Cyclic GMP-dependent Protein Kinase Regulates CCAAT Enhancer-binding Protein β Functions through Inhibition of Glycogen Synthase Kinase-3*

The CCAAT enhancer-binding protein (C/EBPβ) plays an important role in the regulation of gene expression during cell proliferation, differentiation, and apoptosis. We previously showed that C/EBPβ participates in cGMP-regulated transcription of c-fos in osteoblasts (Chen, Y., Zhuang, S., Cassenaer, S., Casteel, D. E., Gudi, T., Boss, G. R., and Pilz, R. B. (2003) Mol. Cell. Biol. 23, 4066–4082). In the present work, we show that cGMP/cGMP-dependent protein kinase (PKG) induced dephosphorylation and activation of C/EBPβ by inhibiting glycogen synthase kinase-3β (GSK-3β). Phosphorylation of GSK-3β on Ser9 negatively regulates the enzyme activity, and we found that PKG phosphorylated this site both in vitro and in vivo; the in vivo phosphorylation occurred rapidly and preceded C/EBPβ dephosphorylation. Previous studies with GSK-3 inhibitors suggest that GSK-3β is a C/EBPβ kinase in resting cells. We determined that GSK-3β phosphorylated C/EBPβ in vitro on Thr189, Ser185, Ser181, and Ser177; C/EBPβ was phosphorylated on these same sites in intact, unstimulated osteoblasts, and phosphorylation was decreased in cGMP-treated cells. Mutation of the GSK-3 phosphorylation sites in C/EBPβ prevented C/EBPβ phosphorylation in resting cells, enhanced C/EBPβ DNA binding, and led to increased target gene transactivation, mimicking the stimulatory effects of cGMP on C/EBPβ. cGMP regulation of C/EBPβ was disrupted by a mutant GSK-3β(Ala9) resistant to cGMP/PKG phosphorylation and inhibition. We conclude that cGMP increases the DNA binding potential of C/EBPβ by preventing the negative effects of GSK-3 phosphorylation.

proteins were purified to homogeneity as judged by SDS-PAGE/Coomassie Blue staining and were DNA binding-competent.
In vitro phosphorylation of GSK-3 was performed in a total volume of 10 l, with 0.3 pmol of PKG II incubated at 30°C with variable amounts of GSK-3␤ in kinase assay buffer (20 mM Tris-HCl, 10 mM MgCl 2 , 5 mM dithiothreitol, pH 7.5) and 5 M [␥- 32 PO 4 ]ATP in the presence or absence of 10 M cGMP for the indicated time. For Western blots with anti-phospho-GSK3␤(Ser 9 ) antibody, reactions were performed using unlabeled ATP. For in vitro phosphorylation of C/EBP␤, 100 pmol of His-tagged wild type or mutant C/EBP␤ was incubated with 1 pmol of either GSK-3␤ or purified catalytic subunit of PKA (gift of S. Taylor) in kinase assay buffer with 5 M [␥- 32 PO 4 ]ATP at 30°C for 30 min. Proteins were separated by SDS-PAGE, transferred to Immobilon-P TM , and analyzed by autoradiography. The phosphorylated C/EBP␤ band was cut out, digested with trypsin, and analyzed by two-dimensional phosphopeptide mapping using high voltage electrophoresis in the first dimension and thin layer chromatography in the second dimension as described (40).
Some in vitro phosphorylation experiments were performed with Myc epitope-tagged wild type and mutant C/EBP␤ proteins expressed in UMR106 cells. Cells were transfected with 1 g of C/EBP␤ expression vector per 6-well dish, C/EBP␤ was isolated by immunoprecipitation with anti-Myc antibody, and precipitates were washed in kinase assay buffer and incubated with purified GSK-3␤ and [␥- 32 PO 4 ]ATP.
Alkaline Phosphatase Treatment of C/EBP␤-UMR106 cells were transfected with 20 ng of expression vector encoding Myc epitopetagged wild type and mutant C/EBP␤ per 6-well dish. Anti-Myc immunoprecipitates were isolated and incubated at 37°C for 30 min with either buffer alone (50 mM Tris-HCl, 100 mM NaCl, 10 mM MgCl 2 , 1 mM dithiothreitol, pH 7.9) or buffer plus 5 units of calf intestinal phosphatase (New England Biolabs), in the presence or absence of the phosphatase inhibitors ␤-glycerolphosphate (10 mM), NaF (10 mM), and imidazole (2 mM).
In Vivo Phosphorylation of C/EBP␤-UMR106 cells were either left untransfected or transfected in 6-well dishes with 50 ng of PKG II and 20 ng of wild type C/EBP␤, mutant C/EBP␤, or pXJ40-Myc empty vector as indicated. Forty h after transfection, cells were transferred to phosphate-free Dulbecco's modified Eagle's medium for 1 h and labeled for 4 h with 32 PO 4 (100 Ci/6-well dish), with 250 M 8-CPT-cGMP added to some cultures for the last 1 h. Cell lysates were subjected to immunoprecipitation with either a C/EBP␤-specific antibody (see Fig. 3) or anti-Myc antibody (see Fig. 4), and immunoprecipitates were analyzed by SDS-PAGE, electroblotting, and autoradiography. The amount of C/EBP␤ present in the immunoprecipitates was determined by blotting with an anti-C/EBP␤ antibody.
