Hormonal control of insulin-like growth factor I gene transcription in human osteoblasts: dual actions of cAMP-dependent protein kinase on CCAAT/enhancer-binding protein delta.

Insulin-like growth factor-I (IGF-I) is essential for somatic growth and promotes bone cell replication and differentiation. IGF-I production by rat osteoblasts is stimulated by activation of cAMP-dependent protein kinase (PKA). In this report, we define two interacting PKA-regulated pathways that control IGF-I gene transcription in cultured human osteoblasts. Stimulation of cAMP led to a 12-fold increase in IGF-I mRNA and enhanced IGF-I promoter activity through a DNA response element termed HS3D and the transcription factor CCAAT/enhancer-binding protein delta (C/EBPdelta). Under basal conditions, C/EBPdelta was found in osteoblast nuclei but was transcriptionally silent. Treatment with the PKA inhibitor H-89 caused redistribution of C/EBPdelta to the cytoplasm. After hormone treatment, the catalytic subunit of PKA accumulated in osteoblast nuclei. Inhibition of active PKA with targeted nuclear expression of PKA inhibitor had no effect on the subcellular location of C/EBPdelta but prevented hormone-induced IGF-I gene activation, while cytoplasmic PKA inhibitor additionally caused the removal of C/EBPdelta from the nucleus. These results show that IGF-I gene expression is controlled in human osteoblasts by two PKA-dependent pathways. Cytoplasmic PKA mediates nuclear localization of C/EBPdelta under basal conditions, and nuclear PKA stimulates its transcriptional activity upon hormone treatment. Both mechanisms are indirect, since PKA failed to phosphorylate human C/EBPdelta in vitro.

Insulin-like growth factor-I (IGF-I), 1 a 70-amino acid secreted protein, plays a fundamental role in growth and development in a variety of vertebrate species (1,2). IGF-I acts by binding to its cell surface receptor, a heterotetrameric ligandactivated tyrosine-protein kinase that is structurally and functionally related to the insulin receptor (3). In bone, IGF-I promotes longitudinal growth in vivo (4,5) and has been shown to enhance osteoblast replication and synthesis of type I collagen in tissue culture (6,7). In addition to the role of IGF-I in growth, its induced expression in bone cells in response to parathyroid hormone (PTH) may explain some of the anabolic effects of PTH within the skeleton (8 -11). IGF-I also may serve as a coupling factor during bone remodeling to balance resorption and new bone formation (9,10).
IGF-I is synthesized and secreted by cultured cells enriched in the osteoblast phenotype derived from rat fetal calvarial bones (12,13), but conflicting evidence exists regarding IGF-I expression by human osteoblasts (14). In rat osteoblasts in primary culture, IGF-I mRNA and protein are induced by PTH and prostaglandin E 2 (PGE 2 ), hormones that stimulate cAMP production (10,15). PTH and PGE 2 enhance IGF-I gene expression in these cells by a mechanism involving cAMP-dependent protein kinase (PKA) and PKA-mediated activation of the transcription factor CCAAT/enhancer binding protein ␦ (C/EBP␦), which binds to a hormone response element in the major IGF-I gene promoter termed HS3D (16 -19). Recently, we showed that PKA stimulates the nuclear translocation of C/EBP␦ in rat osteoblasts by an indirect pathway that does not involve phosphorylation of the transcription factor (20).
The catalytic subunit of PKA is able to modulate gene expression by phosphorylating a variety of cytoplasmic and nuclear targets including several transcription factors (21). In some cell types, the transcriptional actions of PKA require regulated nuclear translocation of the enzyme in response to hormonal stimulation (22,23). In the nucleus, in addition to directly phosphorylating CREB and other transcription factors (24,25), the catalytic subunit of PKA has been shown to potentiate gene transcription by phosphorylating the transcriptional co-activator CBP (26). C/EBP␦ belongs to a family of transcriptional regulators with diverse actions on tissue differentiation, intermediary metabolism, wound healing, and immune responses (27). Members of the C/EBP family are related structurally and contain three well defined protein motifs: a NH 2 -terminal transactivation region, a central basic DNA-binding domain, and a COOHterminal dimerization interface known as the leucine zipper segment (27). The latter two domains are found in a larger family of basic leucine zipper transcription factors with a broad range of biological effects (27,28). C/EBP␦ has been implicated previously in the control of fat cell differentiation and as a transcriptional mediator of the acute inflammatory response (27,29,30).
The current experiments were initiated to assess regulation of IGF-I gene expression in human osteoblasts. We find that IGF-I gene transcription is acutely activated in these cells by PKA-mediated mechanisms that involve two separable but interacting pathways that converge on C/EBP␦. Cytoplasmic PKA maintains the predominant nuclear localization of C/EBP␦ under basal conditions, and nuclear PKA stimulates its transcriptional activity upon hormone treatment. Both effects of PKA are indirect, since C/EBP␦ does not appear to be a substrate for PKA. Our results may have implications for therapeutic strategies in humans designed to enhance bone density using PTH (31) or other agents.

