Role of CREB in transcriptional regulation of CCAAT/enhancer-binding protein beta gene during adipogenesis.

The proximal promoter of the C/EBPbeta gene possesses dual cis regulatory elements (TGA1 and TGA2), both of which contain core CREB binding sites. Comparison of the activities of C/EBPbeta promoter-reporter genes with 5'-truncations or site-directed mutations in the TGA elements showed that both are required for maximal promoter function. Electrophoretic mobility shift and chromatin immunoprecipitation (ChIP) analyses with antibodies specific to CREB and ATF1 showed that these CREB family members associate with the proximal promoter both in vitro and ex vivo. Immunoblotting and ChIP analysis revealed that other CREB family members, CREM and ATF1, are up-regulated and associate with the proximal C/EBPbeta promoter in mouse embryonic fibroblasts (MEFs) from CREB(-/-) mice. ChIP analysis of wild-type MEFs and 3T3-L1 preadipocytes revealed that interaction of phospho-CREB, the active form of CREB, with the C/EBPbeta gene promoter occurs only after induction of differentiation of 3T3-L1 preadipocytes and MEFs. Consistent with the interaction of CREB and ATF1 at the TGA regulatory elements, expression of constitutively active CREB strongly activated C/EBPbeta promoter-reporter genes, induced expression of endogenous C/EBPbeta, and caused adipogenesis in the absence of the hormonal inducers normally required. Conversely, expression of a dominant-negative CREB blocked promoter-reporter activity, expression of C/EBPbeta, and adipogenesis. When subjected to the standard adipocyte differentiation protocol, wild-type MEFs differentiate into adipocytes at high frequency, whereas CREB(-/-) MEFs exhibit greatly reduced expression of C/EBPbeta and differentiation. The low level of expression of C/EBPbeta and differentiation in CREB(-/-) MEFs appears to be due to up-regulation of other CREB protein family members, i.e. ATF1 and CREM.

The increased adipose tissue mass associated with obesity results from an increase in the number and size of adipocytes (1,2). The increase in adipocyte number is due to recruitment of preadipose cells (including preadipocytes) that populate the vascular stroma of adipose tissue (3). This process is mimicked by the mitotic clonal expansion (MCE) 1 of preadipocytes in cell culture that follows induction of differentiation (4 -7). When growth-arrested 3T3-L1 preadipocytes or mouse embryo fibroblasts (MEFs) are treated with differentiation inducers, they synchronously re-enter the cell cycle, undergo ϳ2 rounds of MCE, and then acquire adipocyte characteristics. Recent evidence has shown that this process is a prerequisite for terminal adipocyte differentiation (7,8). C/EBP␤, a B-Zip transcription factor that is expressed prior to MCE, plays an essential role in this process (9) and in subsequent events of the differentiation program (10).
Following MCE, C/EBP␤ initiates a cascade of transcriptional activation (11). C/EBP␤ activates expression of the C/EBP␣ and PPAR␥ genes which function together as pleiotropic transcriptional activators of the large group of genes that produce the adipocyte phenotype (6,12,13). Both of the C/EBP␣ and PPAR␥ genes possess cis-C/E BP regulatory elements in their proximal promoters at which C/EBP␤ binds and coordinately activates transcription (14 -17). Once expressed, C/EBP␣ is thought to maintain expression of both the C/EBP␣ and PPAR␥ genes via transactivation mediated by their respective C/E BP regulatory elements (14, 16 -18). As expression of C/EBP␣ increases (19), the expression of C/EBP␤ declines. Presumably, this dual role facilitates maintenance of the terminally differentiated state.
Several lines of evidence have implicated CREB (cAMP response element-binding protein) as a transcriptional activator in the adipocyte differentiation program. Forced expression of CREB in 3T3-L1 preadipocytes promotes differentiation as evidenced by expression of adipocyte markers and the accumulation of cytoplasmic triglyceride (20). Factors, such as methylisobutylxanthine or forskolin that increase cellular cAMP or cAMP itself (6,(21)(22)(23)(24) in combination with other agents, i.e. glucocorticoid and IGF-1 or high levels of insulin, induce differentiation of 3T3-L1 preadipocytes (25). The cellular target of cAMP, protein kinase A, catalyzes phosphorylation of CREB on serine 133 and thereby, its capacity to activate transcription (26). Finally, there is circumstantial evidence that cAMP, acting through CREB, is responsible for the transcriptional activation of the C/EBP␤ gene (27).