Electrophoretic Mobility Shift Assays-Nuclear extracts were incubated with 5Ј-end-labeled double-stranded oligodeoxynucleotide (oligo-dNT) probes encoding a canonical CRE or C/EBP binding sequence and analyzed by nondenaturing PAGE and autoradiography as described previously (13). The CRE oligo-dNT (5Ј-AGAGATTGC-CTGACGTCAGAGAGCTAG-3Ј) and the canonical C/EBP binding site (5Ј-TGCAGATTGCGCAATCTGCA-3Ј) were from Promega Life Sciences and Santa Cruz Biotechnology, respectively. (The CRE and C/EBP consensus sequences are underlined.) For supershift assays, nuclear extracts were incubated for 20 min at 4°C with specific antibodies prior to the addition of probe (13).
Data Presentation-Results presented in bar graphs represent the mean Ϯ S.D. of at least three independent experiments performed in duplicate. Autoradiographs and Western blots demonstrate a representative experiment performed at least three times with similar results.

RESULTS
cGMP Induces PKG II-dependent Phosphorylation of GSK-3␤ on Ser 9 -Early passage UMR106 rat osteoblasts express both PKG I and PKG II, and c-fos mRNA is synergistically induced by cGMP and calcium (13). This synergistic effect is dependent on PKG activity, is enhanced more by transfection of PKG II than PKG I, and requires cooperation between C/EBP␤ and CREB, with cGMP and calcium regulating C/EBP␤ and CREB activity, respectively (13). In C6 glioma cells, cGMP induces dephosphorylation of C/EBP␤, suggesting that cGMP/PKG regulate the activity of a C/EBP␤ kinase or phosphatase (13). GSK-3 is a candidate kinase, because GSK-3 is active in unstimulated cells, its activity is down-regulated by multiple signal transduction pathways, and GSK-3 can phosphorylate C/EBP␣ and -␤ in vitro (34,41). To investigate the potential role of GSK-3 in mediating the effects of cGMP on C/EBP␤, we examined whether PKG activation induces GSK-3 phosphorylation on a regulatory site that inhibits GSK-3 activity (i.e. Ser 9 on GSK-3␤) (31,32).
Early passage UMR106 cells were incubated in serum-free medium and treated for 20 or 60 min with the membrane-permeable cGMP analog 8-CPT-cGMP, and the phosphorylation state of GSK-3␤ was assessed by Western blotting using an antibody specific for GSK-3␤ phosphorylated on Ser 9 . 8-CPT-cGMP increased GSK-3 Ser 9 phosphorylation to a similar degree as serum (Fig. 1A, upper panel, compare lanes 2 and 3, cells treated with cGMP, and lane 5, serum-stimulated cells, with lanes 1 and 4, untreated cells). Specificity of the anti-phospho-Ser 9 antibody was demonstrated using purified GSK-3␤ phosphorylated by PKA in vitro (data not shown, but Fig. 2B, discussed below, shows similar results). Equal loading of the blot was shown by reprobing with an antibody recognizing GSK-3␤ irrespective of its phosphorylation state (Fig. 1A, middle). UMR106 cells express low levels of GSK-3␣, preventing assessment of the cGMP effect on GSK-3␣ phosphorylation.
As occurs in other cell types (45), we found that PKG expression progressively declined in UMR106 cells after prolonged passage in culture; total PKG activity, representing PKG I and II, was 220 Ϯ 25 pmol/ min/mg protein in early passage cells and Ͻ20 pmol/min/mg protein in late passage cells. In late passage UMR106 cells, 8-CPT-cGMP did not increase GSK-3␤ Ser 9 phosphorylation in either untransfected cells or cells transfected with an empty vector, but transfecting cells with a PKG II expression vector restored the effect of cGMP on GSK-3 phosphorylation (Fig. 1B, compare top left panel, cells transfected with empty vector, with top right panel, cells transfected with PKG II expression vector; total GSK-3 levels and PKG II expression are shown in the second and the lowest panel, respectively). cGMP induced GSK-3␤ phosphorylation within 5 min, and the effect persisted for several hours, but cGMP did not change Akt phosphorylation ( Fig. 1B; phospho-Akt levels are shown in the third panel). Similar results were obtained in PKG-deficient C6 rat glioma cells transfected with PKG II expression vector versus empty vector (data not shown). These results indicate that the effect of cGMP on GSK-3 phosphorylation is mediated by PKG II and independent of Akt.
