was able to enhance the ability of ccn2ϩ/ϩ MEFs to adhere to fibronectin and type I collagen, ccn2Ϫ/Ϫ MEFs were unable to respond to TGF␤ by an increase in adhesion to ECM (Fig. 10). These results provide functional support to our array and signaling data and suggest that adhesive signaling in response to TGF␤ is impaired in ccn2Ϫ/Ϫ MEFs compared with ccn2ϩ/ϩ MEFs. Collectively, our results support the notion that a principal, physiological role of CCN2 is to modulate adhesive responses in fibroblasts to ECM and growth factors.

DISCUSSION
TGF␤ is a potent profibrotic cytokine; yet, as TGF␤ possesses multiple functions, antagonizing TGF␤ in chronic fibrotic diseases is likely to have substantial side effects (34,35). Consequently, identification of a specific modulator of TGF␤ is likely to be of significant impact in the development of selective therapies for disease. Although it has long been hypothesized that CCN2 is a downstream mediator of TGF␤ action (15), the mechanism of action of CCN2 is largely unknown. Functional data indicate that overexpression of CCN2 itself may have little effect (36,37) but may require a co-stimulus for its activity (36,37). CCN2 directly binds to TGF␤ and appears to increase binding to TGF␤ to the TGF␤ receptors and activation of a Smad-responsive promoter in Xenopus cells at extremely low TGF␤ concentrations (38); however, the functional relevance of these observations is unknown. Overall, whether CCN2 modulates particular TGF␤ signaling pathways and the physiological effect of loss of CCN2 to mammalian cells are unclear.
In this study, for the first time we show that CCN2 is a selective mediator of TGF␤ action in MEFs. TGF␤ is able to induce a majority of TGF␤-responsive genes in ccn2Ϫ/ϪMEFs and activate a generic Smad3-responsive promoter; however, a significant minority of TGF␤responsive genes depends on CCN2. TGF␤ induction of type I collagen and ␣-SMA is impaired by loss of CCN2 expression, even at very early time points. TGF␤-induced FAK and Akt phosphorylation is impaired in ccn2Ϫ/Ϫ MEFs. ccn2Ϫ/Ϫ MEFs showed defects in adhesive signaling in response to TGF␤ and whether these responses were relevant to FIGURE 7. The induction of type I collagen, ␣-SMA, but not vinculin, protein is blocked by wortmannin. Wild-type MEFs were cultured in DMEM, 0.5% FBS for 24 h and treated for TGF␤1 (4 ng/ml) for 6 h in the presence or absence of prior treatment for 45 min with 100 nM of the PI 3-kinase inhibitor wortmannin. Protein (␣-SMA, type I collagen, and vinculin) was detected as described under "Materials and Methods." FIGURE 8. The induction of CCN2-dependent transcripts in ccn2؉/؉ MEFs is blocked by wortmannin. MEFs isolated from ccn2ϩ/ϩ mice were cultured in DMEM, 0.5% FBS for 24 h and treated for TGF␤1 (4 ng/ml) for 6 h in the presence or absence of prior treatment for 45 min with or without 100 nM wortmannin. RT-PCR was performed with primer sets, as indicated, detecting transcripts whose induction was determined previously to be CCN2-dependent. Addition of wortmannin blocked the TGF␤ induction of ccn2-dependent genes in ccn2ϩ/ϩ MEFs (* indicates significant induction of mRNA by TGF␤, p Ͻ 0.05). WT, ccn2ϩ/ϩ; WTT, ccn2ϩ/ϩ ϩ TGF␤; WTT ϩ W, ccn2ϩ/ϩ ϩ TGF␤ ϩ wortmannin.
the induction of CCN2-dependent mRNA induction. Indeed, TGF␤ induction of FAK and Akt phosphorylation was severely impaired in ccn2Ϫ/Ϫ MEFs, and the Akt/PI 3-kinase pathway was necessary for the induction of ccn2-dependent genes in ccn2ϩ/ϩ MEFs in response to TGF␤. These results suggest that, in the context of "activated" fibroblasts such as MEFs that endogenously express CCN2 (10), CCN2 is a co-factor required for TGF␤ induction of gene expression. A similar function of CCN2 would be expected in activated fibroblasts in vivo; e.g. in fibrotic cells (16) or in tumor stroma (39).
It is interesting to note that cluster analysis did not reveal categories of genes that were completely dependent on CCN2. It should be noted, however, that consistent with previous observations examining ␣-SMA stress fiber formation post-adhesion (6), CCN2 was not required for TGF␤-induced collagen gel contraction (data not shown). Intriguingly, we found that loss of CCN2 did not affect basal mRNA or protein expression of ␣-SMA or type I collagen, suggesting that CCN2 does not directly mediate the expression of these genes. These results emphasize the selective role of CCN2 as a mediator of TGF␤ action, and are consistent with the notion that induction of mRNAs encoding several procontractile proteins were not affected by loss of CCN2, including vinculin. However, there were genes over-represented in those transcripts affected by the loss of CCN2, namely those involved with transcription and translation. Although beyond the scope of our current study, these results suggest that CCN2 may indirectly affect the expression of downstream target genes, in addition to being required as a cofactor to modulate TGF␤ signaling. It is interesting to note that several of the mRNAs whose expression depended on CCN2, as revealed by array analysis, included the forkheads, which are Akt-responsive (40). In addition, it is likely that activation of downstream transcription factors that are Aktresponsive is impaired in ccn2Ϫ/Ϫ MEFs; indeed, we have shown that activation of a reporter gene driven by multiple copies of a binding  . Enhanced adhesion MEFs to fibronectin and type I collagen in response to TGF␤ is CCN2-dependent. MEFs isolated from ccn2ϩ/ϩ and ccn2Ϫ/Ϫ mice dermal fibroblasts were cultured in DMEM, 0.5% FBS for 24 h, treated with TGF␤1 (4 ng/ml) for 24 h, detached with EDTA, and allowed to adhere to BSA, type I collagen, or fibronectin (4 g/ml each) that had been coated on a 96-well plate. After a blocking step, fibroblasts were allowed to adhere for 40 min, and cell adhesion was assessed as described under "Materials and Methods" (6 wells/data point; average Ϯ S.E. (error bar) are shown). The values obtained from the BSA control experiment, not significantly above background, indicate the very few cells nonspecifically adhering to the plate. element for NFB, a transcription factor responsive to Akt (33), is impaired in the absence of CCN2. Consistent with this notion, a recent report indicated that recombinant CCN2 activated NFB (41).
