CCAAT/Enhancer-binding Protein δ Is a Critical Regulator of Insulin-like Growth Factor-I Gene Transcription in Osteoblasts*

Insulin-like growth factor-I (IGF-I) plays a major role in promoting skeletal growth by stimulating bone cell replication and differentiation. Prostaglandin E2 and other agents that induce cAMP production enhance IGF-I gene transcription in cultured rat osteoblasts through a DNA element termed HS3D, located in the proximal part of the major rat IGF-I promoter. We previously determined that CCAAT/enhancer-binding protein δ (C/EBPδ) is the key cAMP-stimulated regulator of IGF-I transcription in these cells and showed that it transactivates the rat IGF-I promoter through the HS3D site. We now have defined the physical-chemical properties and functional consequences of the interactions between C/EBPδ and HS3D. C/EBPδ, expressed in COS-7 cells or purified as a recombinant protein from Escherichia coli, bound to HS3D with an affinity at least equivalent to that of the albumin D-site, a known high affinity C/EBP binding sequence, and both DNA elements competed equally for C/EBPδ. C/EBPδ bound to HS3D as a dimer, with protein-DNA contact points located on guanine residues on both DNA strands within and just adjacent to the core C/EBP half-site, GCAAT, as determined by methylation interference footprinting. C/EBPδ also formed protein-protein dimers in the absence of interactions with its DNA binding site, as indicated by results of glutaraldehyde cross-linking studies. As established by competition gel-mobility shift experiments, the conserved HS3D sequence from rat, human, and chicken also bound C/EBPδ with similar affinity. We also found that prostaglandin E2-induced expression of reporter genes containing human IGF-I promoter 1 or four tandem copies of the human HS3D element fused to a minimal promoter and show that these effects were enhanced by a co-transfected C/EBPδ expression plasmid. Taken together, our results provide evidence that C/EBPδ is a critical activator of IGF-I gene transcription in osteoblasts and potentially in other cell types and species.

Insulin-like growth factor-I (IGF-I) plays a major role in promoting skeletal growth by stimulating bone cell replication and differentiation. Prostaglandin E 2 and other agents that induce cAMP production enhance IGF-I gene transcription in cultured rat osteoblasts through a DNA element termed HS3D, located in the proximal part of the major rat IGF-I promoter. We previously determined that CCAAT/enhancer-binding protein ␦ (C/EBP␦) is the key cAMP-stimulated regulator of IGF-I transcription in these cells and showed that it transactivates the rat IGF-I promoter through the HS3D site. We now have defined the physical-chemical properties and functional consequences of the interactions between C/EBP␦ and HS3D. C/EBP␦, expressed in COS-7 cells or purified as a recombinant protein from Escherichia coli, bound to HS3D with an affinity at least equivalent to that of the albumin D-site, a known high affinity C/EBP binding sequence, and both DNA elements competed equally for C/EBP␦. C/EBP␦ bound to HS3D as a dimer, with protein-DNA contact points located on guanine residues on both DNA strands within and just adjacent to the core C/EBP half-site, GCAAT, as determined by methylation interference footprinting. C/EBP␦ also formed protein-protein dimers in the absence of interactions with its DNA binding site, as indicated by results of glutaraldehyde cross-linking studies. As established by competition gel-mobility shift experiments, the conserved HS3D sequence from rat, human, and chicken also bound C/EBP␦ with similar affinity. We also found that prostaglandin E 2 -induced expression of reporter genes containing human IGF-I promoter 1 or four tandem copies of the human HS3D element fused to a minimal promoter and show that these effects were enhanced by a co-transfected C/EBP␦ expression plasmid.

Taken together, our results provide evidence that C/EBP␦ is a critical activator of IGF-I gene transcription in osteoblasts and potentially in other cell types and species.
Insulin-like growth factor-I (IGF-I), 1 a conserved 70-residue secreted protein, plays a fundamental role in regulating somatic growth in mammals and other vertebrate species (1,2). IGF-I is synthesized by many cells including osteoblasts (1,2) and can act as a growth and differentiation factor within the skeleton as well as in other tissues (2)(3)(4). Production of IGF-I by skeletal cells is controlled by local and systemic agents, including hormones (5)(6)(7)(8)(9)(10). Both parathyroid hormone and prostaglandin E 2 (PGE 2 ) stimulate IGF-I synthesis in cultured osteoblasts by enhancing IGF-I gene expression (11,12) through mechanisms that are secondary to hormonal induction of cAMP accumulation (5,7,11). Previous studies have shown that PGE 2 stimulates IGF-I gene transcription in osteoblasts by activating promoter 1, the major IGF-I promoter in bone (12)(13)(14) and in most other tissues (15). We have found that induction of IGF-I transcription by PGE 2 is part of a primary hormonal response that does not require ongoing protein synthesis (16), but like other cAMP-activated pathways, does require the catalytic subunit of cAMP-dependent protein kinase (13). We recently mapped a functional cAMP response element to the 5Ј-untranslated region of rat IGF-I exon 1 within a previously footprinted site termed HS3D (16) and identified CCAAT/enhancer-binding protein ␦ (C/EBP␦) as the principal cAMP-activated transcription factor in osteoblasts that binds to and transactivates IGF-I promoter 1 through the HS3D site (17).