To determine whether the effect of cGMP on GSK-3 phosphorylation occurred in other cells expressing endogenous PKG II, we examined human embryonal WI38 fibroblasts. Treating WI38 cells with 8-CPT-cGMP induced a rapid and sustained phosphorylation of GSK-3␤ Ser 9 with no detectable change in Akt Ser 473 phosphorylation (Fig. 1C). Thus, GSK-3␤ phosphorylation on Ser 9 , a site that negatively regulates kinase activity, was increased by cGMP in several cell types, including osteoblasts, glial cells, and embryonal lung fibroblasts.
Purified PKG II Directly Phosphorylates GSK-3␤ Ser 9 in Vitro-Since 8-CPT-cGMP induced GSK-3␤ Ser 9 phosphorylation in intact cells without activation of Akt, PKA, or Erk-1/2, we asked whether PKG II can directly phosphorylate this site in vitro. Incubating purified PKG II with recombinant GSK-3␤ in the presence of [␥-32 PO 4 ]ATP and cGMP led to high levels of 32 PO 4 incorporation into GSK-3; in the absence of cGMP, lower GSK-3 phosphorylation was observed, consistent with low basal PKG II activity ( Fig. 2A, compare lanes 3 and 4, GSK-3␤ incubated with PKG II in the presence and absence of cGMP, respectively, with lane 1, GSK-␤ alone, and lane 2, PKG II alone; note PKG II autophosphorylation). A similar experiment, performed with unlabeled ATP and with the reaction products analyzed by Western blotting using In the three top panels, Western blots were probed with the same antibodies as described in A. The Western blot in the bottom panel was developed using a PKG II-specific antibody (37). C, WI38 lung embryonal fibroblasts were treated for variable times with cGMP and cell lysates analyzed by Western blotting as described in A.
phospho-Ser 9 -specific antibodies, showed that PKG II phosphorylated GSK-3␤ on Ser 9 (Fig. 2B). To determine the affinity of PKG II for GSK-3 as a substrate, phosphorylation reactions were carried out with variable GSK-3␤ concentrations (Fig. 2C). The apparent K m of PKG II for GSK-3␤ was 0.15 Ϯ 0.03 M (mean Ϯ S.D. of three independent determinations), which is similar to the K m of PKG II for the cystic fibrosis transmembrane conductance regulator, a major physiologic PKG II target (46). PKG II appears to have a greater affinity for GSK-3␤ than PKA, whose K m for GSK-3␤ is 7.2 M (35).
Characterization of GSK-3␤ as a C/EBP␤ Kinase-Both C/EBP␣ and C/EBP␤ contain a highly conserved serine/threonine-rich region adjacent to the C-terminal basic leucine zipper domain ( Fig. 4A; the amino acid sequence of the serine-rich region in rat, mouse, and human C/EBP␤ is 100% conserved). Ross et al. (34) observed C/EBP␣ phosphorylation in resting cells, which was decreased in cells treated with lithium, a GSK-3 inhibitor. They also showed GSK-3 phosphorylation of C/EBP␣ in vitro and mapped phosphorylation to three sites that constitute a typical GSK-3 consensus sequence ((S/T)XXX(S/T), where X represents any amino acid; Fig. 4A). C/EBP␤ immunoprecipitated from mammalian cells was also phosphorylated by GSK-3 in vitro, but the phosphorylation sites were not determined (41). Therefore, we decided to analyze C/EBP␤ phosphorylation by GSK-3 in further detail.
Recombinant C/EBP␤ purified from bacteria was incubated with [␥- 32 PO 4 ]ATP in the presence or absence of purified GSK-3␤ or PKA catalytic subunit, with the latter serving as a positive control (Fig. 3A). Maximal 32 PO 4 incorporation into C/EBP␤ was higher in the presence of GSK-3␤ than PKA, and the GSK-3␤-phosphorylated protein migrated with a slightly higher apparent molecular weight than the PKA-phosphorylated protein. A two-dimensional phosphopeptide map of C/EBP␤ phosphorylated by GSK-3 in vitro produced several confluent spots with low mobility in both dimensions (Fig. 3B, middle). In contrast, the map of C/EBP␤ phosphorylated by PKA in vitro demonstrated multiple clearly defined spots with high mobility in one or both dimensions (Fig. 3B, left). The latter map is similar to published data and thus served as a control for complete digestion of C/EBP␤ (40). Using site-directed mutagenesis, PKA has been shown to phosphorylate rat C/EBP␤ at Ser 105 and Ser 240 (in the left panel of Fig. 3B, phosphopeptides eliminated by mutation of Ser 105 and Ser 240 are marked with x and y, respectively) (40). GSK-3, therefore, does not appear to target Ser 105 or Ser 240 in vitro, although Ser 105 together with Ser 101 forms a potential GSK-3 consensus sequence (47).