In the present investigation, we identify and characterize the roles of dual CRE-like cis regulatory elements (and their cognate trans-acting factors) in the C/EBP␤ gene promoter early in the adipocyte differentiation program. We show that CREB is activated by phosphorylation and, along with ATF1, binds to the dual CRE-like elements in the proximal promoter of the C/EBP␤ gene. Several lines of evidence indicate an essential role of the cis-elements and members of the CREB/ATF1 family in the adipocyte differentiation process.
Preparation of Mouse Embryo Fibroblasts (MEFs)-CREBϩ/Ϫ females were crossed with CREBϩ/Ϫ males, and embryos were dissected 14.5 days after detection of vaginal plugs. After the head and internal organs were removed, embryos were minced and incubated in trypsin for 30 min at 37°C. Dispersed cells were harvested (and saved), and fresh trypsin was added to the undispersed clumps. This step was repeated for three times until the embryo was virtually completely digested. After a low speed centrifugation, the pooled dispersed cells were resuspended in DMEM containing 10% (v/v) fetal calf serum and propagated. MEF cells were induced into differentiation using the same protocol employed for 3T3-L1 preadipocytes.
Gene Constructs and Mutations-A DNA fragment containing the mouse C/EBP␤ cDNA (sequence information from GenBank TM ) was used as a probe to screen a FixII mouse (BALB/cJ) genomic DNA library (provided by D. Nathans, Johns Hopkins University School of Medicine). After three rounds of screening, positive clones were selected and amplified. The amplified phage DNA was subjected to restriction digestion and Southern blotting and probed with a 5Ј-proximal promoter fragment of mouse C/EBP␤ cDNA (nucleotides Ϫ341 to ϩ50). Positive bands of appropriate size were cloned and sequenced. The promoter sequences were then cloned into pGL3basic luciferase reporter expression vector (Promega). Serial 5Ј-deleted reporter constructs of the C/EBP␤ promoter were constructed using nested deletion techniques (Exo Mung Bean Deletion Kit, Stratagene).
TGA-site mutant constructs, e.g. MT1, MT2, and DMT, were generated by the PCR method as described previously (28). In TGA1, the core TGA site at Ϫ111 bp was mutated from ATGACGCG to AACTCGCG (mutated bases underlined), and in TGA2 the core TGA site at Ϫ65 bp was mutated from GTGACGCAG to GACTCGCAG. To construct DMT with mutations on both TGA sites, a unique ApaI restriction site between the two TGA sites was utilized. After double digestion with ApaI and XhoI, the DNA fragment containing the wild-type TGA2 site in construct MT1 was replaced with a mutated TGA2 fragment from construct MT2. The authenticity of the mutations was verified by sequencing.
Transfection and Reporter Gene Analysis-Proliferating (ϳ70% confluent) 3T3-L1 preadipocytes were transiently transfected with 2 g of promoter-reporter construct along with 8 g of carrier DNA using the calcium phosphate precipitation method (29). After 2 days post-confluent (about 72 h or more after transfection) in culture, cells were treated with the MDI differentiation inducer mixture. Eight hours after induction, cells were harvested and cell extracts were prepared and assayed for luciferase activity.
Anti-CREB, -ATF1, and -ATF1/2-(which recognizes both CREB and ATF1) were used in EMSA supershift experiments. Anti-CREB was generously provided by Dr. David Ginty (Dept. of Neuroscience, Johns Hopkins University School of Medicine, Baltimore, MD); anti-ATF1 and anti-ATF1/2 were from Dr. Wolfgang Schmid (Division of Molecular Biology of the Cell I, German Cancer Research Center, Heidelberg, Germany). The same antibodies were used for ChIP assays.
