Transforming growth factor-beta inhibits adipocyte differentiation by Smad3 interacting with CCAAT/enhancer-binding protein (C/EBP) and repressing C/EBP transactivation function.

Transforming growth factor (TGF)-beta is a potent inhibitor of adipocyte differentiation. To identify which adipocyte transcription factors might be targeted by TGF-beta, we overexpressed key adipogenic transcription factors, C/EBPbeta, C/EBPdelta, or peroxisome proliferator-activated receptor (PPAR) gamma in NIH3T3 cells and tested the ability of TGF-beta to block adipogenesis. We show that TGF-beta inhibits adipocyte differentiation driven by either C/EBPbeta or C/EBPdelta without affecting C/EBP protein expression levels, suggesting that these C/EBPs are a direct target of TGF-beta action. Because TGF-beta inhibits adipogenesis by signaling through Smad3, we examined physical and functional interactions of Smad3 and Smad4 with C/EBPbeta, C/EBPdelta, and PPARgamma2. C/EBPbeta and C/EBPdelta were found to physically interact with Smad3 and Smad4, and Smad3 cooperated with Smad4 and TGF-beta signaling to repress the transcriptional activity of C/EBPs. Thus, repression of the activity of C/EBPs by Smad3/4 at C/EBP binding sites inhibited transcription from the PPARgamma2 and leptin promoters. In contrast, PPARgamma interacted only very weakly with Smad3 and its transcriptional activity was not repressed by Smad3/4 or in response to TGF-beta. Smad3/4 did not reduce the ability of C/EBP to bind to its cognate DNA sequence, but repressed transcription by inhibiting the transactivation function of C/EBP.

Cell differentiation requires changes in protein expression patterns to allow manifestation of a specialized phenotype from a precursor state. Defined transcription factors, often referred to as "master regulators," are necessary and often sufficient to activate the differentiation process. In mesenchymal differentiation, MyoD, PPAR␥, 1 and CBFA1 represent master regulators that drive muscle, fat, and bone cell differentiation, respectively. The signaling pathways that regulate their function are not well understood.
Several transcription factors play key roles in adipocyte differentiation. C/EBP transcription factors, belonging to the ba-sic region-leucine zipper superfamily, were the first to be recognized as critical for adipogenesis. C/EBP␣ expression induces growth arrest and adipocyte differentiation of mesenchymal cells (1)(2)(3), but the time course of its expression is delayed relative to the earliest adipocyte marker genes. The ␤ and ␦ isoforms of C/EBP were subsequently identified as essential initiators of adipocyte differentiation (4,5), and mice with both genes deleted have grossly defective adipose tissue development (6). Analysis of the gene for aP2, an early adipocyte marker, led to the identification of PPAR␥, a nuclear hormone receptor (7). Like C/EBPs, PPAR␥ can drive adipogenesis in mesenchymal cells (8), and deletion of PPAR␥ blocks adipocyte differentiation in vivo (9). ADD-1/SREBP1c, a basic helix-loophelix transcription factor, may be involved in the generation of the ligand for transcriptional activation of PPAR␥ (10).
Specification of adipocyte differentiation involves cooperation of C/EBPs with PPAR␥2. C/EBP␤ and C/EBP␦ are induced in response to hormonal stimuli and, together, directly activate transcription of the PPAR␥2 gene (11,12) and other genes linked to adipogenesis (13). Whereas many adipocyte gene promoters contain binding sites for C/EBP and PPAR␥, no PPAR␥ binding site has been found in the PPAR␥ promoter (14). Activation of PPAR␥2 transcription and transcriptional activation by ligand results in further activation of adipocyte marker genes. C/EBP␤ and -␦ are down-regulated as differentiation proceeds, and their transcription functions are thought to be replaced by C/EBP␣ (4). C/EBP␣ activates many of the same genes as C/EBP␤ and -␦, and also triggers growth arrest that accompanies full differentiation (1,15). C/EBP␣ cooperates with PPAR␥2 to activate adipocyte gene expression, and both factors are required for adipocyte differentiation (16). Thus, any of these transcription factors could represent targets for regulation by signaling pathways that affect adipogenesis.
One signaling pathway that affects adipocyte differentiation is initiated by TGF-␤. TGF-␤ regulates mesenchymal differentiation, inhibiting osteoblast (17), myoblast (18), and adipocyte differentiation (19,20). TGF-␤ blocks adipocyte differentiation in vitro (19,21), and transgenic overexpression of TGF-␤ in adipose tissue inhibits differentiation in vivo (22). However, TGF-␤ is expressed endogenously in adipose tissue in vivo (23), and in cultured preadipocytes and adipocytes (24 -26). In animal models of obesity (23) and humans with obesity, TGF-␤1 expression is increased, correlating directly with body mass index and increased expression of PAI-1 (plasminogen activator inhibitor-1), which in turn are closely related to insulin resistance (27). These observations contrast with the ability of TGF-␤ to strongly inhibit adipocyte differentiation. However, increased TGF-␤ expression in obese adipose tissue is believed to be related to increased tumor necrosis factor-␣ expression in obese adipose tissue (28). Like TGF-␤, tumor necrosis factor-␣ strongly inhibits adipocyte differentiation in culture (29). Whereas there is already an extensive body of literature on the role of tumor necrosis factor-␣ in adipocyte tissue physiology and insulin resistance, little is as yet known about the role of TGF-␤ in adipose tissue.
