Protein Kinase D Induces Transcription through Direct Phosphorylation of the cAMP-response Element-binding Protein*

Protein kinase D (PKD), a family of serine/threonine kinases, can be activated by a multitude of stimuli in a protein kinase C-dependent or -independent manner. PKD is involved in signal transduction pathways controlling cell proliferation, apoptosis, motility, and protein trafficking. Despite its versatile functions, few genuine in vivo substrates for PKD have been identified. In this study we demonstrate that the transcription factor cAMP-response element-binding protein (CREB) is a direct substrate for PKD. PKD1 and CREB interact in cells, and activated PKD1 provokes CREB phosphorylation at Ser-133 both in vitro and in vivo. A constitutive active mutant of PKD1 stimulates GAL4-CREB-mediated transcription in a Ser-133-dependent manner, activates CRE-responsive promoters, and increases the expression of CREB target genes. PKD1 also enhances transcription mediated by two other members of the CREB family, ATF-1 and CREM. Our results describe a novel mechanism for PKD-induced signaling through activation of the transcription factor CREB and suggest that stimulus-induced phosphorylation of CREB, reported to be mediated by protein kinase C, may involve downstream activated PKD.

The mammalian PKD 3 family of serine/threonine kinases includes the isoforms PKD1 (mouse PKD and human PKC), PKD2, and PKD3 (also named PKC). PKD was originally considered to be a member of the PKC family (1,2) but is now classified in the calcium/calmodulin-dependent kinase group based on sequence similarities in the kinase domain (3). The PKDs share a similar architecture consisting of a C-terminal catalytic domain, an N-terminal regulatory domain that encompasses two cysteine-rich regions (C1a and C1b), and a pleckstrin homology (PH) domain (4 -7). A comparison of the amino acid sequences of PKD1-3 reveals that the highest homology lies in the catalytic domain, followed by C1a, C1b, and PH domain, suggesting that isoform-specific functions may be due to the regulatory part of the kinase (8).
PKD1 can be activated by several mechanisms. The most studied mechanism involves a sequential activation of phospholipase C that results in the generation of the second messengers inositol 1,4,5-triphosphate and diacylglycerol (DAG) and subsequent activation of the classical (␣, ␤I, ␤II, and ␥) and novel (␦, ⑀, , and ) PKC isoforms. Binding of DAG to the C1b domain of PKD1 directs its translocation to the plasma membrane where activated novel PKC phosphorylates PKD1 at Ser-744/748 in the activation loop, causing activation of the enzyme. The activated enzyme can be imported via its C1b motif into the nucleus, where it transiently accumulates before being exported to the cytosol through a CRM1-dependent nuclear export pathway that requires the PH domain of PKD (4 -7). Mutations and deletions in the regulatory domain induce activation of PKD to various extents, and the entire regulatory domain has an inhibitory effect on the kinase activity (9). Indeed, PKD can also be activated through caspase-mediated cleavage of the regulatory domain (10). G␤␥ subunits can activate PKD1 through direct interaction with the PH domain (11). However, a recent report suggested that the ␤ 1 ␥ 2 -mediated activation of PKD required PKC (12). Furthermore, PKD can be activated by bone morphogenetic protein 2 and endothelin-I in a PKC-independent manner, but the mechanisms remain elusive (13,14).
Depending on the cell type and external stimulus, PKD localizes to different cellular compartments (4). This suggests that PKD can regulate several cellular functions according to its cellular localization. PKD was shown to be involved in the fission of carriers from the trans-Golgi network to plasma membrane (15), and PKD-mediated phosphorylation of phosphatidylinositol 4-kinase III␤ results in enhanced vesicular stomatitis virus-G protein transport to the plasma membrane (16). PKD can also participate in cell motility (17,18), and altered PKD1 activity in prostate cancer cells influences cellular aggregation (18). PKD mediates PDGF-induced suppression of epidermal growth factor-triggered c-Jun N-terminal kinase (JNK) activation in a cell-dependent manner (19 -21). In addition, RIN1 phosphorylation by PKD1 has been demonstrated to result in activation of the Ras-MEK-ERK signaling pathway (22). Other studies have revealed a role for PKD in proliferation (23)(24)(25), osteoblastic cell differentiation (13), and apoptosis (4,26). Finally, PKD may affect gene expression by inducing the nuclear exclusion of histone deacetylases (26 -29).
CREB is a 43-kDa transcription factor whose activity is regulated by phosphorylation. Stimulus-induced phosphorylation of Ser-133, the major regulatory site, enables CREB to recruit the co-activators CBP/p300 and stimulate CREB-dependent transcription (34). More than 20 different protein kinases that are compounds of distinct signaling pathways have been described as CREB kinases (34,35).