Chromatin Immunoprecipitation Analysis-ChIP analyses were performed using an assay kit (Upstate Biotechnology, Lake Placid, NY) according to the manufacturer's protocol. Two-day post-confluent 3T3L1 or MEF cells (one 10-cm dish) were treated with MDI for 1 h, after which cells were cross-linked with 1% formaldehyde for 30 min at room temperature. After washing twice with ice-cold PBS containing protease inhibitors, scraping and centrifugation, cell pellets were resuspended in SDS lysis buffer. After incubation for 10 min at 4°C, the cell lysates were sonicated six times with each time being 25 s using a sonicator (550 Sonic Dismembrator, Fisher Scientific). After centrifugation, the supernatant was diluted in ChIP dilution buffer and then incubated overnight at 4°C with anti-CREB, anti-ATF1, anti-ATF1/2 (which can recognize both CREB and ATF1), anti-phospho-CREB, or anti-CREM. Immune complexes were recovered by the addition of 60 l of salmon sperm DNA/ protein A-agarose-50% slurry and incubation for 2 h at 4°C with rotation. Agarose beads were pelleted by gentle centrifugation (1000 rpm at 4°C). The beads were sequentially washed with low and high salt buffer, LiCl buffer, and finally twice with TE (Tris, 10 mM, pH 8.0; EDTA, 1 mM) buffer. After washing, the immune complexes were eluted by incubation for 15 min at 25°C with 200 l of fresh elution buffer (1% SDS, 0.1 M NaHCO 3 ). To reverse the cross-linking of DNA, 8 l of 5 M NaCl were added and incubated for 4 h or overnight at 65°C. After treatment with proteinase K for 1 h at 45°C, DNA was recovered by phenol-chloroform extraction and ethanol precipitation. The DNA pellets were resuspended in 50 l of TE buffer. PCR amplification was carried out for 35 cycles with 2 l of sample DNA solution, and PCR products were separated on 2% agarose gels in 1ϫ TBE. Two primers were used to amply the segment flanking the two TGA sites with forward primer 5Ј-GCCCTCTCGCGC-TC-3Ј and reverse primer 5Ј-GGCTCCGCTGCGTC-3Ј.
Oil Red O Staining-Cell monolayers were washed three times with PBS and then fixed for 2 min with 3.7% formaldehyde. Oil red O (0.5% in isopropanol) was diluted with water (3:2), filtered through a 0.45-m filter, and incubated with the fixed cells for 1 h at room temperature. Cells were washed with water, and the stained fat droplets in the cells were visualized by light microscopy and photographed.
Ecdysone-inducible VP16-CREB and KCREB Expression System-The Ecdysone-inducible expression system was employed to prepare stably transfected 3T3-L1 preadipocytes that could be induced to express VP16-CREB, CREB-DIEDML, LacZ, or KCREB. The open reading frame for KCREB was isolated from the plasmid, pRSV-KCREB, as a HindIII-EcoRI fragment. This fragment was subjected to PCR with a 5Ј primer that introduced a new HindIII site and a consensus Kozak translation initiation sequence (GCCACC) immediately upstream of the first methionine codon. The resulting PCR product was purified by electrophoresis on a 1% agarose gel and ligated into the HindIII and EcoRI sites of the plasmid, pIND. The open reading frame for VP16 (amino acids 412-490) was excised from the plasmid, pVP16 (Arthur Gutierrez-Hartman, University of Colorado Health Sciences Center, Denver, CO) as a HindIII-BamHI fragment. This fragment was also subjected to PCR to introduce a Kozak sequence immediately upstream of the translation start site and was then directly ligated to a BglII-EcoRI fragment containing the DNA binding domain (amino acids 217-327) of CREB-327 excised from the plasmid pRSET-CREB (James Hoeffler, Invitrogen, Carlsbad, CA). The chimeric VP16-CREB gene was ligated into the HindIII and EcoRI sites of pIND. As a control, we generated a chimeric gene composed of the VP16 transactivation domain linked to the (non)DNA binding domain of KCREB. The CREB-DIEDML open reading frame was excised from a plasmid provided by Richard Goodman (Volum Institute) as a HindIII-XbaI fragment and ligated directly into the corresponding sites of pIND. The resulting plasmids were confirmed by restriction enzyme mapping and sequencing.

Dual CREs in the Proximal Promoter of the C/EBP␤ Gene
Mediate Reporter Gene Expression Early in the Adipocyte Differentiation Program-A 300-bp cDNA probe corresponding to the 5Ј-untranslated region of the C/EBP␤ mRNA was used to screen a mouse genomic library to obtain the 5Ј-flanking region of the gene (results not shown). An insert of ϳ6 kb was isolated, mapped, and sequenced. Alignment with the sequence of the C/EBP␤ gene from the NCBI mouse genome data base verified its location in the 5Ј-flanking region of the gene. A series of 5Ј-truncated promoter fragments (beginning at Ϫ2912 and terminating at ϩ33 relative to the start site of transcription) were isolated and inserted into the pGL3basic luciferase reporter vector (Fig. 1A).