EXPERIMENTAL PROCEDURES
Materials-PGE 2 and forskolin were purchased from Sigma, and H-89 was obtained from Calbiochem. PGE 2 was reconstituted in ethanol at 1 mM, and the other drugs were dissolved in Me 2 SO to at least 1000 times the final concentration. Digoxigenin (DIG) RNA labeling mix, blocking reagent for nucleic acid detection, and fluorescein-coupled anti-DIG antibody were purchased from Roche Molecular Biochemicals. Herring sperm DNA was obtained from CLONTECH and sheared by placing in a sonicator bath for 10 min. Polyclonal antibodies to rat C/EBP␦ were raised in chickens and purified in our laboratory, as described (32). Rabbit polyclonal antibodies to catalytic subunit of PKA␣ (PKA␣C, C-20) and to Akt were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA) and Cell Signaling Technology (Beverly, MA), respectively. Cy3-conjugated rabbit anti-chicken IgY was obtained from Jackson ImmunoResearch Laboratories (West Grove, PA), and Alexa 594-conjugated goat anti-rabbit IgG was from Molecular Probes, Inc. (Eugene, OR). Alkaline phosphatase-conjugated goat anti-chicken IgY and alkaline phosphatase-conjugated goat antirabbit IgG were from Southern Biotechnology Associates (Birmingham, AL); 4Ј,6-diamidino-2-phenylindole and Hoechst nuclear stains were from Sigma; and VECTASHIELD Mounting Medium was from Vector Laboratories (Burlingame, CA). The catalytic subunit of PKA and the activation domain of cAMP response element-binding protein (CRE Bad) were gifts from Dr. James R. Lundblad (Oregon Health Sciences University, Portland, OR). All other reagents were purchased from commercial suppliers.
Plasmids-The mutant regulatory subunit of mouse PKA (clone MtR(AB)) was a gift from Dr. G. Stanley McKnight (University of Washington, Seattle, WA). Plasmid pcDNA3/Neo was purchased from Invitrogen (Carlsbad, CA), and the expression vector pcDNA3-rat C/EBP␦ was constructed in our laboratory (19). Human and rat IGF-I promoter 1-luciferase fusion genes were previously described (33,34). Luciferase reporter genes containing the minimal RSV promoter with or without four copies of human HS3D have been described previously (32). The nucleotide sequence of human HS3D is as follows: 5Ј-GAGC-CTGCGCAATGGAATAAAGTC-3Ј (32). The region that binds C/EBP␦ is underlined.
His-tagged human C/EBP␦ was cloned from genomic DNA by nested PCR combined with PCR-mediated mutagenesis. DNA was amplified by PCR using oligonucleotides complementary to the gene sequence reported in GenBank TM : 5Ј-GACAGCCTCGCTTGGACGCAGAGCC-3Ј (top strand) and 5Ј-GGGTCGTTGCTGAGTCTCTCCCGCC-3Ј (bottom strand). The initial PCR product was reamplified using a top strand primer containing XbaI and NdeI sites followed by a His epitope tag (underlined), which preceded the ATG codon (boldface type): 5Ј-AATG-CTCTAGACATATGGCACACCACCACCACCACCACATGAGCGCCG-CGCTCTTCAGCCT-3Ј. The bottom strand primer contained an EcoRI site adjacent to the human C/EBP␦ stop codon (underlined): 5Ј-TCGG-GAATTCCGCGTTACCGGCAGTCTGCTGTC-3Ј. The amplified DNA was purified, digested with XbaI and EcoRI, and inserted into the corresponding sites of pBluescript. After the entire coding region was verified by DNA sequencing on both strands, it was excised by digestion with NdeI and EcoRI and inserted into the corresponding sites of pET-29a(ϩ) (Novagen, Madison, WI) to produce pET29a-His-human C/EBP␦. His-tagged rat C/EBP␦ was generated by PCR-mediated mutagenesis. The His epitope tag (codons underlined) was added to the 5Ј-end of cDNA in pBluescript-rat C/EBP␦ (18) between the NdeI site and the ATG codon (boldface type) using the top strand primer: 5Ј-AA-TGCCATATGGCACACCACCACCACCACCACATGAGCGCCGCGCT-CTTCAGCCT-3Ј. The bottom strand primer was complementary to the internal region of rat C/EBP␦, which contained a NcoI site: 5Ј-CTACA-TTGATTCCATGGCTGCCG-3Ј. The amplified DNA was digested with NdeI and NcoI to obtain the 5Ј-end of His epitope-tagged rat C/EBP␦. The 3Ј portion was excised from the original pBluescript-rat C/EBP␦ plasmid with NcoI and EcoRI, and both fragments were cloned into NdeI-and EcoRI-digested pET29a(ϩ). The intended changes were verified by sequencing.
The EGFP-tagged wild type protein kinase inhibitor peptide (EGFP-PKIwt) was made by sequential subcloning of PCR-amplified fragments corresponding to EGFP and PKI. The coding sequence for EGFP was amplified from pEGFP (CLONTECH) using a top strand primer containing a BamHI site (underlined), which preceded the ATG codon (boldface type): 5Ј-AATAATGGATCCACCATGGTGAGCAAGGGCGAGGA-GCTGTTCACC-3Ј. The bottom strand primer contained an EcoRI site (underlined) in frame with the COOH-terminal amino acid of pEGFP (amino acid 265, boldface type): 5Ј-AATAATGAATTCTCTGGATCCG-GTGGATCCCGGGCCCGCGGTACC-3Ј. The amplified DNA was purified, digested with BamHI and EcoRI, and inserted into the corresponding sites of pcDNA3/Neo, generating pcDNA-EGFP. The coding sequence of PKI␣ (amino acids 2-77; gift of Dr. Richard A. Maurer, Oregon Health Sciences University, Portland, OR) was amplified using a top strand primer containing an EcoRI site (underlined) in frame with amino acid 2 of the PKI sequence: 5Ј-AATAATGAATTCACTGATGTG-GAAACTACGTATGCAG-3Ј. The bottom strand primer contained a XbaI site (underlined) adjacent to the PKI␣ stop codon (boldface type): 5Ј-AATAATTCTAGACTATTAGCTTTCAGACTTGGCTGC-3Ј. The amplified product was purified, digested with EcoRI and XbaI, and inserted into the corresponding sites of pcDNA-EGFP, and the coding region was confirmed by sequencing.