TGF-␤ signals through two types of transmembrane serinethreonine kinase receptors. Ligand binding to the type II TGF-␤ receptor stabilizes complex formation with the type I TGF-␤ receptor and induces activation of the type I receptor (T␤RI) by the type II receptor (T␤RII) kinase (30). Smads then act as signaling effectors (31,32). C-terminal phosphorylation of Smad2 or Smad3 by T␤RI results in a conformational change that promotes heteromerization with Smad4, and stimulates nuclear translocation of Smad complexes. In the nucleus, Smad proteins regulate transcription by binding to DNA and interacting with other transcription factors.
We have shown that Smad3, and not Smad2, mediates inhibition of adipocyte differentiation by TGF-␤ (20). Smad3 also acts as an effector of TGF-␤ inhibition of osteoblast (33) and myoblast differentiation (34). In the latter case, Smad3 physically interacts with MyoD and disrupts its binding to DNA, thus reducing activation of muscle-specific gene expression. In TGF-␤-mediated inhibition of osteoblast differentiation, Smad3 represses CBFA1 function without disrupting its DNA binding, although the mechanism of Smad3-mediated repression of CBFA1 remains to be characterized (33).
In this report, we examined the mechanism by which Smad3 and TGF-␤ inhibit adipocyte differentiation. We found that adipogenesis, driven by either C/EBP␤ or C/EBP␦, could be inhibited by TGF-␤ without a decrease in C/EBP protein levels. C/EBPs physically interacted with both Smad3 and Smad4, whereas PPAR␥2 interacted weakly or not at all with Smads. This interaction correlated with repression of C/EBP-mediated transcription by Smad3 and Smad4 at adipocyte differentiationdependent promoters and multimerized C/EBP binding sites. In contrast, Smad3 and Smad4 did not affect transcription by PPAR␥2. Smad3 and Smad4 cooperated to repress the transcription function of C/EBPs without inhibition of target DNA sequence binding. These data represent the first example of direct inhibition of the transactivation function of a transcription factor by Smads.
Retrovirally infected NIH3T3 cells were made as described (20). Selection with 2 g/ml puromycin was started 48 h postinfection. For differentiation, confluent cells were treated for 48 h with 1 M dexamethasone and 0.5 mM isobutylmethylxanthine in Dulbecco's modified Eagle's medium containing 5 g/ml insulin and 10% fetal bovine serum. Cells expressing PPAR␥2 were also treated with, and maintained in, 5 M troglitizone (Parke-Davis). As needed, 10 ng/ml TGF-␤ was added at the same time as the differentiation inducers and readded when the medium was changed. All cells were maintained in Dulbecco's modified Eagle's medium with 10% fetal bovine serum plus 5 g/ml insulin after the dexamethasone-isobutylmethylxanthine treatment.
Analysis of Lipid Accumulation, RNA, and Protein-NIH3T3 cells, expressing the transcription factor of interest, were analyzed 8 days after initiation of differentiation treatment for lipid by Oil Red O staining (20). Parallel cultures were harvested for RNA or protein extraction. RNA was extracted using the SV total RNA isolation kit (Promega) and processed for Northern analysis (20). For protein extraction, cells were lysed in 300 l/well with FLAG lysis buffer (300 mM NaCl, 20 mM Tris, pH 7.5, 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 20 g/ml leupeptin, 20 g/ml aprotinin). 50 g of clarified lysate was run on SDS-PAGE gels and transferred to polyvinylidene difluoride. Western blotting and development with ECLϩ was performed as directed (Amersham Biosciences), using anti-C/EBP␤, anti-C/EBP␦, and anti-PPAR␥ primary antibodies from Santa Cruz Biotechnology.
Transient Transfections and Reporter Assays-3T3-F442A or 3T3-L1 cells, plated 18 h prior to transfection at 9 ϫ 10 4 cells per well of 6-well dishes, and COS cells, plated at 1 ϫ 10 6 per 10-cm plate, were transfected using LipofectAMINE (Invitrogen). For 3T3-F442A and 3T3-L1 cells, 1 g of DNA was used per well, and for COS cells, 5 g of DNA was used per plate, and the total amount of DNA was constant by addition of pRK5. For luciferase assay, cells were lysed in 300 l/well with 1ϫ reporter lysis buffer (Promega), and assayed using reagents from BD Pharmingen. Assays were performed at least 3 times in duplicate or triplicate. All values are expressed as the -fold induction relative to the basal activity.
GST Interaction Assays and Immunoprecipitations-GST-Smad proteins (44) were prepared and purified on glutathione-Sepharose beads (Amersham Biosciences). 35 S-Labeled C/EBP␣, -␤, -␦, or PPAR␥2 was generated by in vitro translation using the TNT quick-coupled transcription/translation kit (Promega) and [ 35 S]methionine. 5 l of translation mixture, adjusted to 1 ml with GST pull-down buffer, was incubated with 2 g of GST or GST-Smads, and adsorbed proteins were analyzed, as described (33).
