Synergistic Cooperation between Sp1 and Smad3/Smad4 Mediates Transforming Growth Factor β1 Stimulation of α2(I)-Collagen (COL1A2) Transcription*

Transforming growth factor-β1 (TGFβ) is a strong activator of extracellular matrix accumulation. TGFβ stimulates the gene coding for human α2(I)-collagen (COL1A2) by inducing binding of an Sp1-containing complex to an upstream promoter element (TGFβ responsive element or TbRE) that contains a CAGA box. Here we report that the CAGA box of the TbRE is the binding site of the Smad3/Smad4 complex, and that the binding of the complex is required for TGFβ-induced COL1A2 up-regulation. Recombinant Smad3 and Smad4 bind in vitro to the CAGA box of COL1A2; TGFβ treatment of cultured fibroblasts induces Smad3/Smad4 binding to the TbRE; transient overexpression of Smad3 and Smad4 in fibroblasts transactivates TbRE-driven transcription; and COL1A2 gene up-regulation by TGFβ is abolished in cells stably transfected with plasmids that express dominant negative forms of Smad3 or Smad4. In Sp1-deficientDrosophila Schneider cells, there was cooperative synergy between Smad3/Smad4 and Sp1 at the TbRE site. The analysis also emphasized the requirement of both Sp1- and Smad-binding sites for optimal promoter transactivation. In cells stably transfected with a plasmid expressing a dominant negative form of Sp1, the synergy was shown to be promoter-specific and dependent on the binding of Sp1 to the TbRE. Interestingly, overexpression of dominant negative Sp1 was found to block the antagonistic signal of tumor necrosis factor-α on COL1A2 transcription, as well. These results provide the first linkage between the Smad3 and Smad4 proteins and TGFβ stimulation of type I collagen biosynthesis.

The extracellular matrix (ECM) 1 is responsible for the biomechanical and physiological properties of the connective tissue; additionally, it participates in embryonic development and growth through the modulation of several cellular activities (1). The balance between deposition and degradation of the ECM is also a critical aspect of wound healing, tissue repair, and homeostasis (1). It follows that deregulated ECM remodeling is the pathological hallmark of numerous fibrotic and degenerative conditions (2). Transforming growth factor ␤1 (TGF␤) is a key regulator of ECM assembly and remodeling (2)(3)(4). This action is exerted through two complementary pathways, one which reduces matrix degradation and the other which stimulates matrix accumulation (2)(3)(4). In mechanistic terms, TGF␤ inhibits the synthesis of extracellular proteinases while upregulating the production of their inhibitors and structural ECM components (2)(3)(4). The combined action of TGF␤ on the genes implicated in the formation and degradation of the ECM is mostly exerted at the transcriptional level through ill defined intracellular pathways. Elucidating the molecular nature of these pathways is therefore relevant to the understanding of morphogenesis, homeostasis, and tissue repair, as well as of a wide variety of pathologies. The genes coding for the plasminogen activator inhibitor-type 1 (PAI-1) and for the ␣2 chain of type I collagen (COL1A2) are thus far the best characterized experimental models to study the intracellular pathway responsible for TGF␤ stimulation of ECM accumulation (5,6).
Type I collagen is the major structural component of the ECM and consequently, the main contributor to connective tissue physiology and pathology (4,7). Indirect lines of evidence have established a causal relationship between TGF␤ activity and type I collagen production. First, expression of the type I collagen genes correlates with the extracellular distribution of TGF␤ during mouse embryogenesis (8). Second, levels of TGF␤ and type I collagen are elevated in activated cells from patients with fibrotic diseases (9). Third, TGF␤ stimulates type I collagen mRNA accumulation in cultured cells, independently of de novo protein synthesis (10). Fourth, TGF␤ activates transcription of type I collagen promoter/reporter constructs in transient transfection assays (6).
Using a combination of cell transfection and protein DNA binding assays, we have located the TGF␤-responsive element (TbRE) of the human COL1A2 gene between nucleotides 313 and 250, relative to the start site of transcription (11). The TbRE sequence consists of two nearly juxtaposed footprints (boxes 3A and B), the most distal of which represents the 3Ј half of a larger footprinted area (box A) (Fig. 1). The analyses have also demonstrated that binding of the ubiquitous activator Sp1 to box 3A (Ϫ313/Ϫ286) is required for TbRE activity (11,12). The electrophoretic mobility shift assay (EMSA) has documented that TGF␤ treatment of cultured fibroblasts translates into increased intensity of the TbRE-bound complex (TbRC) (11). The TGF␤-induced increase appears to involve Sp1 co-factor(s) since it is not observed when only box 3A is used as a probe in the EMSA (11). At variance with our data, other investigators have reported that CTF/NF-I or AP1, and not Sp1, mediate TGF␤ stimulation of the mouse or human COL1A2 promoter (13,14). As a result of this controversy, the intracellular pathway that mediate TGF␤ induction of ECM accumulation remains undefined.