The EGFP-PKInls was generated by PCR-mediated mutagenesis. The fragment of PKI coding sequence lacking the PKI nuclear export signal (amino acids 2-37) was amplified using the top strand primer described above for the EGFP-wtPKI and the bottom strand primer containing an XhoI site in frame with amino acid 37 of PKI␣: 5Ј-AATAATCTCGAGTTGATTGCTGTTGCCACTTGC-3Ј. The amplified product was purified, digested with EcoRI and XhoI, and inserted into the corresponding sites of pcDNA-EGFP to generate EGFP-PKI-(2-37). A nuclear localization sequence (nls) was introduced at the COOH terminus as follows. Two oligonucleotides were generated, encoding sense and antisense sequences containing two tandem copies of the nls from SV40 large T antigen followed by a stop codon (boldface type) and flanked by a 5Ј XhoI and a 3Ј XbaI overhang (underlined): 5Ј-TCGAG-CCTAAGAAGAAGAGGAAGGTCGGTACCGGAGGAGGTGAAGCAC-CCAAGAAGAAGCGAAAAGTAGGATCCACCGGATAAT-3Ј (sense) and 5Ј-CTAGATTATCCGGTGGATCCTACTTTTCGCTTCTTCTTGG-GTGCTTCACCTCCTCCGGTACCGACCTTCCTCTTCTTCTTAGGC-3Ј (antisense). The oligonucleotides were annealed, phosphorylated by T4 kinase, and ligated into the XhoI and XbaI sites of EGFP-PKI-(2-37). The coding region was confirmed by sequencing.
Cell Culture and DNA-mediated Gene Transfer-Primary normal human osteoblasts cryopreserved in the third passage (lot 0F0630) were purchased from BioWhittaker Inc. (Walkersville, MD). Cells were used for experiments within one or two additional passages in our laboratory (no more than 10 population doublings). The osteoblastic phenotype was verified by the manufacturer by staining cells for alkaline phosphatase expression and by demonstrating mineralization in culture. Cells were propagated at 37°C in humidified air containing 5% CO 2 in osteoblast growth medium (osteoblast basal medium supplemented with 10% fetal bovine serum, 20 -30 mg/liter ascorbic acid, 50 mg/liter gentimicin, and 50 g/liter amphotericin B, all from BioWhittaker Inc.). Cells were transfected at 50 -70% confluent density with Fugene (Roche Molecular Biochemicals) or Effectene (Qiagen Inc., Valencia, CA) using protocols developed by the manufacturers.
In Situ Hybridization-In situ hybridization was performed using a DIG-labeled human IGF-I riboprobe. An 817-base pair human IGF-IA cDNA was excised from pGEM1 (35) with RsaI and EcoRI and subcloned into the HincII and EcoRI sites of pBluescript KS to generate pBS-IGF-IA. The in situ hybridization probe was generated from the KpnI-linearized plasmid by in vitro transcription in the presence of DIG RNA labeling mix using T7 RNA polymerase, following a protocol modified from Raap et al. (36). Confluent human osteoblasts grown in two-well tissue culture slide chambers were incubated in serum-free medium for 18 h followed by treatment with vehicle or forskolin in serum-free medium for 6 h. Cells were rinsed in PBS, fixed in 4% formaldehyde, and dehydrated by incubating successively in 70, 90, and 100% ethanol. After removal of lipids by incubation in xylenes for 10 min, cultures were rehydrated in decreasing concentrations of ethanol followed by PBS, permeabilized with 0.1% (w/v) pepsin in 0.1 M NaCl at 37°C for 20 min, rinsed in PBS, postfixed in 1% formaldehyde for 10 min, and washed again in PBS. The probe was diluted to 5 ng/l in hybridization buffer containing 60% deionized formamide, 300 mM NaCl, 30 mM sodium citrate, 10 mM EDTA, 25 mM sodium phosphate (pH 7.4), 5% dextran sulfate, and 250 g/ml sheared herring sperm DNA (denatured shortly before use). Probe solution (15 l) was added to the fixed permeabilized cells, covered with a 22-cm 2 coverslip, and incubated at 55°C for 16 h. Following hybridization, coverslips were removed in 2ϫ SSC buffer (300 mM NaCl, 30 mM sodium citrate), and three high stringency washes were performed in 50% formamide in 2ϫ SSC buffer at 55°C for 30, 20, and 10 min followed by a wash in PBS. The slides were incubated for 45 min at 25°C in blocking buffer (100 mM Tris-HCl (pH 7.5), 150 mM NaCl, and 0.5% (w/v) blocking reagent) and for 45 min in 1:5 dilution of anti-DIG-fluorescein antibody in blocking buffer. The slides were washed in PBS, dehydrated as before, dried, mounted in VECTASHIELD medium containing 3.8 g/ml 4Ј,6diamidino-2-phenylindole nuclear stain, and viewed by fluorescence microscopy (Nikon Eclipse TE 300). Images were captured with a Photometrics CoolSNAP fx camera (Roper Scientific, Tucson, AZ) and Apple Macintosh G4 computer using IPLab software, version 3.5 (Scanalytics Inc., Fairfax, VA). Image processing was performed using Photoshop 5.5 (Adobe Systems, San Jose, CA).
