Requirement of Smad3 and CREB-1 in Mediating Transforming Growth Factor-β (TGFβ) Induction of TGFβ3 Secretion*

Because increased transforming growth factor-β (TGFβ) production by tumor cells contributes to cancer progression through paracrine mechanisms, identification of critical points that can be targeted to block TGFβ production is important. Previous studies have identified the precise signaling components and promoter elements required for TGFβ induction of TGFβ1 expression in epithelial cells (Yue, J., and Mulder, K. M. (2000) J. Biol. Chem. 275, 30765–30773). To determine how regulation of TGFβ3 expression differs from that of TGFβ1, we identified the precise signaling pathways and transcription factor-binding sites that are required for TGFβ3 gene expression. By using mutational analysis in electrophoresis mobility shift assays (EMSAs), we demonstrated that the c-AMP-responsive element (CRE) site in the TGFβ3 promoter was required for TGFβ-inducible TGFβ3 expression. Electrophoresis mobility supershift assays indicated that CRE-binding protein 1 (CREB1) and Smad3 were the major components present in this TGFβ-inducible complex. Furthermore, by using chromatin immunoprecipitation assays, we demonstrated that CREB-1, ATF-2, and c-Jun bound constitutively at the TGFβ3 promoter (–100 to +1), whereas Smad3 bound at this site only after TGFβ stimulation. In addition, inhibition of JNK and p38 suppressed TGFβ induction of TGFβ3 transactivation, whereas inhibition of ERK and protein kinase A had no effect. Small interfering RNA-CREB1 and small interfering RNA-Smad3 significantly inhibited TGFβ stimulation of TGFβ3 promoter reporter activity and TGFβ3 production. Our results indicate that TGFβ activation of the TGFβ3 promoter CRE site, which leads to TGFβ3 production, is required for TGFβRII, JNK, p38, and Smad3 but was independent of protein kinase A, ERK, and Smad4.

TGF␤ 2 is the prototype of a large superfamily of multifunctional cytokines that are differentially expressed and function in a wide range of target cells (1). TGF␤ suppresses the proliferation of normal cells. However, in malignant cells TGF␤ production often enhances tumor progression, especially once the cells have lost the negative growth control signals imparted by TGF␤. Therefore, identification of critical points that can be targeted to block TGF␤ production in cancer cells is important (2,3). A better understanding of the mechanisms underlying TGF␤ auto-regulated secretion will assist in the identification of these critical branch points.
The TGF␤ family has the following three homologous forms: TGF␤1, TGF␤2, and TGF␤3. Although the three TGF␤s share 60 -80% identity, they are encoded by distinct genes, and their expression is controlled by a different regulatory sequence or promoter (4 -6). They also show different physiological and pathological activities in certain cell types and systems (7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17). Because the expression of TGF␤ is controlled by an auto-feedback loop (18), TGF␤ is an important mechanism for amplifying its own expression and production. Because of the critical role that TGF␤ plays in regulating a broad range of physiological and pathological effects, such as proliferation, cell cycle arrest, apoptosis, angiogenesis, metastasis, and invasiveness in cells and tissues, any changes in the signaling pathways mediating TGF␤3 autoregulation may lead to severe abnormalities. It is therefore important to understand the mechanisms underlying TGF␤ autoregulation, as well as the signal transduction pathways mediating TGF␤-induced TGF␤3 expression.
The autoregulation of TGF␤ appears to be mediated through specific sites in the promoter regions of the distinct isoforms (19). Although these sites have been examined for the TGF␤1 promoter (19), little is known about the relevant regions in the TGF␤3 promoter. The promoter area of the TGF␤3 gene has little structural or functional similarity to the TGF␤1 promoter. Similarly, transcriptional regulation of this gene has significantly diverged from that of the other TGF␤s (6). It has been reported that a region proximal to the TATA box in the 5Ј-flanking region of the TGF␤3 gene might be important in regulating TGF␤3 expression (6), but no conclusive studies have defined the critical sites required for TGF␤ induction of TGF␤3 expression.
Here we have determined that the TGF␤-inducible region of the TGF␤3 promoter contains three transcription factor binding consensus sites: a CRE at Ϫ45 to Ϫ39 (GACGTCA), an SBE at Ϫ49 to Ϫ46 (CAGA), and an activator protein-2 (AP-2) site at Ϫ57 to Ϫ50 (CCCCAGGC). We have investigated the transcriptional regulation of each of these TGF␤3 gene promoter regions in response to TGF␤ exposure, and we have demon-strated that the CRE site is the critical site. We have also defined the signal transduction pathways mediating TGF␤ induction of TGF␤3 secretion, and we demonstrate that Smad3 and CREB-1 are the critical transcription factors involved. Although we have shown previously that JNKs and ERKs were the critical mediators of TGF␤1 production in untransformed epithelial cells (19), here we show that JNKs and p38, but not PKA and ERKs, are the critical upstream mediators of TGF␤3 production. Our studies provide new insights into the TGF␤ auto-feedback regulatory mechanisms that are required for one of the key biological responses to TGF␤, namely stimulation of its own secretion. In this case, however, the results are specific for production of TGF␤3. Furthermore, our results demonstrate that blocking these critical points reduces TGF␤3 production, and as such, our findings suggest important intervention strategies for controlling tumor growth mediated by excessive tumor cell-secreted TGF␤3 in the tumor microenvironment.
