Cyclic AMP- and Cyclic GMP-dependent Protein Kinases Differ in Their Regulation of Cyclic AMP Response Element-dependent Gene Transcription*

The ability of cGMP-dependent protein kinases (cGKs) to activate cAMP response element (CRE)-dependent gene transcription was compared with that of cAMP-dependent protein kinases (cAKs). Although both the type Iβ cGMP-dependent protein kinase (cGKIβ) and the type II cAMP-dependent protein kinase (cAKII) phosphorylated the cytoplasmic substrate VASP (vasodilator- and A kinase-stimulatedphosphoprotein) to a similar extent, cyclic nucleotide regulation of CRE-dependent transcription was at least 10-fold higher in cAKII-transfected cells than in cGKIβ-transfected cells. Overexpression of each kinase in mammalian cells resulted in a cytoplasmic localization of the unactivated enzyme. As reported previously, the catalytic (C) subunit of cAKII translocated to the nucleus following activation by 8-bromo-cyclic AMP. However, cGKIβ did not translocate to the nucleus upon activation by 8-bromo-cyclic GMP. Replacement of an autophosphorylated serine (Ser79) of cGKIβ with an aspartic acid resulted in a mutant kinase with constitutive kinase activity in vitroand in vivo. The cGKIβS79D mutant localized to the cytoplasm and was only a weak activator of CRE-dependent gene transcription. However, an amino-terminal deletion mutant of cGKIβ was found in the nucleus as well as the cytoplasm and was a strong activator of CRE-dependent gene transcription. These data suggest that the inability of cGKs to translocate to the nucleus is responsible for the differential ability of cAKs and cGKs to activate CRE-dependent gene transcription and that nuclear redistribution of cGKs is not required for NO/cGMP regulation of gene transcription.

The cyclic nucleotides, cAMP and cGMP, are intracellular second messengers mediating the actions of a large number of hormones and neurotransmitters. These cyclic nucleotides act to allosterically regulate the action of a small number of important proteins. Unlike cAMP, which acts mainly through cAMP-dependent protein kinases (cAKs), 1 cGMP is able to ac-tivate three classes of proteins: ion channels, phosphodiesterases, and cGMP-dependent protein kinases (cGKs). The cAKs and the cGKs are highly homologous protein kinase families with similar substrate specificities. Phosphorylation of cellular proteins by both families of kinases leads to alterations in calcium mobilization, protein phosphatase activity, ion channel function, gene transcription, smooth muscle contractility, and platelet aggregation (1)(2)(3)(4).
The cGKs are classified into two types based on their historical order of characterization. The type I enzymes (cGKIs) are highly expressed in lung (5), cerebellum (6), platelets (7), and smooth muscle (8). Two type I isoforms, I␣ and I␤, arise from the alternative splicing of a single gene (9 -12). These two forms differ in their amino-terminal autoinhibitory domains but share the same cGMP-binding sites and catalytic domains (4,13). In Purkinje cells (6), smooth muscle cells (14,15), monocytes (16), and neutrophils (17,18), the majority of the cGKI immunoreactivity is soluble and localized to the cytoplasm. A second type of cGK, termed the type II cGK (cGKII), is highly expressed in intestinal microvilli (19) and is encoded by a gene distinct from that encoding cGKI proteins (20,21). While cGKI isoforms are soluble proteins, cGKII is particulate and associated with cellular membranes (19). Both types possess amino-terminal leucine zipper motifs and exist as homodimers in native tissues (13). In contrast to the cAKs, which have separate catalytic and regulatory subunits, each monomer of the cGKs consists of both a regulatory domain and a catalytic domain contained in the same polypeptide (13).
Many mammalian tissues coexpress isoforms of cAK and cGK, where the cAK and cGK proteins are thought to play distinct roles in cellular regulation. The in vitro substrate specificities of the cAKs and the cGKs are very similar, although a number of proteins have been identified as specific substrates for either cAKs or cGKs. For example, the type I regulatory (R) subunit of cAK (22), G-substrate (23), histone H2B (24), and the bovine lung cGMP-binding cGMP-specific phosphodiesterase (25,26) are specific substrates of cGKI in vitro, while the cAMP response element-binding protein (CREB) has been shown to be a specific in vitro substrate of cAK (76). Although in vitro substrate specificity may be an important indicator of in vivo substrate specificity, recent evidence suggests that colocalization of kinase and substrate in the cell is at least an equally important factor (27).
The cAMP signaling pathway is used to regulate the tran-scription of many genes and involves the phosphorylation of specific transcription factors by the C subunit of cAK (28). In the absence of cAMP, cAK exists predominantly as an inactive tetrameric holoenzyme composed of two R subunits and two C subunits. The inactive holoenzyme may be localized diffusely in the cytoplasm or localized to specific subcellular compartments by interaction of the R subunits with protein kinase A anchoring proteins (27). Upon cytoplasmic elevation of cAMP, cAMP binds to each R subunit, causing the holoenzyme complex to dissociate into a homodimer of R subunits and two catalytically active C subunits. Once released, the C subunit can phosphorylate cytoplasmic substrates, and because of its small size (40 kDa), it can also passively diffuse through the nuclear pores into the nucleus (29). In the nucleus, the C subunit can phosphorylate nuclear transcription factors, such as members of the CREB/ATF family, which bind directly to specific enhancer sequences and alter levels of gene transcription (30,31). Substrates of cAK therefore include both cytoplasmic and nuclear proteins.
Like the cAKs, the cGKs are capable of activating gene transcription in a cyclic nucleotide-dependent manner (32)(33)(34)(35). For example, in PC-12 cells, stimulation of the NO/cGMP pathway leads to increased expression of the immediate early genes c-fos and junB. Importantly, this induction can be blocked by the selective cGK inhibitor KT5823, suggesting that this induction is dependent on cGK activity (35).
While the mechanisms by which the cAKs activate gene transcription have been well characterized, little is known about cGK regulation of gene transcription. In this report, we demonstrate that both transiently expressed and endogenous cGKs are localized to the cytoplasm in mammalian cells. Treatment of cells with cyclic nucleotide does not cause nuclear accumulation of the cGKs, while the C subunit of cAKs does translocate to the nucleus. As a result of its cytoplasmic restriction, cGKI␤ is a weaker activator of CRE-dependent gene transcription than cAKII. These data suggest that regulation of gene transcription by cGK involves phosphorylation of cytoplasmic substrate protein(s) that remain to be characterized.

MATERIALS AND METHODS
Screening of Mouse Brain cDNA Library for Full-length Murine cGKI␤ cDNA-cDNA library screening was performed essentially as described previously (20). A 1.0-kilobase pair EcoRI-SalI restriction fragment from pCGKI.6 was isolated and labeled by random primer extension in the presence of [␣-32 P]dATP. The resulting radiolabeled DNA fragments were used to screen a mouse brain cDNA library. Clones mcGKI␤3D and mcGKI2.2 were isolated, and their inserts were restriction-mapped. mcGKI␤3D, which contains the entire open reading frame (ORF) of murine cGKI␤, was fully sequenced in both directions using Sequenase DNA polymerase (U.S. Biochemical Corp.). mcGKI2.2, which lacks the first 1130 bp of the cGKI␤ ORF, was partially sequenced to confirm a 2-bp deletion in the ORF of mcGKI␤3D. The murine cGKI␤ cDNA sequence derived from the sequencing of both clones has been submitted to the GenBank TM data base. Sequence analyses were performed using DNASTAR software.