The Smad proteins have emerged as essential components of the TGF␤ signaling pathways (15). They are generally divided into pathway-restricted (Smads 1-3, 5, 8, and 9), common mediator (Smad4) and inhibitory (Smads 6 and 7) Smads (15)(16)(17). Smad2 and Smad3 are specific mediators of the TGF␤ and activin pathways (15)(16)(17)(18)(19). TGF␤ binds to the type II receptor and this in turn leads to interaction with and phosphorylation of the type I receptor, and activation of its kinase activity. The ligand-bound transmembrane receptor complex then activates the Smad2 or Smad3 proteins which form alternative heterooligomers with Smad4 and translocate into the nucleus (15). The resulting Smad complexes can regulate transcription by binding directly to defined DNA sequence complexes or without binding directly to DNA (16,19). Cooperativity between Smad complexes and other transcription factors at distinct DNAbinding sites has been identified within several TGF␤-responsive elements as well (20 -24). The modular arrangement of distinct DNA-binding sites within a given TGF␤-responsive element is therefore believed to direct binding specificity, as well as cooperativity between other transcription factors and Smad complexes (15).
A recent study has linked TGF␤ inducibility of the PAI-1 gene to the binding of Smad3/Smad4 to cis-acting elements that contain the sequence AG(G/A)CAGACA, a.k.a.: CAGA box (20). The authors of that report also noted that single or multiple CAGA boxes are present in the TGF␤-responsive elements of several other genes, including COL1A2, thus raising the possibility that the Smad3-Smad4 complex might be the common mediator of the TGF␤ signaling for ECM accumulation (20). The putative Smad-binding site (ATGCAGACA) of COL1A2 lies within box B (Ϫ271/Ϫ250) immediately 3Ј of the Sp1 recognition sequence (Fig. 1); consequently, a Smad complex may represent the alleged Sp1 co-factor involved in COL1A2 transactivation (11). Consistent with this observation, transient overexpression of Smad3 and Smad4 has been recently shown to transactivate the Ϫ772 COL1A2 promoter (25). However, this preliminary observation did not link the transactivation to Smad binding to box B, or evaluate the possibility of a functional interaction between the Smads and Sp1 (25).
The present study was designed to further characterize TGF␤ up-regulation of COL1A2 and more generally, to increase our understanding of the TGF␤ signaling pathway that controls ECM accumulation. To this end, we examined whether the Smad proteins are indeed components of the TbRC and if so, what their functional interaction with Sp1 might be. Our results indicate that Sp1 and Smad3/Smad4 cooperate synergistically in transactivating the COL1A2 promoter after binding to the TbRE. Furthermore, they provide additional evidence for the critical role of Sp1 in constitutive COL1A2 expression, and in integrating the transcriptional responses of the gene to antagonistic cytokines.

EXPERIMENTAL PROCEDURES
Cell Cultures, Plasmids, and Reagents-Mouse NIH3T3 fibroblasts were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (HyClone Inc. Logan, UT) and antibiotics (penicillin, 50 units/ml, and streptomycin, 50 g/ml) (11). TGF␤ and TNF␣ were purchased from Roche Molecular Biochemicals (Indianapolis, IN) and added at final concentrations of 7.5 and 20 ng/ml, respectively, 8 h after the cells were placed in medium containing 0.1% fetal calf serum (26). Schneider cells were grown in Drosophila Schneider cell medium with 10% heat-inactivated fetal bovine serum (27). Bacterial expression plasmids (pGEX-Smad2 and pGEX-Smad2dMH2) were generous gifts of Dr. P. ten Dijke (Ludwig Institute for Cancer Research, Uppsala, Sweden), or were derived from the pRK5F-Smad3 and pRKF5-Smad4 plasmids (kindly provided by Dr. R. Derynck, University of California in San Francisco) (18) by subcloning the inserts into the pGEX vector (Amersham Pharmacia Biotech). Deletion of the MH2 coding domain of pGEX-Smad3 led to the generation of pGEX-Smad3⌬MH2. Similarly, the inserts of pRK5F-Smad3 and pRKF5-Smad4 were resubcloned into the mammalian expression vector pcDNA3.1 (Invitrogen, Carlsbad, CA) along with the Flag-tag sequence at the 3Ј end. The pcDNA3.1 vector was also used to subclone the inserts of the dominant negative Smads, pRK5F-Smad3⌬C and pRKF5-Smad4⌬C (generously provided by Dr. R. Derynck) (18). To generate the plasmid