For analysis of Smad-C/EBP interactions in vivo, transfected COS cells were metabolically labeled with [ 35 S]methionine and cysteine and processed for immunoprecipitation as described (45), except that cells were lysed in FLAG lysis buffer. Precipitates were washed twice with HSA (12.5 mM potassium phosphate buffer, pH 7.4, 600 mM NaCl), once with MDB (0.1% SDS, 0.05% Nonidet P-40, 300 mM NaCl, 10 mM Tris, pH 8.3), once with HSA, and once with SA (12.5 mM potassium phosphate buffer, pH 7.4, 300 mM NaCl). For sequential immunoprecipitation of C/EBP immunoprecipitates, 50 l of a buffer containing 1% SDS, 20 mM Tris, pH 7.5, 50 mM NaCl, and 1 mM dithiothreitol was added to the washed beads, and the samples were heated at 95°C for 4 min. The eluted samples were then immunoprecipitated using anti-FLAG antibody. Immunoprecipitated proteins were run on SDS-PAGE, and gels were soaked in Amplify (Amersham Biosciences) prior to autoradiography.
Electrophoretic Mobility Shift Assays and Biotinylated Oligonucleotide Interactions-For electrophoretic mobility shift assay, the following oligonucleotides were synthesized: 2ϫ wild-type C/EBP (top strand only), 5Ј-CTTGGCATATTGCGCAATATGCTTGGCATATTGCGCAAT ATGC-3Ј; 2ϫ mutant C/EBP, 5Ј-CTAGCGATAaaGCGCttTATGCTT-GCGATAaaGCGCttTATGC-3Ј. The mutations, indicated by lowercase letters, are in the residues that are critical for C/EBP binding (46). For biotinylated oligonucleotide binding reactions, the identical oligonucleotides (top strand) were modified by 5Ј addition of biotin.
For each biotinylated oligonucleotide reaction, 30 l of streptavidin magnetic beads (Promega) were washed twice with 2ϫ B&W buffer (10 mM Tris, pH 7.5, 1 mM EDTA, 2 M NaCl), then 100 ng of wild-type or mutant oligonucleotide was bound to the beads in 1ϫ B&W buffer. Beads were washed twice with 2ϫ B&W buffer, once with binding buffer (5% glycerol, 20 mM Tris, pH 7.5, 1 mM EDTA, 1 mM dithiothreitol, 0.15% Triton X-100, 100 mM NaCl, 4 mM MgCl 2 ), blocked for 30 min using 1% bovine serum albumin in binding buffer, and resuspended in 50 l of binding buffer. Transfected COS cells were lysed in 1 ml of FLAG lysis buffer, and 100 l of clarified lysate was used in a 0.5-ml reaction with 150 l of 3ϫ binding buffer, 10 g of poly(dI-dC), and 50 l of DNA-bound streptavidin magnetic beads. Incubations proceeded for 1 h at 4°C with gentle mixing, followed by 5 washes in 1ϫ binding buffer. Bound proteins were analyzed by SDS-PAGE followed by Western blot analysis.

TGF-␤ Inhibits C/EBP␤-and C/EBP␦-mediated Adipocyte
Differentiation-We have previously shown that TGF-␤ signaling inhibits adipocyte differentiation through TGF-␤-activated Smad3 (20). To address whether TGF-␤ inhibits the function of individual adipocyte transcription factors, we generated NIH3T3 cell lines stably infected with retroviruses encoding C/EBP␤, C/EBP␦, C/EBP␣, or PPAR␥. NIH3T3 cells, while unable to differentiate into adipocytes, support adipogenesis when any one of these transcription factors is expressed (4,8). Furthermore, these cells have a functional TGF-␤ signaling system (48,49). Expression of the respective transcription factors was confirmed by Western blot (Fig. 1A). Whereas TGF-␤ treatment did not affect C/EBP expression, PPAR␥ expression in the stable PPAR␥-infected cells consistently decreased upon treatment with TGF-␤.
Adipocyte differentiation of these cells was induced, and cells infected with empty vector were subjected to the same treatments. Consistent with previous reports (4,8), all three transcription factors caused characteristic adipocyte rounding and lipid accumulation, as assessed by microscopic examination and Oil Red O staining of lipid accumulation (Fig. 1B). However, in the presence of 10 ng/ml TGF-␤, the differentiation of C/EBP␦ cells was prevented. TGF-␤ also completely inhibited adipogenesis in NIH3T3 cells that stably expressed C/EBP␣ (data not shown). Cells expressing C/EBP␤ were also inhibited, but to a lesser extent (Fig. 1B). We also observed a low level decrease in differentiation of PPAR␥ expressing cells in response to TGF-␤ (Fig. 1B), which is likely related to the decreased PPAR␥ levels in TGF-␤-treated cells (Fig. 1A).