To examine C/EBP␤ phosphorylation in intact cells, UMR106 cells were incubated with 32 PO 4 for 4 h, and some cultures received 8-CPT-cGMP for 1 h prior to isolating C/EBP␤ by immunoprecipitation. As previously shown in C6 glioma cells (13), cGMP treatment of UMR106 cells decreased 32 PO 4 incorporation into C/EBP␤ by about 50% (Fig. 3C, compare lanes 2 and 3, cells cultured in the absence or presence of 8-CPT-cGMP; lane 1 shows cells subjected to immunoprecipitation with control IgG). To increase the recovery of C/EBP␤, we performed similar experiments in UMR106 cells transfected with small amounts of expression vectors encoding C/EBP␤ and PKG II. As occurred with the endogenous protein, 32 PO 4 incorporation into transfected C/EBP␤ was decreased in cGMP-treated cells (Fig. 3C, compare lanes 4 and 5). Unless stated otherwise, the amounts of transfected C/EBP␤ and PKG II were titrated not to exceed endogenous protein levels by more than 2-3-fold. When the in vivo phosphorylated C/EBP␤ from untreated UMR106 cells was subjected to two-dimensional phosphopeptide mapping, the majority of the radioactivity was recovered in several confluent spots with low mobility in both dimensions (arrow in Fig. 3B, right panel), similar to the spots obtained with C/EBP␤ phosphorylated by GSK-3 in vitro (Fig. 3B, compare middle and right panels). Phosphopeptide maps obtained with C/EBP␤ isolated from cGMP-treated cells showed decreased intensity of these confluent low mobility spots. Similar results were obtained in C6 glioma cells (not shown). Since GSK-3 is active in unstimulated cells (31,32), these results suggest that phosphorylation of C/EBP␤ in untreated cells may be due to GSK-3 and that  1, 3, and 4) and/or purified PKG II (lanes 2-4) were incubated in the presence of [␥- 32 PO 4 ]ATP for 30 min as described under "Experimental Procedures"; to some samples, 10 M cGMP was added (lanes 2 and 3). Reaction products were analyzed by SDS-PAGE/autoradiography; autophosphorylated PKG II is visible on the top. B, purified GSK-3␤ (lanes 1-3) and PKG II (lanes 2 and 3) were incubated in the presence of unlabeled ATP; 10 M cGMP was added to the sample in lane 2. Reaction products were analyzed by SDS-PAGE/Western blotting using antibodies specific for GSK-3␤ phosphorylated on Ser 9 (top), or total GSK-3␤ (bottom). C, variable amounts of GSK-3␤ were incubated with PKG II and [␥- 32 PO 4 ]ATP in the presence of 10 M cGMP for 5 min. Reaction products were analyzed by SDS-PAGE/autoradiography (shown as an inset, with 0.07, 0.13, 0.25, 0.5, 1, 2.5, and 5 pmol of GSK-3␤ in lanes 1-7). The bands representing phosphorylated GSK-3␤ were excised, and the amount of 32 PO 4 in each band was quantitated by scintillation counting and comparison to a standard curve (generated by counting variable amounts of [␥-32 PO 4 ]ATP). A Lineweaver-Burk plot was generated using the mean Ϯ S.D. of three independent experiments, yielding a K m of 0.15 M.
C/EBP␤ phosphorylation may be decreased in cGMP-treated cells because of GSK-3 inhibition. This conclusion is supported by the observation that C/EBP␤ phosphorylation in intact cells is reduced when GSK-3 is inhibited by valproic acid or lithium (13,41).
Mapping GSK-3␤ Phosphorylation Sites in the Serine-rich Region of C/EBP␤-Since the serine-rich region of C/EBP␤ contains a GSK-3 consensus motif that is conserved with the GSK-3 phosphorylation sites described in C/EBP␣ (Fig. 4A), we performed site-directed mutagenesis of C/EBP␤ mutating all of the likely phosphoacceptor sites in this region to alanine residues. Wild type and mutant C/EBP␤ containing Ala 189 , Ala 185 , Ala 181 , or Ala 177 were purified from bacteria and incubated with GSK-3 in the presence of [␥-32 PO 4 ]ATP (Fig. 4B). Compared with wild type protein, 32 PO 4 incorporation into the C/EBP␤ mutants Ala 189 and Ala 185 was severely reduced, whereas incorporation into mutants Ala 181 and Ala 177 was moderately decreased (Fig. 4B, top). Similar amounts of wild type and mutant proteins were present in the assay (Fig. 4B, bottom, shows a Coomassie stain of the unphosphorylated proteins). Thus, GSK-3 phosphorylates multiple sites in C/EBP␤; Thr 189 and Ser 185 appear to be preferred sites, and mutation of either site appears to impair phosphorylation of neighboring site(s). This may be because GSK-3 prefers serine/threonine residues immediately N-terminal to a proline, and GSK-3 is a "hierarchical" kinase whose substrate affinity is enhanced when the substrate is prephosphorylated at the ϩ4-position of the consensus sequence (S/T) 1 XXX(S/T) 4 (32,48). Some substrates require phosphorylation by a "priming" kinase at the ϩ4-position, but others do not, and when sequential overlapping GSK-3 sites are present as in C/EBP␤, GSK-3 can act as its own priming kinase (48).