expressing the dominant negative version of Sp1, the sequence coding for the zinc finger domain of the protein was amplified using specific primers and the pfu Taq polymerase (Stratagene, La Jolla, CA) using as a template the DNA of pSp1778cDNA (a kind gift of Dr. J. Kadonaga, University of California in San Diego) (28). The inserts from pRK5F-Smad3 and pRKF5-Smad4 were subcloned into the pPac vector to generate pPacSmad3 and pPacSmad4 (27). The pPacSp1 construct was generously provided by Dr. D. A. Brenner (University of North Carolina, Chapel Hill) (27), whereas Sp1mut-LUC was generated from an analogous construct kindly provided by Dr. M. Trojanowska (Medical University of South Carolina, Charleston) (29) by subcloning the mutant promoter into the pXP (1) vector (30). Other promoter reporter gene constructs have been described elsewhere (18,23). The Ϫ800PAI/ LUC construct and the PAI-1 cDNA were kind gifts of Dr. D. Rifkin (New York University School of Medicine) (31). Recombinant Sp1 was purchased from Promega (Madison, WI), whereas nuclear factor-specific antibodies and anti-GST antibody were purchased from Santa Cruz Biotechnology (Santa Cruz, CA).
Bacterial Recombinant Proteins-Cells of the XL1-Blue bacterial strain were transformed with the expression vectors, grown and induced according to standard protocols (32). After induction, cells were lysed by sonication and 150 l of gluthathione-Sepharose 4B (Amersham Pharmacia Biotech) were added to the supernatant. After 1 h incubation at room temperature with gentle shaking, the beads were collected and washed and the bound proteins were eluted in 300 l of the manufacturer's recommended buffer. Protein concentration was estimated using a commercial protein detection kit (Bio-Rad). Nuclear extracts were purified from control and TGF␤-treated cells as described previously (11,12). In some EMSAs, 20 ng of recombinant GST fusion proteins were incubated under previously described conditions together with ϳ10,000 cpm of radiolabeled probe (11,12). In other EMSAs, the incubation included 0.5 footprint unit of recombinant Sp1 protein or 5 g of crude nuclear extracts. When appropriate, samples were preincubated for 1 h with antibodies or unlabeled oligonucleotides. DNA-protein complexes were separated from unbound material in 5% polyacrylamide gel electrophoresis and visualized by autoradiography.
Cell Transfections-Conditions for the preparation and transfection of plasmids into mammalian and insect cells by the calcium phosphate procedure have been detailed before (11,27). After transfection, cells were placed in Dulbecco's modified Eagle's medium with 10% fetal calf serum for 24 -36 h and in Dulbecco's modified Eagle's medium with 0.1% fetal calf serum for 8 h followed by addition of TGF␤ for an additional 0.5 or 12 h for nuclear extract preparation or gene expression assays, respectively (11). The activity of the various reporter genes was measured according to the standard protocol and normalized against co-transfected control plasmids (11); they include the reporter genes luciferase (pSVLUC), chloramphenicol acetyltransferase (pSVCAT), and ␤-galactosidase (HspLacZ) which are under the control of the SV40 (SV) or heat-shock protein promoter. Schneider cells were harvested and processed 3 days after transfections following the published protocol (27). Transfections were performed multiple times and in duplicate. The statistical value of the data was evaluated by the Mann-Whitney U test. Stable transfectants were selected by culturing the cells for about 3 weeks in the presence of 400 g/ml G418 (33).
Northern and Western Blots and in Vitro Transcription Assay-Total RNA purified according to Chomczynski and Sacchi (34) was used for Northern blot hybridization as probes for ␣2(I)-collagen, PAI-1, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Rates of COL1A2 and GAPDH transcription were determined and quantified as described previously (35); experiments were performed in triplicate. For Western analysis, cell extracts were fractionated by SDS-polyacrylamide gel electrophoresis on a 12% gel and transferred onto a nitrocellulose membrane. The membrane was probed with the anti-Flag M2 monoclonal antibody (Sigma) at a final dilution of 1:2,000, followed by incubation with horseradish peroxidase-conjugated goat anti-mouse immunoglobulin G (Santa Cruz Biotechnology), as per the manufacturer's recommendations.