The previous impression that CCN2 is a downstream mediator of TGF␤ emerged from the observation that TGF␤ induces CCN2 in fibroblasts but not in epithelial cells, suggesting that CCN2 is a specific mediator of TGF␤ action in fibroblasts (15). However, the expression of CCN2 in MEFs and in fibrotic cells is not dependent on the TGF␤response element of the CCN2 promoter or on Smad3 (10,11). In addition, whereas subcutaneous injection of CCN2 alone into the back of rats causes little or no fibrotic response, co-injection of CCN2 with TGF␤ results in persistent fibrotic responses (35). (Conversely, TGF␤ alone, either in vitro or in vivo, causes a transient fibrotic response that depends on the continuous application of ligand (36,42).) We anticipate in conditions of pathological CCN2 overexpression, such as in the fibrotic disease scleroderma, there is constitutive expression of CCN2 observed even in nonlesional fibroblasts (fibroblasts isolated from unscarred back skin of scleroderma patients; see Ref. 43). In skin of scleroderma patients, scarring expands outward from the established lesion driven by a wave of inflammation, accompanied by an increased local TGF␤ concentration (44). Given the results presented in this study, we would expect that in this context CCN2 would enhance TGF␤ activation of FAK/Akt contributing to the expansion of the fibrotic lesion. Intriguingly, FAK and Akt have both been shown recently to be activated in scleroderma fibroblasts (45,46). Consistent with our findings that PI 3-kinase/Akt is involved with myofibroblast formation (this report and see Ref. 47), it was recently reported that TGF␤ induction of ␣-SMA in 10T1/2 mesenchymal cells was inhibited by PI 3-kinase/Akt antagonism (48). Although beyond the scope of this present study, our results suggest that CCN2 (or FAK/Akt) antagonism would be expected to have a significant impact on retarding the progression of the fibrosis in scleroderma by preventing the TGF␤ induction of adhesive signaling in the region of lesional expansion.
It is interesting to note that both MEFs and nonlesional scleroderma fibroblasts are already somewhat activated because they both express ␣-SMA (6,42). This is in contrast to adult fibroblasts (20). We would hypothesize that CCN2 expression in cells in which CCN2 is basally expressed would be involved physiologically, in effect, in the "superactivation" of cells that are already stimulated by permitting TGF␤ to induce adhesive pathways. This hypothesis is similar to those proposed previously that CCN2 perpetuates TGF␤ responses in scleroderma patients, resulting in sustained, chronic fibrosis (49). Obviously, basal CCN2 expression would not be required for normal, "resting," nonactivated adult fibroblasts to be activated in response to TGF␤, as CCN2 is not basally expressed by normal adult fibroblasts. In this regard, it is interesting to note that studies identifying the contribution of CCN2 to gene expression, using application of neutralizing oligonucleotides recognizing CCN2, have often been performed on cell types, such as rat mesangial cells (e.g. Ref. 50), which express CCN2 endogenously (13).
A key feature of tissue remodeling requires that cells attach and migrate upon ECM. Functional evidence for the role of CCN2 in adhesive signaling in response to TGF␤ was provided by the observation that TGF␤-induced adhesion to ECM was impaired in the absence of CCN2. Indeed, recombinant CCN2 promotes cell adhesion and migration via integrins and heparan sulfate-containing proteoglycans and activates FAK and Akt (30,51). In addition, endogenous CCN2 protein is found in a complex with integrin ␣5, ␣4, and ␤1 and syndecan 4 and that CCN2 is required for optimal adhesive signaling of MEFs on fibronectin (6). TGF␤ activates adhesive signaling through integrin ␣5␤1 (52), which is also a receptor for CCN2 (6, 50, 53). CCN2 directly binds TGF␤ (38).
Our results are consistent with the notion that CCN proteins in general, including CCN2, interact with different proteins and ligands involved in signaling and consequently act to bring together regulatory circuits and consequently modify their activity (3,4).
In conclusion, although CCN2 was discovered over a decade ago, the physiological roles of CCN2 (and the signaling cascades through which CCN2 acts) are largely unknown. Our results, using MEFs, showing that CCN2 is required for the ability of TGF␤ to induce FAK and Akt phosphorylation and for adhesion of ECM, indicate a crucial role for CCN2 in mediating adhesive signaling in response to TGF␤. These results suggest that CCN2 plays a critical role in facilitating tissue remodeling and emphasize the specific role that CCN2 may play in situations where CCN2 is constitutively expressed, such as development, cancer, and fibrosis.