Purification of C/EBP␤ from bacteria requires a denaturation/renaturation step, which could lead to variable amounts of unfolded proteins, which might be differentially recognized by kinases. We therefore performed additional in vitro phosphorylation experiments with C/EBP␤ mutants expressed in mammalian cells. Myc epitope-tagged versions of wild type and mutant C/EBP␤ were expressed at high levels in UMR106 cells, isolated by immunoprecipitation with anti-epitope antibody, and incubated with [␥- 32 PO 4 ]ATP in the presence of purified GSK-3␤. The pattern of 32 PO 4 incorporation into C/EBP␤ mutants containing single amino acid substitutions (i.e. Ala 189 , Ala 185 , Ala 181 , or Ala 177 ) was similar to the pattern observed with the bacterially expressed proteins (Fig. 4C, upper panel 1, 2, 4, and 6). Cell lysates were subjected to immunoprecipitation using an antibody specific for C/EBP␤ (lanes 2-5) or using control IgG (cIg; lanes 1 and 6). Immunoprecipitates were analyzed by SDS-PAGE/electroblotting followed by either autoradiography (top), or Western blotting with an anti-C/EBP␤ antibody (bottom).
All mutant proteins were expressed and immunoprecipitated at levels comparable with the wild type protein (Fig. 4C, bottom; lane 8 shows an immunoprecipitate obtained from cells transfected with empty vector). Thus, mutation of all four potential phosphoacceptor sites in the GSK-3 consensus sequence of C/EBP␤ almost completely eliminated GSK-3 phosphorylation of the protein, and the triple mutant showed that GSK-3 can phosphorylate Thr 189 in vitro.
To determine whether the same sites are targeted by GSK-3 in vivo, UMR106 cells expressing low levels of epitope-tagged C/EBP␤ proteins were incubated with 32 PO 4 , and some cells were treated with cGMP. As described above, 32 PO 4 incorporation into wild type C/EBP␤ was lower in cGMP-treated compared with untreated cells, but minimal 32 PO 4 incorporation into C/EBP␤ M(A) 4 occurred in either condition (Fig. 4D, upper panel, compare lanes 4 and 5, cells transfected with C/EBP␤ M(A) 4 , with lanes 2 and 3, cells transfected with wild type C/EBP␤; the Western blot in the lower panel shows that similar amounts of wild type and mutant protein were immunoprecipitated). Thus, mutation of the four serine/threonine residues constituting a GSK-3 consensus sequence in C/EBP␤ almost completely eliminated in vivo phosphorylation in UMR106 cells cultured in the presence or absence of cGMP, supporting the notion that GSK-3 is the main kinase phosphorylating C/EBP␤ in resting cells.
Compared with wild type C/EBP␤ isolated from UMR106 cells, C/EBP␤ M(A) 4 migrated slightly faster on SDS-PAGE (Fig. 4, D and E).
To determine the effect of C/EBP␤ phosphorylation on the protein's migration, wild type and mutant C/EBP␤ M(A) 4 isolated from resting UMR106 cells were incubated with calf intestinal alkaline phosphatase. Dephosphorylation of wild type C/EBP␤ by this nonspecific phosphatase caused most of the protein to shift to a faster migrating species that co-migrated with the M(A) 4 mutant; phosphatase treatment did not affect migration of the mutant protein (Fig. 4E, compare lanes 1 and 2  with lanes 3 and 4, proteins incubated in the absence or presence of alkaline phosphatase, respectively; in lanes 5 and 6, protein phosphatase inhibitors were included as a control). Thus, migration of C/EBP␤ M(A) 4 is similar to migration of completely dephosphorylated wild type C/EBP␤. The fact that cGMP-induced dephosphorylation of wild type C/EBP␤ is only partial and not associated with a detectable gel shift is discussed below.