TGF-␤ Represses Adipocyte Marker Gene Expression Activated by C/EBP␤ or C/EBP␦-We assessed the effect of TGF-␤ on the expression of several adipocyte differentiation genes. Ectopic expression of any of the three transcription factors induced expression of PPAR␥, aP2, and adipsin mRNAs, but none induced C/EBP␣ expression, consistent with previous reports (4,50). Expression of C/EBP␤ led to the highest levels of adipocyte marker gene expression (Fig. 2). TGF-␤ treatment of these cells moderately decreased the level of PPAR␥ expression and strongly inhibited aP2 and adipsin expression (Fig. 2). NIH3T3 cells expressing C/EBP␦ had lower levels of PPAR␥, aP2, and adipsin than C/EBP␤-expressing cells, consistent with the lower number of lipid-filled cells in cultures expressing C/EBP␦ versus C/EBP␤ (Fig. 1B). Treatment of the C/EBP␦expressing cells with TGF-␤ blocked the expression of PPAR␥, aP2, and adipsin (Fig. 2). C/EBPs can directly activate the PPAR␥ and aP2 promoters (11,12,51) and are critical for induction of adipsin expression (52). Because the levels of C/EBP␤ and -␦ were unaffected by TGF-␤, these results suggest that TGF-␤ inhibits the transcriptional activity of C/EBP␤ and -␦, and that this repression may result in decreased adipocyte marker mRNA levels and reduced adipogenesis.
TGF-␤ treatment of PPAR␥ expressing cells decreased the levels of both viral and endogenous PPAR␥ mRNA (Fig. 2) and protein (Fig. 1A). Expression of aP2 mRNA, which is directly activated by PPAR␥ (7), was only modestly decreased by TGF-␤. This decrease was consistent with the decreased PPAR␥ mRNA and protein levels. In contrast, adipsin and PPAR␥ mRNAs, which require C/EBP transcription factors for full induction, and are not directly induced by PPAR␥2 (14,52), were strongly down-regulated. These data support the notion that TGF-␤ primarily inhibits the activity of C/EBPs, i.e. C/EBP␦ and C/EBP␤, whereas inhibition of PPAR␥ may occur at the level of PPAR␥ protein and mRNA accumulation, rather than its transcriptional activity.
Physical Interaction of Smad3 with C/EBPs-To evaluate a direct role of TGF-␤ in inhibiting the function of the adipogenic transcription factors, we examined their ability to physically interact with Smads. In vitro translated C/EBP␤, C/EBP␦, C/EBP␣, and PPAR␥ were tested for interaction with GSTfused Smad1, -2, -3, or -4 (Fig. 3A). The three C/EBPs had an identical interaction profile, having strong interaction with both GST-Smad3 and GST-Smad4, very weak interaction with GST-Smad1, and no interaction with Smad2. PPAR␥ exhibited a marginal interaction with GST-Smad3 only. Neither the addition of PPAR␥ ligand nor cotranslating PPAR␥ with its partner RXR␣ improved the ability of PPAR␥ to interact with GST-Smads (data not shown).
The interaction of C/EBP␦ with Smad3 and Smad4 was also observed in vivo (Fig. 3B). For this purpose, we expressed C/EBP␦, FLAG-tagged Smad3 or Smad4, and the constitutively active T␤RII/RI chimera (53) in various combinations in COS cells. The high transfection efficiency of COS cells that allows this type of analysis stands in contrast with 3T3-F442A cells, which were not amenable for protein interaction experiments (data not shown). The minimal sensitivity of COS cells to TGF-␤, coincident with their very low levels of TGF-␤ receptors, requires coexpression of an activated TGF-␤ receptor rather than TGF-␤ treatment, to activate transfected Smads. Transfected cells were 35 S-labeled, and subjected to immunoprecipitation with anti-FLAG or anti-C/EBP␦, or sequential immunoprecipitation with anti-C/ EBP␦ followed by anti-FLAG (Fig. 3B). Smad3 or Smad4 were detected in anti-C/EBP␦ immunoprecipitations, but only when C/EBP␦ was expressed. T␤RII/RI expression did not stimulate this interaction, as has been observed for interactions of Smads with transcription factors expressed in COS cells (33). Identical results were obtained when C/EBP␤ was expressed instead of C/EBP␦ (data not shown). The inability of the anti-Smad or anti-C/EBP antibodies to immunoprecipitate denatured proteins generated in sequential immunoprecipitation precluded a similar analysis of endogenous protein interactions in 3T3-F442A cells (data not shown).
We next made truncation mutants of C/EBP␦ consisting of its transcription activation domain or basic region plus leucine zipper, required for DNA binding and dimerization, and assessed their abilities to interact with GST-fused Smad3 and -4. The interactions of Smad3 and Smad4 with the basic region plus leucine zipper segment were strong, but we also detected a weak, yet specific interaction with the transcription activation domain (Fig. 4A). Conversely, C/EBP␦ primarily interacted with the MH1 (N) domain of Smad3, although there was also a very weak interaction with the MH2 (C) domain (Fig. 4B). In transfected cells, C/EBP␦ interacted with the MH2 domain as well as, or perhaps better than, the MH1 domain of Smad3 (Fig. 4C). This stronger interaction of the MH2 domain in vivo versus in vitro has also been observed with the interaction of Smad3 with MyoD (34), and may reflect the stabilizing participation of additional proteins in the transcription machinery. Neither truncated Smad3 protein could interact as strongly with C/EBP␦ as full-length Smad3, consistent with an interaction of both Smad3 domains with C/EBP␦.