Electrophoretic Mobility Shift Assays (EMSAs)-EMSAs were performed as described (24). Nuclear protein extracts were prepared from either IEC4-1 or CCL64 cells by a method described previously (24). Briefly, the cells were cultured in 10-cm dishes and treated with TGF␤3 (10 ng/ml) for 1 h. The cells were then disrupted in 500 l of lysis buffer A (25 mM HEPES (pH 7.8), 50 mM KCl, 0.5% Nonidet P-40, 100 M dithiothreitol, 10 g/ml leupeptin, 25 g/ml aprotinin, and 1 mM phenylmethylsulfonyl fluoride). After a 1-min centrifugation (16,000 ϫ g, 4°C), the pellet containing the nuclei was washed once with 500 l of buffer B (buffer A without Nonidet P-40), resuspended in 120 l of extraction buffer (buffer B but with 500 mM KCl and 10% glycerol), and incubated with shaking at 4°C for 30 min. The nuclear extracts were stored at Ϫ70°C until analysis. The DNA binding reaction (for EMSA) was carried out at room temperature for 30 min in a mixture containing 4 g of nuclear protein, 1 g of poly(dI-dC), and 15,000 cpm of 32 P-labeled doublestranded oligonucleotide. The samples were fractionated through a 5% polyacrylamide gel. Gels were dried and exposed to x-ray film for further analysis.
Chromatin Immunoprecipitation (ChIP) Assays-ChIP assays (25,26) were performed using the ChIP assay kit from Upstate Biotechnology, Inc. (Lake Placid, NY). In brief, HaCaT cells were incubated in the absence or presence of TGF␤3 (10 ng/ml) for 1 h. The cells were then cross-linked with formaldehyde. Nuclear extracts were sonicated to obtain ϳ250-bp fragments of DNA. The resulting DNA was immunoprecipitated overnight with antibodies to CREB-1, CREB-2, ATF-2, c-Jun, JunD, Smad3, or Smad4 (Santa Cruz Biotechnology) and salmon sperm/agarose beads (catalog number 16-157; Upstate Biotechnology, Inc.). The resulting DNA sample incubated with IgG or with no antibody was used as a negative control. The supernatant from a sample was used as total DNA input. The DNA was dissolved in 30 l of water, and 2 l was employed for PCR analysis using the primers for the TGF␤3 promoter region containing AP-2, SBE, and CRE sites (5Ј-GTGCAGCAAAAGAGGCTGCG-TGCG-3Ј and 5Ј-TCTATTTCTCTCTGCTGAAAT-3Ј). The products were resolved in a 3% agarose gel and visualized with ethidium bromide.
siRNA Transfections-siRNA-CREB-1 was purchased from Santa Cruz Biotechnology. siRNA-Smad3 was purchased from Cellogenetics, Inc. (Baltimore). The siRNA products were annealed duplexes of RNA. Transfection of siRNAs into cells was performed according to the manufacturer's instructions. A nontargeting siRNA was used as a control for nonsequencespecific effects of the transfected siRNAs (Dharmacon, Inc., Dallas).
Construction of TGF␤3 Promoter Luciferase Reporter Plasmid-The T␤3-P221/110-Luc encodes a TGF␤3 promoter region from Ϫ221 to ϩ110. The nucleotide sequence was PCRamplified. The primers used were as follows: sense, 5Ј-actc-gaACGCGTtgtggcaggagtgattccaaga-3Ј (the capital letters indicate an MluI restrict enzyme digestion site), and antisense, 5Ј-gcacgcAGATCTcttggacttgactctctgctcc-3Ј (the capital letters indicate a BglII restrict enzyme digestion site). The PCR products and the basic pGL3 luciferase reporter vector (Promega, Madison, WI) were digested using MluI and BglII, and the products were gel-purified and inserted into the vector using Quick Ligase (New England Biolabs, Beverly, MA). The sequence of the resulting T␤3-P221/110-Luc luciferase reporter plasmid was confirmed by sequencing in both directions. The
Real Time Quantitative RT-PCR-RNA was isolated from HaCaT cells and their siRNA-CREB-1 and siRNA-Smad3 transfectants as described above using TRIzol reagent by Invitrogen. Total cDNA was prepared by RT-PCR using random hexamer primers purchased from Invitrogen and the isolated RNA as the template. Real time RT-PCR was performed using SYBR Green I agent from Qiagen Inc. (Valencia, CA), with the cDNA prepared above as the template. 18 S rRNA was used as an internal control. The TGF␤3 primers used were as follows: sense, 5Ј-TGTGTGCGCCCCCTCTA-3Ј; antisense, 5Ј-GGTTCATGGACCCACTTCCA-3Ј.
Quantitation of TGF␤3-The quantitative determination of TGF␤3 was performed. IEC4-1 cells, and their siRNA-CREB-1 or siRNA-Smad3 transfectants, were seeded in 12-well plates and grown in the absence or presence of TGF␤1 for 72 h. Conditioned medium (CM) was then collected for quantitative determination of TGF␤3 using an ELISA kit (DuoSet ELISA Development System, R & D Systems, Inc., Minneapolis, MN). The system specifically detects TGF␤3 with no cross-reactivity or interference with TGF␤1. The cells were trypsinized and counted. "Total" TGF␤3 was quantitated according to the manufacturer's instructions. Briefly, the samples were acid-activated using 1 N HCl to activate latent TGF␤3 and then neutralized by addition of 1.2 N NaOH, 0.5 M HEPES solution to convert all TGF␤3 to the form detectable using the antibody in the ELISA kit.
Statistical Analysis-Data were analyzed and statistically significant differences determined using the Student's t test. The results are expressed as the mean Ϯ S.E.