Effect of C/EBP␤ Phosphorylation on Transcriptional Activation of a Target Gene-We previously showed that C/EBP␤ cooperates with CREB to stimulate transcription from CRE-containing promoters and that the transactivation potential of C/EBP␤ at the CRE is regulated by cGMP (13). To determine whether the effect of cGMP is mediated by GSK-3 phosphorylation of C/EBP␤, we performed two sets of experiments. First, we compared the wild type and phosphorylation-deficient mutant C/EBP␤ M(A) 4 with respect to their ability to transactivate a CRE-driven luciferase reporter gene. UMR106 cells were co-transfected with pCRE-Luc and increasing amounts of expression vector encoding either wild type or mutant C/EBP␤, with some cells treated with 8-CPT-cGMP (Fig. 5A). As described previously (13), transfection of wild type C/EBP␤ led to transactivation of the CRE-dependent reporter gene, and at each level of C/EBP␤ expression, cGMP further enhanced reporter gene activity (Fig. 5A, compare black and gray bars, cells cultured in the presence and absence of cGMP, respectively; p Ͻ 0.05 at each level of wild type C/EBP␤ vector DNA). At similar expression levels of wild type and mutant C/EBP␤ M(A) 4 (Fig. 5B), the mutant protein activated the reporter gene more than wild type C/EBP␤ in the absence of cGMP, and cGMP did not affect reporter gene activity in cells expressing C/EBP␤ M(A) 4 (p Ͻ 0.05 for the comparison between untreated cells transfected with wild type and mutant C/EBP␤ at each level of vector DNA). Thus, mutation of GSK-3 phosphorylation sites in C/EBP␤ mimicked the  (38). The serine-rich region of rat C/EBP␤ (amino acid residues 171-194), including a putative GSK-3 consensus sequence ((S/T)XXX(S/ T)XXX(S/T)) is enlarged; a homologous region of C/EBP␣ is shown below, which contains three sites phosphorylated by GSK-3 (in boldface type) (34). The sequence of the C/EBP␤ serine-rich region is identical in the rat, mouse, and human protein (Thr 189 in rat C/EBP␤ corresponds to Thr 188 in the mouse and Thr 235 in the human protein). B, wild type (WT; lane 1) and mutant C/EBP␤ proteins with the indicated amino acid substitutions (lanes 2-5) were purified from bacteria, and equal amounts of protein were incubated with purified GSK-3␤ and [␥-32 PO 4 ]ATP as described under "Experimental Procedures." Phosphorylation products were analyzed by SDS-PAGE and autoradiography (top). Unphosphorylated proteins were analyzed by SDS-PAGE and Coomassie Blue staining (bottom). C, Myc epitope-tagged wild type (lane 1) and mutant C/EBP␤ constructs (lanes 2-7) were expressed in transfected UMR106 cells and isolated by immunoprecipitation with antiepitope antibody. Immunoprecipitates were incubated with purified GSK-3␤ and [␥-32 PO 4 ]ATP as described under "Experimental Procedures," and reaction products were analyzed by SDS-PAGE/electroblotting/autoradiography (top). The membrane was later probed with a C/EBP␤-specific antibody, demonstrating comparable amounts of C/EBP␤ proteins present in the immunoprecipitates (bottom).  2 and 3), or mutant C/EBP␤ M(A) 4 (lanes 4 and 5); cells were labeled with 32 PO 4 , and some cultures were treated with 8-CPT-cGMP (lanes 3 and 5) as described in Fig. 3B. Cell lysates were subjected to immunoprecipitation using anti-epitope antibody, and immunoprecipitates were analyzed by SDS-PAGE/electroblotting/autoradiography (top), or Western blotting with an anti-C/EBP␤ antibody (bottom). E, wild type (lanes 1, 3, and 5) and mutant C/EBP-␤ M(A) 4 (M; lanes 2, 4, and 6) were isolated from transfected, untreated UMR106 cells as described in C. The immunoprecipitated C/EBP-␤ proteins were incubated with buffer alone (lanes 1 and 2), with purified alkaline phosphatase (AP), or with alkaline phosphatase plus phosphatase inhibitor mixture (APϩInh) as described under "Experimental Procedures." Reaction products were analyzed by SDS-PAGE and Western blotting using a C/EBP␤-specific antibody. Migration of phosphorylated and dephosphorylated C/EBP␤ is indicated by two arrows. effect of cGMP, leading to enhanced transactivation of a target gene, but rendered the protein insensitive to cGMP.
Next, we examined C/EBP␤-dependent transcription in cells expressing the constitutively active GSK-3␤(Ala 9 ) mutant (32). UMR106 cells were co-transfected with pCRE-Luc, wild type C/EBP␤, and either wild type or mutant GSK-3␤(Ala 9 ); some cells received 8-CPT-cGMP (Fig. 5C). In cells co-transfected with wild type GSK-3␤, cGMP stimulated C/EBP␤-dependent reporter gene activity 2-fold, but co-transfection of the constitutively active mutant GSK-3␤(Ala 9 ) prevented the transcriptional effect of cGMP (p Ͻ 0.05 for the comparison between cGMP-treated cells transfected with wild type versus mutant GSK-3␤). These findings are consistent with the interpretation that cGMP targets GSK-3 and that the effect of cGMP on C/EBP␤-mediated transactivation is explained by GSK-3 inhibition and dephosphorylation of C/EBP␤ in cGMP-treated cells. These results support our earlier finding that inhibition of GSK-3 with valproate mimics the effect of cGMP on C/EBP␤ transactivation (13).