Immunocytochemistry-Confluent human osteoblast cultures were preincubated in serum-free medium for 20 h, followed by the addition of drugs or vehicle (ethanol or Me 2 SO or both diluted 1:1000) in serumfree medium for 2-6 h. H-89 was added to cells 15 min prior to the addition of forskolin or PGE 2 . After drug treatment, cells were rinsed twice with PBS, fixed in 4% paraformaldehyde, permeabilized with a 1:1 mixture of acetone and methanol, and blocked with 10% bovine serum albumin in PBS. After two washes with PBS, cells were incubated with primary antibodies, either polyclonal chicken anti-C/EBP␦ (1:500 dilution) or polyclonal rabbit anti-PKA␣C (1:100) in PBS plus 3% bovine serum albumin for 2 h at 25°C. Cells were then washed three times in PBS and incubated with 100 ng/ml Hoechst dye and the appropriate labeled secondary antibodies, either Cy3-conjugated rabbit anti-chicken IgY (1:400) or Alexa 594-conjugated goat anti-rabbit IgG (1:1000), for 2 h in the dark. Finally, cells were washed in PBS and were examined and imaged as described for in situ hybridization.
Reporter Gene Analysis-Primary human osteoblasts were seeded in six-well plates at 96,000 cells/well and transfected the next day with 2 g of firefly luciferase reporter genes or 1 g each of reporter gene and expression plasmid (pcDNA3, pcDNA3-C/EBP␦, dominant negative PKA, EGFP, EGFP-PKIwt, or EGFP-PKInls). Transfections with promoter-reporter genes also included 1 ng of a vector containing the Renilla luciferase gene under control of the cytomegalovirus immediate early enhancer/promoter (pRL-CMV, Promega Corp., Madison, WI). Renilla luciferase activity was used to normalize for transfection efficiency. At 24 -48 h after transfection, cells were washed with PBS, incubated in serum-free medium for 18 h, and treated with vehicle or drugs in serum-free medium for an additional 6 h. Luciferase activity was measured using the Dual-Luciferase Reporter Assay (Promega). Cultures were washed in PBS, scraped into 200 l/well of passive lysis buffer and cleared by brief centrifugation. The entire supernatant was then assayed using an AutoLumat LB 953 luminometer (Berthold Systems, Inc., Aliquippa, PA). Light emission was measured by integration over 10 s of the enzymatic reactions. Firefly luciferase values were divided by Renilla luciferase to yield the relative enzymatic activity for each experimental point. All experiments were performed in duplicate and repeated at least three times except for dominant negative PKA transfections, which were performed twice.
Protein Extraction and Immunoblotting-Confluent osteoblast cultures were deprived of serum for 20 h and then treated with PGE 2 for 4 h. Cytoplasmic and nuclear protein extracts were prepared as de- The most 5Ј transcription start site has been designated ϩ1 and is indicated by a bent arrow (61). The translation initiation codon is marked by a vertical arrow. Two promoter-reporter plasmids are diagrammed below the gene map. The wild type fusion gene (hIGF-I) consists of DNA from Ϫ1630 to ϩ322 joined to firefly luciferase cDNA; hIGF-I⌬ lacks sequences from ϩ112 to ϩ197. Both plasmids have been described previously (33). B and C, expression of hIGF-I and hIGF-I⌬ luciferase fusion genes after transfection into primary human osteoblasts. Cells were transfected and treated, and luciferase activity was measured as described under "Experimental Procedures." Where indicated, cells were co-transfected with pcDNA3/Neo (vector) or dominant-interfering subunit of PKA in the same expression plasmid (dnPKA). Luciferase values obtained after transfection of hIGF-I luciferase fusion gene in the absence of treatment have been arbitrarily given a value of 1. Forskolin (F) and H-89 were each used at a final concentration of 10 M. The single asterisk indicates a significant decrease in reporter gene activity compared with untreated conditions (p Յ 0.02 for all); the double asterisk denotes a significant increase in luciferase expression compared with all other conditions (p Յ 0.04). The results in B and C are from three independent experiments, with each performed in duplicate.
Preparation of Recombinant C/EBP␦ Proteins-Human and rat Histagged C/EBP␦ were purified from bacteria as follows. Plasmids pET29a-His-human C/EBP␦ and pET29a-His-rat C/EBP␦ were transformed into the BL21(DE3) strain of Escherichia coli. Cultures were grown to an A 600 of 0.7 in 1 liter of Circlegrow (Bio 101, Vista, CA) containing 50 g/ml kanamycin and then were induced to express recombinant proteins by the addition of isopropyl-1-thio-␤-D-galactopyranoside (Sigma; final concentration of 1 mM) and incubation for 4 h. Bacterial pellets were harvested by centrifugation, washed in PBS, and resuspended in 30 ml of binding buffer (50 mM Tris, pH 7.5, 300 mM NaCl, 5% glycerol, 0.4 mM phenylmethylsulfonyl fluoride, 2 mM ␤-mercaptoethanol, 50 mM imidazole). Cells were lysed by passaging twice through a French Pressure Cell Press (Spectronic Instruments, Rochester, NY) at 16,000 p.s.i., and bacterial debris was removed by centrif-ugation. The His-tagged proteins were purified by FPLC (Amersham Pharmacia Biotech). Proteins were bound to nickel-nitrilotriacetic acidagarose columns (Qiagen) in binding buffer, washed with 30 column volumes of binding buffer, and eluted with a continuous imidazole gradient from 50 to 500 mM into 20 fractions. The protein-containing fractions were pooled, dialyzed overnight against 5 liters of dialysis buffer (25 mM Tris, pH 8.5, 1 mM EDTA, 5% glycerol, 0.6 mM phenylmethylsulfonyl fluoride, 1 mM dithiothreitol), and passed through the HiTrap Q column (Amersham Pharmacia Biotech) in dialysis buffer. After the column was washed with 5 volumes of dialysis buffer, proteins were eluted in a continuous salt gradient up to 1 M NaCl. Ten fractions were collected, analyzed for purity, aliquoted, and stored at Ϫ80°C until use. Proteins were diluted to reduce the salt concentration to 200 mM prior to use.