The CRE Consensus Site in T␤3-P61/35 Is Critical for TGF␤3
Responsiveness-It has been reported that a sequence from Ϫ490 to ϩ100 in the 5Ј-flanking region of the TGF␤3 gene might be important in regulating TGF␤3 expression (6). This region contains two ubiquitously expressed Sp-1 transcription factor DNA-binding sites (Ϫ420 to Ϫ415 and Ϫ363 to Ϫ358), an AP-2 site (Ϫ57 to Ϫ50), and a CRE site (Ϫ45 to Ϫ39). The CRE site has been shown to be involved in TGF␤3 transcription, although its TGF␤-inducible expression was not examined (6). In addition to the AP-2 and CRE sites, we also identified an SBE site (Ϫ49 to Ϫ46), between the AP-2 and CRE sites, which can also be induced by TGF␤ (27). We presumed that the SBE and/or CRE sites, but not the Sp-1 sites, in the TGF␤3 promoter region might be important for TGF␤ induction of TGF␤3 gene expression because Sp-1 has relatively little effect on the activation of the promoter (6). To address this issue, we constructed a T␤3-P221/110-Luc luciferase reporter plasmid using the DNA sequence from the TGF␤3 5Ј-flanking region (Ϫ221 to ϩ110) of the promoter. This sequence contains the AP-2, SBE, and CRE sites. We treated IEC4-1 cells with TGF␤ to stimulate the auto-induction loop, and we measured T␤3-P221/110-Luc reporter activity. Here we show that TGF␤ affected a time-dependent increase in T␤3-P221/110-Luc activity. The TGF␤ stimulation of T␤3-P221/110-Luc activity reached a peak 24 h after TGF␤3 treatment (Fig. 1A). These results indicate that T␤3-P221/110 is important in TGF␤ autoinduction of TGF␤3 gene transcription.
The IEC4-1 cells were used for these studies because they represent a good model for TGF␤ regulation in untransformed epithelial cells because of their high level of TGF␤ sensitivity and their well characterized TGF␤ pathways (21,22,28,29). In addition, these cells have been shown to produce high levels of endogenous TGF␤ (20).
To determine whether the AP-2 (Ϫ57 to Ϫ50), SBE (Ϫ49 to Ϫ46), and/or CRE (Ϫ45 to Ϫ39) sites within the Ϫ221 to ϩ110 region of the TGF␤3 promoter were involved in TGF␤ induction of TGF␤3 gene transcription, we examined the DNA binding activity of a sequence containing these three sites (T␤3-P61/35) in response to TGF␤3. As shown in Fig. 1B, TGF␤ significantly increased complex formation at T␤3-P61/35 in a time-dependent manner, with peak levels occurring 1 h after TGF␤ treatment. Thus, our results indicate that a TGF␤-inducible complex formed at T␤3-P61/35 in response to TGF␤ treatment.
During the TGF␤ signaling process, TGF␤ binds to RII, which then recruits RI into the active complex (30). To investigate whether the TGF␤ receptors were required for auto-regulation of TGF␤3 expression, we performed EMSAs with T␤3-P61/35 after expression of a dominant-negative TGF␤ receptor II (DN RII) (22). As shown in Fig. 2A, the T␤3-P61/35 DNA binding activity stimulated by either TGF␤1 or TGF␤3 was significantly suppressed in the DN RII-expressing IEC4-1 cells ( Fig. 2A, lanes 6 and 7 versus lanes 2 and 3). In contrast, DN RII had no effect in the PMA-induced cells ( Fig. 2A, lane 8 versus lane 4). These results indicate that RII is required for TGF␤3 auto-loop regulation.
To determine which DNA-binding sites are required for TGF␤3 production, we performed site-directed mutagenesis at the relevant sites, and we then repeated the EMSAs using labeled versions of these as probes. As shown in Fig. 2B, mutation of the CRE site completely abolished DNA binding activity at T␤3-P61/35 (lanes 10 -12), whereas mutation of the AP-2 (lanes 4 -6) or SBE (lanes 7-9) sites had no effect. The TGF␤inducible complex formed at the wild-type T␤3-P61/35 is shown for comparison (Fig. 2B, lanes 2 and 3). These results indicate that DNA binding activity at the CRE site is critical for mediating the TGF␤3 auto-induction loop. As shown in Fig. 2B (lanes 1, 4, 8, and 10), a 10-fold excess of unlabeled probe completely competed for DNA binding of the radioisotope-labeled probe, indicating that the bands observed represented specific binding to T␤3-P61/35.
To further confirm that the CRE site in the TGF␤3 promoter is indeed required for the TGF␤3 gene transcription, we constructed Mut-T␤3-P221/110-Luc, a CRE site-mutated T␤3-P221/110-Luc luciferase reporter plasmid. We transfected IEC4-1 cells with Mut-T␤3-P221/110-Luc, followed by treatment of the IEC4-1 transfectants with TGF␤. As shown in Fig.  2C, CRE mutation completely abolished the luciferase activity induced by TGF␤. This result confirms that this CRE site in the TGF␤3 promoter region is critical for TGF␤3 gene transcription.
CREB-1 and Smad3, but Not Smad4, Are Bound at the T␤3-P61/35-It is known that CREB-1 can recognize and bind to CRE consensus sites in various gene promoters. As we have demonstrated that the CRE site in T␤3-P61/35 is critical for TGF␤ auto-induction of TGF␤3, we examined whether the CREB-1 protein was a component in the transcription factor complex bound at this T␤3-P61/35. As shown in Fig. 3A by EMSAs, TGF␤-inducible complex formation (lanes 1 and 2) was reduced when the CREB-1 antibody was present (lane 4) but not when rabbit IgG was present (lane 3). A supershift was also observed using the CREB-1 antibody (Fig. 3A, lane 4). Thus, CREB-1 was in the transcription factor complex bound to the T␤3-P61/35 region.