Smads and Sp1
Bind to the TbRE-It has been reported that TGF␤ stimulates PAI-1 gene transcription through the binding of Smad3/Smad4 hetero-oligomers to three promoter sites containing the so-called CAGA box (20). The same study also noted that a single CAGA box is present in the TbRE of the COL1A2 promoter, immediately 3Ј of the Sp1-binding sites ( Fig. 1). Based on this evidence, we tested whether or not the CAGA box in the TbRE is indeed an authentic Smad-binding site.
To this end, purified recombinant Smad2, Smad3, and Smad4-GST fusion proteins were incubated with (CAGA) 3 , an oligonucleotide that contains three copies of the CAGA box of COL1A2. The GST fusion products included the full-length Smad2, Smad3, and Smad4 proteins, and truncated versions of Smad2 and Smad3 without MH2 domains. The MH2 domain mediates protein multimerization of activated Smads, in addition to inhibiting DNA binding by the MH1 domain of inactive Smad proteins (21,36,37). Deletion of the MH2 domains was therefore required to test the ability of the MH1 domains of Smad2 and Smad3 to bind to (CAGA) 3 . The EMSA documented that only the samples that contain recombinant Smad4 and truncated Smad3 proteins yield retarded bands which, in addition, are supershifted by the respective antibodies ( Fig. 2A). Competition with the unlabeled (CAGA) 3 oligonucleotide, but not an unrelated sequence, further documented the binding specificity ( Fig. 2A). Another EMSA was performed using nuclear extracts purified from untreated and TGF␤-treated fibroblasts in order to confirm the relationship between intracellular activation of Smads and Smad3/Smad4 binding to (CAGA) 3 . This second EMSA documented the appearance of a new retarded band in the TGF␤-treated sample, whose specificity was documented by the supershift with the anti-Smad3 and anti-Smad4 antibodies (Fig. 2B). Incidentally, none of the other bands were affected by the antibodies in the sample without TGF␤ treatment (data not shown). Altogether, the results indicated that Smad3 and Smad4, but not Smad2, bind to the CAGA motif of the TbRE, and that the binding is inducible by TGF␤.
The next set of EMSAs employed the natural TbRE sequence in place of the multimerized CAGA motif, and recombinant Smad3 and Smad4 proteins with or without recombinant Sp1. Like the previous test, recombinant Smad3 did not contain the MH2 domain. Incubation of the TbRE probe with recombinant Smad3 and Smad4 yielded no visible complex even in the presence of 400 ng of each protein (data not shown). Similarly, recombinant Sp1 alone yielded only a smear suggesting that stabilization of the complex may depend on the presence of all three proteins (Fig. 3A). Indeed, incubation of the TbRE probe together with recombinant Sp1, Smad3, and Smad4 yielded two retarded bands whose specificity was documented by being affected by the anti-Sp1 and anti-Smad3 antibodies, but not by the anti-Smad4 antibody (Fig. 3A). This last result is at vari-ance with the observed binding of recombinant Smad4 to the (CAGA) 3 oligonucleotide ( Fig. 2A). One possible explanation for the discrepancy is that binding of Smad3 may be favored over Smad4 when a single CAGA motif is available, and/or in the presence of Sp1 bound to the cognate site in the TbRE. Assuming that this interpretation is correct, the finding would imply that binding of the Smad3-Smad4 complex to the TbRE is mediated by Smad3.
To further confirm the combined participation of Smad and Sp1 proteins in the formation of the TbRE-bound complex, we employed nuclear extracts from untreated and TGF␤-treated cells that were transfected with Smad3 and Smad4 expression plasmids. The EMSA revealed that the patterns of nuclear extracts from untreated and TGF␤-treated cells are identical, and that the retarded bands are equally affected by antibodies against the Smad or Sp1 proteins (Fig. 3B). On the other hand, preincubation with anti-Sp1 or anti-Smads antibodies decreased the intensity of the same bands without an apparent supershift (Fig. 3B). In contrast to the results obtained with the (CAGA) 3 probe, the same anti-Smad antibodies had no effect on the binding to the TbRE of nuclear extracts purified from TGF␤-treated cells that were not transfected with Smad expression plasmids (data not shown). This point notwithstanding, the results of the EMSAs using purified recombinant products were consistent with those performed with Smad proteins transiently expressed in mammalian cells.