The C/EBP family comprises a diverse group of transcriptional regulators with actions on tissue development and regeneration, inflammation, and intermediary metabolism (18). These proteins are members of the basic leucine zipper family of transcription factors (18,19) and share strong amino acid similarity in their COOH-domains, which contain motifs responsible for protein dimerization and DNA binding (18). The first C/EBP proteins to be characterized, C/EBP␣ and C/EBP␤ (20 -22), function as transcriptional activators and play major roles in adipocyte differentiation and in regulating gene expression in the liver and other tissues (18,(23)(24)(25)(26). C/EBP␦ also has been implicated in the control of adipogenesis and in mediating the acute phase response to inflammatory stimuli (18,23,24). Its potential role in controlling hormone-activated IGF-I synthesis in bone cells had not been described until our recent report (17).
The current experiments were designed to assess interactions between C/EBP␦ and the HS3D DNA element of the major IGF-I promoter from both physical-chemical and functional perspectives. We find that C/EBP␦ binds to the HS3D site from the rat IGF-I gene with an affinity equivalent to that of a known high affinity C/EBP element from the rat albumin promoter (21,27) and that, like C/EBP␣ and C/EBP␤ (21,22,28), it can form protein-protein dimers in the absence of DNA. C/EBP␦ also binds to HS3D sites from the human and chicken IGF-I genes with high affinity and functions as a HS3D-de-pendent cAMP-inducible transcription factor for the major human IGF-I promoter. Taken together, our results provide evidence that C/EBP␦ is a critical regulator of IGF-I gene transcription in osteoblasts and potentially in other cell types and species.
C/EBP␦ and C/EBP␤ expression vectors pcDNA3-C/EBP␦, and pcDNA3-C/EBP␤ have been described (17). Bacterial expression plasmids for full-length and internally truncated C/EBP␦ were constructed as follows. The 5Ј end of the C/EBP␦ coding region was first modified by polymerase chain reaction by introducing a BamHI site (underlined) adjacent to the ATG codon (bold) with oligonucleotides, 5Ј-GCGGATC-CATGAGCGCCGCTCTTTTCAG-3Ј and 5Ј-TGTGATTGCTGTTGAA-GAGGT-3Ј. The amplified fragment was purified and digested by BamHI and NcoI and then was inserted into BamHI-and NcoI-digested pBS-C/EBP␦ (17) to make pBS-C/EBP␦/BamATG. After sequencing the amplified portion, the entire C/EBP␦ coding region was inserted into BamHI-and SalI-digested pET29a(ϩ) (Novagen, Madison, WI) to make pET29a-C/EBP␦. In this plasmid, the C/EBP␦ coding sequence has been fused in-frame to an NH 2 -terminal 33-residue S tag. To generate the internally truncated bacterial C/EBP␦ expression plasmids, pET29a-C/ EBP␦-⌬SacII and pET29a-C/EBP␦-⌬NcoI, pET29a-C/EBP␦ was digested with SacII or NcoI, and the plasmid-containing fragments were religated. In recombinant protein C/EBP␦-⌬SacII, amino acids 23 through 152 of C/EBP␦ have been eliminated, whereas in C/EBP␦-⌬NcoI, residues 2 through 68 have been deleted. A bacterial expression plasmid for C/EBP␤ was constructed by directionally cloning rat C/EBP␤ (22) into NcoI-and SalI-digested pET29a(ϩ) to generate pET29a-C/EBP␤. In this plasmid, the C/EBP␤ coding sequence has been fused in frame to an NH 2 -terminal 27 amino acid S-tag.
Gene Transfer Experiments-Transfection studies using primary rat osteoblasts were performed as described previously (13,16). IGF-I promoter 1-luciferase fusion genes were co-transfected with C/EBP expression plasmids or the empty expression vector and with a vector carrying the ␤-galactosidase gene under control of the SV40 promoter (Promega Corp., Madison, WI) to normalize for transfection efficiency. After transfection, the cells were incubated for 48 h until reaching confluent density. Cells then were rinsed with serum-free medium and treated for 6 h with vehicle (ethanol diluted 1:1000 in serum-free medium) or 1 M PGE 2 (in serum-free medium). After incubation, the medium was aspirated and cultures were rinsed with phosphate-buffered saline and lysed in cell lysis buffer (Promega Corp.), and luciferase activity was measured as described (13,16).