Smad3 Inhibits Transcriptional Activation of Adipocyte Marker Genes by C/EBPs-The adipogenic function of C/EBP␤ has been shown to depend on the transcription activation domain of C/EBP␤ (4). The association of Smad3 and Smad4 with C/EBPs (Figs. 3 and 4) and the inhibition of differentiation of cells expressing C/EBPs by TGF-␤ (Figs. 1 and 2) suggests that Smad3, in cooperation with Smad4, may decrease the ability of C/EBP␤ or C/EBP␦ to activate transcription. We tested this hypothesis using promoters from the PPAR␥2 and leptin genes, two adipocyte differentiation-induced promoters known to be activated by C/EBPs.
As shown in Fig. 5A, C/EBP␦ activated transcription of a promoter segment of PPAR␥2 containing 190 bp upstream from the transcription start. Smad3 inhibited C/EBP␦-mediated transcription, and this inhibition was enhanced in the presence of activated TGF-␤ receptor. TGF-␤ itself exerted only a minimal decrease (data not shown), consistent with the high numbers of reporter plasmids and high expression levels of transcription factors in these transient transfection/reporter assays (e.g. Refs. 54 and 55). Smad4 enhanced the ability of Smad3 to down-regulate C/EBP-mediated transcription of the PPAR-␥ promoter segment. Similar results were seen using a promoter segment containing 159 base pairs upstream from the transcription start site of the leptin gene (Fig. 5B). C/EBP␤-induced transcription was inhibited by Smad3 and by TGF-␤ signaling. This inhibition was more dramatic when Smad4 was coexpressed with Smad3.
TGF-␤ Signaling and Smad3 Inhibit Transcription at C/EBP Binding Sites-The down-regulation of C/EBP␤-or ␦-activated transcription from the PPAR␥2 and leptin promoters by TGF-␤ and Smads did not exclude functional interactions with other DNA binding transcription factors, besides C/EBP␤ and -␦. We therefore tested if Smad3 repressed transcription from an artificial promoter that is specifically activated by C/EBP, i.e. an artificial promoter with three tandem C/EBP binding sequences. Luciferase reporter assays were again carried out in 3T3-F442A preadipocytes.
As shown in Fig. 6, A-C, C/EBP␤, -␦, and -␣ activated transcription from this 3ϫ C/EBP promoter, and their activities were mildly inhibited by an activated TGF-␤ receptor. C/EBPinduced transcription was inhibited by Smad3, and strongly inhibited by Smad3 and Smad4, but not Smad4 alone, and their inhibition was enhanced when the activated TGF-␤ receptor was coexpressed. Smad3 and Smad4 also decreased the basal transcription in the absence of cotransfected C/EBPs, presumably because of inhibition of endogenous C/EBP activity. This repression was not because of decreased C/EBP expression (Fig. 6, A and B, bottom panels). The inhibition of the C/EBPs by Smad3/4 was not restricted to 3T3-F442A preadipocytes, but was also observed in 3T3-L1 preadipocytes (Fig. 6D).
We also tested the effects of Smad3 and Smad4 on transcription by PPAR␥2. We used a luciferase reporter containing 3 PPAR␥ binding sites, similar to the 3ϫ C/EBP luciferase reporter, and tested the ability of Smad3 and Smad4 to repress transcription activated by PPAR␥2 in 3T3-F442A cells. Smad3 and Smad4 did not inhibit the activation of the 3ϫ PPRE reporter by PPAR␥2 and RXR␣ (Fig. 6E), despite their ability to inhibit C/EBP␦ activation of the 3ϫ C/EBP luciferase reporter. These data support the idea that TGF-␤ inhibits adipocyte differentiation by inhibiting the abilities of C/EBPs, but not PPAR␥2, to activate target gene expression.
DNA Binding of Smad3 Is Not Required for Repression of C/EBP Activity-Transcriptional activation by Smad3 involves Smad3 binding to DNA. However, Smad3 DNA binding is not required in the one example of Smad3-mediated repression where this requirement was tested (33). To determine whether Smad3 needs to bind DNA to inhibit transcription by C/EBP, we tested the ability of the R47D mutant of Smad3, which is unable to bind to DNA (56), to inhibit C/EBP␦-mediated transcription from 3ϫ C/EBP-luc. As shown in Fig. 7A, Smad3 R47D repressed transcriptional activation by C/EBP␦, similarly to the wild-type Smad3. However, Smad4 did not potentiate the inhibition by Smad3 R47D, as compared with wildtype Smad3. Similar results (data not shown) were obtained using the Smad3⌬LG mutant, which likewise does not bind DNA (56).