To identify promoter sequences necessary for activation of reporter gene expression during adipogenesis, the 5Ј-truncated promoter-luciferase constructs were transiently transfected into 3T3-L1 preadipocytes when cells reached ϳ70% confluence. Two days after achieving confluence the preadipocytes were subjected to our standard adipocyte differentiation protocol (25), which involved exposure to medium containing MDI (methylisobutylxanthine, dexamethasone, and insulin). Fig. 1B illustrates the protocol for analysis of reporter gene activity. Thereafter, cells were harvested every 2 h during the early phase (0 -14 h) of the differentiation program, and cell extracts were prepared for luciferase activity measurements. Maximal luciferase activity (ϳ10-fold greater than non-induced controls) occurred ϳ8 h following induction after which activity began to decline (results not shown). As shown in Fig. 1C deletions between Ϫ2912 bp and Ϫ117 bp promoter had little effect on reporter activity. However, deletion of the region between Ϫ117 bp and Ϫ107 bp resulted in a reduction of luciferase activity of ϳ50%. Further deletion of the segment between Ϫ107 bp and Ϫ50 bp virtually abolished reporter activity. These findings indicated that two regions in the proximal promoter, between Ϫ117 bp and Ϫ107 bp and between Ϫ107 bp and Ϫ50bp, were crucial for promoter-reporter activity induced by the differentiation inducers. Using the TESS (transcription element search system) to scan the promoter sequence of the C/EBP␤ gene, two CRE-like elements were identified. Both elements share the same half-CRE core sequence TGACG, located at Ϫ111/Ϫ107 bp (referred to as TGA1) and at Ϫ65/Ϫ61 bp (referred to as TGA2).
The importance of the two CRE-like elements for promoter activity was verified by site-directed mutagenesis. Consistent with the results of the 5Ј-deletion experiments, mutation of each of the core CRE consensus sequences of the TGA1 and TGA2 sites (TGA 3 ACT) resulted in a 50 -60% reduction of promoter activity (Fig. 1D). Mutation of both sites virtually abolished promoter activity. The two CRE-like elements are located just upstream of a consensus TATA box (Fig. 1A). In the experiments described below the Ϫ117-bp construct was used as a minimal promoter-reporter.
Because members of the CREB family of transcription factors are known to bind to sequences of this type, supershift experiments were conducted with antibodies directed against these family members. Co-incubation of antibody, directed against CREB with labeled TGA1 or TGA2 oligonucleotides and nuclear extracts from 3T3-L1 preadipocytes after exposure to MDI for 0, 2, and 8 h, produced supershifted bands (Fig. 2B). Although most of the protein-TGA complexes were removed/ supershifted by this antibody, a significant fraction remained suggesting the presence of another TGA site-binding protein of similar size in the nuclear extract. This additional TGA-binding protein appears to be ATF1, a member of the CREB family of transcription factors. Thus, preincubation with antibody against ATF1 produced a weakly supershifted band (Fig. 2C,  lanes 3 and 6), however, antibody that recognizes both CREB and ATF1 completely supershifted the band formed in the absence of specific antibody (Fig. 2C, lanes 2 and 5). Taken together, these findings indicate that both CREB and ATF1 interact with the TGA oligonucleotides and that CREB is the predominant isoform present in nuclear extracts from 3T3-L1 preadipocytes.
Expression of C/EBP␤ Is Correlated with Phosphorylation of CREB-The expression of CREB, ATF1, phospho-CREB, and C/EBP␤ was analyzed following treatment of 3T3-L1 preadipocytes to differentiation inducers. Cell lysates, prepared at various times after treatment with MDI, were subjected to immunoblotting with antibodies against each of the three proteins (Fig. 3). CREB and ATF1 were expressed constitutively over the entire time course (up to 48 h after induction) even prior to exposure to differentiation inducers and did not correlate with the onset of expression of C/EBP␤. C/EBP␤ was rapidly expressed, i.e. at 30 min, and reached a maximum between 2 and 4 h after induction. Correlated with the expression of C/EBP␤ was the appearance of phospho-CREB and phospho-ATF1. Phospho-CREB and phospho-ATF1 were detected as early as 5 min after induction, remained high until 2-4 h and then declined. Chromatin immunoprecipitation (ChIP) experiments (shown below) revealed that phospho-CREB binds to the proximal promoter of the C/EBP␤ gene in 3T3-L1 preadipocytes within 1 h after induction.