Construction of Murine cGKI␤ Mammalian Expression Vector-The pCMV.mcGKI␤ mammalian expression vector was constructed by the polymerase chain reaction (PCR) method using the oligonucleotides 5Ј-GGA GAT CTC CAC CAT GGG CAC CCT GCG GGA TTT AC-3Ј and 5Ј-GGA TCC AAG CTT ACA TTA GAA GTC TAT GTC-3Ј with cDNA clone mcGKI␤3D as a template. The resulting 2100-bp PCR fragment containing the coding region of murine cGKI␤ flanked by a BglII site and a BamHI site was digested with BglII and BamHI and ligated into the BglII site of pCMV.Neo (36) to create pCMV.mcGKI␤. During sequencing of cDNA clone mcGKI␤3D, a two-nucleotide deletion ( Fig. 1) was discovered in the ORF sequence. To correct the deletion in pCM-V.mcGKI␤, a 600-bp BsrGI-AflII restriction fragment from cDNA clone mcGKI2.2, containing the correct sequence, was ligated into BsrGI-and AflII-digested pCMV.mcGKI␤. pCMV.mcGKI␤ was sequenced to confirm the entire coding region sequence.

Transient Transfection of Mammalian Cells and in Vitro Kinase
Activity Determination-10-cm plates of CV-1 or HEK293 cells were transiently transfected using a calcium phosphate coprecipitation method (37). 48 h after application of DNA precipitates, plates were washed twice with ice-cold phosphate-buffered saline (PBS). Following the addition of 200 l of homogenization buffer (10 mM sodium phosphate (pH 7.0), 1 mM EDTA, 1 mM dithiothreitol, 250 mM sucrose) containing 1 mM phenylmethylsulfonyl fluoride, 1 g/ml pepstatin A, and 1 g/ml leupeptin (Sigma), cells were scraped into separate tubes and sonicated twice for 10 s. For kinase activity determinations, cyclic nucleotide (50 M) was added to or omitted from separate tubes containing a phosphotransferase assay mixture consisting of 20 mM Tris (pH 7.5), 10 mM MgAc, 500 M IBMX, 200 M ATP, 11 nM [␥-32 P]ATP (ICN) (specific activity ϭ 200 -300 cpm/pmol), 10 mM NaF, 10 mM dithiothreitol, and the synthetic phosphate acceptor peptide Kemptide (LRRASLG; 150 M) or H2Btide (RKRSRAE; 110 M). When assaying cGKI␤ activity, protein kinase inhibitor (PKI) peptide (1 M) was added to all tubes. The assay was initiated by the addition of cell extracts (0.2 mg/ml), and the phosphotransfer reaction was allowed to proceed for 10 min at 30°C. The assay was terminated by spotting aliquots onto P81 phosphocellulose (Whatman). The P81 phosphocellulose was washed in 10 mM phosphoric acid and counted.
pCMV.FC/CD, the expression plasmid for the chimeric cGKI deletion mutant, was constructed by replacing the amino-terminal regulatory domain (amino acids 1-355) of cGKI␤ with an amino-terminal Flag epitope (DYKDDDDK) followed by the amino-terminal (amino acids 1-20) of murine C␣ (39). The two-step PCR method was used to create this chimeric cDNA. Initially, a C␣ PCR fragment containing a BamHI site and an amino-terminal Flag epitope was generated using the primers 5Ј-GGG GGA TCC ACC ATG GAC TAC AAG GAC GAC GAT GAC AAG GGC AAC GCC GCG GCC GCC AAG AA-3Ј and 5Ј-AAG TAC TCC GGA GTC CCA C-3Ј with pGEM-4.C␣ (40) as a template. The resulting PCR fragment was ligated into pGEM-T (Promega) to create pGEM-T.Flag-C␣1. To generate the cAK/cGK chimera, PCR fragments coding for the Flag-tagged amino terminus of murine C␣ and the carboxyl terminus of murine cGKI␤ were amplified in separate PCRs. A PCR fragment coding for the amino terminus of murine C␣ was generated using primer 5Ј-CAG GAA ACA GCT ATG AC-3Ј and the chimeric primer 5Ј-GGC TTC ATA TTT TGC TTT TGC TAG GAA CTC TTT CAC GCT-3Ј with pGEM-T.Flag-C␣1 as a template. A PCR fragment coding for the carboxyl terminus of murine cGKI␤ was generated using primer 5Ј-AGG CAC GCT TCC ATC AAC-3Ј and the chimeric primer 5Ј-AGC GTG AAA GAG TTC CTA GCA AAA GCA AAA TAT GAA GCC-3Ј with pCMV.mcGKI␤ as a template. The partially overlapping PCR fragments were isolated, combined with flanking primers 5Ј-GGG GGA TCC ACC ATG GAC TAC AAG GAC GAC GAT GAC AAG GGC AAC GCC GCG GCC GCC AAG AA-3Ј and 5Ј-AGG CAC GCT TCC ATC AAC-3Ј, and amplified. The resulting fragment was cut with BamHI and BsrGI, isolated, and ligated into the BglII-and BsrGI-digested pCMV.mcGKI␤ to create pCMV.FC/CD. pCMV.FC/CD was restrictionmapped and sequenced to confirm the amplified sequence.
Construction of Murine cGKII and cGKIIG2A Mammalian Expression Vectors and Generation of Polyclonal Antibodies to Murine cGKII-pCM-V.mcGKII and pCMV.mcGKIIG2A were constructed by PCR. PCR fragments were generated using the following oligonucleotides as forward primers: 5Ј-GAG ATC TGC TAG CCC ACC ATG GGA AAT GGT TCA GTG-3Ј (cGKII) and 5Ј-GAG ATC TGC TAG CCC ACC ATG GCA AAT GGT TCA GTG-3Ј (cGKIIG2A). The oligonucleotide 5Ј-CTC TAT CGA GGG CCC AAG-3Ј was used as the reverse primer, and pCMV.His 6 cGKII (41) was used as the template in the PCRs to amplify 780-bp DNA fragments coding for the amino termini of mcGKII and mcGKIIG2A. pCMV.His 6 cGKII was digested with BglII. The full-length cGKII insert was subcloned into the BglII site of pSP73 (Promega) to create pSP73.His 6 cGKII. The PCR fragments described above were digested with BglII and NsiI, isolated, and ligated individually into pSP73.His 6 cGKII, which had been fully digested with NsiI and partially digested with BglII, generating pSP73.mcGKII and pSP73.cGKIIG2A. pSP73.mcGKII and pSP73.cGKIIG2A were individually cut with BglII, and 2.4-kilobase pair fragments were isolated and ligated into BglII-cut pCMV-.Neo to generate pCMV.mcGKII and pCMV.mcGKIIG2A, respectively. Both pCMV.mcGKII and pCMV.mcGKIIG2A were restriction-mapped, and the amplified regions were fully sequenced. His 6 cGKII was expressed in Spodoptera frugiperda (Sf9) cells and purified as described previously (41). Purified His 6 cGKII was used to immunize rabbits for polyclonal antibody production (Research Genetics Inc.).
Luciferase Assays-CV-1 and HEK293 cells were grown separately on 10-cm plates to 30 and 50% confluency, respectively, and transfected using a standard calcium phosphate method (37) with 0.5 g of the cAMP-responsive reporter construct human chorionic gonadotropinluceriferase (HCG-luciferase) (42) as well as the indicated amounts of pRSV.␤gal and expression vectors. The total amount of plasmid DNA was brought to 30 g with the parental vector pCMV.Neo. 24 h after transfection, cells were incubated for an additional 24 h in Dulbecco's modified Eagle's medium (DMEM) with or without added cyclic nucleotides. Following treatment, cells were washed twice with ice-cold PBS, scraped into homogenization buffer, sonicated, and assayed for both luciferase and ␤-galactosidase activities as described (43).
Identification of a cDNA Clone Encoding Murine VASP, Construction of a Flag-tagged VASP Mammalian Expression Vector, and Western Blot Analysis-A full-length cDNA sequence coding for murine VASP (I.M.A.G.E. Consortium clone identification no. 354921/GenBank TM accession no. W45954) (44) was identified in a search of the expressed sequence tag data base for protein sequences homologous to human VASP using the BLAST algorithm (45). This I.M.A.G.E. Consortium (LLNL) cDNA clone was obtained from Genome Systems Inc. It was sequenced in both directions using Sequenase DNA polymerase (U.S. Biochemical), and sequence analyses were performed using DNASTAR software.