Smad3/Smad4 and Sp1 Synergistically Transactivate TbRE-driven Transcription-The conclusions drawn from the DNA binding assays were tested functionally in transient transfection experiments using mammalian and insect cells. Mouse NIH3T3 fibroblasts were co-transfected with the Ϫ378COL1A2/LUC construct and with either Smad2 and Smad4 or with Smad3 and Smad4 expression plasmids. As predicted from the DNA binding assays, the former combination had no effect on luciferase gene expression and the Smad3/ Smad4 combination induced it about 1.8-fold (Fig. 4A). The Smad3/Smad4 stimulation was directly linked to the presence of the CAGA motif in the TbRE by documenting loss of transactivation using a reporter construct that harbors the mutant CAGA box (Fig. 4B). Smad3/Smad4 inducibility was also tested using the heterologous TK promoter driven by three copies of the region that encompasses the TbRE (boxes A and B, Fig. 1) in order to corroborate previous evidence of the autonomous nature of the TbRE-mediated response to TGF␤ (11). As predicted, Smad3/Smad4 stimulated TbRE-driven transcription also in the context of the heterologous promoter and at a higher level (ϳ3.3-fold) than the natural promoter (Fig. 4B). The higher transactivation of the heterologous promoter with multiple response elements compared with the natural promoter with only one copy of the response element is in line with results obtained with other TGF␤-inducible genes, such as PAI-1 (20).
The next set of co-transfections were designed to characterize the functional relationships between Sp1 and Smad3/ Smad4 binding to the TbRE. The experiments were performed in Drosophila Schneider cells, rather than mammalian fibroblasts, in order to assess the effect in an Sp1-deficient background. Luciferase reporter constructs containing different COL1A2 promoter sequences were transfected in Schneider cells together with a fixed amount of the Sp1 expression vector. The reporter constructs included Ϫ378COL1A2/LUC, as well as a plasmid in which the TbRE is directly linked to the TATA box of COL1A2 (⌬12/LUC) and a plasmid that contains only the TATA box of COL1A2 (Ϫ54COL1A2/LUC) (Fig. 5A). Sp1 stimulated Ϫ378COL1A2/LUC expression about 10-fold compared to the empty vector (Fig. 5A). The ⌬12/LUC plasmid exhibited the same level of Sp1 inducibility as Ϫ378COL1A2/LUC, but higher basal activity (Fig. 5A). This last result is consistent with the proposed presence of a repressor-binding site in the deleted region of ⌬12/LUC (29). As expected, Sp1 failed to induce the basal Ϫ54COL1A2/LUC promoter construct (Fig. 5A).
Additional cell transfections were performed with varying amounts of the Sp1 expression plasmid in order to titrate the effect of the nuclear factor. Increasing amounts of the Sp1 expression plasmid activated the Ϫ378COL1A2/LUC reporter construct up to 46-fold; furthermore, the titration test revealed that 100 ng of the Sp1 expression plasmid transactivate the reporter construct significantly and below the saturation level (Fig. 5A). Schneider cells were therefore co-transfected with various combinations of increasing amounts of Smad3 and Smad4 expression vectors and with or without 100 ng of the Sp1 expression plasmid. Co-expression of Smad3 and Smad4 proteins had no effect on the reporter gene; by contrast, coexpression of Smad3 or Smad4 together with Sp1 resulted in a more significant increase by the former compared to the latter combination (Fig. 5B). Most importantly, the combination of all three proteins induced the reporter gene substantially and consistently with synergistic cooperation between Sp1 and Smad3/Smad4 (Fig. 5B). Promoter constructs harboring TbRE mutations in either the Sp1 recognition sequence or the CAGA box corroborated the specificity of the effect, in addition to documenting the requirement of both binding sites for optimal promoter stimulation (Fig. 5C). Unlike the promoter with the mutated Sp1-binding site, the construct with the mutant CAGA box exhibited higher activation by the Sp1 and Smad3/ Smad4 combination than by Sp1 alone (Fig. 5C).
Dominant-negative Smads and Sp1 Block TGF␤ Stimulation of COL1A2-The use of dominant-negative (DN) forms of transcription factors has proven to be a useful means to study gene regulation in cultured cells (33). We therefore decided to establish clonal cell lines stably transfected with DN forms of Smad3, Smad4, or Sp1 in order to examine the effects of TGF␤ treatment on the expression of the endogenous COL1A2 gene. The DN proteins included Smad variants without the last 39 amino acids, and a truncated molecule containing only the zinc finger (DNA-binding) domain of Sp1 (18,38).
Expression of mutant Smad proteins in the stably transfected cells was monitored by Western blot analysis using the M2 antibody against the Flag tag (Fig. 6A). RNA was purified from TGF␤-treated or untreated cells which were stably trans-fected with DN-Smad3, DN-Smad4, or the empty expression vector. Northern blot hybridizations to COL1A2 and GAPDH cDNAs documented the specific block of TGF␤-induced COL1A2 mRNA accumulation by the DN-Smad proteins (Fig.  6B). Nuclear run-on assays correlated most of the inhibition at the transcriptional level (Fig. 6C).