Nuclear Protein Extracts-Confluent osteoblast cultures were deprived of serum for 20 h. Cells then were rinsed with serum-free medium and incubated with vehicle (ethanol diluted 1:1000) or 1 M PGE 2 for up to 4 h. Medium was aspirated, and cultures were rinsed twice with phosphate-buffered saline at 4°C. Cells were harvested with a cell scraper and gently pelleted, and the pellets were washed with phosphate-buffered saline. Nuclear extracts were prepared by the method of Lee et al. (31) with minor modifications (13,16,17). Cells were lysed in hypotonic buffer (10 mM HEPES, pH 7.4, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM dithiothreitol) with 1% Triton X-100, phosphatase inhibitors (1 mM sodium orthovanadate, 10 mM sodium fluoride, 0.4 M microcystin CL), and protease inhibitors (0.5 mM phenylmethylsulfonyl fluoride, 1 g/ml pepstatin A, 2 g/ml leupeptin, 2 g/ml aprotinin). Nuclei were pelleted and resuspended in hypertonic buffer containing 0.42 M NaCl, 0.2 mM EDTA, 25% glycerol, and the phosphatase and protease inhibitors indicated above. Soluble proteins released by a 30-min incubation at 4°C were collected by centrifugation at 12000 ϫ g for 5 min, and the supernatant was dialyzed for 2 h against 2000 volumes of buffer (20 mM HEPES, pH 7.4, 100 mM KCl, 0.1 mM EDTA, 0.5 mM dithiothreitol, 1 mM sodium orthovanadate, 20% glycerol) containing the protease inhibitors listed above. Protein concentrations were determined using a modified Bradford assay (Bio-Rad).
C/EBP␦ and C/EBP␤ were expressed in transiently transfected COS-7 cells (17). Cells were grown in 150-mm-diameter tissue culture dishes and were transfected by calcium-phosphate precipitation using 20 g of the expression plasmid pcDNA3-C/EBP␦. Twenty-four h later, the medium was changed, and after an additional 24 h, cells were harvested, and nuclear extracts were prepared as described above.
Antibodies-Polyclonal antibodies to C/EBP␦ and C/EBP␤ were prepared in chickens (Gallina Biotech, Alberta, CA) using purified Stagged fusion proteins as antigens. IgY fractions were isolated from eggs by precipitation with polyethylene glycol followed by affinity purification using bacterial-derived antigen immobilized on Sepharose CL-4B (Amersham Pharmacia Biotech).
Western Blotting-Western immunoblotting was performed after transfer of electrophoresed proteins to nitrocellulose membranes. Membranes were incubated in blocking buffer consisting of 5% nonfat dry milk and 2% fetal bovine serum in TBS-T (20 mM Tris-Cl, pH 7.6, 137 mM NaCl, 0.05% Tween 20) for 1 h at 25°C. Affinity-purified antibody prepared against full-length bacterially expressed C/EBP␦ or C/EBP␤ (diluted 1:500 in blocking buffer) was then added for 1 h at 25°C. After washing the membranes in TBS-T, secondary antibody (rabbit antichicken IgY diluted 1:1000 in blocking buffer) was added for 1 h at 25°C. Subsequent steps were performed as described previously (17). Immunoreactive bands were visualized by enhanced chemiluminesence and exposure to x-ray film.
Preparation of Recombinant C/EBP Proteins-Recombinant proteins were generated in bacteria as follows. Plasmids pET29a-C/EBP␦, pET29a-C/EBP␦-⌬SacII, pET29a-C/EBP␦-⌬NcoI, and pET29a-C/EBP␤ were transformed into the BL21(DE3) strain of Escherichia coli. Bacterial cultures were grown to an A 600 of 0.6 in 50 ml of Circlegrow (Bio101, Vista, CA) containing 30 g/ml kanamycin and 34 g/ml chloramphenicol and then were induced to express recombinant proteins by addition of isopropyl-1-thio-␤-D-galactopyranoside (Sigma) to a final concentration of 1 mM for 3 h. Bacterial pellets were harvested by centrifugation, then were resuspended into 5 ml of binding/wash buffer (20 mM Tris-HCl, pH 7.5, 1.5 M NaCl, 1% Triton X-100) containing 6 M urea. Cells were lysed by one cycle of freezing and thawing followed by sonication. Bacterial debris was removed by centrifugation. The Stagged proteins were purified using S-agarose (Novagen), according to the manufacturer's protocol. Purified proteins were eluted in binding/ wash buffer supplemented with 2 M guanidine thiocyanate and 2 M urea, followed by dialysis against 20 mM HEPES, pH 7.9, 100 mM KCl, 2 mM EDTA, 20% glycerol, 0.5 mM phenylmethylsulfonyl fluoride, 0.5 mM dithiothreitol for 2 h at 4°C. The S tag was cleaved using biotinylated thrombin as described by the supplier (Novagen). Recombinant proteins were aliquoted and stored at Ϫ80°C until use.