The nucleotide sequence of the murine cDNA derived from a mouse embryo (embryonic day 13.5-14.5) cDNA library differs from that of the published murine VASP genomic sequence (46), in that it contains a G instead of an A at nucleotide position 955, a three-nucleotide deletion from position 1192 to 1194, and an A instead of a G at nucleotide position 1867. The first difference changes threonine 209 to an alanine, the second difference eliminates glutamine 292, and the third difference is located in the 3Ј-untranslated region. The murine VASP cDNA sequence has been submitted to the GenBank TM data base.
A mammalian expression plasmid encoding amino-terminal Flagtagged murine VASP protein was constructed using the PCR method. A PCR fragment containing an amino-terminal Flag epitope (DYKD-DDDK) was generated using the primers 5Ј-GGA TCC GGT ACC TCC ACC ATG GAC TAC AAG GAC GAC GAT GAC AAG GGC GGA GGT ATG AGC GAG ACG GTC ATC TGT TCC-3Ј and 5Ј-GGA TCC CTC GAG TCA AGG AGA ACC CCG CTT CCT CAG-3Ј with clone 354921 as a template. The resulting PCR fragment was digested with BamHI, isolated, and ligated into BglII digested pCMV.Neo to create pCMV. Flag-VASP. pCMV.Flag-VASP was restriction-mapped, and the inserted DNA was fully sequenced.
For phosphorylation studies, 10-cm plates of CV-1 cells at 30% confluency were transfected using a calcium phosphate transfection method as described previously (37). Each plate was transfected with 20 g of pCMV.Flag-VASP. 24 h after transfection, the media was changed, and the cells were incubated for 24 h in DMEM without serum. Following treatment with cyclic nucleotides, cells were quickly washed three times with 5 ml of ice-cold PBS, scraped into 400 l of ice-cold homogenization buffer, and sonicated for 10 s. Extracts were quickly diluted with 500 l of RIPA buffer (20 mM sodium phosphate (pH 7.0), 300 mM NaCl, 2% sodium deoxycholate, 2% Triton X-100, 2% SDS, 2 mM EDTA, 2 mM EGTA, 100 mM NaF, 10 mM sodium pyrophosphate) and vortexed. Fractions were denatured in SDS-PAGE buffer at 95°C for 5 min, resolved on 10% SDS-PAGE gels, and transferred to 0.45-m nitrocellulose membranes (BA-85; Schleicher and Schuell). Membranes were blocked for 1 h in TBST (50 mM Tris (pH 7.5), 150 mM NaCl, and 0.05% Tween 20) supplemented with 5% nonfat dried milk and subsequently incubated with a 1:1000 dilution of an anti-Flag epitope antibody (M2) (Eastman Kodak Co.) in TBST supplemented with 5% nonfat dried milk for 1 h. Filters were washed three times for 10 min with TBST and incubated with an 35 S-labeled sheep anti-mouse antibody (0.5 Ci/ml) (Amersham Pharmacia Biotech) in PBS supplemented with 0.5% bovine serum albumin and 0.1% Triton X-100 as the secondary antibody for 1 h. Following the final set of three 10-min washes with TBST, the blots were dried and quantitated. Phosphor-Imager quantitation was performed in a PhosphorImager apparatus and analyzed with ImageQuant software (Molecular Dynamics, Inc., Sunnyvale, CA).
Immunofluorescence Microscopy-CV-1 cells were grown in DMEM containing 10% fetal calf serum in eight-well tissue culture chambers on poly-L-lysine-coated glass slides (Lab-Tek) to 30% confluency. Cells were transiently transfected using a standard calcium phosphate method (37) with 0.1 g of pCMV.mcGKI␤; 0.01 g of pCMV.Flag-C␣3; and 0.04 g of pCMV.RII␣, 0.1 g of pCMV.mcGKI␤S79D, or 0.1 g of pCMV.FC/CD. Total plasmid concentration was maintained at 0.25 g by the addition of the parental vector, pCMV.Neo. Following a 12-h incubation with DNA precipitates, cells were washed once with DMEM containing 10% fetal calf serum and grown for 24 h. Indicated cells were stimulated with 8-Br-cAMP (1 mM) or 8-Br-cGMP (1 mM) and 3-isobutyl-1-methylxanthine (500 M) in DMEM for various times at 37°C. Following stimulation, cells were washed twice with ice-cold PBS and fixed with 4% formaldehyde in PBS for 10 min at room temperature followed by a 1:1 mixture of methanol and acetone for 5 min. After washing three times with PBS, cells were incubated with a rabbit polyclonal antibody generated against the carboxyl-terminal 15 amino acids of cGKI (anti-cGMP-PK CT) (Upstate Biotechnology, Inc.) at a 1:1000 dilution or an anti-Flag epitope antibody (M2) (Eastman Kodak) at a 1:2000 dilution in blocking buffer (PBS supplemented with 1% bovine serum albumin, 3% goat serum, and 0.1% saponin (Sigma)) for 1 h at room temperature. After four washes with wash buffer (PBS supplemented with 0.1% saponin), a 1:1000 dilution of Cy3-F(abЈ) 2 fragment goat anti-rabbit IgG (Jackson) or a 1:3000 dilution of Cy3-F(abЈ) 2 fragment goat anti-mouse IgG (Jackson) was incubated with the cells for 1 h in the dark in blocking buffer. Prior to examination by fluorescence microscopy, cells were washed four times for 2 min in wash buffer and twice for 2 min in PBS.
CV-1 cells transiently transfected with 0.2 g of pCMV.mcGKII or 0.2 g of pCMV.mcGKIIG2A or A7r5 cells were grown, stimulated, fixed, and examined as above. Treated cells were incubated with anti-cGKII serum or anti-cGMP-PK CT at a 1:1000 dilution in blocking buffer for 1 h at room temperature. After four washes with wash buffer, a 1:1000 dilution of Biotin-SP-F(abЈ) 2 fragment goat anti-rabbit IgG (Jackson) was incubated with the cells for 1 h in blocking buffer. After four washes with wash buffer, a 1:1000 dilution of Alexa-488-conjugated streptavidin (Molecular Probes, Inc., Eugene, OR) was incubated with the cells for 1 h in blocking buffer in the dark. Finally, cells were washed four times for 2 min in wash buffer and twice for 2 min in PBS.
In control experiments, to verify the specificity of the anti-cGMP-PK CT antibody for endogenous cGKI in A7r5 cells, the anti-cGMP-PK CT antibody (40 nM) was preincubated for 1 h at room temperature with a peptide coding for the carboxyl-terminal 18 amino acids of cGKI (5 M) (STRESSGEN) prior to incubation with cells.

RESULTS
Cloning and Sequencing of Murine cGKI␤-A single fulllength cGKI␤ cDNA clone was isolated from a mouse brain cDNA library. The cDNA was fully sequenced and shown to contain 2841 bp (Fig. 1). The murine cGKI␤ cDNA contains a short 89-bp 5Ј-untranslated region, an ORF of 2061 nucleotides, and a 691-bp 3Ј-untranslated region. The predicted murine cGKI␤ protein contains 686 amino acids with a calculated molecular mass of 77.8 kDa and shows greater than 99% amino acid identity to human cGKI␤ (9). Despite 139 nucleotide differences in the ORF sequence, there are only three amino acid differences between murine cGKI␤ and human cGKI␤: Glu 242 to Asp, Thr 280 to Gln, and Asn 671 to Ser (Fig. 1).