By using an oriented peptide library approach, Nishikawa et al. (36) determined the preferred substrate phosphorylation motif of PKD. This and similar studies revealed that PKD seems to preferentially phosphorylate substrates with a LXRXX(T*/ S*) consensus motif, where T* and S* denote the phosphoacceptor sites threonine or serine (36 -38). Doppler et al. (38) used this knowledge and developed an antibody against a phosphopeptide with the PKD consensus motif. They identified Hsp27 as a novel substrate of PKD, but the antibody also detected an approximate 45-kDa protein in extracts from bombesin-, bradykinin-, PDGF-, phorbol ester-, or pervanadate-stimulated NIH3T3 cells (38). Interestingly, all these stimuli induce activation of PKD (39 -42), as well as CREB Ser-133 phosphorylation (34), and the molecular mass of 45 kDa corresponds well with that of CREB (43 kDa). These observations prompted us to test whether CREB is a PKD substrate. Our results show that CREB is an in vitro and in vivo PKD substrate and that PKD activates CREB-dependent transcription in a Ser-133-dependent manner. The biological consequences of these findings are discussed.
Plasmids-The plasmids GAL4-CREB, GAL4-CREB S133A, and pG5E1bLuc have been described previously (45). The empty expression vector pSR and the plasmid pSR C␣, containing the catalytic C␣ subunit from PKA, were the kind gifts from Dr. T. Jahnsen and Dr. K. Taskén (46). The GST-CREB plasmid (amino acids 1-261 of CREB) was a kind gift from Dr. Michael Comb (47). The plasmids pcDNA3.1, pEGFP-C1, pCMV-CREB, pCMV-K-CREB, and the pGEX vector were purchased from Invitrogen, Clontech, and GE Healthcare, respectively. EGFP-MK5L337 has been described previously (48) Constitutive active PKD (744/748 E/E mutant) and dominant negative PKD1 (K618N mutant) were cloned as XbaI/EcoRI fragments in pcDNA3.1 (40,41). GAL4-CREBS133A was digested with KpnI and StuI, generating an ϳ200-bp fragment containing the part of CREB surrounding S133A. This fragment was used to replace the corresponding area in GST-CREB, creating the GST-CREBS133A plasmid. pCRE-luc was purchased from Clontech, and pmutCRE-luc has been previously described (49). VP16-CREB was created by subcloning of the ϳ1-kb EcoRI and BamHI fragment from CREB-pGADP into VP16 (Clontech). CREB-pGADP was prepared as follows; PCR was performed with GAL4-CREB as a template, using the primers 5Ј-ATGACATGGAATTCGGAGCAGACAAC-3Ј and 5Ј-CTT-GRGGCAGTAAGGATCCTTAAGTG-3Ј, which includes a restriction site for EcoRI and BamHI respectively. The resulting PCR product was digested with EcoRI and BamHI and subcloned into pGADT7 (Clontech). To clone the PH-KD domain of PKD1 (corresponding to the fragment generated by cleavage with caspase-3, consisting of amino acid residues 427-918 of mouse PKD1), PCR on full-length PKD cDNA was performed using the primers 5Ј-GCA CAC GAA GCG GTC GAC CAG CAC TGT G-3Ј and 5Ј-CCA GGT CTG ATA GAG CTC TAG CCA AGG GTG ACT C-3Ј. The PCR product was digested with SalI and SacI and cloned in the corresponding sites of pSL1180. The PH-KD sequences were then excised with EcoRI and XhoI and ligated into the corresponding sites of pcDNA3.1 to generate the eukaryotic expression vector for the PH-KD fragment of PKD1.
Sequencing-Cycle sequencing was performed using the Big Dye sequencing kit (PerkinElmer Life Sciences). Sequencing reactions were analyzed on an ABI377 Prism Sequencer (PerkinElmer Life Sciences).
In Vitro Kinase Assay-In vitro kinase assay was performed as outlined before (43). GST fusion proteins were purified from Escherichia coli BL21 extracts using glutathione-agarose beads according to the instructions of the manufacturer. Equal amounts of GST-CREB and GST-CREB mutants were used in all experiments. Distinct proteolytic bands were obtained, which is in agreement with others (50,51).
Transient Transfection and Luciferase Assay-Transfection of COS-1 cells was performed by using the DEAE-dextran method (52). Transfection studies in SK-N-DZ, A431-PKD1, and HEK293 cells were done using Lipofectamine 2000 according instructions of the manufacturer. Luciferase assays were as described previously (49). As neither empty expression vector nor dominant negative (dn) PKD1 affected CREB-mediated transcription, we chose to use dnPKD1 as a reference to compare the activity of CREB in the presence of activated PKD1.
Immunoprecipitation-Confluent 10-cm plates with HEK293 cells transfected with appropriate constructs were washed twice in ice-cold PBS and harvested in a cold lysis buffer containing 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, and Complete protease inhibitor mixture (Roche Applied Science) (53). Lysates were clarified by centrif-ugation at 10,000 ϫ g for 10 min. Lysates were precleared with random but irrelevant antibody for 1 h at 4°C followed by precipitation of the nonspecific complexes with protein G-Sepharose (Amersham Biosciences). The precleared lysates were then incubated with anti-CREB protein for 1 h at 4°C, followed by precipitation of the complexes with protein G-Sepharose. The immune complexes were washed three times in ice-cold lysis buffer (see above) and twice in ice-cold 50 mM Tris, pH 8.0, and subsequently subjected to immunoblot analysis.