Assay for Formation of C/EBP Homodimers-Two g of truncated recombinant C/EBP␦ protein (C/EBP␦-⌬NcoI) was incubated with or without 0.01% glutaraldehyde for 10 min at 25°C. Samples were separated by 8% polyacrylamide gel electrophoresis followed by Coomassie Blue staining.
DNA-Protein Binding Studies-Gel mobility shift experiments followed previously published methods (16,17). Oligonucleotides and competitors are listed in Table I. Radiolabeled double-stranded DNA probes were synthesized by annealing complementary oligonucleotides followed by fill-in of single-stranded overhangs with dCTP, dGTP, dTTP, and [␣-32 P]dATP (800 Ci/mmol, NEN Life Science Products) using the Klenow fragment of DNA polymerase I. Nuclear protein extracts or recombinant proteins were preincubated for 30 min on ice with 2 g of poly(dI-dC) without or with unlabeled specific or nonspecific DNA competitors in 25 mM HEPES, pH 7.6, 60 mM KCl, 7.5% glycerol, 0.1 mM EDTA, 5 mM dithiothreitol, and 0.05% bovine serum albumin. After the addition of 5 ϫ 10 4 cpm of DNA probe for 30 min on ice, samples were applied to 4 -12 and 4 -20% nondenaturing polyacrylamide gradient gels (Novex, San Diego, CA) or a 5% nondenaturing polyacrylamide gel. The dried gels were exposed to x-ray film at Ϫ80°C with intensifying screens. Quantitative DNA-protein binding studies were performed with a constant amount of protein (50 ng of bacterial recombinant C/EBP␦ or 1.3 g of COS-7 cell nuclear extract) and increasing concentrations of radiolabeled probe (0.5 to 100 nM). After electrophoresis, gels were dried, and the radioactivity in bands representing protein-bound DNA and free probe was measured by phosphoimager (Molecular Imager System, Bio-Rad). The dissociation constant (K d ) was calculated from these data as the negative reciprocal of the slope after results were graphed by Scatchard plot analysis.
Deoxynuclease I (DNase I) footprinting was performed as described (32,33). End-labeled double-stranded DNA probes flanking the HS3D site in rat IGF-I promoter 1 were generated by polymerase chain reaction, using one end-labeled and one unlabeled oligonucleotide primer (the primers were 5Ј-CTAAATCCCTCTTCTGCTTG-3Ј and 5Ј-AGATA-GAGCCTGCGCAAT-3Ј) as previously described (31,32). Graded amounts of recombinant bacterial C/EBP␦ protein were preincubated for 15 min with poly(dI-dC) in 25 mM HEPES, pH 7.6, 60 mM KCl, 7.5% glycerol, 0.1 mM EDTA, 5 mM dithiothreitol, and 0.05% bovine serum albumin, followed by the addition of labeled probe (4.0 ϫ 10 5 cpm/ sample) and incubation for 60 min on ice. The reaction mixture was then treated with DNase I (final concentration 0.23 g/ml, Worthington Biochemical Corp., Freehold, NJ) in 2.5 mM MgCl 2 and 2.5 mM CaCl 2 for 1 min at 25°C. Nuclease treatment was terminated by addition of 20 mM EDTA, 200 mM NaCl, 1% sodium dodecyl sulfate, and 10 g of yeast transfer RNA, followed by phenol-chloroform extraction and ethanol precipitation. Samples were analyzed after electrophoresis on an 8% polyacrylamide, 8 M urea gel and autoradiography for 16 h at Ϫ80°C with an intensifying screen.
Methylation interference assays were performed by published methods (34,35). Double-stranded DNA probes labeled at one end were synthesized as described above. Labeled probes were methylated by incubation with 0.2% dimethylsulfate in 50 mM sodium cacodylate and 1 mM EDTA at 25°C for 4 min followed by 2 cycles of ethanol precipitation. Recombinant bacterial C/EBP␦ protein (100 ng) was incubated with methylated labeled DNA for 30 min on ice, and the DNA-protein complex and free probe were separated by electrophoresis on a 5% polyacrylamide gel. The wet gels were exposed to x-ray film, and protein-bound and free probes were isolated and eluted. Eluted DNA was cleaved by 1 M piperidine for 30 min at 95°C followed by 3 cycles of lyophilization and reconstitution. Samples were analyzed after electrophoresis on an 8% polyacrylamide, 8 M urea gel and autoradiography for 6 h at Ϫ80°C with an intensifying screen.