Effect of cGKI␤ Overexpression on CRE-dependent Gene Transcription-To determine whether cGKI␤ was capable of regulating CRE-dependent gene transcription, we examined the ability of cGKI␤ to transactivate the cAMP-responsive HCG promoter. CV-1 cells ( Fig. 2A) or HEK293 cells (Fig. 2B) were transfected with a constant amount of the cGKI␤ expression vector along with the HCG-luciferase reporter plasmid (42). Control cells were transfected with the HCG-luciferase reporter plasmid alone. Transfected cells were treated with or without 8-Br-cGMP (1 mM) for 20 h. CV-1 and HEK293 cells were chosen for this experiment because regulation of CRE-dependent gene transcription by cAK has been characterized by transfection experiments previously in these cell lines (36,43,47). In both CV-1 and HEK293 cells, transfection of wild type cGKI␤ only minimally stimulated luciferase gene transcription in the absence of cyclic nucleotide treatment (Fig. 2, A and B). CV-1 and HEK293 cells transfected with the HCG-luciferase reporter plasmid alone showed minimal responses to 8-Br-cGMP treatment (Fig. 2, A and B). When CV-1 cells overex-pressing cGKI␤ were treated with 8-Br-cGMP (1 mM), a small but reproducible increase in luciferase activity was observed ( Fig. 2A). In HEK293 cells, overexpression of cGKI␤ increased luciferase activity following 8-Br-cGMP treatment 13-fold (Fig.  2B). Thus, the HCG-luciferase reporter plasmid is more sensitive to overexpression of cGKI␤ in HEK293 cells than in CV-1 cells. The greater sensitivity of the luciferase reporter assay in HEK293 cells may be due to a number of factors including the levels of accessory transcription factors (48,49) or cellular phosphatases (50,51). The data from both CV-1 and HEK293 cells indicate that cGKI␤ is capable of inducing CRE-dependent gene transcription in mammalian cell lines.
cGKI␤ Is a Relatively Weak Activator of CRE-dependent Gene Transcription-For comparison of cGKI␤ and cAK activation of CRE-dependent gene transcription, the same HCG-luciferase reporter assay was employed. To generate comparable levels of cGKI␤ and cAKII in transfected cells, the quantity of each expression vector transfected was carefully titrated to produce similar amounts of cGKI␤ and cAKII protein and kinase activity (data not shown). For cGK expression, CV-1 cells were transfected with 10 g of pCMV.mcGKI␤. For cAK expression, cells were transfected with both 1 g of pCMV.Flag-C␣3 and 4 g of pCMV.RII␣ (52). Control cells were transfected with the parental pCMV.Neo vector alone. Extracts from cGKI␤-and cAKII-transfected cells possess similar levels of cGKI␤ and cAKII protein by quantitative Western blotting using purified proteins as standards (data not shown). The extracts from cGKI␤-and cAKII-transfected cells also showed similar specific activities with in vitro kinase assays in which the nonspecific peptide Kemptide (150 M) was used as the phosphoacceptor (extracts from cGKI␤ transfected cells ϭ 0.73 nmol/min/mg and extracts from cAKII transfected cells ϭ 0.84 nmol/min/mg). The maximal cGMP-or cAMP-stimulated protein kinase activities of extracts from cGKI␤-transfected cells and extracts from FIG. 1. Nucleotide and predicted amino acid sequence of murine cGKI␤. The sequence represents a composite of the sequences derived from mcGKI␤3D and mcGKI2.2. Amino acid sequence of murine cGKI␤ inferred from the composite nucleotide sequence is represented below the DNA sequence with the one-letter amino acid codes. Nucleotide numbers are indicated at the left of the sequence and amino acid numbers at the right of the sequence. Amino acid residues from the human cGKI␤ sequence (9) that are divergent from the predicted murine cGKI␤ amino acid sequence are shown individually below the murine cGKI␤ amino acid sequence. The two nucleotides (1672 and 1673) deleted in mcGKI␤3D but present in mcGKI2.2 are underlined. (GenBank accession no. AF084547) cAKII-transfected cells were at least 10-fold higher than endogenous kinase activity in extracts from control CV-1 cells (data not shown).
Once similar levels of kinase activity were obtained, transcriptional regulation was examined (Fig. 2C). In cells expressing cGKI␤ or cAKII, stimulated luciferase activity was halfmaximal at 100 M 8-Br-cGMP and 30 M 8-Br-cAMP, respectively. Cotransfection with cGKI␤ gave a maximal 2-fold stimulation of reporter gene expression over basal levels, whereas cotransfection with cAK gave a maximal 14-fold stimulation (Fig. 2C). cAKII-transfected cells showed significantly higher luciferase activity at all cyclic nucleotide concentrations examined (Fig. 2C). Although CV-1 cells express both endogenous cGKI and cAK, cells transfected with the reporter gene alone showed no response to 8-Br-cGMP treatment but a significant response to 8-Br-cAMP treatment (4-fold) (data not shown). A reduced potency of cGKI␤ relative to cAKII was also observed in HEK293 cells, where cotransfection with cGKI␤ gave a maximal 13-fold induction in comparison with the 250fold induction seen with the C subunit of cAK (data not shown). cAK's maximal induction of luciferase activity was 52-fold higher than cGKI␤'s in CV-1 cells and 25-fold higher than cGKI␤'s in HEK293 cells (Fig. 2D). Our results in both CV-1 and HEK293 cells indicate that when compared with cAK, cGKI␤ is only a weak activator of CRE-dependent gene transcription.
VASP Is Phosphorylated Efficiently by Both cGKI␤ and cAK in Vivo-Since cGKI␤ and cAKII were expressed at equal amounts, but showed a 10-fold difference in their ability to regulate transcription, experiments were designed to compare the ability of cGKI␤ and cAK to phosphorylate a mutually recognized substrate in vivo.
A cytoplasmic substrate for cGK and cAK has been identified in human platelets (53). This 46-kDa protein, termed VASP (vasodilator-and A kinase-stimulated phosphoprotein), has been shown to be widely expressed and localized to cytoplasmic focal adhesions and stress fibers (54). The human VASP protein was demonstrated to be phosphorylated at three sites to varying extents by both cGK and cAK. Phosphorylation of one of these sites, Ser 157 , caused a shift in mobility from 46 to 50 kDa in a SDS-PAGE gel. VASP was chosen as a good substrate for comparing cGK and cAK activity, since Ser 157 is phosphorylated comparably by both cGK and cAK in vivo (55).
A full-length murine VASP cDNA was sequenced (see "Ma- terials and Methods"), and the murine VASP protein predicted by the ORF was 87% identical to the human VASP protein (56). All three human VASP cyclic nucleotide-dependent phosphorylation sites are conserved in murine VASP (46). To differentiate transfected VASP from endogenous VASP, we constructed an amino-terminal Flag-tagged VASP mammalian expression vector, pCMV.Flag-VASP (see "Materials and Methods").
To verify that Flag-tagged murine VASP was expressed and properly localized, cells were transiently transfected with pC-MV.Flag-VASP. Indirect immunofluorescence analysis with the M2 anti-Flag antibody showed that Flag-tagged VASP localizes to the cytoplasm in the presence or absence of 8-Br-cAMP (data not shown). Western blot analysis of cell extracts with the M2 anti-Flag antibody detected a single band with an apparent molecular mass of 46 kDa (Fig. 3A), the size expected for nonphosphorylated VASP. No band was seen in nontransfected CV-1 cell extracts (data not shown).
When Flag-tagged VASP-transfected cells were treated with 8-Br-cAMP, two immunoreactive species were detected on Western blots. The most rapidly migrating species corresponded to the previously observed 46-kDa VASP band, and an additional 50-kDa VASP band corresponded to a phosphorylated form of the VASP protein (Fig. 3A). Phospho-VASP was observed following 5 min of cyclic nucleotide treatment and remained at the maximal level for at least 1 h (Fig. 3A). Formation of the 50-kDa band was specifically due to phosphorylation of Ser 153 (the murine VASP residue equivalent to Ser 157 of human VASP) (56) by endogenous cAK because its detection was completely blocked by cotransfection of an expression vector for mouse PKI protein (data not shown), which specifically inhibits cAKs but not cGKs (57). Formation of the 50-kDa band was also prevented by site-directed mutagenesis of Ser 153 to an alanine (data not shown).