PKA Assay-Purified, recombinant catalytic subunit of PKA was diluted to 10 g/ml in 100 g/ml bovine serum albumin. Recombinant CREBad, His-human C/EBP␦, or His-rat C/EBP␦ proteins (500 ng each) were mixed on ice with 0.5 Ci of [␥-32 P]ATP in assay buffer (100 M ATP, 10 mM MgCl 2 , 250 g/ml bovine serum albumin, 12.5 mM Tris-Cl, pH 7.5). The recombinant catalytic subunit of PKA (10 ng) was added, and reactions were allowed to proceed for 2 min at 30°C. The reactions were stopped after being placed on ice by the addition of EDTA to 80 mM final concentration. After boiling for 5 min in SDS sample buffer, the samples were separated by SDS-polyacrylamide gel electrophoresis. Gels were stained with SYPRO Orange (Molecular Probes), dried, and exposed to Phospho Screens (Eastman Kodak Co.) for 30 min followed by detection using Molecular Imager FX imaging system and Quantity One software.
Statistical Analysis-Data are presented as the means Ϯ S.E. Statistical significance was determined using Student's t test for paired samples. Results were considered statistically different when p Ͻ 0.05.

FIG. 3. Activation of PKA potentiates C/EBP␦-stimulated gene expression through the HS3D site in human osteoblasts.
Expression of luciferase genes is shown under control of a minimal RSV promoter (no enhancer) or four copies of human HS3D fused to the minimal RSV promoter (4xhHS3D) after transfection into primary human osteoblasts. Cells were transfected and treated, and luciferase activity was measured as described under "Experimental Procedures." Luciferase values measured after transfection of a reporter gene containing the minimal RSV promoter in the absence of drug treatment have been arbitrarily given a value of 1. Forskolin (F) and H-89 were each used at a final concentration of 10 M. A, the asterisk indicates a significant increase in luciferase activity after forskolin treatment (p ϭ 0.015). B, cells were co-transfected with 4xhHS3D and either pcDNA3/Neo (vector) or a rat C/EBP␦ cDNA in the same plasmid. The single asterisk indicates a significant increase in luciferase expression after forskolin or C/EBP␦ (p Յ 0.04). The double asterisk denotes a significant increase in reporter gene activity after incubation of cells with forskolin compared with all other treatments (p Յ 0.01). The results in A and B are from three independent experiments, with each performed in duplicate.

FIG. 4. Expression of C/EBP␦ in human osteoblasts. A,
Western immunoblots for C/EBP␦ and Akt using nuclear (16 g) and cytoplasmic (40 g) protein extracts from primary human osteoblasts treated with vehicle (C) or 1 M PGE 2 (P) for 4 h. B, immunocytochemistry for C/EBP␦ in primary human osteoblasts after incubation with vehicle (cont), 1 M PGE 2 , or 10 M H89 for 4 h (left panels). The right panels show nuclei stained with Hoechst dye.

Hormonal Activation of IGF-I Gene Transcription in Human
Osteoblasts Involves PKA and C/EBP␦-In previous studies, we and others have shown that IGF-I gene expression was regulated in primary cultures of rat osteoblasts by hormones such as PTH and PGE 2 that enhanced production of cAMP (10,15). These hormones rapidly induced IGF-I gene expression through a PKA-dependent mechanism that involved a DNA response element in the 5Ј-untranslated region of rat IGF-I exon 1 termed HS3D (17,18) and the transcription factor C/EBP␦ (18,19,32). We found that PKA activated C/EBP␦ by inducing its nuclear translocation (20), leading to binding at the HS3D site and transcriptional stimulation of the major IGF-I promoter (19,32,37). We now have evaluated hormonal mechanisms controlling IGF-I gene expression in low passage cultured human osteoblasts. Fig. 1 shows results of in situ hybridization experiments. As seen in the figure, IGF-I mRNA was minimally detectable in human osteoblasts under control conditions (Ͻ2% of cells weakly positive) but was readily seen after treatment of cells for 6 h with forskolin (ϳ24% of cells positive). Thus, in human osteoblasts, IGF-I gene expression is acutely induced through a cAMP-stimulated pathway.