RESULTS
Our previous studies defined HS3D as an atypical cAMP response element located in the 5Ј-untranslated region of rat IGF-I exon 1 that mediated hormonally activated IGF-I gene transcription in primary rat osteoblasts (13,16). We subsequently identified C/EBP␦ as the cAMP-regulated transcription factor responsible for hormonally stimulated gene expression in these cells (17). The current experiments were designed to investigate the physical-chemical properties of the interac-  2 and 4) demonstrate that each antiserum recognizes the respective antigen. Right panel, Western immunoblots of nuclear protein extracts from COS-7 cells transiently transfected with an expression plasmid for rat C/EBP␤ (lanes 5 and 7) or for rat C/EBP␦ (lanes 6 and 8) and probed with antibodies to each protein. tions between C/EBP␦ and the HS3D site and to determine whether C/EBP␦ was involved as a mediator of cAMP-activated transcription in IGF-I genes from species other than rats.
HS3D Is a High Affinity Binding Site for C/EBP␦-Quantitative gel-mobility shift assays were used to determine the affinity of C/EBP␦ for the HS3D DNA element following the methods outlined in "Experimental Procedures." In the first series of experiments, nuclear extracts from COS-7 cells expressing C/EBP␦ (Fig. 1) were used as the source of recombinant protein, and DNA-protein binding reactions were performed with a constant quantity of nuclear protein (1.3 g) and a 200-fold concentration range of 32 P-labeled double-stranded rat HS3D oligonucleotide (0.5-100 nM; Fig. 2A). Binding was saturable, with an EC 50 of ϳ5 nM DNA. The calculated K d of 4.78 nM was very similar to the value obtained in parallel experiments using the previously described high affinity C/EBP binding site from the rat albumin promoter (K d of 5.56 nM; Fig. 2B).
Analogous studies were performed using full-length recombinant C/EBP␦ expressed and purified from E. coli (Fig. 3, lane  1). As seen with COS-7 nuclear protein extracts, binding of 32 P-labeled double-stranded rat HS3D to bacterially derived C/EBP␦ was saturable, with an EC 50 of ϳ10 nM and a calculated K d of 7.83 nM, approximately half that obtained using the albumin C/EBP site as the labeled DNA probe (K d of 15.67 nM; Fig. 4).
Further evidence that HS3D functioned as a high-affinity binding site for C/EBP␦ was obtained from a series of crosscompetition gel-mobility shift experiments, using nuclear extracts from COS-7 cells expressing C/EBP␦. As seen in Fig. 5, unlabeled double-stranded HS3D and albumin D-site oligonucleotides competed identically with a 32 P-labeled HS3D probe for binding to C/EBP␦ and competed equivalently with 32 Plabeled albumin D-site DNA. Based on these results and on the information shown in Figs. 2 and 4, we conclude that HS3D is a high affinity binding site for full-length C/EBP␦. C/EBP␦ Binds to HS3D as a Dimer-Previous studies had shown that other members of the C/EBP family, including C/EBP␣ and C/EBP␤, were able to bind to idealized palindromic recognition sites as dimers (21,22,28). We performed experiments to assess the stoichiometry of interactions between C/EBP␦ and the nonpalindromic HS3D sequence. Bacterial fusion proteins were purified containing full-length and internally truncated rat C/EBP␦ fused to an NH 2 -terminal S-tag (Fig. 3, lanes 1 and 3). The truncated recombinant protein, C/EBP␦-⌬SacII, lacked amino acids 23 through 152 of C/EBP␦, which compose part of the transcriptional activation domain (18). Its absence would not be predicted to alter DNAprotein binding parameters. Both proteins could bind to the labeled HS3D probe, as indicated by results of gel-mobility shift experiments pictured in Fig. 6, lanes 2 4. HS3D is a high affinity binding site for bacterially expressed C/EBP␦. Quantitative gel-mobility shift experiments were performed with 32 P-labeled rat HS3D (A) or rat albumin D (Alb D) site double-stranded oligonucleotides (B), and full-length C/EBP␦ was expressed and purified from E. coli, as described under "Experimental Procedures." DNA binding was quantified by phosphoimager, and results of three experiments were plotted as shown. Binding curves are illustrated in the left panels, and Scatchard plots are illustrated in the right panels (B/F, protein-bound CPM/ unbound CPM). than did complexes with full-length protein (Fig. 6, W:W, compare lanes 4 and 2). When both C/EBP␦ isoforms were mixed together before the addition of the labeled oligonucleotide probe, an additional DNA-protein band of intermediate mobility (W:T) was detected after gel electrophoresis and autoradiography (lane 3), indicating formation of a relatively stable heterooligomeric complex containing full-length and truncated C/EBP␦ species. Thus, C/EBP␦ interacted with HS3D as a dimer.