Effect of C/EBP␤ Phosphorylation on Its DNA Binding Activity-To examine the DNA binding activity of the phosphorylation-deficient mutant C/EBP␤ M(A) 4 , we performed electrophoretic mobility shift assays with double-stranded oligo-dNT probes encoding either a CRE or a canonical C/EBP binding sequence. With the CRE probe, nuclear extracts from UMR106 cells transfected with wild type or mutant C/EBP␤ M(A) 4 produced a distinct protein-DNA complex not present in cells transfected with empty vector (Fig. 6A, compare lanes 3 and 4, cells transfected with wild type or mutant C/EBP␤, with lane 2, cells transfected with empty vector; the right panel shows a shorter exposure). The protein-DNA complex formed by mutant C/EBP␤ M(A) 4 was more intense and demonstrated altered mobility compared with the complex formed by wild type C/EBP␤. Preincubation of nuclear extracts with excess unlabeled oligo-dNT probe resulted in competition of all specific protein-DNA complexes (Fig. 6A, lane 1, shows results for wild type protein, but similar results were obtained with the mutant protein). The addition of excess unlabeled oligo-dNT with an unrelated sequence (e.g. SP1 consensus sequence) had no effect on formation of the protein-DNA complexes (data not shown). Preincubation of nuclear extracts with a C/EBP␤-specific antibody resulted in disappearance of the protein-DNA complexes formed by the transfected proteins and appear-  4 , respectively. C, cells were transfected with pCRE-Luc, pRSV-␤Gal, and PKG II, and some cultures were treated with 8-CPT-cGMP as described in A, but cells were co-transfected with 5 ng of wild type C/EBP␤ and either wild type or mutant GSK-3␤(Ala 9 ) as indicated. The luciferase/␤-galactosidase ratio measured in untreated cells co-transfected with wild type GSK-3␤ was assigned a value of 1.  5), and some cells were treated with 250 M 8-CPT-cGMP (lanes 3 and 5) for 1 h prior to harvesting. Electrophoretic mobility shift assays were performed with radioactively labeled oligo-dNT probes encoding either a canonical CRE sequence (top) or a canonical C/EBP binding site (middle). Bottom, nuclear extracts were analyzed by Western blotting using a Myc epitope-specific antibody.
ance of "supershifted" complexes (Fig. 6A, lanes 6 and 7; supershifted complexes are marked with an asterisk). Thus, although the protein-DNA complex formed by mutant C/EBP␤ M(A) 4 migrated differently from the complex formed by the wild type protein on nondenaturing PAGE, it was nevertheless recognized by the C/EBP␤-specific antibody.
In UMR106 cells transfected with wild type C/EBP␤, the amount of C/EBP␤-specific complexes formed with the CRE probe increased in cells treated with 8-CPT-cGMP, compared with untreated cells (Fig. 6B, top, compare lanes 2 and 3, cells cultured in the absence and presence of cGMP). In contrast, the mutant C/EBP␤ M(A) 4 demonstrated increased DNA binding in the absence of cGMP that was insensitive to cGMP (Fig. 6B, top, lanes 4 and 5). Wild type and mutant C/EBP␤ were expressed at the same level (Fig. 6B, lower panel). These results are consistent with increased transcriptional activation of a CRE-dependent reporter gene by the mutant C/EBP␤ M(A) 4 (Fig. 5). Similarly, binding of wild type C/EBP␤ protein to an oligo-dNT probe encoding a canonical C/EBP binding sequence was enhanced when cells were treated with cGMP, and the mutant C/EBP␤ M(A) 4 bound more effectively than wild type, but binding of the mutant protein was not modulated by cGMP (Fig. 6B, middle, compare lanes 4 and 5, mutant C/EBP␤ M(A) 4 , with lanes 2 and 3, wild type C/EBP␤; the lower panel shows expression of both proteins on a Western blot). Thus, mutation of the GSK-3 phosphorylation sites in C/EBP␤ increases DNA binding to both CRE and canonical C/EBP binding sequences but renders DNA binding insensitive to cGMP regulation. Consistent with these results, alkaline phosphatase treatment of nuclear extracts increases the C/EBP␤ DNA binding (41). Thus, increased DNA binding of C/EBP␤ in cGMPtreated cells can be explained by dephosphorylation of C/EBP␤, related to down-regulation of GSK-3 activity by cGMP-induced GSK-3 Ser 9 phosphorylation (Fig. 1).

DISCUSSION
We previously found that cGMP modulates the activity and phosphorylation state of C/EBP␤ in osteoblasts and neuronal cells (13). In the present work, we found that GSK-3␤ constitutively phosphorylates C/EBP␤ on specific sites in resting cells, thereby restraining C/EBP␤ DNA binding and transactivation potential. The effects of cGMP on C/EBP␤ can be explained by inhibition of GSK-3 activity, resulting in dephosphorylation of C/EBP␤, leading to increased DNA binding and transactivation of target genes.