To demonstrate activation of transiently expressed cGKI␤ by 8-Br-cGMP, cells were cotransfected with expression vectors for Flag-tagged VASP, cGKI␤, and PKI. cGKI␤ and endogenous cAK are expressed at similar levels in extracts from CV-1 cells transfected with 2 g of pCMV.mcGKI␤, as determined by in vitro kinase assays (data not shown). Western blot analysis of extracts from untreated cells using an anti-Flag antibody detected two bands, a major 46-kDa band and a minor 50-kDa band (Fig. 3B), indicating that cGKI␤ has significant basal activity toward VASP in vivo. Within 5 min of 8-Br-cGMP treatment, nearly all of the remaining 46-kDa band was converted to the 50-kDa form and remained in that form for at least 1 h (Fig. 3B). To exclude the possibility that 8-Br-cGMP was cross-activating endogenous cAK, we cotransfected sufficient PKI expression vector to completely block cAK activation (data not shown). Coexpression of PKI had no effect on 8-Br-cGMP-dependent increases in VASP phosphorylation (data not shown). These findings suggest that both cGKI␤ and cAK are equally capable of phosphorylating protein substrates in cells treated with membrane soluble cyclic nucleotides. Thus, the lack of significant regulation of CRE-dependent gene transcription by cGKI␤ is not due to a deficit in cGKI␤ activation.
Effect of Cyclic Nucleotides on the Subcellular Distribution of cGKI␤ and the C Subunit of cAKII-Nuclear pore complexes mediate the transport of proteins between the cytoplasm and the nucleus of mammalian cells. These pore complexes allow passive diffusion of small proteins (Ͻ40 -65 kDa) in and out of the nucleus, while larger proteins must be actively transported (58). cAKs consist of separate R subunits and C subunits, while cGKs possess a linker sequence, which connects the regulatory domain and catalytic domain into a single polypeptide chain. Therefore, unlike the cAKs, activation of the cGKs by cGMP does not result in the dissociation of the regulatory and the catalytic components of the enzyme and does not cause the release of a small catalytic subunit (13). The actual size of active cGKs is increased further by dimerization of cGK subunits (13). Hence, while the free C subunit of cAK (40 kDa) is able to passively diffuse into the nucleus to phosphorylate specific transcription factors (29), cGKI␤ (150 kDa) would be expected either to be restricted to the cytoplasm or to require a nuclear localization sequence (NLS) for transport to the nucleus.
Because CREB is primarily a nuclear protein (28), a restricted cytoplasmic localization could explain cGKI␤'s minimal ability to activate CRE-dependent gene transcription. To examine the subcellular localization of cGKI␤, we transiently transfected CV-1 cells with pCMV.mcGKI␤. CV-1 cells express low levels of endogenous cGKI and possess a flat morphology. Anti-cGKI␤ serum, which recognizes the carboxyl-terminal 15 amino acids of murine cGKI (anti-cGMP-PK CT), detected a single major band with an apparent molecular mass of 75 kDa in extracts of CV-1 cells transfected with cGKI␤. The amount of cGKI␤ was more than 10-fold higher than the small amount of endogenous cGKI found in control CV-1 cell extracts (data not shown). Staining of transfected CV-1 cells with the anti-cGMP-PK CT antibody revealed diffuse, cytoplasmic immunofluorescence (Fig. 4A). The diffuse staining showed no specific association with major cellular structures. The addition of 8-Br-cGMP to the medium did not lead to a change in the cytoplasmic staining pattern or an increase in nuclear staining (Fig. 4B). Furthermore, no change in the staining pattern was observed at other times following 8-Br-cGMP treatment (30 min, 2 h, and 4 h; data not shown). Treatment with 8-Br-cGMP had no apparent effect on the intensity of the cGKI␤ staining and suggested that the ability of the antibody to recognize activated cGKI␤ was not altered by cyclic nucleotide treatment. Nontransfected cells showed a low but detectable level of cytoplasmic staining probably due to endogenous cGKI. Treatment with 8-Br-cGMP had no effect on this staining pattern either (data not shown). A similar restriction of cGKI␤ to the cytoplasm was observed in transfected COS-1, HEK293, and baby hamster kidney mammalian cell lines (data not shown).
In contrast, as shown previously (30), the C subunit of cAK was found to quickly diffuse into the nucleus upon treatment of cells with 8-Br-cAMP (Fig. 4, C and D). Thus, unlike the C subunit of cAK, which can translocate into the nucleus upon activation, cGKI␤ is restricted to the cytoplasm independent of its activation state.
In CV-1 cells, immunofluorescence microscopy revealed that both endogenous CREB and a GFP-hCREB chimera protein were targeted to the nucleus. No significant immunofluorescence was discernible in the cytoplasm (data not shown). Hence, the restricted cytoplasmic localization of cGKI␤ is entirely consistent with its being a weak activator of CRE-dependent gene transcription.
cGKII Does Not Translocate to the Nucleus upon Activation by cGMP-To determine the subcellular localization of the cGKII isoform, rabbit antiserum was prepared using fulllength His-tagged cGKII purified from SF9 cells as the antigen (41). Anti-cGKII serum recognized a single band with an apparent molecular mass of 86 kDa in extracts from cGKII-transfected HEK293 cells. No bands were detected in extracts from control or cGKI␤-transfected HEK293 cells. Preimmune serum from the rabbit did not detect bands in any extracts (data not shown).
Immunostaining for transfected wild type murine cGKII in CV-1 cells using cGKII antiserum, revealed a crescent-shaped, perinuclear staining pattern (Fig. 5A). This staining was specific for cGKII, since no significant anti-cGKII immunofluorescence was observed in cells transfected with parental vector alone (data not shown). Treatment with 8-Br-cGMP for 1 h did not affect the staining pattern and did not cause an increase in nuclear staining (Fig. 5B).
To rule out the possibility that cGKII was not translocating to the nucleus because it was strongly membrane-bound, we mutated the penultimate glycine of murine cGKII to an alanine. The cGKIIG2A mutant is a nonmyristoylated and soluble protein that is enzymatically similar to wild type cGKII (59). In contrast to wild type cGKII, the cGKIIG2A mutant was localized diffusely in the cytoplasm (Fig. 5C). As described for wild type cGKII, treatment with 8-Br-cGMP had no effect on the subcellular distribution of the cGKIIG2A protein (Fig. 5D). These data indicate that restriction to the cytoplasm is not cGKI␤-specific but a general property of both cGKI␤ and cGKII.
Localization of Endogenous cGKI in A7r5 Smooth Muscle Cells-Since overexpression of proteins can result in aberrant subcellular localization, several cell lines were screened for detectable expression of cGKI and cGKII. The anti-cGMP-PK CT antibody detected a single cGKI band with an apparent molecular mass of 75 kDa in cell extracts from N1E-115 neuroblastomas, N2A neuroblastomas, NG-108 neuroblastomas, CV-1 cells, HEK293 cells, A10 smooth muscle cells, and A7r5 smooth muscle cells. The highest levels were detected in extracts from NG-108 neuroblastomas and A7r5 smooth muscle cells. No band was detected in extracts from COS-1 cells or Y1 adrenal tumor cells or when identical blots were probed with anti-cGKII serum (data not shown).
Localization of endogenous cGKI in A7r5 smooth muscle cells by indirect immunofluorescence revealed diffuse cytoplasmic staining (Fig. 6A), and treatment with 1 mM 8-Br-cGMP had no effect on this staining pattern (Fig. 6B). The cytoplasmic staining was specific for cGKI, since it was abolished by preabsorption of the anti-cGMP-PK CT antibody with a peptide coding for the carboxyl-terminal 18 amino acids of cGKI (Fig. 6C).