Immunoblot Analysis-Immunoblot analyses were performed on immune complexes or extracts derived from cells grown in 6-well trays. The cells were washed twice in PBS, harvested in 80 l of lysis buffer (0.25 M dithiothreitol, 1:1 NuPAGE lithium dodecyl sulfate sample buffer (four times), and double distilled H 2 O), heated to 70°C for 10 min, and sonicated for 4 s. Samples were analyzed by SDS-PAGE (4 -12% NuPAGE; Invitrogen) and transferred to 0.45-m pore size polyvinylidene difluoride membrane (Millipore). The blots were blocked with 5% nonfat dry milk (Nestle ), 0.1% Tween 20 (blocking buffer) for 1 h at room temperature and incubated with primary antibodies (1:1000) overnight at 4°C in blocking buffer. The blots were subsequently incubated for 1 h in room temperature with alkaline phosphate-conjugated anti-rabbit antibodies (DAKO, Denmark) in blocking buffer before exposure to chemiluminescence substrate CDP-Star. To verify equal protein loading, the membranes were stripped for 5 min in 0.2 M NaOH, washed three times in PBS with 0.1% Tween 20 (PBST), blocked, and reprobed with anti-CREB antibody.
RNA Isolation and Reverse Transcriptase (RT)-PCR-Total RNA from transfected HEK293 cells was isolated using the Nucleospin RNA II purification kit (Clontech and Takara Bio Inc., Shiga, Japan) according to the manufacturers' protocol. Two g of RNA was reverse-transcribed using iScript cDNA synthesis kit (Biocompare, Bio-Rad) and subjected to PCR. The primers for APRT and the PCR conditions have been described previously (49). To amplify Nur77 cDNA, the following primers were used: 5Ј-TCTGCTCAGGCCTGGTGCTAC-3Ј (forward) and 5Ј-GGCACCAAGTCCTCCAGCTTG-3Ј (reverse). The PCR cycling conditions were 30 times at 94°C for 30 s, 60°C for 30 s, and 72°C for 30 s. PCR products were visualized on an ethidium bromide-stained agarose gel.

PKD1 Phosphorylates CREB in Vitro at Two Residues-PKD
is a protein kinase that has unique substrate specificity, with a consensus sequence LXRXX(T*/S*), where T*/S* denotes the target threonine and serine for phosphorylation, respectively (36 -38). Interestingly, CREB Ser-133 is contained within the PKD consensus motif LSRRPS*, and a study with antibodies against phosphopeptide with PKD consensus motif detected a 45-kDa protein in extracts from bombesin-, bradykinin-, PDGF-, phorbol 12-myristate 13-acetate-, or pervanadatestimulated NIH3T3 cells (38). All these stimuli induce activation of PKD as well as phosphorylation of CREB at Ser-133 (34, 39 -42). This prompted us to test whether purified activated PKD1 or its isoform PKD2 could phosphorylate CREB in vitro. To ensure that purified PKD was active, myelin basic protein (MBP) was included in the experiment. The in vitro kinase assays showed that both PKD1 and PKD2 phosphorylated the positive control MBP as well as GST-CREB, but not the negative control GST (Fig. 1A). Several bands are seen; however, it is generally known that purified GST-CREB always generates different proteolytic fragments (50,51). Mutating Ser-133 into nonphosphorylatable alanine significantly reduced PKD1-mediated phosphorylation of the largest fragment, whereas the phosphorylation levels of the smallest fragment were not affected (Fig. 1B, lanes 4 and 5). The smallest proteolytic fragment was phosphorylated by PKD but not by PKA (Fig. 1B, lanes 3 and 4). These findings suggest that PKD1 phosphorylates CREB at other residues in addition to Ser-133. PKD is a phorbol ester-activated kinase, and Ser-89 and Ser-121 have been suggested as putative phorbol esterinduced phosphoacceptor sites (54,55). To test whether these sites are targets for PKD1, double mutants in which Ser-133 along with Ser-89 or Ser-121, respectively, were replaced with alanine were generated by site-directed mutagenesis. The double mutants CREBS89A/S133A and CREBS121A/S133A could still be phosphorylated by activated PKD1, indicating that these sites are unlikely to be PKD phosphorylation sites in vitro (Fig. 1B). A closer examination of the amino acid sequence of CREB also revealed Ser-98 as a putative PKD target, as the motif contains a leucine and arginine in position Ϫ5 and Ϫ3, respectively. To explore whether Ser-98 is a PKD phosphoacceptor site, site-directed mutagenesis was performed to create the CREBS98A and CREBS98A/S133A mutants. CREBS98A, but not the double mutant, could be phosphorylated by PKD1 (Fig. 1C, upper panel). The smallest proteolytic fragment of CREBS98A was not phosphorylated by PKD1 (compare lanes 1 and 2 in Fig. 1C), suggesting that this fragment contains Ser-98. The amount of GST-CREB proteins used in the kinase assay were run on SDS-PAGE in order to evaluate the protein level. As shown in Fig. 1C (lower panel), the levels of GST fusion proteins used in the assays are comparable, confirming that the lack of autoradiography bands for GST-CREBS98A/S133A is because of mutation of phosphoacceptor sites for PKD and not the absence of protein.