PKG Phosphorylation of GSK-3␤ on Ser 9 -GSK-3 is abundant and highly active in unstimulated cells, and activation of several signal transduction pathways, including PI3K/Akt, cAMP/PKA, or Ras/Erk-1/2, decreases GSK-3 activity through phosphorylation of Ser 21 in GSK-3␣ or Ser 9 in GSK-3␤ (31,32). GSK-3␣ and -3␤ are thought to have overlapping functions and are typically concordantly regulated, although their expression varies among different cell types, and UMR106 cells express predominantly GSK-3␤ (31,32). We found that cGMP induced GSK-3␤ Ser 9 phosphorylation to a similar extent as serum stimulation; the cGMP effect was not explained by activation of Akt, PKA, or Erk-1/2 but was dependent on PKG activity. In neuronal and endothelial cells, cGMP can activate the PI3K/Akt pathway, whereas in platelets, cGMP inhibits PI3K activity (43,44,49,50); we found no significant effect of cGMP on Akt phosphorylation in osteoblasts and fibroblasts, suggesting that the effect of cGMP on PI3K/Akt is cell type-specific. cGMPinduced GSK-3 Ser 9 phosphorylation occurred in UMR106 osteoblasts, C6 glioma cells, and WI38 embryonal fibroblasts; it occurred in cells expressing endogenous PKG and was absent in PKG-deficient cells but was restored by transfection of PKG II into these cells. PKG II phosphorylated Ser 9 of GSK-3␤ in vitro with a K m similar to the K m for cystic fibrosis transmembrane conductance regulator, an established PKG II substrate, and lower than the K m reported for PKA phosphorylation of GSK-3␤ (35,46). GSK-3 as a C/EBP␤ Kinase-Our results and those from other laboratories (41) indicate that C/EBP␤ is a physiological GSK-3 substrate and that basal C/EBP␤ phosphorylation is due to GSK-3 activity. We found that the phosphopeptide map of C/EBP␤ phosphorylated in vivo, in resting UMR106 osteoblasts and C6 glioma cells, was similar to the map of C/EBP␤ phosphorylated by GSK-3 in vitro. Similarly, C/EBP␤ is phosphorylated in unstimulated NIH 3T3 fibroblasts and human macrophages on a tryptic peptide that includes the GSK-3 consensus sequence described in Fig. 4 (51, 52). Thus, C/EBP␤ phosphorylation on these sites may be quite ubiquitous, as is the expression and activity of GSK-3 (32). Mutation of phosphoacceptor sites within the GSK-3 consensus sequence of the C/EBP␤ serine-rich region abolished C/EBP␤ phosphorylation in intact cells as well as in vitro phosphorylation by GSK-3. Finally, dephosphorylation of C/EBP␤ occurred in intact cells when GSK-3␤ was inhibited by cGMP/PKG II or valproate, and C/EBP␤ dephosphorylation induced by cGMP or valporate led to increased DNA binding and target gene activation by the transcription factor (see Ref. 13 for results with the GSK-3 inhibitor valproate). In experiments where we transfected wild type or mutant C/EBP␤, we were careful to transfect only low levels of C/EBP␤ proteins, and cGMP treatment of UMR106 cells reduced phosphorylation of endogenous C/EBP␤ to a similar degree as transfected C/EBP␤, suggesting that the level of transfected C/EBP␤ did not exceed the regulatory capacity of the cells.
C/EBP␤ Dephosphorylation after Inhibition of GSK-3-In C6 glioma cells, we showed that cGMP-induced dephosphorylation of C/EBP␤ required PKG activity (13), as did cGMP-induced phosphorylation of GSK-3 on the inhibitory Ser 9 site. C/EBP␤ dephosphorylation was detectable at 15 min, was maximal at 1 h, and persisted for Ͼ2 h (13). The kinetics of cGMP-induced GSK-3␤ Ser 9 phosphorylation were similar in C6 and UMR106 cells and consistent with GSK-3 inhibition causing C/EBP␤ dephosphorylation, because GSK-3␤ phosphorylation was detectable at 5 min, peaked at 20 min, and persisted for several hours. C/EBP␤ dephosphorylation implies the activity of protein phosphatase(s), and we cannot exclude the possibility that cGMP may also increase the activity of C/EBP␤ phosphatase(s), in addition to decreasing GSK-3 activity. Two major serine/threonine protein phosphatases, PP1 and PP2B (calcineurin), however, are negatively regulated by cGMP/PKG (53)(54)(55). Future work will be aimed at identifying the protein phosphatase(s) responsible for dephosphorylating C/EBP␤ on the sites targeted by GSK-3. The finding that cGMP induced only partial (ϳ50%) dephosphorylation of C/EBP␤ without a detectable gel shift of the protein suggests that the four GSK-3 phosphorylation sites may turn over with different kinetics, perhaps due to differential phosphatase sensitivities, leading to cGMP-induced dephosphorylation of some but not all sites.
Dephosphorylation of C/EBP␤ due to inhibition of GSK-3 activity was also observed by Piwien-Pilipuk et al. (41), who showed that growth hormone induces C/EBP␤ dephosphorylation in 3T3-F442A preadipocytes and increases DNA binding through activation of PI3K and Akt, leading to Akt-mediated Ser 9 phosphorylation and GSK-3 inhibition. These results and ours suggest that C/EBP␤ is an in vivo substrate of GSK-3 and that inhibition of GSK-3 activity leads to C/EBP␤ dephosphorylation and activation. Similarly, insulin stimulates PI3K and Akt activity and induces dephosphorylation of C/EBP␣ through inhibition of GSK-3 in 3T3-L1 adipocytes (34). Other transcription factors targeted by GSK-3 are c-Jun, c-Myc, NF-B, and the adipocyte determination-and differentiation-dependent factor 1, most of which are negatively regulated by GSK-3 phosphorylation (32, 33, 56).