Generation of Full-length Constitutively Active cGKI␤ Mutants-The difference in the ability of cGKs and cAKs to activate CRE-dependent gene transcription could be due to differences in their substrate specificities, their subcellular localizations, or both. To define the mechanism(s) by which cGKs and cAKs differentially regulate gene transcription, a full-length and a small sized constitutively active cGKI␤ mutant were generated. In efforts to produce a full-length constitutively active kinase, three different classes of cGKI␤ point mutants were generated. For the first class of mutant, mutations were made in cGKI␤ analogous to those that have been identified in the C subunit of cAK and result in constitutive activity of the C subunit even in the presence of excess R subunit (60 -64). These residues, His 68 and Trp 197 , are conserved between the C subunit of cAK and the catalytic domain of cGKI. His 419 and Trp 530 of cGKI␤ were mutated to glutamine and arginine, respectively, in accordance with the mutations found for the R subunit-insensitive C subunit mutants (60 -64). A second class of constitutively active mutant was designed based on the observation that regulatory domain arginine residues in both cAK and cGK may mimic substrate site arginines. These pseudosubstrate arginine residues have been shown to be important in the high affinity of several kinases' regulatory domains for their catalytic domains (65). Specifically, proteolytic cleavage of cGKI␤ after its pseudosubstrate arginine (Arg 75 ) generates a kinase with high constitutive activity (66). Thus, to generate the second class of constitutively active mutant, we mutated Arg 75 to a neutral residue, alanine. Finally, for the third class of mutation, it was reported that cGKI␤ undergoes autophosphorylation at two sites in its regulatory domain in the presence of cyclic nucleotides. Phosphorylation of the second site, Ser 79 , leads to a large increase in basal kinase activity (67). For this third class of constitutively active mutant, we mutated Ser 79 to an acidic residue, aspartic acid, to mimic the effect of autophosphorylation. One advantage of the second and third approaches is that mutations in the regulatory domain are less likely to affect the substrate specificity or specific activity of the enzyme.
The effects of these mutations were tested by measuring the basal and cGMP-activated kinase activities in extracts made from HEK293 cells transiently transfected with the mutant expression vectors (Fig. 7). Extracts from HEK293 cells transfected with pCMV.Neo or a mammalian expression vector encoding a catalytically inactive mutant (cGKI␤K404R) showed no significant cGMP-dependent kinase activity (Fig. 7). Except for the cGKI␤H419Q mutant, which shows a 5-fold reduction in specific activity, the cGMP-dependent kinase activities (Fig. 7) and protein expression levels (data not shown) of all of the constitutively active mutants were similar to wild type cGKI␤. Both the cGKI␤R75A and cGKI␤S79D mutants exhibited high basal activity when compared with wild type cGKI␤. The increases in basal kinase activity of the cGKI␤R75A and cGKI␤S79D mutants were 12-and 14-fold, respectively. Specifically, mutation of Ser 79 to an aspartic acid increased the basal kinase activity from 5 to 72% of the total cGMP-dependent kinase activity (Fig. 7). These results support the hypotheses that Arg 75 is important in the high affinity of cGKI␤'s regulatory domain for its catalytic domain (65) and that autophosphorylation of Ser 79 is capable of activating cGKI␤ in the absence of cGMP (67). Of interest, the cGKI␤H419Q and cGKI␤W530R mutants showed only small increases in basal kinase activity (Fig. 7), suggesting that the amino acid residues in the cAK catalytic domain responsible for the tight interaction between cAK's R subunit and C subunit are less important to inhibition of cGKI's catalytic domain by its regulatory domain. The cGKI␤S79D mutant showed the highest basal activity and was therefore employed in further studies of cGKI regulation of transcription.
Generation of a Constitutively Active cGKI Deletion Mutant-An additional mutant was sought to determine the effect of cGKI's protein size on its ability to regulate transcription. Previous studies have shown that truncation of cGKI's autoinhibitory regulatory domain results in a fully active, cGMPindependent catalytic domain that possesses a similar substrate specificity as the cGKI holoenzyme (68). Because of its small size (40 kDa), it was predicted that the catalytic domain of cGKI would be able to passively diffuse into the nucleus and directly phosphorylate Ser 133 of CREB in a manner similar to the C subunit of cAK (29).
The entire amino-terminal regulatory domain of cGKI␤ was deleted, and only the catalytic domain was expressed in HEK293 cells. To increase the stability of the cGKI catalytic domain, we appended the amino terminus of the C subunit of cAK to the cGKI catalytic domain. Previously, the first 23 amino acids of cAK's C subunit were reported to stabilize PKC's catalytic domain and allow for a high level of expression in mammalian cells (69). To assess its stability and the extent of its activation in transfected cells, the basal and cGMP-dependent kinase activities of this chimera were examined in extracts made from transiently transfected HEK293 cells (Fig. 8A). Wild type cGKI␤, the cGKI␤S79D mutant, and the catalytic domain of cGKI were found to be expressed at equivalent levels by Western blot analysis (data not shown) and in vitro kinase assays (Fig. 8A). As expected, the truncated catalytic domain of  -cGMP; b). Cells were fixed as in Fig. 4 and then labeled with anti-cGMP-PK CT antibody as the primary antibody, Biotin-SP-F(abЈ) 2 fragment goat anti-rabbit IgG as the secondary antibody, and Alexa-488-conjugated streptavidin as the fluorescent marker. Cytoplasmic immunofluorescence was abolished by preabsorption of the anti-cGMP-PK CT antibody with a peptide coding for the carboxyl-terminal 18 amino acids of cGKI (Blocked; c).
In Vivo Activity of Constitutively Active cGKI␤ Mutants-To determine if the cGKI␤S79D mutant and the catalytic domain of cGKI have increased basal kinase activity in vivo, Flagtagged VASP was transfected alone or cotransfected with cGKI␤, cGKI␤S79D, or the catalytic domain. The level of VASP phosphorylation on Ser 153 was low in serum-starved CV-1 cells expressing Flag-tagged VASP alone with 3% of the total VASP in the 50-kDa form (Fig. 8, B and C). Expression of the cGKI␤S79D mutant converted 30% of VASP to the 50-kDa form, while the catalytic domain converted 70% of VASP to the 50-kDa form (Fig. 8, B and C). Wild type cGKI␤ stimulated VASP phosphorylation significantly less, with only 6% conversion of VASP to the 50-kDa form (Fig. 8, B and C). Thus, both the cGKI␤S79D mutant and the catalytic domain of cGKI showed constitutive activity in vivo, with the catalytic domain being slightly more active toward VASP than the cGKI␤S79D mutant.
Subcellular Localization of Constitutively Active cGKI␤ Mutants-In agreement with the earlier localization studies, which used 8-Br-cGMP to activate wild type cGKI␤, the subcellular localization of the cGKI␤S79D mutant was cytoplasmic in the absence or presence of 8-Br-cGMP (Fig. 9, A and B). Unlike full-length cGKI␤, the Flag-tagged catalytic domain of cGKI was found both in the nucleus and in the cytoplasm (Fig.  9C). The catalytic domain was evenly distributed between the nucleus and the cytoplasm, consistent with the passive diffusion previously reported for the C subunit of cAK (29). The catalytic domain did not concentrate preferentially in the nucleus (Fig. 9C). As expected, the addition of 8-Br-cGMP to the medium had no discernible effect on its localization (data not shown). The catalytic domain of cGKI was detected both in the cytoplasm and in the nucleus with the M2 anti-Flag antibody (data not shown) and the Anti-cGMP-PK CT antibody (Fig. 9C), demonstrating that the anti-cGMP-PK CT antibody was capable of detecting cGKI in the nucleus. These data corroborate the cytoplasmic restriction of active full-length cGKI␤ and strongly suggest that cGKI␤ is restricted to the cytoplasm because of its large size.