We next performed a series of transient transfection experiments using IGF-I promoter-luciferase reporter plasmids (diagrammed in Fig. 2A) to determine if IGF-I gene transcription was under hormonal control in human osteoblasts. As illustrated in Fig. 2B, activity of the wild type human IGF-I pro-moter 1-reporter gene (hIGF-I promoter) was stimulated 3-fold in cells treated with forskolin for 6 h. This increase was eliminated, and promoter activity was reduced to Ͻ50% of basal levels by the PKA inhibitor H-89 or by co-transfection with a dominant-interfering regulatory subunit of PKA (Fig. 2B). By contrast, a human IGF-I promoter 1-reporter plasmid lacking the site homologous to rat HS3D (hIGF-I⌬ promoter) was Ͻ50% as active as the wild type hIGF-I promoter under basal conditions, and its activity was not changed by forskolin, H-89, or the dominant negative PKA subunit (Fig. 2C, and data not shown). Enzymatic activity of a luciferase reporter gene containing four tandem copies of the 19-base pair core human HS3D site cloned 5Ј to a minimal Rous sarcoma virus promoter (4xhHS3D) was stimulated 7-fold after treatment of osteoblasts with forskolin for 6 h; activity of a plasmid containing the minimal promoter alone fused to luciferase (no enhancer) was not increased (Fig. 3A). Incubation of osteoblasts with H-89 eliminated the response of the 4xhHS3D plasmid to forskolin, indicating that the transgene was regulated by PKA. Co-transfection with a rat C/EBP␦ expression plasmid induced the 4xhHS3D-reporter gene by 6-fold under basal conditions, and treatment with forskolin for 6 h resulted in a further 2.6-fold increase in luciferase activity. Incubation of osteoblasts with H-89 reduced reporter gene expression to below basal levels in the absence or presence of forskolin (Fig. 3B). Based on these results, we conclude that a hormonally activated pathway in- volving PKA and potentially inducing C/EBP␦ regulates IGF-I gene transcription through the HS3D site in human osteoblasts.
PKA Activity Is Required for Nuclear Localization of C/EBP␦-To assess mechanisms of C/EBP␦ activation by PKA, we first determined if C/EBP␦ was expressed in human osteoblasts. As illustrated in Fig. 4A, C/EBP␦ was detected by Western immunoblotting in nuclear protein extracts but not in cytoplasmic extracts of human osteoblasts under basal conditions and after incubation of cells for 4 h with PGE 2 , a hormone that stimulates cAMP production and IGF-I gene transcription in rat osteoblasts (15). Levels of C/EBP␦ protein appeared to be unchanged by PGE 2 treatment. Adequate separation of nuclear and cytoplasmic proteins was verified by detection of the enzyme Akt solely in the cytoplasmic fractions. The finding that C/EBP␦ is concentrated in cell nuclei before and after hormone treatment was confirmed by immunocytochemistry (Fig. 4B), although it appears that a small amount of the transcription factor is cytoplasmic. However, inhibition of PKA activity by H-89 resulted in the appearance of C/EBP␦ in the cytoplasm under both control conditions and after treatment of cells with PGE 2 (Fig. 4B and data not shown). A similar loss of nuclear C/EBP␦ was observed in human osteoblasts transfected with an expression plasmid encoding a dominant-interfering PKA regulatory subunit (data not shown). These results demonstrate that active PKA is required for nuclear localization of C/EBP␦ in human osteoblasts and that in the absence of hormonal stimulus there is sufficient basal PKA activity to maintain nuclear expression of C/EBP␦. These observations contrast with rat osteoblasts, where under basal conditions C/EBP␦ is cytoplasmic, and hormone treatment stimulates its nuclear translocation (20).
Both Cytoplasmic and Nuclear PKA Are Required for Activation of C/EBP␦ in Human Osteoblasts-Despite its nuclear localization in human osteoblasts, C/EBP␦ does not appear to be transcriptionally competent prior to hormonal stimulation of PKA in these cells, as evidenced by minimal expression of IGF-I mRNA under basal conditions (see Fig. 1). This suggests that one PKA-mediated signal is needed for nuclear maintenance of C/EBP␦ and that another is required for its activation. To identify this second signal, we first looked at the subcellular distribution of the catalytic subunit of PKA before and after hormone treatment (Fig. 5). As seen by Western immunoblotting in Fig. 5A, in human osteoblasts, PKA was predominantly cytoplasmic under basal conditions but became primarily nuclear after treatment of cells with PGE 2 for 4 h. Similar results were observed by immunocytochemistry. The catalytic subunit of PKA was cytoplasmic in the absence of hormone but accumulated in the nucleus during incubation with PGE 2 for 4 h (Fig. 5B), although its nuclear expression was never as complete as that of C/EBP␦. In contrast, in rat osteoblasts PKA remained predominantly cytoplasmic after hormone treatment ( Fig. 5A and data not shown). Thus, PGE 2 induces nuclear translocation of the catalytic subunit of PKA in human but not in rat osteoblasts.