PKD-stimulating Agents Induce CREB Ser-133 Phosphorylation-Several growth factors like PDGF and regulatory peptides, as well as oxidative stress, have been shown to result in PKC-mediated activation of PKD (7). Pervanadate is a stim- ulus that mimics oxidative stress (56), whereas tumor-promoting phorbol esters, like TPA, are agents that bypass membrane receptors and mimic DAG, resulting in PKC activation and phosphorylation of PKD at Ser-744/Ser-748 (40,41,(57)(58)(59). PDGF, TPA, and vanadate were chosen as stimuli to activate endogenous PKD in A431-PKD1 and/or SK-N-DZ cells.
First, we checked whether PKD was activated by PDGF, TPA, or vanadate in our cell lines by comparing the phospho-PKD1 (Ser-744/748) levels in extracts of unstimulated and stimulated cells. Both PDGF, TPA, and pervanadate clearly induced PKD1 and CREB Ser-133 phosphorylation in A431-PKD1 cells, which may suggest that PKD has a putative role as a CREB kinase induced by all three stimuli in these cells ( Fig. 2A, left panel). The same experiment was performed with TPA as a stimulus in SK-N-DZ cells ( Fig. 2A, right panel). Also in these cells, stimulation with TPA resulted in both CREB and PKD1 phosphorylation, suggesting that the PKD-induced CREB phosphorylation is not cell-specific.
To further evaluate the contribution of PKD in stimuli-induced CREB phosphorylation, we assayed the phosphoserine 133-CREB levels in extracts of PDGF-or TPA-stimulated A431-PKD1 and TPA-stimulated SK-N-DZ cells in the absence or presence of the inhibitors Gö6976 (inhibits PKC␣ and -␤ and PKD) and Gö6983 (inhibits PKC␣, -␤, -␥, -␦, or -). As can be seen in Fig. 2B, both inhibitors reduced the level of phosphorylated CREB in both PDGF-and TPAstimulated cells, but not the total level of protein (Fig. 2B, upper versus lower panel). The presence of the inhibitors alone did not influence the phosphorylation of CREB neither in A431-PKD1 cells (Fig. 2B, left) nor SK-N-DZ cells (results not shown). The inhibition with Gö6976 reduced phosphorylation of CREB more than Gö6983, which may suggest the involvement of PKD in PDGF-or TPA-provoked CREB phosphorylation. PKCs (␣, ␤, ␥, ␦, or ), although to a lesser extent than PKD, are also involved in PDGF-or TPA-induced CREB phosphorylation because the phospho-Ser-133 levels were diminished in the presence of Gö6983. The phosphospecific CREB antibodies cross-react with phosphorylated ATF-1. As can been seen in Fig. 2A, the PKD activators also induced phosphorylation of ATF-1, suggesting that this transcription factor may also be a substrate for PKD. Similarly to CREB, a stronger decrease in PDGF-provoked ATF-1 phosphorylation was observed with Gö6976, which also inhibits PKD, than with Gö6983, which does not affect kinase activity of PKD. Also TPA induced ATF-1 phosphorylation in a PKD-dependent manner (results not shown). These results suggest that PDGF-and TPA-induced activated PKD may participate in the signaling pathways that lead to phosphorylation of CREB.
PKD can also be activated in a PKC-independent manner, involving genotoxic agents (10). Doxorubicin-induced caspase- 3-mediated proteolytic cleavage of PKD1 generates fragments consisting of the N-terminal region and the PH domain together with the kinase domain, the latter fragment possesses full catalytic activity (60). We wanted to explore whether doxorubicin could trigger PKD-mediated CREB phosphorylation. For this purpose, we assayed the phosphoserine 133-CREB levels in extracts of doxorubicin-stimulated A431-PKD1 cells in the absence or presence of Gö6976 or Gö6983. The results show that doxorubicin induces phosphorylation of CREB and ATF-1. Reduced levels of phosphorylated CREB and ATF-1 were observed in both Gö6976-and Gö6983-treated cells. Again, a stronger decrease in CREB and ATF-1 phosphorylation was monitored in the presence of Gö6976 compared with Gö6983 (Fig. 2C). This indicates that PKD is involved in doxorubicin-induced CREB phosphorylation. PKD1 could also phosphorylate CREB at Ser-98 in vitro (see Fig. 1). Phosphospecific antibodies raised against phospho-Ser-98 detected in vitro PKD1-phosphorylated CREB (Fig. 1D), but we were unable to detect phosphorylation of this site in cells treated with several PKD1 activating stimuli (results not shown) (see "Discussion").