C/EBP␣ and C/EBP␤ have been shown to form protein-protein dimers in solution even in the absence of DNA (21,22,28). To assess the potential for C/EBP␦ to self-associate, protein cross-linking studies were performed with the purified truncated recombinant protein, C/EBP␦-⌬NcoI (Fig. 3, lane 2), in the absence or presence of low concentrations of glutaraldehyde (0.01%) for 10 min at 25°C. As shown in Fig. 7, only the monomeric protein of ϳ32 kDa was visualized after gel electrophoresis in the absence of cross-linker, whereas a larger band of ϳ65 kDa additionally was observed after incubation with glutaraldehyde. Similar results were seen with full-length C/EBP␦. Thus, like C/EBP␣ and C/EBP␤, C/EBP␦ is able to form protein-protein dimers, which can be assembled into oligomeric DNA-protein complexes in the presence of a high affinity element like HS3D.
Identifying DNA-Protein Contact Sites-We initially characterized the HS3D element by in vitro DNaseI footprinting with rat liver nuclear protein extracts (33) and subsequently identified a qualitatively similar hormone-inducible DNA-protein interaction using rat osteoblast proteins (13). To determine whether recombinant C/EBP␦ could recognize the same segment of DNA as osteoblast-derived nuclear proteins, in vitro DNaseI footprinting was performed with end-labeled doublestranded DNA probes derived from rat IGF-I promoter 1 and graded concentrations of C/EBP␦ expressed and purified from bacteria. As seen in Fig. 8, recombinant C/EBP␦ protected the HS3D site (nucleotides 200 -214 on both strands) from nuclease digestion, effectively recapitulating what was observed previously with rat osteoblast nuclear proteins (13).
We next performed in vitro dimethylsulfate footprinting to define some of the nucleotides directly involved in protein-DNA binding. As depicted in Fig. 9, two guanosine residues on the upper DNA strand and three on the lower strand that collec- Autoradiographic exposure was for 4 h at Ϫ80°C with intensifying screens. W, wild type; T, truncated C/EBP␦ (C/EBP␦-⌬SacII); F, free probe. A schematic diagram of full-length and truncated C/EBP␦ proteins appears below the autoradiogram.

FIG. 5. Cross-competition between HS3D and an albumin D (Alb D)-site oligonucleotide.
Gel-mobility shift experiments were performed with 32 P-labeled rat HS3D or albumin D-site doublestranded probes, nuclear protein extracts from COS-7 cells transiently transfected with a C/EBP␦ expression plasmid, and the indicated competitor DNAs at 15, 50, 100, and 200-fold molar excess (HS3D and albumin D-site) or at 200-fold molar excess (Oct-1). Autoradiography was performed for 6 h at Ϫ80°C with intensifying screens. DNA binding was quantified by phosphoimager and was plotted as shown in the lower panels (squares and solid lines, HS3D competitor; triangles and dotted lines, albumin D-site competitor). Similar results were seen in two additional independent experiments. tively span 8 bp were required for binding by recombinant full-length C/EBP␦. Methylation of these residues inhibited binding, resulting in the accumulation of modified DNA in the unbound fraction and its subsequent cleavage by dimethylsulfate and piperidine. These results are in good agreement with our previous analysis of the HS3D site by site-directed mutagenesis (16,17). Table II depicts an alignment of the HS3D region of the rat IGF-I gene with analogous portions of the human, chicken, and chum salmon genes (36 -38). The 25-bp segment shown is highly conserved, with 1 nucleotide substitution, a G to C transversion, and 1 deletion in the human and chicken DNAs compared with rat, and 2 substitutions and 1 deletion in salmon. The human and chicken sequences are identical, whereas salmon HS3D differs by only a single nucleotide. Cross-competition gel-mobility shift studies were performed to determine the relative affinity of the human/chicken HS3D region for C/EBP␦ expressed in COS-7 cells. As seen in Fig. 10, both rat and human/chicken 32 Plabeled double-stranded probes gave rise to DNA-protein complexes of identical mobility, and both unlabeled HS3D oligonucleotides competed equivalently for binding to C/EBP␦ with either 32 P-labeled DNA sequence. Thus the human and chicken HS3D sequences behave as high-affinity C/EBP␦ binding sites.

The HS3D Site Is Structurally and Functionally Conserved in IGF-I Promoters from Different Species-
Functional analyses of the potential role of HS3D in mediating hormonal regulation of human IGF-I gene transcription were performed with a chimeric human IGF-I promoter 1-luciferase fusion plasmid. Transient transfection experiments using rat primary osteoblast cultures showed that a fragment of human promoter 1 from Ϫ1630 to ϩ322 with respect to the most 5Ј transcription start site (36) mediated a 6-fold increase in reporter gene activity after incubation of cells with 1 M PGE 2 for 6 h (Fig. 11A). Co-transfection of the same plasmid with an expression vector for C/EBP␦ led to a 5-fold rise in luciferase activity under basal conditions when compared with co-transfections with the empty expression plasmid and stimulated a further 3-fold increase in IGF-I gene activation after treatment with PGE 2 (Fig. 11A).