To address the potential roles of nuclear and cytoplasmic PKA in regulating C/EBP␦ activity in human osteoblasts, we sought to differentially inhibit the enzyme in each subcellular compartment by targeted expression of the protein kinase inhibitor peptide (PKI). Cells were transfected and treated with PGE 2 for 4 h prior to fixation and processing for immunocytochemistry. As shown in Fig. 6A, a fusion protein consisting of EGFP followed by native PKI (EGFP-PKIwt) was expressed exclusively in the cytoplasm of human osteoblasts, while a chimeric PKI protein containing two copies of the SV40 nls at its COOH terminus (EGFP-PKInls) was found only in the nucleus. EGFP was seen in both nucleus and cytoplasm. Results in Fig. 6A demonstrate that the EGFP-PKIwt protein caused a change in the subcellular localization of C/EBP␦ from the nucleus to a more diffuse distribution in both cytoplasmic and nuclear compartments in the majority of transfected cells (41 of 50 cells counted, graphed in Fig. 6B). By contrast, the EGFP-PKInls fusion protein or EGFP alone did not change the predominantly nuclear location of C/EBP␦ (Fig. 6B). These results and those in Fig. 4B indicate that cytoplasmic PKA is responsible for maintaining nuclear localization of C/EBP␦. We next evaluated effects of differential inhibition of PKA activity on IGF-I gene transcription in human osteoblasts. As shown in Fig. 7A, co-transfection with either PKI plasmid prevented the hormone-stimulated increase in hIGF-I promoter activity and reduced luciferase values to basal levels or below. In conjunction with the observations of Fig. 6, we interpret these results as demonstrating that in human osteoblasts cytoplasmic PKA is required for nuclear localization of C/EBP␦ and that nuclear PKA is needed to stimulate its transcriptional activity. In contrast, in rat osteoblasts, where PKA did not accumulate in the nucleus after hormone treatment (Fig. 5A), FIG. 7. Differential regulation of IGF-I promoter by cytoplasmic versus nuclear expression of PKI in human and rat osteoblasts. Shown is expression of an hIGF-I promoter-luciferase reporter gene in human osteoblasts (Ob) (A) and of rat IGF-I (rIGF-I) promoterluciferase reporter gene in rat osteoblasts (B). Cells were co-transfected with luciferase reporter genes and EGFP (vector), EGFP-PKIwt (PKIwt), or EGFP-PKInls (PKInls). Cells were treated and luciferase activity was measured as described under "Experimental Procedures." PGE 2 was used at a final concentration of 1 M. For each cell type, luciferase values obtained after co-transfection with EGFP have been arbitrarily given a value of 1. The single asterisk indicates a significant decrease in reporter gene activity compared with untreated conditions (p Յ 0.04 for all); the double asterisk denotes a significant increase in luciferase expression compared with untreated conditions (p Յ 0.03). The results in each graph are from three independent experiments, with each performed in duplicate.
EGFP-PKIwt inhibited PGE 2 -induced transcription of a rat IGF-I promoter, but EGFP-PKInls was ineffective (Fig. 7B). Thus, in rat osteoblasts, it appears that cytoplasmic PKA regulates both the nuclear translocation and transcriptional activity of C/EBP␦.
Human C/EBP␦ Is Not a Substrate for PKA in Vitro-In previous studies, we and others have demonstrated that rat C/EBP␦ was not a direct substrate for PKA (20,38). However, inspection of the amino acid sequence of human C/EBP␦ revealed several differences from the rat protein, including species-specific changes that created putative PKA phosphorylation sites at residues 165-167 (Arg-Ser-Ser) and 178 -181 (Arg-Glu-Lys-Ser). To test the hypothesis that human C/EBP␦ was a substrate for PKA, a series of in vitro kinase assays was performed with purified, recombinant catalytic subunit of PKA and recombinant His epitope-tagged human and rat C/EBP␦. As shown in Fig. 8, neither human nor rat C/EBP␦ was phosphorylated by PKA under the experimental conditions described under "Experimental Procedures." In contrast, a recombinant protein containing the activation domain of the cAMP-regulated transcription factor, CREB, was readily modified. These results demonstrate the specificity of the in vitro kinase assay and show that neither human nor rat C/EBP␦ appear to be high affinity substrates for PKA. DISCUSSION Hormonal factors and mechanical forces influence bone remodeling by coordinately regulating signal transduction pathways in osteoblasts and osteoclasts (39,40). In osteoblasts, both systemic PTH and locally produced PGE 2 activate G proteincoupled receptors that, among other effects, enhance cAMP production, leading to increased IGF-I gene and protein expression through PKA-mediated mechanisms (10,15,17). Locally produced IGF-I in turn promotes osteoblast replication and differentiation (7,41). We now find in human osteoblasts in primary culture that PKA stimulates IGF-I gene transcription through two separable but interacting pathways involving the transcription factor C/EBP␦. Cytoplasmic PKA provides signals for nuclear localization of C/EBP␦ under basal conditions, and nuclear PKA promotes its transcriptional activity upon hormone treatment. Both effects of PKA are indirect, since C/EBP␦ does not appear to be a substrate for PKA.
Abundant published evidence supports the idea that IGF-I is produced by osteoblast-enriched cultures from fetal rat calvarial bones (9,10,15,42,43), but contradictory results have been reported for human osteoblasts. IGF-I has been detected in conditioned culture media from primary human osteoblasts by some investigators (44) but not by others (45,46). Equivalently conflicting results have been reported for IGF-I mRNA expression (47)(48)(49)(50). However, IGF-I transcripts have been detected previously by in situ hybridization in cells within human adult bone (51), a finding corroborated here in cultured osteoblasts after treatment of cells with forskolin. Our observation of increased IGF-I mRNA abundance in human bone cells after elevation of cAMP supports an earlier report by Okazaki et al. (47) and suggests that IGF-I gene expression in bone is influenced by the local hormonal milieu.
In the major rat IGF-I gene promoter, the HS3D element is essential for binding of C/EBP␦ and for hormone-regulated gene activation (19). The DNA sequence is conserved between rat and human IGF-I genes (17 of 19 identical residues in the DNA multimerized in the 4xhHS3D reporter gene (32)). Previous competition gel mobility shift experiments have determined that C/EBP␦ binds to human HS3D with an affinity only slightly less than measured for the rat DNA element (32), and as confirmed here, C/EBP␦ is able to transactivate promoters containing human HS3D sequences. Since a similar level of DNA sequence conservation is observed for HS3D sites in chicken and salmon IGF-I genes (32), it is conceivable that this region mediates hormonal regulation of IGF-I gene expression in bone through C/EBP␦-like proteins in multiple vertebrate species.