We further explored whether Smad3 and Smad4 might be present in the complex at T␤3-P61/35, because Smads are known to play important roles in TGF␤3 signaling and because an SBE site is present in T␤3-P61/35. As shown in Fig. 3A, lane 5, the anti-Smad3 antibody reduced the TGF␤-inducible DNA binding, indicating that Smad3 was present in the transcription factor complex bound to T␤3-P61/35. This did not occur when phorbol myristate acetate was used as the inducer (data not shown). Because activated Smad2 or Smad3 can form a dimer with Smad4 prior to stimulation of transcriptional activation of target genes through SBE sites, it was of interest to determine whether Smad4 was also in this complex bound to T␤3-P61/35. As shown in Fig. 3A, lane 6, the anti-Smad4 antibody had no effect on DNA binding of T␤3-P61/35, suggesting that Smad4 may not be involved in the transcription factor complex bound to T␤3-P61/35. To verify the specificity of the anti-Smad3 and anti-Smad4 antibodies used, a full palindromic SBE sequence was used in the supershift EMSAs. As shown in Fig. 3B, a 10-fold excess of unlabeled cold probe completely competed for TGF␤ induction of the DNA binding of the radioisotopelabeled probe, indicating that the band observed was specific for the consensus palindromic SBE site (lane 1). Both Smad3 and Smad4 were present in the transcription factor complex bound to the consensus SBE site (Fig. 3A, lanes 5 and 6). The results in Fig. 3A, lane 2, indicate that in the absence of TGF␤, there was no binding to the consensus palindromic SBE site in IEC4-1 cells, whereas TGF␤ could induce DNA binding at this site, as shown in lane 3 (Fig. 3B). As shown in Fig. 3B (lane 4), addition of rabbit IgG in the EMSAs had no effect on TGF␤ induction of DNA binding at the consensus palindromic SBE site. Overall, our results indicate that the anti-Smad3 and anti-Smad4 antibodies used (Fig. 3A) were capable of competing effectively for DNA binding at the consensus SBE site in the EMSA supershift assays.
Smad3 Is Bound at the T␤3-P61/35 Only in Response to TGF␤ Stimulation-Because we have demonstrated that both CREB-1 and Smad3 were required for DNA binding at T␤3-P61/35 in response to TGF␤ treatment, we performed ChIP analyses to confirm these findings, as well as to explore which other potential transcription factors bind to this TGF␤3 promoter region. It has been reported that CREB-1, ATF-2, and c-Jun are able to bind to consensus CREs (20,31,32). In addition, CREB-2 is a member of the CREB family, and JunD is a close homologue of c-Jun, but it was not clear whether these transcription factors were effective in TGF␤ induction of DNA binding at the CRE site in the TGF␤3 promoter (Ϫ45 to Ϫ39). Therefore, we examined whether CREB-1, CREB-2, ATF-2, c-Jun, JunD, Smad3, and/or Smad4 could bind to this TGF␤3 promoter region (Ϫ45 to Ϫ39). To determine the transcription factors bound to the endogenous TGF␤3 promoter, we performed ChIP analyses using noncancerous TGF␤-responsive HaCaT human keratinocytes. As shown in Fig. 3C, in the absence of TGF␤, the transcription factors that bound to the CRE site in the TGF␤3 promoter (Ϫ45 to Ϫ39) were CREB-1, ATF-2, and c-Jun. In contrast, CREB-2, JunD, Smad3, and Smad4 were not present in the complex bound to this CRE site under these conditions (Fig. 3C, upper panel). After addition of TGF␤, however, the DNA binding activity of CREB-1 and ATF-2, but not that of c-Jun, was increased (Fig. 3C, lower  panel). Most significantly, Smad3 was detectable in the complex bound at the CRE site in the TGF␤3 promoter (Ϫ45 to Ϫ39) only in response to TGF␤ treatment (Fig. 3C, lower  panel). Thus, although CREB-1, ATF-2, and c-Jun all bound constitutively to the CRE site in the TGF␤3 promoter (Ϫ45 to Ϫ39), CREB-1 and ATF-2 binding was also inducible by TGF␤, and Smad3 only bound to the TGF␤3 promoter CRE after TGF␤ treatment. These results suggest that Smad3, as well as the CREB/ATF family members CREB-1 and ATF-2, may play a role in TGF␤ induction of TGF␤3 promoter DNA binding activity.
Smad3 Is Required for TGF␤ Induction of TGF␤3 Promoter DNA Binding Activity-We have demonstrated above that Smad3 is present in the transcription factor complex bound to T␤3-P61/35 after induction by TGF␤. Because there is an SBE site upstream of the CRE site in T␤3-P61/35, which was not required for TGF␤-inducible DNA binding to T␤3-P61/35 (Fig.  2B), we employed CCL64-Smad3C cells, and their parental CCL64-L20 cells, in order to confirm whether Smad3 was required for TGF␤-inducible complex formation at the T␤3-P61/35. CCL64-Smad3C cells stably express a dominant-negative form of Smad3 (33). In this cell line, TGF␤ is unable to induce phosphorylation of Smad3, and the inhibitory effect of TGF␤ on cell growth is blocked (33). Thus, normal Smad3 function in this cell line is lost. As shown in Fig. 4A by EMSAs, TGF␤ induction of complex formation at T␤3-P61/35 was blocked in the CCL64-Smad3C cells, compared with control L20 cells (lane 5 versus lane 2). In contrast, the PMA-induced complex formation at T␤3-P61/35 was not diminished, when compared with control L20 cells (Fig. 4A, lane 6 versus lane 3). Thus, Smad3 is required for TGF␤ induction of DNA binding at the CRE site of the TGF␤3 promoter region.
We also examined TGF␤ induction of T␤3-P221/110-Luc luciferase activity after transfection of L20 or Smad3C cells with T␤3-P221/110-Luc. As shown in Fig. 4B, TGF␤ induced a 5.9fold increase of T␤3-P221/110-Luc luciferase activity in the L20 cells, but only a 1.6-fold increase in the Smad3C cells. Statistic  1, 4, 7, and 10). TGF␤3-inducible DNA binding to the TGF␤3 promoter was abolished when mCRE was used as the probe (lanes 10 -12) but not when mAP-2 (lane 4 -6) or mSBE (lanes 7-9) was used as probes. Three independent experiments were carried out and representative figures are shown. C, IEC4-1 cells were seeded in 24-well plates and grown to 70 -80% confluence. Cells were transfected with 0.4 g of T␤3-P221/110-Luc or Mut-T␤3-P221/110-Luc and 0.1 g of Renilla luciferase control reporter (pRL-SV40) per well using Lipofectamine TM 2000 (Invitrogen), as described in the user manual. 24 h after transfection, the cells of the TGF␤3 group were exposed to TGF␤3 (10 ng/ml), whereas the control group was not treated. The cells were harvested another 24 h later, and luciferase assays were performed as described under the "Materials and Methods." Data are plotted as mean Ϯ S.E. of triplicate samples for each of three independent experiments. OCTOBER 6, 2006 • VOLUME 281 • NUMBER 40

JOURNAL OF BIOLOGICAL CHEMISTRY 29483
analyses indicated a significant difference between the fold changes of the two cell groups (p Ͻ 0.01). Collectively, our results demonstrate that Smad3 not only binds to T␤3-P61/35 but is also functionally required for TGF␤3 autoregulation.