To establish a specific role for the human/chicken HS3D site FIG. 7. C/EBP␦ dimerizes in the absence of DNA. Bacterially expressed C/EBP␦-⌬NcoI (3 g) was incubated with or without 0.01% glutaraldehylde for 10 min at 25°C. Samples were separated by 8% polyacrylamide gel electrophoresis followed by Coomassie Blue staining. Similar results were observed in two additional experiments.
FIG. 8. C/EBP␦ binds to the HS3D site in rat IGF-I promoter 1 as assessed by DNaseI footprinting. In vitro DNaseI footprinting was performed as described under "Experimental Procedures" with end-labeled DNA derived from rat IGF-I promoter 1 and graded concentrations of bacterially generated C/EBP␦ (5, 15, 50, 100, 200 ng). Autoradiographic exposure was for 16 h at Ϫ80°C with intensifying screens. The location of the footprint on each DNA strand is indicated and was calibrated by including a DNA sequencing ladder in adjacent lanes of the 6% polyacrylamide gel. The numbers to the left of each panel correspond to coordinates of rat IGF-I exon 1 as described (15).
in mediating PGE 2 -activated and C/EBP␦-regulated transcription, a reporter gene containing 4 tandem copies of the 19-bp natural human/chicken HS3D region cloned 5Ј to a minimal RSV promoter was transfected into rat osteoblasts. As seen in Fig. 11B, treatment with 1 M PGE 2 stimulated a 4-fold increase in luciferase activity but had no significant effect on a reporter plasmid containing the minimal RSV promoter alone. Co-transfection with an expression plasmid for C/EBP␦ led to a 17-fold rise in luciferase activity under basal conditions as compared with co-transfections with the empty expression vector, and stimulated an additional 2-fold increase in promoter activity after incubation with PGE 2 (Fig. 11B). DISCUSSION The studies described in this report define the physicalchemical properties and functional consequences of interac-tions between HS3D, the DNA element in IGF-I promoter 1 that mediates stimulation of IGF-I gene transcription by cAMP or PGE 2 in osteoblasts (13,16), and C/EBP␦, the key transcription factor responsible for cAMP-activated IGF-I expression in these cells (17). We show that C/EBP␦, expressed in COS-7 cells or purified as a recombinant protein from E. coli, bound to HS3D with an affinity at least equivalent to that of the albumin D-site, a known high affinity C/EBP binding sequence (21,27), and that both DNA elements competed equally for C/EBP␦. C/EBP␦ bound to HS3D as a dimer, with protein-DNA contact points located on guanine residues on both DNA strands within and just adjacent to the core C/EBP half-site, GCAAT. C/EBP␦ also formed protein-protein dimers in the absence of interactions with its DNA binding site, as indicated by results of glutaraldehyde cross-linking experiments. In conjunction with functional studies, demonstrating cAMP-inducible transactivation by C/EBP␦ of human IGF-I promoter 1 and of a reporter gene with four tandem copies of the conserved human/chicken HS3D site, our results provide evidence that C/EBP␦ is a critical activator of IGF-I gene transcription in rodent osteoblasts and, potentially, in other cell types and species.
The chemical properties of C/EBP␦ assessed here resemble those of the related factors, C/EBP␣ and C/EBP␤. With the addition of our new studies, it is now clear that all three proteins can rapidly form dimers in dilute solution without a requirement for the presence of the specific DNA binding site (Refs. 21, 22, and 28; and this report). Dimerization is mediated by the COOH-terminal leucine zipper, which consists a heptad of leucine repeats within a relatively preserved 35 amino acid core (18). This region and the adjacent basic DNA binding domain are the most conserved portions of C/EBPs, having ϳ60% identity among C/EBP␣, ␤, and ␦ (18). Dimerization appears to be a prerequisite for DNA binding, since in previous studies with C/EBP␣, mutation of any of the leucine residues blocked recognition of a high affinity C/EBP element (28).
Prior results using the COOH-terminal portion of C/EBP␣ in dimethylsulfate footprinting experiments with an idealized dyad-symmetrical high affinity C/EBP site had identified four nucleotide contact points and several other sites of enhanced DNA cleavage (34). Our observations with full-length C/EBP␦ and HS3D DNA are very similar. We detected the same four protected nucleotides and mapped an additional protection to a more 5Ј guanine on the lower DNA strand (Fig. 9). The slight differences between results may be explained potentially by variation in the DNA binding sites, although the 8-bp central region is identical, or by the different proteins used, a COOHterminal fragment of C/EBP␣ previously (34) versus full-length C/EBP␦ here.