The ability of DN-Sp1 to interfere with the activity of the wild type protein was assessed in NIH3T3 fibroblasts. Expression of DN-Sp1 in the stably transfected fibroblasts was monitored by EMSA using Sp1 and AP1 recognition sequences. As expected, the intensity of the Sp1 complex was much weaker or nearly absent in two randomly selected DN-Sp1 transfected clones than in control cells; by contrast, the intensity of the AP1 complex was the same in control as in DN-Sp1 transfectants (Fig. 7A). In accordance to the available evidence, the lack of a faster migrating band in the DN-Sp1 extracts may conceivably reflect the block of protein multimerization and DNA binding by the mutant molecule (39). Northern blot analysis of the same stably transfected clones untreated or treated with TGF␤ revealed that the DN-Sp1 protein abrogated COL1A2 stimulation by the cytokine, in addition to reducing the steady-state levels of COL1A2 mRNA in the untreated cells as well (Fig.  7B). This result is consistent with the data of our previous study that have indirectly implicated Sp1 in constitutive and TGF␤-inducible expression of COL1A2 using mithramycin, a broad inhibitor of nuclear protein binding to CG-rich sequences (12). The effect of DN-Sp1 was specific, in that the mutant proteins did not block TGF␤-induced stimulation of the PAI-1 gene (Fig. 7B) (20). The nuclear run-on assay documented that the TGF␤ stimulation of COL1A2 is mostly blocked at the transcriptional level (Fig. 7C).
Aside from mediating the TGF␤ response, we have shown that binding of Sp1 to the TbRE is also implicated in the inhibitory action of tumor necrosis factor ␣ on COL1A2 transcription (26). We therefore reasoned that DN-Sp1 stably transfected fibroblasts may also be unresponsive to TNF␣ down-regulation of COL1A2 expression. Accordingly, we performed an additional Northern analysis of RNA purified from untreated and TNF␣-treated DN-Sp1-transfected cells. The results confirmed the role of Sp1 in integrating the two antagonistic pathways by showing comparable levels of COL1A2 mRNA in TNF␣-treated and untreated cells stably transfected with DNSp1 (Fig. 8A).
In Schneider cells, the wild type COL1A2 promoter construct is activated by Sp1 in the absence of Smad3/Smad4, but not vice versa (Fig. 5A). Likewise, the combination of Sp1 and Smad3/Smad4 stimulate the activity of the COL1A2 promoter with the CAGA mutation, but not of the construct with the Sp1 mutation (Fig. 5C). These results raised the possibility that Sp1 may act downstream of the Smads. To test this postulate, NIH3T3 cells stably transfected with DN-Sp1 were co-transfected with the (AϩB) 3 TK-CAT reporter construct and the Smad3 and Smad4 expression plasmids. We chose the heterologous TK promoter with multiple TbRE copies instead of the less inducible COL1A2 promoter in order to maximize our ability of detecting changes in reporter gene transactivation (see Fig. 4, A and B). Unlike the 2.8-fold induction of the control sample, overexpression of Smad3 and Smad4 in the DN-Sp1 cells had no effect on reporter gene activity (Fig. 8B). The inability of Smads to rescue the block of TGF␤ responsiveness in the DN-Sp1 cells is promoter-specific, in that overexpression of the Smad proteins in the same cells stimulated the cotransfected PAI-1 promoter construct about 2-fold (Fig. 8C). Altogether the results of these transfection experiments demonstrated that both Sp1 and Smad3/Smad4 are required for TGF␤ responsiveness of the COL1A2 promoter, and that the response is the result of promoter-and sequence-specific synergism between the two nuclear factors. DISCUSSION Transcription of the type I collagen gene is modulated by a wide variety of substances, such as retinoic acid, phorbol esters, prostaglandins, glucocorticoids, and cytokines (4). Abnormal regulation of type I collagen expression is responsible for the excessive matrix deposition that characterizes many fibrotic disorders, including scleroderma and liver cirrhosis (2). Cytokines secreted by immunoinflammatory cells at pre-lesional sites are believed to cause deregulated activation of the type I collagen genes in fibrotic conditions (2,4). TGF␤ is a potent stimulator of type I collagen gene activity and an important cytokine involved in ECM formation and remodeling, and in the initiation and progression of fibrosis. Hence, the type I collagen genes represent a useful paradigm to study the signal transduction pathway induced by TGF␤ to stimulate ECM accumulation. Despite this long-established correlation, very little is known about the molecular and cellular mechanisms that mediate the TGF␤-induced stimulation of type I collagen production. Results presented here indicate that synergistic cooperation between Sp1 and Smad3/Smad4 mediates TGF␤ stimulation of COL1A2 transcription.