PKD1 Phosphorylates CREB at Ser-133 in Vivo-To corroborate the role of PKD1 as a CREB kinase, we examined direct interaction between PKD1 and CREB in vivo by co-immunoprecipitation experiments. HEK293 cells were transfected with pCMV-CREB and either empty vector pcDNA3.1 or an expression vector encoding a constitutive active (ca) PKD1 mutant, and immunoprecipitation was performed using antibodies against CREB. As shown in Fig. 3A, left panel, caPKD1 co-immunoprecipitates with CREB. The precipitation is specific, as the immunoprecipitation of empty vector-transfected cells resulted in no band. As an additional test for specificity of the PKD1-CREB interaction, we tested whether a non-CREB kinase co-immunoprecipitated under these experimental conditions. For this purpose, we chose the mitogenactivated protein kinase MK5. This protein does not phosphorylate MK5 in vitro, nor does it interact with CREB in a mammalian two-hybrid assay. In addition, MK5 did not stimulate CREB-mediated transcription (results not shown). We overexpressed pCMV-CREB and either EGFP vector or EGFP-MK5L337, a constitutive active kinase (48). As shown in Fig.  3A, right panel, MK5 does not coimmunoprecipitate with CREB, suggesting that the co-immunoprecipitation obtained with PKD-CREB is not only because of overexpression. The interaction was also detected between endogenous CREB and exogenous PKD in COS-1 cells, and finally the interaction was confirmed by co-immunoprecipitation of endogenous PKD1 and CREB in A431-PKD1 cells (results not shown). We also performed the reciprocal experiment using anti-PKD antibodies to immunoprecipitate, but without success, which may be because of the antibody quality (result not shown).
To test whether activated PKD1 can also phosphorylate CREB in vivo at Ser-133, we co-transfected A431-PKD1 and COS-1 cells with an expression plasmid encoding caPKD1, dnPKD1, or pcDNA3.1 together with either an expression plasmid for GAL4-CREB or for GAL4-CREBS133A fusion protein.
Western blot analysis with phospho-Ser-133-specific CREB antibody performed on extracts of transfected cells revealed that ectopic expression of activated PKD1 increased phosphorylation of GAL4-CREB at Ser-133 in both cell lines (Fig. 3B,  lane 3 in left and right panels). No increase in phosphorylation of GAL4-CREB was detected in the presence of empty vector or dnPKD (Fig. 3B, lanes 1 and 2, respectively, in left and right  panels). To ensure that increased phospho-Ser-133 CREB levels in the presence of caPKD1 were not the result of unequal loading, the membranes were stripped and re-probed with antibodies against CREB (Fig. 3B, middle lane in left and right panels). The presence of endogenous or exogenous PKD1 was finally verified by antibodies against PKD1, which confirmed equal exogenous expression of both dnPKD and caPKD. We also included similar experiments where we used phospho-Ser-  1-3 and 4 -6, respectively). The fusion protein contains the DNA binding domain of GAL4 (residues 1-147) fused to full-length wild-type CREB or CREB in which serine-133 is replaced by alanine. Western blot was performed, and the level of phosphorylation was determined using a phosphoserine 133-specific antibody. To verify equal loading and blotting of the samples, the membrane was stripped and re-incubated with anti-CREB antibodies (middle panel). Finally, the expression of PKD was determined by PKD-specific antibodies (lower panel).
98-specific antibody. However, the phospho-Ser-98-specific antibodies were unable to detect phosphorylation of this site in vivo in the presence of ectopically expressed activated PKD1. In conclusion, these experiments confirm the involvement of PKD1 in CREB phosphorylation of Ser-133 in vivo.
Activated PKD1 Enhances CREB-mediated Transcription-CREB is a substrate for a plethora of different kinases, and signal-induced phosphorylation of Ser-133 is necessary but not always sufficient to activate CREB-mediated transcription (reviewed in Refs. 34 and 61). This urged us to investigate whether phosphorylation of CREB by PKD1 increased the transcriptional potentials of CREB. Transient transfection studies performed in SK-N-DZ and A431-PKD1 cells revealed that coexpression of caPKD1, but not dnPKD1, stimulated GAL4-CREB-dependent transcription (Fig. 4A). Empty vector or an expression vector with wild-type PKD1 was unable to stimulate CREB-driven transcription (results not shown). PKD1-induced transcription activation was mediated by CREB and not by the  5 g), and either dnPKD1 (1 g) or caPKD1 (1 g). An empty GAL4 vector (pM) was used as a control. The luciferase activity in cell extracts was determined the day after the transfection. The luciferase activity in cells transfected with pM and dnPKD1 was arbitrary set as 1, and the luciferase activities in the other cells are represented as relative luciferase units (RLU). The results are the average of three independent parallels Ϯ S.D., and the figure is representative for at least three independent experiments. B, SK-N-DZ cells were transfected with the reporter plasmid CRE-luc or pmutCRE-luc (1 g), and expression plasmid encoding dnPKD1 (1 g), caPKD1 (1 g), pSR (1 g), or pSR-C␣ (1 g). The luciferase activity in cells transfected with GAL4-CREB and empty vector was arbitrary set as 1, and the luciferase activities are represented as fold induction. C, COS-1 cells were transfected with the expression plasmid for GAL4 (0.5 g) or GAL4-CREB (0.5 g) and either dnPKD1 (1 g), caPKD1, or PH-KD (1 g). The luciferase activity was determined as described in legend of A. D, HEK293 cells were transfected with the expression plasmid for GAL4-CREB (0.5 g), GAL4-CREM (0.5 g), and GAL4-ATF-1 (0.5 g) and either dnPKD (1 g) or caPKD1 expression plasmid (1 g). The luciferase activity was determined as described in legend of A. MAY 18, 2007 • VOLUME 282 • NUMBER 20 basal transcriptional machinery or the GAL4 moiety because luciferase activities in the presence of the DNA-binding domain of GAL4 were comparable whether or not activated PKD1 was co-expressed (Fig. 4A, bars labeled pM). The induction of CREB-mediated transcription was Ser-133-dependent, because caPKD1 failed to stimulate GAL4-CREBS133A-driven transcription. Similar results were obtained with co-transfection studies in COS-1 cells (results not shown), indicating that PKD1-induced CREB activation is not cell-specific. To evaluate the amplitude of caPKD1-induced increase in CREB transactivation potential, we included an experiment where we compared the transactivation potential of caPKD1 to the C␣-subunit of protein kinase A (PKA), the first identified CREB kinase (35,62). As shown in Fig. 4B, caPKD1 induced CREB transactivation potential to comparable levels as PKA. Ectopic expression of the PH kinase fragment of PKD1 also stimulated GAL4-CREB-mediated transcription (Fig. 4C).