FIG. 2. Electrophoretic mobility shift and supershift analyses with wt and mutant oligonucleotides corresponding to the two C/EBP␤ promoter TGA sites and nuclear extracts of 3T3-L1 cells following induction of differentiation.
A, EMSA with wt and mutant TGA-site oligonucleotides. Where indicated a 20-fold excess of an unlabeled consensus CREB binding-site oligonucleotide probe was used. B, supershift-EMSA with wt TGA-site oligonucleotides with anti-CREB antibody. C, supershift-EMSA with wt TGA-site oligonucleotides with antibody that recognizes CREB, CREB and ATF1 (ATF1/2), or with ATF1-specific antibody. PI serum refers to preimmune serum.

Effect of Constitutively Active CREB (VP16-CREB) and Dominant-negative CREB (A-CREB) on C/EBP␤ Promoter-reporter
Gene Activity-The effect of constitutively active CREB on transcription driven by the C/EBP␤ promoter was further evaluated in co-transfection experiments with 3T3-L1 preadipocytes. Constitutively active CREB (VP16-CREB) was co-transfected with the minimal (Ϫ117 bp) C/EBP␤ promoter-reporter construct containing wt or mutated TGA1 and TGA2 regulatory elements. VP16-CREB, which possesses the CRE binding domain of CREB, strongly activated reporter gene expression (Fig. 4A). Deletion of the DNA binding domain in VP16-CREB-dLZ prevented activation of the reporter gene (Fig. 4A). Moreover, mutation of the TGA sites abrogated transactivation by VP16-CREB. It can be concluded that activation of the C/EBP␤ promoter-reporter gene requires interaction between CREB and the CREs.
Dominant-negative A-CREB possesses a leucine zipper that allows dimerization with CREB family members but lacks a functional DNA binding domain. Instead, A-CREB contains an acidic region that replaces the basic DNA binding region. The acidic extension forms a heterodimeric coiled-coil with the basic region of endogenous CREB stabilizing the interaction. Thus, heterodimers formed cannot bind to CRE elements (33). As shown in Fig. 4B, co-transfection of the A-CREB expression vector with the Ϫ117-bp C/EBP␤ promoter-reporter construct inhibited reporter gene expression by preadipocytes treated with differentiation inducers. The fact, that A-CREB markedly inhibits the ability of the minimal promoter to drive reporter gene expression, provides additional evidence that CREB/ ATF1 transcriptionally activates the C/EBP␤ gene.

CREB(Ϫ/Ϫ) Mouse Embryonic Fibroblasts (MEFs) Exhibit Reduced Expression of C/EBP␤ and Adipogenesis-When
subjected to the same protocol that triggers differentiation of 3T3-L1 preadipocytes, MEFs can be induced to differentiate into adipocytes (9). In a previous study we showed that disruption of the C/EBP␤ gene in MEFs prevented adipogenesis (9). Given that CREB activates transcription of the C/EBP␤ gene in 3T3-L1 preadipocytes, we suspected that disruption of the CREB gene would block expression of C/EBP␤ and thereby, differentiation. To test this hypothesis CREB(Ϫ/Ϫ) MEFs were employed. MEFs were isolated from both E14.5 CREB(Ϫ/Ϫ) and wt littermate mouse embryos. The genotypes of the MEFs were verified by PCR of genomic DNA (not shown), and disruption of CREB expression was assessed by Western blotting. Expression of C/EBP␤ (Fig. 5A) and the accumulation of cytoplasmic triglyceride as assessed by Oil Red O staining (Fig. 5B) were markedly reduced in CREB(Ϫ/Ϫ) MEFs compared with

. Effect of constitutively active CREB (VP16-CREB) and dominant-negative CREB (A-CREB) on the expression of wt and mutant C/EBP␤ promoter (DMT)-reporter genes.
A, preadipocytes were co-transfected with either a minimal (Ϫ117 bp) wt or mutant (DMT; both TGA-1 and -2 mutated) C/EBP␤ promoter-reporter constructs along with: empty vector, VP16-CREB, or VP16-CREBdLZ (lacking the CREB DNA binding domain). B, preadipocytes were cotransfected with the minimal (Ϫ117 bp) wt C/EBP␤ promoter-reporter construct along with empty vector or dominant negative A-CREB. Two days later the cells were treated or not with differentiation inducers. that in wt MEFs. The fact that expression of C/EBP␤ and the accumulation of triglyceride were not totally blocked was apparently due to compensatory up-regulation of ATF1 and CREM (Fig. 5A). The latter CREB family members might be expected to partially compensate for the loss of CREB protein, because they both can transactivate, albeit less strongly, the C/EBP␤ gene promoter (shown below).