The work of Dennler et al. (20) on TGF␤ stimulation of the PAI-1 gene was the first to suggest that Smad proteins may be involved in COL1A2 modulation, as well. There are, however, important differences in the organization of the respective TGF␤-responsive elements. The TGF␤-responsive element of the PAI gene contains three CAGA boxes; two of them have the same sequence, AGACAGACA, and the third, AGCCAGACA, differs from the other for a single nucleotide (20). There is only one CAGA box in the COL1A2 element and of as yet different composition, ATGCAGACA. We have shown here that the single and slightly different CAGA box of COL1A2 is also a Smadbinding site.
Bacterially expressed Smad3 and Smad4 proteins bind to three copies of the CAGA box of the COL1A2 promoter, and binding of Smad3 and Smad4 to (CAGA) 3 is induced in TGF␤treated fibroblasts. At variance with the last result was our inability to detect Smad3/Smad4 binding to the TbRE using the same anti-Smad antibodies. On the other hand, we did observe binding of both Sp1 and Smad3/Smad4 to the TbRE with nuclear extracts prepared from untreated and TGF␤-treated cells transiently transfected with Smad3 and Smad4 expression vectors. We have no explanation for the failure of detecting Smad3/Smad4 in the untransfected cells; among other possibilities, it might be due to the low amount of endogenous Smad proteins. Although Smad3 and Smad4 bind to (CAGA) 3 , Smad3 is probably the only component of the Smad3-Smad4 complex that actually binds to the TbRE. This suggestion rests solely on the finding that Sp1 and Smad3 are the only proteins detected in the TbRE-bound complex which was assembled in the presence of recombinant Sp1 and recombinant Smad4 and Smad3 without MH2 domains. With this limitation in mind, it is tempting to speculate that Smad4 may bind to the TbRE under different conditions. For example, TNF␣ inhibition of COL1A2 transcription is also accompanied by increased intensity of the TbRC (26). Conceivably, the change could be related to the binding of Smad4 in association with other Smad partner(s). Work in progress is examining this and other possibilities.
Following the original suggestion of Dennler et al. (20) Chen et al. (25) reported that transient overexpression of Smad3 and Smad4, but not Smad1 or Smad2, stimulates transcription of a reporter gene driven by the Ϫ772 COL1A2 promoter. The authors did not, however, analyze if Smad3/Smad4 transactivates through the CAGA box of the TbRE, how Smad3/Smad4 affects endogenous COL1A2 expression, and whether Smad3/Smad4 may interact with other nuclear factor(s) to stimulate transcription (25). While our work was under review, Chen et al. (40) reported that optimal TGF␤ stimulation of the COL1A2 promoter transiently transfected in cultured cells depends on the integrity of the Smad3/Smad4-binding site in the TbRE. Here we have independently confirmed that Smad3 and Smad4 stimulate transactivation of the shorter Ϫ378 COL1A2 promoter. We have also associated COL1A2 transactivation specifically to the binding of Smad3/Smad4 to the CAGA box of the TbRE. Moreover, we have further proven the importance of Smad3 and Smad4 by documenting the inability of endogenous COL1A2 to respond to TGF␤ stimulation in cells that are stably transfected with DN-Smad3 or DN-Smad4 expressing plasmids.
Smad3 involvement in the TGF␤ signaling pathway that induces ECM accumulation is in apparent contradiction with recent findings using Smad3 Ϫ/Ϫ mice (41). Although the mutant animals produce less matrix, this problem can be reversed by administration of exogenous TGF␤; additionally, the null mice exhibit accelerated wound healing (41). There is, however, an important difference between the two studies. In ours, we have examined the COL1A2 response to TGF␤ in cultured fibroblasts, whereas the analysis of the Smad3 mutant mice involved a complex network of signals among epithelial cells, keratinocytes, fibroblasts, and monocytes. It is therefore plau-sible that TGF␤ signal may act through redundant pathways in the Smad3 Ϫ/Ϫ mice. An illustrative example of redundant signaling is the TAK1-MKK6-p-38K pathway which has been recently shown to mediate in part the growth inhibitory effects of TGF␤ (42).