PKD1 and CREB
CREB is a member of a family of proteins that also consists of CREM (cAMP-response element modulator) and ATF-1 (activating transcription factor). CREB and ATF-1 are expressed ubiquitously, whereas CREM has a more restricted expression pattern. All the members of the CREB family have similar functional domains and can be regulated by phosphorylation (61). Moreover, the phospho-Ser-133 antibodies that cross-react with the corresponding phospho-Ser-63 in ATF-1 also detected the phosphorylated form of ATF-1 in cells exposed to the PKD activators TPA and doxorubicin (Fig. 2). Therefore, we wanted to investigate whether PKD1 could transactivate the other CREB family members. HEK293 cells were transfected with dnPKD or caPKD and GAL4-CREB, GAL4-CREM, or GAL4-ATF-1. The transcriptional activity of all the tested CREB family members was induced in the presence of caPKD, albeit with different amplitude, CREB being most responsive to caPKD1 (Fig. 4D).
From these experiments we conclude that constitutively activated PKD1, as well as the kinase-active PH-KD fragment, can transactivate GAL4-CREB in a Ser-133-dependent manner. Moreover, caPKD1 can enhance the transcriptional activities of GAL4-CREM and GAL4-ATF-1.
PKD1 Induces Expression of CRE-driven Promoter-The CREB family of transcription factors regulates promoters containing an 8-bp palindromic sequence known as the cAMPresponsive element (CRE) (61). We wanted to explore whether PKD1 could stimulate the activity of a CRE-driven promoter. SK-N-DZ cells were transfected with a reporter plasmid under the control of a cAMP-responsive promoter (pCRE-luc). The same reporter plasmid but with mutations in the CRE (pmut-CRE-luc), which prevent CREB binding (49), was included as control for the specificity of caPKD1-induced transcription. As depicted in Fig. 5, co-expression of caPKD1 and CRE-luc resulted in a moderate (2.5-fold) but reproducible transactivation of the reporter. The transactivation was dependent on a functional CRE element, as pmutCREluc showed no significant increase in trans-activity by co-expression of caPKD1 compared with dnPKD1. Co-expression of the catalytic subunit of PKA (C␣) resulted in a 3.5-fold increase in luciferase activity compared with the activity measured in cells co-transfected with empty vector (pSR). These results suggest that PKD1 can induce a cAMP-responsive promoter by modulating the activity of CRE-binding proteins.
Activated PKD1 Induces Transcription of Nur77-Finally, we wanted to investigate whether PKD1 could stimulate transcription of cellular genes through direct phosphorylation of CREB. For this purpose, we ectopically expressed caPKD and monitored the expression of the CREB-responsive gene Nur77 (63). High transfection efficiency is favored in order to evaluate the effect of transient ectopic expression of PKD1 on transcription of endogenous genes. Therefore, we transfected the easily transfectable HEK293 cells with either caPKD, the dominant negative CREB mutants A-CREB or K-CREB, or combinations of these. A-CREB is a nonphosphorylatable mutant where Ser-133 is mutated into alanine, whereas K-CREB is a non-DNA binding CREB mutant that can dimerize with endogenous wildtype CREB and inhibit its binding to CRE (61,64). TPA is well known PKD activator, and TPA stimulation results in CREB Ser-133 phosphorylation in several cells (34). Untreated or TPA-treated HEK293 cells were therefore included as controls. Total RNA was isolated, and a semi-quantitative RT-PCR was performed to determine the mRNA levels of Nur77. As shown in Fig. 6, TPA and the presence of caPKD1 clearly augmented Nur77 transcript levels. The PKD1-induced increase of Nur77 expression is mediated through CREB, as co-transfection with the dominant negative A-CREB or K-CREB resulted in reduced PKD1-induced Nur77 expression (Fig. 6). RT-PCR with APRT primers was included as a control and shows that the RNA quality and quantity was the same among the different samples. In accordance, caPKD1 also stimulated transcription of the ICER gene in a CREB-dependent manner (result not shown). These findings underscore that PKD1 can stimulate gene expression through direct activation of the transcription factor CREB.