Binding of CREB Family Members to Chromatin-associated C/EBP␤ Promoter Sequences in 3T3-L1 Preadipocytes and MEFs and the Effect of CREB Family Members on Transcription Driven by the C/EBP␤ Promoter-
The binding of CREB family members to chromatin-associated C/EBP␤ promoter TGA elements was investigated by ChIP analysis. Two-day post-confluent 3T3-L1 (or MEF) cells were treated (or not) with adipocyte differentiation inducers for 1 h and then with formaldehyde to cross-link DNA-protein complexes. After sonication, chromatin fragments were immunoprecipitated with CREB, ATF1, ATF1/2 (which can recognize both CREB and ATF1), phospho-CREB, or CREM antibody, and the immunoprecipitated fragmented DNA was subjected to PCR to amplify C/EBP␤ promoter DNA (between Ϫ140 bp and ϩ33bp) containing the dual TGA regulatory elements.
As illustrated in Fig. 6A, antibodies to CREB, ATF1, or ATF1/2 (which recognizes both CREB and ATF1) immunoprecipitated TGA element-containing promoter fragments from both 3T3-L1 preadipocytes and MEFs, whereas preimmune IgG did not. These results indicate specific association of these proteins with the proximal C/EBP␤ gene promoter. In all cases, the CREB family members associated with the promoter fragments both before and after exposure to differentiation inducers, although in some cases the association was stronger after exposure to differentiation inducers. Consistent with the results of EMSA showing that these homologues bind to TGA regulatory elements in vitro (Fig. 2), the results of ChIP analysis show that most of the CREB family members associate with the TGA-containing region of the C/EBP␤ promoter ex vivo. Similar findings were obtained with anti-phospho-CREB antibody. However, because phosphorylation of CREB is dependent upon kinase activation by the differentiation inducers, the association of phospho-CREB with the C/EBP␤ gene promoter was detected only after exposure of the cells to differentiation FIG. 6. Effect of differentiation inducers on chromatin immunoprecipitation of C/EBP␤ promoter DNA sequences and effect of CREB, CREM, and ATF1 on transcription driven by the C/EBP␤ promoter. A, formaldehyde cross-linked fragmented, chromatin-associated DNA from 1 ϫ 10 7 3T3-L1 preadipocytes, wild-type MEFs (WT-MEF), or CREB(Ϫ/Ϫ) MEFs (KO-MEF) was immunoprecipitated with preimmune (PI) serum or antibodies to CREB, phospho-CREB, ATF1, CREB, and ATF1 (ATF1/2) or CREM. Fragmented DNA was subjected to PCR amplification along with specific primers flanking the two TGA (TGA1 and TGA2) sites. In all cases only the expected single ϳ200-bp PCR product was observed. B, reporter gene expression in 3T3-L1 preadipocytes that had been co-transfected with a C/EBP␤ promoter-luciferase construct and expression vectors for CREB, CREM, or ATF1. Luciferase assays were conducted 48 h after cotransfection. The -fold induction is expressed as activity of CREB, CREM, or ATF1 relative empty vector.

FIG. 7. Effect of constitutively active and dominant-negative forms of CREB on C/EBP␤ expression and adipogenesis. 3T3-L1
preadipocyte cells expressing VP16-CREB, CREB-DIEDML, or KCREB or control cells (stably transfected with the plasmids, pVgRXR and pIND-LacZ) were grown to confluence as described under "Experimental Procedures." Ponasterone was added to medium at a final concentration of 10 M to induce expression of LacZ and each CREB isoform. Some cells were treated with 10 g/ml insulin, 1 M dexamethasone, and 3 M isobutylmethylxanthine (MDI) as indicated. A, expression of LacZ, or each CREB isoform (indicated above each lane) was induced with ponasterone for 36 h. Some cells also received MDI. Cell lysates were subjected to immunoblotting for LAP and LIP isoforms of C/EBP␤. B, preadipocytes were treated with ponasterone for 36 h to induce the expression of LacZ or CREB isoforms. An equal amount of cell lysates was subjected to Western blotting analysis to verify the expression of CREB isoforms. C, preadipocytes were treated with ponasterone for 8 days to induce expression of LacZ or CREB isoforms. Some cells were treated with MDI for 48 h and then refed with medium containing 10 g/ml insulin every 2 days. On day 8 cells were stained with Oil Red O as an indicator of triglyceride accumulation. Representative photomicrographs are shown. inducers (Fig. 6A). Although CREB family members can bind to the C/EBP␤ promoter regardless of prior treatment with differentiation inducers (Fig. 6A), their capacity to drive reporter gene expression mediated by the C/EBP␤ gene promoter is dependent upon "activation" by differentiation inducers (Fig.  1C). This and the fact that the differentiation inducers activate phosphorylation of CREB indicate that binding to the promoter is not dependent upon phosphorylation. These findings are consistent with those of others (34,35) that phosphorylation of CREB facilitates its interaction with CBP/p300, a component of the RNA polymerase II transcriptional initiation complex.