Because Smad4 plays an important role in mediating Smad-dependent TGF␤ responses, usually by forming dimers with Smad3 or Smad2, it was conceivable that Smad4 might be involved in the TGF␤ induction of T␤3-P221/110-Luc luciferase activity. Accordingly, we co-transfected a DN-Smad4 plasmid with T␤3-P221/ 110-Luc into IEC4-1 cells. As shown in Fig. 4C, DN-Smad4 expression did not block T␤3-P221/110-Luc activity either in the absence or presence of TGF␤. Thus, Smad4 is not required for TGF␤ induction of T␤3-P221/110-Luc luciferase activity, suggesting that Smad4 is not required for the TGF␤ induction of TGF␤3 transcription mediated through Smad3.
JNK and p38, but Not PKA or the ERKs, Are Required for TGF␤3 Autoregulation-We have demonstrated previously that TGF␤ induction of DNA binding to T␤3-P61/35 occurs through the TGF␤ receptors ( Fig. 2A). Accordingly, it was of interest to explore whether downstream signaling events were also involved in TGF␤ induction of TGF␤3 transcription. Furthermore, it is known that mitogen-activated protein kinases (MAPKs) and PKA are upstream kinases of CREB-1 (34 -36). Upon activation, these kinases translocate to the nucleus where they phosphorylate CREB-1 and initiate its DNA binding to the consensus CRE site. To determine whether JNK, ERK, p38, or PKA were required for TGF␤3 auto-regulation, we employed the following specific inhibitors: SP600125 as a selective JNK inhibitor, SB203580 as a selective p38 inhibitor, PD98059 as a selective MEK inhibitor, and H89 as a selective inhibitor of PKA. Our results indicate that SP600125 and SB203580 effectively suppressed TGF␤ induction of T␤3-P61/35 DNA binding activity (Fig. 5A) and transactivation (Fig. 5B) in a dose-dependent manner. However, PD98059 affected neither DNA binding to T␤3-P61/35 nor T␤3-P221/110-Luc activity (Fig. 5,  A and B) at concentrations previously demonstrated to be effective in blocking TGF␤ activation of ERKs in IEC4-1 cells (37). All three inhibitors had little effect on basal T␤3-P61/35 DNA binding activity (data not shown).
Most unexpectedly, H89 (10 M) had no effect on either DNA binding to T␤3-P61/35 or T␤3-P221/110-Luc luciferase activity, either in the absence or presence of TGF␤ (Fig. 5, C and  D). H89 has been shown previously to be effective in selectively blocking PKA in several cell types over the concentration range of 5-20 M (38 -42). Thus, a concentration of 10 M H89 would be expected to completely block PKA activity in our system. Therefore, our results suggest that the signaling pathways mediating TGF␤3 auto-regulation are independent of PKA.

siRNA-CREB-1 and siRNA-Smad3 Both Suppress TGF␤ Induction of T␤3-P61/35 DNA Binding and T␤3-P221/110-Luc
Luciferase Activity-Because we have shown that CREB-1 and Smad3 were both present in the complex at T␤3-P61/35, it was FIGURE 3. CREB-1 and Smad3 are the primary components present in the transcription factor complex binding at T␤3-P61/35. Nuclear protein was extracted from IEC4-1 cells, and EMSA supershift assays using the antibodies (Ab) indicated were performed as described under the "Materials and Methods." A, cells were incubated in the absence or presence of TGF␤3 (10 ng/ml). The probe used was T␤3-P61/35, and the arrow indicates the supershifted band. B, cells were incubated in the absence or presence of TGF␤3 (10 ng/ml). The probe used was the full palindromic SBE sequence as described under the "Materials and Methods." C, ChIP assays were performed as described under the "Materials and Methods." The antibodies indicated were incubated with cross-linked DNA isolated from HaCaT cells that had been treated in the absence (upper panel ) or presence (lower panel ) of TGF␤. The TGF␤3 promoter region Ϫ100 to ϩ1 was amplified using the template ChIP DNA obtained with the antibodies indicated. of interest to confirm our results using siRNA-CREB-1 and siRNA-Smad3 transfectants in the EMSA and luciferase reporter experiments. Commercially available double-stranded siRNA was purchased and transfected into IEC4-1 cells to block target gene protein expression. As shown in Fig. 6A, both CREB-1 and Smad3 were detectable using either an anti-CREB-1 (lane 1, 1st panel) or an anti-Smad3 antibody (lane 1, 3rd panel) . Fig. 6A, lane 2 of the 1st and 3rd panels, indicates that the scrambled control siRNAs had no effect on expression levels of either CREB-1 (lane 2, 1st panel) or Smad3 (lane 2, 3rd panel). Fig. 6A (lane 3, 1st panel) also indicates that the expression levels of CREB-1 in IEC4-1 cells transfected with siRNA-CREB-1 were significantly suppressed compared with untransfected IEC4-1 cells (Fig. 6A, lane 1) or to IEC4-1 cells transfected with the scrambled control siRNA (lane 2). In contrast, as expected, siRNA-Smad3 did not suppress CREB-1 levels (Fig. 6A, lane 4, panel 1). Fig. 6A (lane 3) also demonstrates that siRNA-Smad3 completely blocked Smad3 expression (lane 4), whereas siRNA-CREB-1 (lane 3) was ineffective. ␤-Actin was used as a loading control (Fig. 6A, 2nd and 4th panels). Thus, siRNA-CREB-1 and siRNA-Smad3 could effectively block the expression of their respective target proteins.