Activation of CRE-dependent Gene Transcription in Cells Overexpressing Constitutively Active cGKI␤ Mutants-To correlate the localization of the various cGKI␤ mutants with their ability to increase CRE-dependent gene transcription, the abilities of wild type cGKI␤, the cGKI␤S79D mutant, and the catalytic domain of cGKI to regulate CRE-dependent gene transcription were measured using the HCG-luciferase reporter. Expression of cGKI␤ or the cGKI␤S79D mutant did not stimulate CRE-dependent transcription in CV-1 cells (Fig. 10A) under conditions where a substantial increase in the phosphorylation level of the cytoplasmic substrate VASP was observed (Fig. 8, B and C). In contrast, the catalytic domain, which could enter the nucleus, activated transcription from the HCG-luciferase reporter construct 4-fold (Fig. 10A). Unlike full-length cGKI␤ and the cGKI␤S79D mutant, the catalytic domain was effective at both elevating the phosphorylation state of VASP (Fig. 8, B and C) and inducing CRE-dependent gene transcrip-  4). 24 h after transfection, cells were serum starved for an additional 24 h. Extracts were generated and blotted as in Fig. 3. C, PhosphorImager quantitation of level of Flag-tagged VASP phosphorylation. Statistical analysis of experiment was as described for B. The percentage of VASP phosphorylation was calculated by dividing the quantity of the 50-kDa band by the sum of the 46-kDa band and the 50-kDa band.  Fig. 4 and then labeled with the anti-cGMP-PK CT antibody as the primary antibody and Cy3-F(abЈ) 2 goat anti-rabbit as the secondary antibody. tion in CV-1 cells (Fig. 10A), suggesting that the ability of the catalytic domain to passively enter the nucleus allows it to be a more efficient activator of CRE-dependent gene transcription.
Regulation of HCG-luciferase activity was also examined in HEK293 cells, because the reporter assay was shown to be more sensitive in this cell line. When assayed for induction of luciferase gene expression upon transient transfection into HEK293 cells, wild type cGKI␤ showed essentially no stimulation of luciferase activity in the absence of nucleotide treatment (Fig. 10B). In contrast, transfection of the cGKI␤S79D mutant elevated the basal luciferase activity 66-fold (Fig. 10B). This induction is consistent with the induction observed in HEK293 cells expressing wild type cGKI␤ and treated with 8-Br-cGMP (13-fold). In HEK293 cells, cGKI␤S79D was significantly less potent as a regulator of gene transcription than the catalytic domain of cGKI (Fig. 10B). Whereas the cGKI␤S79D mutant stimulated CRE-dependent gene transcription 66-fold, the catalytic domain of cGKI induced luciferase activity 1071-fold (Fig. 10B). The C subunit of cAK and the catalytic domain of cGKI activated CRE-dependent gene transcription to a similar extent (Fig. 10C), with the 3-fold difference in activation being primarily due to the low stability of the cGKI catalytic domain compared with the C subunit of cAK (data not shown). As observed in the CV-1 cells, full-length cGKI␤ was cytoplasmic in the presence or absence of 8-Br-cGMP in HEK293 cells (data not shown). Therefore, both the CV-1 and HEK293 cell data strongly support the conclusion that active full-length cGKI␤ is a relatively weak activator of CRE-dependent gene transcription primarily because it is restricted to the cytoplasm.
Characterization of cGKI␤ ATP-binding Domain Mutant-Unlike CV-1 cells, transfection of full-length cGKI␤ into HEK293 cells confers readily measurable CRE-dependent transcriptional responses to 8-Br-cGMP. When cotransfected with HCG-luciferase into HEK293 cells, wild type cGKI␤ can mediate a 13-fold increase in luciferase activity in response to 8-Br-cGMP (Fig. 2B). The magnitude of induction is dependent on the amount of cGKI␤ expression vector transfected (data not shown) and the 8-Br-cGMP concentration in the media (Fig.  11). Although immunofluorescence microscopy localized a majority of the cGKI immunoreactivity to the cytoplasm in cGKI␤transfected HEK293 cells, it was possible a small fraction of the total cGKI␤ was present in the nucleus and responsible for the small increase in transcription (data not shown). Typically, proteins larger than 40 -65 kDa require a basic NLS to be actively transported to the nucleus. Inspection of the cGKI␤ protein sequence by other investigators for a potential NLS revealed a single cluster of basic amino acids in the ATPbinding domain (KILKKRHI; residues 404 -411) (33,34).
To determine whether the induction of CRE-dependent gene transcription by full-length cGKI␤ in HEK293 cells was dependent on the putative NLS sequence, this sequence was altered to the corresponding amino acids from the C subunit of cAK. The amino acids encoding this basic sequence are not highly conserved in the C subunit of cAK (KILDKQKV; resi- dues 72-79) (39). Using site-directed mutagenesis, we changed two of the basic amino acids in cGKI␤'s putative NLS to the equivalent residues in the C subunit of cAK(Lys 407 to Asp and Arg 409 to Gln). These specific amino acid substitutions were chosen because the C subunit of cAK is not actively transported to the nucleus (29), and these mutations were unlikely to perturb the structure of the kinase since they are found in the C subunit of cAK.
To determine the effect of these mutations on the ability of cGKI␤ to activate CRE-dependent gene transcription, we cotransfected wild type cGKI␤ or the cGKI␤K407D/R409Q mutant along with a CRE-dependent reporter construct into HEK293 cells. While cGKI␤ increased luciferase activity 10-fold following 8-Br-cGMP treatment, the activated cGKI␤K407D/R409Q mutant only induced a 4-fold increase in gene transcription (Fig. 12A). Therefore, substitution of two basic amino acids in the putative NLS sequence reduced but did not eliminate cGKI␤'s ability to activate CRE-dependent gene transcription.
Because this basic sequence is located within the ATP-binding region of the catalytic domain, it was possible that the mutations negatively affected cGKI␤'s catalytic activity. To determine the relative effect of the two amino acid substitutions on the catalytic activity of the cGKI␤K407D/R409Q mutant, the basal and cGMP-dependent kinase activities of this mutant were compared with those of wild type cGKI␤ in extracts made from the transiently transfected HEK293 cells. Although expression levels of wild type cGKI␤ and the cGKI␤K407D/K409Q mutant were similar as determined by Western blot analysis (data not shown), cGMP-dependent kinase activity was 60% less in the extract from the cells overexpressing the cGKI␤K407D/R409Q mutant as compared with the extract from the cells overexpressing wild type cGKI␤ (Fig. 12B).
The cGKI␤K407D/R409Q mutant showed a reduction in the phosphorylation of substrates in vivo, as determined by VASP transfection experiments (Fig. 12, C and D). Western blot analysis of extracts from cells transfected with VASP and PKI alone detected a single 46-kDa band (Fig. 12C). When identically transfected cells were treated with 8-Br-cGMP (1 mM) for 1 h, both a 46-kDa band and a 50-kDa band were detected (Fig.  12C). Formation of the 50-kDa phosphorylated VASP band with 8-Br-cGMP treatment was due to phosphorylation of Ser 153 by endogenous cGKI, since its formation was not blocked by coexpression of PKI, although PKI did completely block 8-Br-cAMP induction of the 50-kDa band (data not shown). While 94% of Flag-tagged VASP was converted to the 50-kDa form in cells expressing wild type cGKI␤ treated with 8-Br-cGMP, only 56% was converted in similarly treated cells expressing cGKI␤K407D/R409Q (Fig. 12, C and D). Hence, the cGKI␤K407D/R409Q mutant has significantly reduced basal and cGMP-stimulated kinase activity in vivo. These data suggest that mutation of basic residues in the putative NLS decreases the ability of cGKI␤ to activate gene transcription, not because it prevents localization of cGKI␤ in the nucleus but because it decreases the catalytic activity of the kinase. Thus, mutagenesis of the putative NLS of cGKI␤ did not specifically alter either cGKI␤'s localization or its ability to regulate transcription. DISCUSSION cAK and cGK regulation of CRE-dependent gene transcription was investigated in this study to determine if these two kinases differentially regulate this process. Because of the large size of active cGKs, we sought to determine whether, like the C subunits of cAKs, cGKs would translocate to the nucleus following activation by cyclic nucleotides. Analyses of the subcellular localization of endogenous cGKI in A7r5 cells as well as transfected cGKI␤ and cGKII indicate that the cGKs localize to the cytoplasm regardless of their activation state. The inability of cGKI␤ to translocate to the nucleus and directly phosphorylate CREB renders it a relatively weak activator of CRE-dependent gene transcription and suggests one mechanism by which cAK and cGK differentially regulate gene transcription. These findings strongly imply that the restricted cytoplasmic localization of the cGKs is an important mechanism for selective regulation of nuclear functions by the two families of cyclic nucleotide-dependent protein kinases.