We also assessed the effects of Smad3NL, i.e. the Smad3 MH1 domain with the linker segment, and Smad3C, consisting of the MH2 domain, to inhibit activation of the C/EBP reporter by C/EBP. Smad3C exhibited a stronger inhibition than Smad3, and Smad3NL, which can also interact with C/EBP, did not inhibit C/EBP transcriptional activation, and tended to slightly enhance it (Fig. 7B). These results are consistent with the inability of DNA binding mutations in the MH1 domain to impair the repression of C/EBP by Smad3 (Fig. 7A).
Smad3 Does Not Inhibit DNA Binding of C/EBP-Smad3 could repress C/EBP transcription through interference with the DNA binding of C/EBP, similarly to the inhibition of MyoD activity by Smad3 (34). Alternatively, Smad3 could target the C/EBP transactivation domain, which mediates transcription once C/EBP is bound to DNA. Because Smad3 interacted strongly with the basic region plus leucine zipper segment of C/EBP that binds DNA, we examined the abilities of Smad3 and -4 to block DNA binding of C/EBP␦. Transcription from the transactivation domain of VP16 is not influenced by TGF-␤/Smad3 signaling (34). Thus, when fused to a DNA binding segment, alterations in transcription from the DNA binding site are likely to correlate with changes in DNA binding. A chimera of VP16 and the basic region plus leucine zipper segment of C/EBP␦ did not activate transcription from the 3ϫ C/EBP promoter (data not shown). We therefore fused the VP16 transactivation domain to full-size C/EBP␦, and this chimera activated transcription from C/EBP binding sites to a higher level than C/EBP␦ itself (Fig. 8A). Coexpression of an activated TGF-␤ receptor, Smad3 or -4, or Smad3/4, had little or no effect on transcription by VP16-C/ EBP␦ at the 3ϫ C/EBP promoter; the minor repression may have been because of repression of the C/EBP␦ activation domain in VP16-C/EBP␦ (see below). This is in contrast with the strong repression of C/EBP␦ by Smad3/4 in the same assay (Fig. 8A). These data suggest that Smad3 and/or Smad4 has little or no effect on DNA binding of C/EBP␦.
We also used a biotinylated oligonucleotide binding assay to assess whether excess Smad3 or -4 affected C/EBP␦ binding to its DNA sequence. C/EBP␦ and/or Smad3 or -4 were expressed in COS cells and their binding to an oligonucleotide comprising a wild-type or mutant C/EBP binding sequence was assessed by Western blotting (Fig. 8B). C/EBP␦ specifically bound the wildtype but not the mutant oligonucleotide. Smad3 or Smad4, expressed at high levels (as high as 10-fold excess, data not shown), did not reduce C/EBP␦ binding to a single (not shown) or double C/EBP binding sequence; if anything, Smad3 slightly increased binding. A small amount of Smad3 could be detected specifically bound to C/EBP at the DNA, but not in the absence of C/EBP or to the mutant oligonucleotide in the presence of C/EBP. The low level binding of Smad3 through C/EBP to the C/EBP binding sequence is similar to the interaction of Smad3 through CBFA1 to the CBFA1-binding oligonucleotide (33). This weak, yet specific protein interaction is likely stabilized by the many protein interactions in the multiprotein transcription machinery (57). Smad4 binding to C/EBP at the DNA was barely detectable and similar to background.
Finally, we assessed whether 10 ng/ml TGF-␤ affected the binding of endogenous C/EBP in 3T3-F442A cells to a 32 Plabeled oligonucleotide containing two C/EBP binding sites, in an electrophoretic mobility shift assay. A complex was detected that was specifically competed by excess C/EBP oligonucleotide, but not by mutant C/EBP oligonucleotide, and could be displaced and supershifted by a C/EBP␤ antibody (Fig. 8C). TGF-␤ treatment did not affect the formation of this complex. These nuclear extracts were able to form a TGF-␤-inducible complex on a Smad-binding element oligonucleotide, which could be supershifted with Smad3 or Smad4 antibody (data not shown), indicating that the cells were responsive to TGF-␤. However, anti-Smad3 or anti-Smad4 did not supershift the C/EBP complex (Fig. 8C), suggesting an unstable interaction or a conformation unfavorable for antibody-Smad interaction. This result is consistent with the low amount of interaction detected in the oligonucleotide pull-down assay (Fig. 8B), and is similar to what has been seen for Smad3-Cbfa1 interaction (33).