CREB-1 is activated by phosphorylation at Ser-133, which is critical for its biological function. Although we have demonstrated that siRNA-CREB-1 effectively suppressed CREB-1 protein expression (Fig. 6A), it was of interest to further investigate whether siRNA-CREB-1 suppressed TGF␤ induction of CREB-1 phosphorylation specifically at Ser-133. Accordingly, we incubated IEC4-1 cells and their respective transfectants with or without TGF␤ and performed Western immunoblotting by using an antibody specific for the activated form of phospho-CREB-1 (Fig. 6B, upper panel). As shown in Fig. 6B, siRNA-CREB-1 significantly blocked both basal and TGF␤-induced phosphorylation of CREB-1 at Ser-133. Densitometric scanning of the phospho-CREB-1 bands from Western analysis demonstrated that TGF␤ could induce a 3-fold increase in phosphorylation of CREB-1 in IEC4-1 cells or IEC4-1 cells transfected with the siRNA control (Fig. 6B, lower panel). However, TGF␤ only induced a 1.4-fold increase of phosphorylation in CREB-1 in IEC4-1 cells transfected with siRNA-CREB-1, indicating that siRNA-CREB-1 suppressed TGF␤ induction of phosphorylation of the protein. Interestingly, siRNA-Smad3 also suppressed TGF␤ induction of CREB-1 phosphorylation, with only a 1.6-fold increase being observed. Thus, Smad3 might also be involved in the pathway mediating TGF␤ induction of CREB-1 phosphorylation.
As we have demonstrated that both CREB-1 and Smad3 are involved in the transcription factor complex formed at the CRE site in T␤3-P61/35 in response to TGF␤ treatment, we performed EMSAs using siRNA-CREB-1 and siRNA-Smad3 to FIGURE 4. Smad3 is required for TGF␤ induction of T␤3-P61/35 DNA binding and T␤3-P221/110-Luc luciferase reporter activity. A, CCL64 L20 or CCL64 Smad3C cells were incubated in the absence or presence of TGF␤3 (10 ng/ml) or PMA (100 ng/ml) for 1 h. Nuclear protein was harvested, and EMSAs were performed as described under the "Materials and Methods." Three independent experiments were performed, and representative figures are shown. B, CCL64 L20 or CCL64 Smad3C cells were transfected with T␤3-P221/110-Luc. 24 h after transfection, the cells in the TGF␤3 group were exposed to TGF␤3 (10 ng/ml), whereas the control group was not treated. The cells were harvested after an additional 24 h, and luciferase assays were performed as described under the "Materials and Methods." Data are plotted as the mean Ϯ S.E. of triplicate samples for each of three independent experiments. Numbers on top of each bar indicate the fold increase compared with control. C, IEC4-1 cells were co-transfected with T␤3-P221/110-Luc and DN-Smad4. 24 h after transfection, the cells in the TGF␤3 group were exposed to TGF␤3 (10 ng/ml), whereas the control group was not treated. The cells were harvested after an additional 24 h, and luciferase assays were performed as described under the "Materials and Methods." EV, empty vector. Data are plotted as the mean Ϯ S.E. of triplicate samples for each of three independent experiments. Numbers on top of each bar indicate the fold increase compared with control.  6 versus lane 2). siRNA-CREB-1 or siRNA-Smad3 significantly blocked TGF␤ induction of T␤3-P61/35 DNA binding activity (Fig. 6C, lane 7 versus 3 and lane 8  versus 4), indicating that both CREB-1 and Smad3 were required for the DNA binding at T␤3-P61/35. These results were confirmed by quantitative analyses of the bands using densitometry (data not shown). siRNA-CREB-1 also repressed basal levels of T␤3-P61/35 DNA binding activity (Fig. 6C, lane 3 versus 2), which is consistent with the results shown in Fig. 6B, indicating that siRNA-CREB-1 also inhibited basal levels of phospho-CREB-1. Thus, both CREB-1 and Smad3 were required for TGF␤ induction of T␤3-P61/35 DNA binding activity.
siRNA-CREB-1 and siRNA-Smad3 Both Suppress TGF␤ Induction of TGF␤3 Expression-Although we have demonstrated that CREB-1 and Smad3 were involved in TGF␤ induction of TGF␤3 promoter DNA binding and transcriptional activities, it was of interest to extend our observations to exam- 24 h after transfection, the cells were pretreated with either SB203580, PD98059, or SP600125 at the concentrations indicated for 30 min and then incubated in the absence or presence of TGF␤3 (10 ng/ml) for an additional 24 h. The cells were harvested, and luciferase assays were performed as described under the "Materials and Methods." Data plotted are the mean Ϯ S.E. of triplicate samples from one of three independent experiments. Asterisks indicate a statistically significant difference ( p Ͻ 0.05) in the fold changes between the absence and presence of TGF␤ in the inhibitor-treated IEC4-1 cells compared with the control IEC4-1 cells (no inhibitors). C, IEC4-1 cells were pretreated with H89 at the concentrations indicated for 30 min and were then incubated in the absence or presence of TGF␤3 (10 ng/ml) for 1 h. Nuclear protein was harvested, and EMSAs were performed as described under the "Materials and Methods." Three independent experiments were performed, and a representative figure is shown. D, IEC4-1 cells were transfected with T␤3-P221/110-Luc. 24 h after transfection, the cells were pretreated with H89 at the concentrations indicated for 30 min and were then incubated in the absence or presence of TGF␤3 (10 ng/ml) for an additional 24 h. The cells were harvested, and luciferase assays were performed as described under the "Materials and Methods." Data plotted are the mean Ϯ S.E. of triplicate samples from one of three independent experiments. FIGURE 6. siRNA-CREB-1 and siRNA-Smad3 both suppress TGF␤ induction of T␤3-P61/35 DNA binding and T␤3-P221/110-Luc luciferase reporter activity. A, IEC4-1 cells transfected with the indicated siRNA plasmids were harvested. Cell lysates were subjected to Western blotting using the CREB-1 and Smad3 antibodies described under the "Materials and Methods." As a loading control, ␤-actin was also detected in the same membrane using an anti-␤-actin antibody. The siRNA-Control (siRNA-Ctrl) used was an siRNA-scrambled sequence from Ambion. B, upper panel, siRNA-transfected IEC4-1 cells as described in A were incubated in the absence or presence of TGF␤3 (10 ng/ml) for 30 min. The cells were then harvested, and lysates were prepared, and Western blotting was performed with phospho-CREB-1 and phospho-Smad3 antibodies, as described under the "Materials and Methods." As a loading control, ␤-actin was also detected in the same membrane using an anti-␤-actin antibody. The siRNA-control used was an siRNA-scrambled sequence from Ambion.  OCTOBER 6, 2006 • VOLUME 281 • NUMBER 40 ine whether CREB-1 and Smad3 were necessary for TGF␤ induction of TGF␤3 expression. We therefore performed real time RT-PCR to examine TGF␤3 mRNA expression levels after knockdown of CREB-1 or Smad3. As shown in Fig. 6E, TGF␤ treatment induced a 5.7-fold increase in TGF␤3 mRNA expression in control cells, confirming the presence of an active TGF␤3 auto-loop. Furthermore, siRNA-CREB-1 or siRNA-Smad3 decreased the TGF␤ induction of TGF␤3 mRNA expression to levels of only 3.1-fold, a statistically significant decrease compared with that for siRNA-Ctrl (p Ͻ 0.01). These results indicate that CREB-1 and Smad3 are required for TGF␤ induction of TGF␤3 mRNA expression.