In this study, constitutively active cGKI mutants were generated by two approaches to determine if cAK's greater ability to activate CRE-dependent gene transcription was due to differences in cGK's and cAK's substrate specificities or differences in their subcellular localizations. First, mutation of an autophosphorylation site serine to an aspartic acid to mimic phosphorylation resulted in a constitutively active cGKI without significantly changing the size of the protein. Autophosphorylation of cGKI␤ at Ser 79 has been shown to activate the enzyme and to produce an elongation of the enzyme, suggesting that both cGMP binding and autophosphorylation activate the enzyme by a similar mechanisms (70). When this activated form of full-length cGKI␤, cGKI␤S79D, was expressed and its ability to transactivate a CRE-responsive promoter was measured, only a minimal increase in luciferase expression was found when compared with catalytic domain of cGKI. Like activated wild type cGKI␤, the cGKI␤S79D mutant was also restricted to the cytoplasm, suggesting a mechanism for its poor transactivating ability.
Each subunit of a cGK dimer consists of an amino-terminal regulatory domain and a carboxyl-terminal catalytic domain. Deletion of cGKI's regulatory domain generates a monomeric, constitutively active kinase. In contrast to the cGKI␤S79D mutant, the smaller, constitutively active catalytic domain of cGKI was found in both the cytoplasm and the nucleus. The catalytic domain strongly activated CRE-dependent gene transcription, demonstrating that the phosphotransferase activity  (Fig. 12A) were assayed for kinase activity in the presence (gray bar) or absence (black bar) of cGMP (50 M) using the heptapeptide substrate H2Btide (110 M). Protein kinase inhibitor peptide (1 M) was included in all assay tubes to inhibit endogenous cAK activity. C, Western blot analysis of Flag-tagged VASP phosphorylation. CV-1 cells were transiently transfected with pCMV.Flag-VASP (20 g), pCMV.Flag-PKI (2 g) and no cGKI␤ expression vector (Neo; lanes 1 and 2), 1 g of pCMV.mcGKI␤ (WT; lanes 3 and 4), or 1 g of pCMV.cGKI␤K407D/ R409Q (Mut; lanes 5 and 6). 24 h after transfection, cells were serum-starved for an additional 24 h and then treated for 1 h with (ϩ) or without (Ϫ) 8-Br-cGMP (1 mM). Extracts were generated and blotted as in Fig. 3. D, PhosphorImager quantitation of level of Flag-tagged VASP phosphorylation. Statistical analysis of experiment described for C. Black bar, without 8-Br-cGMP; gray bar, plus 8-Br-cGMP. of cGKI is capable of recognizing members of the nuclear CREB-like transcription factor family. Likewise, the catalytic domain of cGKI activated gene transcription to a similar extent as the free C subunit of cAK, implying that cGKI's restricted cytoplasmic localization and not its unique substrate specificity is the major reason for cGKI's weak transactivating ability. These data are entirely consistent with the conclusion that cGKI␤ is a weak activator of CRE-dependent gene transcription because the large size of the active kinase prevents nuclear translocation.
During the course of these experiments, a report appeared describing the identification of an NLS in the ATP-binding domain of human cGKI␤ that is only functional upon activation of the kinase by cGMP (34). The identified NLS (KILKKRHI) does not closely resemble the well characterized monopartite nuclear targeting sequence of SV40 large T antigen (PKKKRKV) or the bipartite motif of nucleoplasmin (71). In an attempt to compare these findings with our own, the murine cGKI␤ expression vector was transiently transfected into baby hamster kidney cells. Localization of murine cGKI␤ in baby hamster kidney cells by indirect immunofluorescence revealed diffuse cytoplasmic staining either in the presence or absence of 8-Br-cGMP. Similar results were obtained in HEK293 cells, COS-1 cells, and CV-1 cells, suggesting that the restricted cytoplasmic localization of cGKI␤ was common in mammalian cell lines. The discrepancy between our results and those published previously is not due to the overexpression of cGKI␤, since similar results were also obtained with endogenous cGKI in A7r5 cells. Finally, mutagenesis of the putative NLS reported by Gudi et al. (34) in our murine cGKI␤ shows no specific effect on its ability to transactivate a CRE-dependent reporter construct (Fig. 12). Instead, the decreased ability of this ATPbinding site mutant to transactivate a CRE-dependent reporter construct correlated with its decreased catalytic activity in vitro and in vivo.
In this study, we demonstrate that cGKI␤ is a weak inducer of CRE-dependent gene transcription. When HEK293 cells were transiently transfected with cGKI␤, 8-Br-cGMP treatment elevated CRE-dependent gene transcription 13-fold (Fig.  2B). 8-Br-cGMP treatment did not significantly activate CREdependent transcription in the absence of transfected cGKI␤ (Fig. 2B). Transcriptional activation was also mediated by the cGKI␤S79D constitutively active mutant alone, implying that cGKI␤ kinase activity by itself was capable of elevating CREdependent transcription and that the 8-Br-cGMP was not affecting other signaling systems. Transient coexpression of PKI did not block cGKI␤-dependent increases in gene transcription, suggesting that this transactivation is independent of cAK activity. Experiments using a GAL4-CREB fusion construct suggest that transcriptional regulation by cGKI is at least partially mediated through CREB, since the cGKI␤S79D mutant is capable of transactivating pGAL4-luc when coexpressed with the GAL4-CREB fusion protein (data not shown). These findings suggest that the pathway(s) by which cGKI␤ activates CRE-dependent gene transcription may involve specific cytoplasmic cGK substrates that can signal through CREB.
The exact mechanism(s) by which cGKs weakly regulate CRE-containing promoters is unknown. It is possible that cGK activation results in the activation of other protein kinases that are able to translocate to the nucleus and phosphorylate CREB. A number of growth factor-stimulated kinases including the RSKs (72) and MAPKAP kinase 2 (73) as well as CaM kinases (74) have recently been shown to translocate to the nucleus and phosphorylate CREB. Alternatively, although it is difficult to determine, it is possible that a small fraction of CREB is cytoplasmic and in equilibrium with the majority of nuclear CREB bound to DNA. Finally, novel transcription factors that can be phosphorylated in the cytoplasm and translocate to the nucleus may be the substrate of cGKs. In this regard, it is interesting to note that C/EBP␤ can be phosphorylated in the cytoplasm of PC12 cells by cAK, and following phosphorylation it translocates to the nucleus and increases gene transcription (75).
Since the cGKs and the cAKs are highly homologous protein kinase families that possess similar substrate specificities, many questions remain regarding the unique roles these kinases may play in cells that coexpress cAK and cGK isoforms. The results of this study suggest that the physiological role(s) of these cyclic nucleotide-dependent protein kinases are more distinct than currently appreciated and that, in comparison with cAKs, cGKs play only a minor role in the regulation of CRE-dependent transcription. The findings as well as the reagents generated within this study should be useful in characterization of the specific in vivo role(s) of the cyclic nucleotidedependent protein kinase families.