Smad3 Represses the Transactivation Function of C/EBPs-Because Smad3 and Smad4 did not inhibit C/EBP binding to DNA, we assessed whether Smad3 and -4 inhibited the transcription function of C/EBPs. We fused C/EBP␦ to the DNA binding domain of GAL4 and measured the activation of transcription of the FR-luc reporter by this fusion protein. TGF-␤/ Smad3 signaling does not affect the binding of the Gal4 DNA binding domain to the Gal4 binding sites in FR-luc (34); so this reporter system allows measurements of transactivation function under conditions of constant DNA binding. As shown in Fig. 9, A and B, TGF-␤ signaling repressed the transactivation function of GAL4-fused C/EBP␦. Smad3 also repressed GAL4-C/EBP␦-activated transcription of FR-luc, and this repression was stronger with an activated TGF-␤ receptor. Smad4 only weakly suppressed transcription by GAL4-C/EBP␦ (Fig. 9A). The repression was dose-responsive (Fig. 9B), and greatest when Smad3 and -4 were coexpressed (Fig. 9, A and  B). This profile of repression by TGF-␤/Smad3 signaling was similar to the repression of C/EBP␦ function at the C/EBP binding site (Fig. 6). GAL4-C/EBP␣ was also repressed by Smad3 or Smad3/4 (Fig. 9C), similarly to GAL4-C/EBP␦. For unknown reasons, GAL4-C/EBP␤ was unable to activate transcription of FR-luc in 3T3-F442A cells (data not shown). Together, the data show that Smad3 cooperates with Smad4 to repress the transactivation function of C/EBPs, without affecting C/EBP binding to the promoter DNA sequences of target genes, and that this may represent a mechanism by which TGF-␤ blocks adipogenesis. DISCUSSION

TGF-␤/Smad3 Signaling Inhibits Adipogenic Differentiation
Primarily through Functional Repression of C/EBP␤ and C/EBP␦-We previously observed that inhibition of adipogenesis by TGF-␤ was accompanied by reduced mRNA levels for PPAR␥, C/EBP␣, and ADD1/SREBP1c, but not of C/EBP␤ and -␦ (20). These data suggested that TGF-␤ targets the transcription factor cascade upstream of PPAR␥, possibly by repressing the functions of C/EBP␤ and -␦. We therefore expressed the adipogenic transcription factors individually in NIH3T3 cells, and evaluated the effect of TGF-␤ on the adipocyte differentiation program activated by each transcription factor. This analysis revealed that TGF-␤ repressed adipogenesis directed by C/EBP␤ or -␦ (or C/EBP␣; data not shown), without decreasing their levels. This indicates that TGF-␤ signaling represses the function of C/EBP␤ and -␦, a conclusion confirmed by the TGF-␤/Smad3-mediated repression of transcription by C/EBPs at synthetic and natural promoters, including the PPAR␥2 promoter. Taken together, these observations let us conclude that adipogenic differentiation is inhibited by TGF-␤ at the level of C/EBP␤ and -␦ function, upstream from PPAR␥ expression.
The differentiation of NIH3T3 cells driven by ectopic PPAR␥ expression was also mildly inhibited by TGF-␤. However, TGF-␤ also decreased PPAR␥ mRNA and protein levels. This decrease in PPAR␥2 expression may occur post-transcriptionally, because C/EBP␤ or -␦ expression from the same retroviral promoter was not affected by TGF-␤. Also, the induction of aP2 expression was relatively unaffected by TGF-␤, when considering the reduced level of PPAR␥ expression. The differentiationdependent expression of aP2 depends primarily on PPAR␥, with C/EBP playing a contributory role (58), whereas C/EBP is critical for PPAR␥ and adipsin expression (14,52). The relative insensitivity of aP2 expression to TGF-␤, in contrast to the strong repression of PPAR␥ and adipsin, in NIH3T3 cells expressing PPAR␥, suggests that the transcription function of PPAR␥ is less affected by TGF-␤ than that of C/EBP. Accordingly, TGF-␤/Smad3 signaling did not repress the transcription activity of PPAR␥. Together, our data suggest that the repression of C/EBP␤ and -␦ by TGF-␤/Smad3 signaling prevents induction of PPAR␥ expression, and that the function of PPAR␥ itself is not a direct target of repression by TGF-␤.
Mechanism of Smad3-mediated Repression of C/EBPs-The mechanism of TGF-␤-induced transcriptional activation through cooperation of Smads with sequence-specific transcription factors, and the role of Smad corepressors in reducing activation, are well documented. In contrast, little is known about mechanisms of TGF-␤-mediated repression of transcription. In epithelial cells, TGF-␤/Smad3 signaling can repress transcription by the androgen receptor (59), and activin and Smad3 repress C/EBP␤-induced transcription from the haptoglobin promoter (60). In mesenchymal cells, TGF-␤/Smad3 signaling represses the functions of CBFA1 in osteoblastic differentiation (33), and of myogenic basic helix-loop-helix transcription factors (34). A mechanism of TGF-␤/Smadmediated repression was clearly demonstrated only in the latter case. In response to TGF-␤, Smad3 represses MyoD func-FIG. 8. Smad3/4 and TGF-␤ do not inhibit C/EBP binding to DNA. A, Smad3 or Smad3/4 do not inhibit transcription by VP16-C/ EBP␦. 