DISCUSSION
PKD is a family of protein kinases involved in several processes, including protein transport, motility, apoptosis, prolif- eration, and gene expression (4 -8). The only known contribution of PKD to gene expression relies on PKD-mediated phosphorylation of HDAC5 and HDAC7, resulting in their nuclear export and release of transcriptional repression (26 -29). PKD has also been shown to induce expression of 11␤-hydroxylase and aldosterone synthase reporters, but the mechanism was not investigated (65). In this study, we have found that CREB represents a novel substrate for PKD1. CREB is an inducible transcription factor whose activity is regulated by phosphorylation (35). Our results show that direct phosphorylation of CREB at Ser-133 by PKD1 enhances the transcriptional activity of CREB and stimulates the expression of CREB-responsive genes, illustrating a new mechanism by which PKD can affect cellular gene expression.
Our in vitro kinase assay revealed Ser-98 as an additional PKD phosphoacceptor site (Fig. 1C). Calmodulin kinase II and PKA have been shown previously to phosphorylate Ser-98 in vitro (66). PKA-mediated Ser-98 phosphorylation required relatively high concentrations of enzyme (1 g of PKA per 3 g of CREB), as determined by titration studies, and substituting Ser-98 by Ala had no effect on the PKA-induced transcriptional activity of CREB (67). In our study, we used 500 ng of PKD per 5 g of GST-CREB, meaning a ratio as high as 1:10. The sequence surrounding Ser-98 is LKRLFSG, which fits well the PKD1 consensus motif LXRXX(T/S). In contrast, the PKA consensus site is RRPSYR (68,69) and displays less similarity with the sequence surrounding Ser-98, thus making Ser-98 a more plausible phosphoacceptor site for PKD1 in vivo. However, our phospho-Ser-98-specific antibodies detected in vitro phosphorylated GST-CREB, confirming their specificity, but we were not able to detect in vivo phosphorylation of this residue even under different experimental conditions such as time-and dosedependent studies with TPA or co-expression with caPKD1 (results not shown). Moreover, transient transfection studies with GAL4-CREB or GAL4-CREBS98A revealed no significant difference in transactivation potential when caPKD1 was coexpressed. Therefore, we conclude that the Ser-98 phosphorylation shown in Fig. 1

occurs in vitro but not in vivo.
PKC has been shown to phosphorylate CREB in vitro (54,55,70), and several studies have shown the involvement of the PKC pathway in TPA-induced CREB phosphorylation (34,61,71,72). So far, no studies have addressed whether PKC is a genuine CREB kinase in vivo. The finding that PKD can be activated in parallel with or downstream of PKC raises the possibility that some cellular responses that arise from PKC activation may be ultimately mediated by PKD (7). PKD research is hampered by the lack of specific activators and chemical inhibitors. However, to examine whether PKD is involved in stimulus-induced activation of CREB, we monitored the CREB Ser-133 phosphorylation status in cells treated with known, but unspecific, PKD activators. We found that stimulation of cells with PDGF, TPA, doxorubicin, and pervanadate resulted in increased CREB Ser-133 phosphorylation. For the three former stimuli, the differential inhibitory specificity of Gö6976 (PKC␣ and -␤ and PKD) and Gö6983 (PKC␣, -␤, -␥, -␦, or -) indicated the involvement of PKD as well as PKC␣, -␤, -␥, -␦, or -in PDGF, TPA-, and doxorubicin-induced CREB Ser-133 phosphorylation. To further explore the role of PKD1 in PDGF-, TPA-, and pervanadate-induced CREB phosphorylation, we initiated siRNA studies against PKD1 in A431-PKD1 cells. We were not able to see any significant reduction of stimuli-induced CREB phosphorylation in the cells transfected with siRNA against PKD1, despite an almost complete depletion of PKD1 expression (results not shown). The inability of siRNA directed against PKD1 to prevent PDGF-, TPA-, and pervanadate-induced CREB phosphorylation could be the result of the engagement of different signaling pathways converging to CREB or of functional redundancy of the PKD isoforms. Stimulus-induced PKD activation depends on activation of PKC, and different PKC isoforms can mediate CREB phosphorylation independently of PKD (e.g. see results with Gö6983). Hence, reducing PKD expression by siRNA will not interfere with PKC-triggered phosphorylation of CREB. Moreover, we found that the presence of H89 reduced TPA-induced CREB phosphorylation in SK-N-DZ cells. 