Although CREM did not appear to associate with C/EBP␤ promoter sequences in 3T3-L1 preadipocytes as detected by ChIP analysis (Fig. 6A), CREM was found to associate with C/EBP␤ promoter sequences in MEFs upon treatment with differentiation inducers as detected by ChIP analysis (Fig. 6A). It should be noted, however, that both CREM and ATF1 were overexpressed by CREB(Ϫ/Ϫ) MEFs (Fig. 5A). The interaction of CREM with the C/EBP␤ promoter sequences was increased in CREB(Ϫ/Ϫ) MEFs before by treatment with differentiation inducers as detected by ChIP analysis (Fig. 6A).
Members of the CREB family of transcription factors not only bind to the C/EBP␤ promoter ex vivo but also activate reporter gene expression driven by the promoter in 3T3-L1 preadipocytes. As shown in Fig. 6B expression vectors for CREB, CREM, and ATF1, co-transfected into 3T3-L1 preadipocytes with a minimal C/EBP␤ promoter-reporter, strongly activate reporter gene expression.
Constitutively Active CREB (VP-16CREB or CREB-DIEDML) Activates, and Dominant-negative CREB (K-CREB) Inhibits, Expression of C/EBP␤ and Adipogenesis-To verify that CREB activates the expression of C/EBP␤ and adipogenesis, the ecdysone-inducible system was employed to control expression of CREB proteins in 3T3-L1 preadipocytes. 3T3-L1 preadipocyte lines were established that harbor expression vectors in which constitutively active CREB (VP16-CREB and CREB-DIEDML), dominant-negative CREB (K-CREB), or a control vector (LacZ) are under the control of the ecdysone promoter. The effects of expression of the transgenes, induced by treatment of growth-arrested preadipocytes with ponasterone for 36 h, were compared with preadipocytes induced to differentiate using the standard differentiation protocol. As illustrated in Fig. 7 (A and C), expression of constitutively active VP16-CREB or CREB-DIEDML (Fig. 7B) was sufficient to induce the expression of C/EBP␤ and adipogenesis (as indicated by Oil Red O staining of cytoplasmic triacylglycerol) without induction of differentiation with the inducers (MDI) normally required. In contrast, expression of dominant active K-CREB inhibited the expression of C/EBP␤ protein and adipogenesis induced by treatment of the preadipocytes with differentiation inducers (KCREB plus MDI in Fig. 7C). Expression of LacZ by control cells had no effect on expression of C/EBP␤ or adipogenesis in cells treated or not with differentiation inducers. Taken together, these findings demonstrate that active CREB protein is sufficient and required for the expression of C/EBP␤ protein and the induction of adipocyte differentiation.
Effect of CREB, ATF-1, ATF-2, or c-Jun Antisense Oligonucleotides on Adipogenesis by 3T3-L1 Preadipocytes-To gain further insight into the dependence of adipogenesis on members of the CREB family of transcription factors, 3T3-L1 preadipocytes were treated with antisense oligonucleotides corresponding to sequences in the CREB, ATF-1, ATF-2, CREM, or c-Jun mRNAs. As expected, the CREB antisense oligonucleotides virtually abolished adipogenesis as indicated by the accumulation of cytoplasmic triglyceride (Fig. 8). ATF-1, and to a lesser extent ATF-2 antisense oligonucleotides, decreased but did not eliminate the expression of cytoplasmic triglyceride. CREM and c-Jun antisense oligonucleotides had no detectable effect on triglyceride accumulation (Fig. 8). These findings further document the requirement of CREB and ATF-1 for the differentiation of 3T3-L1 preadipocytes into adipocytes. DISCUSSION In this study, we investigated the mechanism by which CREB activates expression of C/EBP␤ early in the adipocyte differentiation program (7,9,11,36). Once activated (10), C/EBP␤ triggers a cascade of transcriptional events, notably activation of the C/EBP␣ and PPAR␥ genes (14,15,17). C/EBP␣ and PPAR␥ then activate the transcription of the set of genes that give rise to the adipocyte phenotype (6,12,13).