3T3-F442A cells were transfected with VP16-C/EBP␦ or C/EBP␦, and combinations of Smad3, Smad4, or T␤RII/RI, and the activation of the 3ϫ C/EBP-luc reporter was measured. B, Smad3 or Smad4 do not interfere with C/EBP␦ binding to DNA. COS cells were transfected with plasmids for C/EBP␦, FLAG-tagged Smad3, and/or Smad4, using a 4-fold excess of Smad plasmid DNA relative to C/EBP␦ DNA. Biotinylated oligonucleotides with two consensus C/EBP binding sequences (W), or a mutated sequence that does not bind C/EBP (M), were incubated with cell lysates. Bound proteins were precipitated by streptavidin-linked magnetic beads and subjected to Western analysis. Equal amounts of cell lysates were blotted directly to show the levels of Smads and C/EBP␦. C, TGF-␤ does not decrease C/EBP binding to DNA in gel-shift analysis (electrophoretic mobility shift assay). 3T3-F442A cells, grown under differentiation conditions, were untreated or treated for 1 h with 10 ng/ml TGF-␤. Nuclear extracts were incubated with a 32 P-labeled 2ϫ C/EBP binding sequence oligonucleotide. wt 2ϫ C/EBP, unlabeled 2ϫ C/EBP binding sequence oligonucleotide at indicated molar excess; mut 2ϫ C/EBP, unlabeled mutant 2ϫ C/EBP oligonucleotide at the indicated molar excess; ␣-C/EBP␤, incubation with anti-C/ EBP␤ to disrupt the C/EBP complex and generate a supershift (ss); ␣-Smad3, ␣-Smad4, complexes incubated with these antibodies. tion through physical interaction with the helix-loop-helix domain of MyoD, interfering with its dimerization with E12/47, thus impairing MyoD binding to DNA and blocking transcriptional activation. We now show that in mesenchymal cells, TGF-␤ represses the functions of C/EBP␤ and -␦ through Smad3, resulting in inhibition of adipogenic differentiation by TGF-␤. The physical association of Smad3 and -4 with C/EBPs provides the basis for this functional repression. In contrast, Smad3 can only marginally, if at all, associate with PPAR␥ and did not reduce its transcriptional activity. Physical association per se is not predictive of Smad-mediated repression versus activation. Indeed, Smads interact with various transcription factors to activate transcription (31,32); and interaction of Smads with AMLs (acute myeloid leukemia transcription factors) results in coactivation or repression of transcription, depending on the cell and promoter sequence contexts (33,42).
There is as yet no evidence for differential physical interactions that mediate transcriptional activation versus repression.
Because Smad3 interacts with the DNA binding domain of C/EBP (Fig. 4A), we evaluated whether Smad3 or -4 impaired DNA binding of C/EBP␤ or -␦. In contrast to MyoD (34), Smad3 did not decrease DNA binding of C/EBPs, suggesting a different mechanism of repression. Similarly, repression of CBFA1 transcription by Smad3 was not accompanied by decreased DNA binding of CBFA1 (33). The physical interactions of Smad3 and -4 with C/EBP are reminiscent of c-Jun (44,56), another bZIP transcription factor. Smad3 interacts in vitro with the DNA binding domains of both C/EBP or c-Jun. Smad3 did not disrupt and, instead, increased the DNA binding of c-Jun. However, in contrast to the transcriptional cooperativity of Smad3 with c-Jun, Smad3 repressed the C/EBP activity.
We also showed that direct DNA binding of Smad3 is not required for repression of C/EBP function by Smad3. DNA binding-defective Smad3 mutants repressed C/EBP transcription (Fig. 7), repression occurred at C/EBP binding sites without an adjacent Smad binding DNA sequence (Figs. 5-7), and Smad3 only interacted with the C/EBP binding site through C/EBP (Fig. 8). Similarly, Smad3 represses transcription by CBFA1 without the need for DNA binding (33). These observations stand in contrast with the required DNA binding of Smad3 in TGF-␤/Smad3-mediated transcriptional activation, e.g. in the cooperativity of Smad3 with c-Jun (56). It remains to be explored whether this lack of requirement of Smad3 binding to DNA is a general aspect of Smad3-mediated repression.
In contrast to repression of MyoD, Smad3 represses C/EBP transcription by repressing its transactivation function (Fig. 9). This is the first evidence for Smad-mediated repression of the transactivation function of a transcription factor. How Smad3 represses the transactivation function is as yet unclear. One possibility would be that Smad3 recruits histone deacetylases, because its MH1 domain is able to recruit deacetylase activity (61). However, the MH1 domain was dispensable for repression of C/EBP, and the MH2 domain by itself potently repressed C/EBP transcription. Furthermore, trichostatin A, which inhibits class I and II histone deacetylase activities, did not reverse Smad3-mediated repression of transcription by C/EBP (data not shown).
Another possibility would be that Smad3 interferes with the function of CBP/p300 as coactivator for C/EBP. Indeed, C/EBP␤ interacts with, and is transcriptionally activated by, p300 (62), and Smad3 also interacts with CBP/p300 as coactivators (41,63). However, p300 did not reverse Smad3 repression of C/EBP (data not shown), in contrast to the partial reversion observed on the haptoglobin promoter in hepatoma cells (60). Instead, increased p300 levels repressed, rather than enhanced, C/EBP-mediated transcription from the PPAR␥ promoter (data not shown). Furthermore, a Smad3-binding, dominant-negative segment of p300 (64) did not inhibit the repression of C/EBP by Smad3 (data not shown). Perhaps CBP or p300 contribute to Smad3-mediated repression of C/EBP function in our system, and/or as yet unidentified cofactors may be involved. Further research will address the mechanism and co-factor(s) involved in mediating this repression.