4 Previous studies have shown that TPA can also activate the CREB kinase Msk1 (73) and that the PKA inhibitor H89 inhibits Msk1 as well (74). Hence, TPA-induced CREB phosphorylation may involve both PKD and Msk1, and ablation of PKD1 expression by siRNA will not prevent Msk1 to phosphorylate CREB in response to TPA. Moreover, PKD1, PKD2, and PKD3 are regulated in similar ways and can functionally replace each other. Indeed, a recent study using PKD1-, PKD3-, and PKD1/3-deficient DT40B cells revealed a functional redundancy between PKD1 and PKD3 in phosphorylation of HDAC5. Exogenous expression of the third isoform, PKD2, which is normally not expressed in DT40B cells, also rescued normal FIGURE 6. caPKD induces expression of the endogenous Nur77 gene in a CREB-dependent manner. HEK293 cells were transfected with empty vector (2 g) or expression vectors for caPKD (2 g) and/or the dominant negative CREB variants K-CREB (2 g) or A-CREB (2 g). As control, HEK293 cells were stimulated with 100 nM TPA for 3 h or left untreated. Twenty four hours after transfection, total RNA was isolated and reverse-transcribed to cDNA. PCR was performed with specific primers against Nur77 or the housekeeping gene APRT. The PCR products were analyzed on an agarose gel in the presence of ethidium bromide. The intensity of each band was determined by densitometry, and the ratio of the band intensities of Nur77:APRT in the control cells (untreated) was arbitrary set as 100%. The ratios obtained for the other samples were related to the value obtained for the control cells. The values shown in the lower part of the figure represent the average (ϮS.D.) of two independent experiments. HDAC5 phosphorylation in the PKD null cells (28). Similar functional redundancy may also explain that siRNA against PKD1 only resulted in 50% nuclear exclusion of HDAC5 in phenylephrine-treated NRMV cells (14). Thus, siRNA-mediated down-regulation of PKD1 in PDGF/TPA/pervanadatetreated cells will not influence the PKD2 and PKD3 levels, which can still mediate CREB Ser-133 phosphorylation. In conclusion, because of the lack of specific inhibitors/activator of PKD and possible PKD redundancy, we cannot exclude the contribution of other kinases in PDGF-, TPA-, pervanadate-, or doxorubicin-induced CREB phosphorylation.
In vivo CREB binding has been demonstrated for up to 4,000 different promoters of genes encoding proteins involved in transcription, metabolism, cell proliferation, differentiation, apoptosis, and the secretory pathway (75). Interestingly, PKD also contributes to these processes, implying that PKD may exert some of its function by inducing the expression of specific genes through modulating the activity of CREB. TPA-induced expression of Nur77 is mediated by PKD (26,28,29). Several studies have proven that activated PKD elicits nuclear exclusion of HDAC5 and HDAC7 (26,27,29), revealing a mechanism for the transactivation of Nur77 gene expression by PKD (26,27,29). Our study clearly reveals an additional mechanism for PKD-induced expression of Nur77, involving PKD-mediated phosphorylation of the transcription factor CREB. Indeed, the involvement of CREB in TPA-induced Nur77 expression is also seen in HeLa cells (73), and Nur77 is also up-regulated in PC12 cells in the presence of the constitutive active VP16-CREB (63). These findings support the notion that PKD can exert a dual role in transcriptional regulation as follows: PKDmediated translocation of phosphorylated HDACs and/or PKD-mediated transactivation of CREB-responsive promoters.
Another possible link between PKD and CREB may be in the onset of pathological cardiac hypertrophy. A recent study using transgenic mice with constitutive active PKD1, under the control of a heart-specific promoter, confirmed a role of PKD in this pathological condition. Pathological cardiac hypertrophy is associated with reactivation of the fetal gene program, and the hearts of the transgenic mice revealed increased expression of atrial natriuretic factor, brain natriuretic peptide, ␣-skeletal actin, and ␤-myosin heavy chain (14). PKD1 has been shown to mediate agonist-dependent cardiac hypertrophy through nuclear export of HDAC5 (14,27). Interestingly, both atrial natriuretic factor and brain natriuretic peptide have a CRE site (75) and may be CREB-responsive genes. Moreover, perturbed expression of the CREB-responsive c-fos gene has also been associated with cardiac hypertrophy (76,77). Thus, the putative role of PKD as CREB kinase may also be involved in cardiac hypertrophy when PKD is aberrantly expressed. Another plausible biological connection between PKD and CREB may be as a cellular response to oxidative stress. In this study, we found that pervanadate resulted in phosphorylation of CREB. Previous studies by others showed that oxidative stress induced by H 2 O 2 activates PKD via a PKC-and Src-dependent pathway and results in the nuclear transport of PKD (78,79). Moreover, as H 2 O 2 can induce phosphorylation of CREB in several cell lines (34), this may suggest a possible role of PKD as a CREB kinase in response to oxidative stress. CREB is also involved in differen-tiation of oligodendrocytes, which was shown to require PKCdependent CREB phosphorylation (71). The actual CREB kinase in this study was not investigated but could involve PKD as TPA was used as a stimulus.
In conclusion, our results identified CREB as a new PKD substrate. Moreover, the CREB-PKD connection may form a biological relevant link in physiological processes where both proteins have been shown to be involved.