In past studies, we identified C/EBP␦ as a downstream target of PKA action in the cytoplasm of primary cultures of rat calvarial osteoblasts (19,32). Activation of PKA induced the rapid translocation of C/EBP␦ from the cytoplasm to the nucleus, although the transcription factor was not phosphorylated by PKA (20). We now find in human osteoblasts that C/EBP␦ plays a similar role in coupling PKA to IGF-I gene activation. However, in human cells C/EBP␦ is primarily nuclear at all times, yet it is only capable of inducing IGF-I gene transcription after hormonal stimulation. In addition, inhibition of basal PKA activity in the cytoplasm resulted in redistribution of C/EBP␦ out of the nucleus. The mechanisms by which cytoplasmic PKA normally maintains nuclear expression of C/EBP␦ in human osteoblasts are unknown, and we have been unable to ascertain if C/EBP␦ resides primarily in the nucleus of these cells under basal conditions or if it shuttles between subcellular compartments. Addressing this question awaits development of specific inhibitors of nuclear import.
Since nuclear localization of C/EBP␦ in human osteoblasts did not lead to induction of IGF-I gene expression until cells were treated with forskolin or PGE 2 , it was apparent that additional PKA-dependent steps were required to activate the transcriptional potential of nuclear C/EBP␦. We identified one critical step as hormone-induced nuclear translocation of the catalytic subunit of PKA. Nuclear translocation of PKA in response to cAMP has been shown previously in several other cell types, including bovine epithelial, rat thyroid, and neuroblastoma cells (22,52,53). The catalytic subunit of PKA does not possess a classical nuclear localization signal, and little is known about mechanisms of its entry into the nucleus. Microinjection experiments have suggested that passive diffusion of the enzyme can occur and that its relatively small size of ϳ40-kDa should allow the protein to pass through the nuclear pore without any need of assistance from nuclear import receptors (54). Nuclear accumulation of the catalytic subunit appears to be reversible (52,53), and in several studies overexpressed PKI promoted its export to the cytoplasm (55,56). Interestingly, we detected little nuclear accumulation of PKA in rat osteoblasts, perhaps because of species-specific or agespecific differences in expression of different subtypes of the enzyme. In rat calvarial osteoblasts, cytoplasmic PKA appeared to be sufficient to stimulate nuclear translocation of C/EBP␦ and to activate its transcriptional capabilities (Ref. 20 and data presented here).
The most intriguing finding in this report was the demonstration that nuclear PKA was required for stimulation of IGF-I gene transcription in human osteoblasts. Inhibition of nuclear PKA by targeted expression of PKI completely abolished PGE 2 -induced IGF-I promoter activity yet had no effect on nuclear localization of C/EBP␦. In contrast, nuclear PKI did not block hormone-stimulated IGF-I gene transcription in rat osteoblasts, consistent with the observation that the catalytic subunit of PKA did not translocate to the nucleus of these cells. In both cell types, forced expression of PKI in the cytoplasm prevented hormonal activation of IGF-I transcription and caused loss of C/EBP␦ from the nucleus or prevented its nuclear translocation. The mechanisms of activation of C/EBP␦ by nuclear or cytoplasmic PKA are unknown, but do not appear to involve direct phosphorylation, in marked contrast to the "classical" model of phosphorylation by nuclear PKA of the cAMPregulated transcription factor CREB (57). Other transcription factors, specifically TTF-1 and Pit-1, also are indirectly regulated by PKA. These latter proteins appear to require PKA in the nucleus to stimulate transcription of target genes, but mutating all possible PKA phosphorylation sites within these transcription factors did not affect their regulation (58,59). For Pit-1, it has been suggested that hormonal stimulation involves PKA-mediated recruitment of the transcriptional co-activator CBP through its direct phosphorylation (26), since mutation of a PKA site in CBP completely abrogated hormone-induced transcription of a reporter gene containing Pit-1 response elements (26). There is little information on the co-activators that potentially interact with C/EBP␦ to promote IGF-I gene transcription, but p300 has been shown to bind to the related protein, C/EBP␤, and to potentiate its activity (60). Fig. 9 presents a working model for control of IGF-I gene expression in human osteoblasts by hormones that activate cAMP and PKA. In the absence of hormonal stimulation, basal PKA activity in the cytoplasm is sufficient to promote and maintain expression of C/EBP␦ in the nucleus but is not able to activate C/EBP␦-mediated gene transcription. Under basal conditions, it is not known if C/EBP␦ in the nucleus is a dimer or if it is bound to its target sites on DNA (we have shown that C/EBP␦ is able to dimerize spontaneously in vitro, even in the absence of DNA (32)). Upon hormone treatment, the catalytic subunit of PKA dissociates from its regulatory subunit, diffuses to the nucleus, and indirectly modifies C/EBP␦ and/or activates its transcriptional partners, leading to stimulation of IGF-I gene expression. We cannot exclude the possibility that cytoplasmic PKA provides another signal necessary for transcription factor function in addition to maintaining C/EBP␦ in the nucleus. Elucidating the steps involved in this pathway controlling IGF-I expression in bone should provide opportunities for therapeutic interventions in osteoporosis and other skeletal disorders. FIG. 9. Model of hormonal regulation of IGF-I gene transcription in human osteoblasts. The left panel shows that basal PKA activity in the cytoplasm is required for nuclear expression of C/EBP␦. The right panel illustrates that upon hormonal stimulation and increased production of cAMP, the catalytic subunit of PKA translocates to the nucleus, where it stimulates the transcriptional activity of C/EBP␦ and induces IGF-I gene expression. The cytoplasm is shown in yellow, and the nucleus is in green. AC, adenylate cyclase; C, catalytic subunit of PKA; cA, cAMP, Gs, stimulatory G protein; R, regulatory subunit of PKA; ␦, C/EBP␦.