Smad3 and CREB-1 Mediate TGF␤3 Expression
siRNA-CREB-1 and siRNA-Smad3 Both Suppress Endogenous TGF␤3 Production-To measure endogenous TGF␤3 production in response to TGF␤3 stimulation, we performed ELISA quantitative analyses on CM from IEC4-1 cells, which produce high levels of TGF␤3 (20). Because the TGF␤3 ELISA is specific for TGF␤3, with no cross-reactivity with TGF␤1 or TGF␤2, we used TGF␤1 to treat the cells and performed ELISAs to evaluate TGF␤3 concentrations in CM collected from IEC4-1 cells. As shown in Fig. 6F, basal TGF␤3 production was reduced in siRNA-CREB-1 transfected IEC4-1 cells. Upon TGF␤1 stimulation, TGF␤3 production in IEC4-1 cells was increased by more than 6-fold. This induction of TGF␤3 by TGF␤1 was significantly suppressed in siRNA-CREB-1-and siRNA-Smad3-transfected IEC4-1 cells. The results are consistent with our previous observations in EMSAs and reporter assays. Furthermore, the results indicate that a TGF␤3 autocrine loop exists in IEC4-1 cells and that both CREB-1 and Smad3 binding at the CRE site of the TGF␤3 promoter are critical for the physiological production of TGF␤3 in these cells.
Schematic Model for TGF␤3 Autoregulation-Based upon the results in the preceding figures, we have defined the signal transduction pathways that mediate TGF␤3 autoregulation, specifically focusing on TGF␤ induction of TGF␤3 secretion, as depicted in the schematic diagram in Fig. 7. According to this model, CREB-1, ATF-2, and c-Jun constitutively bind at the TGF␤3 promoter CRE site. TGF␤ activates Smad3 as well as JNKs and p38 through the TGF␤ receptors (43)(44)(45). JNKs and p38 activate CREB-1 and ATF-2 (45)(46)(47), which thereafter bind at the CRE site (Ϫ45 to Ϫ39) in the TGF␤3 promoter region T␤3-P61/35. Phospho-Smad3 is also bound at this CRE site. Based upon our results, the activation of Smad3, CREB-1, and ATF-2 by TGF␤ results in an increase in TGF␤3 transcriptional activity and a subsequent increase in TGF␤3 production. c-Jun is also constitutively bound to this TGF␤3 promoter CRE site. However, the DNA binding activity of c-Jun is not TGF␤-inducible, suggesting that c-Jun may play a role in basal TGF␤3 production in response to stimuli other than TGF␤.

DISCUSSION
Although three mammalian isoforms of TGF␤, TGF␤1, TGF␤2, and TGF␤3, share 60 -80% identity at the amino acid level, the promoter regions of these isoforms are highly variable, suggesting that their expression is regulated by distinct mechanisms. Our previous results indicated that the proximal AP-1 site in the TGF␤1 promoter region was critical for TGF␤1 auto-induction and that the JNK and ERK cascades, as well as Smad3 and Smad4, were required for TGF␤1 auto-regulation in untransformed epithelial cells (19). However, the signaling pathways mediating TGF␤ regulation of TGF␤3 production have not been investigated previously. In this study, we found that TGF␤3 auto-induction was mediated by the CRE site located between Ϫ61 and Ϫ35 in the TGF␤3 promoter region and that CREB-1 and Smad3, as well as JNKs and p38, but not PKA, ERKs, or Smad4, were the critical activators of TGF␤3 expression. In addition, we found that CREB-1, ATF-2, and c-Jun are constitutively bound at this CRE site. In contrast, Smad3 binding at the TGF␤3 promoter CRE site was only observed after TGF␤ stimulation. siRNA-CREB-1 and siRNA-Smad3 significantly inhibited the TGF␤-inducible effects on complex formation at the TGF␤3 promoter CRE, T␤3-P221/ 110-Luc activity, TGF␤3 mRNA expression, and TGF␤3 secretion. Because these data demonstrate that both CREB-1 and Smad3 are critical for mediating TGF␤3 auto-induction, our results provide the first evidence of the physiological relevance of these transcriptional factors, as well as of the CRE site, in mediating this critical biological response to TGF␤. Furthermore, DN RII or DN Smad3, but not DN Smad4, effectively blocked the DNA binding activity induced by TGF␤, confirming that RII and Smad3 are required for this TGF␤3 auto-loop regulation.