We identified dual CRE-like elements (i.e. TGA1 and TGA2 (Fig. 1A)) within the proximal promoter of the C/EBP␤ gene to which CREB family members, i.e. CREB and ATF1, bind (Fig.  2, A-C) and activate transcription (Figs. 1D and 4A). CREB and ATF1 are constitutively expressed by 3T3-L1 preadipocytes and MEFs throughout the differentiation program, even in growth-arrested cells, prior to induction of differentiation. EMSA and supershift experiments show that, although CREB and ATF1 together account for virtually all binding to the TGA-1 and -2 elements, CREB accounts for most of the binding activity (Fig. 2, A-C).
Nuclear extracts from CREB(Ϫ/Ϫ) MEFs exhibit greatly reduced binding to oligonucleotide probes corresponding to the TGA regulatory elements in the C/EBP␤ promoter (data not shown). The remaining residual binding activity is due to CREM and ATF1 (Fig. 6A), whose expression is markedly upregulated in CREB(Ϫ/Ϫ) MEFs (Fig. 5A). Likewise, the expression of C/EBP␤ and adipogenesis are greatly reduced in CREB(Ϫ/Ϫ) MEFs (Fig. 5, A and B). The fact that the reduction of the level of C/EBP␤ and adipogenesis correspond closely to the reduction of DNA binding activity to the TGA elements provides compelling evidence that CREB and its cis regulatory elements play important roles in the early phase of the adipocyte differentiation program. In an analogous manner CREM has been shown to compensate for CREB deficiency in human adrenocortical cancer cell line H295R.
Other protein families, notably AP1 (Fos/Jun) protein, share similar DNA consensus binding sites with CREB/ATF1 family. Nevertheless, reduction in expression of Jun using antisense Jun had no effect on adipogenesis (Fig. 8). It has also been shown that overexpression of DeltaFosB inhibits adipogenesis while increasing bone formation (37). Likewise, overexpression of Finkel-Biskis-Reilly osteosarcoma virus v-Fos inhibits adipocyte differentiation (38).
Immediately following induction of differentiation CREB and ATF1 undergo phosphorylation (Fig. 3). This is followed by the phosphorylation of C/EBP␤ (18). Whereas phosphorylation of CREB and ATF1 enhances binding to regulatory elements in the C/EBP␤ gene promoter (Fig. 6A), the primary effect of phosphorylation of CREB and ATF1 appears to be transactivation of the C/EBP␤ gene. This is consistent with our findings that activation of reporter gene expression and phosphorylation of CREB and ATF1 occur only after treatment with differentiation inducers (Figs. 1C, 1D, and 3). Chromatin immunoprecipitation experiments with anti-phospho-CREB antibody showed that phospho-CREB, which occurs rapidly upon induction (Fig. 3), binds to the proximal promoter of the C/EBP␤ gene in intact preadipocytes (Fig. 6A). Furthermore, expression of constitutively active CREB, in which the CREB transcriptional activation domain is replaced by VP16, a strong viral transcriptional activator, activated expression of C/EBP␤ and differentiation (Fig. 7A).
Phosphorylation of CREB is known (39 -46) to be activated by two components, i.e. cAMP and insulin, of the inducer mixture used to trigger differentiation of 3T3-L1 preadipocytes and MEFs (25). Thus, it appears that protein kinase A and a downstream kinase of the IGF-1 receptor signal transduction pathway phosphorylate CREB early in the differentiation program. CREB is known to be phosphorylated by several protein kinases, including protein kinase A and AKT (26,47,48). Consistent with involvement of the IGF-1 receptor signal transduction pathway, preadipocytes are known to possess a high level of IGF receptors but few insulin receptors (49). Moreover, it has been shown that in preadipocytes insulin (at the high level used in the differentiation inducer mixture) acts through the IGF receptor (49,50). Recently, it was shown (51) that RhoA/Rho kinase blocks adipogenesis and promotes myogenesis in mesenchymal precursor cells. This observation also suggests that blockade of the IGF-1 signaling pathway blocks activation of CREB.