Our study has also re-emphasized the critical importance of Sp1 in COL1A2 gene regulation. Previous work has shown that mutations in the Sp1-binding sites of the TbRE are associated with reduced COL1A2 promoter activity, a finding later supported by the lower COL1A2 expression in cells treated with mithramycin (11,12). This conclusion was confirmed here by associating the decrease of COL1A2 mRNA levels with the stable expression of DN-Sp1. We have also shown that overexpression of Smad3/Smad4 cannot activate TbRE-driven transcription in DN-Sp1-transfected cells. Two lines of evidence indicate that the requirement of Sp1 is promoter-specific. First, the PAI-1 promoter is activated by overexpression of Smad3 and Smad4 in cells stably transfected with the DN-Sp1 plasmid; second, steady-state levels of PAI-mRNA increase after TGF␤ treatment of the same stably transfected cells. Altogether, the results suggest that Sp1 probably lies downstream of Smad3/Smad4 in the TGF␤ pathway leading to type I collagen production. We also demonstrated that transient overexpression of Smad3 and Smad4 in Schneider cells has a synergistic effect on Sp1-induced TbRE-driven transcription. A similar finding has been also reported for the p21 gene (24). There are, however, mechanistic differences between how these nuclear factors interact on the p21 versus the COL1A2 promoter.
Synergy between Sp1 and Smad3/Smad4 was demonstrated despite the fact that the TGF␤-responsive element of the p21 gene contains multiple Sp1-binding sites with no CAGA box or other known Smad-binding sites (24). Synergy was also observed by co-transfecting Smad3, Smad4, and the GAL4-Sp1 fusion protein together with a reporter gene under the control of the GAL4-binding site (24). Among other alternatives, the authors suggested that Smad proteins may transactivate the p21 promoter by functionally interacting with Sp1 (24). In the case of the COL1A2 promoter, Smad3 and Smad4 cannot turn on transcription without Sp1 in either Schneider cells or fibroblasts stably transfected with DN-Sp1. Moreover, synergy between Sp1 and Smad3/Smad4 is greatly reduced in Schneider cells co-transfected with the reporter plasmid harboring the mutant CAGA box, and is totally eliminated in those co-transfected with the construct driven by the mutant Sp1-binding site. Furthermore, the Smad3-Smad4 complex cannot by itself transactivate the collagen promoter. Finally, recombinant Smad3 and Smad4 do not bind to the TbRE in the absence of recombinant Sp1. Altogether, these findings indicate that Smad3/Smad4 activity is dependent on the binding of Sp1 to the TbRE, and suggest that synergy may in part be due to physical interaction between these factors. They also raise the possibility that Sp1 binding to box 3A may be a prerequisite for Smad3/Smad4 binding to the CAGA box. Future analyses will test the validity of these hypotheses.
Sp1 is also important for TNF␣ down-regulation of COL1A2. First, the responsive element of TNF␣ includes the TbRE, and thus contains the Sp1-binding sites of box 3A (26). Second, COL1A2 down-regulation by TNF␣ is abolished in cells stably transfected with DNSp1 (Fig. 8). Thus, there is indirect support for the original observation that Sp1 is the point of convergence of the TGF␤ and TNF␣ signaling pathways on the COL1A2 gene (26). Our recent work has demonstrated the importance of C/EBP proteins in the TNF␣ down-regulation of the COL1A2 gene (33). It is therefore possible that Sp1 may stabilize the C/EBP complex by interacting with these factors, an idea in line with previous evidence of functional interaction between Sp1 and C/EBP-␤ in other genes (43,44). Additionally, TNF␣ may also reverse the transcriptional stimulation of the TbRC through an undefined mechanism. Possible alternatives include changing the Smad4 partner to an inhibitory Smad, modifying the complex post-transcriptionally or a combination of both mechanisms (15). Along these lines, a recent report has implicated activation of the NFB/RelA-dependent pathways by a variety of pathogenic and proinflammatory stimuli in the suppression of TGF␤ signaling through Smad7 activation (45).
Although the underlying mechanism remains unclear, convergence between different signaling pathways is not unique to the COL1A2 gene. There are in fact many other examples of cross-talk between signaling pathways that exert agonistic or antagonistic effects on gene expression. Examples of signal integration involving the Smads include TGF␤ and vitamin D, and TGF␤ and interferon ␥, as well as the repression of TGF-␤ anti-mitogenic responses by oncogenic Ras (46 -48). In our model of ECM accumulation, Sp1 is the common integrator of the cytokine-induced responses and the C/EBPs and Smads are the components selectively induced in each pathway. Work in progress is testing this model, in addition to delineating how Sp1 integrates the antagonistic signals induced by TGF␤ and TNF␣ in order to modulate COL1A2 transcription in opposite ways.