In this study, we have shown for the first time that Smad3 does not bind to the SBE site in the T␤3-P61/35 region upon TGF␤ stimulation, but instead it binds to the CRE site, where FIGURE 7. Schematic model for TGF␤3 autoregulation. CREB-1, ATF-2, and c-Jun constitutively bind at the TGF␤3 promoter CRE site and are likely required for basal expression of TGF␤3. Upon TGF␤3 stimulation, phosphorylated Smad3 is also recruited to this CRE site, which, together with abovementioned transcription factors, significantly enhances the TGF␤3 promoter transactivation, thereby resulting in an increase in TGF␤3 production. Overall, phosphorylation of CREB-1 and Smad3 are the critical events in mediating TGF␤3 autoinduction.
CREB-1 also binds. Moreover, Smad4 was not present in this binding complex. Because Smad3 knockdown significantly blocked TGF␤ induction of T␤3-P61/35 DNA binding, T␤3-P221/110-Luc activity, and TGF␤3 secretion, Smad3 is not only present in the complex bound at T␤3-P61/35 CRE, but it is also functionally required for TGF␤-inducible TGF␤3 secretion. There are several reports that indicate that Smad3, but not Smad4, associates with both the CREB-1-binding protein, as well as the structurally related p300 protein, in response to TGF␤ exposure (48 -50). In keeping with these previous findings, and our results herein, it is conceivable that Smad3 may actually form a complex with CREB-1 in response to TGF␤ and activate TGF␤3 secretion in the absence of Smad4.
We also found that CREB-1 and ATF-2, but not CREB-2, were constitutively bound at the TGF␤3 promoter CRE site. Although CREB-1, CREB-2 (also termed ATF-4), and ATF-2 all belong to the ATF/CREB family, they do not share much similarity other than a basic region-leucine zipper (bZip) motif (51). Our data indicate that CREB1, but not CREB-2, is involved in TGF␤3 secretion mediated by TGF␤.
It has been shown that the ATF/CREB family of proteins can form heterodimers with AP-1 proteins to activate target gene transcription (51). Similarly, previous reports have indicated that c-Jun can bind to a consensus CRE site. For example, it has been reported that TGF␤ could stimulate CREB-dependent transcription by increasing the amount of c-Jun present in the CREB-1-binding protein-containing complex bound at a consensus CRE site (52). In addition, overexpression of c-Jun not only activated a reporter gene containing a consensus CRE but also activated a portion of the c-Fos promoter only containing a CRE site (52). Our results also suggest that c-Jun might be involved in TGF␤3 regulation.
It should be pointed out that although Smads are critical for TGF␤ signaling, TGF␤ also stimulates other intracellular signaling pathways, such as JNKs, ERKs, and p38 (19,43,44,(53)(54)(55)(56). Such MAPKs are required for the signaling of TGF␤ effects on growth, apoptosis, and gene expression (53)(54)(55)(56). MAPKs are also kinases that can function upstream of CREB (25). In this study, selective inhibitors of JNKs and p38 effectively blocked DNA binding to the TGF␤3 promoter region T␤3-P61/35, indicating that JNKs and p38 were required for TGF␤ activation of complex formation at this site. Most notably, the selective MEK1 inhibitor failed to suppress the DNA binding to T␤3-P61/35, although there are reports that ERKs are involved in CREB activation, particularly in specific neuronal cell cultures (26,31). Although others have shown that JNK and p38 are involved in TGF␤ signaling, this is the first report of their requirement in TGF␤ induction of TGF␤3 gene transcription.
It has been reported that ATF-2, which is known to be a nuclear target of p38 and JNKs, was phosphorylated in the N-terminal activation domain in response to TGF␤ (47). In addition, TGF␤-mediated induction of fibronectin requires activation of JNK, which in turn modulates the activity of ATF-2 and c-Jun (57). Moreover, TGF␤ can induce the phosphorylation of ATF-2 via p38 and TGF␤-activated kinase-1 (TAK1), follow by ATF-2 complex formation with Smad3 and Smad4 (58). The binding between ATF-2 and Smad3/4 is mediated via the MH1 domain of the Smad proteins and the basic leucine zipper domain of ATF-2 (58). Furthermore, ATF-2 can cooperate with Smad3 to regulate the rate of chondrocyte maturation in response to TGF␤ (59). In our report, ATF-2 displayed TGF␤-inducible DNA binding activity at the TGF␤3 promoter CRE site, suggesting that ATF-2 may also play a role in TGF␤ induction of TGF␤3 production through the signaling pathways described above. Thus, our findings provide evidence of co-regulation by the MAPK and Smad signaling pathway components in mediating this biological response to TGF␤, similar to the synergistic contribution of these signaling pathways in mediating other TGF␤ responses (53,56,60).
It is well documented that the autocrine and/or paracrine effects of TGF␤1 play a crucial role in tumor invasion and progression. However, although increased expression of TGF␤3 has been observed in a variety of late stage carcinomas, reports regarding the paracrine effects of TGF␤3 in modulating the tumor microenvironment and the immune system of the host are scarce. It has been reported that expression of TGF␤3 correlates with the progression of osteosarcomas (61). In addition, TGF␤3 has been shown to contribute to the formation of tumor stroma, induction of angiogenesis, and modulation of extracellular matrix, suggesting that excessive secretion of TGF␤3 by the tumor and/or stromal cells would also foster cancer progression (62). Identification of the factors that are required for TGF␤3 auto-regulation improves our understanding of the mechanisms underlying TGF␤3 secretion. Our results may enable the design of novel strategies to regulate TGF␤3 secretion in pathological conditions by manipulating factors that are involved in this pathway. For example, blockade of TGF␤3 production in late stage solid cancers using siRNA approaches may prevent the invasiveness and metastatic nature of such cancers. Future studies can address such potential therapeutic strategies.