Our previous results defined HS3D as a functionally important component in the major IGF-I promoter from the rat (13,16). The current studies demonstrate that human IGF-I promoter 1 also is activated by PGE 2 and that a multimerized HS3D element from human or chicken IGF-I promoter functions as a PGE 2 -induced and C/EBP␦-regulated hormone response element. Based on these observations and on the similarity of HS3D sites in IGF-I genes from human, rat, chicken, and salmon (Table II), we tentatively predict that regulation of IGF-I transcription via C/EBP␦ also is conserved and postulate that C/EBP␦ may be a critical intermediate in the hormonal control of IGF-I synthesis in osteoblasts in several species. Detailed analysis of recently generated C/EBP␦-deficient mice (24) should provide additional insights into the role of this transcription factor in regulating production of IGF-I in bone and other tissues.
Other C/EBP isoforms also may play roles in regulating IGF-I gene expression. We had shown previously that C/EBP␤ FIG. 9. Identifying the nucleotide contact points for C/EBP␦ on HS3D DNA. Methylation interference assays were performed as described under "Experimental Procedures." Lanes contain control (C), bound (B), and free (F) DNA, as indicated. Autoradiographic exposure was for 6 h at Ϫ80°C with intensifying screens. Guanine residues important for binding of C/EBP␦ are in bold type and are marked by asterisks on the autoradiograph and on the DNA sequence.  (38) could bind to the HS3D site and, in co-transfection experiments, could transactivate a rat IGF-I promoter 1-luciferase reporter gene (17). Because both C/EBP␣ and C/EBP␤ are expressed in liver, fat, and other tissues (18, 20 -22) where IGF-I mRNA also is synthesized (2,15), it is reasonable to expect that these proteins also may modulate IGF-I gene transcription under different physiological conditions. We previously found that C/EBP␦ was activated and IGF-I transcription was stimulated in primary rat osteoblasts by a cyclic AMP-dependent protein kinase-dependent pathway that did not require ongoing protein synthesis (16). Preliminary experiments have shown that C/EBP␦ can be translocated from the cytoplasm to the nucleus of osteoblasts after PGE 2 treat-ment, even when protein synthesis is blocked. 2 Goals for the future will be to characterize the pathways responsible for induction of C/EBP␦ activity in these cells and to determine the pathways through which C/EBP␦ stimulates IGF-I gene transcription. FIG. 10. The HS3D site of the human IGF-I promoter binds C/EBP␦. Gelmobility shift experiments were performed with 32 P-labeled rat and human HS3D probes, nuclear extracts from COS-7 cells transiently transfected with a C/EBP␦ expression plasmid, and unlabeled double-stranded competitor DNAs at 50, 100, 150, and 200-fold molar excess (except for Oct-1, which was used at 200fold molar excess). Autoradiographic exposure was for 8 h at Ϫ80°C with an intensifying screen. Representative autoradiographs are shown in the top panel. The mean of two experiments has been plotted in the lower panels after quantitation by phosphoimager (squares and solid lines, rat HS3D competitor; triangles and dotted lines, human HS3D competitor).
FIG. 11. C/EBP␦ overexpression stimulates human IGF-I promoter activation in rat osteoblasts through the HS3D site. A, osteoblast-enriched cultures were transiently transfected with a promoterless control plasmid, pOLuc (no promoter), or the chimeric promoterreporter gene, hIGF1630-Luc (hIGF-I promoter) along with either no expression plasmid, the empty vector, or an expression plasmid for C/EBP␦. After treatment with control medium (containing ethanol vehicle) or 1 M PGE 2 for 6 h, cytoplasmic extracts were prepared, and luciferase activity was determined and normalized for transfection efficiency using a co-transfected ␤-galactosidase reporter plasmid. B, osteoblast-enriched cultures were transiently transfected with recombinant luciferase reporter genes containing the minimal RSV promoter or the RSV promoter plus 4 copies of a 19-nucleotide human HS3D oligonucleotide (RSV promoter ϩ 4X hHS3D). Each promoter-reporter gene was transiently co-transfected with either no expression plasmids, the empty vector, or an expression plasmid for C/EBP␦. After treatment with control medium (containing ethanol vehicle) or 1 M PGE 2 for 6 h, cytoplasmic extracts were prepared, and luciferase activity was measured and normalized for transfection efficiency using a co-transfected ␤-galactosidase reporter. Results are shown in panels A and B from 3 independent experiments where n ϭ 9. In panels A and B, asterisks indicate the following: *, significantly different from control cells (p Ͻ 0.05); **, significantly different from vector-transfected cells (p Ͻ 0.05).